Statistical Physics EPS

Basic Definitions
Microstate: A complete description of the system in terms of microscopic variables.

Macrostate: An incomplete description of a system in terms of macroscopic variables.

Statistical weight: The statistical weight of a macrostate is the number of microstates in

the macrostate. Denoted Ω.

Ensemble: An ensemble is a probability distribution for the microstates of a system.

Microcanonical Ensemble
Fundamental postulate: In an isolated system, all accessible microstates are equally likely.

Microcanonical ensemble: The probability for an accessible microstate of an isolated system

at equilibrium is:
1
pj = .
Ω
Boltzmann entropy: Applies to systems obeying the microcanonical ensemble:

SB = kB log Ω.

Configuration: The list of all the occupancies of the single particle states [n0 , n1 , . . . , nj , . . .]
specifies a configuration. The probability for a configuration is
Ωn X
pn = , Ω= Ωn ,
Ω n

where the sum is over all configurations. The statistical weight of a configuration is
N!
Ωn = Q .
i ni !

Equilibrium configuration: The most likely configuration.

Boltzmann distribution: The most-likely distribution of particles among the single-particle

states is:
N
nj = e−α−βj = e−j /kb T ,
Z
P −βj
where Z = j e is the partition function.
P P
How to derive: maximize Boltzmann entropy subject to constraints: j nj = N , j j nj = U ,
where j denotes a single-particle state. To relate β to T , use the first law of thermodynamics.
To relate α with Z, use the constraint on the particle number.
The probability that a particle is in the j th single particle state is given by:

e−βj
pj = .
Z

Stirling’s approximation: log x! ≈ x log x − x for large x.

1
Combinations with repetition: The number of ways of choosing n items from a set of k
where each item can be chosen multiple times and order does not matter is:

(n + k − 1)!
.
n!(k − 1)!
Think in terms of dividers:
× × | × || × . . . | × ×
where × denotes a single item and (k − 1) dividers divide them into k “types”. This is useful
for indistinguishable particles.

Canonical Ensemble
Canonical ensemble: Applies to a system in thermal equilibrium with a heat bath of temper-
ature T . The probability for each microstate is
1 −j /kB T X
pj = e , Z≡ e−j /kB T .
Z j

How to derive: Consider Ω of the combined system (heat bath and the system). The com-
bined system is isolated. The statistical weight of a microstate j is Ω(U − j ), since the heat
bath will have energy U − j . The rest hopefully follows.

Gibbs entropy: The entropy of a system obeying some ensemble described by some pj is
given by X
S = −kB pj log pj .
j

How to derive: Consider an assembly of N identical systems, all obeying the same ensemble.
The occupancy for a single microstate is N pj . Write down Ω and use Stirling’s approximation.

Internal energy: The internal energy is the mean energy of the microstates of the system.
In terms of the partition function:
X ∂ log Z ∂ log Z
U = hEi = p j j = − = kB T 2 .
j
∂β ∂T

Fluctuation in energy: The variance in energy is:

∂ 2 log Z
(∆E)2 = hE 2 i − hEi2 = = kB T 2 C V
∂β 2

Helmholtz function: In terms of the partition function:

F = U − T S = −kB T log Z.

2
Partition function factorization: The partition function factorizes for weakly interacting
particles. For a system of N such particles,
X
N
Z = Z(1) , Z(1) = e−βj .
j

Grand Canonical Ensemble

Grand canonical ensemble: Applies to systems with variable particle numbers in thermal
equilibrium with a heat bath. A microstate is characterized by energy j and particle number
Nj . The probability that the system is in the j th microstate is:
1 X
pj = e−(j −µNj )/kB T , Z = e−(j −µNj )/kB T .
Z j

How to derive: Consider an assembly of NA systems. Maximize entropy subject to constraints

(nj denotes the number of systems in microstate j):
X X X
n j = NA , nj j = NA U, nj Nj = NA N,
j j j

where U is the mean energy per system and N is the mean particle number per system. Solve
for nj , and pj = nj /NA . To relate β to temperature and λ to chemical potential, calculate
Gibbs entropy.

