R A F T: Notes On Statistical Mechanics

Download as pdf or txt
Download as pdf or txt
You are on page 1of 177

K.P.N.

Murthy

Notes on
STATISTICAL MECHANICS
August 28, 2017

T
AF
DR
It was certainly not by design that the particles fell into order
They did not work out what they were going to do,
but because many of them by many chances
struck one another in the course of infinite time
and encountered every possible form and movement,
that they found at last the disposition they have,
and that is how the universe was created.
Titus Lucretius Carus (94 BC - 55BC)
de Rerum Natura
Everything existing in the universe is the fruit of chance and ne-
cessity. Democritus (370 BC)
The moving finger writes; and, having writ,
moves on : nor all your piety nor wit
shall lure it back to cancel half a line
nor all your tears wash out a word of it
Omar Khayyam (1048 - 1131)
Whatever happened,
T
happened for good.
Whatever is happening,
AF
is happening for good.
Whatever will happen,
will happen for good.
DR

Bhagavat Gita
”· · · Ludwig Boltzmann, who spent much of his life studying sta-
tistical mechanics, died in 1906, by his own hand. Paul Ehrenfest,
carrying on the work, died similarly in 1933. Now it is our turn ....
to study statistical mechanics. Perhaps it will be wise to approach
the subject rather cautiously.· · · ”
David Goodstein, States Matter, Dover (1975) (opening lines)
All models are wrong, some are useful. George E P Box
DR
AF
T
Contents

Quotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . i

1. Micro-Macro Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1 Aim of Statistical Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2 Micro - Macro Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.1 Boltzmann Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2.2 Boltzmann-Gibbs-Shannon Entropy . . . . . . . . . . . . . . . . 4
1.2.3 Heat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.4 Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.5 Helmholtz Free Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2.6 Energy Fluctuations and Heat Capacity . . . . . . . . . . . . 6
1.3 Micro World : Determinism and T
Time-Reversal Invariance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
AF
1.4 Macro World : Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Books . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.6 Extra Reading : Books . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.7 Extra Reading : Papers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
DR

2. Maxwell’s Mischief . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.1 Experiment and Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2 Sample space and events . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Probabilities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.4 Rules of probability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Random variable . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.6 Maxwell’s mischief : Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.7 Calculation of probabilities from an ensemble . . . . . . . . . . . . . 19
2.8 Construction of ensemble from probabilities . . . . . . . . . . . . . . . 19
2.9 Counting of the elements in events of the sample space :
Coin tossing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.10 Gibbs ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.11 Why should a Gibbs ensemble be large ? . . . . . . . . . . . . . . . . . 22

3. Binomial, Poisson, and Gaussian . . . . . . . . . . . . . . . . . . . . . . . . . . 27


3.1 Binomial Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Moment Generating Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
iv Contents

3.3 Binomial → Poisson . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31


3.4 Poisson Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Binomial → Poisson à la Feller . . . . . . . . . . . . . . . . . . . . 33
3.5 Characteristic Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.6 Cumulant Generating Function . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.7 The Central Limit Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.8 Poisson → Gaussian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.9 Gaussian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38

4. Isolated System: Micro canonical Ensemble . . . . . . . . . . . . . . . 41


4.1 Preliminaries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
4.2 Configurational Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.3 Ideal Gas Law : Derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.4 Boltzmann Entropy −→ Clausius’ Entropy . . . . . . . . . . . . . . . 44
4.5 Some Issues on Extensitivity of Entropy . . . . . . . . . . . . . . . . . . 44
4.6 Boltzmann Counting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
4.7 Micro canonical Ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.8 Heaviside and his Θ Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.9 Dirac and his δ Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.10 Area of a Circle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.11 Volume of an N -Dimensional Sphere . . . . . . . . . . . . . . . . . . . . . 50
T
4.12 Classical Counting of Micro states . . . . . . . . . . . . . . . . . . . . . . .
4.12.1 Counting of the Volume . . . . . . . . . . . . . . . . . . . . . . . . . .
52
52
AF
4.13 Density of States . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.13.1 A Sphere Lives on its Outer Shell : Power Law can
be Intriguing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.14 Entropy of an Isolated System . . . . . . . . . . . . . . . . . . . . . . . . . . 53
DR

4.15 Properties of an Ideal Gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54


4.15.1 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.15.2 Equipartition Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.15.3 Pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.15.4 Ideal Gas Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.15.5 Chemical Potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.16 Quantum Counting of Micro states . . . . . . . . . . . . . . . . . . . . . . 57
4.16.1 Energy Eigenvalues : Integer Number of Half Wave
Lengths in L . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.17 Chemical Potential : Toy Model . . . . . . . . . . . . . . . . . . . . . . . . . 60

5. Closed System : Canonical Ensemble . . . . . . . . . . . . . . . . . . . . . 63


5.1 What is a Closed System ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.2 Toy Model à la H B Callen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
5.3 Canonical Partition Function . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.3.1 Derivation à la Balescu . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
5.4 Helmholtz Free Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.5 Energy Fluctuations and Heat Capacity . . . . . . . . . . . . . . . . . . 68
Contents v

5.6 Canonical Partition Function : Ideal Gas . . . . . . . . . . . . . . . . . 69


5.7 Method of Most Probable Distribution . . . . . . . . . . . . . . . . . . . 70
5.8 Lagrange and his Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.9 Generalisation to N Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.10 Derivation of Boltzmann Weight . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.11 Mechanical and Thermal Properties . . . . . . . . . . . . . . . . . . . . . . 77
5.12 Entropy of a Closed System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
5.13 Free Energy to Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
5.14 Microscopic View : Heat and Work . . . . . . . .X .............. 80
5.14.1 Work in Statistical Mechanics : W = pi dEi . . . . . . 81
X i
5.14.2 Heat in Statistical Mechanics : q = Ei dpi . . . . . . 82
i
5.15 Adiabatic Process - a Microscopic View . . . . . . . . . . . . . . . . . . 82
5.16 Q(T, V, N ) for an Ideal Gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

6. Open System : Grand Canonical Ensemble . . . . . . . . . . . . . . . 87


6.1 What is an Open System ? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
6.2 Micro-Macro Synthesis : Q and G . . . . . . . . . . . . . . . . . . . . . . . . 90
6.3 Statistics of Number of Particles . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3.1 Euler and his Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3.2 Q : Connection to Thermodynamics . . . . . . . . . . . . . . .
T 92
6.3.3 Gibbs-Duhem Relation . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3.4 Average number of particles in an open system, hN i . 93
AF
6.3.5 Probability P (N ), that there are N particles in an
open system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.3.6 Number Fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
DR

6.3.7 Number Fluctuations and Isothermal Compressibility 95


2
6.3.8 Alternate Derivation of the Relation : σN /hN i2 =
kB T kT /V . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
6.4 Energy Fluctuations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99

7. Quantum Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105


7.1 Occupation Number Representation . . . . . . . . . . . . . . . . . . . . . . 105
7.2 Open System and Q(T, V, µ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
7.3 Fermi-Dirac Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
7.4 Bose-Einstein Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
7.5 Classical Distinguishable Particles . . . . . . . . . . . . . . . . . . . . . . . 109
7.6 Maxwell-Boltzmann Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
7.6.1 QMB (T, V, N ) → QMB (T, V, µ) . . . . . . . . . . . . . . . . . . 110
7.6.2 QMB (T, V, µ) → QMB (T, V, N ) . . . . . . . . . . . . . . . . . . 111
7.7 Thermodynamics of an open system . . . . . . . . . . . . . . . . . . . . . 111
7.8 Average number of particles, hN i . . . . . . . . . . . . . . . . . . . . . . . . 112
7.8.1 Maxwell-Boltzmann Statistics . . . . . . . . . . . . . . . . . . . . . 112
7.8.2 Bose-Einstein Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 113
vi Contents

7.8.3 Fermi-Dirac Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113


7.8.4 Study of a System with fixed N Employing Grand
Canonical Formalism . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
7.9 Fermi-Dirac, Maxwell-Boltzmann and Bose-Einstein Statis-
tics are the same at High Temperature and/or Low Densities 114
7.9.1 Easy Method : ρΛ3 → 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 114
7.9.2 Easier Method : λ → 0 . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.9.3 Easiest Method Ω(n b 1 , n2 , · · · ) = 1 . . . . . . . . . . . . . . . . . 117
7.10 Mean Occupation Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
7.10.1 Ideal Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.10.2 Ideal Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.10.3 Classical Indistinguishable Ideal Particles . . . . . . . . . . . 118
7.11 Mean Ocupation : Some Remarks . . . . . . . . . . . . . . . . . . . . . . . . 119
7.11.1 Fermi-Dirac Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.11.2 Bose-Einstein Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.11.3 Maxwell-Boltzmann Statistics . . . . . . . . . . . . . . . . . . . . . 120
7.11.4 At High T and/or Low ρ all Statistics give the same
hnk i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.12 Occupation Number : Distribution and Variance . . . . . . . . . . . 121
7.12.1 Fermi-Dirac Statistics and Binomial Distribution . . . . 122
7.12.2 Bose-Einstein Statistics and Geometric Distribution . 122
T
7.12.3 Maxwell-Boltzmann Statistics and Poisson Distribution124
AF
8. Bose-Einstein Condensation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.1 Introduction
X . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
8.2 hN i = hnk i . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127
k
DR

8.3 Summation to Integration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128


8.4 Graphical Inversion and Fugacity . . . . . . . . . . . . . . . . . . . . . . . . 130
8.5 Treatment of the Singular Behaviour . . . . . . . . . . . . . . . . . . . . . 131
8.6 Bose-Einstein Condensation Temperature . . . . . . . . . . . . . . . . . 133
8.7 Grand Potential for bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
8.8 Energy of Bosonic System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.8.1 T > TBEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
8.8.2 T ≤ TBEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
8.9 Specific Heat Capacity of bosons . . . . . . . . . . . . . . . . . . . . . . . . 137
CV
8.9.1 for T > TBEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
N kB
1 dλ 3 g3/2 (λ)
8.9.2 Third Relation : =− ............ 138
λ dT 2T g1/2 (λ)
CV
8.9.3 for T < TBEC . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
N kB
8.10 Mechanism of Bose-Einstein Condensation . . . . . . . . . . . . . . . . 140

9. Elements of Phase Transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147


Contents 1

10. Statistical Mechanics of Harmonic Oscillators . . . . . . . . . . . . . 157


10.1 Classical Harmonic Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . 157
10.1.1 Helmholtz Free Energy . . . . . . . . . . . . . . . . . . . . . . . . . . . 158
10.1.2 Thermodynamic Properties of the Oscillator System . 159
10.1.3 Quantum Harmonic Oscillator . . . . . . . . . . . . . . . . . . . . 159
10.1.4 Specific Heat of a Crystalline Solid . . . . . . . . . . . . . . . . 161
10.1.5 Einstein Theory of Specific Heat of Crystals . . . . . . . . 164
10.1.6 Debye Theory of Specific Heat . . . . . . . . . . . . . . . . . . . . 165
10.1.7 Riemann Zeta Function . . . . . . . . . . . . . . . . . . . . . . . . . . 167
10.1.8 Bernoulli Numbers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169

T
AF
DR
DR
AF
T
1. Micro-Macro Synthesis

1.1 Aim of Statistical Mechanics


Statistical mechanics provides a theoretical bridge that takes you from the
micro world1 , to the macro world2 . The chief architects of the bridge were
Ludwig Eduard Boltzmann (1844 - 1906), James Clerk Maxwell(1831-1879),
Josiah Willard Gibbs(1839-1903) and Albert Einstein(1879-1953).
Statistical Mechanics makes an attempt to derive the macroscopic prop-
erties of an object from the properties of its microscopic constituents and the
interactions amongst them. It tries to provide a theoretical basis for the em-
pirical thermodynamics - in particular the emergence of the time-asymmetric
Second law of thermodynamics from the time-symmetric microscopic laws of
classical and quantum mechanics. T
• When do we call something, a macroscopic object, and something a mi-
AF
croscopic constituent ?
The answer depends crucially on the object under study and the properties
under investigation. For example,
DR

• if we are interested in properties like density, pressure, temperature etc.


of a bottle of water, then the molecules of water are the microscopic
constituents; the bottle of water is the macroscopic object.
• in a different context, an atom constitutes a macroscopic object; the
electrons, protons and neutrons form its microscopic constituents.
• A polymer is a macroscopic object; the monomers are its microscopic
constituents.
• A society is a macroscopic object; men, women, children and perhaps
monkeys are its microscopic constituents.
Statistical mechanics asserts that if you know the properties of the micro-
scopic constituents and how they interact with each other, then you can make
predictions about its macroscopic behaviour.
1
of Newton, Schrödinger, Maxwell, and a whole lot of other scientists who devel-
oped the subjects of classical mechanics, quantum mechanics, and electromag-
netism
2
of Carnot, Clausius, Kelvin, Helmholtz and several others responsible for the
birth and growth of thermmodynamics
4 1. Micro-Macro Synthesis

1.2 Micro - Macro Connections

1.2.1 Boltzmann Entropy

The first and the most important link, between the micro(scopic) and the
macro(scopic) worlds is given by,
b
S = kB ln Ω(E, V, N ). (1.1)
It was proposed by Boltzmann3 . S stands for entropy and belongs to the
macro world described by thermodynamics. Ω b is the number of micro states
of a macroscopic system . kB is the Boltzmann constant 5 that establishes
4

correspondence of the statistical entropy of Boltzmann to the thermodynamic


entropy of Clausius 6 .

1.2.2 Boltzmann-Gibbs-Shannon Entropy

More generally we have the Boltzmann-Gibbs-Shannon entropy given by,


X
S = −kB pi ln pi . (1.2)
i

The micro states are labelled by i and the sum is taken over all the micro
T
states of the macroscopic system; The probability is denoted by pi .
An interesting issue : Normally we resort to the use of probability only when
AF
we have inadequate data about the system or incomplete knowledge of the phe-
nomenon under study. Thus, such an enterprise is sort of tied to our ignorance.
Keeping track of the positions and momenta of some 1030 molecules via Newton’s
equations is undoubtedly an impossible task. It was indeed the genius of Boltzmann
DR

which correctly identified that macroscopic phenomena are tailor-made for a statis-
tical description. It is one thing to employ statistics as a convenient tool to study
macroscopic phenomena but quite another thing to attribute an element of truth to
such a description. But then this is what we are doing precisely in Eq. (1.2) where
we express entropy, which is a property of an object, in terms of probabilities. It is
definitely a bit awkward to think that a property of an object is determined by what
we know or what we do not know about it! But remember in quantum mechanics
3
engraved on the tomb of Ludwig Eduard Boltzmann (1844-1906) in Zentralfried-
hof, Vienna.
4
For example an ordered set of six numbers, three positions and three momenta
specify a single particle. An ordered set of 6N numbers specify a macroscopic
system of N particles. The string labels a micro state.
5
kB = 1.381×10−23 joules (kelvin)−1 . We have kB = R/A where R = 8.314 joules
(kelvin)−1 is the universal gas constant and A = 6.022 × 1023 (mole)−1 is the
Avagadro number.
6
Note that the quantity ln Ωb is a number. Claussius’ entropy is measured in units
of joules/kelvin. Hence the Boltzmann constant is kB and has units joules/kelvin.
It is a conversion factor and helps you go from one unit of energy (joule) to
another unit of energy (kelvin).
1.2 Micro-Macro Connections 5

the observer, the observed, and the observation are all tied together : the act of
measurement places the system, at the time of observation, in one of eigenstates
of the observable; we can make only probabilistic statement about the eigenstate
the wave function of the system would collapse into. For a discussion on this subtle
issue see the beautiful tiny book of Carlo Rovelli7 .
Two important entities we come across in thermodynamics are heat and
work. They are the two means by which a thermodynamic system transacts
energy with its surroundings or with another thermodynamic system. Heat
and work are described microscopically as follows.

1.2.3 Heat
X
q= Ei dpi . (1.3)
i

The sum runs over all the micro states. Ei is the energy of the system when
it is in micro state i. The probability that the system can be found inPmicro
state i given by pi . We need to impose an additional constraint that i dpi
is zero to ensure that the total probability is unity.

1.2.4 Work
The statistical description of work is given by
X
T
AF
W = pi dEi . (1.4)
i
Ei is the energy of the system when it is in the micro state labelled by the
index i.
DR

1.2.5 Helmholtz Free Energy


An important thermodynamic property of a closed system is Helmholtz free
energy given by,
F = −kB T ln Q(T, V, N ). (1.5)
8
Helmholtz free energy F (T, V, N ) of thermodynamics is related to the canon-
ical partition function Q(T, V, N ) of statistical mechanics. This is another
important micro-macro connection for a closed system. The canonical parti-
tion function is the sum of the
P Boltzmann weights of the micro states of the
closed system. Q(β, V, N ) = i exp(−βEi ).
7
Carlo Rovelli, Seven Brief Lessons in Physics, Translated by Simon Carnell and
Erica Segre, Allen Lane an imprint of Penguin Books (2015)pp.54-55.
8
In thermodynamics Helmholtz free energy, F (T, V, N ) is expressed by the Legen-
dre transform
  of U (S, V, N ), where we transform S to T employing the relation
∂U
T = and U to F : F = U − T S;
∂S V,N
6 1. Micro-Macro Synthesis

1.2.6 Energy Fluctuations and Heat Capacity


The next equation that provides a micro-macro synthesis is given by,
2
σE = kB T 2 CV . (1.6)
This relation connects the specific heat at constant volume (CV ), which we
come across when we study thermodynamics, to the statistical fluctuations
2
(σE ) of energy in a closed system.
We shall see of several such micro-macro connections in the course of
study of statistical mechanics. We can say the aim of statistical mechanics
is to synthesise the macro world from the micro world. This is not an easy
task. Why ? Let us spend a little bit of time on this question.

1.3 Micro World : Determinism and


Time-Reversal Invariance
The principal character of the micro world is determinism and time-reversal
invariance. Determinism implies that the entire past and the entire future is
frozen in the present. The solar system is a good example. If you know the
positions and momenta of all the planets now, then the Newtonian machinery
is adequate to tell you where the planets shall be a thousand years from now
T
and where they were some ten thousand years ago.
In the micro world of Newton and Schrödinger, we can not tell which
AF
direction the time flows. Both directions are equally legitimate and equally
unverifiable. Microscopic laws do not distinguish the future from the past.
They are time-reversal invariant; i.e. the equations are invariant under trans-
DR

formation of t → −t.

1.4 Macro World : Thermodynamics


On the other hand, the macro world obeys the laws of thermodynamics.
• The zeroth law tells us of thermal equilibrium; it provides a basis for
the thermodynamic property temperature. It is the starting point for the
game of thermodynamics.
• The first law is an intelligent articulation of the law of conservation of
energy; it provides a basis for the thermodynamic property called the in-
ternal energy. You can put in energy into a system or take it out, by heat
or by work. Thus the change in the internal energy (dU ) of a thermody-
namic system equals the energy transacted by it with the surroundings
or with another thermodynamic system by heat d¯q and work d¯W . Thus
dU = d¯q + d¯W . The bar on d reminds us that these quantities are small
quantities but not exact differentials. Heat and work are not properties of
a thermodynamic system; rather they describe thermodynamic processes.
1.4 Micro World : Thermodynamics 7

• The second law tells us that come what may, an engine can not de-
liver work equivalent to the energy it has drawn by heat from a heat
source. However an engine can draw energy by work and deliver exactly
equivalent amount by heat. The second law is a statement of this ba-
sic asymmetry between ”heat→ work” and ”work→ heat” conversions.
The Second law provides a basis for the thermodynamic property called
entropy. The entropy of a system increases by dS = d¯q/T when it ab-
sorbs reversibly d¯q amount of energy by heat at constant temperature.
The second law says that in any process entropy increases or remains
constant : ∆S ≥ 0.
• The third law tells that entropy vanishes at absolute zero tempera-
ture. Notice that in thermodynamics only change in entropy is defined :
dS = d¯q/T . Since dS is an exact differential we can assert S as a ther-
modynamic variable and it describes a property of the system. The third
law demands that this property is zero at absolute zero and hence we can
assign a value S at any temperature. We can say the third law provides
a basis for absolute zero temperature on entropy scale.
Of these, the second law is tricky. It breaks the time-symmetry present
in the microscopic descriptors. Macroscopic behaviour is not time-reversal
invariant. There is a definite direction of time - the direction of increasing
entropy. T
How do we comprehend the time asymmetric macroscopic behaviour
emerging from the time symmetric microscopic laws ?
AF
Let us make life simpler by attributing two aims to statistical mechanics.
The first is to provide a machinery for calculating the thermodynamic prop-
erties of a system on the basis of the properties of its microscopic constituents
e.g. atoms and molecules, and their mutual interactions.
DR

Statistical Mechanics has been eminently successful in this enterprise.


This is precisely why we are studying this subject.
The second aim is to derive the Second law of thermodynamics. Statistical
Mechanics has not yet had any spectacular success on this count. However,
some recent developments in non linear dynamics and chaos, have shown there
is indeed an unpredictability in some (non-linear) deterministic system; we
now know that determinism does not necessarily imply predictability. This
statement, perhaps, provides the raison d’etre for the ’statistics’ in statistical
mechanics.
In these lectures I shall not address the second issue - concerning the
emergence of time asymmetry observed in macroscopic phenomena. I shall
leave this question to the more knowledgeable and better equipped physi-
cists/philosophers. Instead we shall concentrate on how to derive the prop-
erties of a macroscopic system from those of its microscopic constituents and
their interactions.
I shall tell you of the elementary principles of statistical mechanics. I shall
be as pedagogical as possible. Stop me when you do not understand.
8 1. Micro-Macro Synthesis

I shall cover topics in


• micro canonical ensemble that provides a description of isolated system;
• canonical ensemble, useful in the study of closed system;
• grand canonical ensemble that describes open system.
I shall discuss equilibrium fluctuations of
• energy in canonical ensemble and relate them to heat capacity;
• number of molecules in an open system and relate them to isothermal
compressibility.
Within the framework of grand canonical ensemble I shall discuss Bose-
Einstein, Fermi-Dirac and Maxwell Boltzmann statistics.
I shall deal with ideal gas, and classical and quantum harmonic oscillators.
I shall discuss Bose Einstein condensation in some details with an emphasis
on the mechanism.
While on quantum harmonic oscillators, I shall discuss statistical me-
chanics of phonons emerging due to quantization of displacement field in a
crystalline lattice, and photons arising due to quantization of electromagnetic
field.
Maxwell’s demon is a mischievous entity. The demon has a simple goal.
Let me explain briefly.
T
For every thermodynamic property, we have, in statistical mechanics, a
corresponding statistical quantity. We take the average of the statistical quan-
AF
tity over a suitable ensemble9 and relate it to its thermodynamic counter part.
Thus entropy is a statistical entity. It is a mere number. It depends on the
number of micro states accessible to a macroscopic system, consistent with
the constraints. If entropy is statistical then the second law which states that
DR

entropy invariable increases in an irreversible process, should be a statistical


statement. If so, there is a non-zero probability that the second law would be
violated. Maxwell constructed a demon that violates the second law ... a de-
mon that extracts work from an equilibrium system. I shall discuss Maxwell’s
demon and its later incarnations.
Toward the end, I shall discuss some recent developments in thermody-
namics - work fluctuation theorems and second law violations. I shall assume
you are all comfortable with calculus. I shall also assume you have a nodding
acquaintance with thermodynamics.
Let me end this section by giving a list of some books and articles on
statistical mechanics and thermodynamics that have caught my fancy. This
list is by no way exhaustive.
9
micro canonical, for isolated system, canonical for closed system and grand
canonical for open systems
1.5 Books 9

1.5 Books

• R K Pathria, Statistical Mechanics, Second Edition, Butterworth -


Heinemann (1996). A popular book. Starts with a beautiful historical
account of the subject. Contains a very large collection of interesting,
non-trivial problems, some of which are taxing ! You will learn how to
do statistical mechanics from this book.
• Donald A McQuarrie, Statistical Mechanics, Harper & Row (1976). A
beautiful book with emphasis on applications. Contains excellent prob-
lems; physical chemists or chemical physicists will feel more at home with
this book.
• R Balescu,, Equilibrium and Non-Equilibrium Statistical Mechanics,
Wiley (1975). An insightful book with emphasis on conceptual issues.
Irreversibility is explained neatly. I shall tell you how to derive canonical
partition function, borrowing from this book.
• David Goodstein, States of Matter, Dover (2002). A Delightful and
entertaining text. You are reminded of Feynman’s writing when you read
this book. The discussion on dimensional analysis is excellent. This book
is a must in your bookshelf.
• Debashish Chowdhury and Dietrich Stauffer, Principles of Equi-
librium Statistical Mechanics, Wiley-VCH (2000). An easy to read and
T
enjoyable book. Discusses applications to several fields. Contains a good
collection of interesting problems.
AF
• F Rief, Fundamentals of statistical and thermal physics, McGraw-Hill
(1965). One of the best text books on statistical thermodynamics. Rief
develops thermal physics entirely in the vocabulary of statistical mechan-
ics. As a result after reading this book, you will get an uneasy feeling that
DR

thermodynamics has been relegated to the status of an uninteresting ap-


pendix to statistical mechanics. My recommendation : read this book for
learning statistical - thermodynamics; then read Callen and/or Van Ness
for learning thermodynamics. Then you will certainly fall in love with
both statistical mechanics and thermodynamics separately!
• Palash B Pal, An Introductory Course of Statistical Mechanics, Narosa
(2008). A book with a broad perspective and with emphasis on relativis-
tic systems.
• D Chandler, Introduction to Modern Statistical Mechanics, Oxford Uni-
versity Press, New York (1987). A book that connects neatly the tra-
ditional to modern methods of teaching statistical mechanics; gives an
excellent and simple introduction to re-normalization groups. A great
book for the teachers also.
• Claude Garrod, Statistical Mechanics and Thermodynamics, Oxford
University Press (1995). A good book at an introductory level; neatly
organized; pedagogic; nice problems and exercises. I shall borrow from
this book for teaching micro canonical ensemble.
10 1. Micro-Macro Synthesis

• Kerson Huang, Statistical Mechanics, Second Edition, Wiley India


(2011). Generations after generations of physicists have learnt statis-
tical mechanics from this book. It is perhaps one of a very few books
that take kinetic theory and Boltzmann transport equation as a start-
ing point. Historically, statistical mechanics originated from the work of
Ludwig Eduard Boltzmann. After all, it is Boltzmann transport equa-
tion - with its collision term obeying stosszahlansatz10 that establishes
irreversible march toward thermal equilibrium - an important point that
is either overlooked or not adequately emphasized in most of the text
books on statistical mechanics.
The book contains three parts; the first is on thermodynamics; the second
on statistical mechanics; and the third on special topics in statistical
mechanics.
Do not learn thermodynamics from this book; you will lose interest. The
other two parts are excellent - especially the third on special topics.
I would recommend retain your copy of the first edition of this book.
Huang has omitted in his second edition, several interesting discussions
present in the first edition.
• Kerson Huang, Introduction to Statistical Physics, Second Edition,
CRC Press (2009). I think Huang has made an hurried attempt to ’mod-
ernize’ his earlier classic : ”Statistical Mechanics”. I do not recommend
T
this book to students taking their first course in statistical mechanics.
However a discerning teacher will find this book very useful. Beware of
AF
typos, several of them.
• J W Gibbs, Elementary Principles in Statistical Mechanics, Schribner,
New York (1902). A seminal work on statistical mechanics. It is indeed
a thrillng experience to read statistical mechanics in the words of one of
DR

its masters. The book looks at canonical formalism in a purely logical


fashion. Read this book after becoming a bit familiar with statistical
mechanics.
• Avijit Lahiri, Statistical Mechanics : An Elementary Outline, Revised
Edition, Universities Press (2008). A neat and well planned book, though
somewhat idiosyncratic. Focuses on bridging the micro world described
by quantum mechanics to the macro world of thermodynamics. Concepts
like mixed states, reduced states etc. provide the basic elements in the de-
velopment of the subject. The book contains a good collection of worked
out examples and problems.
• Joon Chang Lee, Thermal physics - Entropy and Free Energies, World
Scientific (2002). Joon Chang Lee presents statistical thermodynamics in
an unorthodox and distinctly original style. The presentation is so simple
and so beautiful that you do not notice that the book is written in an
awful English and at several places, the language is flawed.
10
collision number assumption or also known as molecular chaos
1.5 Books 11

• James P Sethna, Entropy, Order Parameters, and Complexity, Claren-


don Press, Oxford (2008). James Sethna covers an astonishingly wide
range of modern applications; a book, useful not only to physicists, but
also to biologists, engineers, and sociologists. I find exercises and foot-
notes very interesting; often more interesting than the main text!
• C Kittel, and H Krömer, Thermal physics, W H Freeman (1980). A
good book; somewhat terse. I liked the parts dealing with entropy, and
Boltzmann weight; contains a good collection of examples.
• Daniel V Schröder, An Introduction to Thermal Physics, Pearson
(2000). Schröder has excellent writing skills. The book reads well. Con-
tains plenty of examples. Somewhat idiosyncratic.
• M Glazer, and J Wark, Statistical Mechanics : A Survival Guide, Ox-
ford University Press (2010). This book gives a neat introduction to sta-
tistical mechanics; well organized; contains a good collection of worked-
out problems; a thin book and hence does not threaten you !
• H C Van Ness, Understanding Thermodynamics, Dover (1969). This
is an awesome book; easy to read and very insightful. In particular, I
enjoyed reading the first chapter on the first law of thermodynamics,
the second on reversibility, and the fifth and sixth on the second law.
My only complaint is that Van Ness employs British Thermal Units.
Another minor point : Van Ness takes the work done by the system as
T
positive and that done on the system as negative. Engineers always do
this. Physicists and chemists employ the opposite convention. For them
AF
the sign coincides with the sign of change of internal energy caused by
the work process. If the transaction leaves the system with higher energy,
work done is positive; if it results in lowering of energy, the work done is
negative.
DR

• H B Callen, Thermodynamics, John Wiley (1960). Callen sets the stan-


dard for a text book. This book has influenced generations of teachers
and students alike, all over the world. Callen is a house hold name in
the community of physicists. The book avoids all the pitfalls in the his-
torical development of thermodynamics by introducing a postulational
formulation.
• H B Callen, Thermodynamics and an Introduction to thermostatistics,
Second Edition, Wiley, India (2005). Another classic from H B Callen.
He has introduced statistical mechanics without undermining the inner
strength and beauty of thermodynamics. In fact, the statistical mechanics
he presents, enhances the beauty of thermodynamics.
The simple toy problem with a red die (the closed system) and two white
dice (the heat reservoir), and the restricting sum to a fixed number (con-
servation of total energy) motivates beautifully the canonical formalism.
The pre-gas model introduced for explaining grand canonical ensemble of
fermions and bosons is simply superb. I also enjoyed the discussions on
12 1. Micro-Macro Synthesis

the subtle mechanism underlying Bose condensation. I can go on listing


several such gems. The book is full of beautiful insights.
A relatively inexpensive, Wiley student edition of the book is available
in the Indian market. Buy your copy now !
• Gabriel Weinreich, Fundamental Thermodynamics, Addison Wesley
(1968). Weinreich is eminently original; has a distinctive style. Perhaps
you will feel uneasy when you read this book for the first time. But very
soon, you will get used to Weireich’s idiosyncracy; and you would love
this book.
• N D Hari Dass, Principles of Thermodynamics, CRC Press, Taylor and
Francis (2014). A beautiful book on thermodynamics; perhaps the most
recent to arrive at the market place. Hari Dass is a great story teller.
He has the magic to convert prosaic things into exciting entities. If you
are a beginner and plan to learn thermodynamics, I will recommend this
book to you. There is pedagogy; titbits of history scattered all through;
and no dearth of rigor. You will learn thermodynamics correctly. I will
recommend this book to an expert also. I am sure he will find something
new and something surprising in every chapter.
The only complaint I have is, the book gives a feeling that the last word
on thermodynamics has been said. I don’t think so. We are still struggling
to know what heat is. Thermodynamics tells us that heat is a process
T
by which energy is transacted amongst macroscopic bodies. This helps
because you will not mistake heat for a property of the system. But then
AF
this statement does not say anything about what heat is. We know heat
is something hidden somwhere in the 1024 or so of trajectories; you can
not compute these trajectories : not only because the number is obscenely
large but also because they are inherently unpredictable due to sensitive
DR

dependence on the in initial conditions. It is in this region of the phase


space where chaos reigns we have to look for a meaning of heat. I would
like to hear of these in the words of Hari Dass, perhaps in the next edition
of his book !
• C B P Finn, Thermal Physics, Nelson Thornes (2001). Beautiful; con-
cise; develops thermodynamics from first principles. Finn brings out the
elegance and power of thermodynamics.
• Max Planck, Treatise on Thermodynamics, Third revised edition, Dover;
first published in the year 1897. Translated from the seventh German
edition (1922). A carefully scripted master piece; emphasises chemical
equilibrium. I do not think any body can explain irreversibility as clearly
as Planck does. If you think third law of thermodynamics is irrelevant,
then read the last chapter. You may change your opinion.
• E Fermi, Thermodynamics, Dover (1936). A great book from a great
master; concise; the first four chapters (on thermodynamic systems, first
law, the Second law, and entropy) are superb. I also enjoyed the parts
covering Clapeyron and van der Waal equations.
1.6 Extra Reading : Books 13

• J S Dugdale, Entropy and its physical meaning, Taylor and Francis


(1998). An amazing book. Dugdale de-mystifies entropy. This book is
not just about entropy alone, as the name would suggest. It teaches you
thermodynamics and statistical mechanics. A book that cleverly avoids
unnecessary rigour.
• M W Zamansky, and R H Dittman, Heat and Thermodynamics, an
intermediate textbook, Sixth edition, McGraw-Hill (1981). A good and
dependable book for a first course in thermodynamics.
• R Shanthini, Thermodynamics for the Beginners, Science Education
Unit, University of Peredeniya (2009). Student-friendly. Shanthini has
anticipated several questions that would arise in the mind of an oriental
student when he or she learns thermodynamics for the first time. The
book has a good collection of worked out examples. A bit heavy on heat
engines.
• Dilip Kondepudi and Ilya Prigogine, Modern Thermodynamics :
From heat engines to Dissipative Structures, John Wiley (1998). Clas-
sical, statistical, and non equilibrium thermodynamics are woven into a
single fabric. Usually chemists present thermodynamics in a dry fash-
ion. This book is an exception; it tells us learning thermodynamics can
be fun. Contains lots of interesting tit-bits on history. Deserves a better
cover design; the present cover looks cheap.
T
AF
1.6 Extra Reading : Books

• Nicolaus Sadi Carnot, Reflexions sur la puissance motrice du feu et sur


les machines propres á déveloper cette puissance, Paris (1824); for English
DR

translation see Sadi carnot, Reflections on the motive power of fire and
on machines fitted to develop that power, in J Kestin (Ed.) The second
law of thermodynamics, Dowden, Hutchinson and Ross, Stroudsburg, PA
(1976)p.16
• J Kestin (Ed.), The second law of thermodynamics, Dowden, Hutchinson
and Ross (1976)
• P Atkin, The Second Law, W H Freeman and Co. (1984)
• G Venkataraman, A hot story, Universities Press (1992)
• Michael Guillen, An unprofitable experience : Rudolf Clausius and the
second law of thermodynamics p.165, in Five Equations that Changed the
World, Hyperion (1995)
• P Atkins, Four Laws that drive the Universe, Oxford university Press
(2007).
• Christopher J T Lewis, Heat and Thermodynamics : A Historical
Perspective, First Indian Edition, Pentagon Press (2009)
• S G Brush, Kinetic theory Vol. 1 : The nature of gases and of heat,
Pergamon (1965) Vol. 2 : Irreversible Processes, Pergamon (1966)
14 1. Micro-Macro Synthesis

• S G Brush, The kind of motion we call heat, Book 1 : Physics and the
Atomists Book 2 : Statistical Physics and Irreversible Processes, North
Holland Pub. (1976)
• I Prigogine, From Being to Becoming, Freeman, San Francisci (1980)
• K P N Murthy, Excursions in thermodynamics and statistical mechan-
ics, Universities Press (2009)

1.7 Extra Reading : Papers

• K K Darrow, The concept of entropy, American Journal of Physics 12,


183 (1944).
• M C Mackay, The dynamical origin of increasing entropy, Rev. Mod.
Phys. 61, 981 (1989).
• T Rothman, The evolution of entropy, pp.75-108, in Science á la mode
: physical fashions and fictions Princeton University Press (1989)
• Ralph Baierlein, Entropy and the second law : A pedagogical alterna-
tive, American Journal of Physics 62, 15 (1994)
• G. Cook, and R H Dickerson, Understanding the chemical potential,
American Journal of Physics 63, 738 (1995).
• K. P. N. Murthy, Ludwig Boltzmann, Transport Equation and the Sec-
ond Law, arXiv: cond-mat/0601566 (1996)T
• Daniel F Styer, Insight into entropy, American Journal of Physics 68,
AF
1090 (2000)
• T P Suresh, Lisha Damodaran, and K M Udayanandan, Gibbs’ Paradox
: Mixing and Non-Mixing Potentials, Physics Education 32(3), Article
No. 4 (2016).
DR

• B J Cherayil, Entropy and the direction of natural change, Resonance


6, 82 (2001)
• J K Bhattacharya, Entropy á la Boltzmann, Resonance 6, 19 (2001)
• J Srinivasan, Sadi Carnot and the second law of thermodynamics, Res-
onance 6 42 (2001)
• D C Shoepf, A statistical development of entropy for introductory
physics course, American Journal of Physics 70, 128 (2002).
• K P N Murthy, Josiah Willard Gibbs and his Ensembles, Resonance
12, 12 (2007).
2. Maxwell’s Mischief

2.1 Experiment and Outcomes

Toss a coin : You get either ”Heads” or ”Tails”. The experiment has two
outcomes. Consider tossing of two coins. Or equivalently toss a coin twice.
There are four outcomes : An outcome is an ordered pair. Each entry in the
pair is drawn from the set {H, T }.
We can consider, in general, tossing of N coins. There are 2N outcomes.
Each outcome is an ordered string of size N with entries drawn from the set
{H, T }.

