You are on page 1of 83

Statistical Physics

Lecture Notes

Abstract
These notes represent the material covered in the Part II lecture Statistical Physics.
They are largely based on the more extended lecture notes by David Tong [2]. The main
purpose of these notes to provide as close as possible a one-to-one representation of the
course as it appears on the black board in the lecture room. This comes, at times, at the
expanse of uniformly using complete sentences and instead using short phrases and the
like.
Readers interested in more details as well as a wider range of subjects, will find David
Tong’s lecture notes an excellent source and may also find the following books of interest.
• E. Mandl, “Statistical Physics” .
• L. D. Landau & E. M. Lifshitz, “Statistical Physics” .
• F. Reif, “Fundamentals of Statistical and Thermal Physics” .
• M. Kardar, “Statistical Physics of Particles”, “Statistical Physics of Fields”; see also
Kardar’s webpage [3].
• A. B. Pippard, “The Elements of Classical Thermodynamics” .
Example sheets for this course will be available on the web page

http://www.damtp.cam.ac.uk/user/examples

Cambridge, January 2014

Ulrich Sperhake

1
CONTENTS 2

Contents
A The fundamentals of statistical physics 3
A.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
A.2 The microcanonical ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
A.2.1 Entropy and the 2nd law of thermodynamics . . . . . . . . . . . . . . . . 5
A.2.2 Temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
A.2.3 The Two-State system . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
A.2.4 Pressure, Volume, 1st law of thermodynamics . . . . . . . . . . . . . . . . 12
A.3 The canonical ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
A.3.1 The partition function . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
A.3.2 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
A.3.3 Free energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
A.4 The grand canonical ensemble . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
A.4.1 The chemical potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
A.4.2 The grand canonical ensemble . . . . . . . . . . . . . . . . . . . . . . . . 21
A.4.3 The grand canonical potential . . . . . . . . . . . . . . . . . . . . . . . . 21

B Classical Gases 23
B.1 From QM to classical . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
B.2 Ideal gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
B.2.1 Equipartition of energy . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
B.2.2 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
B.2.3 The ideal gas in the GrCE . . . . . . . . . . . . . . . . . . . . . . . . . . 27
B.3 Maxwell distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
B.4 Diatomic gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
B.5 Interacting gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
B.5.1 Mayer f function and B2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
B.5.2 Van der Waals equation of state . . . . . . . . . . . . . . . . . . . . . . . 34

C Quantum Gases 36
C.1 Density of states . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
C.1.1 Relativistic systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
C.2 Photons: Blackbody Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
C.2.1 Planck distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
C.2.2 Cosmic Microwave Background (CMB) . . . . . . . . . . . . . . . . . . . 40
C.2.3 The birth of QM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
C.3 Phonons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
C.4 The diatomic gas revisited . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
C.5 Bosons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
C.5.1 Bose-Einstein (BE) distribution . . . . . . . . . . . . . . . . . . . . . . . 44
C.5.2 QM gas at high T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
C.5.3 Bose Einstein condensation . . . . . . . . . . . . . . . . . . . . . . . . . 47
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 3

C.5.4 Heat capacity: A first look at phase transitions . . . . . . . . . . . . . . 49


C.6 Fermions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
C.6.1 Ideal Fermi gas . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
C.6.2 Degenerate Fermi gas and the Fermi surface . . . . . . . . . . . . . . . . 53
C.6.3 Fermi gas at low T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
C.6.4 White Dwarfs and the Chandrasekhar limit . . . . . . . . . . . . . . . . 56
C.6.5 Pauli paramagnetism (not lectured) . . . . . . . . . . . . . . . . . . . . . 57

D Classical Thermodynamics 59
D.1 Temperature and the 0th law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
D.2 The 1st law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
D.3 The 2nd law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
D.3.1 The Carnot cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
D.3.2 Thermodynamic temperature scale and ideal gas . . . . . . . . . . . . . . 63
D.3.3 Entropy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
D.4 Thermodynamic potentials: Free Energy, Enthalpy . . . . . . . . . . . . . . . . 67
D.4.1 Maxwell’s relations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
D.5 The 3rd law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

E Phase transitions 70
E.1 Liquid-gas transition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
E.1.1 Phase equilibrium . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
E.1.2 The Clausius-Clapeyron Equation . . . . . . . . . . . . . . . . . . . . . . 72
E.1.3 The critical point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
E.2 The Ising model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
E.2.1 Mean-field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
E.2.2 Critical exponents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
E.3 Landau Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
E.3.1 Second order phase transitions . . . . . . . . . . . . . . . . . . . . . . . . 81
E.3.2 First order phase transitions . . . . . . . . . . . . . . . . . . . . . . . . . 82

A The fundamentals of statistical physics


A.1 Introduction
Science works in layers, e.g.:
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 4

History
Economy
Psychology
Biology
Chemistry
Atomic Physics
Nuclear Physics
Particle Physics

Choose one area, consider the neighbours, ignore the rest.


