Chapter 7

The Grand Partition Function

Reminder:

Z = ∑ e -𝛽E(s) (1)
s
1
P(s) = e -𝛽E(s) (2)
Z
1
⏨=
X ∑ X(s)e -𝛽E(s) (3)
Z s
X is some quantity where expectation value is given with partition function

Basic Assumptions:

P(s 2 ) 𝛺 R (s 2 ) e SR (s2 )/k

= = SR (s1 )/k
= e [SR (s2 )-SR (s1 )]/k
P(s 1 ) 𝛺 R (s 1 ) e
subscript R denotes entropy or multiplicity of the resevoir

1
dS R = ( dU R + PdVR - 𝜇dN R ) (4)
T
volume often changes negligibly so disregard pdV term

P(s 2 ) e -[E(s2 )-𝜇N(s2 )]/kt

=
P(s 1 ) e -[E(s1 )-𝜇N(s1 )]/kt

Gibbs Factor = e -[E(s)-𝜇N(s)]/kt (5)

Grand Parition function / Gibbs sum:

Ƶ = ∑ e -[E(s)-𝜇N(s)]/kt (6)
s
Distribution functions
Fermi-Dirac

probability of a state being occupied is given by:

1 1
P(n) = e -𝛽[nϵ-𝜇n] = e -𝛽n[ϵ-𝜇]
Ƶ Ƶ
for fermions n must be one or zero for each state.
Therefore, the grand partition function becomes:

Ƶ = e 0 + e -𝛽[𝜖-𝜇] = 1 + e -(ϵ-𝜇)/kt

Recall from (3):

1
⏨=
X ∑ X(s)e -𝛽E(s)
Z s

1
n
⏨= ∑ n(s)e -(ϵ-𝜇)/kt
Ƶ n
For Fermions n can only be zero or one. Therefore:

e -(ϵ-𝜇)/kt 1
n
⏨= -(ϵ-𝜇)/kt
= (ϵ-𝜇)/kt
1 + e e +1
This is simply the Fermi-Dirac Distribution

1
n FD =
⏨ (ϵ-𝜇)/kt
e +1

Bose - Einstein

Ƶ = ∑ e -[E(s)-𝜇N(s)]/kt = 1 + e -(ϵ-𝜇)/kt + e -2(ϵ-𝜇)/kt + e -3(ϵ-𝜇)/kt + ....

s
Recall:
1
= ∑ 1 + x + x2 + x3 + ...
1-x
Therefore
 1
Ƶ =
1 - e -(ϵ-𝜇)/kt

We now want to find the average occupancy:

ne -n(ϵ-𝜇)/kt
n
⏨= ∑
n Ƶ
Now it's time for Clever Math Tricks to get rid of the n in the sum:

let x = (ϵ - 𝜇) / kt

-nx
n =
⏨ ∑ ne =
-1
∑ ∂
e -nx =
-1 ∂
∑ e -nx =
-1 ∂Ƶ
n Ƶ Ƶ n ∂x Ƶ ∂x n Ƶ ∂x

computing the parial

∂ -1
n = - 1 - e -x
⏨ 1 - e -x
∂x

1
= = n
⏨BE
e (ϵ-𝜇)/kt - 1

Degenerate Fermi Gas

Basic Assumptions:

p2
ϵ =
2m

h
p =
𝜆

n𝜆
L =
2
 Putting Them together
Imagine a large cloud of fermions - large enough to treat as a sphere in energy
space.
Assume T = 0

U = 2
h2
ϵ = 2
n x2 + n y2 + n z2
8L m
h2 2
ϵf = 2
nmx
8L m

1 4 3
N = * 𝜋 n mx
8 3

U = 2 ∑ ∑ ∑ ϵ(n)
nx ny nz
For large number of states treat as sphere in n-space:

𝜋/2 𝜋/2 nmx h2n2

U = 2∫ d𝜙∫ d𝜃∫ dn 2
*n 2 sin 𝜃
0 0 0 8L m
2 nmx
𝜋h
U = 2
∫0 n 4 dn
8L m
h2n2 n3
U = *
8L 2 m 5
Now Substitute for definitions of N and the Fermi Energy

3
U = N ϵf
5
For T ≠ 0

We define g such that:

𝜋 (8m) 2/3 3N
g(ϵ) = 3
V ϵ = ϵ
2h 2 ϵ F3/2
(makes the math look better)
 𝜖F
N = ∫0 g(𝜖)d𝜖 at T = 0
𝜖F
N = ∫0 g(𝜖) n
⏨FD d𝜖 for any T

Therefore to get U:

∞
U = ∫0 𝜖 g(𝜖) n
⏨FD d𝜖
Then the Sommerfeld Expansion happens.
(I have to much self respect to type that out)

3 𝜋2 (kT) 2
U = Nϵ F + N + ....
5 4 𝜖F

For Blackbody Radiation:

Start with the idea of the partition function

Recall: chemical potential is always zero for photons, because it takes no
energy to create or destroy them

Z = ∑ e -𝛽E(s)
s
Energies for each state are given by oscillator energies:
E n = 0, hf, 2hf, ....