Roll a die : You get one of the six outcomes :




• ;

;

• ;
• •
;

T •

;
• • •


AF
 
• • • • • • • • •

Consider throwing of N dice. There are then 6N outcomes. Each outcome is


an ordered string of N entries drawn from the basic set of six elements given
above.
DR

Select randomly an air molecule in this room and find its position
and momentum : Consider the air molecule to be a point particle. In clas-
sical mechanics, a point particle is completely specified by its three position
(q1 , q2 , q3 ) and three momentum (p1 , p2 , p3 ) coordinates. An ordered set of
six numbers
{q1 , q2 , q3 , p1 , p2 , p3 }
is the outcome of the experiment. A point in the six dimensional phase space
represents an outcome of the experiment or the microstate of the system of
single molecule. We impose certain constraints e.g. the molecule is always
confined to this room. Then all possible strings of six numbers, consistent
with the constrains, are the outcomes of the experiment.

2.2 Sample space and events

The set of all possible outcomes of an experiment is called the sample space.
Let us denote it by the symbol Ω.
16 2. Maxwell’s mischief

• Ω = {H, T } for the toss of a single coin.


• Ω = {HH, HT, T H, T T } for the toss of two coins.
A subset of Ω is called an event. Let A ⊂ Ω denote an event. When you
perform the experiment if you get an outcome that belongs to A then we say
the event A has occured.
For example consider tossing of two coins. Let event A be described by the
statement that the first toss is H. Then A consists of the following elements:
{HH, HT }.
The event corresponding to the roll of an even number in a dice, is the
subset
 
• • • • • • 
; ; .
 
• • • • • •

2.3 Probabilities
Probability is defined for an event. What is the probability of the event {H}
in the toss of a coin ? One-half. This would be your immediate response.
The logic is simple. There are two outcomes : ”Heads” and ”Tails”. We have
no reason to believe why should the coin prefer ”Heads” over ”Tails” or
T
vice versa. Hence we say both outcomes are equally probable. What is the
probability of having at least one ”H” in a toss of two coins ? The event
AF
corresponding this statement is {HH, HT, T H} and contains three elements.
The sample size contains four elements. The required probability is thus 3/4.
All the four outcomes are equally probable 1 . Then the probability of an
event is the number of elements in that event divided by the total number of
DR

elements in the sample space. For e.g., the event A of rolling an even number
in a game of dice, P (A) = 3/6 = 0.5. The outcome can be a continuum. For
example, the angle of scattering of a neutron is a real number between zero
and π. We then define an interval (θ1 , θ2 ), where 0 ≤ θ1 ≤ θ2 ≤ π, as an
event. A measurable subset of a sample space is an event.

2.4 Rules of probability


The probability p that you assign to an event, should be obey the following
rules.
p≥0
p(A ∪ B) = p(A) + p(B) − p(A ∩ B)
p(φ) = 0 p(Ω) = 1 (2.1)
1
Physicists have a name for this. They call it the axiom (or hypothesis or assump-
tion) of Ergodicity. Strictly ergodicity is not an assumption; it is absence of an
assumption required for assigning probabilities to events.
2.5 Random variable 17

In the above φ denotes a null event and Ω, a sure event.


How does one assign probability to an event ?
Though this question does not bother the mathematicians, the physicists
should worry about this2 . They should find the right way to assign probabil-
ities to get the phenomenon right. We can say that the subject of statistical
mechanics mainly deals with finding the right way to characterize a micro
state, the sample space, the events and the assigning of probabilities to the
events, depending on the system and phenomenon under investigation.

2.5 Random variable


The next important concept in probability theory is random variable , de-
noted by the symbol x = X(ω) where ω denotes an outcome and x a real
number. Random variable is a way of stamping an outcome with a number :
Real number, for a real random variable. Integer, for an integer random vari-
able. Complex number, for a complex random variable3 . Thus the random
variable x = X(ω) is a set function.
Consider a continuous random variable x = X(ω). We define a probability
density function f (x) by
f (x)dx = P (ω|x ≤ X(ω) ≤ x + dx) T (2.2)
In other words f (x)dx is the probability of the event (measurable subset) that
AF
contains all the outcomes to which we have attached a real number between
x and x + dx.
Now consider a continuous random variable defined between a to b with
a < b. We define a quantity called the ”average” of the random variable x
DR

as Z b
µ= dx x f (x).
a
µ is also called the mean, expectation, first moment etc.
Consider a discrete random variable n, taking values from say 0 to N . Let
P (n) define the discrete probability. We define the average of the random
variable as
2
Maxwell and Boltzmann attached probabilities to events in some way; we got
Maxwell-Boltzmann statistics.
Fermi and Dirac had their own way of assigning probabilities to Fermions e.g.
electrons, occupying quantum states. We got Fermi-Dirac statistics.
Bose and Einstein came up with their scheme of assigning probabilities to
Bosons, populating quantum states; and we got Bose-Einstein statistics.
3
In fact, we stamped dots on the faces of die; this is equivalent to implementing
the idea of an integer random variable : attach an integer between one and six
to each outcome.
For a coin, we stamped ”Heads” on one side and ”Tails” on the other. This is
in the spirit of defining a random variable; we have stamped figures instead of
numbers.
18 2. Maxwell’s mischief
N
X
µ= n P (n).
n=0
But then, we are accustomed to calculating the average in a different
way. For example I am interested in knowing the average marks obtained
by the students in a class, in the last mid-semester examination. How do I
calculate it ? I take the marks obtained by each of you, sum them up and
divide by the total number of students. That is it. I do not need notions
like probabilities, probability density, sum over the product of the random
variable and the corresponding probability, integration of the product of the
continuous random variable and its probability density function etc.
Historically, before Boltzmann and Maxwell, physicists had no use for
probability theory in their work. Newton’s equations are deterministic. There
is nothing chancy about a Newtonian trajectory. We do not need probabilistic
description in the study of electrodynamics described by Maxwell equations;
nor do we need probability to comprehend and work with Einstein’s relativity
- special or general.
However mathematicians had developed the theory of probability as an
important and sophisticated branch of mathematics.
It was Ludwig Eduard Boltzmann who brought, for the first time, the
idea of probability into physical sciences; he was championing the cause of
T
kinetic theory of heat and atomic theory of matter. Boltzmann transport
equation is the first ever equation written for describing the time evolution
AF
of a probability distribution.

2.6 Maxwell’s mischief : Ensemble


DR

However, Maxwell, had a poor opinion about a physicist’s ability to com-


prehend mathematicians’ writings on probability theory, in general, and the
meaning of average as an integral over a probability density, in particular.
After all, if you ask a physicist to calculate the average age of a student
in the class, he’ll simply add the ages of all the students and divide by the
number of students.
To be consistent with this practice, Maxwell proposed the notion of an
ensemble of outcomes of an experiment (or an ensemble of realisations of a
random variable). Let us call it Maxwell ensemble 4 .
Consider a collection of a certain number of independent realisations of
the toss of a single coin. We call this collection a Maxwell ensemble if it it
obeys certain conditions, see below.
Let N denote the number of elements in the ensemble. Let n1 denote the
number ”Heads” and n2 number of ’Tails”. We have n1 +n2 = N . If n1 = N p,
4
Later we shall generalise the notion of Maxwell ensemble and talk of ensemble
as a collection identical copies of a macroscopic system. We shall call it a Gibbs
ensemble
2.9 Counting of elements of an event 19

and hence n2 = N q, then we say the collection of outcomes constitutes a


Maxwell ensemble.
Thus an ensemble holds information about the outcomes of the exper-
iment constituting the sample space; it also holds information about the
probability of each outcome. The elements of an ensemble are drawn from
the sample space; however each element occurs in an ensemble as often as to
reflect correctly its probability.
For example consider an experiment of tossing a p-coin5 with p =
0, 75; then a set of four elements given by {H, H, H, T } is a candi-
date for an ensemble underlying experiment. . A set of eight elements
{H, H, H, H, H, H, T, T } is also an equally valid ensemble for this exper-
iment. Thus the size of an ensemble is somewhat arbitrary. If the number of
times each outcome occurs in the ensemble is consistent with its probability
it would suffice.

2.7 Calculation of probabilities from an ensemble


Suppose we are given the following ensemble : {H, T, H, H, T }. By looking at
the ensemble, we can conclude that the sample space contains two outcomes
{H, T }.
T
We also find that the outcome H occurs thrice and T occurs twice. Hence
We also conclude that the probability of H is 3/5 and that of T is 2/5.
AF
2.8 Construction of ensemble from probabilities
DR

We can also do the reverse. Given the outcomes and their probabilities, we
can construct an ensemble. Let ni denote the number of times an outcome
i occurs in an ensemble. Let N denote the total number of elements of the
ensemble. Choose ni such that ni /N equals pi ; note that we have assumed
that pi is already known.

2.9 Counting of the elements in events of the sample


space : Coin tossing
Consider tossing of N identical coins or tossing of a single coin N times. Let
us say the coin is fair. In other words P (H) = P (T ) = 1/2.
Let Ω(N ) denote the set of all possible outcomes of the experiment. An
outcome is thus a string N entries, each entry being ”H” or ”T”. The number
b ). We have
of elements of this set Ω(N ) is denoted by the symbol, Ω(N
5
a p-coin is one for which the probability of ’Heads’ is p and that of the ’Tails’ is
q = 1 − p.
20 2. Maxwell’s mischief

b ) = 2N .
Ω(N
Let Ω(n1 , n2 ; N ) denote a subset of Ω(N ), containing only those outcomes
with n1 ’Heads’ and n2 ’Tails’. Note n1 + n2 = N . How many outcomes are
there in the set Ω(n1 , n2 ; N ) ?
b 1 , n2 ; N ) denote the number of elements in the event Ω(n1 , n2 ; N ).
Let Ω(n
b 1 , n2 ; N ).
In what follows I shall tell you how to derive an expression6 for Ω(n
Take one outcome belonging to the event Ω(n1 , n2 ; N ). There will be
n1 ’Heads’ and n2 ’Tails” in that outcome . Imagine for a moment that
all these ’Heads’ are distinguishable. If you like, you can label them as
H1 , H2 , · · · , Hn1 . Carry out pemutation of all the ’Heads’ and produce
n1 ! new configurations. From each of these new configurations, produce n2 !
configurations by carrying out the permutations of the n2 ’Tails’. Thus from
one outcome belonging to Ω(n1 , n2 ; N ), we have produced n1 ! × n2 ! new con-
figurations. Repeat the above for each element of the set Ω(n1 , n2 ; N ), and
b 1 , n2 ; N )n1 !n2 ! configurations. A moment of thought will tell you
produce Ω(n
that this number should be the same as N !. Therefore
b 1 , n2 ; N ) × n1 ! × n2 ! = N !
Ω(n
Let us work out an example explicitly to illustrate the above. Consider
tossing of five coins. There are 25 = 32 outcomes/microstates listed below.
The number in the brackets against each outcome denotes the number of
”Heads” in that outcome. T
AF
Table 2.1. List of 32 outcomes of tossing five coins. Given in brackets is
the number of ”Heads” in each outcome
DR

1. H H H H H (5) 17. T H H H H (4)


2. H H H H T (4) 18. T H H H T (3)
3. H H H T H (4) 19. T H H T H (3)
4. H H H T T (3) 20. T H H T T (2)
5. H H T H H (4) 20. T H T H H (3)
6. H H T H T (3) 21. T H T H T (2)
7. H H T T H (3) 22. T H T T H (2)
8. H H T T T (2) 23. T H T T T (1)
9. H T H H H (4) 25. T T H H H (3)
10. H T H H T (3) 26. T T H H T (2)
11. H T H T H (3) 27. T T H T H (2)
12. H T H T T (2) 28. T T H T T (1)
13. H T T H H (3) 29. T T T H H (2)
14. H T T H T (2) 30. T T T H T (1)
15. H T T T H (2) 31. T T T T H (1)
16. H T T T T (1) 32. T T T T T (0)

The outcomes numbered 4, 6, 7, 10, 11, 13, 18, 19, 20, 25 are the ones
with three ”Heads” and two ”tails”). These are the elements of the event
Ω(n1 = 3; n2 = 2; N = 5).
Take outcome No. 4. Label the three heads as H1 , H2 and H3 . Carry out
permutations of the three ”Heads” and produce 3! = 6 elements. These are
Take an element from the above set. Label the ’Tails’ as T1 and T2 . Carry
6
I remember I have seen this method described in a book on Quantum Mechanics
by Gasiorowicz. Check it out
2.9 Counting of elements of an event 21

Table 2.2. In outcome no. 4, see Table (2.1), the Heads are permuted

(4) H H H T T

H2 H3 H1 T T
H1 H2 H3 T T
H3 H1 H2 T T
H1 H3 H2 T T
H3 H2 H1 T T
H2 H1 H3 T T

out permutation of the ’Tails’ and produce 2! = 2 elements. Do this for each
of the above six elements.
Thus from the outcome No, 4, we have produced 3! × 2! = 12 out-
comes, listed below. Repeat the above exercise on the outcomes numbered

Table 2.3. The twelve outcomes produced from outcome No. 5

(4) H H H T T

H2 H3 H1 T1 T2
H1 H2 H3 T1 T2
H2 H3 H1 T2 T1
H1 H2 H3 T2 T1
H3 H1 H2 T1 T2
H1 H3 H2 T1 T2
H3 H1 H2 T2 T1
H1 H3 H2 T2 T1
H3 H2 H1 T1 T2
H2 H1 H3 T1 T2
H3 H2 H1 T2 T1
H2 H1 H3 T2 T1

T
6, 7, 10, 11, 13, 18, 19, 20, 25. Table below depicts the results for outcome
no. 6.
AF
Table 2.4. The twelve outcomes produced from outcome No. 6
DR

(6) H H T H T

H2 H3 T1 H1 T2
H1 H2 T1 H3 T2
H2 H3 T2 H1 T1
H1 H2 T2 H3 T1
H3 H1 T1 H2 T2
H1 H3 T1 H2 T2
H3 H1 T2 H2 T1
H1 H3 T2 H2 T1
H3 H2 T1 H1 T2
H2 H1 T1 H3 T2
H3 H2 T2 H1 T1
H2 H1 T2 H3 T1

Thus we produce Ω(n b 1 = 3, n2 = 2; N = 5) × n1 ! × n2 ! outcomes.


This number is just the number of permutations of N = 5 objects labelled
H1 , H2 , H3 , T1 , T2 and it equals N !. Therefore,

b 1 = 3, n2 = 2; N = 5) = 5! = 10
Ω(n
3!2!
b 1 , n2 ; N ) is called the binomial coefficient7
Ω(n
7
We have the binomial expansion given by
X⋆ N ! n1 n2
(a + b)N = a b
n1 !n2 !
{n1 ,n2 }
22 2. Maxwell’s mischief

2.10 Gibbs ensemble

Following Gibbs, we can think of an ensemble as consisting of large number


of identical mental copies of a macroscopic system 8 . All the members of an
ensemble are in the same macro state9 . However they can be in different micro
states. Let the micro states of the system under consideration, be indexed
by {i = 1, 2, · · · }. The number of elements of the ensemble in micro state j
divided by the size of the ensemble equals t the probability of the system to
be in micro state j. It is intuitively clear that the size of the ensemble should
be large (→ ∞) so that it can capture exactly the
probabilities of different micro states of the system10 . Let me elaborate
on this issue, see below.

2.11 Why should a Gibbs ensemble be large ?


b 1 , n2 ; N ) is maximum11 ?
What are the values of n1 and n2 for which Ω(n
It is readily shown that for n1 = n2 = N/2 the value of Ω(n b 1 , n2 ; N )
b m
is maximum. Let us denote this number by the symbol Ω (N ). We have,
Ωb m (N ) = Ω(n
b 1 = N/2, N2 = n/2; N ).
Thus we have T
The sum runs over all possible values of {n1 , n2 }. The superscript ⋆ on the
AF
summation sign should remind us that only those values of n1 and n2 consistent
with the constraint n1 + n2 = N are permitted.
8
For example the given coin is a system. Let p denote the probability of ”Heads”
and q = 1−p the probability of ”tails”. The coin can be in a micro state ”Heads”
or in a micro state ”Tails”.
DR

9
This means the values of p and q are the same for all the coins belonging to the
ensemble.
10
If you want to estimate the probability of Heads in the toss of a single coin
experimentally then you have to toss a large number of identical coins. Larger
the size of the ensemble more (statistically) accurate is your estimate .
11
you can find this in several ways. Just guess it. I am sure you would have guessed
the answer as n1 = n2 = N/2. We know that the binomial coefficient is largest
when n1 = n2 = N/2 if N is even, or when n1 and n2 equal the two integers
closest to N/2 for N odd. That is it.
b 1 , n2 ; N ) with respect
If you are more sophisticated, take the derivative of Ω(n
to n1 and n2 with the constraint n1 + n2 = N , set it zero; solve the resulting
equation to get the value of N for which the function is an extremum. Take the
second derivative and show that the extremum is a maximum.
You may find it useful to take the derivative of logarithm of Ω(n b 1 , n2 ; N );
employ Stirling approximation for the factorials : ln(m!) = m ln(m) −
m for large m. Stirling approximation to large factorials is described in the
next section.
You can also employ any other pet method of yours to show that, for n = N/2,
b
the function Ω(n; N ) is maximum.
2.11 Why should Gibbs ensemble be large ? 23
X⋆ N!
b )=
Ω(N = 2N ; (⋆ ⇒ n1 + n2 = N ) (2.3)
n1 ! n2 !
{n1 ,n2 }

b m (N ) = Ω(n
b 1 = n2 = N/2; N ) = N!
Ω (2.4)
(N/2)! (N/2)!
b m (N ) for large values of N . We employ Stirling’s first
Let us evaluate Ω
approximation : N ! = N N exp(−N ) and get
12

b m (N ) = =  N N exp(−N ) N
Ω 2 = 2 (2.5)
(N/2)(N/2) exp(−N/2)
The above implies that almost all the outcomes of the experiment belong
to the event with equal number of ’Heads’ and ’Tails’. The outcomes with
unequal number of ’Heads’ and ’Tails’ are so overwhelmingly small that the
difference falls within the small error arising due to the first Stirling approx-
imation.
For estimating the tiny difference between Ω(Nb ) and Ω b m (N ), let us em-
13
ploy the second Stirling approximation :
12
First Stirling Approximation : N ! ≈ N N exp(−N ).
We have, N ! = N × (N − 1) × · · · × 3 × 2 × 1. Therefore,
ln N ! = ln 1 + ln 2 + ln 3 + · + ln N T
XN Z N
N
= ln(k) ≈ ln x dx = (x ln x − x) 1 = N ln N − N − 1
AF
k=1 1

≈ N ln N − N
N ! ≈ N N exp(−N )
DR

13
A better approximation√ to large factorials is given by Stirling’s second formula
: N ! ≈ N N exp(−N ) 2πN . We have
Z ∞ Z ∞
Γ (N + 1) = N ! = dx xN e−x = dx eF (x)
0 0
′ N N
F (x) = N ln x − x; F (x) = − 1 ; F ′′ (x) = − 2
x x
Set F ′ (x) = 0; this gives x⋆ = N . At x = x⋆ the function F (x) is maximum.
(Note : F ′′ (x = x⋆ ) is negative). Carrying out Taylor expansion and retaining
only upto quadratic terms, we get,
(x − x⋆ )2 ′′ (x − N )2
F (x) = F (x⋆ ) + F (x = x⋆ ) = N ln N − N −
2 2N
We have,
Z ∞ Z 
∞ 
(x − N )2
N! = dx eF (x) = N N e−N dx exp −
0 0 2N
√ Z ∞ √
= N N e−N N √ dx exp(−x2 /2) ∼
N→∞ N N e−N 2πN
− N
24 2. Maxwell’s mischief

N ! = N N exp(−N ) 2πN

and get
b m (N ) = Ω(n
Ω b 1 = n2 = N/2; N )
r
N 2
=2 ×
πN
r
b 2
= Ω(N ) × (2.6)
πN
b 1 , n2 ; N ) as entropy of the event
Let us define S(Ω(n1 , n2 ; N )) = ln Ω(n
Ω(n1 , n2 ; N ). For n1 = n2 = N/2 we get an event with largest entropy. Let
us denote it by the symbol SB (N ).
Let SG = S(Ω(N )) = N ln 2 denote the entropy of the sure event. It
is the logarithm of the number of outcomes of the experiment of tossing N
independent and fair coins.
Taking logarithm of the terms on both sides of Eq. (2.6) we get,
1 1
SB = SG − ln N − ln(2/π) (2.7)
2 2
∼ 1
N →∞ SG − ln N (2.8)
2
T
The entropy of the sure event is of the order of N ; the entropy of the event
with N/2 ”Heads” and N/2 ’Tails” is less by an extremely small quantity
AF
of the order of ln(N ). Hence when you toss a very large number of coins
independently you will get almost always N/2 ’Heads’ and N/2 ’Tails”.
For example take a typical value for N = 1023 . We have SG = 0.69 × 1023
and SB = 0.69 × 1023 − 24.48.
DR

Note that only when N is large, we have SB equals SG . It is precisely be-


cause of this, we want the number of elements to be large, while constructing
a Gibbs ensemble. We should ensure that all the micro states of the system
are present in the ensemble in proportions, consistent with their probabilities.
For example I can simply toss N independent fair coins just once and if
N is large then I am assured that there shall be (N/2) ± √ ǫ ’Heads’ and
(N/2) ∓ ǫ ’Tails’, where ǫ is negligibly small : of the order of N . Consider
the probability for random variable n to take values outside the interval
 
N N N N
−ǫ , +ǫ
2 2 2 2

where ǫ is an arbitrarily small number. Let us denote this probability as


Pout (N ). One can show that in the limit N → ∞, the probability Pout (N )
goes to zero.
Take ǫ = 1/100 and calculate Pout (N ) for N = 103 , 104 , 105 , 106 , and
8
10 . The table below depicts the results, and we see that Pout nearly zero for
large N .
2.11 Why should Gibbs ensemble be large ? 25

Table 2.5. Probability that in a toss of N coins, the number of Heads deviates by
more than one percent from N/2

ǫ =
1/100; Pout = 
N N N N
N 1-P 2
−ǫ 2
≤n≤ 2
+ǫ 2

103 0.752
4
10 0.317
5
10 0.002
10 6
1.5 × 10−23
107 2.7 × 10−2174

In statistical mechanics we consider a macroscopic system which can be


in any of its numerous microscopic states. These micro states are analogous
to the outcomes of an experiment. In fact the system switches spontaneously
from one micro state to another when in equilibrium. We can say the set
of all micro states constitute a micro state space which is the analogue of
T
the sample space. We attach a real number representing the numerical value
of a property e.g. energy, to each micro state. This is analogous to random
AF
variable. We can define several random variables on the same micro state
space to describe different macroscopic properties of the system.
DR
DR
AF
T
3. Binomial, Poisson, and Gaussian

3.1 Binomial Distribution


Consider a system consisting of one coin. It has two micro states : H and T.
The probability for the system to be in micro state H is p and that in micro
state T is q = 1 − p.
Consider the case with p = 0.6 and hence q = 1 − p = 0.4. A possible
Maxwell ensemble of micro states is A popular book. Starts with a b

{T, H, H, H, T, H, H, T, H, T }.

Notice that the ensemble contains ten elements. Six elements are H and four
are T. This is consistent with the given probabilities: P(H) = 6/10; and
T
P(T ) = 4/10.
However a Gibbs ensemble is constructed by actually tossing N identical
AF
and independent coins. In the limiet N → ∞, sixty percent of the coins shall
be in micro state H and forty, T. To ensure this we need to take the size of
the ensemble N , to be very large. How large ? You will get an answer to this
question in what follows.
DR

Let us say, we attempt to construct the ensemble by actually carrying out


the experiment of tossing identical coins or by tossing the same coin several
times independently. What is the probability that in the experiment there
shall be n1 ’Heads’ and hence n2 = (N − n) ’Tails’ ? Let us denote this by the
symbol B(n1 , n2 ; N ), where N is the number of independent identical coins
tossed or number of times a coin is tossed independently. It is readily seen,
N!
B(n1 , n2 ; N ) = pn1 q n2 ; n1 + n2 = N. (3.1)
n1 ! n2 !
B(n1 , n2 ; N ) is called the Binomial distribution. Let n1 = n, n2 = N − n, we
can write the Binomial distribution for the single random variable n as,
N!
B(n; N ) = pn q N −n
n!(N − n)!
Figure (3.1) depicts Binomial distribution for N = 10, p = 0.5 (Left) and
0.35 (Right). First moment of n : What is average value of n ? The
average, also called the mean, the first moment, the expectation value etc. is
denoted by the symbol hni and is given by,
28 3. Binomial, Poisson, and Gaussian

0.25

0.3
0.2
0.25
0.15
B(n)

B(n)
0.2

0.1 0.15

0.1
0.05
0.05

0 0
0 5 10 0 5 10
n n
N!
Fig. 3.1. Binomial distribution : B(n; N ) = pn (1 − p)N−n with N = 10;
n!(N − n)!
B(n; N ) versus n; depicted as sticks; (Left) p = 0.5; (Right) p = .35.

N
X N
X N!
hni = n B(n; N ) = n T pn q N −n ,
n=0 n=1
n! (N − n)!
N
AF
X (N − 1)!
= Np pn−1 q N −1−(n−1) ,
n=1
(n − 1)! [N − 1 − (n − 1)]!

N
X −1
(N − 1)!
DR

= Np pn q N −1−n ,
n=0
n!(N − 1 − n)!
= N p(p + q)N −1 = N p.
Second factorial moment of n : The second factorial moment of n
is defined as hn(n − 1)i. It is calculated as follows.
N
X
hn(n − 1)i = n(n − 1)B(n; N ),
n=0
N
X N!
= n(n − 1) pn q N −n ,
n=2
n!(N − n)!
N
X (N − 2)!
= N (N − 1)p2 pn−2 q (N −2)−(n−2) ,
n=2
(n − 2)![(N − 2) − (n − 2)!
N
X −2
(N − 2)!
= N (N − 1)p2 pn q (N −2)−n .
n=0
n![(N − 2) − n]!
3.2 Moment Generating Function 29

= N (N − 1)p2 (q + p)N −2 = N (N − 1)p2 . (3.2)


Moments of n : We can define higher moments. The k-th moment is
defined as
N
X
Mk = hnk i = nk B(n). (3.3)
n=0

Variance of n : An important property of the random variable is variance.


It is defined as,
N
X N
X
σn2 = (n − M1 )2 B(n), = n2 B(n) − M12 , = M2 − M12 . (3.4)
n=0 n=0

We have,
hn(n − 1)i = N (N − 1)p2 ,
hn2 i − hni = N 2 p2 − N p2 ,
hn2 i = N 2 p2 − N p2 + N p,
σn2 = hn2 i − hni2 = N pq. (3.5)
The square-root of variance is called the standard deviation. A relevant quan-
tity is the relative standard deviation. It is given by the ratio of the standard

σn 1 q
T
deviation to the mean. For the Binomial random variable, we have,
r
AF
= √ . (3.6)
hni N p

The relative standard deviation is inversely proportional to N . It is small
for large N . It is clear that the number of elements N , in a Gibbs ensemble
DR

should be large enough to ensure that the relative standard deviation is as


small as desired. Let me now describe a smart way of generating the moments
of a random variable.

3.2 Moment Generating Function

Let B(n) denote the probability that n coins are in micro state ”Heads” in
an ensemble of N coins. We have shown that,
N!
B(n) = pn q N −n . (3.7)
n!(N − n)!
The moment generating function is defined as
N
X
e
B(z) = z n B(n). (3.8)
n=0
30 3. Binomial, Poisson, and Gaussian

The first thing we notice is that B(ze = 1) = 1. This guarantees that the
probability distribution B(n) is normalised. The moment generating func-
tion is like a discrete transform of the probability distribution function. We
transform the variable n to z.
Let us now take the first derivative of the moment generating function
with respect to z. We have,
e XN
dB e ′ (z) =
=B n z n−1 B(n),
dz n=0

N
X
e ′ (z) =
zB n z n B(n). (3.9)
n=0
. Substitute in the above z = 1. We get,
e ′ (z = 1) = hni.
B (3.10)
e evaluated at z = 1 generates the first moment.
Thus the first derivative of B
e
Now take the second derivative of B(z) to get
e XN
d2 B
= n(n − 1)z n−2 B(n),
dz 2 n=0

z2
e
d2 B
dz 2
=
XN
z n n(n − 1) B(n).
T (3.11)
AF
n=0
Substitute in the above z = 1 and get,

e
d2 B = hn(n − 1)i, (3.12)
dz 2
DR

z=1
For the Binomial random variable, we can derive the moment generating
function :
XN XN
e n N!
B(z) = z B(n) = (zp)n q N −n , = (q + zp)N . (3.13)
n=0 n=0
n! (N − n)!
The moments are generated as follows.
dBe
= N (q + zp)N −1 p, (3.14)
dz

dBe
hni = = N p, (3.15)
dz z=1

e
d2 B
= N (N − 1)(q + zp)N −2 p2 , (3.16)
dz 2

e
d2 B
hn(n − 1)i = = N (N − 1)p2 . (3.17)
dz 2 z=1
3.3 Binomial → Poisson 31

3.3 Binomial → Poisson

When N is large, it is clumsy to calculate quantities employing Binomial


distribution. Consider the following situation.
I have N molecules of air in this room of volume V . The molecules are dis-
tributed uniformly in the room. In other words the number density, denoted
by ρ is same at all points in the room. Consider now an imaginary small
volume v < V completely contained in the room. Consider an experiment
of choosing randomly an air molecule from this room. The probability that
the molecule shall be in the small volume is p = v/V ; the probability that
it shall be out side the small volume is q = 1 − (v/V ). There are only two
possibilities. We can use Binomial distribution to calculate the probability
for n molecules to be present in v.
Consider first the problem with V = 10M 3 , v = 6M 3 and N = 10. The
value of p for the Binomial distribution is 0.6. The probability of finding n
molecules in v is then,
10!
B(n; N = 10) = (0.1)n (0.9)10−n . (3.18)
n!(10 − n)!
The table below gives the probabilities calculated from the Binomial distri-
bution.
T
AF
Table 3.1. Probabilities calculated from Binomial distribution :
B(n; N = 10, p = .1)
DR

n B(n; 10) n B(n; 10)

0 0.0001 6 0.2508
1 0.0016 7 0.2150
2 0.0106 8 0.1209
3 0.0425 9 0.0403
4 0.1115 10 0.0060
5 0.2007 − −

Consider the same problem with v = 10−3 M 3 and N = 105 . We have


p = 10−4 and N p = 10. Immediately we recognise that Binomial distribution
is not appropriate for this problem. Calculation of the probability of finding
n molecules in v involves evaluation of factorial of 100000.
32 3. Binomial, Poisson, and Gaussian

What is the right distribution for this problem and problems of this kind ?
To answer this question, consider what happens to the Binomial distribution
in the limit of N → ∞, p → 0, and N p = µ, a constant1 . Note that

N p = N v/V = ρv = constant.