Fundamental laws, large numbers
→ emergent phenomena, e.g. traffic jams, temperature

Statistical Physics: translate microphyics (fundamental laws)


→ macrophysics (temperature, color, . . . )

We will see that this can be done quite rigorously for many laws: ideal gas law, Wien’s dis-
placement, . . .

Not all macrosystems are understood at micro level: black holes, high T super conductors

Note: We have large numbers N ∼ 1023 6= 1

A.2 The microcanonical ensemble

E, N, ...
Isolated system: no exchange of energy, particles with outside
world
P R
We do this QM, but applies to classical systems as well ( → )

time independent Schrödinger eq.: Ĥ|ψi = E|ψi


ψ Eigenstate
E Eigenvalue

For N ∼ 1023 degrees of freedom: − impossible to solve


− unnecessary to solve: system jumps from state to state . . .
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 5

our view: mixed state with probability p(n) for state |ni
X
expectation value hÔi = p(n)hn|Ô|ni
n

equilibrium: probability distribution is time independent!


e.g. leave system for a while

Fundamental assumption: For an isolated system in equilibrium, all accessible mi-


crostates are equally likely

“accessible” = same energy E (for now)

Ω(E) = # of states with energy E


( 1
Ω(E)
if |ni has energy E
⇒ p(n) =
0 otherwise

23 )
Comments: • Ω(E) is absurdly large! E.g. 1023 2-state particles ⇒ Ω(E) = 2(10

• In QM, energy levels are discrete. For N ∼ 1023 they are finely spaced
→ almost like continuum.
We implicitly define Ω(E) as the # of states with energy ∈ [E, E + ∆E);
∆E ≪ measurement accuracy, ≫ level spacing

• p(n) has nothing to do with quantum uncertainty!


it’s just our ignorance.

A.2.1 Entropy and the 2nd law of thermodynamics


Def.: Entropy of system S(E) = kB ln Ω(E)
J
kB = 1.381 · 10−23 K
“Boltzmann’s constant”

Recall: N 2-state particles ⇒ Ω = 2N


⇒ S∼N
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 6

additive: consider 2 separate systems E2


E1
⇒ Ω(E1 , E2 ) = Ω1 (E1 ) Ω2 (E2 )
⇒ S(E1 , E2 ) = S1 (E1 ) + S2 (E2 )

The second law


bring the two systems together
E1 E2
They exchange energy: E1 → Ẽ1 , E2 → Ẽ2

Etot = E1 + E2 = Ẽ1 + Ẽ2

X
⇒ Ω(Etot ) = Ω1 (Ẽ1i ) Ω2 (Etot − Ẽ1i )
{Ẽi }
" #
X S1 (Ẽ1i ) S2 (Etot − Ẽ1i )
= exp +
kB kB
{Ẽi }

discreteness of QM energy levels: see comment above!

when the systems were separate, we had Ω(E1 , E2 ) states.


each such state is also one of the Ω(Etot ) states when the system is combined!

⇒ Ω(Etot ) ≥ Ω(E1 , E2 )

⇒ S(Etot ) ≥ S(E1 , E2 ) = S1 (E1 ) + S2 (E2 )

⇒ ∆S ≥ 0

For large N: recall that S ∼ N


" #
X S1 (Ẽ1i ) S2 (Etot − Ẽ1i )
⇒ the above sum Ω(Etot ) = exp +
kB kB
{Ẽ1i }

is a sum of exponentials of N ∼ 1023 .