Grand potential: From the Gibbs entropy, it follows that

−kB T log Z = U − T S − µN ≡ Φ.

Indistinguishable Particles
Counting: Particles share a common volume and occupy a shared set of single particle states.
Microstates are described in terms of occupancies of shared single particle states.

Density of states: The density of states in k-space is

 3
L V
= 3.
2π (2π)
How to derive: Solve SE for a free particle. Construct a set of orthogonal wavevectors and
simply count.
The number of states in an element of k-space is
 3
L
dg = d3 k.
2π
The number of states with wavenumber k to k + dk is (where k ≡ kkk)
V 2
dg = k dk = g(k)dk.
2π 2

Density of states in energy: (for a free particle)

2πV 3/2
g() = (2m) 1/2 .
h3

3
Classical Ideal Gas
Mean occupancy: The mean occupancy is the average number of particles in the state:
N −j /kB T
fj = N pj = e .
Z
How to derive: Each particle is in a heat bath, so each obey the canonical ensemble. Have to
assume N pj  1 for all j - the gas must be dilute.

Partition function: Can approximate sum over all states j with an integral over energy
since energy levels are very close
X Z ∞
−j /kB T
Z= e ≈ Dg()e−/kB T d.
j 0

Factorization: The partition function factorizes, but since the particles are indistinguishable,
we have to divide by the number of permutations
N 3/2
Z(1)

2πmkB T
Z(N ) = , Z(1) = DV .
N! h2

Maxwell-Boltzmann distribution:
r  3/2
2 m 2
n(v)dv = N v 2 e−mv /2kB T .
π kB T

How to derive: n() = Dg()f (), so switch from energy to velocity.

Quantum Statistics
Fermions: Half-integer spin particles, obey Pauli exclusion.

Bosons: Integer-spin particles, do not obey Pauli exclusion.

Occupation of states: The trick is to treat each single-particle state as a system obeying
the grand canonical ensemble. It can exchange particles and energy with other states (that are
also systems). The probability that the jth state has occupancy nj is

1
Pr(nj ) = exp(nj (µ − j )β).
Zj

Each microstate is completely described by its occupancy.

Partition function: The partition function for the jth single particle state system is
X
Zj = exp(nj (µ − j )β).
nj

Fermi-Dirac distribution: The mean occupancy of a state with energy  is

1
fF D = .
e(−µ)/kB T + 1

4
Bose-Einstein distribution: The mean occupancy of a state with energy  is

1
fBE = .
e(−µ)/kB T − 1
How to derive: For Fermi-Dirac, nj ∈ {0, 1} and for Bose-Einstein, nj ∈ N. For both cases,
simply calculate Zj and hence hnj i.

Planck’s radiation law: The number density of photons is

V ω 2 dω
n(ω)dω =

π 2 c3 e~ω/kB T − 1

Spectral energy density is

n(ω) ~ ω 3 dω
u(ω)dω = ~ω dω = 2 3 ~ω/k T .
V π c e B −1
Stefan-Boltzmann law: Total energy density is
4
8π 5 h kB T

U= .
15c3 h

The energy flux is

2π 5 kB
4
j ∗ = σT 4 ,
. σ=
15h3 c2
Bose-Einstein condensation: We must have 0 > µ for the density of states to converge.
Setting µ = 0 = 0 yields
   3/2
3 2πmkB T
Nµ=0 = ζ V .
2 h2
This gives a maximum number for a given temperature, but we can still cool down the system.
Hence, we make the correction:
Z ∞
1
N = n0 + fBE ()g()d, n0 = −µ/k T
0 e B −1

Fermi energy: The chemical potential at T = 0 defines the Fermi energy:



F = µ  .
T =0

The total number of particles is related to F by

(
Z F
1  < F
N= Dg()d, since at T = 0, fF D () = .
0 0  > F