We shall show below that in this limit, Binomial goes over to Poisson
distribution.

3.4 Poisson Distribution

We start with
e
B(z) = (q + zp)N . (3.19)
We can write the above as
 N
e zp
B(z) = qN 1 + ,
q

 N
zp
= (1 − p)N 1+ , T (3.20)
q
AF
 N  N
1 zN p 1
= 1 − Np 1+ . (3.21)
N q N
In the above replace N p by µ and q by 1 to get,
DR

 µ N  zµ N
e
B(z) = 1− 1+ .
N N
In the limit of N → ∞ we have by definition2 ,
e
B(z) ∼ exp(−µ) exp(zµ),

= Pe (z). (3.22)
e
Thus in the limit N → ∞, p → 0 and N p = µ, we find B(z) → Pe (z), given
by
Pe (z) = exp[−µ(1 − z)]. (3.23)
1
Note that for a physicist, large is infinity and small is zero.
2
exponential function is defined as
 x N
limit
exp(x) = N→∞ 1+
N
3.4 Poisson Distribution 33

The coefficient of z n in the power series expansion of Pe (z) gives P (n),


µn
P (n) = exp(−µ). (3.24)
n!
The above is called the Poisson distribution 3 . Thus in the limit of N → ∞,
p → 0, N p = µ, the Binomial distribution goes over to Poisson distribution.
Figure (3.2) depicts Poisson distribution for µ = 1.5 and 9.5.

0.35 0.14

0.3 0.12

0.25 0.1

0.2 0.08

0.15 0.06

0.1 0.04

0.05 0.02

0 0
−5 0 5 0 5 T 10 15 20
AF
Fig. 3.2. Poisson distribution with mean µ, depicted as sticks. Gaussian distribu-
tion with mean µ and variance σ 2 = µ depicted by continuous line.
(Left) µ = 1.5; (Right) µ = 9.5. For large µ Poisson and Gaussian coincide
DR

3.4.1 Binomial → Poisson à la Feller

Following Feller 4 , we have


3
We shall come across Poisson distribution in the context of Maxwell-Boltzmann
statistics. Let nk denote the number of ’indistinguishable classical’ particles in a
single-particle state k. The random variable nk is Poisson-distributed.
4
William Feller, An Introduction to PROBABILITY : Theory and its Applica-
tions, Third Edition Volume 1, Wiley Student Edition (1968)p.153
34 3. Binomial, Poisson, and Gaussian

B(n; N ) N ! pn q N −n (n − 1)! (N − n + 1)!


= ,
B(n − 1; N ) n! (N − n)! N ! pn−1 q N −n+1
p (N − n + 1)
= ,
nq
N p − p (n − 1)
= ,
nq
 
N →∞
N p=µ
µ
p→0 ∼ . (3.25)
n
Thus we get for large N , B(n; N ) = B(n − 1; N )µ/n.
Start with B(n = 0; N ) = q N . We have
qN = (1 − p)N (3.26)
 N
Np
= 1− (3.27)
N
 µ N
= 1− (3.28)
N
N →∞
∼ exp(−µ). (3.29)
Thus for N → ∞, we get B(n = 0; N ) = P (n = 0; µ) = exp(−µ). We get,
P (n = 1; N ) = µ exp(−µ),
T (3.30)
AF
µ2
P (n = 2; N ) = exp(−µ), (3.31)
2!
µ3
P (n = 3; N ) = exp(−µ). (3.32)
3!
DR

Finally prove by induction


µn
P (n; N ) = exp(−µ).
n!
The next item in the agenda is on Gaussian distribution. It is a continuous
distribution defined for −∞ ≤ x ≤ +∞. Before we take up the task of
obtaining Gaussian from Poisson (in the limit µ → ∞), let us learn a few
relevant and important things about continuous distribution.

3.5 Characteristic Function


Let x = X(ω) be a continuous random variable, and f (x) its probability
density function. The Fourier transform of f (x) is called the characteristic
function of the random variable x = X(ω) :
Z +∞
φX (k) = dx exp(−ikx) f (x).
−∞
3.7 Central Limit Theorem 35

Taylor expanding the exponential in the above, we get


X∞ Z X∞
(−ik)n ∞ (−ik)n
φX (k) = dx xn f (x) = Mn . (3.33)
n=0
n! −∞ n=0
n!
Thus the characteristic function generates the moments.

3.6 Cumulant Generating Function


The logarithm of the characteristic function is called the cumulant generating
function. ψX (k) = ln φX (k). Let us write the above as,

!
X (−ik)n
ψX (k) = ln 1 + Mn ,
n=1
n!
∞ ∞
!n
X (−1)n+1 X (−ik)m
= Mm . (3.34)
n=1
n m=1
m!
We now express ψX (k) as a power series in k as follows
X∞
(−ik)n
ψX (k) = ζn . (3.35)
n=1
n! T
where ζn is called the n-th cumulant.
AF
From the above equations we can find the relation between moments and
cumulants.
DR

3.7 The Central Limit Theorem


Let {Xi : i = 1, 2, · · · , N } denote N random variables and let
N
X
Y = Xi (3.36)
i=1

be their sum. Note Y is a random variable and its probabiity density function
is formally given by,
Z +∞ Z +∞
fY (y) = dx1 dx2 · · ·
−∞ −∞

Z N
!
+∞ X
··· dxN f (x1 , x2 , · · · , xN ) δ y− xi (3.37)
−∞ i=1

If the random variables are mutually independent then their joint density can
be written as the product of their individual probability densities, see below.
36 3. Binomial, Poisson, and Gaussian

f (x1 , x2 , · · · , xN ) = f1 (x1 )f2 (x2 ) · · · fN (xN ) (3.38)


where the symbol fi (xi ) denotes the probability density function of the ran-
dom variable Xi . Therefore we have,
Z +∞ Z +∞
fY (y) = dx1 f1 (x1 ) dx2 f2 (x2 ) · · ·
−∞ −∞

Z N
!
+∞ X
··· dxN fN (xN ) δ y − xi (3.39)
−∞ i=1

Now, multiply both sides of the equation above by exp(−iky) and integrate
over y from −∞ to +∞. The Left Hand Side gives the characteristic function
of the random variable Y .
Z +∞
dy exp(−iky) fY (y) = φY (k) (3.40)
−∞

Thus we get,
Z +∞ Z +∞ Z +∞
φY (k) = dy exp(−iky) dx1 f1 (x1 ) dx2 f2 (x2 ) · · ·
−∞ −∞ −∞

···
Z +∞
dxN fN (xN ) δ y −
T
N
X
xi
!
(3.41)
AF
−∞ i=1

In the right hand side of the above, the integral over y can be carried out
making use of a property of Dirac-delta function5 , and we get,
DR

5
Z +∞
dx f (x) δ(x − x0 ) = f (x0 )
−∞
3.7 Central Limit Theorem 37
Z +∞ Z +∞
φY (k) = dx1 f1 (x1 ) dx2 f2 (x2 ) · · ·
−∞ −∞

Z +∞
··· dxN fN (xN ) exp[−ik(x1 + x2 + · · · + xN )]
−∞

Z +∞ Z +∞
= dx1 exp(−ikx1 )f1 (x1 ) dx2 exp(−ikx2 )f2 (x2 ) · · ·
−∞ −∞

Z +∞
··· dxN exp(−ikxN )f (xN )
−∞

= φ1 (k) × φ2 (k) × · · · × φN (k) (3.42)


where φi (k) is the characteristic function of Xi and it is given by,
Z +∞
φi (k) = dx exp(−ikx)fi (x) (3.43)
−∞
If the random variables are also identically distributed, then we have
N
φY (k) = [φX (k)] T
where φX (k) is the characteristic function of the common distribution.
(3.44)
AF
Now consider the scaled random variable,
X1 + X2 + · · · + XN
Y = (3.45)
N
DR

The characteristic function of Y can be obtained following the procedure


described above except that we have the scaled variable k/N . Thus,
N
φY (k) = [φX (k → k/N )]
= exp [N ln φX (k → k/N )] ,
" ∞ #
X (−ik)n ζn
= exp N
n=1
n! N n
"∞ #
X (−ik)n ζn
= exp ,
n=1
n! N n−1
 
k2 σ2 2
= exp −ikµ − + O(1/N )
2! N
 
∼ k2 σ2
N →∞ exp −ikµ − . (3.46)
2! N
Thus the characteristic function of Y , in the limit N → ∞ is given by φY (k) =
exp(−ikµ − (k 2 /2!)σ 2 /N ). This is indeed the characteristic function of the
Gaussian random variable with mean µ and variance σ 2 /N , see below.
38 3. Binomial, Poisson, and Gaussian

Thus the sum of N independent and identically distributed random vari-


ables (with finite variance) tends to have a Gaussian distribution for large
N . This is called the central limit theorem.

3.8 Poisson → Gaussian


Let us show now that for large values of the mean the Poisson distribution
tends to a Gaussian. Start with the moment generating function of the Pois-
son random variable:

Pe(z; µ) = exp[−µ(1 − z)].

Substitute z = exp(−ik) and get,

Pe(k; µ) = exp [−µ {1 − exp(−ik)}] .

Carry out the power series expansion of the exponential function and get,
"∞ #
X (−ik)n
Pe (k; µ) = exp µ . (3.47)
n=1
n!
We recognize the above as the cumulant expansion of a distribution for which
T
all the cumulants are the same µ. For large value value of µ, it is adequate to
AF
consider only small values of k. Hence we retain only terms upto quadratic
in k. Thus for k small, we have,
 
e k2
P (k) = exp −ikµ − µ . (3.48)
2!
DR

The above is the Fourier transform or the characteristic function of a Gaus-


sian random variable with mean as µ and variance also as µ.
Thus in the limit µ → ∞, Gaussian distribution with mean and variance
both equal to µ is a good approximation to Poisson distribution with mean
µ, see Fig. 2.

3.9 Gaussian
A Gaussian of mean µ and variance σ 2 is given by
 
1 (x − µ)2
G(x) = √ exp − . (3.49)
σ 2π 2σ 2
The characteristic Rfunction is given by the Fourier transform formally ex-
e +∞
pressed as G(k) = −∞ dx exp(−ikx)G(x). The integral can be worked out,
 
e
and I leave it as an exercise. We get, G(k) = exp −ikµ − k 2 σ 2 /2 . Consider
a Gaussian of mean zero and variance σ 2 . It is given by
3.9 Gaussian 39

1 1 x2
g(x) = √ exp − 2 . (3.50)
σ 2π 2σ
The width of the Gaussian distribution is 2σ. The Fourier transform of g(x)
is denoted ge(k) and is given by
 
1
g (k) = exp − k 2 σ 2 .
e (3.51)
2
The Fourier transform is also a Gaussian with zero mean and standard devi-
ation 1/σ 2 . The width of e
g(k) is 2/σ. The product of the width of g(x) and
the width of its Fourier transform eg(k) is 4.
If g(x) is sharply peaked then its Fourier transform ge(k) will be broad
and vice versa.

T
AF
DR
DR
AF
T
4. Isolated System: Micro canonical Ensemble

4.1 Preliminaries

Our aim is to study an isolated system of N number of point particles with


an energy E and confined to a volume V . The particle do not interact with
each other. We shall see how to count the number of micro states, denoted
by the symbol Ω,b of the system. Ωb will be, in general, a function of energy E,
volume V and the number of particles N . We shall carry out both classical
as well as quantum counting of the micro states and find that both lead to
the same expressions for entropy.
Before we address the full problem, we shall consider a simpler problem
of counting the micro states taking into account only the spatial coordinates
T
neglecting completely the momentum coordinates. Despite this gross simpli-
fication, we shall discover that the machinery of statistical mechanics helps
AF
us derive the ideal gas law1 .
1
I must tell you of a beautiful derivation of the ideal gas law by Daniel Bernoulli
(1700-1782). It goes as follows. Bernoulli imagined air to be made of billiard
balls all the time in motion, colliding with each other and with the walls of
DR

the container. When a billiard ball bounces off the wall, it transmits a certain
momentum to the wall and Bernoulli imagined it as pressure. It makes sense.
First consider air contained in a cube of side one meter. There is a certain amount
of pressure felt by the wall. Now imagine the cube length to be doubled with
out changing the speeds of the molecule. In modern language this assumption
is the same as keeping the temperature constant. The momentum transferred
per collision remains the same. However since each billiard ball molecule has
to travel twice the distance between two successive collisions with the wall, the
force on the wall should be smaller by a factor of two. Also pressure is force per
unit area. The area of the side of the cube is four times more now. Hence the
pressure should be less by a further factor of four. Taking into account both these
factors, we find the pressure should be eight times less. We also find the volume
of cube is eight times more. From these considerations, Bernoulli concluded that
the product of pressure and volume must be a constant when there is no change
in the molecular speeds - a brilliant argument based on simple scaling ideas.
42 4. Isolated System: Micro canonical Ensemble

4.2 Configurational Entropy


Consider placing a single particle in a volume V divided into two equal halves.
Let ǫ = V /2. There are two ways, see Fig. (4.1). We have,

Fig. 4.1. Two ways of keeping a particle in a box divided into two equal parts.
ǫ = V /2

b V b = kB ln(2).
Ω(V, N = 1, ǫ = V /2) = = 2 and, S = kB ln Ω (4.1)
ǫ
Now consider distributing two distinguishable particles in these two cells each
of volume ǫ = V /2, see Fig. (4.2). We then have

T
AF
DR

Fig. 4.2. Four ways of keeping two distinguishable particles in a box divided into
two equal halves. ǫ = V /2.

 2
b V b = 2kB ln(2). (4.2)
Ω(V, N = 2, ǫ = V /2) = = 4 andS = kB ln Ω
ǫ
For N particles we have,
 N
b V b = N kB ln(2). (4.3)
Ω(V, N, ǫ = V /2) = = 2N , and S = kB ln Ω
ǫ
Let us now divide the volume equally into V /ǫ parts and count the number of
 N
b V
ways or organizing N (distinguishable) particles. We find Ω(V, N) = ,
ǫ
and
4.3 Ideal Gas Law : Derivation 43

b = N kB ln(V /ǫ)
S = kB ln Ω
= N kB ln V − N kB ln ǫ. (4.4)

4.3 Ideal Gas Law : Derivation

Differentiate S given by Eq. (4.4), with respect V . We get,


 
∂S N kB
= . (4.5)
∂V E,N V
From thermodynamics2 we have
 
∂S P
= . (4.6)
∂V E,N T
Thus we see from Eq. (4.5) and Eq. (4.6)
P V = N kB T. (4.7)
2
In thermodynamics we start with U ≡ U (S, V ) for a given quantity of say an
ideal gas. We have formally,
   
∂U ∂U
dU = dS + dV.
∂S V ∂V S T
We identify the physical meaning of the partial derivatives as follows.
AF
Thus internal energy can change in a (quasi static) reversible process either
by heat (= T dS), or by work (= −P dV ). Hence we have the first law of thermo-
dynamics
dU = T dS − P dV.
DR

We have then,
   
∂U ∂U
T = ; P =− .
∂S V ∂V S
Let us now consider S ≡ S(U, V ), a natural starting point for statistical
mechanics. We have,
   
∂S ∂S
dS = dU + dV
∂U V ∂V U
To express the partial derivatives in the above in terms of T and P , we rearrange
the terms in the first law equation (4.6) as,
1 P
dS = dU + dV.
T T
Equating the pre-factors of dU and dV in the above two equation, we get,
   
∂S 1 ∂S P
= ; = .
∂U V T ∂V U T
44 4. Isolated System: Micro canonical Ensemble

4.4 Boltzmann Entropy −→ Clausius’ Entropy

From Eq. (4.4), we have,


N kB
dS = dV. (4.8)
V
Employing the equation of state : P V = N kB T , which we have derived, we
can rewrite the above as
P dV
dS = . (4.9)
T
Consider an isothermal process in an ideal gas. We have dU = 0. This implies
T dS = P dV . When the system absorbs a certain quantity of heat q isother-
mally and reversibly, we have q = T dS = P dV . Equating P dV to q in Eq.
(4.9), we get
q
dS = , (4.10)
T
which shows that Boltzmann entropy is consistent with the thermodynamic
entropy.

4.5 Some Issues on Extensitivity of Entropy T


AF
The expression for entropy given below,
S(V, N ) = N kB ln V − N kB ln ǫ, (4.11)
is not extensive. If I double the value of V and of N , I expect S to be doubled.
DR

It does not. Mathematically S is extensive if it is a first order homogeneous


function V and N . In other words we should have S(λV, λN ) = λS(V, N ).
The above expression for entropy does not satisfy this rule. This is called
Gibbs’ paradox. More precisely Gibbs formulated the paradox in terms of
entropy of mixing of like and unlike gases. We shall see these in details later
when we consider closed system described by canonical ensembles. .

4.6 Boltzmann Counting

To restore the extensive property of entropy, Boltzmann introduced an ad-hoc


notion of indistinguishable particles. N ! permutations of the particles, should
all be counted as one micro state since they are indistinguishable. Hence,
 N
b 1 V
Ω(V, N ) = . (4.12)
N! ǫ
4.7 Micro canonical Ensemble 45

b
S(V, N ) = kB ln Ω(V, N ),

= kB [N ln V − N ln N + N − N ln ǫ] ,

 
V
= N kB ln + N kB − N kB ln ǫ. (4.13)
N
In the above, I have expressed ln N ! = N ln N − N , employing Stirling’s first
formula for a large factorial. We find that this prescription of Boltzmann
restores the extensivity of entropy; i.e. we find
S(λV, λN ) = λS(V, N ). (4.14)
Boltzmann counting, at best, can be considered as a patch work. You don’t
pull down a well-built wall because there is a small crack in it. Instead you
cover the crack by pasting a paper over it. A good formalism is not dismissed
because of a flaw3 . You look for a quick fix. Boltzmann counting provides
one. In fact non extensivity of entropy is a pointer to a deeper malady. The
fault is not with statistical mechanics but with classical formalism employed
to describe ideal gas. For the correct resolution of the Gibbs’ paradox we
have to wait for the arrival of quantum mechanics.

4.7 Micro canonical Ensemble


T
AF
Time has come for us to count the micro states of an isolated system of N non
interacting point particles confined to a volume V , taking into considerations
DR

the positions as well as the momenta of all the particles.


Each particle for its specification requires six numbers : three positions
and three momenta. The entire system can be specified by a string of 6N
numbers. In a 6N dimensional phase space the system is specified by a point.
The phase space point is all the time moving. We would be interested de-
termining the region of the phase space accessible to the system when it is
in equilibrium. If we are able to count the phase space volume, then we can
employ the first micro-macro connection proposed by Boltzmann and get an
expression for entropy as a function of energy, volume and the number of
particles.
The system is isolated. It does not transact energy with the surround-
ings; not does it transact matter. Hence its energy remains a constant. The
3
Desparate and often elaborate patch work are not new to physicists. They have
always indulged ’papering’ when cracks appear in their understanding of science.
A spectacular example is the entity aether proposed to justify the wave nature of
light; Maxwell’s work showed light is a wave. Waves require medium for propa-
gation. Hence the medium aether, with exotic properties, was porposed to carry
light.
46 4. Isolated System: Micro canonical Ensemble

potential energy is zero since the particles do not interact with each other.
The kinetic energy is given by
3N
X p2i
E= . (4.15)
i=1
2m
The system is thus confined to the surface of a 3N dimensional sphere. We
need a formula for the volume of an hyper-sphere in 3N dimensions. To this
end we need to know of Heaviside4 theta function and Dirac5 delta function.

4.8 Heaviside and his Θ Function

Define a function
 ǫ

 0 for −∞ ≤ x ≤ − ;

 2







 
1 1 ǫ ǫ
f (x; ǫ) = x + for − ≤ x ≤ + ; (4.16)

 ǫ 2 2 2







 ǫ

1 for + ≤ x ≤ +∞.
2
T
AF
where ǫ > 0.
Define
lim.
Θ(x) = ǫ → 0 f (x; ǫ).
DR

Θ(x) is called the step function, Heaviside step function, unit step function
or theta function. It is given by,

 0 for −∞ ≤ x < 0;
Θ(x) = (4.17)

1 for 0 < x ≤ +∞.
Figure (4.3) depicts f (x; ǫ) for ǫ = 2, 1, 1/2 and the theta function obtained
in the limit of ǫ → 0.

4.9 Dirac and his δ Function

Start with the function f (x; ǫ) defined by Eq. (4.16). Take the derivative of
the function. We find that the derivative is 1/ǫ, when −ǫ/2 < x + ǫ/2 and
zero otherwise, Define,
4
Oliver Heaviside(1850-1925)
5
Paul Adrien Maurice Dirac(1902-1984)
4.9 Dirac and his δ Function 47

1 1
0.8 0.8
0.6 0.6
0.4
0.4
0.2
0.2
0
0
−2 −1 0 1 2 −2 −1 0 1 2

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 T 0
−2 −1 0 1 2 −2 −1 0 1 2
AF
Fig. 4.3. f (x; ǫ) versus x. (Top Left) ǫ = 2; (Top Right) ǫ = 1; (Bottom Left)
ǫ = 1/2; (Bottom Right) Θ(x)
DR


 0 for −∞ ≤ x < −ǫ/2;




df 
g(x; ǫ) = = 1 for −ǫ/2 < x + ǫ/2; (4.18)
dx 
 ǫ




0 for +ǫ/2 < x ≤ +∞.
Figure(4.4) depicts g(x; ǫ) for ǫ = 2, 1, 1/2, 1/4.
The Dirac-delta function is defined as,
δ(x) = limit
ǫ→0 g(x; ǫ). (4.19)
Consider the following integral.
Z +∞
I= dx g(x; ǫ). (4.20)
−∞

We find that the integral is the same for all values of ǫ. This gives us an
important property of the Dirac-delta function:
48 4. Isolated System: Micro canonical Ensemble

4 4

3 3

2 2

1 1

0 0
−2 −1 0 1 2 −2 −1 0 1 2

4 4

3 3

2 2

1 1

0 T 0
−2 −1 0 1 2 −2 −1 0 1 2
AF
df
Fig. 4.4. g(x; ǫ) = versus x. (Top Left) ǫ = 2; (Top Right) ǫ = 1; (Bottom
dx
Left) ǫ = 1/2; (Bottom Right) ǫ = 1/4
DR

Z +∞
dx δ(x) = 1. (4.21)
−∞

4.10 Area of a Circle

Let us demonstrate how to use the theta function and delta function to derive
an expression for a circle of radius R. Let us denote the area of a circle by
the symbol V2 (R) - the ’volume’ of a two dimensional ’sphere’ of radius R.
A little thought will tell you,
Z +∞ Z +∞ 2
!
X
2 2
V2 (R) = dx1 dx2 Θ R − xi . (4.22)
−∞ −∞ i=1

Let
yi = xi /R for i = 1, 2.
4.10 Area of a Circle 49

Then,
Z Z 2
!
+∞ +∞ X
2 2
V2 (R) = R dy1 dy2 Θ R (1 − yi2 . (4.23)
−∞ −∞ i=1
We have
Θ(λx) = θ(x) ∀ λ > 0.
Therefore,
Z Z 2
!
+∞ +∞ X
2
V2 (R) = R dy1 dy2 Θ 1 − yi2 (4.24)
−∞ −∞ i=1

= R2 V2 (R = 1). (4.25)
We can now write Eq. (4.22) as
Z +∞ Z 2
!
+∞ X
2
V2 (R = 1)R = dx1 dx2 Θ R2 − x2i . (4.26)
−∞ −∞ i=1
Now differentiate both sides of the above equation with respect to the variable
R. We have already seen that the derivative of a Theta function is the Dirac-
delta function. Therefore
Z +∞ Z +∞ 2
!
X
V2 (R = 1)2R = 2R dx1
−∞
dx2 δ R −2
−∞
2
xi . T (4.27)
i=1
AF
Now multiply both sides of the above equation by exp(−R2 )dR and integrate
over the variable R from 0 to ∞. We get,
Z ∞ Z ∞ Z +∞
2 2
V2 (R = 1) exp(−R )2RdR = exp(−R ) 2RdR dx1
DR

0 0 −∞

Z 2
!
+∞ X
2
dx2 δ R − x2i . (4.28)
−∞ i=1
Z ∞ Z ∞ Z ∞
V2 (R = 1) dt exp(−t) = dx1 dx2 exp(−x21 − x22 ). (4.29)
0 −∞ −∞
Z +∞ 2
V2 (R = 1) × 1 = dx exp(−x2 ) . (4.30)
−∞
 Z ∞ 2 Z ∞ 2
V2 (R = 1) = 2 dx exp(−x2 ) = dx x−1/2 exp(−x) . (4.31)
0 0
Z ∞ 2
V2 (R = 1) = x(1/2)−1
exp(−x)dx = [Γ (1/2)]2 = π. (4.32)
0
Thus V2 (R) = V2 (R = 1) × R2 = πR2 , a result we are all familiar with.
50 4. Isolated System: Micro canonical Ensemble

4.11 Volume of an N -Dimensional Sphere

The volume of an N - dimensional sphere of radius R is formally given by


the integral,
Z +∞ Z +∞ Z +∞ N
!
X
2 2
VN (R) = dx1 dx2 · · · dxN Θ R − xi , (4.33)
−∞ −∞ −∞ i=1

Change the coordinate system from

{xi : i = 1, N } to {yi = xi /R : i = 1, N }.

dxi = Rdyi ∀ i = 1, N ;

" N
#! N
!
X X
2
Θ R 1− yi2 = Θ 1− yi2 .
i=1 i=1

We have,
Z Z Z N
!
+∞ +∞ +∞ X
VN (R) = RN dy1 dy2 · · · T dyN Θ 1 − yi2 ; (4.34)
−∞ −∞ −∞ i=1
AF
= VN (R = 1)RN . (4.35)
DR

where VN (R = 1) is the volume of an N - dimensional sphere of radius unity.


To find the volume of N -dimensional sphere of radius R, we proceed as
follows.
Z +∞ Z +∞ N
!
X
N 2 2
VN (R = 1)R = dx1 · · · dxN Θ R − xi . (4.36)
−∞ −∞ i=1

Differentiate both sides of the above expression with respect to R and get,
Z +∞ Z +∞
N −1
N VN (R = 1)R = dx1 dx2 · · ·
−∞ −∞
Z N
!
+∞ X
··· dxN δ R2 − x2i 2R. (4.37)
−∞ i=1

Now, multiply both sides by exp(−R2 )dR and integrate over R from 0 to ∞.
The Left Hand Side:
Z ∞
LHS = N VN (R = 1) dR exp(−R2 )RN −1 . (4.38)
0
4.11 Volume of an N -Dimensional Sphere 51

Let x = R2 ; then dx = 2RdR. This give


1 dx
dR = .
2 x1/2
We get,
Z ∞
N N
LHS = VN (R = 1) x 2 −1 exp(−x)dx,
2 0

 
N N
= VN (R = 1) Γ ,
2 2


N
=Γ + 1 VN (R = 1). (4.39)
2
The Right Hand Side :
Z ∞ Z +∞ Z +∞
RHS = dR exp(−R2 ) dx1 dx2 · · ·
0 −∞ −∞
Z N
!
+∞ X
2
··· dxN δ R − x2i 2R; (4.40)
−∞ i=1

t = R2 ; dt = 2RdR,
Z ∞ Z Z
T
AF
+∞ +∞
RHS = dt exp(−t) dx1 dx2 · · ·
0 −∞ −∞
Z N
!
+∞ X
dxN δ t− x2i ,
DR

−∞ i=1

Z +∞ Z +∞ Z +∞  
= dx1 dx2 · · · dxN exp −(x21 + x22 + · · · x2N ) ,
−∞ −∞ −∞
Z ∞ N
2
= dx exp(−x )
−∞

= π N/2 . (4.41)
Thus we get
π N/2
VN (R = 1) = . (4.42)
Γ N2 + 1

π N/2
VN (R) =  RN . (4.43)
Γ N2 + 1
52 4. Isolated System: Micro canonical Ensemble

4.12 Classical Counting of Micro states

Consider an isolated system of N non-interacting point particles. Each par-


ticle requires 3 position coordinates and 3 momentum coordinates for for its
specification. A string of 6N numbers denotes a micro state of the system.
Let us first find the volume of the phase space accessible to the system. The
X3N
p2i
integral over 3N spatial coordinates gives V N . We have, E = . The
i=1
2m
volume of the phase space of the √ system with energy ≤ E is the volume of a
3N dimensional sphere of radius 2mE.

4.12.1 Counting of the Volume

Let us measure the volume of the phase space in units of h3N , where h is
Planck constant. We have
∆x∆px ≥ h.
Thus h3N is the volume of a ”minimum uncertainty” cube. Thus we have

b V N (2πmE)3N/2
Ω(E, V, N ) = 3N . (4.44)
h Γ 3N
2 +1
T
AF
4.13 Density of States

Let g(E) denote the density of (energy) states. g(E)dE gives the number of
micro states with energy between E and E + dE. In other words,
DR

Z E
b
Ω(E) = g(E ′ )dE ′ . (4.45)
0
From the above, we find
!
b
∂ Ω(E, V, N )
g(E, V, N ) = . (4.46)
∂E
V,N

b
Let us take the partial derivative of Ω(E, V, N ) with respect to E and get,
V N (2πm)3N/2 3N (3N/2)−1
g(E, V, N ) = E . (4.47)
h3N Γ ( 3N
2 + 1)
2
Let us substitute N = 1 in the above and get the single particle density of
states, g(E, V ) as,
V π
g(E, V ) = (8m)3/2 E 1/2 . (4.48)
h3 4
4.14 Entropy of an Isolated System 53

4.13.1 A Sphere Lives on its Outer Shell : Power Law can be


Intriguing

In the limit of N → ∞, the volume of a thin outer shell tends to the volume
of the whole sphere. This intriguing behaviour is a consequence of the power
law behaviour.
VN (R) − VN (R − ∆R) RN − (R − ∆R)N
= ,
VN (R) RN
 N
∆R
= 1− 1− = 1 for N → ∞. (4.49)
R
Consider the case with R = 1 and ∆R = 0.1. The percentage of the total
volume contained in the outermost shell of an N dimensional sphere for
N = 1, 2, · · · 10 is given in the table below. . Hence in the limit of N → ∞

Table 4.1. Percentage of volume of an N dimenional sphere contained in a thin


outer shell between R = 0.90 and 1

VN (R = 1) − VN (R = 0.9) VN (R = 1) − VN (R = 0.9)
N × 100 N × 100
VN (R = 1) VN (R = 1)

1 10.000% 8
T 57.000%
AF
2 19.000% 9 61.000%
3 27.000% 10 65.000%
4 34.000% 20 88.000%
5 41.000% 40 99.000%
6 47.000% 60 99.000%
DR

7 52.000% 80 99.980%
100 99.997%

the number of micro states with energy less than or equal to E is nearly the
same as the number of micro states with energy between E − ∆E and E.

4.14 Entropy of an Isolated System

From Eq. (4.44) we see that


     
3 E 3 4πm 3
S(E, V, N ) = N kB ln V + ln + ln + . (4.50)
2 N 2 3h2 2
We find that the above expression for entropy is not extensive :

S(λE, λV, λN ) 6= λS(E, V, N ).


54 4. Isolated System: Micro canonical Ensemble

To restore extensivity of entropy we shall follow Boltzmann’s prescription


and divide Ω(E, V, N ), see Eq. (4.44), by N !.

b V N 1 (2πmE)3N/2
Ω(E, V, N ) = 3N . (4.51)
h N ! Γ 3N 2 +1
The corresponding entropy is then,
       
V 3 E 3 4πm 5
S(E, V, N ) = N kB ln + ln + ln + . (4.52)
N 2 N 2 3h2 2
It is clear that the expression for entropy given above is extensive.

4.15 Properties of an Ideal Gas


4.15.1 Temperature
The temperature of an ideal gas, as a function of E, V , and N , is given by
 
∂S 1 3N kB 2E
= = ⇒ T = . (4.53)
∂E V,N T 2E 3N kB

4.15.2 Equipartition Theorem


The energy of the system is thus given by,
 
T
AF
1
E = 3N kB T . (4.54)
2
The above is called equi-partition theorem. Each quadratic term6 in the
Hamiltonian carries an energy of kB T /2.
DR

4.15.3 Pressure
The pressure of an isolated system of ideal gas as a function of E, V , and N ,
is given by,
 
∂S P N kB
= = , (4.55)
∂V E,N T V

N kB T
P = . (4.56)
V
Substituting in the above T as a function of E, V , and N , see Eq. (4.53), we
get,
2E
P = . (4.57)
3V
3N
X
6 p2i
For an ideal gas H = . There are 3N quadratic terms in the Hamiltonian.
i=1
2m
4.15 Properties of an Ideal Gas 55

4.15.4 Ideal Gas Law

See the expression for P given in Eq.(4.56). We have the ideal gas law, P V =
N kB T.

4.15.5 Chemical Potential

An expression for the chemical potential as a function of E, V , and N is


derived as follows.
 
∂S µ
=− . (4.58)
∂N E,V T
Therefore,
   
V 3 4πmE
µ = −kB T ln − kB T ln . (4.59)
N 2 3N h2
Substituting in the above, the expression for T in terms of E and N , from
Eq. (4.53), we get micro canonical chemical potential7 ,
   
2E V E 4πmE
µ=− ln − ln . (4.60)
3N N N 3N h2
T
In the above expression for the micro canonical chemical potential, let us
substitute
AF
E = 3N kB T /2
and get canonical chemical potential8 ,
   
V 3 2πmkB T
µ = −kB T ln − kB T ln ,
DR

N 2 h2

 
V
= −kB T ln + 3kB T ln(Λ). (4.61)
N
where Λ is the thermal or quantum wavelength9 given by,
7
Chemical potential for an isolated system; it is expressed as a function of E, V ,
and N
8
chemical potential of a closed system : it is expressed as a function of T , V , and
N.
9
Consider a particle with energy kB T . We have
p2 √
E = kB T = ; p2 = 2mkB T ; p = 2mkB T .
2m
The de Broglie wavelength associated a particle having momentum p is
h h
Λ= = √ .
p 2mkB T
56 4. Isolated System: Micro canonical Ensemble

h
Λ= √ . (4.62)
2πmkB T
Let the number density be denoted by ρ. We have thus ρ = N/V . We can
write the chemical potential in a compact form, as follows.
 