Such sums are dominated by their maximum value!
S1 + S2
Say, for some energy Ẽ1i = E∗ , the exponent is twice as large as
kB
for any other E.
Then this term is ∼ eN times larger than all other terms.
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 7

⇒ All terms but the one with Ẽ1i = E∗ are negligible


Setting Ẽ1 = Ẽ1i , Ẽ2 = Etot − Ẽ1i , the exponent is
1 h i
S1 (Ẽ1 ) + S2 (Ẽ2 ) ,
kB
and it is maximal for Ẽ1i = E∗ if
∂ h i ∂S
1 ∂S2
i
S 1 ( Ẽ1 ) + S 2 ( Ẽ2 ) = − =0
∂ Ẽ1 ∂ Ẽ1 ∂ Ẽ2

∂S1 ∂S2
⇒ − =0 (∗)
∂E E∗ ∂E Etot −E∗

Then the total entropy is


S(Etot ) ≈ S1 (E∗ ) + S2 (Etot − E∗ ) ≥ S1 (E1 ) + S2 (E2 )

⇒ Subsystems 1 and 2 have nearly determined energies E∗ , Etot − E∗


after contact.
Note: If E1 6= E∗ , subsystem 1 will hardly ever return from energy E∗ to E1

“contact vastly enhances the number of accessible states”

Second Law: energy is rearranged such that S1 (Ẽ1 ) + S2 (Ẽ2 ) is maximal

A.2.2 Temperature
Note: We are slightly departing here from the microcanonic assumption E = const.
This is to be viewed as an ensemble of systems with different E.

1 ∂S
Def.: Temperature T : =
T ∂E

Why is this a good definition?

Does it describe coffee?


1) Units ok thanks to Boltzmann’s constant
2) Consider Eq. (∗): Energy rearranged such that S1 (Ẽ1 ) + S2 (Ẽ2 ) max.
Now assume E1 = Ẽ1 , E2 = Ẽ2 , i.e. no energy transfer at all
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 8


∂S1 ∂S2
⇒ − = 0 ⇒ T1 = T2
∂E E1 ∂E E2
⇒ No energy transfer corresponds to equal T1 = T2 (before contact) as expected.
3) Let us assume small energy transfer: δE1 = −δE2
!  
∂S1 ∂S2 ∂S1 ∂S2 1 1
⇒ δS ≈ δE1 + δE2 = − δE1 = − δE1
∂E E1 ∂E E2 ∂E E1 ∂E E2 T1 T2

δS > 0 because S maximized


if T1 > T2 ⇒ δE1 < 0
⇒ energy goes from hot to cold

Summary: T looks a good definition


equal T ⇒ equilibrium
we will evaluate T for ideal gas later; our def. is correct

Heat capacity

∂E
Def.: C ≡
∂T

Comments: • we should call this “energy capacity”

• C is nice: can be measured!

• consider E a function of T

∂S ∂S ∂E C
⇒ = =
∂T ∂E ∂T T

∂S
⇒ C=T
∂T
δQ
we will see this as C = in thermodynamics
δT
where we also specify what’s kept constant: CV , CP , . . .
Z T2
C(T )
we can measure entropy differences: ∆S = dT
T1 T

. . . beats counting ∼ exp(1023 ) states


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 9

C
Def.: specific heat capacity c =
N

∂S 1 ∂2S ∂ 1 ∂T ∂ 1 1
Note: = ⇒ 2
= = =− 2
∂E T ∂E ∂E T ∂E ∂T T T C
∂2S
Almost all substances have C > 0 ⇒ <0
∂E 2
⇒ The extremum in Eq. (∗) is really a max.

⇒ “thermodynamically stable systems”

Exception: black holes; Hawking radiation

A.2.3 The Two-State system

Stirling’s formula

We often have ln N!
 
1 1
Stirling: ln N! = N ln N − N + ln(2πN) + O (Examples)
2 N

ln p
N
X
For now: ln N! = ln p
p=1
Z N Z N
≈ ln p dp = 1 ln p dp
1 1 1 2 3
... p
Z N
p
= [p ln p]N
1 − dp = N ln N − (N − 1)
1 p

⇒ ln N! ≈ N ln N − N (lower limit!)
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 10

Two-spin-state system

N particles: non-interacting, 2 states: ↑, ↓

Let E↓ = 0, E↑ = ǫ

N↑ particles with spin up ⇒ N↓ = N − N↑

⇒ E = N↑ ǫ

what is Ω(E)?

pick N↑ particles from N

N
 N!
⇒ Ω(E) = N↑
=
N↑ ! (N − N↑ !)