Z = 1 + e -𝛽𝜖 + e -2𝛽𝜖 + e -3𝛽𝜖 + ...

1
= ∑ 1 + x + x2 + x3 + ...
1-x
1
Z =
1 - e -𝛽ϵ
 1 ∂Z hf
⏨= -
E =
Z ∂𝛽 e hf/kt - 1

dividing by energy hf gives planck distribution

1
n
⏨Pl = hf/kt
e -1
Planck Distribution gives the occuppancy of one mode ( mode = single particle state)

Now the Box of Radiation:

ϵ = pc
n𝜆
L =
2
2L
𝜆=
n
h
p =
𝜆
hcn
ϵ =
2L

U = 2∑ ∑ ∑ 𝜖 n
⏨Pl (𝜖) = ∑ hcn 1
nx ny nz L e hcn/2Lkt - 1
extra factor of 2 for polarization states
Then we pretend the sums are interals like before except n goes to infinity

𝜋/2 𝜋/2 ∞
U = 2∫ d𝜙∫ d𝜃∫ dn n 2 sin 𝜃 𝜖 ⏨
n Pl
0 0 0
𝜋/2 𝜋/2 ∞ hcn 1
U = 2∫ d𝜙∫ d𝜃∫ dn n 2 sin 𝜃
0 0 0 L e hcn/2Lkt - 1

U ∞ 8𝜋𝜖 3 / (hc) 3
= ∫0 d𝜖
V e 𝜖/kt - 1

Stuff under the integral is the energy density of a state.Gives the relative intensity of
radiation as a function of photon energy (frequency basically)
 8𝜋 𝜖3
u(𝜖) =
(hc) 3 e 𝜖/kt - 1

Back to the Integral:

U 8𝜋 (kT) 4 ∞ x3
= ∫0 dx
V (hc) 3 ex - 1
Spectrum Peaks at x = 2.82 or E = 2.82kt

0 2.5 5 7.5 10 12.5 15 17.5

Total Energy Density

U 8 𝜋 5 (kT) 4
=
V 15 (hc) 3

Here's Stefan's Law

(The Derivation for this one sucks and I don't want to do it)

P 2 𝜋 5 (kT) 4
= 3 2
= 𝜎 T4
A 15 h c

2 𝜋5 k4 W
𝜎 = = 5.67 × 10 -8
15 h 3 c 2 2
m K
4
 P = 𝜎eAT 4
e = emissivity which is a number between zero and 1.
1 is a perfect blackbody.

Now for Phonons

Basic Assumptions

(Phonons in a Box)
hv hvn
𝜖 = hf = =
𝜆 2L

1
n
⏨Pl =
e hf/kt - 1

U = 3∑ ∑ ∑ 𝜖 n
⏨Pl (𝜖)
nx ny nz
factor of 3 for each possible polarization state (I don't know what that means either)

If the crystal is a cube then the number of atoms in any direction is the cube root of the total

3
1 ≥ n ≤ N) n = number of atoms in a given direction

Now for the Debye thingy where we assume that we have a sphere
instead of a cube.

1/3
6N
R = n max =
𝜋

Now it's more of the same/similar:

𝜋/2 𝜋/2 nmx

U = 3∫ d𝜙∫ d𝜃∫ dn 𝜖 n
⏨Pl
0 0 0
 3𝜋 nmx hv n3
U = ∫0 dn
2 2L e hvn/2LkT - 1
This integral can not be solved analytically, but we're going to give
it plastic surgery to clean it up a bit

hvn
let x =
2LkT
1/3
hvn mx hv 6N TD
x mx = = =
2LkT 2kT 𝜋V T

1/3
hv 6N
TD =
2k 𝜋V

9NkT 4 T D /T x3
U = 3
∫0 dx
TD ex - 1
This is a lot better. (This is probably the one to memorize?)

Here's some limits

U = 3NkT and C V = 3Nk when T ≫ T D

3
3𝜋 4 NkT 4 12 𝜋 4 T
U = 3
and C V = Nk when T ≪ TD
5 TD 5 TD

For metals when T ≪ T D

12 𝜋 4 Nk T
C = 𝛾T +
5 TD
𝜋 2 Nk 2
𝛾 = 𝛾 is the contribution from free electrons
2𝜖 F

This is the function for intermediate temperatures

 T D /T x 4 exp(x)
C V = 9Nk ∫ dx
0 (exp(x) - 1) 2
(prolly doesnt matter that much)

I think that's everything. Thanks for reading.

I may have driven myself half mad trying to type this tonight, only time will tell whether or not
it was worth it. (It definitely wasn't)