V
µ = −kB T ln + 3kB T ln(Λ),
N

= kB T ln ρ + kB T ln Λ3 ,

= kB T ln(ρΛ3 ). (4.63)
• When the density of particles is small, and temperatures are high, then
ρΛ2 << 1. Note that the thermal wavelength Λ is inversely proportional
to (square root of ) temperature. Hence it is small at high temperatures.
Hence at high temperatures and/or low densities, the chemical potential
is negative and large.
• At ρΛ3 = 1, the chemical potential is zero.
• At low temperatures and high densities, ρΛ3 >> 1. The chemical po-
tential is positive.
Figure (4.5) shows typical classical behaviour of chemical potential with
T
change of temperature.
AF
0.6
DR

0.4

0.2

µ 0

−0.2

−0.4

−0.6

−0.8

−1
0 0.5 1
T 1.5

Fig. 4.5. µ versus T for a classical ideal gas. µ = −(3T /2) ln(T ). We have set
kB = h = ρ/(2πm)3/2 = 1
4.16 Quantum Counting of Micro states 57
ρ
I have set kB = h = = 1 for producing the above graph.
(2πm)3/2
From the figure we observe that at high temperatures µ is negative; at low
temperatures it is positive; as T decreases µ increases, reaches a maximum
and then decreases to zero at zero temperature.

4.16 Quantum Counting of Micro states

We have done classical counting of micro states and showed that for an iso-
lated system of a single particle confined to a volume V , the number of micro
states with energy less than ǫ, is given by
3/2
b V ) = V (2πmǫ)
Ω(ǫ, , (4.64)
h3 Γ ((3/2) + 1)
We have obtained the above by substituting N = 1 in the equation,

b 1 V N (2πmE)3N/2
Ω(E, V, N ) =  .
h3N N! 3N
Γ +1
2
10

10
Now let us do quantum counting of micro states T .
A fundamental entity in quantum mechanics is the wave function ψ(q, t), where
AF
q is the position vector. The wave function is given a physical interpretation that

ψ ⋆ (q, t)ψ(q, t)dq

gives the probability of finding the system in an elemental volume dq around


DR

the point q at time t. Since the system has to be somewhere, for, otherwise we
would not be interested in it, we have the normalization,
Z
ψ ⋆ (q, t)ψ(q, t)dq = 1,

where the integral is taken over the entire coordinate space - each of the x, y
and z coordinates extending from −∞ to +∞.
A central problem in quantum mechanics is the calculation of ψ(q, t) for the
system of interest. We shall be interested in the time independent wave function
ψ(q) describing a stationary states of the system.
How do we get ψ(q) ?
Schrödinger gave a prescription : Solve the equation Hψ(q) = Eψ(q), with ap-
propriate boundary conditions.
We call this the time independent Schrödinger equation.
H is the Hamiltonian operator
~2
H=− ▽2 +U (q). (4.65)
2m
The first operator on the right is kinetic energy and the second the potential
energy.
58 4. Isolated System: Micro canonical Ensemble

4.16.1 Energy Eigenvalues : Integer Number of Half Wave


Lengths in L

Let me tell you how to obtain the energy eigenvalues without invoking
Schrödinger equation. Consider a particle confined to a one dimensional box
of length L. We recognise the segment L must contain integral number of
half wave lengths - so that the wave function vanishes at the boundaries of
the one dimensional box. In other words,
λ
L = n× : n = 1, 2, , · · · (4.68)
2
2L
λ= : n = 1, 2, , · · · (4.69)
n
Substitute the above in the de Broglie relation
h h
p= = n : n = 1, 2, · · · .
λ 2L
This yields
p2 h2
ǫn = = n2 : n = 1, 2, · · · .
2m 8mL2
Consider a particle in an L × L × L cube - a three dimensional infinite well.
The energy of the system is given by T
h2
ǫnx ,ny ,nz = (n2 + n2y + n2z ),
AF
8mL2 x
where nx = 1, 2, · · · , , ny = 1, 2, · · · , and nz = 1, 2 · · · , .
E in the Schrödinger equation is a scalar ... a real number... called energy. It
DR

is an eigenvalue of the Hamiltonian operator : we call it energy eigenvalue.


The Schrödinger equation is a partial differential equation. Once we impose
boundary condition on the solution, then only certain discrete energies are per-
mitted. We call these energy eigenvalues.
Energy Eigenvalue by Solving Schrödinger Equation Once we specify
boundary conditions, then a knowledge of the Hamiltonian is sufficient to deter-
mine its eigenvalues and the corresponding eigenfunctions. There will be usually
several eigenvalues and corresponding eigenfunctions for a given system.

For a single particle in a one dimensional infinite well


~2 ∂ 2
H=− . (4.66)
2m ∂x2
Solve the one dimensional Schrödinger equation with the boundary condi-
tion : the wave function vanishes at the boundaries. Show that the energy
eigenvalues are given by
h2
ǫn = n2 ; n = 1, 2, · · · (4.67)
8mL2
4.16 Quantum Counting of Micro states 59

The ground state is (nx , ny , nz ) = (1, 1, 1); it is non degenerate; the energy
eigenvalue is
3h2
ǫ1,1,1 = . (4.70)
8mL2
The first excited state is three-fold degenerate. The corresponding energy
eigenvalue is
3h2
ǫ2,1,1 = ǫ1,2,1 = ǫ1,1,2 = . (4.71)
4mL2
We start with,
h2
ǫ= (n2 + n2y + n2z ).
8mL2 x
We can write the above as,
8mL2 ǫ
n2x + n2y + n2z = = R2 . (4.72)
h2
(nx , ny , nz ) represents a lattice point in the three dimensional space. The
equation n2x + n2y + n2z = R2 says we need to count the number of lattice
points that are at a distance R from the origin. It is the same as the number
of lattice points that are present on the surface of a sphere of radius R in the
T
positive quadrant; note that the x, y, and z coordinates of the lattice points
are all positive. It is difficult to count the number of lattice points lying on
AF
the surface of a sphere. Instead we count the number of points contained in
a thin spherical shell. To calculate this quantity we first count the number of
points inside a sphere of radius
 1/2
DR

8mL2 ǫ
R=
h2

b
and take one-eighth of it. Let us denote this number by Ω(ǫ). We have,
  3/2
b 1 4 π 8mL2 ǫ
Ω(ǫ) = π R3 = . (4.73)
8 3 6 h2
We recognize V = L3 and write the above equation as
b V ) = V π (8mǫ)3/2 = V 4π (2mǫ)3/2 ,
Ω(ǫ,
h3 6 h3 3

V 4π (2πmǫ)3/2 V (2πmǫ)3/2
= 3 3/2
= 3 √ ,
h 3 π h (3/2)(1/2) π

V (2πmǫ)3/2 V (2πmǫ)3/2
= 3
= 3 . (4.74)
h (3/2)(1/2)Γ (1/2) h Γ ((3/2) + 1)
60 4. Isolated System: Micro canonical Ensemble

The above is exactly the one we obtained by classical counting, see Eq. (4.64)
Notice that in quantum counting of micro states, the term h3 comes naturally,
while in classical counting it is hand-put11 .
b V ) with
The density of (energy) states is obtained by differentiating Ω(ǫ,
respect to the variable ǫ. We get
V π
g(ǫ, V ) =(8m)3/2 ǫ1/2 . (4.75)
h3 4
The important point is that the density of energy states is proportional to
ǫ1/2 .

4.17 Chemical Potential : Toy Model


Following Cook and Dickerson12 consider an isolated system of two identical,
distinguishable and non-interacting particles occupying non-degenerate en-
ergy levels { 0, ǫ, 2ǫ, 3ǫ, · · · }, such that the total energy of the system is 2ǫ.
b
Let Ω(E = 2ǫ, N = 2) denote the number of micro states of the two-particle
system with total energy E = 2ǫ. Aim is to calculate Ω(E b = 2ǫ, N = 2)
We label the two particles as A and B. The micro states with total energy
2ǫ are given in Table (4.2 : LEFT). We find that Ω(E b = 2ǫ, N = 2) = 3. The

b
S(E = 2ǫ, N = 2) = kB ln Ω(E
T
entropy of the two-particle system with energy E = 2ǫ is given by

= 2ǫ, N = 2) = kB ln(3).
AF
Now add a particle, labelled C, such that the energy of the system does
DR

Table 4.2. LEFT : Three micro states of a two-particles system with total energy
2ǫ. RIGHT : Six micro states of a three-particles system with total energy 2ǫ
0 ǫ 2ǫ
0 ǫ 2ǫ A,B - C
B,C - A
A - B
C,A - B
B - A
A B,C -
- A,B -
B C,A -
C A,B -

b
not change. Let Ω(E = 2ǫ, N = 3) denote the number of micro states of the
11
We wanted to count the phase space volume. We took h3 as the volume of a
six-dimensional cube. We considered the six-dimensional phase space (of a single
particle) as filled with non-overlapping exhaustive set of such tiny cubes. We
have to do all these because of Boltzmann ! He told us that entropy is logarithm
of number of micro states. We need to count the number of micro states.
12
See G Cook and R H Dickerson, Understanding the chemical potential, American
Journal of Physics 63(8), 737 (1995)
4.17 Chemical Potential : Toy Model 61

three-particle system with a total energy of E = 2ǫ. Table (4.2 : RIGHT)


lists the micro states.
b
We find Ω(E = 2ǫ, N = 3) = 6. The entropy is given by

b
S(E = 2ǫ, N = 3) = kB ln Ω(E = 2ǫ, N = 3) = kB ln(6).

We find S(E = 2ǫ, N = 3) > S(E = 2ǫ, N = 2). Note that


 
∂U
µ= .
∂N S,V

In other words, µ is the change in energy of the system when one particle
is added in such a way that the entropy and volume of the system remain
unchanged.
Let us now remove ǫ amount of energy from the three-particles system
and count the number of micro states. Table(4.3) shows that the number of
micro states of the three particles system with energy 2ǫ − ǫ = ǫ is three.

Table 4.3. Three micro states of a three-particles system with total energy ǫ
0 ǫ
A,B C T
B,C A
C,A B
AF

Thus to a two-particle system with energy 2ǫ, if we add a particle, we


DR

should remove ǫ of energy to keep entropy the same. Therefore µ = −ǫ.


DR
AF
T
5. Closed System : Canonical Ensemble

5.1 What is a Closed System ?

A closed system is one which does not exchange material with the surround-
ings. However, it does exchange energy. It is in thermal contact with the
surroundings. We idealize the surroundings as a ”heat bath”1 .
Thus, a closed system in thermal equilibrium, is characterized by T , V
and N . The system is not isolated. Hence its micro states are not all equi-
probable.

5.2 Toy Model à la H B Callen T


Consider2 a fair red die representing the system and two fair white dice
AF
representing the surroundings. A string of three numbers, each lying between
1 and 6 constitutes a micro state of the three dice system. There are 63 = 216
micro states and they are all equally probable3 . In particular the system-
dice(the red one) shall be in one of its six micro states with equal probability.
DR

Let us now impose the condition that the three dice add to 6. Under
this condition, let us enquire if the six micro states of the dice are equally
probable.
Let P (k) denote the probability that the red die shows up k given the
three dice add to 6. Because of the condition imposed, the red die can be in
any one of the four states, {1, 2, 3, 4} only; and these four micro states are
not equally probable. The probabilities can be calculated as follows.
We find that for the three-dice system, there are 10 micro states with the
property that the three dice add to 6. These are listed in Table(1).

1
A ”heat bath” transacts energy with the system, but its temperature does not
change.
2
Herbert B Callen, Thermodynamics and an Introduction to Thermostatistics,
Second Edition, Wiley (2006)
3
All micro states are equally probable : Ergodicity.
64 5. Closed System : Canonical Ensemble

Table 5.1. The ten micro states of Callen’s three-dice system with the constraint
that they add to ten

No. W R W No W R W

1 1 1 4 6 2 2 2
2 2 1 3 7 3 2 1
3 3 1 2 8 1 3 2
4 4 1 1 9 2 3 1
5 1 2 3 10 1 4 1

The ten micro states are equally probable. These are the micro states of the uni-
verse - which consists of the system and its surroundings. The universe constitutes
an isolated system.
Of these ten micro states of the universe, there are four micro states for which
the system die shows up in its microstate 1; therefore P (1) = 0.4. Similarly we can
calculate the other probabilities :

P (2) = 0.3; P (3) = 0.2; P (4) = 0.1; P (5) = P (6) = 0.0

The important point is that the system is not cound in all its microstates with
equal probabiity. T
From the above toy model, we can say that if we consider the system and its
AF
surroundings together to constitute the universe and demand that the universe has
a fixed energy, then the system will not be in its micro states with equal probability.
What is the probability of a micro state of a closed system ? We shall calculate
the probability in the next section employing two different methods. The first in-
volves Taylor expansion of S(E). I learnt of this, from the book of Balescu4 . The
DR

second is based on the method of most probable distribution, described in several


books5 .

5.3 Canonical Partition Function

5.3.1 Derivation à la Balescu


A closed system, its boundary and the bath - constitute the universe; the universe
is an isolated system. We know for an isolated system, all micro states are equally
probable. Let E denote the energy of the universe. It remains a constant.
Now consider a particular micro state of the closed system. Let us label it as
C. Let its energy be E(C). Note that E(C) < < E . When the closed system is in
4
R Balescu, Equilibrium and non-equilibrium statistical mechanics, Wiley (1975).
5
see e.g. R. K. Pathria, Statistical Mechanics, Second Edition Butterworth Heine-
mann (2001)p.45
5.3 Canonical Partition Function 65

b − E(C)) micro states


its micro state C, the surroundings can be in any one of Ω(E
of the universe6 .
For the universe, which is an isolated system, all the micro states are equally
probable. Thus we can say that the probability of finding the closed system in its
micro state C is given by,
b − E(C))
Ω(E
P (C) = , (5.1)
bt

bt .
where we have denoted the total number of micro states of the universe as Ω
b − E(C)). Therefore
We have S(E − E(C)) = kB ln Ω(E
 
b 1
Ω (E − E(C)) = exp S(E − E(C)) . (5.2)
kB
Also since E(C) < < E , we can Taylor expand S(E − E(C)) around E retaining
only the first two terms. We get,
 
∂S
S(E − E(C)) = S(E ) − E(C) ,
∂E E=E
1
= S(E ) − E(C) . (5.3)
T
Substituting the above in the expression for P (C), we get,
 
exp [S(E )/kB ] E(C)
P (C) = exp − ,
ct
Ω kB T
= α exp[−βE(C)] ,
T (5.4)
AF
where α is a constant and β = 1/(kB T ). We can evaluate α completely in terms of
the properties of the closed system by the normalization condition for the proba-
bilities, see below.
X
P (C) = 1 , (5.5)
DR

C
X
α exp [−βE(C)] = 1 , (5.6)
C
1 X
Q(T, V, N ) = = exp [−βE(C)] . (5.7)
α C

where Q(T, V, N ) is called the canonical partition function.

Canonical Partition Function : Transform of Density of


States
We start with
6
I am considering that a micro state of the universe can be thought of as a simple
juxtaposition of the micro state of closed system and the micro state of the
surroundings. The system and the surroundings interact at the boundaries and
hence there shall exist micro states of the isolated system which can not be neatly
viewed as the system micro state juxtaposed with the surroundings micro state.
Such micro states are so few in number we shall ignore them.
66 5. Closed System : Canonical Ensemble
X
Q(β, V, N ) = exp[−βEi (V, N )] , (5.8)
i

where, β = 1/[kB T ], and the sum runs over all the micro states of a closed system
at temperature T , volume V and number of particles N . Let Ω(E, b V, N ) denote
b
the density of (energy) states. In other words Ω(E, V, N )dE is the number of micro
states having energy between E and E + dE. The canonical partition function, see
Eq. (5.8), can be written as an integral over energy,
Z ∞
Q(β, V, N ) = b
dE Ω(E, V, N ) exp [−βE(V, N )] . (5.9)
0

We see that the canonical partition function is a ’transform’ of the density of states .
The ”variable” energy is transformed to the ”variable” temperature. The transform
helps us go from a micro canonical (ensemble) description (of an isolated system)
with independent variables E, (V and N ) to a canonical (ensemble) description (of
a closed system) with independent variables T , (V , and N . The density of states is
a steeply increasing function of E. The exponential function exp(−βE) decays with
E for any finite value of β. The decay is steeper at higher value of β or equivalently
at lower temperatures. The product shall be, in general, sharply peaked at a value
of E determined by β.
When β is small (or temperature is large) the integrand would peak at a large
value of E. When β is high (at low temperatures) it would peak at a low value of
E.

5.4 Helmholtz Free Energy


T
AF
The internal energy U of thermodynamics is obtained by averaging the statistical
energy E over a canonical ensemble. A closed system will invariably be found with
an energy U = hEi but for extremely small (relative) fluctuations around U ; these
fluctuations are proportional to the inverse of the square root of the number of
DR

molecules. Consider,
Z ∞
Q(β, V, N ) = b
dE Ω(E, V, N ) exp [−βE(V, N )] . (5.10)
0

In the above replace the integral over E by the value of the integrand, evaluated at
E = hEi = U . We get,
b
Q = Ω(E b
= U, V, N ) exp(−βU ) ; ln Q = ln Ω(U, V, N ) − βU ,

b
−kB T ln Q = U − T kB ln Ω(U, V, N )

= U − T S(U, V, N ). (5.11)
7
We identify the right hand side of the above as (Helmholtz) free energy :
 
1 ∂S
F (T, V, N ) = U − T S(U, V, N ) ; = . (5.12)
T ∂V V,N
7
Legendre Transform : Start with U (S, V, N ). Concentrate on the dependence of
U on S. Can this dependence be described in an equivalent but alternate way ?
Take the slope of the curve U (S) at S. We have
5.4 Helmholtz Free Energy 67

 
∂U
T (S, V, N ) = .
∂S V,N

We can plot T against S; but then it will not uniquely correspond to the given
curve U (S). All parallel curves in the U -S plane shall lead to the same T (S).
It is the intercept that will tell one curve from the other, in the family. Let us
denote the intercept by the symbol F . We have
 
U (S, V, N ) − F ∂U
= T ; F (T, V, N ) = U (S, V, N ) − T S; T = .
S ∂S V,N
The equation for micro canonical temperature, is inverted and entropy is ex-
pressed as a function of T , V , and N . Employing the function S(T, V, N ) we get
U (T, V, N ) and F (T, V, N ). The above is called Legendre transform. S transforms
to ’slope’ T and U (S) transforms to the intercept F (T ). We call F (T, V, N ) as
Helmholtz free energy.
Example: Consider U expressed as a function of S, V , and N :
S3
U (S, V, N ) = α .
NV
We get  
∂U 3S 2
T = =α .
∂S V,N NV
Inverting, we get, r
S=
NV T

.
T
AF
We then get internal energy as a function of temperature,
r  3/2
NV T
U= .
3α 3
DR

The Helmholtz free energy is then expressed as


 1/2  3/1
NV T
F = U − T S = −2 .
α 3
ENTHALPY : H(S, P, N ). Start with U (S, V, N ). Carry out the Legendre trans-
form of V → −P and U (S, V, N ) → H(S, P, N ).
 
∂U
H(S, P, N ) = U + P V ; P = .
∂V S,N
GIBBS FREE ENERGY : G(T, P, N ). Start with U (, V, N ). Transform S → T ,
V → −P , and U → G(T, P, N ).
   
∂U ∂U
G(S, P, N ) = U − T S + P V ; T = ; P = .
∂S V,N ∂V S,N
GRAND POTENTIAL : G(T, V, µ). Start with U (S, V, N ). Transform S → T ,
N → µ, and U → G(T, P, N ).
   
∂U ∂U
G(T, V, µ) = U − T S − µN , T = ,µ = .
∂S V,N ∂N S,N
68 5. Closed System : Canonical Ensemble

Thus we get a relation between (the microscopic description enshrined in)


the canonical partition function (of statistical mechanics) and (the macroscopic
description given in terms of) (Helmholtz) free energy (of thermodynamics) :
F (T, V, N ) = −kB T ln Q(T, V, N ). Statistical mechanics aims to connect the micro
world (of say atoms and molecules) to the macro world (of solids and liquids). In
other words it helps you calculate the macroscopic properties of a system say a solid,
in terms of the properties of its microscopic constituents (atoms and molecules) and
their interactions.
Boltzmann started the game of statistical mechanics by first proposing a micro
- macro connection for an isolated system, in the famous formula engraved on his
tomb: S = kB ln Ω. b You will come across several micro-macro connections in this
course on statistical mechanics. The formula, F (T, V, N ) = −kB T ln Q(T, V, N ),
provides another important micro - macro connection.

5.5 Energy Fluctuations and Heat Capacity


P
The average energy of a system is formally given by, hEi = i Ei pi , where pi
is the probability of the micro state i and Ei is the energy of the system when in
micro state i. For a closed system, pi = Q−1 exp(−βEi ) , where Q(T, V, N ) is the
(canonical) partition function given by
X
Q= exp(−βEi ).
i

We have,
P
T
AF
i Ei exp(−βEi )
hEi = P ,
i exp(−βEi )
1 X
= Ei exp(−βEi ) ,
Q i
DR

1 ∂Q ∂ ln Q
=− =− . (5.13)
Q ∂β ∂β
We identify hEi with the internal energy, usually denoted by the symbol U in
thermodynamics.
We have,
1 ∂Q
U =− , (5.14)
Q ∂β
   2
∂U 1 ∂2Q 1 ∂Q  
=− + = − hE 2 i − hEi2 = −σE
2
. (5.15)
∂β V Q ∂β 2 Q ∂β
Now write,
∂U ∂U ∂T
= × = CV (−kB T 2 ) . (5.16)
∂β ∂T ∂β
We get the relation between the fluctuations of energy of an equilibrium system and
the reversible heat required to raise the temperature of the system by one degree
Kelvin :
2
σE = kB T 2 C V . (5.17)
5.6 Canonical Partition Function : Ideal Gas 69

The left hand side of the above equation represents the fluctuations of energy when
the system is in equilibrium. The right hand side is about how the system would
respond when you heat it8 . Note CV is the amount of reversible heat you have to
supply to the system at constant volume to raise its temperature by one degree
Kelvin. The equilibrium fluctuations in energy are related to the linear response;
i.e. the response of the system to small perturbation9 .

5.6 Canonical Partition Function : Ideal Gas


I shall derive an expression for the canonical partition function of an ideal gas of N
molecules confined to a volume V and at temperature T . I shall do the derivation
by a method that involves the density of states
We first derive an expression for the density of (energy) states, denoted by g(E)
from micro canonical ensemble. g(E)dE is the number of micro states of an isolated
system with energy between E and E + dE. Formally, we have
b
∂Ω
g(E) = (5.18)
∂E

N
b= V 1 (2πmE)3N/2
Ω . (5.19)
N ! h3N Γ ( 3N
2
+ 1)
Therefore the density of (energy) states is given by,
T
b
∂Ω V N 1 (2πm)3N/2 3N 3N
g(E) = = E 2 −1 ,
AF
∂E N ! h3N Γ ( 3N
2
+ 1) 2

V N 1 (2πm)3N/2 3N
= E 2 −1 (5.20)
DR

N ! h3N Γ ( 3N
2
)
where we have made use of the relation
     
3N 3N 3N
Γ +1 = Γ . (5.21)
2 2 2
The partition function is obtained as a ”transform” of the density of states where
the variable E transformed to the variable β.
Z
V N 1 (2πm)3N/2 ∞ 3N
Q(β, V, N ) =  dE exp(−β E) E 2 −1 . (5.22)
N ! h3N Γ 3N 2 0

Consider the integral,


Z ∞
3N
I= dE exp(−βE)E 2 −1 . (5.23)
0

Let,
8
Notice that σ 2 is expressed in units of Joule2 . The quantity kB T 2 is expressed
in units of Joule-Kelvin. CV is in Joule/Kelvin. Thus kB T 2 CV has units of
Joule2 .
9
first order perturbation.
70 5. Closed System : Canonical Ensemble

x = βE then dx = βdE, (5.24)


Z ∞
1 3N
I = 3N/2 dx x 2 −1 exp(−x), (5.25)
β 0

Γ ( 3N
2
)
= . (5.26)
β 3N/2
Substituting the above in the expression for the partition function we get,
VN 1 VN 1
Q(T, V, N ) = 3N
(2πmkB T )3N/2 = , (5.27)
N! h N ! Λ3N
where
h
Λ= √ . (5.28)
2πmkB T
First check that Λ has the dimension of length. Λ is called thermal wavelength
or quantum wavelength. It is approximately
√ the de Broglie wavelength of a particle
with energy p2 /2m = kB T . We have p = 2mkB T . Therefore

h h h
Λ= = √ ≈ √ .
p 2mkB T 2πmkB T
A wave packet can be constructed by superimposing plane waves of wavelengths
in the neighbourhood of Λ. Thus a wave is spatially localized in a volume of the
T
order of Λ3 . Consider a gas with number density ρ. The inter particle distance is of
the order of ρ1/3 . Consider the situation ρΛ3 << 1. The particles are far apart.
AF
The wave packets do not overlap. Classical description will suffice. Quantum effects
manifest when ρΛ3 >> 1 : when density is large and temperature is low.
DR

5.7 Method of Most Probable Distribution


Let us now derive an expression for the canonical partition function employing the
method of most probable distribution. Consider an isolated system repre-
senting the universe. For convenience we imagine it to be a big cube. It contains
molecules moving around here and there, hitting against each other and hitting
against the wall. The universe is in equilibrium10 . Let the temperature be T . The
universe attains that temperature for which its entropy is maximum, under the
constraints of energy, volume and number of particles, imposed on it.
Let us imagine that the universe, represented by a big cube, is divided into a set
of small cubes of equal volumes by means of imaginary walls. Each cube represents
a macroscopic part of the universe.
Each small cube is, in its own right, a macroscopic object with a volume V .
Since the the walls of a small cube permits molecules and energy to move across,
the number of molecules in a cube, is not a constant. It shall fluctuate around a
10
Remember that an isolated system left to itself will eventually reach a state of
equilibrium whence all its macroscopic properties are y unchanging with time;
also a macroscopic (intensive) property property shall be the same at all regions
in the system.
5.7 Method of Most Probable Distribution 71

mean value; the fluctuations, however, are extremely small. What remains constant
is the chemical potential, µ.
The above observations hold good for energy also. Energy is not a constant. It
fluctuates around a mean value; the fluctuations are small; what remains fixed is
the temperature, T .
Let A denote the number of cubes contained in the big cube.
The universe - the big cube, has a certain amount energy say E and certain
number of molecules N and a certain volume V and these quantities are constants.
You can immediately see that what we have is Gibbs’ grand canonical ensemble
of open systems : each small cube represents an open system. and is a member of a
grand canonical ensemble. All the members are identical as far as their macroscopic
properties are concerned. This is to say the volume V , the temperature T and
chemical potential µ are all the same for all the members.
Now, let us imagine that the walls are made impermeable to movement of
molecules across. A cube can not exchange matter with its neighbouring cubes.
Let us also assume that each cube contains exactly N molecules. However energy
in a cube is not fixed. Energy can flow from one cube to its neighbouring cubes.
This constitutes a canonical ensemble11 .
Aim :
To find the probability for the closed system to be in its micro state i.
First, we list down all the micro states of the system. Let us denote the micro
states as {1, 2, · · · }. Note that the macroscopic properties T , V , and N are the
same for all the micro states. In fact the system switches from one micro state to
another, incessantly. Let Ei denote the energy of the system when it is in micro
state i. The energy can vary from one micro state to another.
T
To each cube, we can attach an index i. The index i denotes the micro state
of the closed system with fixed T , V and N . An ordered set of A indices uniquely
specifies a micro state of the universe.
AF
Let us take an example. Let the micro states of the closed system be denoted by
the indices {1, 2, 3}. There are only three micro states. Let us represent the isolated
system by a big square and construct nine small squares, each of which represents
a member of the ensemble. Each square is attached with an index which can be 1,
DR

2 or 3. Thus we have a micro state of the universe represented by the following

Table 5.2. A micro state with occupation number representation (2, 3, 4)


3 1 2

2 3 3

2 3 1

11
Usually, canonical ensemble is constructed by taking a system with a fixed value
of V and N and assembling a large number of them in such a way that each is in
thermal contact with its neighbours. Usually these are called mental copies of the
system. The system and its mental copies are then isolated. The isolated system
constitutes the universe. All the mental copies are identical macroscopically in
the sense they all have the same value of T , V and N . Also other macroscopic
properties defined as averages of a stochastic variable e.g. energy are also the
same for all the mental copies. But these mental copies will often differ, one from
the other, in their microscopic properties.
72 5. Closed System : Canonical Ensemble

In the above micro state, there are two squares with index 1, three with index
2 and four with index 3. Let {a1 = 2, a2 = 3, a3 = 4} be the occupation number
string of the micro state. There are several micro states having the same occupation
number string. I have given below a few of them.

133 233 121 112 123


212 312 232 223 123
323 321 232 333 233

Table 5.3. A few micro states with the same occupation number representation of
(2, 3, 4). There are 1260 micro states with the same occupation number representa-
tion

Notice that all the micro states given above have the same occupation number
string {2, 3, 4}. How many micro states are there with this occupation number string
? We have
b 3, 4) = 9! = 1260
Ω(2, (5.29)
2!3!4!

number string {2, 3, 4}


T
I am not going to list all the 1260 of the microstates belonging to the occupation

Let me generalize and say that a string of occupation numbers is denoted by


AF
the symbol ã = {a1 , a2 , · · · }, where a1 + a2 + · · · = A. We also have an additional
constraint namely a1 E1 + a2 E2 + · · · = E .
b
Let Ω(ã) b 1 , a2 , · · · ) denote the number of micro states of the universe
= Ω(a
belonging to the string ã. For a given string, we can define the probability for the
DR

closed system to be in its micro state indexed by i as


ai (ã)
pi (ã) = (5.30)
A
Note, the string ã = {a1 , a2 · · · } obeys the following constraints.
A
X
ai (ã) = A ∀ strings ã (5.31)
i=1

A
X
ai (ã)Ei = E ∀ strings ã (5.32)
i=1

Note that the value of pi varies from one string to another. It is reasonable to
obtain the average value of pi over all possible strings ã. We have
X  ai (ã) 
Pi = P(ã) (5.33)

A

where
5.8 Lagrange method 73

b
Ω(ã)
P(ã) = P . (5.34)
b
ã Ω(ã)

We are able to write the above because all the elements of an an ensemble have the
same probability given by inverse of the size of the ensemble. In the above,

b A!
Ω(ã) = , (5.35)
a1 !a2 ! · · ·
where we have used a simple notation ai = ai (ã) ∀ i = 1, 2, · · · .
b
Let us take a look at Ω(ã) b
for various strings ã. For large A the number Ω(ã)
will be overwhelmingly large for a particular string, which we shall denote as ã⋆ .
We can ensure this by taking A → ∞. Note that A should be large enough so
that even a micro state of smallest probability is present in the ensemble atleast
once.
Thus we can write
b ⋆)
ai (ã⋆ ) Ω(ã
Pi = ,
A Ω(ã b ⋆)

ai (ã⋆ )
= ,
A

a
= i (5.36)
A
T b
Thus the problem reduces to finding that string ã⋆ for which Ω(ã) is a maximum.
Of course there are two constraints on the string. They are
X
AF
aj (ã) = A ∀ ã; (5.37)
j
X
aj (ã)Ej = E ∀ ã. (5.38)
j
DR

We need to find the maximum (or minimum) of a function of a many variable


under one or several constraints on the variables. In the above example there are
two constraints. We shall tackle this problem employing the Lagrange method of
undetermined multipliers and to this we turn our attention below.

5.8 Lagrange and his Method


Let me pose the problem through a simple example.
A mountain be described by h(x, y) where h is a function of the variable x and
y. h is the elevation of the mountain at a point (x, y) on the plane.
I want to find out (x⋆ , y ⋆ ) at which h is maximum.
We write
∂h ∂h
dh = dx + dy = 0 (5.39)
∂x ∂y
If dx and dy are independent then dh = 0 if and only if
74 5. Closed System : Canonical Ensemble

∂h
=0 (5.40)
∂x
∂h
=0 (5.41)
∂y
We have two equations and two unknowns. In principle we can solve the above two
equations and obtain (x⋆ , y ⋆ ) at which h is maximum.
Now imagine there is a road on the mountain which does not necessarily pass
through the peak of the mountain. If you are travelling on the road, then what is
the highest point you will pass through ? In the equation
∂h ∂h
dh = dx + dy = 0 (5.42)
∂x ∂y
the infinitesimals dx and dy are not independent. You can choose only one of them
independently. The other is determined by the constraint which says that you have
to be on the road.
Let the projection of the mountain-road on the plane be described by the curve

g(x, y) = 0.

This gives us a constraint


∂g ∂g
dx + dy = 0 (5.43)
∂x ∂y
From the above we get,
 
∂g
∂x
dy = −   dx
T (5.44)
AF
∂g
∂y
We then have,
DR

∂h ∂h
dh = dx + dy = 0 (5.45)
∂x ∂y
 
∂g
∂h ∂h 
= dx + −  ∂x   dx = 0, (5.46)
∂x ∂y ∂g
∂y
     
 ∂h 
∂h ∂y ∂g
= −    dx = 0. (5.47)
∂x  ∂g  ∂x
∂y

In the above dx is an arbitrary non-zero infinitesimal. Hence the above equality


holds good if and only if the terms inside the square bracket is zero. We have,
∂h ∂g
−λ = 0, (5.48)
∂x ∂x
where we have set,
 
∂h
∂y
λ=  . (5.49)
∂g
∂y
5.9 Generalisation to N Variables 75

We have an equation similar to Eq. (5.48) involving partial derivative with respect
to the variable y, which follows from the definition, see Eq. (5.49), of the Lagrange
undetermined multiplier, λ.
Thus we have two independent equations,
∂h ∂g
−λ = 0, (5.50)
∂x ∂x
∂h ∂g
−λ = 0. (5.51)
∂y ∂y
We can solve and and get x⋆ ≡ x⋆ (λ) and y ⋆ = y ⋆ (λ). The value of x and y at
which h(x, y) is maximum under constraint g(x, y) = 0 can be found in terms of
the unknown Lagrange multiplier λ.
Of course we can determine the value of λ by substituting the solution
(x⋆ (λ), y ⋆ (λ)) in the constraint equation : g(x⋆ (λ), y ⋆ (λ)) = 0.