N!
⇒ S(E) = kB ln = kB [ln N! − ln N↑ ! − ln(N − N↑ )!]
N↑ ! (N − N↑ )!
≈ kB [N ln N − N − N↑ ln N↑ + N↑ − (N − N↑ ) ln(N − N↑ ) + N − N↑ ]

= kB [(N − N↑ ) ln N + N↑ ln N − N↑ ln N↑ − (N − N↑ ) ln(N − N↑ )]
 
N − N↑ N↑ E
= −kB (N − N↑ ) ln + N↑ ln N↑ =
N N ǫ
    
E E E E
= −kB N 1− ln 1 − + ln
Nǫ Nǫ Nǫ Nǫ

S(E)
special cases: S(0) = 0
 

S = kB N ln 2 maximum
2
S(Nǫ) = 0
0 Nε / 2 Nε E
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 11

 
1 ∂S kB Nǫ
Temperature: = = ... = ln −1
T ∂E ǫ E
N↑ E 1
⇒ = = ǫ/(k T )
N Nǫ e B +1
N↑ 1
For T → ∞: =
N 2


What happens for E > ?
2
T < 0 ⇒ as we increase E, Ω(E) decreases
1
view as going through 0 to negative values
T

 
∂E ∂ Nǫ Nǫ2 eǫ/(kB T ) C
Heat capacity: C = = =
∂T ∂T eǫ/(kB T ) +1 kB T 2 (eǫ/(kB T ) + 1)2
ǫ
• C max near T ∼
kB
• T → 0 ⇒ C ∼ e−ǫ/(kB T ) → 0
T
“gap to first excited state”

⇒ heating “a bit” does nothing


1
• T →∞ ⇒ C ∼ → 0 “half the states are
T2
already ↑”

Schottky anomaly

for normal substances: C dominated by phonons or free electrons; spin negligible

⇒ C increases with T .
for special cases at low T : e.g. paramagnetic salts

spin contribution significant ⇒ C like Fig.


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 12

A.2.4 Pressure, Volume, 1st law of thermodynamics


Consider volume V of system

⇒ S(E, V ) = kB ln Ω(E, V )
 
1 ∂S
= ; “keep V constant”
T ∂E V

1 ∂S
Recall: we defined = and concluded that
T ∂E
2 systems keep their energies E1 , E2 if they have same T
 
∂S
Def: “Pressure”: p≡T
∂V E

Repeat the argument for T equilibrium

⇒ 2 systems keep their volume if they have the same p

   
∂S ∂S 1 p
First Law: from our definitions: dS = dE + dV = dE + dV
∂E V ∂V E T T

⇒ dE = T dS − pdV

Note: pdV = p A dx = F dx
dx
= work done by system
p
A
sign: dV < 0

⇒ we work on the system

⇒ dE > 0

energy conservation: T dS = heat δQ added to the system; cf. Sec. D

pdV = work done by the system

⇒ The system’s change in energy is equal to the heat added

plus the work done on the system


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 13

   
∂E ∂S
heat capacity: we now write CV = =T
∂T V ∂T V

↑  
δQ
Strictly but it’s the same since dV = 0
δT V
 
∂S ∂E
analagously: Cp = T (don’t use here!)
∂T p ∂T

Ludwig Boltzmann: did a lot of this in the absence of proof for atoms!

S = kB ln Ω on his tomb stone

his work received a lot of criticism;

truely appreciated after his suicide in Trieste in 1906

A.3 The canonical ensemble


Closed system: can exchange energy, but no matter, with outside world

• closed system S
S
• at equilibrium temperature T

• coupled to large reservoir R T R



changes in T of reservoir negligible

What is the number of states of the total system S + R?


 
X X SR (Etot − En )
Ω(Etot ) = ΩR (Etot − En ) = exp
n n
kB

where n = state of system S with energy En

ΩR = # of states of reservoir; SR = kB ln ΩR

Note: We sum over all states n, not over the energies En of S!

Otherwise: degeneracy factor g(En )


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 14

R large ⇒ En ≪ Etot
  X  
X SR (Etot − En ) SR (Etot ) ∂SR En
⇒ Ω(Etot ) = exp ≈ exp −
k B k B ∂E k
n n | {ztot} B
1
=
T
X
⇒ Ω(Etot ) ≈ eSR (Etot )/kB e−En /(kB T )
n

⇒ # of states where S sits in state |ni: eSR /kB e−En /(kB T )

Each of the Ω(Etot ) states equally likely

e−En /(kB T )
⇒ probability that S is in state |ni: p(n) = X
e−Em /(kB T )
m

“Boltzmann distribution”, “canonical ensemble”

Comments: • Reservoir only plays a role through T

• p ∼ e−E/(kB T ) ⇒ high-energy states unlikely

• E of system large ⇒ fewer states of reservoir to distribute its energy

• T → 0 forces system into ground state (lowest E)

A.3.1 The partition function


1
Def.: β ≡
kB T

X
Def.: “partition function” Z≡ e−βEn
n

= sum of probabilities of |ni up to normalization

e−βEn
⇒ Boltzmann distribution: p(n) =
Z

Z is the most important quantity in statistical physics!