5.9 Generalisation to N Variables


Let f (x1 , x2 , · · · xN ) be a function of N variables. The aim is to maximize f under
one constraint g(x1 , x2 , · · · , xN ) = 0.
We start with
XN
∂f
df = dxi = 0 (5.52)
∂x i
i=1 T
for maximum. In the set {dx1 , dx2 , · · · dxµ , · · · dxN }, not all are independent. They
AF
are related by the constraint
XN
∂g
dxi = 0 (5.53)
i=1
∂x i
DR

We pick up one of the variable, say xµ and write


 
∂g
XN
∂xi
dxµ = −   dxi (5.54)
i=1,i6=µ
∂g
∂xµ
Substitute the above in the expression for df . We get,
XN  
∂f ∂g
−λ dxi = 0 (5.55)
i=1;i6=µ
∂xi ∂xi

where
 
∂f
∂xµ
λ=   (5.56)
∂g
∂xµ
There are only N − 1 values of dxi . We have eliminated dxµ . Instead we have the
undetermined multiplier λ. Since dxi : i = 1, N and i 6= µ are all independent of
each other we can set each term in the sum to zero. Therefore
76 5. Closed System : Canonical Ensemble

∂f ∂g
−λ = 0 ∀ i 6= µ (5.57)
∂xi ∂xi
From the definition of λ we get
∂f ∂g
−λ =0 (5.58)
∂xµ ∂xµ
Thus we have a set of N equations
∂f ∂g
−λ = 0 ∀ i = 1, N (5.59)
∂xi ∂xi
There are N equations and N unknowns. In principle we can solve the equation
and get
x⋆i ≡ x⋆i (λ) ∀ i = 1, N,
where the function h is maximum under constraint

g(x1 , x2 , · · · xN ) = 0.

The value of the undetermined multiplier λ can be obtained by substituting the


solution in the constraint equation.
If we have more than one constraints we introduce separate Lagrange multipliers
for each constraint. Let there be m ≤ N constraints. Let these constraints be given
by
gi (x1 , x2 , · · · xN ) = 0 ∀i = 1, m.
We introduce m number of Lagrange multipliers, λi : i = 1, m and write
∂f
∂xi
− λ1
∂g1
∂xi
− λ2
∂g2
∂xi
· · · − λm
∂gm
∂xi
= 0 ∀ i = 1, N
T (5.60)
AF
where the m ≤ N .
DR

5.10 Derivation of Boltzmann Weight


Let us return to our problem of finding Pi - the probability that a closed equilibrium
system (with macroscopic properties T, V, N ) will be found in it micro state i with
energy Ei . Employing the method of most probable distribution, we have found
that,
a⋆i
Pi = (5.61)
A
where A is the number of elements of the canonical ensemble and a⋆j = aj (ã⋆ ).
b
ã⋆ is that string for which Ω(ã) is maximum, under two constraints.

b A!
Ω(ã) = (5.62)
a1 !a2 ! · · ·
The two constraints are
X
aj (ã) = A (5.63)
j
X
aj (ã)Ej = E (5.64)
j
5.11 Mechanical and Thermal Properties 77

b
We extremize entropy given by ln Ω(ã).
X X
b 1 , a2 , · · · ) = ln A! −
ln Ω(a aj ln aj + aj (5.65)
j j

We introduce two Lagrange multipliers α and β and write


b 1 , a2 , · · · )
∂ ln Ω(a ∂(a1 + a2 + · · · − A)
−α
∂ai ∂ai
∂(a1 E1 + a2 E2 + · · · − E )
−β =0 (5.66)
∂ai
Let a⋆j denote the solution of the above equation. We get,
− ln a⋆j − α − 1 − βEj = 0 ∀ j = 1, 2, · · · (5.67)
The above can be written in a convenient form
a⋆j = γ exp(−βEj ) (5.68)
where γ = exp(−α − 1). We thus have,
Pj = η exp(−βEj )
where η = γ/A.
Thus we get the probability that a closed system shall be found in its micro
state j in terms of the constants η (which can be expressed as a function of the
Lagrange multiplier α) and β (which is the Lagrange multiplier for the constraint
on the total energy of the isolated system).
The task now is to evaluate the constants η and
T
P β. The constant η can be
evaluated by imposing the normalization condition : j Pj = 1. The closed system
AF
has to be in one of its micro state with unit probability. Thus we have,
1
Pj = exp(−βEj ) (5.69)
Q
X
DR

Q(β, V, N ) = exp(−βEj ) (5.70)


j

We call Q the canonical partition function.


What is the nature of the Lagrange multiplier β ? On physical ground we
identify
1
β= .
kB T
This we do by making a correspondence to thermodynamics. I shall refer you to
• Donald A McQuairrie, Statistical Mechanics, Harper and Row (1976)pp.35-44,
for this.

5.11 Mechanical and Thermal Properties


In statistical mechanics, we have a random variable corresponding to a property e.g.
energy, E. We take an average of this random variable over a canonical ensemble and
denote it by hEi. This quantity corresponds to the internal energy, usually denoted
by the symbol U , of thermodynamics : U = hEi. This establishes a micro-macro
connection.
78 5. Closed System : Canonical Ensemble

Energy of a closed system varies, in general, from one micro state to another.
An equilibrium system visits a sequence of micro states dictated by Newtonian dy-
namics. Averaging over the visited microstates during the observation time gives
us the time average. Instead, invoking ’ergodicity’, we average over a Gibbs’ en-
semble and equate it to the thermodynamics property. Thus, energy averaged over
a canonical ensemble, gives internal energy of thermodynamics, see below.
exp(−βEi )
pi = P (5.71)
i exp(−βEi )

P
X
Pi Ei exp(−βEi )
hEi = Ei pi = (5.72)
i i exp(−βEi )

We are able to establish a connection between statistical mechanics and thermody-


namics for a mechanical property like energy; notice energy is a ’private’ property
of a micro state; in other words I can attach a numerical value for the property
called energy to each microstate. Hence we are able to average this property over
a canonical ensemble and get the internal energy.
How do we calculate a thermal property like entropy ? First we notice, entropy
is not a ’private’ property of a micro state. We can not attach a numerical value for
entropy to a microstate. Entropy is a ”social” property or a ’collective’ property.
Hence entropy can not be easily calculated the way we calculated energy. For an
b
isolated system, we could write an expression for entropy as S = kB ln Ω(E, V, N )
since all the micro states are equally probable12
What is the expression for entropy of a closed system ? To this issue we turn
our attention below. T
AF
5.12 Entropy of a Closed System
P
Consider the expression − i pi ln pi in the context of canonical ensemble. In an
DR

earlier lecture we talked of the number of micro states of an isolated system. The
isolated system is constructed as follows : Assemble a large number of closed sys-
tems. Each closed system is in thermal contact with its neighbouring closed systems.
Let η = {i : i = 1, 2, · · · } label the micro states of a closed system. Let A denote
the number of closed systems in the ensemble. The ensemble is isolated. We have
thus an isolated system. We call it the universe. We describe a micro state of the
universe by specifying the index i for each if its members. The index comes from
the set η defied earlier. The micro states of the universe system are all equally
probable; we group them and denote a group by specifying the string {a1 , a2 , · · · }
where ai is the number of elements of the ensemble, having the index i. We have

b 1 , a2 , · · · ) = A!
Ω(a (5.73)
a1 !a2 ! · · ·
b For convenience we
The aim is to find {a⋆i : i = 1, 2, · · · } that maximizes Ω.
b
maximize ln Ω. Let us consider the limit ai → ∞ ∀ i. Also consider the variables
pi = ai /A. Then
12
The probability of a microstate in a micro canonical ensemble is pi =
b
1/Ω(E, b = −kB P pi ln pi
V, N ) ∀ i. Hence kB ln Ω i
5.13 Free Energy to Entropy 79
X X
b = A ln A − A −
ln Ω ai ln ai + ai (5.74)
i i
X X
= ai ln A − ai ln ai (5.75)
i
X a 
i
=− ai ln (5.76)
A
X
= −A pi ln pi (5.77)
i
b
ln Ω X
=− pi ln pi (5.78)
A i

P of one of the A number of closed systems constituting the


The above is the entropy
the universe. Thus, − i pi ln pi provides a natural formula for the entropy of a
system whose micro states are not equi-probable.
Physicists would prefer to measure entropy in units of Joules per Kelvin. For,
that is what they have learnt from Clausius, who defined
q
dS = ,
T
where dS is the entropy gained by a system when it receives an energy of q Joules
by heat in a reversible process at a constant temperature of T Kelvin. Hence we
define,
X
S = −kB pi ln pi . (5.79)
i
T
This expression for entropy is natural for a closed system described by a canonical
AF
ensemble. We call it Boltzmann-Gibbs-Shannon entropy.

5.13 Free Energy to Entropy


DR

We start with the thermodynamic relation


F (T, V, N ) = U − T S (5.80)
We recognize that F (T, V, N ) is the Legendre transform of the fundamental equa-
tion U (S, V, N ) where we transform the variable S in favour of T defined as
 
∂U
T = (5.81)
∂S V,N
We re-write the expression for F as,
S
− = β(F − U ) (5.82)
kB
We make use of the following,
X
Q= exp(−βEi )
i

1
pi = exp(−βEi ) ;
Q
80 5. Closed System : Canonical Ensemble

F = −kB T ln Q

1 X
U = hEi = Ei exp(−βEi )
Q i

and write,
" #
S 1 X
− = β −kB T ln Q − Ei exp(−βEi )
kB Q i
" #
1 X 1 X
= − ln Q exp(−βEi ) − β Ei exp(−βEi )
Q i Q i
X  1  
= exp(−βEi ) − ln Q − Ei β
i
Q
X    
1 exp(−βEi )
= exp(−βEi ) ln
i
Q Q
X
= pi ln pi (5.83)
i

from which we get the Boltzmann-Gibbs-Shannon entropy,


X
S = −kB pi ln pi .
i
T
AF
5.14 Microscopic View : Heat and Work
The thermodynamic energy U is identified with statistical energy E averaged over
a suitable Gibbs ensemble. We use micro canonical ensemble for isolated system,
DR

canonical ensemble for closed system, and grand canonical ensemble for open sys-
tem. Let pi : i = 1 , 2 · · · denote formally an ensemble. pi is the probability of a
micro state of the system under consideration.
For example if the system is isolated, then all micro states are equally probable.
We simply count the number of micro states of the system; let us say there are Ω b
micro states. Inverse of this shall be the probability of a micro state.
For an isolated system pi = 1/Ω. b
1
For a closed system, pi = exp(−βEi ).
Q
1
For an open system pi = exp(−βEi + βµNi )
Q X
We can write, in general, U = pi Ei . This equation suggests that the inter-
i
nal energy of a closed system can be changed by two ways.
1. change {Ei : i = 1, 2, · · · } keeping {pi : i = 1, 2, · · · } the same. This we call
as work. Note that the energy eigenvalue change when we change the boundary
condition, i.e. when we move the boundary thereby changing the volume.
2. change {pi i = 1, 2, · · · } keeping {Ei P: i = 1, 2, · · · } the same. The changes
in pi should be done in such way that i pi = 1. This we call as heat.
5.14 Microscopic View : Heat and Work 81

Thus we have,
X X⋆
dU = pi dEi + Ei dpi (5.84)
i i

where the superscript ⋆ in the second sum should remind us that all dpi s are not
independent and that they should add to zero.
In the first sum we change Ei by dEi ∀ i keeping pi ∀ i unchanged.
In thePsecond sum we change pi by dpi ∀ i keeping Ei unchanged for all i and
ensuring i dpi = 0.

X
5.14.1 Work in Statistical Mechanics : W = pi dEi
i

We have
X X ∂Ei
pi dEi = pi dV
i i
∂V

P 
∂ pi Ei
i
= dV
∂V

∂hEi
= dV
∂V T
AF
∂U
= dV
∂V

= −P dV = W (5.85)
DR
82 5. Closed System : Canonical Ensemble
X
5.14.2 Heat in Statistical Mechanics : q = Ei dpi
i

Method-1
We start with the first term on the right hand side of Eq. (5.84).
X
S = −kB pi ln pi , (5.86)
i
X⋆
dS = −kB [1 + ln pi ] dpi , (5.87)
i
X⋆
T dS = d¯q = −kB T ln pi dpi , (5.88)
i
X⋆ e−βEi
= −kB T [−βEi − ln Q] dpi since pi = (5.89)
i
Q
X⋆
= Ei dpi (5.90)
i

Method-2
Alternately, we recognize that change of entropy is brought about by change
of probabilities of microstates and vice versa, keeping their respective energies the
same. Thus,
X X ∂pi
Ei dpi = Ei dS (5.91)
∂S
i

=

i
P
i Ei pi

dS =
∂hEi
dS =
∂U
T
dS = T dS = q (5.92)
AF
∂S ∂S ∂S

5.15 Adiabatic Process - a Microscopic View


DR

In an adiabatic process, the system does not transact energy by heat. Hence the
probabilities {pi } of the micro states do not change during an adiabatic process.
The only way the internal energy can change is through change of {ǫi }. For a
particle confined to a box, quantum mechanics tells us that the energy eigenvalue
is inversely proportional to the square of the length of the box.
1
ǫi ∼ (5.93)
L2

∼ V −2/3 (5.94)

∂ǫi
∼ V −5/3 (5.95)
∂V
First law of thermodynamics tells us dU = d¯Q + d¯W ; For an adiabatic process
d¯Q − 0; if the process is reversile then d¯W = −P dV . Therefore,
5.16 Q(T, V, N ) for an Ideal Gas 83
X ∂ǫi
dU = pi dV (5.96)
i
∂V

−P dV ∼ V −5/3 dV (5.97)

P V 5/3 = Θ, a constant. (5.98)

5.16 Q(T, V, N ) for an Ideal Gas


I shall derive an expression for the canonical partition function of an ideal gas of
N molecules confined to a volume V and at temperature T .
Formally the partition function is given by,
Z +∞ Z +∞
VN 1
Q(T, V, N ) = dp 1 dp2 · · ·
N ! h3N −∞ −∞

Z +∞  
β 
··· dp3N exp − p21 + p22 · · · + p23N (5.99)
−∞ 2m

=
VN 1
Z +∞ 
dp exp −
T
β 2
p
3N
(5.100)
AF
N ! h3N −∞ 2m

Z +∞  3N
VN 1 1 p2
DR

= dp exp − (5.101)
N ! h3N −∞ 2 mkB T
Consider the integral
Z +∞  
1 p2
I = dp exp − (5.102)
−∞ 2 mkB T
since the integrand is an even function, we can write the above as
Z +∞  
1 p2
I =2 dp exp − (5.103)
0 2 mkB T
Let
p2
x= (5.104)
2mkB T
Therefore,
p
dx = dp (5.105)
mkB T

r
mkB T 1
dp = dx (5.106)
2 x1/2
84 5. Closed System : Canonical Ensemble

The integral can now expressed as


Z ∞

I = 2mkB T dx x−1/2 exp(−x)
0

Z
√ ∞
= 2mkB T dx x(1/2)−1 exp(−x)
0

 
√ 1
= 2mkB T Γ
2
√ √
= 2πmkB T since Γ (1/2) = π (5.107)
The canonical partition function is thus given by,
VN 1
Q(β, V, N ) = (2πmkB T )3N/2 (5.108)
N ! h3N
We write the above as
VN 1
Q(T, V, N ) = (5.109)
N ! Λ3N

h
Λ(T ) = √ . (5.110)
2πmkB T

Energy
T
AF
 
V
ln Q = N ln + N − 3N ln(Λ) (5.111)
N

∂ ln Q ∂ ln Λ
DR

−hEi = = −3N (5.112)


∂β ∂β

ln Λ = ln(h/ 2πm) + (1/2) ln β (5.113)

3N kB T
U = hEi = (5.114)
2
The above shows that energy is equi-partitioned amongst the 3N degrees of freedom;
each degree of freedom carries kB T /2 of energy.

Heat capacity
The heat capacity is given by
∂U 3N kB
CV = =
∂T 2
We have nR = N kB , where n is the number of moles and R the universal gas
constant. Thus heat capacity per mole of the substance - the so-called molar specific
heat is given by,
CV 3R
= (5.115)
n 2
5.16 Q(T, V, N ) for an Ideal Gas 85

R = 8.315 J (mol)−1 K −1 = 1.9873 Cal. (mol)−1 K −1 ; 3R/2 = 2.9809 ≈


cal. (mol)−1 K −1

T
AF
DR
DR
AF
T
6. Open System : Grand Canonical Ensemble

Grand Canonical Ensemble is useful in the study of open systems.

6.1 What is an Open System ?


An open system is one which exchanges energy and matter with its surroundings.
The surroundings act as a heat bath as well as a particle (or material) bath.
• A heat bath transacts energy with the system. The temperature of the heat
bath does not change because of the transaction.
• A material (or particle) bath transacts matter (or particles) with the system.
The chemical potential of the material bath does not change because of the
transaction.
T
The system is in thermal equilibrium with the surroundings1 . The system is also
in diffusional equilibrium with the surroundings2 . The system can be described by
AF
its temperature, T , volume, V and chemical potential, µ. Notice that T , µ, and
V are independent properties of an open system. Since the system is not isolated,
its micro states are not equi-probable. Our aim is to calculate the probability of a
micro state of the open system.
DR

Let us take the open system, its boundary and surroundings and construct a
universe, which constitutes an isolated system. We are interested in constructing
an isolated system because, we want to start with the only assumption we make in
statistical mechanics : all the micro states of an isolated system are equally probable.
We have called this the ergodic hypothesis.
Let E denote the total energy of the universe. E >> E, where E is the (average
or typical) energy of the open system under study.
Let N denote the total number of particles in the universe. N >> N, where
N is the (average or typical) number of particles in the open system. V is the total
volume of the universe. The universe is characterized by E , N , and V and these are
held at constant values.
We want to describe the open system in terms of its own micro states. Let C
be a micro state of the open system. Let
• E(C) be the energy of the open system when it is in its micro state C and
• let N (C) be the number of particles in the open system when it is in its micro
state C.
1
equality of temperature signals thermal equilibrium.
2
equality of chemical potential ensures diffusional or material equilibrium. Of
course, equality of pressure shows mechanical equilibrium.
88 6. Open System : Grand Canonical Ensemble

When the open system is in micro state C the universe can be in any one of
its innumerable micro states3 such that the energy of the surroundings is E − E(C)
and the number of particles in the surroundings is N − N (C).
Let ΩC = Ω(E − E(C), N − N (C)), denote the subset of micro states of the
universe having a common property given by the following. When the universe is
in a micro state belonging to ΩC , then the system is in the chosen micro state C
and the surroundings can be in any of its microstates having energy E − E(C) and
number of particles N −N (C). Let Ω bC denote the number of elements in the subset
(event) ΩC . Following Boltzmann we define a statistical entropy associated with
the event ΩC as
bC
SC = kB ln Ω
b − E(C), N − N (C))
= kB ln Ω(E
= S(E − E(C), N − N (C)). (6.1)
Since E(C) << E , and N (C) << N , we can Taylor-expand S retaining only
the first two terms. We have
   
S E − E(C), N − N (C) = S E , N

 
∂S
−E(C)
∂E
N E,N

 
−N (C)
∂S
∂N
T


E E,N
(6.2)
AF
From the first law of thermodynamics we have, for a reversible process,
dE = T dS − P dV + µdN (6.3)
from which it follows,
DR

1 P µ
dS = dE + dV − dN (6.4)
T T T
We have
S ≡ S(E, V, N ) (6.5)

     
∂S ∂S ∂S
dS = dE + dV + dN (6.6)
∂E V,N ∂V E,N ∂N E,V

3
The picture I have is the following. I am visualizing a micro state of the isolated
system as consisting of two parts. One part holds the signature of the open
system; the other holds the signature of the surroundings. For example a string
of positions and momenta of all the particles in the isolated system defines a
micro state. This string consists of two parts. The first part s contains the string
of positions and momenta of all the particle in the open system and the second
part contains the positions and momenta of all the particles in the surroundings.
Since the system is open the length system-string is a fluctuating quantity and
so is the length of bath-string. However the string of the isolated system is of
fixed length. I am neglecting those micro states of the isolated system which hold
the signature of the interaction between the system and the surroundings at the
boundaries.
6.1 What is an Open System ? 89

Comparing the coefficients of dE, dV and dN , in equations (6.4) and (6.6), we get,
     
∂S 1 ∂S P ∂S µ
= ; = ; =− (6.7)
∂E V,N T ∂V E,N T ∂N E,V T
Therefore,
   
1 µ
S E − E(C), N − N (C) = S E , N − E(C) + N (C) (6.8)
T T
The probability of the micro state C is given by
 
b E − E(C), N − N (C)

P (C) = (6.9)
ΩbTotal
We are able to write the above because of the postulate of ergodicity : All micro
states of the universe - an isolated system, are equally probable. We have,
  
1 1
P (C) = exp S E − E(C), N − N (C)
ΩbTotal kB

 
1 S(E , V, N ) E(C) µN (C)
= exp − +
bTotal
Ω kB kB T kB T
T
AF
= α exp[−β {(E(C) − µN (C)}]. (6.10)
In the above, the constant α can be determined by the normalization condition,
X
P (C) = 1, where the sum runs over all the micro states of the open system. We
DR

c
have,
1
P (C) = exp[−β{E(C) − µN (C)}] (6.11)
Q
where the grand canonical partition function is given by
X
Q(T, V, µ) = exp [−β {E(C) − µN (C)}] (6.12)
C

The fugacity is given by λ = exp(βµ); then we can write the grand canonical
partition function as,
X N(C)
Q(T, V, λ) = λ exp[−βE(C)] (6.13)
C

• Collect those micro states of a grand canonical ensemble with a fixed value
of N . Then these micro states constitute a canonical ensemble described the
canonical partition function, Q(T, V, N ).
• Collect all the micro states of a grand canonical ensemble with a fixed energy.
and fixed number of particles. Then these micro states constitute a microcanon-
ical ensemble.
90 6. Open System : Grand Canonical Ensemble

Thus we can write the grand canonical partition function as,



X
Q(T, V, λ) = λN Q(T, V, N ) (6.14)
N=0

Let Q1 (T, V ) denote single-particle canonical partition function. We have


QN
1
Q(T, V, N ) = QN (T, V ) = . (6.15)
N!
Therefore,
X∞
λN Q N1
Q(T, V, µ) = = exp(λQ1 ) (6.16)
N=0
N !

Grand canonical ensemble is an extremely useful ensemble. The reason is that


the constraint of constant N required for calculating the canonical ensemble is often
mathematically awkward4 .

6.2 Micro-Macro Synthesis : Q and G


• Entropy is the thermodynamic counter part of the statistical mechanical micro
canonical partition function - the density of states. We have S(E, V, N ) =
b
kB ln Ω(E, V, N ).
T
• Helmholtz free energy is the thermodynamic counter part of the statistical
mechanical canonical partition function, Q(T, V, N ). We have F (T, V, N ) =
AF
−kB T ln Q(T, V, N ).
What is the thermodynamic counter part of the grand canonical ensemble ?
Let us call it, the ”grand” potential and denote it by the symbol G. G is a function
of T, V, andµ. G(T, V, µ), is obtained from U (S, V, N ) by Legendre transform of
DR

S → T and N → µ. We have,
G(T, V, µ) = U (S, V, N ) − T S − µN (6.17)

   
∂U ∂U
T = : µ= (6.18)
∂S V,N
∂N S,V

Some authors e.g. Donald A McQuairrie, Statistical Mechanics, Harper and Row
(1976) would like to identify P V as the thermodynamic counter part of the grand
canonical ensemble. The correspondence is

G = −P V = −kB T ln Q.
4
We shall experience this while trying to evaluate the canonical partition function
for fermions and bosons. We will not be able to carry out the sum over occupation
numbers because of the constraint that they add up to a constant N . Hence we
shall multiply the restricted sum by λN and sum over all possible values of N .
This would remove the restriction and we shall express the partition function as
sum over (micro states) of product (over occupation numbers). We shall interpret
λ as fugacity. The chemical potential and fugacity are related : λ = exp(βµ).
All these can be viewed as mathematical tricks. The language of grand canonical
ensemble gives a physical meaning to these mathematical tricks.
6.2 Micro-Macro Synthesis : Q and G 91

Let us establish such a relation between G and Q.


G = −kB T ln Q : We follow the same method we employed for establishing the
connection between Helmholtz free energy and canonical partition function. We
have,
X
Q(T, V, λ) = exp [−β (Ei − µNi )] (6.19)
i

where the sum runs over all the micro states i of the open system. Ei is the energy
of the system when in micro state i, and Ni is the number of particles in the system
when in its micro state i.
We replace the sum over micro states by sum over energy and number of par-
ticles. Let g(E, N ) denote the density of states. We have then,
Z Z
Q(T, V, µ) = dE dN g(E, N ) exp[−β(E − µN )] (6.20)

The contribution to the integrals come overwhelmingly from a single term at We


then get,
Q(T, V, µ) = g(hEi, hN i) exp[−β(hEi − µhN i)] (6.21)
Let us denote hEi by E and hN i by N . Taking logarithm on both sides,
ln Q = ln g(E, N ) − β(E − µN ) (6.22)
We then have,
−kB T ln Q = −T [kB ln g(E, N )] + E − µN = E − T S − µN = G (6.23)
P N′ ′
Alternately, we start with Q = N ′ λ Q(T, V, N ) In principle, the number of
T
particles in an equilibrium open system is not a fixed number. It fluctuates from one
micro state to another. However the fluctuations are very small; it can be shown
AF
that the relative fluctuations are inversely proportional to the size of the system.
In the above expression for Q, only one term contributes overwhelmingly to the

sum over N ′ . Let the value of N ′ for which the λN Q(T, V, N ′ ) is maximum be N .
Hence the sum over N can be replaced by a single entry with N ′ = N .

DR

Q(T, V, µ) = λN Q(T, V, N ) (6.24)


µN
ln Q(T, V, µ) = βµN + ln Q(T, V, N ) = + ln Q(T, V, N ) (6.25)
kB T
kB T ln Q = µN + kB T ln Q(T, V, N ) (6.26)
While studying Canonical ensembles we have shown that
F (T, V, N ) = −kB T ln Q(T, V, N ).
Therefore we can write the above equation as,
kB T ln Q = µN − F (T, V, N ) (6.27)

−kB T ln Q = F − µN = U − T S − µN = G (6.28)
Recall the discussions on Legendre Transform : We start with U ≡ U (S, V, N ).
Transform S in favour of the ”slope” T (partial derivative of U with respect to S).
We get the ”intercept” F (T, V, N ) as U − T S. Let us carry out one more transform
: N → µ. i.e. transform N in favour of the slope µ (partial derivative of U with
respect to N ); µ is the chemical potential. We get the ”intercept” G(T, V, µ) - the
grand potential. We have G(T, V, µ) = U − T S − µN . Thus we have, G(T, V, µ) =
−kB T ln Q(T, V, µ)
92 6. Open System : Grand Canonical Ensemble

6.3 Statistics of Number of Particles


The grand canonical partition function Q of statistical mechanics (micro world)
corresponds to the grand potential (G) of thermodynamics (macro world). We have
G(T, V, µ) = −kB T ln Q.
Some authors5 would like to identify P V as the thermodynamic counterpart
of grand canonical partition function : P V = kB T ln Q. Let us first derive this
relation. To this end, let me tell you of a beautiful formula proposed by Euler, in
the context of homogeneous function.

6.3.1 Euler and his Theorem


U is a homogeneous function of S, V and N . U is an extensive property; so are
S, V and N . Extensivity means that U is a first order homogeneous function of S,
V , and N . Therefore U (λS, λV, λN ) = λU (S, V, N ) where λ is a constant6 . Euler’s
trick consists of differentiating both sides of the above equation with respect to λ.
We get,
∂U ∂(λS) ∂U ∂(λV ) ∂U ∂(λN )
+ + = U (S, V, N ) (6.29)
∂(λS) ∂λ ∂(λV ) ∂λ ∂(λN ) ∂λ

∂U ∂U ∂U
S +V +N = U (S, V, N ) (6.30)
∂(λS) ∂(λV ) ∂(λN )
T
The above is true for any value of λ. In particular it is true for λ = 1. Substitute
in the above λ = 1 and get,
AF
∂U ∂U ∂U
S +V +N = U (S, V, N ) (6.31)
∂S ∂V ∂N

T S − P V + µN = U (6.32)
DR

6.3.2 Q : Connection to Thermodynamics


We proceed as follows. From Euler relation : U = T S − P V + µN , we have −P V =
U − T S − µN. The RHS of this equation is grand potential. Hence,
−P V = G(T, V, µ) = −kB T ln Q(T, V, µ)
P V = kB T ln Q(T, V, µ) (6.33)

6.3.3 Gibbs-Duhem Relation


Now that we are on the Euler’s formula, let me digress a little bit and see if the
equations we have derived, can be used to establish a relation amongst the intensive
properties T , P and µ of the system.
Derivation from U(S, V, N)
5
Donald McQuairrie, Statistical Mechanics, Harper and Row (1976)
6
do not confuse λ here with fugacity introduced earlier. Unfortunately I have used
the same symbol. You should understand the meaning of λ in the contetzr it is
used.
6.3 Statistics of Number of Particles 93

We proceed as follows.
U = T S − P V + µN, (6.34)
dU = T dS − P dV + µdN + SdT − V dP + N dµ (6.35)
From the first law of thermodynamics, we have dU = T dS − P dV + µdN . Hence,
N dµ + SdT − V dP = 0. It follows then,
S V
dµ = − dT + dP = −sdT + vdP (6.36)
N N
where s is the specific entropy - entropy per particle and v is specific volume -
volume per particle.

6.3.4 Average number of particles in an open system, hN i



X
We start with, Q(T, V, λ) = λN QN (T, V ). Take the partial derivative of Q with
N=0
respect to λ and get,
  X∞ ∞
∂Q 1 X
= N λN−1 QN T, V = N λN QN (T, V )
∂λ T,V N=0
λ N=0
" ∞
#
Q 1 X N
= N λ QN (T, V )
λ Q N=0
Q
λ
hN i
=
T (6.37)
AF
Thus the average number of particles in an open system can be formally expressed
as,
1 ∂Q ∂ ln Q
hN i = λ =λ (6.38)
DR

Q ∂λ ∂λ

6.3.5 Probability P (N ), that there are N particles in an open


system
We start with

X ∞
X QN
Q(T, V, λ) = λN QN (T, V ) = λN 1

N=0 N=0
N!
= exp(λQ1 ) (6.39)
The average of N can be formally expressed as,
∞ N
1 X (λQ1 )N X λN Q N
1
hN i = N = (6.40)
Q N=0 N! N=0
Q N !

We have already shown Q = exp(λQ1 ). Therefore,



X (λQ1 )N exp(−λQ1 )
hN i = N (6.41)
N=0
N!
94 6. Open System : Grand Canonical Ensemble

We have earlier shown that hN i = λQ1 = ζ say. Then,


N
X X∞
ζ N exp(−ζ)
hN i = N = N P (N ) (6.42)
N=1
N! N=1

ζN
P (N ) thus is a Poisson distribution : P (N ) = exp(−ζ. For Poisson distribution,
N!
the mean equals variance. Therefore
2
σN = hN 2 i − hN i2 = hN i = ζ = λ Q1 . (6.43)

6.3.6 Number Fluctuations


In an open system, the energy E and the number of molecules N are random
variables. Energy fluctuates when the system goes from micro state to another.
The number of molecules fluctuates from one micro state to the other. Let us now
derive an expression for the variance of N :

σ 2 = hN 2 i − hN i2 .

To this end, we start with


X  
Q(T, V, µ) = exp − β{E(c) − µN (c)} (6.44)
c

In the above
c denotes a micro state of the open system
T
AF
E(c) denotes the energy of the open system when in micro state c
N (c) denotes the number of particles of the open when in micro state c
Let us now take the partial derivative of all the terms in the above equation, with
respect to the variable µ, keeping the temperature and volume constant. We have,
   
DR

∂Q X
= βN (c) exp − β{E(c) − µN (c)} = βhN iQ(T, V, µ) (6.45)
∂µ T,V c
 2  "     #
∂ Q ∂Q ∂hN i
= β hN i +Q (6.46)
∂µ2 T,V ∂µ T,V ∂µ T,V
7
The left hand side of the above equation equals β 2 hN 2 iQ.
Substituting this in the above, we get,
7
  X  
∂Q
= βN (c) exp − β{E(c) − µN (c)}
∂µ T,V c
 2
 X  
∂ Q
= β 2 [N (c)]2 exp − β{E(c) − µN (c)}
∂µ2 T,V c

= β 2 hN 2 iQ.
6.3 Statistics of Number of Particles 95
 
∂hN i
β 2 hN 2 iQ = β 2 hN i2 Q + βQ (6.47)
∂µ T,V

 
2 2 2 ∂hN i
σ = hN i − hN i = kB T (6.48)
∂µ T,V

Since hN i = λ Q1 , we have,
∂hN i dλ
= Q1 = βλQ1 (6.49)
∂µ dµ
We have shown
 
2 ∂hN i
σN = kB T (6.50)
∂µ T,V
∂hN i
Substituting in the above = λβQ1 , we get,
∂µ
2
σN = kB T Q1 βλ (6.51)
= λQ1 = hN i (6.52)
consistent with what we have shown earlier : the distribution of N is Poissonian
with mean=variance=λQ1
In what follows, we shall derive an expression for the fluctuations of N in terms
T
of isothermal compressibility - a property measurable experimentally.
AF
6.3.7 Number Fluctuations and Isothermal Compressibility
Let us express the number fluctuations in terms of experimentally measurable prop-
erties 8 of the open system. Let us define
DR

V
v= .
hN i
It is called specific volume. It is the volume per particle. We have,
8
We shall show that,
 
∂hN i hN i2
= kT (6.53)
∂µ T,V V
where kT denotes isothermal compressibility - an experimentally measurable
property. Isothermal compressibility is defined as
 
1 ∂V
kT = − (6.54)
V ∂P T
96 6. Open System : Grand Canonical Ensemble
   
∂hN i ∂(V /v)
= (6.55)
∂µ T,V ∂µ T,V

 
V ∂v
=− 2 (6.56)
v ∂µ T,V

 
hN i2 ∂v
=− (6.57)
V ∂µ T,V

In the above we can express,


     
∂v ∂v ∂P
= (6.58)
∂µ T,V ∂P T,V ∂µ T,V

Employing Gibbs-Duhem relation, we find 9 ,


 
∂P hN i
= (6.60)
∂µ T V
Therefore,
     
∂v hN i ∂v 1 ∂v
∂µ T,V
=
V ∂P T,V
=
v ∂P T,V
= −kT T (6.61)
AF
Finally we get,
 
∂hN i hN i2
= kT (6.62)
∂µ T,V V
DR

 
∂hN i
σ 2 = kB T (6.63)
∂µ T,V

hN i2
= kB T kT (6.64)
V

σ2 kB T
= kT (6.65)
hN i2 V
9
Gibbs - Duhem relation reads as hN idµ = V dP − SdT. At constant temperature,
we have,
hN idµ = V dP,
 
∂P hN i
= (6.59)
∂µ T V
6.3 Statistics of Number of Particles 97

Thus the fluctuations in the number of molecules of an open system is directly


proportional to the isothermal compressibility. Hence we expect isothermal com-
pressibility to be positive 10 .
The number fluctuations are small in the thermodynamic limit; they are of the
order of inverse of the square-root of the number of particles in the system.
Thus equilibrium fluctuations are related to an appropriate susceptibility which
measures the response of the system to a small external perturbation. When heated,
the system responds by raising its temperature. Energy absorbed by heat divided
by the increase in temperature is called the heat capacity. The volume does not
change during this process. We saw that the equilibrium energy fluctuations are
proportional heat capacity.
Thus, the susceptibility is heat capacity for energy fluctuations; it is isothermal
compressibility for the number fluctuations. These are special cases of a more gen-
eral principle called Fluctuation-Dissipation theorem enunciated by Albert Einstein
in the context of Brownian motion.
However, close to first order phase transition, isothermal compressibility di-
verges. The fluctuations in the number of particles become large; they increase to
the order of the system size.