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 15

Z is multiplicative: Consider two systems 1, 2


X (1) (2) Xh (1) (2)
i
⇒ Z= e−β[Em +En ] = e−βEm e−βEn
n,m n,m

Xh (1)
i Xh (2)
i
−βEm −βEn
= e e = Z1 Z2
m n

In QM we have 2 probabilities: - QM

- our ignorance of the system


−β Ĥ
e
⇒ density matrix: ρ̂ =
Z
⇒ p(φ) = hφ|ρ̂|φi = probability of state |φi

We won’t use that

Derivations from Z
X X En e−βEn
average energy: hEi = (pn En ) =
n n
Z


⇒ hEi = − ln Z
∂β


energy fluctuations: ∆E 2 = (E − hEi)2 = hE 2 i − hEi2

∂2 ∂
⇒ ... ⇒ ∆E 2 = 2
ln Z = − hEi
∂β ∂β

∂hEi ∂β ∂hEi 1
heat capacity: CV = = = ∆E 2
∂T V ∂T ∂β kB T 2

⇒ ∆E 2 = kB T 2 CV

Comments: 1) large fluctuations ∼ large heat capacity:


manifestation of “fluctuation-dissipation theorem”
∆E 1
2) Recall: CV ∼ N, E ∼ N ⇒ ∼√
E N
⇒ for large N: E peaked near hEi, essentially constant

in thermodynamic limit: microcanonic ≈ canonic; E = hEi


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 16

The two-spin-state system revisited


X βǫ
single particle: Z1 = e−βEn = 1 + e−βǫ = 2e−βǫ/2 cosh
n
2

N
Y βǫ
⇒Z= Zk = 2N e−N βǫ/2 coshN
k=1
2
 
∂ Nǫ βǫ Nǫ sinh x ex − e−x
⇒ hEi = − ln Z = . . . = 1 − tanh = =
∂β 2 2 1 + eβǫ cosh x ex + e−x
CV = . . .

Note: the partition function automatically handles the combinatorics

A.3.2 Entropy
MiCE: S = kB ln Ω(E) = kB ln of # states

Now: probability distribution over states with different E


S1 S2 ... SW−1 SW
Trick: take W identical systems, all coupled to R

For each system: states |ni, say n = 1 . . . N T R

⇒ # of systems in state |ni: p(n) W , if W large

Consider reservoir + all systems as microcanical (Etot fixed)


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 17

⇒ Ωtot = ΩR ΩS

what is ΩS ?

1) List all states n = 1 . . . N p(1) W p(2) W p(N) W


slots slots slots
2) create p(n) W slots for state n ... ... ...
...
P
3) of all slots = W
n=1 2 3 ... N
4) W ! = # of permutations of W systems

= # of ways to put them into slots

5) [p(n) W ]! = # of permutations of p(n) W systems

= # of ways to reshuffle systems in state n

without changing the physical setup

W!
⇒ ΩS = Y = # of different ways to get p(1) W systems in state |1i,
[p(n)W ]!
n p(2) W systems into state |2i, etc.

⇒ Stot = kB (ln ΩR + ln ΩS ) we ignore SR
X
⇒ SS = kB ln ΩS = (Stirling)
. . . = −kB W p(n) ln p(n) = entropy of W systems
n

X
⇒ one system: S = −kB p(n) ln p(n) due to Gibbs
n

Comments: • S = function of probability distribution

• MiCE: prob. distr. = f (Energy) ⇒ S = S(E)

• CE: prob. distr. = f (T )


1
• p(n) = ⇒ ... ⇒ S = kB ln Ω(E)
Ω(E)
" #
e−βEn kB X −βEn e−βEn kB β X
• p(n) = ⇒ S=− e ln = En e−βEn + kB ln Z
Z Z n Z Z n


⇒ S = kB (T ln Z)
∂T
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 18

MiCE vs. CE
X
S = kB ln Ω(E) vs. S = −kB p(n) ln p(n)
n

different probability distributions!

But for N → ∞ physical observables agree in both ensembles. How?


X X
Consider Z = e−βEn = Ω(Ei ) e−βEi
n {Ei }

N → ∞ ⇒ Ω(Ei ) e−βEi is strongly peaked at Ei = E∗ “e±N effect”

⇒ sum in Z dominated by E∗ term with max. condition


 
∂  −βE

Ω(E)e =0
∂E E∗

⇒ Z ≈ Ω(E∗ )e−βE∗
1 ∂S ∂ ln Ω ∂ ln Ω
We’ll need: = = kB ⇒ β=
T ∂E ∂E ∂E

⇒ hEi = − ln Z
∂β

=− [ln Ω − βE∗ ]
∂β
∂E∗ ∂ ln Ω ∂E∗
=− + E∗ + β
∂β ∂E∗ ∂β
= E∗ (!)