6.3.8 Alternate Derivation of the Relation :


2
σN /hN i2 = kB T kT /V
I had, earlier, derived a relation is between the number fluctuation and isothermal
compressibility. I had followed closely Pathria 11 . Here is an alternate derivation.

σ 2 = kB T
∂hN i

T
Start with a fluctuation-dissipation relation 12 derived earlier,

AF
(6.66)
∂µ V,T
We consider the reciprocal,
 
∂µ
DR

(6.67)
∂hN i V,T

We observe that µ is a function13 of T and P . We are keeping T a constant. Hence


µ can change only when P changes.
Gibbs-Duhem relation tells us,
hN idµ = V dP − SdT. (6.68)
When T is a constant, we have dT = 0. This gives us
10
The relative fluctuations of energy in a canonical ensemble is proportional to
heat capacity at constant volume. Hence we expect heat capacity to be positive.
11
R. K. Pathria, Statistical Mechanics, Second Edition, Butterworth and Hene-
mann (1996).
12
relating number fluctuations to response to small changes in the chemical poten-
tial.
13
Gibbs and Duhem told us that the three intensive properties T , P , and µ are not
all independent. Only two of them are independent. The third is automatically
fixed by the other two.
98 6. Open System : Grand Canonical Ensemble

V
dµ = dP (6.69)
hN i

   
∂µ V ∂P
= (6.70)
∂hN i T,V
hN i ∂hN i T,V

 
∂µ V
= (6.71)
∂P T,V hN i
hN i
Let us now define ρ = , which denotes the particle density : number of particles
V
per unit volume. We have,
     
∂P ∂P 1 ∂P
= = (6.72)
∂hN i V,T ∂(ρV ) V,T V ∂ρ V,T
The density can be changed either by changing hN i and/or V . Here we change ρ
infinitesimally, keeping V constant. As a result the pressure changes infinitesimally.
Let me repeat : both these changes happen at constant V and T .
As far as P is concerned, it does not care whether ρ has changed by change of
hN i or of V . Hence it is legitimate to write
   
∂P ∂P
= (6.73)
∂ρ V,T ∂(hN i/V ) hNi,T

=−
V2

∂P

T (6.74)
AF
hN i ∂V hNi,T

Thus we get,
   
∂µ V ∂P
=
DR

∂hN i V,T hN i ∂hN i V,T

 
1 ∂P
=
hN i ∂ρ V,T

 
V2 ∂P
=− (6.75)
hN i2 ∂V hNi,T

Take a reciprocal of the above and get,


   
∂hN i hN i2 ∂V
=− 2 (6.76)
∂µ V,T V ∂P hNi,T
Then we get,
  
2 hN i2 ∂V hN i2
σN = kB T − 2 = kB T kT
V ∂P T,hNi V

2
σN kB T
= kT (6.77)
hN i2 V
6.4 Energy Fluctuations 99

6.4 Energy Fluctuations


We start with the grand canonical partition function and its partial derivatives, the
first and second, with respect to the variable β :
X N
Q= λ i exp(−βE1 )
i
  X
∂Q
=− λNi Ei exp(−βEi )
∂β λ,V i
  X
∂2Q
= λNi Ei2 exp(−βEi )
∂β 2 λ,V i

The first and second moment of energy are given bu,


 
1 ∂Q
hEi = −
Q ∂β λ,V

 
1 ∂2Q
hE 2 i =
Q ∂β 2 λ,V

To derive an expression for the fluctuations of energy in a grand canonical ensemble,


we proceed as follows.
  "   #!
∂hEi
∂β λ,V
=−
∂ 1 ∂Q
∂β Q ∂β λ,V
T
AF
λ,V
 2  "   #2
1 ∂ Q 1 ∂Q
=− +
Q ∂β 2 λ,V Q ∂β λ,V
2 2

= − hE i − hEi
DR

2
= −σE:OPEN
We proceed in the same fashion and  obtain 
a similar expression for the fluctuations
∂hEi 2
of energy in a canonical ensemble14 = −σE:CLOSED . Note for an open
∂β N,V
system, the partial derivative of energy with respect to β is taken at constant λ
and V . For a closed system, the partial derivative of energy with respect to beta is
2 2
taken at constant N and V . What is the relation between σE:OPEN and σE:CLOSED ?
The key to answering to this question is in the following entities : • U (T, V, N )
with N constant and
• U (T, V, λ) = U (T, V, N (T, V, λ)) with N depending on T , V , and λ and with λ
kept constant instead of N .
The central relation15 is,
X ∂Q X ∂2Q X 2
14
Q= exp(−βEi ); =− Ei exp(−βEi ); 2
= Ei exp(−βEi );
i
∂β i
∂β i
 2
1 ∂Q ∂hEi 1 ∂2Q 1 ∂Q 2
hEi = − ; =− + = −σE:CLOSED
Q ∂β ∂β Q ∂β 2 Q ∂β
15
Equation (6.78) above can be derived from first principles. Formally we start
with U (T, V, λ) ≡ U (T, V, N (T, V, λ)). We have,
100 6. Open System : Grand Canonical Ensemble
       
∂U ∂U ∂U ∂N
= + × . (6.78)
∂T λ,V ∂T N,V ∂N T,V ∂T λ,V
 
∂U U (T + ∆T, V, N (T + ∆T, V, λ)) − U (T, V, N (T, V, λ))
= ∆Tlimit
→0
∂T λ,V ∆T
In the above we express
 
∂N
N (T + ∆T, V, λ) = N (T, V, λ) + ∆T = N + (∆N )λ,V ,
∂T λ,V
 
∂N
where we have used the short-hand notation (∆N )λ,V = ∆T . There-
∂T λ,V
fore
 
∂U U (T + ∆T, V, N + (∆N )λ,V ) − U (T, V, N )
= ∆Tlimit
→0
∂T λ,V
∆T
We can write
 
∂U
U (T + ∆T, V, N + (∆N )λ,V ) = U (T, V, N ) + ∆T
∂T V,N
 
∂U
+(∆N )λ,V
∂N T,V

Therefore, T
     
∂U ∂U (∆N )λ,V ∂U
= + ×
AF
∂T λ,V ∂T V,N ∆T ∂N T,V
     
∂U ∂N ∂U
= + ×
∂T V,N ∂T λ,V ∂N T,V
DR
6.4 Energy Fluctuations 101

The last term in Eq. (6.78) above involving the partial derivative of N with respect
to T can be evaluated, see box below.

Start with N (T, V, µ(T, λ)), where µ = kB T ln λ. We have,


       
∂N ∂N ∂N ∂µ
= + × (6.79)
∂T λ,V ∂T µ,V ∂µ T,V ∂T λ,V
The first term on the right hand side of Eq. (6.79) above can be evaluated, see
the box below.

Consider N (T, V, µ). At constant V we can write,


   
∂N ∂N
dN = dT + dµ
∂T µ ∂µ T
dN = 0 implies,
   
∂N ∂N
dT = − dµ
∂T µ ∂µ T

     
∂N ∂N ∂µ
=− ×
∂T µ
∂µ T
∂T N

Therefore Eq.(6.79) can be written as,



∂N

=−

∂N

×

∂µ

+

∂N

×

∂µ

T
AF
∂T λ,V ∂µ T,V ∂T N,V ∂µ T,V ∂T λ,V

  "    #
∂N ∂µ ∂µ
= −
DR

∂µ T,V ∂T λ,V ∂T N,V

  "   #
∂N µ ∂µ
= − (6.80)
∂µ T,V
T ∂T N,V
 
∂µ µ
In the last step we have made use of the relation : = , see box
∂T λ,V T
below.
 
∂µ
λ = exp(βµ); µ = kB T ln λ; = kB ln λ = µ/T
∂T λ
102 6. Open System : Grand Canonical Ensemble

Consider the term within square brackets on the right hand side of Eq. (6.80);
we can show that "   #  
µ ∂µ 1 ∂U
− = ,,
T ∂T N,V T ∂N T,V
see box below.

At constant volume we have dU = T dS + µdN. Therefore,


    "   #
∂U ∂S µ ∂S
= µ+T =T +
∂N T,V ∂N T,V T ∂N T,V
"   #
µ ∂µ
=T −
T ∂T N,V

In the above, the last step follows from one of Maxwell’s relations, see box below.

Maxwell’s Relation
F = U − TS
dF = −P dV + µdN − SdT
T
= µdN − SdT (at constant V )
AF
We have
   
∂F ∂F
=µ ; = −S
∂N T ∂T N
   
∂2F ∂µ ∂2F ∂S
DR

= ; =−
∂T ∂N ∂T N ∂N ∂T ∂N T
2 2
   
∂ F ∂ F ∂µ ∂S
= ⇒ =−
∂T ∂N ∂N ∂T ∂T N ∂N T

     
∂N ∂N 1 ∂U
Therefore Eq. (6.80) can be written as, = .
∂T λ,V ∂µ T,V T ∂N T,V
Now we can go back to Eq. (6.78) and write it as
     2  
∂U ∂U 1 ∂U ∂N
= + (6.81)
∂T λ,V ∂T N,V T ∂N T,V ∂µ T,V
6.4 Energy Fluctuations 103

Make use of the relation16 ,   2


∂N σN
=
∂µ T,V kB T
and re-write Eq. (6.81) as
     2
∂U ∂U 1 2 ∂U
= + σ
∂T λ,V ∂T N,V kB T 2 N ∂N T,V

     2
∂U ∂U ∂U
kB T 2 = kB T 2 2
+ σN
∂T λ,V
∂T N,V
∂N T,V

 2
2 ∂hEi
σE:OPEN = kB T 2 C V + 2
σN (6.82)
∂hN i T,V

Thus the energy fluctuations in an open system has an extra component given by
the second term in the above, arising due to fluctuations in the number of particles.

T
AF
DR

16

X
Q= exp[−β(E − µN )]
i

∂Q X ∂2Q X 2
=β N exp[−β(E − µN )]; 2
= β2 N exp[−β(E − µN )]
∂µ i
∂µ i

1 ∂Q 1 ∂2Q
= βhN i; = β 2 hN 2 i
Q ∂µ Q ∂µ2
   2
∂hN i ∂ 1 ∂Q 1 ∂2Q 1 ∂Q 
β = = − = β 2 hN 2 i − hN i2
∂µ ∂µ Q ∂µ Q ∂µ2 Q ∂µ

∂hN i 1 2
= σ
∂µ kB T N
DR
AF
T
7. Quantum Statistics

7.1 Occupation Number Representation


Let {1, 2, · · · } label the single-particle quantum states of a system. Let {ǫi : i =
1, 2, · · · } denote the corresponding energies. Notice that there can be several quan-
tum states with the same energy.
We have N non-interacting quantum particles occupying the single particle
quantum states. A micro state of the macroscopic system of N (quantum) particles
in a volume V is uniquely specified by a string of numbers {n1 , n2 , · · · }, where ni
is the number of particles in the single- particle quantum state i. Thus, a string of
occupation numbers uniquely describes a micro state of a quantum system1 . Note
that such a string should obey the constraint : n1 + n2 + · · · = N. The energy of
a micro state is n1 ǫ1 + n2 ǫ2 + · · · . The canonical partition function can now be
written as
Q(T, V, N ) =
X⋆ T
exp [−β(n1 ǫ1 + n2 ǫ2 + · · · )] (7.1)
AF
{n1 ,n2 ,··· }

where the sum runs over all possible micro


P states i.e. all possible strings of occu-
pation numbers obeying the constraint i ni = N. To remind us of this constraint
I have put a star as a superscript to the summation sign.
DR

7.2 Open System and Q(T, V, µ)


The presence of the constraint renders evaluation of the sum a difficult task. A way
out is to transform the variable N in favour of λ, called the fugacity; λ is related
to the chemical potential µ; the relation is λ = exp(βµ). The transform defines the
grand canonical partition function2 , see below.
1
for classical Q
particles, the same string of occupation numbers represent a collec-
tion of N !/ i ni ! micro states.
2
we have already seen that a partition function is a transform. We start with
b
density of states Ω(E) and transform the variable E in favour of β = 1/[kB T ]
to get canonical partition function:
Z ∞
Q(β, V, N ) = b
dE Ω(E, V, N ) exp(−βE(V, N )).
0

We can consider the grand canonical partition function as a transform of the


canonical partition function with the variable N transformed to fugacity λ (or
chemical potential µ = kB T ln(λ).)
106 7. Quantum Statistics

X
Q(β, V, λ) = λN Q(β, V, N ) (7.2)
N=0

This provides a description of an open system with its natural variable : temper-
ature, volume and chemical potential. The system can exchange energy as well as
matter with the environment. Thus energy E, and the number of particles N , of
the system would be fluctuating around the means hEi and hN i respectively, where
the angular bracket denotes averaging over a grand canonical ensemble of the open
system.
Thus we have,

X X⋆
Q(T, V, µ) = λN exp [−β(n1 ǫ1 + n2 ǫ2 + · · · )]
N=0 {n1 ,n2 ,··· }

X X⋆
= λn1 +n2 +··· exp [−β(n1 ǫ1 + n2 ǫ2 + · · · )]
N=0 {n1 ,n2 ,··· }

X X⋆
= exp [−β(ǫ1 − µ)n1 − β(ǫ2 − µ)n2 + · · · )]
N=0 {n1 ,n2 ,··· }

X X⋆
= [exp {−β (ǫ1 − µ)}]n1 ×
N=0 {n1 ,n2 ,··· }

[exp {−β (ǫ2 − µ)}]n2 × · · ·



X X⋆
= xn
N=0 {n1 ,n2 ,··· }
1
1
× xn
2 ×···
2 T (7.3)
AF
where a short hand notation : xi = λ exp(−βǫi ) = exp[−β(ǫi − µ)] is introduced
while writing the last line. We have a restricted sum over strings of occupation num-
bers. The restriction is that the occupation numbers constituting a string should
add to N . We then take a sum over N from 0 to ∞. which removes the restriction.
DR

To appreciate this, see Donald A McQuairrie, Statistical Mechanics,


Harper and Row(1976)p.77;Problem:4-6. I have worked out this problem below.


X X⋆
Consider evaluating the sum, I1 = xn1 n2
1 x2 , where ni = 0, 1, 2 for
N=0 {n1 ,n2 }
i = 1, 2, and ⋆ ⇒ n1 + n2 = N. Write down all the terms in this sum explicitly and
2 X
Y 2
show it is equivalent to xn
i ..
i=1 n=0
Let n1 and n2 denote two integers. Each of them can take values 0, 1, or 2.
Consider first the restricted sum :
X∞ X⋆ n n
I1 = x1 1 x2 2 (7.4)
N=0 {n1 ,n2 }

where the star over the summation sign reminds us of the restriction n1 + n2 = N .
Thus we get Table (1).
7.2 Open System and Q(T, V, µ) 107

N n1 n2 xn1 n2
1 x2

0 0 0 1

1 1 0 x1
0 1 x2

2 2 0 x21
0 2 x22
1 1 x1 x2

3 3 0 -
2 1 x21 x2
1 2 x1 x22
0 3 -

- - -
2 2 x21 x22
- - -

5 - - -
Table 1. Terms in the restricted sum

T
AF
From the table, we can write down I1 as,
I1 = 1 + x1 + x2 + x21 + x22 + x1 x2 + x21 x2 + x1 x22 + x21 x22 (7.5)
Now consider the product over the unrestricted sums,
DR

2 X
Y 2
I2 = xn
i (7.6)
i=1 n=0
2
Y
= (1 + xi + x2i )
i=1

= (1 + x1 + x21 ) × (1 + x2 + x22 )
= 1 + x2 + x22 + x1 + x1 x2 + x1 x22 + x21 + x21 x2 + x21 x22 = I1 (7.7)
In fact we need not have restricted n1 and n2 to the values 0, 1, and 2. We
could have allowed them to run from zero and infinity, see below.=
108 7. Quantum Statistics
∞ X
X ∞ X
X N

xn1 n2
1 x2 = xn N−n
1 x2
{n1 ,n1 }
N=0 N=0 n=0

X N 
X n
x1
= xN
2
N=0 n=0
x2
X∞
1 − (x1 /x2 )N+1
= xN
2
N=0
1 − (x1 /x2 )
X∞
xN+1
2 − x − 1N+1
=
N=0
x2 − x1
 
1 x2 x1
= −
x − 2 − x1 1 − x2 1 − x1
 
1 x2 (1 − x1 ) − x − 1(1 − x2 )
=
x2 − x1 (1 − x1 )(1 − x2 )
   
1 1
= (7.8)
1 − x1 1 − x2
QED

T
AF
We can now write the grand canonical partition function of N quantum particles
occupying single-particle quantum states determined by volume V .
all
YX all
YX
Q(T, V, µ) = xn
i = [λ exp(−βǫi )]n
DR

i n=0 i n=0
all
YX  n
= exp {−β (ǫi − µ)} (7.9)
i n=0

7.3 Fermi-Dirac Statistics


We have n = 0, 1. This is a consequence of Pauli’s exclusion principle: No single
quantum state can have more than one fermion.
Y 
QF D (T, V, µ) = 1 + exp[−β(ǫi − µ) (7.10)
i

7.4 Bose-Einstein Statistics


We have n = 0, 1, · · · ∞. Let exp[−β(ǫi − µ)] < 1 ∀ i. This condition is met if
ǫi > µ ∀ i. Then we have
Y 1
QBE = (ǫi > µ ∀ i) (7.11)
i
1 − exp[−β(ǫ i − µ)]
7.6 Maxwell-Boltzmann Statistics 109

7.5 Classical Distinguishable Particles


If the particles are classical and distinguishable, then the string {n1 , n2 , · · · } will
not uniquely specify a micro state. The string will have a set of classical micro
states associated with it. The number of micro states associated with the string is
given by,

b 1 , n2 , · · · ) = N!
Ω(n (7.12)
n1 !n2 ! · · ·
We then have

X X⋆ N!
QCS (T, V, µ) = λN [exp(−βǫ1 )]n1 ×
N=0
n1 !n2 ! · · ·
{n1 ,n2 ··· }

[exp(−βǫ2 )]n2 × · · ·


" #N ∞
X N
X X
= λ exp(−βǫi ) = λN [Q1 (T, V )]N (7.13)
N=0 i N=0

where Q1 (T, V ) is the single-particle partition function.


We have already seen that the partition function for classical distinguishable
non-interacting point particles leads to an entropy which is not extensive. This
is called Gibbs’ paradox. To take care of this, Boltzmann asked us to divide the
number of micro states by N !, saying that the particles are indistinguishable.
T
The non-extensivity of entropy indicates a deep flaw in classical formulation
of statistical mechanics. Classical particles do not simply exist in nature. We have
AF
only quantum particles : Bosons and Fermions. It is quantum mechanics that would
set the things right eventually. But quantum mechanics had not arrived yet, in the
times of Boltzmann.
Boltzmann corrected the (non-extensive entropy) flaw by introducing a strange
notion of indistinguishability of particles. Genius he was, Boltzmann was almost
DR

on the mark. Division by N ! arises naturally in quantum mechanical formulation,


because of the symmetry of the wave function
Accordingly in the derivation of grand canonical partition function, we shall
divide the classical degeneracy by N !, as recommended by Boltzmann, see below,
and call the resulting statistics as Maxwell-Boltzmann statistics.

7.6 Maxwell-Boltzmann Statistics


Formally we have,

X X⋆ 1
QM B = λN exp[−β(n1 ǫ1 + n2 ǫ2 + · · · )]
N=0
n1 !n2 ! · · ·
{n1 ,n2 ,··· }


X X ⋆ [λ exp(−βǫ1 )]n1 [λ exp(−βǫ2 )]n2
= × × ···
N=0 {n1 ,n2 ,··· }
n1 ! n2 !
110 7. Quantum Statistics

X X ⋆ [exp(−β(ǫ1 − µ)]n1 [exp(−β(ǫ2 − µ)]n2
= × × ···
N=0 {n1 ,n2 ,··· }
n1 ! n2 !


X X ⋆ xn1 xn2
1 2
= ···
n1 ! n2 !
N=0 {n1 ,n2 ,··· }


! ∞
!
X xn
1
X xn
2
= ···
n=0
n! n=0
n!

= exp(x1 ) exp(x2 ) · · ·

Y Y
= exp(xi ) = exp[λ exp(−βǫi )]
i i

Y
= exp[exp{−β(ǫi − µ)}] (7.14)
i

We can also express the grand canonical partition function for classical indistin-
guishable ideal gas as,
X 
QM B = exp(x1 ) exp(x2 ) · · · = exp xi
X
i
T 
AF
= exp exp[−β(ǫi − µ)]
i
X
= exp[λ exp(−βǫi )]
i
DR

= exp[λQ1 (T, V )] (7.15)

7.6.1 QM B (T, V, N ) → QM B (T, V, µ)


We could have obtained the above in a simple manner, by recognising that
[QM B (T, V, N = 1)]N
QM B (T, V, N ) = (7.16)
N!
Then we have,

X [QM B (T, V, N = 1)]N
Q(T, V, λ) = λN
N=0
N!

= exp[λQM B (T, V, N = 1)] (7.17)


7.7 Q(T, V, µ) and G(T, V, µ) 111

7.6.2 QM B (T, V, µ) → QM B (T, V, N )


We start with QM B (T, V, µ) = exp[λQM B (T, V, N = 1)]. Taylor expanding the
exponential,

X QN
M B (T, V, N = 1)
QM B (T, V, µ) = λN (7.18)
N=0
N!

The coefficient of λN is the canonical partition function3 and hence is given by

QM B (T, V, N = 1)N
QM B (T, V, N ) =
N!

7.7 Thermodynamics of an open system


We can write the grand canonical partition function for the MB,BE, and FD statis-
tics as


 exp[exp{−β(ǫi − µ)}] Maxwell − Boltzmann









Y
Q= 1 (7.19)
 1 − exp [−β(ǫi − µ)] BoseEinstein
i 









1 + exp {−β(ǫi − µ)} Fermi − Dirac
T
AF
The connection to thermodynamics is established by the expression for grand
potential denoted by the symbol G(T, V, µ). We have,
G(T, V, µ) = −kB T ln Q(T, V, µ) (7.20)
DR

Recall from thermodynamics that the G is obtained as a Legendre transform of


U (S, V, N ) : S → T ; N → µ; and U → G.
G(T, V, µ) = U − T S − µN (7.21)

   
∂U ∂U
T = ; µ= (7.22)
∂S V,N ∂N S,V

From the above, we get, dG = −P dV − SdT − N dµ It follows,


     
∂G ∂G ∂G
P =− ; S=− ; N =− (7.23)
∂V T,µ ∂T V,µ ∂µ T,µ
If we have an open system of particles obeying Maxwell-Boltzmann, Bose-Einstein,
or Fermi-Dirac statistics at temperature T and chemical potential µ, in a volume
V , then the above formulae help us calculate the pressure, entropy and average
3
Recall that

X
QM B (T, V, λ) = λN QM B (T, V, N )
N=0
112 7. Quantum Statistics

number of particles in the system. In fact, in the last section on grand canonical
ensemble, we have derived formal expressions for the mean and fluctuations of
the number of particles in an open system; we have related the fluctuations to
isothermal compressibility - an experimentally measurable property.
The grand potential for the three statistics is given by,
G(T, V, µ) = −kB T ln Q
 P
 −kB T i exp[−β(ǫi − µ)] Maxwell − Boltzmann



 P
= kB T i ln [1 − exp {−β(ǫi − µ)}] Bose − Einstein (7.24)




 P
−kB T i ln [1 + exp {−β(ǫi − µ)}] Fermi − Dirac

7.8 Average number of particles, hN i

7.8.1 Maxwell-Boltzmann Statistics


Y
Q(T, V, µ) = exp [exp {−β (ǫi − µ)}} (7.25)
i
G(T, V, µ) = −kB T ln Q
X
= −kB T exp [−β(ǫi − µ)] (7.26)

∂G
 X
i
T
AF
=− exp[−β(ǫi − µ) (7.27)
∂µ T,V i
  X
∂G
hN i = − = exp[−β(ǫi − µ)]
∂µ T,V i
DR

X
=λ exp(−βǫi ) = λQ1 (T, V, µ) (7.28)
i

In the above Q1 is the single-particle canonical partition function. The above result
on hN i is consistent with formal result from grand canonical formalism :
The grand canonical partition function is formally given by the transform of
canonical partition function. The thermodynamic variable N is transformed to fu-
gacity λ (or to chemical potential µ = kB T ln λ. Thus we have,

X
Q(T, V, µ) = λN Q(T, V, N )
N=0
X∞
QN
= λN , 1

N=0
N!
X∞
QN
1
= exp(βµN )
N=0
N!
= exp(λQ1 ) (7.29)
The grand potential is given by
7.8 Average number of particles, hN i 113

G(T, V, µ) = −kB T ln Q(T, V, µ)


= −kB T λ Q1
= −kB T exp(µ/kB T )Q1 (7.30)
The average number of particles in the system is then given by,
 
∂G
hN i = − = λQ1 (T, V ) (7.31)
∂µ T,V

7.8.2 Bose-Einstein Statistics

Y 1
Q(T, V, µ) = (7.32)
i
1 − exp[−β(ǫi − µ)]

X
ln Q = − ln [1 − exp {−β(ǫi − µ)}] (7.33)
i

  X
∂G exp[−β(ǫi − µ)]
hN i = − = (7.34)
∂µ T,V i
1 − exp[−β(ǫi − µ)]

7.8.3 Fermi-Dirac Statistics

Q=
Y
1 + exp[−β(ǫi − µ)]
T (7.35)
AF
i

X
G = −kB T ln Q = −kB T ln [1 + exp{−β(ǫi − µ)}] (7.36)
DR

  X
∂G exp[−β(ǫi − µ)]
hN i = − = (7.37)
∂µ T,V i
1 + exp[−β(ǫi − µ)]

7.8.4 Study of a System with fixed N Employing Grand


Canonical Formalism
In my next lecture, I shall derive an expression for the mean of occupation
P number,
nk , of a single particle quantum state. Then we can express hN i = k hnk i. These
results on hN i shall be the same as the one derived in the last three subsections.
Grand canonical formalism is not the best suited for studying ideal gas - classical
or quantum. Recall, we introduced an adhoc and awkward - indistinquishability-
factor of N !, suggested by Boltzmann, to restore the extensivity of entropy while
studying closed system. In an open system the number of particles fluctuate and
the issue of non-extensivity becomes more awkward.
We shall adopt the following strategy. We shall employ grand canonical formal-
ism, but with a fixed N . The price we have to pay is that in such an approach, the
chemical potential is no longer an independent property. It becomes a function of
temperature. Let me explain.
114 7. Quantum Statistics

Let us say we keep µ fixed and change the temperature4 . The mean number
of particles in the system changes when temperature changes. To restore the value
of hN i to the fixed value N , we change the chemical potential. In what follows,
we shall investigate the behaviour of particles under Maxwell-Boltzmann, Bose-
Einstein, and Fermi-Dirac statistics employing grand canonical formalism but with
a fixed N ; This implies µ is not any more an independent property; it is a function
of T .
I must say there is nothing unphysical about this strategy. We are studying
a physical system enclosed by a non-permeable wall - a wall that does not permit
particle exchange. The chemical potential µ, is a well defined property of the system.
It is just that µ is not any more under our control5 . The system automatically selects
the value of µ depending on the temperature.

7.9 Fermi-Dirac, Maxwell-Boltzmann and Bose-Einstein


Statistics are the same at High Temperature and/or
Low Densities
I shall show below that three statistics are the same at high temperatures and / or
low densities by an easy method, an easier method and perhaps the easiest method.

7.9.1 Easy Method : ρΛ3 → 0 T


AF
For ease of notation let
ǫi − µ
ηi = (7.38)
kB T
and write the grand canonical partition function under Maxwell-Boltzmann, Bose-
DR

Einstein, and Fermi-Dirac statistics as



 exp[exp(−ηi )] Maxwell − Boltzmann






Y  1
Bose − Einstein
Q= 1 − exp(−ηi ) (7.39)


i 






1 + exp(−ηi ) Fermi − Dirac
Let us make an estimate of ηi at high temperatures and low densities.
We had shown earlier that the canonical partition function for an ideal gas can
be written as, see notes,
VN 1
Q(T, V, N ) = (7.40)
N ! Λ3N
4
This is permitted in grand canonical formalism since T and µ are independent
properties of the open system. T is determined by the heat bath and µ is deter-
mined by the particle bath or reservoir.
5
µ is not determined by us externally by adjusting the particle bath.
7.9 All Statistics are Same at Hight T , Low ρ 115

where the thermal wave length is given by


h
Λ= √ (7.41)
2πmkB T
The free energy is given by
F (T, V, N ) = −kB T ln Q(T, V, N ) (7.42)

= −kB T [−3N ln Λ + N ln V − N ln N + N ] (7.43)


When you take the partial derivative of the free energy with respect to N keeping
temperature and volume constant, you get the chemical potential6 . Therefore,
 
∂F
µ= = kB T [3 ln Λ − ln V + 1 + ln N − 1] (7.48)
∂N T,V
 
= kB T ln Λ3 + ln(N/V ) (7.49)

= kB T ln(ρΛ3 ) (7.50)

µ
= ln(ρΛ3 ) (7.51)
kB T
where ρ = N/V is the number density, i.e. number of particles per unit volume.
Thus we get,
ǫi T
ηi = − ln(ρΛ3 ) (7.52)
kB T
AF
We have earlier shown that classical limit obtains when ρΛ3 → 0. Thus in this limit,
ηi → ∞ or exp(−ηi ) → 0. Let xi = exp(−ηi ). Let us express the grand canonical
partition function for the three statistics in terms of the small parameter xi .

 exp(xi ) Maxwell − Boltzmann


DR





Y  1
Bose − Einstein
Q= 1 − xi (7.53)

i 







1 + xi Fermi − Dirac
6
In thermodynamics we have
F = U − TS (7.44)

dF = dU − T dS − SdT (7.45)
From the first law of thermodynamics we have dU = T dS − P dV + µdN . Sub-
stituting this in the expression for dF above we get,
dF = −P dV + µdN − SdT (7.46)
Therefore,
 
∂F
µ= (7.47)
∂N T,V
116 7. Quantum Statistics

In the limit xi → 0 we have exp(±xi ) = 1 ± xi and (1 − xi )−1 = 1 + xi . Therefore


in the limit xi → 0, we have

 exp(xi ) xi∼
→0 1 + xi Maxwell − Boltzmann







Y 1 ∼
1 + xi Bose − Einstein
Q= 1 − xi xi →0 (7.54)

i 







1 + xi = 1 + xi Fermi − Dirac
For all the three statistics, the grand canonical partition function take the same
expression. Bosons, Fermions and classical indistinguishable particles behave the
same way when ρΛ3 → 0.
When do we get ρΛ3 → 0 ?
Note that Λ is inversely proportional to square-root of the temperature.
h
Λ= √
2πmkB T

Hence Λ → 0, when T → ∞. For a fixed temperature (Λ is constant), ρΛ3 → 0


when ρ → 0. For a fixed ρ, when T → ∞ ( the same as Λ → 0), then ρΛ3 → 0.

• Classical behaviour obtains at low densities and/or high temperatures.


• Quantum effects manifest only at low temperatures and/or high densities.

7.9.2 Easier Method : λ → 0


T
AF
Another simple way to show that the three statistics are identical in the limit of high
temperatures and low densities is to recognise (see below) that ρΛ3 → 0 implies
λ → 0. Here λ = exp(βµ), is the fugacity. Let us show this first.
DR

We have shown that


µ
= ln(ρΛ3 ) (7.55)
kB T
Therefore the fugacity is given by
 
µ
λ = exp = exp(ln[ρΛ3 ]) = ρΛ3 (7.56)
kB T
Thus ρΛ3 → 0 implies λ → 0.
In the limit λ → 0 we have, for Maxwell-Boltzmann statistics,
Y
QM B = exp[λ exp(−βǫi )] (7.57)
i

Y

λ→0 (1 + λ exp(−βǫi )) (7.58)
i

In the limit λ → 0, for Bose-Einstein statistics, we have,


7.10 Mean occupation number 117
Y 1
QBE = (7.59)
i
1 − λ exp(−βǫi )

Y

λ→0 [1 + λ exp(−βǫi )] (7.60)
i

For Fermi-Dirac statistics, we have exactly,


Y
QF D = [1 + λ exp(−βǫi )] (7.61)
i

Thus in the limit of high temperatures and low densities Maxwell Boltzmann statis-
tics and Bose Einstein statistics go over to Fermi - Dirac statistics.

b 1 , n2 , · · · ) = 1
7.9.3 Easiest Method Ω(n

We could have shown easily that in the limit of high temperature and low density,
the three statistics are identical by considering the degeneracy factor


 1 Bose − Einstein and Fermi − Dirac statistics

b
Ω= (7.62)

 1
 Maxwell − Boltzmann statistics
n1 !n2 ! · · · T
When the temperature is high the number of quantum states that become available
AF
for occupation is very large; When the density is low the number of particles in the
system is low. Thus we have a very few number of particles occupying a very large
of quantum states. In other words the number of quantum states is very large
compared to the number of particles.
Hence the number of micro states with ni = 0 or 1 ∀ i are overwhelmingly large
DR

compared to the those with ni ≥ 2. In any case, in Fermi-Dirac statistics ni is


always 0 or 1.
In other words, when particles are a few in number(number density is low) and
the accessible quantum levels are large in number (temperature is high), then micro
states with two or more particles in one or more quantum states are very rare. In
other words bunching of particles is very rare at low densities and high tempera-
tures. Almost always every quantum state is either unoccupied or occupied by one
particle. Very rarely will you find two or more particles in a single quantum state.
Hence the degeneracy factor is unity at high temperatures and/or low densities.
Hence all the three statistics are identical in the limit of high temperatures and/or
low densities.

7.10 Mean Occupation Number


Let us consider the single particle quantum state k with energy ǫk . Let nk be the
number of particles occupying the state k. nk is called the occupation number. It
is a random variable. Here we shall calculate the statistics of the random variable
nk .
118 7. Quantum Statistics

Let hnk i denote the average occupation number; the averaging is done over
a grand canonical ensemble of micro states. Formally we have for Bosons and
Fermions,
 
YX all all
X all
n X
 xi  nxn
k nxn
k
i6=k n=0 n=0 n=0
hnk i =   = all
(7.63)
Y all
X all
X X
 xn
i
 xn
k
xn
k
i6=k n=0 n=0 n=0

7.10.1 Ideal Fermions


For Fermions, n = 0, 1. Hence
xk 1 1
hnk iF D = = −1 = (7.64)
1 + xk xk + 1 exp[β(ǫk − µ)] + 1

7.10.2 Ideal Bosons


For Bosons n = 0, 1, 2, · · · , ∞. In other words, a single particle quantum state
can hold any number of particles. In fact, we shall see later, that at low enough
temperature all the particles would condense into the lowest energy state, i.e. the
ground state. We call this Bose-Einstein condensation. T
We carry out the summation7 , in the numerator and and the denominator of
the expression for hnk i, see above, analytically and get,
AF
xk xk 1 1
hnk iBE = (1 − xk ) = = −1 = (7.68)
(1 − xk )2 1 − xk xk − 1 exp[β(ǫk − µ)] − 1
DR

7.10.3 Classical Indistinguishable Ideal Particles


  !
YX ∞ n X∞
 x i  xn
k X∞
xn
n n k
n! n! n!
i6=k n=0 n=0 n=0
hnk iM B =   ! = ∞ n (7.69)
YX ∞ X∞ X xk
 xni  xn
k
n! n! n=0
n!
n=0
i6=k n=0

7
Consider
1
S(x) = 1 + x + x2 + · · · = (7.65)
1−x

dS 1
= 1 + 2x + 3x2 + 4x3 + · · · = (7.66)
dx (1 − x)2

dS x
x = x + 2x2 + 3x3 + · · · = (7.67)
dx (1 − x)2
7.11 Mean occupation : Some Remarks 119

In the above the summation in the numerator and the denominator are evaluated
analytically as follows. We start with the definition,
X∞
xn
= exp(x) (7.70)
n=0
n!

Differentiate both sides of the above equation with respect to x. You get
X∞ ∞ ∞
nxn−1 1 X nxn X nxn
exp(x) = = = (7.71)
n=0
n! x n=0 n! n=0
n!

Therefore,
xk exp(xk ) 1
hnk iM B = = xk = exp[−β(ǫk − µ)] = (7.72)
exp(xk ) exp[β(ǫk − µ)]

7.11 Mean Ocupation : Some Remarks


We can write for all the three statistics,

 +1 Fermi − Dirac


1
hnk i = where a = 0 Maxwell − Boltzmann (7.73)
exp[β(ǫk − µ)] + a 


−1 Bose − Einstein
T
Variation of hnk i with energy is shown in the figure, see next page. Note that the
ǫk − µ
AF
x axis is and the y axis is hnk i.
kB T

7.11.1 Fermi-Dirac Statistics


DR

We see that for Fermi-Dirac statistics the occupation number never exceeds unity.
When ǫk − µ is negative and |ǫk − µ| is large the value of hnk i tends to unity. For
ǫ = µ and T 6= 0, we have hnk i = 1/2.
It is also clear that at high temperature the chemical potential of a Fermion must
be negative, because its behaviour coincides with that of Bosons 8 and classical
indistinguishable particles9 .

7.11.2 Bose-Einstein Statistics


For Bosons we must have
ǫk > µ ∀ k.
In particular the lowest value of energy, say ǫ0 corresponding to the ground state
of the macroscopic system, must be greater than µ. Let us take ǫ0 = 0 without
loss of generality. Hence for Bosons µ must be negative. Also µ is a function of
8
for Bosons the chemical potential is negative at all temperature, and zero at
zero temperature and at temperatures less than a critical temperature called
Bose-Einstein condensation temperature.
9
for classical indistinguishable particles the chemical potential is negative at high
temperature, positive at low temperatures and zero at zero temperature.
120 7. Quantum Statistics

temperature. As we lower T the chemical potential increases and at T = TBEC , µ


becomes zero. TBEC is called the Bose-Einstein condensation temperature. At this
temperature the occupancy of the ground state becomes infinitely high. This leads
to the phenomenon of Bose-Einstein Condensation .

1.5
←− M axwell − Boltzmann
1
← Bose − E instein
hnk i

0.5
F ermi − Dirac →
0

−0.5

−1
T
AF
−3 −2 −1 0 1 2 3 4 5
ǫk − µ
kB T
DR

Fig. 7.1. Average occupation number of a quantum state under Bose-Einstein,


Fermi-Dirac, and Maxwell-Boltzmann statistics

7.11.3 Maxwell-Boltzmann Statistics


For the indistinguishable classical particles hnk i takes the familiar exponential decay
form,
hnk i = exp[−β(ǫk − µ)].

7.11.4 At High T and/or Low ρ all Statistics give the same hnk i
When
ǫk − µ
→ ∞,
kB T
all the three statistics coincide. We have already seen that at high temperatures
classical behaviour obtains. Then, the only way
7.12 Occupation Number : Distribution and Variance 121
ǫk − µ
kB T
can become large at high temperature (note in the expression T is in the denomi-
nator) is when µ is negative and its magnitude also should increase with increase
of temperature.
Thus for all the three statistics, at high temperature, the chemical potential µ
is negative and its magnitude must be large.
But then we know µ
= ln(ρΛ3 ).
kB T
This means that ρΛ3 << 1 for classical behaviour to emerge10 . This is in complete
agreement with our earlier surmise that classical behaviour obtains at low ρ and/or
high T . Hence all the approaches are consistent with each other and all the issues
fall in place.

7.12 Occupation Number : Distribution and Variance


In the last lecture, I introduced a random variable nk , which denotes the number
of particles in the quantum state k of energy ǫk . We call nk a random variable
because it takes values that are, in general, different for different micro states. For
Bose-Einstein and Fermi-Dirac statistics, a string of occupation numbers specifies
completely a micro state. We found that for Bosons and Fermions, the average value
of nk can be expressed as,
all
X
nxn
T
AF
k
n=0
hnk i = all
(7.74)
X
xn
k
n=0
DR

where, xk = exp[−β(ǫk − µ)].


Formally
all
X
hnk i = nP (n) (7.75)
n=0

In the above P (n) ≡ P (nk = n) is the probability that the random variable nk
takes a value n. Comparing the above with the first equation, we find
1
P (n) ≡ P (nk = n) = all
xn
k (7.76)
X
xm
k
m=0

where the denominator ensures that the total probability is normalized to unity.
Here, I shall work out explicitly P (n) for Fermi-Dirac, Bose-Einstein, and
Maxwell-Boltzmann statistics. We shall calculate the variance of the random vari-
able nk ; we shall derive the relative fluctuations : the standard deviation divided
by the mean, for the three statistics.
10
Note that ln(x) = 0 for x = 1 and is negative for x < 1. As x goes from 1 to 0,
the quantity ln(x) goes from 0 to −∞.
122 7. Quantum Statistics

7.12.1 Fermi-Dirac Statistics and Binomial Distribution


In Fermi-Dirac statistics the random variable nk can take only two values : n = 0
and n = 1. Thus,

 1

 for n = 0
 1 + xk


P (n) = (7.77)



 x

 k
for n = 1
1 + xk
We have
1
X xk
hnk i = nP (n) = ,
n=0
1 + xk
consistent with the result obtained earlier.
For convenience of notation let us denote the mean of the random variable nk
by the symbol ζ. We have,
xk
ζ = hnk i = .
1 + xk
We thus have,

 1 − ζ for n = 0
P (n) = (7.78)

ζ for n = 1
T
Thus Fermi-Dirac statistics defines a quantum coin with ζ and 1 − ζ as the proba-
bilities of ”Heads” (n = 0) and ”Tails” (n = 1) respectively.
AF
Formally,
1
X
hnk i = nP (n) = ζ (7.79)
n=0
DR

1
X
hn2k i = n2 P (n) = ζ (7.80)
n=0

σ 2 = hn2k i − hnk i2 = ζ(1 − ζ) (7.81)


The relative fluctuations of the random variable nk is defined as the standard
deviation σ divided by the mean ζ. Let us denote the relative fluctuation by the
symbol η. For Fermi-Dirac statistics we have,
r
σ 1
ηF D = = −1 (7.82)
ζ ζ

7.12.2 Bose-Einstein Statistics and Geometric Distribution


For Bosons,
P (nk = n) = P (n) = xn
k (1 − xk ) (7.83)
from which it follows,
7.12 Occupation Number : Distribution and Variance 123

X xk
ζ= nP (n) = (7.84)
n=0
1 − xk

consistent with the result obtained earlier. Inverting the above, we get,
ζ
xk = (7.85)
1+ζ
Then the probability distribution of the random variable nk can be written in a
convenient form
ζn
P (n) = (7.86)
(1 + ζ)n+1
The distribution is geometric, with a constant common ratio ζ/(ζ + 1). We have
come across geometric distribution earlier11 .
Calculation of variance is best done by first deriving an expression for the
moment generating function given by,

X
P̃ (z) = z n P (n) (7.87)
n=0

∞  n
1 X n ζ
= z (7.88)
1 + ζ n=0 1+ζ

=
1

1

T (7.89)
AF
1+ζ ζ z
1−
1+ζ

1
DR

= (7.90)
1 + ζ(1 − z)
Let us now differentiate P̃ (z) with respect to z and in the resulting expression set
z = 1. We shall get hnk i, see below.
∂ P̃ ζ
= (7.91)
∂z (1 + ζ(1 − z))2


∂ P̃
=ζ (7.92)
∂z z=1
Differentiating twice with respect to z and setting z = 1 in the resulting expression
shall yield the factorial moment hnk (nk − 1)i, see below.
11
Let me recollect : The simplest problem in which geometric distribution arises is
in coin tossing. Take a p-coin. Toss the coin until ”H” appears. The number of
tosses is a random variable with a geometric distribution P (n) = q n−1 p. We can
write this distribution in terms of ζ = hni = 1/p and get P (n) = (ζ − 1)n−1 /ζ n .
124 7. Quantum Statistics

∂2P 2ζ 2
= (7.93)
∂z 2 [1 + ζ(1 − z)]3

∂ 2 P
= hnk (nk − 1)i = 2ζ 2 (7.94)
∂z 2 z=1
hn2k i = 2ζ 2 + ζ (7.95)
2
σ = hn2k i − hnk i2 = ζ 2 + ζ (7.96)
r
σ 1
ηBE = = +1 (7.97)
ζ ζ
For doing the problem , You will need the following tricks.

X
S(x) = xn = 1 + x + x2 + x3 + · · ·
n=0
1
= (7.98)
1−x
X∞
dS
= nxn−1 = 1 + 2x + 3x2 + 4x3 + · · ·
dx n=1
1
= (7.99)
(1 − x)2
X∞
dS
x = nxn = x + 2x2 + 3x3 + 4x4 + · · ·
dx

=
n=1
x
T (7.100)
AF
(1 − x)2
  X∞
d dS
x = n2 xn−1 = 1 + 22 x + 32 x2 + 42 x3 + · · ·
dx dx n=1
DR

2x 1
= + (7.101)
(1 − x)3 (1 − x)2
  X∞
d dS
x x = n2 xn = x + 22 x2 + 32 x3 + 42 x4 + · · ·
dx dx n=1

2x2 x
= + (7.102)
(1 − x)3 (1 − x)2
You can employ the above trick to derive power series for ln(1 ± x), see below.
Z
dx x2 x3 x4
= − ln(1 − x) = x + + + ··· (7.103)
1−x 2 3 4

7.12.3 Maxwell-Boltzmann Statistics and Poisson Distribution


For Maxwell-Boltzmann statistics we have,
7.12 Occupation Number : Distribution and Variance 125

hnk i = xk (7.104)

xn
k 1
P (nk = n) ≡ P (n) = (7.105)
n! exp(xk )

ζn
= exp(−ζ) (7.106)
n!
The random variable nk has Poisson distribution. The variance equals the mean.
Thus the relative standard deviation is given by
σ 1
ηM B = = √ (7.107)
ζ ζ
We can now write the relative fluctuations for the three statistics in one single
formula as,

 +1 for Fermi − Dirac Statistics


r 

1
η= − a with a = 0 for Maxwell − Boltzmann Statistics (7.108)
ζ 




−1 for Bose − Einstein Statistics
Let us look at the ratio,
P (n)
r= T.
P (n − 1)
For the Maxwell-Boltzmann statistics, r = ζ/n. The ratio r is inversely proportional
AF
to n. This is the normal behaviour; inverse dependence of r on n is what we should
expect.
On the other hand, for Bose-Einstein statistics, the ratio is given by

P (n) ζ
DR

r= =
P (n − 1) ζ+1
r is independent of n. This means, a new particle will get into any of the quantum
states, with equal probability irrespective of how abundantly or how sparsely that
particular quantum state is already populated. An empty quantum state has the
same probability of acquiring an extra particle as an abundantly populated quantum
state.
Thus, compared to classical particles obeying Maxwell-Boltzmann statistics,
Bosons exhibit a tendency to bunch together. By nature, Bosons like to be together.
Note that this ”bunching-tendency” is not due to interaction between Bosons. We
are considering ideal Bosons. This bunching is purely a quantum mechanical effect;
it arises due to symmetry property of the wave function.
For Fermions, the situation is quite the opposite. There is what we may call
an aversion to bunching; call it anti-bunching if you like. No Fermion would like to
have another Fermion in its quantum state.
DR
AF
T
8. Bose-Einstein Condensation

8.1 Introduction
For bosons we found that the grand canonical partition function is given by,
Y 1
Q(T, V, µ) = (8.1)
i
1 − exp[−β(ǫi − µ)]

The correspondence with thermodynamics is established by the expression for grand


potential denoted by the symbol G(T, V, µ). We have,
G(T, V, µ) = −kB T ln Q(T, V, µ)

X
= kB T ln [1 − exp {−β(ǫi − µ)}] (8.2)
i
T
Recall, from thermodynamics, that G is obtained by Legendre transform of
AF
U (S, V, N ) : S → T ; N → µ; and U → G.
   
∂U ∂U
G(T, V, µ) = U − T S − µN ; T = ; µ= (8.3)
∂S V,N ∂N S,V
DR

From the above, we get,


dG = −P dV − SdT − N dµ (8.4)
It follows,
     
∂G ∂G ∂G
P =− ;S = − ;N = − .
∂V T,µ ∂T V,µ ∂µ T,µ

If we have an open system of bosons at temperature T and chemical potential µ, in


a volume V , then the above formulae help us calculate the pressure, entropy and
number of bosons in the system. In fact we have calculated the fluctuations in the
number of particles in the open system and related it to isothermal compressibility
- an experimentally measurable property.

X
8.2 hN i = hnk i
k

For bosons, we found that the average occupancy of a (single-particle) quantum


state k, is given by,
128 8. Bose-Einstein Condensation

λ exp(−βǫk )
hnk i = (8.5)
1 − λ exp(−βǫk )
where λ is fugacity. We have
λ = exp(βµ). (8.6)
In the above µ is the chemical potential and equals the energy change due to
addition of a single particle under constant entropy and volume :
 
∂U
µ= . (8.7)
∂N S,V
The average number of particles is given by,
X X λ exp(−βǫk )
hN i = hnk i = (8.8)
1 − λ exp(−βǫk )
k k

We would like to study Bosonic system with a fixed number of bosons at var-
ious temperatures i.e. a closed system of bosons described by canonical ensemble.
However we would like to employ grand canonical ensemble formalism in our study.
In grand canonical ensemble formalism we can vary T and µ independently by
choosing appropriate heat bath and particle bath. If T is changed at a fixed value
of µ the average number of particles in the system changes.
However if we want to keep N constant, we lose control over µ. When we
change the temperature, the chemical potential should change in such a way that
the average number of particles remains at the value chosen by us. In other words
in a closed system, µ is a function of temperature.
T
In what follows we shall study a closed system of ideal bosons employing the
grand canonical ensemble formalism in which the chemical potential depends on
AF
temperature; the dependence is such that the average number of bosons in the
system remains the same at all temperatures
DR

8.3 Summation to Integration


X Z
(·) → dǫ (·) g(ǫ)
k
Let us now convert the sum over quantum states to an integral over energy. To
this end we need an expression for the number of quantum states in infinitesimal
interval dǫ around ǫ. Let us denote this quantity by g(ǫ)dǫ. We call g(ǫ) the density
of (energy) states. Thus we have,
Z ∞
λ exp(−βǫ)
N = g(ǫ)dǫ (8.9)
0 1 − λ exp(−βǫ)
We need an expression for the density of states. We have done this exercise earlier.
In fact we have carried out classical counting and quantum counting and found
both lead to the same result. The density of states is given by,
 3/2
2m
g(ǫ) = V 2π ǫ1/2 . (8.10)
h2
We then have,
8.3 Sum to integration 129
 3/2 Z ∞
2m λ exp(−βǫ) 1/2
N = V 2π ǫ dǫ. (8.11)
h2 0 1 − λ exp(−βǫ)
We note that 0 ≤ λ < 1. This suggests that the integrand in the above can be
expanded in powers of λ. To this end we write
X∞
1
= λk exp(−kβǫ). (8.12)
1 − λ exp(−βǫ) k=0

This gives us

X ∞
X
λ exp(−βǫ)
= λk+1 exp[−β(k + 1)ǫ] = λk exp[−βkǫ]. (8.13)
1 − λ exp(−βǫ)
k=0 k=1

Substituting the above in the integral we get,


 3/2 X
∞ Z ∞
2m k
N = V 2π λ exp(−kβǫ)ǫ1/2 dǫ,
h2 0
k=1

 3/2 X
∞ Z ∞
2m exp(−kβǫ)(kβǫ)1/2 d(kβǫ)
= V 2π λk ,
h2 k=1 0 β 3/2 k3/2

 3/2 X
∞ Z ∞
2mkB T λk
= V 2π exp(−x)x1/2 dx,
h2 k3/2 0


2mkB T
3/2 X

k=1

λk
T
AF
= V 2π Γ (3/2),
h2 k3/2
k=1

 3/2 X∞
2mkB T 1 λk
= V 2π Γ (1/2) ,
DR

h2 2 k=1
k3/2

 3/2 ∞
2mkB T √ X λk
=V π π ,
h2 k3/2
k=1

 3/2 X

2πmkB T λk
=V . (8.14)
h2 k=1
k3/2

We have earlier defined a thermal wave length denoted by the symbol Λ. This is
the de Broglie wave length associated with a particle having thermal energy of the
order of kB T . It is also called quantum wavelength. It is given by, see earlier notes,
h
Λ= √ . (8.15)
2πmkB T
The sum over k, in the expression for N given by Eq. (8.14) is usually denoted by
the symbol g3/2 (λ):
X∞
λk λ2 λ3
g3/2 (λ) = 3/2
= λ+ √ + √ +··· (8.16)
k=1
k 2 2 3 3
130 8. Bose-Einstein Condensation

V
Thus we get N = g3/2 (λ) We can write it as,
Λ3
N Λ3
= ρΛ3 = g3/2 (λ).
V
(8.17)
It is easily verified, see below, that at high temperature we get results consistent
with Maxwell Boltzmann statistics :
The fugacity λ is small at high temperature. For small λ we can replace g3/2 (λ)
by λ. We get N = λV /Λ3 . This result is consistent with Maxwell-Boltzmann statis-
tics, as shown below.
For Maxwell-Boltzmann statistics, hnk i = λ exp(−βǫk ). Therefore,
X
N = hnk i
k
X
=λ exp(−βǫk )
k
 3/2 Z ∞
2m
= λ 2πV dǫ ǫ1/2 exp(−βǫ)
h2 0
 3/2 Z ∞
2m (βǫ)1/2 exp(−βǫ) d(βǫ)
= λ 2πV
h2 0 β 3/2
 3/2
2mkB T
= λ 2πV Γ (3/2)
h2 T
 3/2
2mkB T 1
= λ 2πV Γ (1/2)
AF
h2 2
 3/2
2mkB T √
= λ πV π
h2
DR

 3/2
2πmkB T
=λV
h2

V
= λ (8.18)
Λ3

8.4 Graphical Inversion and Fugacity


Let us consider a system of ideal bosons confined to a volume V at temperature T .
To ensure that the number of bosons is at a fixed value N , we must keep the system
at a value of fugacity such that the resulting grand canonical ensemble average of
N equals the chosen value of N . What is the value of λ that would ensure this ?
To answer this question we proceed as follows.
First we plot the so-called Bose function : g3/2 (λ) versus λ, see Fig. 8.1.
For given values of N , V T we can find the value of fugacity by graphical
inversion :
8.5 Treatment of the Singular Behaviour 131

• Draw a line parallel to the x axis at y = N Λ3 /V and


• read off the value of λ at which the line cuts the curve g3/2 (λ).
Once we get the fugacity, we can determine all other thermodynamic properties of
the open system employing the formalism of grand canonical ensemble.
So far so good.
But then we realise that the above graphical inversion scheme does not permit
evaluation of the fugacity of a system with N Λ3 /V greater than 2.612.
This is absurd.
There must be something wrong with what we have done.

8.5 Treatment of the Singular Behaviour


We realise that when N Λ3 /V approaches 2.612, the fugacity λ approaches unity;
the chemical potential µ approaches zero1 . We have already seen that at µ = 0 the
occupancy of the ground state diverges. The singular behaviour of the ground state
occupancy was completely lost when we replaced the sum over quantum states by
an integral over energy : the weighting function is the density of states, given by,
g(ǫ) ∼ ǫ1/2 . The density of states vanishes at zero energy. Hence we must take care
of the singular behaviour separately.
We have,
X λ exp(−βǫk ) λ X λ exp(−βǫk )
N = = + . (8.19)
1 − λ exp(−βǫk ) 1 − λ k≥1 1 − λ exp(−βǫk )
k≥0 T
In the above, we have separated the ground state occupancy and the occupancy of
AF
all the excited states. Let N0 denote the ground state occupancy. It is given by the
first term,
λ
N0 = (8.20)
1−λ
DR

The occupancy of all the excited states is given by the second term, where the sum
is taken only over the indices k representing the excited states. Let Ne denote the
occupancy of excited states. It is given by,
X λ exp(−βǫk
Ne = (8.21)
1 − λ exp(−βǫk )
k

In the above, the sum over k can be replaced by an integral over energy. In the
integral over energy, we can still keep the lower limit of integration as zero, since
the density of states giving weight factors for occupancy of states is zero at zero
energy. Accordingly we write
N = N0 + Ne (8.22)

λ V
= + 3 g3/2 (λ) (8.23)
1−λ Λ
We thus have,
1
The chemical potential approaches the energy of the ground state. With out loss
of generality, we can set the ground state at zero energy; i.e. ǫ0 = 0.
132 8. Bose-Einstein Condensation

N Λ3 Λ3 λ
= + g3/2 (λ) (8.24)
V V 1−λ
Let us define the number of density - number of particles per unit volume, denoted
by the symbol ρ. It is given by
N
ρ= (8.25)
V
The function λ/(1 − λ) diverges at λ = 1, as you can see from Figure (8.2).
Hence the relevant curve for carrying out graphical inversion should be the one
that depicts the sum of the singular part (that takes care of the occupancy of the
ground state) and the regular part (that takes care of the occupancy of the excited
states). For a value of Λ3 /V = .05 we have plotted both the curves and their sum
in Figure (8.3). Thus for any value of ρΛ3 we can now determine the fugacity by
graphical inversion.
We carry out such an exercise and obtain the values of λ for various values
of ρΛ3 and Fig. (8.4) depicts the results. It is clear from Fig. (8.4) that when
ρΛ3 > 2.612, the fugacity λ is close unity.

How close can the fugacity λ get to unity ?

Let us postulate2
a
λ=1− .
N
where a is a number. To determine a we proceed as follows.
We have,
λ
=
N
−1
T (8.26)
AF
1−λ a

N
≈ if N >> a (8.27)
a
DR

We start with,
Λ3 λ
ρΛ3 = + g3/2 (λ) (8.28)
V 1−λ
Sunstitute λ = 1 − a/N in the above and get3 ,
ρΛ3
ρΛ3 = + g3/2 (1) (8.29)
a
Thus we get,
ρΛ3
a= (8.30)
ρΛ3 − g3/2 (1)

2 a
We have reasons to postulate λ = 1− . This is related to the mechanism under-
N
lying Bose-Einstein condensation; we shall discuss the details later. In fact, fol-
lowing Donald A McQuarrie, Statistical Mechanics, Harper and Row (1976)p.173
a
we can make a postulate λ = 1 − . This should also lead to the same conclu-
V
sions.
3
g3/2 (1 − a/N ) ≈ g3/2 (1),
8.6 Bose-Einstein Condensation Temperature 133

Thus λ is less than unity and can be very close to unity; the value of 1 − λ can
be as small as the inverse of the total number of particles in the system. Precisely
1 − λ can be as small as a/N .
The point ρΛ3 = g3/2 (1) = 2.612 is a special point indeed. What is the physical
significance of this point ? To answer this question, consider the quantity ρΛ3
as a function of temperature with ρ kept at a constant value. The temperature
dependence of this quantity is shown below.
 3
h
ρΛ3 = ρ √ (8.31)
2πmkB T
At high temperature for which ρΛ3 < g3/2 (1) = 2.612, we can determine the
value of λ from the equation g3/2 (λ) = ρΛ3 by graphical or numerical inversion.
At low temperatures for which ρΛ3 > 2.612, we have λ = 1 − a/N where
ρΛ3
a= (8.32)
ρΛ3 − g3/2 (1)
The quantity λ/(1 − λ) is the number of particles in the ground state. At
temperatures for which ρΛ3 > 2.612, we have,
λ
N0 = (8.33)
1−λ

N
= (8.34)
a T
AF
N0 1
= (8.35)
N a

1
= 1− g3/2 (1) (8.36)
DR

ρΛ3
We can write the above in a more suggestive form by defining a temperature TBEC
by
ρΛ3BEC = g3/2 (1) (8.37)
Therefore,
 3  √ 3
N0 1 ρΛ3BEC ΛBEC T
= =1− = 1 − = 1 − √ (8.38)
N a ρΛ3 Λ TBEC
 3/2
T
= 1− fer T < TBEC (8.39)
TBEC
We have depicted the behaviour of the fractional number of particles in the ground
state as a function of temperature in Figure (8.5).

8.6 Bose-Einstein Condensation Temperature


Thus we can define the temperature at which Bose-Einstein condensation takes
place as,
134 8. Bose-Einstein Condensation
 3
N h
√ = 2.612 (8.40)
V 2πmkB T
 2/3
h2 N
kB TBEC = (8.41)
2πm 2.612 V
At T = TBEC , Bose- Einstein condensation sets in and the ground state occupancy
becomes anomalously larger and larger as temperature decreases further.

8.7 Grand Potential for bosons


The grand potential for bosons is given by
X
G(T, V, µ) = −kB T ln Q(T, V, µ) = kB T ln[1 − λ exp(−βǫk )] (8.42)
k

Now we shall be careful and separate the singular part and regular part to get,
X
G = kB T ln(1 − λ) + kB T ln[1 − λ exp(−βǫk )] (8.43)
k

In the above, convert the sum over k by an integral over dǫ by the prescription
below,
X  3/2 Z ∞
2m T
(·) −→ V 2π 2
(·) ǫ1/2 dǫ, (8.44)
k
h 0
AF
We get,
G = kB T ln(1 − λ)
DR

 3/2 Z ∞
2m
−kB T V 2π dǫ ln (1 − λ exp(−βǫ)) ǫ1/2 (8.45)
h2 0

We have

X
ln[1 − λ exp(−βǫ)] = − λk exp(−kβǫ) (8.46)
k=1

Then we have,
8.8 Energy of Bosonic System 135

G = kB T ln(1 − λ)

 3/2 X
∞ Z ∞
2m
−kB T V 2π λk dǫ ǫ1/2 exp(−kβǫ)
h2 0
k=1

= kB T ln(1 − λ) −

 3/2 X
∞ Z ∞
2m (kβǫ)1/2 exp(−kβǫ)
kB T V 2π λk d(kβǫ)
h2 0 k3/2 β 3/2
k=1

 3/2 X∞
2mkB T λk
= kB T ln(1 − λ) − kB T V 2π Γ (3/2)
h2 k3/2
k=1

 3/2 X

2πmkB T λk
= kB T ln(1 − λ) − kB T V
h2 k=1
k3/2

V
= kB T ln(1 − λ) − kB T g3/2 (λ) (8.47)
Λ3
Thus we have,
T
AF
V
G(T, V, λ) = kB T ln(1 − λ) − kB T g3/2 (λ) (8.48)
Λ3

8.8 Energy of Bosonic System


DR

In an earlier class, I have derived an expression for the average occupancy of a


single-particle-quantun-state; it is given by
λ exp(−βǫi )
hni i = (8.49)
1 − λ exp(−βǫi )
This immediately suggests that the (average) energy of the system is given by
X X ǫi λ exp(−βǫ )i)
U = hEi = hni i ǫi = (8.50)
i i
1 − λ exp(−βǫi )

We can also derive4 the above relation employing Grand canonical formalism.
4
An open system is described by a grand canonical partition function. It is for-
mally given by,
X
Q(β, V, µ) = exp[−β(Ei − µNi )] (8.51)
i

In the above Ei is the energy of the open system when in micro state i; Ni is
the number of particles in the open system when in micro state i. Let γ = βµ.
136 8. Bose-Einstein Condensation

Let us now go to continuum limit by converting the sum over micro states by
an integral over energy and get,
3 1
U = V kB T 3 g5/2 (λ) (8.56)
2 Λ

8.8.1 T > TBEC


Let us now investigate the energy of the system at T > TBEC . When temperature
is high, the number of bosons in the ground state is negligibly small. Hence the
total energy of the system is the same as the one given by Eq. (8.56).
We can write Eq. (8.56) in a more suggestive form, see below. We have,
ρΛ3 = g3/2 (λ)
N Λ3
= g3/2 (λ)
V
N Λ3
V = (8.57)
g3/2 (λ)
Substituting the above expression for V in Eq. (8.56) we get
3N kB T g5/2 (λ)
U = (8.58)
2 g3/2 (λ)

8.8.2 T ≤ TBEC T
For temperatures less that TBEC , the ground state gets populated anomalously. The
AF
bosons in the ground state do not contribute to the energy. For T ≤ TBEC , we have
µ = 0. This means λ = 15 . Substituting λ = 1 in Eq. ((8.56) we get,
3 1
U = V kB T 3 ζ(5/2) (8.59)
2 Λ
DR

Then we get,
X
Q(β, V, µ) = exp(−βEi ) exp(+γNi ) (8.52)
i

We differentiate Q with respect to the variable β, keeping γ constant. (Note


that T and µ are independent in an open system described by grand canonical
ensemble). We get
∂Q X
=− ǫi exp[−βǫi + γNi ] (8.53)
∂β i

1 ∂Q ∂ ln Q
− =− = hEi = U (8.54)
Q ∂β ∂β
For bosons, we have,
Y 1 X X ǫi λ exp(−βǫi )
Q= ln Q = − ln[1 − λ exp(−βǫi )]U = (8.55).
i
1 − λ exp(−βǫi ) i i
1 − λ exp(−βǫi )

5
we have postulated that λ = 1 − O(1/N ) for T ≤ TBEC .
8.9 Specific Heat Capacity of bosons 137

We also have,
N (ΛBEC )3
V = (8.60)
ζ(3/2)
Hence for T < TBEC , we have
3 (ΛBEC )3 ζ(5/2)
U = N kB T
2 Λ3 ζ(3/2)
 3/2
3 ζ(5/2) T
= N kB T (8.61)
2 ζ(3/2) TBEC
Thus we have,
 
 3 g5/2 (λ)

 N k B T for T > TBEC

 2 g3/2 (λ)


U = (8.62)



    3/2

 3 g5/2 (1) T
 N kB T for T < TBEC
2 g3/2 (1) TBEC

8.9 Specific Heat Capacity of bosons

8.9.1
CV
N kB
for T > TBEC
T
AF
Let us consider first the case with T > TBEC . We have
U 3 g5/2 (λ)
= T (8.63)
N kB 2 g3/2 (λ)
DR

 
1 ∂U CV ∂ 3T g5/2 (λ)
= = (8.64)
N kB ∂T N kB ∂T 2 g3/2 (λ)
To work out the derivative in the above, we need the following :
∂   3
First Relation : g3/2 (λ) = − g3/2 (λ).
∂T 2T
Proof :
We start with ρΛ3 = g3/2 (λ). Therefore,
138 8. Bose-Einstein Condensation
 
∂ ∂Λ ∂ h
[g3/2 (λ)] = 3 ρΛ2 = 3 ρΛ3 √ (8.65)
∂T ∂T ∂T 2πmkB T

3 h 1
=− ρΛ2 √
2 2πmkB T 3/2

3 h
=− ρΛ2 √
2T 2πmkB T

3 3
=− ρΛ3 = − g3/2 (λ) (8.66)
2T 2T
————–Q.E.D
∂ 1
Second Relation : [gn/2 (λ)] = g(n/2)−1 (λ).
∂λ λ
Proof : ∞
X λk
We have by definition, gn/2 (λ) = . Therefore,
k=1
kn/2
"∞ # ∞ ∞
∂ ∂ X λk X kλk−1 1X λk
[gn/2 (λ)] = n/2
= n/2
= (n/2)−1
∂λ ∂λ k=1 k k=1
k λ k=1 k
T =
1
λ
g(n/2)−1 (λ) (8.67)
AF
— Q.E.D

1 dλ 3 g3/2 (λ)
8.9.2 Third Relation : =−
DR

λ dT 2T g1/2 (λ)

Proof :
We proceed as follows :
 
∂ ∂ dλ 3 1 dλ
[g3/2 (λ)] = [g3/2 (λ)] : − g3/2 (λ) = g1/2 (λ) (8.68)
∂T ∂λ dT 2T λ dT
1 dλ 3 g3/2 (λ)
From the above we get, =− –Q.E.D
λ dT 2T g1/2 (λ)
We have,
8.9 Heat capacity below TC 139
 
CV ∂ 3T g5/2 (λ)
=
N kB ∂T 2 g3/2 (λ)
 
3 g5/2 (λ) 3T ∂ g5/2 (λ)
= +
2 g3/2 (λ) 2 ∂T g3/2 (λ)
" #
3 g5/2 (λ) 3T g5/2 (λ) ∂g3/2 (λ) 1 ∂g5/2 (λ) dλ
= − 2

2 g3/2 (λ) 2 g3/2 (λ) ∂T g3/2 (λ) ∂λ dT
"     #
3 g5/2 (λ) 3T g5/2 (λ) 3 1 1 dλ
= − 2
− g3/2 (λ) − g3/2 (λ)
2 g3/2 (λ) 2 g3/2 (λ) 2T g3/2 (λ) λ dT
"   #
3 g5/2 (λ) 3T g5/2 (λ) 3 1 dλ
= − 2
− g3/2 (λ) −
2 g3/2 (λ) 2 g3/2 (λ) 2T λ dT
"   #
3 g5/2 (λ) 3T g5/2 (λ) 3 3 g3/2 (λ)
= − 2
− g3/2 (λ) +
2 g3/2 (λ) 2 g3/2 (λ) 2T 2T g1/2 (λ)
3 g5/2 (λ) 9 g5/2 (λ) 9 g3/2 (λ)
= + −
2 g3/2 (λ) 4 g3/2 (λ) 4 g1/2 (λ)
15 g5/2 (λ) 9 g3/2 (λ)
= − (8.69)
4 g3/2 (λ) 4 g1/2 (λ)

CV
8.9.3 for T < TBEC T
N kB
AF
Now, let us consider the case with T < TBEC . We have,
 3/2
U 3 g5/2 (1) T
= T (8.70)
N kB 2 g3/2 (1) TBEC
DR

 3/2
1 3 g5/2 (1) 5 T
CV = (8.71)
N kB 2 g3/2 (1) 2 TBEC
Thus we have,

 15 g5/2 (λ) 9 g3/2 (λ)

 − for T > TBEC

 4 g3/2 (λ) 4 g1/2 (λ)


1
CV = (8.72)
N kB 
  3/2



 15 g5/2 (1) T
 for T < TBEC
4 g3/2 (1) TBEC
The specific heat is plotted against temperature in Figure (8.6). The cusp in the
heat capacity at T = TBEC is the signature of Bose-Einstein condensation. Asymp-
totically T → ∞, the heat capacity tends to the classical value consistent with
equi-partition.
140 8. Bose-Einstein Condensation

8.10 Mechanism of Bose-Einstein Condensation


Let the ground state be of energy ǫ0 ≥ 0. For example consider particle in a three
dimensional box of length L. The ground state is

(nx , ny , nx ) = (1, 1, 1).