∂ ∂ ∂ 2 ∂
entropy: S = kB (T ln Z) = kB [T (ln Ω − βE∗ )] ∂T = −kB β ∂β

∂T ∂T

= kB ln Ω − kB βE∗ − kB T kB β 2 (ln Ω − βE∗ )
∂β
| {z }
=−E∗

= kB ln Ω(E∗ )

⇒ CE like MiCE at energy E∗ !


A THE FUNDAMENTALS OF STATISTICAL PHYSICS 19

Maximizing entropy

both MiCE and CE can be obtained from a variational principle


X
MiCE: Consider Gibbs entropy S = −kB p(n) ln p(n)
n

with p(n) 6= 0 for states |ni with energy E


X
constraint: p(n) = 1
n
" #
X
vary S + αkB p(n) − 1 ; α = Lagrange multiplier
n
" #
∂ X X
⇒ − p(n) ln p(n) + α p(n) − α = 0
∂p(m) n n
1
⇒ − ln p(m) − p(m) +α=0
p(m)
⇒ ln p(m) = α − 1

⇒ p(m) = eα−1 = const MiCE!

CE: keep hEi fixed (example sheet)

A.3.3 Free energy


Def.: Free energy F := E − T S “available energy”

Mathematically: Legendre trafo

⇒ dF = dE − d(T S) = T dS − pdV − T dS − SdT = −SdT − pdV



∂F ∂F
⇒ S=− , p= −
∂T V ∂V T

∂ ∂ ∂ ∂
Recall E=− ln Z , S = kB (T ln Z) , = −kB T 2
∂β ∂T ∂β ∂T
∂ ∂
⇒ F = E − T S = kB T 2 ln Z − kB T ln Z − kB T 2 ln Z
∂T ∂T
⇒ F = −kB T ln Z
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 20

A.4 The grand canonical ensemble


A.4.1 The chemical potential
Consider additional quantities of the system: particle number N , electric charge q

⇒ S = S(E, V, N) = kB ln Ω(E, V, N)

1 ∂S ∂S
recall: = , p=T
T ∂E V,N ∂V E,N

∂S
Def.: chemical potential µ = −T
∂N V,E

repeat argument for T -equilibrium

⇒ systems do not (net-)exchange particles if µ1 = µ2 “chemical equilibrium”

1 p µ
1st law: dS = dE + dV − dN ⇒ dE = T dS − pdV + µdN
T T T
⇒ µ = “energy cost to add one particle”

For electric charge we would get the electrostatic potential



∂S ∂E
Comment: µ = −T , but from first law: µ = Why?
∂N V,E ∂N S,V

in general: let x, y, z be variables with one constraint



∂x ∂y ∂z
⇒ · · = −1
∂y z ∂z x ∂x y
 −1
∂y ∂z
for us: x = E , y = N , z = S , constraint: V = const ; use =
∂z x ∂y x

If we work at constant temperature rather than energy:



∂F
we use dF = −SdT − pdV + µdN ⇒ µ =
∂N T,V
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 21

A.4.2 The grand canonical ensemble


GrCE: We now let the system also exchange particles with its environment

⇒ fixed T and µ

Def.: Let a state |ni correspond to energy En and particle number Nn


X
Grand canonical partition function Z(T, µ, V ) = e−β(En −µNn )
n

e−β(En −µNn )
⇒ (Sec.
. . .3) ⇒ p(n) =
Z

X ∂
Entropy: S = −kB p(n) ln p(n) ⇒ . . . ⇒ S = kB (T ln Z)
n
∂T


one also gets: hEi − µhNi = − ln Z
∂β
1 ∂
hNi = ln Z
β ∂µ
1 ∂2 1 ∂hNi
∆N 2 = 2 2
ln Z =
β ∂µ β ∂µ

A.4.3 The grand canonical potential


Def.: Grand canonical potential Φ = F − µN

dΦ = −SdT − pdV − Ndµ

View Φ = Φ(T, V, µ)

∂ ∂
with F = E − T S , Φ = E − T S − µN , = −kB T 2
∂β ∂T
∂ ∂
⇒Φ=− ln Z − T kB (T ln Z)
∂β ∂T

⇒ Φ = −kB T ln Z
A THE FUNDAMENTALS OF STATISTICAL PHYSICS 22

Def.: “extensive” quantities scale as system size: E, N, V , S


1 ∂S ∂S ∂S
“intensive” quantities size independent: = , p=T , µ = −T
T ∂E ∂V ∂N

F is extensive: F (T, λV, λN) = λ F (T, V, N)

Φ is extensive: Φ(T, λV, µ) = λ Φ(T, V, µ):

only one extensive independent variable!

⇒ Φ must be ∼ V !

∂Φ
We know = −p ⇒ Φ = −p(T, µ) V
∂V T,µ

A lot of this developed by Josiah Willard Gibbs (1839-1903)


B CLASSICAL GASES 23

B Classical Gases
gas = particles flying around in a box

classical, but QM is often “there” in the background

we’ll typically use the CE


p~2
one particle: H = + U(~q) = Ekin + Epot
2m
state = point in phase space {(qi , pi )}
Z
P R 1
→ ⇒ partition function for one particle Z1 = 3 e−βH(~p,~q) d3 p d3 q
h
h = 6.6 · 10−34 J s
why h in a classical formula?

B.1 From QM to classical


p̂2
One particle in 1 dim. in QM: Hamilton operator: Ĥ = + U(q̂)
2m
Ĥ |ni = En |ni
↑ տ
Eigenstate energy
Z Z
Recall identity operator: 1 = dq |qihq| , 1 = dp |pihp|

X X
⇒ Z1 = e−βEn = hn|e−β Ĥ |ni
n n

X Z Z
−β Ĥ
= hn| dq |qihq|e dq ′ |q ′ihq ′ |ni
n

Z ( )
X X
dq dq ′ hq|e−β Ĥ |q ′ i [hq ′ |nihn|qi] |nihn| = 1 , hq ′ |qi = δ(q − q ′ )

=
n
n
Z
⇒ Z1 = dqhq|e−β Ĥ |qi “ = Tr e−β Ĥ ”
B CLASSICAL GASES 24

1
Recall from QM: e eB̂ = eÂ+B̂+ 2 [Â,B̂]+...

for us: Â = q̂, B̂ = p̂, [q̂, p̂] = i~


 
p̂2
−β +U (q̂) p̂2
−β Ĥ
⇒e =e 2m
= e−β 2m · e−βU (q̂) + O(~)
Classical limit: ~ → 0 !
Z
p̂2
⇒ Z1 = dqhq|e−β 2m e−βU (q̂ |qi |qi eigenstates to q̂ ⇒ U(q̂)|qi = U(q)|qi

↑ ↑
operator number
Z
p̂2
⇒ Z1 = dq e−βU (q) hq|e−β 2m |qi

Z  2


′ −βU (q) −β 2m ′ ′
= dq dp dp e hq|pihp|e |p ihp |qi

1 i
Note: hq|pi = hp|qi∗ = √ e ~ pq , hp|p′i = δ(p − p′ )
2π~
Z
1
In 3 dims.: Z1 = d3 q d3 pe−βH(~p,~q)
(2π~)3

B.2 Ideal gas


Def.: gas = N particles trapped in box of volume V

“ideal” = particles do not interact, i.e. U(q) = 0

“monatomic” = particles have no structure (i.e. no vibration, rotation)


Z Z
1 3 3 p
~
−β 2m
2
⇒ Z1 (V, T ) = d q d pe d3 q = V
(2π~)3
Z ∞
r
−ax2 π
We’ll often use e dx =
−∞ a
Z  3/2
V p2 p2
y p2 mkB T β 1
x
−β 2m −β 2m z
−β 2m
⇒ Z1 (V, T ) = dpx dpy dpz e e e =V a = =
(2π~)3 2π~2 2m 2mkB T
B CLASSICAL GASES 25

s
2π~2
Def.: thermal de Broglie wavelength λ=
mkB T

“average de Broglie wavelength at T ”

V
⇒ Z1 =
λ3

Z1N VN
N indistinguishable particles: Z(N, V, T ) = = 3N
N! λ N!
exchange two particles → no new state! Hence N!
Recall: F = −kB T ln Z = −kB T [N ln V − 3N ln λ − ln N!]

∂F NkB T
p= − =+ ⇒ ideal gas law: pV = NkB T
∂V T V

Comments: • all ~ have disappeared!

• equations linking p, T, V are called “equations of state” (EOS)


∂S

• T as defined above ∂E is really a “good” temperature!