The ground state energy is

3h2
ǫ1,1,1 = ǫ0 = .
8mL2
The chemical potential is always less than or equal to ǫ0 . As temperature de-
creases, the chemical potential increases and comes closer and closer to the ground
state energy ǫ0 ≥ 0. Let us estimate how close µ can get to ǫ0 . In other words, we
want to estimate the smallest possible value of (ǫ0 − µ)/[kB T ]. To this end, consider
the expression for the average number of bosons in the ground state. Let us denote
this by N0 . It is given by,
1
N0 =   (8.73)
ǫ0 − µ
exp −1
kB T
As temperature goes to zero, the chemical potential goes toward the ground
ǫ0 − µ
state energy. For a non-zero value of T , when is small, we can write
kB T

exp

ǫ0 − µ
kB T

=1+
ǫ0 − µ
kB T
T (8.74)
AF
Substituting this in the expression for N0 , given above, we get,
kB T
N0 = (8.75)
ǫ0 − µ(T )
DR

N0 goes to zero6 as T → ∞. At high temperature, the ground state occupancy is


extremely small, as indeed it should.
Therefore we have,
ǫ0 − µ 1
= (8.76)
kB T N0
The largest value that N0 can take is N , i.e. when all the particles condense
into the ground state. In other words, the smallest value that 1/N0 can take is 1/N .
Hence (ǫ0 − µ)/[kB T ] can not be smaller than 1/N . The smallest possible value it
can take is 1/N - inverse of the number of particles in the entire system.
ǫ0 − µ 1
≥ (8.77)
kB T N
Thus, measured in units of kB T , the chemical potential shall always be less that
the ground state energy at any non-zero temperature. At best, the quantity (ǫ − µ),
expressed in units of thermal energy (kB T ), can only be of the order of 1/N .
But, remember. For a physic.ist, small is zero and large is infinity
6
For large T , the numerator is large; but the denominator is also large. Note that
µ(T ) is negative and large for large T . In fact the denominator goes to infinity
faster than the numerator.
8.10 Mechanism of Bose-Einstein Condensation 141

Therefore the chemical potential can never take a value close to any of the
excited states, since all of them invariably lie above the ground state. In a sense,
the ground state forbids the chemical potential to come close to any energy level
other than the ground state energy. It sort of guards all the excited states from a
close visit of µ. As T → 0, the number of bosons in the ground state increases.
This precisely is the subtle mechanism underlying Bose-Einstein con-
densation.

T
AF
DR
5

4.5

3.5

3 g3/2 (λ = 1) = ζ(3/2) = 2.612


2.5

1.5

0.5

0
0 0.2 0.4 0.6 0.8 1
λ
Fig. 8.1. g3/2 (λ) versus λ. Graphical inversion to determine fugacity

T
AF
DR
100

90

80

70

60

50

40
λ
30 →
1 −λ
20

10
λ
0
0 0.2 0.4 0.6 0.8 1

Fig. 8.2. Singular part of N T


8
AF
7
DR

3 g3/2 (λ = 1) = ζ(3/2) = 2.612


2
Λ3 λ
+ g3/2 (λ) →
1 V 1 −λ
0
0 0.2 0.4 0.6 0.8 1
λ
Fig. 8.3. ρΛ3 versus λ. The singular part [Λ3 /V ][λ/(1−λ)] (the bottom most curve),
the regular part g3/2 (λ) (the middle curve) , and the total (ρΛ3 ) are plotted. For
this plot we have taken Λ3 /V as 0.05
1

0.9

0.8

0.7

0.6
λ

0.5

0.4

3 Λ3 λ
0.3 ρΛ = + g3/2 (λ)
3
V 1 −λ
0.2 Λ
= .05
0.1 V
2.612
0
0 0.5 1 1.5 2 2.5 3 3.5 4

ρΛ3
Fig. 8.4. Fugacity λ versus ρΛ3
T
AF
1.5

3 43/2
N0 T
=1−
DR

1
N TBEC
N0
N

0.5

0
0 0.5 1 1.5
T
TBEC
Fig. 8.5. Ground state occupation as a function of temperature
2

1.8

1.6
Classical : 3NkB /2
1.4

1.2
Nk B
CV

0.8

0.6

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
T
TBEC
Fig. 8.6. Heat capacity in the neighbourhood of Bose - Einstein condensation
temperature

T
AF
DR
DR
AF
T
9. Elements of Phase Transition

I shall provide an elementary introduction to phase transition. The topics covered


include phase diagram of a normal substance; coexiatance curves : (i) sublimation
curve, (ii) melting curve, and (iii) vapour pressure curve; triple point; critical point;
first order phase transition; latent heat; second order phase transition; critical phe-
nomena; derivation of Gibbs - Duhem relation starting from Gibbs free energy;
Clausius-Clapeyron equation; Anomalous expansion of water upon freezing; and an
exotic behaviour of Helium-3 at low temperatures.

T
AF
DR
148 9. Elements of Phase Transition

T
AF
DR
9. Elements of Phase Transition 149

T
AF
DR
150 9. Elements of Phase Transition

T
AF
DR
9. Elements of Phase Transition 151

T
AF
DR
152 9. Elements of Phase Transition

T
AF
DR
9. Elements of Phase Transition 153

T
AF
DR
154 9. Elements of Phase Transition

T
AF
DR
9. Elements of Phase Transition 155

T
AF
DR
DR
AF
T
10. Statistical Mechanics of Harmonic
Oscillators

10.1 Classical Harmonic Oscillators


Consider a closed system of 3N harmonic oscillators at temperature T . The os-
cillators do not interact with each other and are distinguishable. Let us derive an
expression for the single-oscillator partition function.
The energy of an harmonic oscillator is given by
p2 1
E= + mω 2 q 2 (10.1)
2m 2
where q and p are the position and momentum respectively, of the one dimensional
harmonic oscillator. ω is the characteristic frequency of the oscillator and m its
mass. A simple pendulum executing small oscillations is a neat example of an
harmonic oscillator.
We have,
Z
1 +∞
Z +∞   2
p 1
T 
AF
2 2
Q1 (T ) = dq dp exp −β + mω q (10.2)
h −∞ −∞ 2m 2
We can write the above in a convenient way as a product of two Gaussian integrals,
one over dq and the other over dp, as
 
DR

Z  
+∞  
1  1 q2 
Q1 (T ) = dq exp − r !2  ×
h −∞  2 kB T 
 
mω 2

Z " #
+∞
1 p2
dp exp − √ 3 (10.3)
−∞ 2 mkB T
Let σ1 and σ2 denote the standard deviations of of the two zero-mean Gaussian
distributions. These are given by,
r
kB T
σ1 = (10.4)
mω 2

σ2 = mkB T (10.5)

kB T
σ1 σ2 = (10.6)
ω
We have normalization identity for a Gaussian
158 10. Statistical Mechanics of Harmonic Oscillators
Z  
+∞
1 x2 √
dx exp − 2
= σ 2π (10.7)
−∞ 2 σ
Therefore,
1 √ √ kB T
Q1 (T ) = (σ1 2π)(σ2 2π) = (10.8)
h ~ω
If all the oscillators are identical i.e. they all have the same characteristic frequency
of oscillations, then
 3N
kB T
Q3N (T ) = (10.9)

On the other hand if the oscillators have distinct characteristic frequency {ωi :
i = 1, 2, · · · , 3N }, then
3N
Y kB T
Q3N (T ) = (10.10)
i=1
~ωi

where we have considered 3N harmonic oscillators with 3N characteristic frequen-


cies.

10.1.1 Helmholtz Free Energy


The free energy of a system of 3N non-interacting, identical classical harmonic
oscillators is given by

kB T

T 


AF
F (T, V, N ) = −3N kB T ln = 3N kB T ln (10.11)
~ω kB T
If the oscillators have different frequencies then
X3N  
kB T
DR

F (T, V, N ) = −kB T ln (10.12)


i=1
~ωi

If N is large we can define g(ω)dω as the number of harmonic oscillators with


frequencies in an interval dω around ω. The sum can be replaced by an integral,
Z ∞  
kB T
F (T ) = −kB T ln g(ω)dω (10.13)
0 ~ω
We have the normalization
Z ∞
g(ω)dω = 3N (10.14)
0

Once we know of free energy, we can employ the machinery of thermodynamics


and get expressions for all other thermodynamic properties of the system, see below.
10.1 Classical Harmonic Oscillators 159

10.1.2 Thermodynamic Properties of the Oscillator System

F (T, V, N ) = U − T S (10.15)

dF = dU − T dS − SdT

= −SdT − P dV + µdN (10.16)


Thus for a system of identical, non-interacting classical harmonic oscillators
 
∂F
P =− =0 why? (10.17)
∂V T,N

   
∂F ~ω
µ= = kB T ln (10.18)
∂N T,V
kB T

     
∂F kB T
S =− = N kB ln +1 (10.19)
∂T V,N ~ω
We also have,
 
∂ ln Q
U =−
∂β V,N
= 3N kB T, T (10.20)
AF
consistent with equipartition theorem which says each quadratic term in the Hamil-
tonian carries kB T /2 of energy. The Hamiltonian of a single harmonic oscillator has
two quadratic terms - one in position q and the other in momentum p.
We also find that the results are consistent with the Dulong and Petit’s law
which says that the heat capacity at constant volume is independent of temperature:
DR

 
∂U
CV = = 3N kB = 3nR (10.21)
∂T V

CV
= 3R ≈ 6 calories (mole)−1 (kelvin)−1 (10.22)
n
More importantly, the heat capacity is the same for all the materials; it depends
only on the number of molecules or the number of moles of the substance and not
on what the substance is. The heat capacity per mole is approximately 6 calories
per Kelvin.

10.1.3 Quantum Harmonic Oscillator


Now let us consider quantum harmonic oscillators. The energy eigenvalues of a
single one dimensional harmonic oscillator is given by
 
1
ǫn = n + ~ω : n = 0, 1, 2, · · · (10.23)
2
The canonical partition function for a single (quantum) harmonic oscillator is then,
160 10. Statistical Mechanics of Harmonic Oscillators

X
Q1 (β) = exp(−β~ω/2) [exp(−β~ω)]n
n=0

exp(−β~ω/2)
= (10.24)
1 − exp(−β~ω)
The partition function of a collection of 3N non-interacting quantum harmonic
oscillators is then given by
exp(−3N β~ω/2)
QN (T ) = (10.25)
[1 − exp(−β~ω)]3N
If the harmonic oscillators are all of different frequencies, the partition function is
given by
3N
Y exp(−β~ωi /2)
Q(T ) = (10.26)
i=1
1 − exp(−β~ωi )
The free energy is given by,
F (T, V, N ) = −kB T ln Q3N (T )

 
1
= 3N ~ω + kB T ln {1 − exp(−β~ω)} (10.27)
2
For 3N independent harmonic oscillators with different frequencies we have

F =
3N 
X ~ωi
+ kB T ln {1 − exp(−β~ωi )}
 T (10.28)
AF
i=1
2

Z ∞  

= dω + kB T ln {1 − exp(−β~ω)} g(ω) (10.29)
2
DR

Z ∞
dω g(ω) = 3N (10.30)
0
We can obtain the thermodynamic properties of the system from the free energy.
We get,
 
∂F 1
µ= = ~ω + kB T ln [1 − exp(−β~ω)] (10.31)
∂N T,V 2

 
∂F
P =− =0 (10.32)
∂V T,N

   
∂F β~ω
S=− = 3N kB − ln {1 − exp(−β~ω)} (10.33)
∂T V,N exp(β~ω) − 1

 
∂ ln Q ~ω ~ω
U =− = 3N + (10.34)
∂β 2 exp(β~ω) − 1
10.1 Classical Harmonic Oscillators 161

The expression for U tells that the equipartition theorem is the first victim of quan-
tum mechanics : Quantum harmonic oscillators do not obey equipartition theorem.
The average energy per oscillator is higher than the classical value of kB T . Only for
T → ∞, we have kB T >> ~ω, the ”quantum” results coincide with the ”classical”
results. Let x = k~ω
BT
. Then Eq. (10.34) can be written as
U x x
= + (10.35)
3N kB T 2 exp(x) − 1
We have plotted U measured in units of 3N kB T as a function of oscillator energy
~ω measured in units of kB T . It is seen that when x → 0 (T → ∞), this quantity
tends to unity, see also below, implying that classical result of equipartition of
energy obtains at high temperature. Quantum effects start showing up only at low
temperatures.
In the limit x → 0 (T → ∞), we have,
U ∼ x x
x→0 + x2 x3
(10.36)
3N kB T 2 x + 2 + 6 + O(x4 )
x 1
= + 2 (10.37)
2 1 + x2 + x6 + O(x3 )
x x x2 x2
= +1− − + + O(x3 ) (10.38)
2 2 6 4
x2 x2
= 1− + + O(x3 ) (10.39)
6 4
= 1+
x2
12
+ O(x3 )
T (10.40)
AF
= 1 (10.41)
The heat capacity at constant volume is given by
 
∂U
CV = (10.42)
DR

∂T V,N

 2
3N ~ω exp[β~ω]
= (10.43)
kB T (exp[β~ω] − 1)2
The second victim of quantum mechanics is the law of Dulong and Petit. The heat
capacity depends on temperature and on the oscillator frequency. The heat capacity
per mole will change from substance to substance because of its dependence on the
oscillator frequency. Only in the limit of T → ∞ (the same as β → 0), do we get
the classical results.
The temperature dependence of heat capacity is an important ”quantum” out-
come. We find that the heat capacity goes to zero exponentially as T → 0. However
experiments suggest that the fall is algebraic and not exponential. The heat capacity
is found to go to zero as T 3 . This is called T 3 law.

10.1.4 Specific Heat of a Crystalline Solid


In the above we studied the behaviour of a collection of independent identical
harmonic oscillators in a canonical ensemble. We shall see below how such a study
162 10. Statistical Mechanics of Harmonic Oscillators

is helpful toward understanding of the behaviour of specific heat of crystalline solid


as a function of temperature.
A crystal is a collection of say N atoms organized in a regular lattice. Let
{x1 , x2 , · · · , x3N } specify the 3N positions of these atoms. For example we can
consider a periodic array of atoms arranged at regular intervals along the three
mutually perpendicular directions, constituting a three dimensional cubic structure.
We can think of other structures like face-centred cubic (FCC), body centred cubic
(BCC) lattices.
An atom vibrates around its lattice location say (x̄i , x̄i+1 , x̄i+2 ). It does not
make large excursions away from its lattice location. We must bear in mind that
the atoms are not independently bound to their lattice position. They are mutually
bound1 .
Consider a total of N atoms organized in a three dimensional lattice. Each atom
executes small oscillations about its mean position. In the process of oscillations
each atom pulls and pushes its neighbours; these neighbours in turn pull and push
their neighbours and so on. The disturbance propagates in the crystal. We can set
up equations of motion for the three coordinates of each of the atoms. We shall
have 3N coupled equations. Consider the Hamiltonian of a solid of N atoms with
position coordinates are {x1 , x2 , · · · x3N }. When the system of atoms is in its lowest
energy, the coordinates are x1,0 , x2,0 , · · · , x3N,0 . Let V (x1 , x2 , · · · x3N ) denote the
potential energy. We express the potential energy under harmonic approximation,
as
V (x1 , x2 , · · · x3N ) = V (x1,0 , x2,0 , · · · , x3N,0 ) +

3N 
X ∂V


∂xi
(xi − xi,0 ) +
T
AF
i=1 x1,0 ,x2,0 ,··· ,x3N,0

3N X
X 3N  2 
1 ∂ V
(xi − xi,0 )(xj − xj,0 ) (10.44)
2 ∂xi ∂xj x1,0 ,x2,0 ,···x3N,0
DR

i=1 j=1

The first term gives the minimum energy of the solid when all its atoms are in their
equilibrium positions. We can denote this energy by V0 .
The second set of terms involving the first order partial derivatives of the poten-
tial are all identically zero by definition : V has a minimum at {xi = xi,0 ∀i = 1, 3N }
1
To appreciate the above statement, consider a class room wherein the chairs
are already arranged with constant spacing along the length and breadth of the
class room. The students occupy these chairs and form a regular structure. This
corresponds to a situation wherein each student is bound independently to his
chair.
Now consider a situation wherein the students are mutually bound to each
other. Let us say that the students interact with each other in the following
way : Each is required to keep an arm’s length from his four neighbours. If the
distance between two neighbouring students is less, they are pushed outward;
if more, they are pulled inward. Such mutual interactions lead to the student
organizing themselves in a two dimensional regular array
I shall leave it to you to visualize how such mutual nearest neighbour inter-
actions can give rise to three dimensional arrays.
10.1 Classical Harmonic Oscillators 163

The third set of terms involving second order partial derivatives describe har-
monic vibrations. We neglect the terms involving higher order derivatives and this
is justified if only small oscillations are present in the crystalline .
Thus under harmonic approximations we can write the Hamiltonian as,
X3N  2 X 3N X3N
1 dξi
H = V0 + + αi,j ξi ξj (10.45)
i=1
2 dt i=1 j=1

where
ξ = xi − x̄i (10.46)

 
1 ∂2V
αi,j = (10.47)
2 ∂xi ∂xj x̄1 ,x̄2 ,··· ,x̄3N

We shall now introduce a linear transformation from the coordinates {ξi : i =


1, 3N } to the normal coordinates {qi : i = 1, 3N }. We choose the linear transfor-
mation matrix such that the Hamiltonian does not contain any cross terms in the
q coordinates.
3N
X 1 
H = V0 + m q̇ 2 + ωi2 qi2 (10.48)
i=1
2

where {ωi : i = 1, 3N } are the characteristic frequencies of the normal modes


of the system. These frequencies are determined by the nature of the potential
T
energy function V (x1 , x2 , · · · x3N ). Thus the energy of the solid can be considered
as arising out of a set of 3N one dimensional, non interacting, harmonic oscillators
AF
whose characteristic frequencies are determined by the nature of the atoms of the
crystalline solid, the nature of their mutual interaction, the nature of the lattice
structure etc..
Thus we can describe the system in terms of independent harmonic oscillators
by defining a normal coordinate system, in which the equations of motion are
DR

decoupled. If there are N atoms in the crystals there are 3N degrees of freedom.
Three of the degrees of freedom are associated with the translation of the whole
crystal; and three with rotation. Thus, there are strictly 3N − 6 normal mode
oscillations. If N is of the order of 1025 or so, it doesn’t matter if the number of
normal modes is 3N − 6 and not 3N .
We can write the canonical partition function as
3N
Y exp(−β~ωi /2)
Q= (10.49)
i=1
1 − exp(−β~ωi )

There are 3N normal frequencies. We can imagine them to be continuously dis-


tributed. Let g(ω)dω denote the number of normal frequencies between ω and
ω + dω. The function g(ω) obeys the normalization
Z ∞
g(ω)dω = 3N (10.50)
0

We have,
164 10. Statistical Mechanics of Harmonic Oscillators
3N 
X 
β~ωi
− ln Q = + ln {1 − exp(−β~ωi )} (10.51)
i=1
2

Z ∞  
β~ω
= + ln {1 − exp(−β~ω)} g(ω)dω (10.52)
0 2
The problem reduces to finding the function g(ω). Once we know g(ω), we
can calculate the thermodynamic properties of the crystal. In particular we can
calculate the internal energy U and heat capacity, see below.
Z ∞ 
~ω ~ω exp(−β~ω)
U = + g(ω)dω
0 2 1 − exp(−β~ω)

Z ∞  
~ω ~ω
= + g(ω)dω (10.53)
0 2 exp(β~ω) − 1

Z ∞
(β~ω)2 exp(β~ω)
C V = kB g(ω)dω (10.54)
0 [exp(β~ω) − 1]2
T
The problem of determining the function g(ω) is a non-trivial task. It is precisely
here that the difficulties lie. However, there are two well known approximations to
AF
g(ω). One of them is due to Einstein and the other due to Debye.

10.1.5 Einstein Theory of Specific Heat of Crystals


DR

Einstein assumed all the 3N harmonic oscillators to have the same frequency. In
other words,
g(ω) = 3N δ(ω − ωE ) (10.55)
where ωE is the Einstein frequency or the frequency of the Einstein oscillator. The
Einstein formula for the heat capacity is then given by
 2
~ω exp(~ωE /[kB T ])
CV = 3N kB (10.56)
kB T (exp(~ωE /[kB T ]) − 1)2
Let us define
~ωE
ΘE = (10.57)
kB
and call ΘE as Einstein temperature. Verify that this quantity has the unit of
temperature. In terms of Einstein temperature we have,
 2
ΘE exp(ΘE /T )
CV = 3N kB (10.58)
T [exp(ΘE /T ) − 1]2
10.1 Classical Harmonic Oscillators 165

• Show that in the limit of T → ∞, the heat capacity


of the Einstein solid tends to the value 3N kB = 3R =
6 cal (mole)−1 K −1 predicted by Dulong and Petit.

• Show that in the low temperature limit,


 2
∼ ΘE
CV T →0 3N kB exp(−ΘE /T ) (10.59)
T

Experiments suggest T 3 decay of CV with temperature. In the next class I shall


discuss Debye’s theory of heat capacity. We will find that Debye’s theory gives the
T 3 law.

10.1.6 Debye Theory of Specific Heat


Debye assumed a continuous spectrum of frequencies, cut off at an upper limit
ωD . Let us call it Debye frequency. Debye assumed based on an earlier work of
Rayleigh, that g(ω) = αω 2 , where the proportionality constant depends on the
speed of propagation of the normal mode, its nature2 , its degeneracy3 . From the
normalization condition,
Z ωD
α
0
ω 2 dω = 3N T (10.60)
AF
3
we get, α = 9N/ωD . Thus we have
 9N


 3
ω 2 for ω ≤ ωD
 ωD
g(ω) = (10.61)
DR





0 for ω > ωD
Let us now calculate CV under Debye’s theory. We start with
Z ∞
(β~ω)2 exp(β~ω)
CV (T ) = kB g(ω)dω (10.62)
0 [exp(β~ω) − 1]2
Let
x = β~ω (10.63)

~ωD
ΘD = (10.64)
kB
ΘD is called the Debye temperature. Then we have,
 3 Z ΘD /T
T x4 exp(x)
CV = (3N kB ) × 3 dx (10.65)
ΘD 0 [exp(x) − 1]2
2
transverse or longitudinal
3
transverse mode is doubly degenerate and longitudinal mode is non-degenerate,
etc..
166 10. Statistical Mechanics of Harmonic Oscillators

Let us consider the integral in the above expression and write,


Z Θ/T
x4 exp(x)
I = dx (10.66)
0 (exp(x) − 1)2
Integrating by parts4 we get,
 4 Z ΘD /T
1 ΘD x3
I = +4 dx (10.67)
exp(ΘD /T ) − 1 T 0 exp(x) − 1
The expression for heat capacity is then,
  
ΘD 1
CV = (3N kB ) −3
T exp(ΘD /T ) − 1
 3 Z ΘD /T #
T x3
+ 12 dx (10.68)
ΘD 0 exp(x) − 1

Let us now consider the behaviour CV in the limit of T → ∞. we have T >> ΘD .


We can set exp(ΘD /T ) ≈ 1+(ΘD /T ); also in the integral we can set exp(x) = 1+x.
Then we get,
"  3 Z ΘD /T #
T
CV = 3N kB −3 + 12 x2 dx (10.69)
ΘD 0

= 3N kB (−3 + 4) T (10.70)
AF
= 3N kB (10.71)
In the low temperature limit we have T << ΘD . We start with,
  
ΘD 1
CV = 3N kB −3
T exp(ΘD /T ) − 1
DR

 3 Z ΘD /T #
T x3
+ 12 dx (10.72)
ΘD 0 exp(x) − 1

In the limit T → 0, the first term inside the square bracket goes to zero like
exp(−ΘD /T ). The upper limit of the integral in the second term inside the square
bracket can be set to ∞. From standard integral tables5 , we have,
Z ∞
x3 π4
dx = (10.73)
0 exp(x) − 1 15
Thus we have in the low temperature limit,
 3
∼ 12π 4 T
CV T →0 N kB (10.74)
5 ΘD
4
Take u(x) = x4 and dv(x) = exp(x)dx/[exp(x) − 1]2
5
The integral equals Γ (4)ζ(4), where Γ (·) is the gamma function and ζ(·) is the
Riemann zeta function. Γ (4) = 3! = 6 and ζ(4) = π 4 /90. See e.g. G B Arfken
and H J Weber, Mathematical Methods for Physicists, Fourth Edition, Academic
Press, INC, Prism Books PVT LTD (1995).
10.1 Classical Harmonic Oscillators 167

10.1.7 Riemann Zeta Function


In an earlier class, we came across an integral,
Z ∞
x3
dx (10.75)
0 exp(x) − 1
The value of the integral is π 4 /15.
This is a particular case of a more general result based on Riemann zeta func-
tion,
Z ∞
xp
dx = Γ (p + 1)ζ(p + 1) (10.76)
0 exp(x) − 1
where Γ (·) is the usual gamma function defined as
Z ∞
Γ (z) = xz−1 exp(−x)dx for Real(z) > 0, (10.77)
0

and ζ(·) is the Riemann zeta function, see below. Note ζ(2) = π 2 /6 and ζ(4) =
π 4 /90, etc..
Riemann zeta function is defined as

X
ζ(p) = n−p (10.78)
n=1

Take f (x) = x−p and then


 −p+1 ∞





x
−p + 1

for p 6= 1

T
AF
Z ∞ 
 1

x−p dx = (10.79)
1 
 ∞




 ln x for p = 1
DR

The integral and hence the series is divergent for p ≤ 1 and convergent for p > 1.

10.1.8 Bernoulli Numbers


Bernoulli numbers Bn are defined by the series
X∞
x xn
= Bn (10.80)
exp(x) − 1 n=0
n!

which converges for |x| < 2π.


By differentiating the power series repeatedly and then setting x = 0, we obtain
 n  
d x
Bn = . (10.81)
dxn exp(x) − 1 x=0

• Show that B0 = 1; B1 = −1/2; B2 = 1/6; B4 = −1/30;


B6 = 1/42; B2n+1 = 0 ∀ n ≥ 1.
168 10. Statistical Mechanics of Harmonic Oscillators

Euler showed that,



2(2n)! X −2n
B2n = (−1)n−1 p , n = 1, 2, 3, · · · (10.82)
(2π)2n p=1

2(2n)!
= (−1)n−1 ζ(2n), n = 1, 2, 3, · · · (10.83)
(2π)2n

• Employing the above relation between Bernoulli numbers and


Riemann zeta function, show that
π2 π4
ζ(2) = ζ(4) = (10.84)
6 90

π6 π8
ζ(6) = ζ(8) = (10.85)
945 9450

T
AF
DR
Index

adiabatic process 82 delta function 46, 48


Avijit Lahiri 10 density of states 52, 66, 69, 90, 105,
128
Balescu, R 9, 64 determinism 6
Bernouilli numbers 167 Dittman, R. H. 12
Bernoulli, Daniel 41 Dugdale, J. S. 12
Binomial distribution 27, 31, 32, 122 Dulong and Petit’s law 161
Boltzmann entropy 44, 88
Boltzmann, Ludwig Eduard 3, 4, 13, Einstein temperature 164
14, 18, 60, 68, 88, 109, 113 Einstein’s theory of specific heat 164
Boltzmann-Gibbs-Shannon entropy 4, Einstein, Albert 3, 97
80 ensemble 18, 19, 22
Bose-Einstein condensation 118–120, entropy 90
132, 133, 139–141, 145 Entropy : Statistical 4
T
Bose-Einstein statistics 111, 113, 114, Entropy : Thermodynamics 4
116, 117, 119, 122, 125 equi-partition theorem 84, 139
AF
Bosons 109, 116, 118, 119 ergodicity 16, 87, 89
bosons 90, 127, 128, 134, 136, 141 Euler theorem 92
event 16
Callen, H B 9, 11
DR

canonical ensemble 44, 63, 66, 71, 79, factorial moment 28, 123
80, 89, 90, 97, 128 Feller, Willam 33
canonical partition function 64–66, Fermi, Enrico 12
68–70, 84, 89–91, 105, 163 Fermi-Dirac statistics 108, 111, 113,
Carlo Rovelli 5 114, 116, 117, 119, 122
Carnot, Sadi 3 Fermions 109, 116, 118
Central Limit Theorem 38 fermions 90
Chandler, D 9 Finn, C. B. P. 12
characteristic function 34, 37, 38 first law of thermodynamics 88, 115
chemical potential 56, 71, 90, 97, 105, Fourier transform 34, 38
111–115, 119, 120, 127, 128, 131, 140, fugacity 89, 90, 105, 128, 130–132
141
Chowdhury, Debashish 9 Gabriel, Weinreich 12
Clausius 3, 79 Gaussian distribution 33, 34, 38
closed system 63–66, 71, 72, 77, 79, 80, geometric distribution 122, 123
157 Gibbs ensemble 27
cumulant 35, 38 Gibbs ensemble 18, 22, 24
cumulant generating function 35 Gibbs, J. W. 3, 10, 22
Gibbs-Duhem relation 92, 96, 97
Debye frequency 165 Glazer, M. 11
Debye temperature 165 Goodstein, David 9
Debye’s theory of specific heat 165 grand canonica partition function 114
170 Index

grand canonical ensemble 71, 80, 87, moment generating function 29, 30,
90, 128, 131 38, 123
grand canonical partition function 89, moments 35
90, 92, 99, 105, 108, 109, 115, 116, 127,
135 Newton 3
grand potential 90–92, 111, 112, 127
open system 8, 71, 80, 87, 88, 91, 94,
harmonic oscillators 157–159, 163 95, 111, 114, 127, 131, 135
Helmholtz 3
Helmholtz free energy 66, 68, 90, 91, Palash, B. Pal 9
158 Pathria, R. K. 9, 97
Huang, Kerson 10 Pauli’s exclusion principle 108
Planck, Max 12
ideal gas 83 Poisson distribution 32, 33, 38, 94, 125
ideal gas law 41, 43 Prigogine, Ilya 12
isolated system 64, 69–71, 78, 80 probability density function 17, 34
isothermal compressibility 95, 97, 112,
127 quantum statistics 105

random variable 17, 29, 122, 123


Kelvin 3 Rief, F 9
Kittel, C. 11 Riemann zeta function 166, 167
Kondepudi, Dilip 12
Krömer, K. 11 sample space 15, 19
Schröder, Daniel V 11
Lagrange method 73, 77 T
Schrodinger 3
Lee, Joon Chang 10 Second law of thermodynamics 3, 7, 8
Legendre transform 79, 90, 111, 127 Sethna, James P. 11
AF
Shantini, R. 12
Maxwell’s demon 8 Stauffer, Dietrich 9
Maxwell’s ensemble 27 Stirling Formula 23
Maxwell, James Clerk 3, 18
DR

Maxwell-Boltzmann statistics 33, 109, thermal wave length 129


111, 112, 114, 116, 117, 124, 125, 130 theta function 46, 48
McQuairrie, Donald A 9, 77, 90, 92, time-reversal invariance 6
106
method of most probable distribution Van Ness, H. C. 9, 11
70
micro canonical ensemble 45, 66, 69, Wark, J 11
80, 90
microcanonical ensemble 89 Zamansky, M. W. 12

You might also like