• in the lab: deviations at higher densities


expected: U 6= 0

B.2.1 Equipartition of energy


 
∂ ∂ ∂ 1 3
energy: E=− ln Z = − [−3N ln λ] = 3N ln β = NkB T
∂β ∂β ∂β 2 2
Assume the gas existed in D spatial dimensions
VN D
⇒ ... ⇒ Z = ⇒ E= NkB T
λDN N! 2
NkB T
⇒ D degrees of freedom; each contributes to E
2
NkB T
equipartition of energy: each degree of freedom contributes to E
2
breaks down for QM systems at small T
B CLASSICAL GASES 26

p2 √ p
For one particle: hEi = ⇒ p∼ mE ∼ mkB T
2m
h
de Broglie wavelength λdB = ∼ λ up to factors of 2 or π
p

∂E 3
heat capacity: CV = = NkB
∂T V 2

Comment on kB : Why is it so small?

energy of gas E ∼ NkB T

E, T have units where numbers are O(1) ⇒ NkB = O(1)


1
# stars in universe ∼ 1022 < N (!) ⇒ kB ∼ ∼ 10−23
N

Chemist Notation

Def.: Avogadro’s number NA ≡ # atoms in 12 g of Carbon12

1 mol ≡ NA atoms

ideal gas law: pV = NkB T = nRT


N J
with: n ≡ , R ≡ NA kB ≈ 8 “Universal gas constant”
NA K mol

B.2.2 Entropy

Recall entropy for CE: S = kB (T ln Z)
∂T
s
∂ ∂ 2π~2 ∂ 1
ln λ = ln = ln T −1/2 = −
∂T ∂T mkB T ∂T 2T


⇒ S = kB ln Z + kB T [N ln V − 3N ln λ − ln N!]
∂T
 
VN 3N
= kB ln 3N − ln N! + kB T Stirling : ln N! ≈ N ln N − N
λ 2T
 
VN 3
≈ kB ln 3N N + N + NkB
λ N 2
B CLASSICAL GASES 27

 
V 5
⇒ S = NkB ln 3 + “Sackur-Tetrode equation”
λN 2

Comments: • S has ~ ; classically we only measure ∆S ⇒ ~ drops out

• S measures the factor N! in Z

Gibbs noticed this before QM:

Mix “red” and “blue” gas ⇒ entropy increases

Mix “red” and “red” gas ⇒ entropy does not increase → N!

B.2.3 The ideal gas in the GrCE


view as subvolume inside larger gas

" #
X  X X
Zideal (µ, V, T ) = eβNn µ e−βEn = eβN µ e−βEm
n N m
տ
states for fixed N
∞ ∞
"  N #
X  βµN  X  N V 1
= e Zideal (N, V, T ) = eβµ
N =0 N =0
λ3 N!
 
eβµ V
= exp
λ3

1 ∂ eβµ V
⇒ hNi = N = ln Z =
β ∂µ λ3
 3 
λ N
⇒ µ = kB T ln
V
V
= average volume per particle
N
λ ≈ average de Broglie wavelength of particle
B CLASSICAL GASES 28

V
λ3 ≈ ⇒ QM effects important
N
V
classical limit ⇒ λ3 ≪ ⇒ µ<0
N

∂E
Comments: • µ= = energy cost of adding particle at constant S, V !
∂N S,V

extra particle ⇒ more ways of distributing energy

⇒ S would increase unless E decreases

⇒ E decreases ⇒ µ < 0
• µ > 0 possible for some special cases

1 ∂2
Fluctuations: ∆N 2 = ln Z = N
β 2 ∂µ2
∆N 1
⇒ =√ → 0 in thermodynamic limit
N N

eβµ V
Recall: pV = −Φ = kB T ln Z = kB T = kB T N ideal gas law !
λ3

B.3 Maxwell distribution


Goal: velocity distribution
Z
1 − β~
p 2
3 3 ~v = p~

1 particle, ideal gas: Z1 = e 2m d qd p
(2π~)3 m
Z
m3 V v2
m~
= e−β 2 d3 v
(2π~)3
Z
4πm3 V mv 2
= v 2 e−β 2 dv
(2π~)3

mv 2

⇒ probability that atom speed in [v, v + dv]: p(v) = N v 2e 2kB T “Maxwell
 3/2 distribution”
m
N = 4π
2πkB T
B CLASSICAL GASES 29
p(v)
Z ∞
3kB T
⇒ hv 2 i = p(v) v 2dv = . . . = Xe
0 m

equipartition of energy
ր
He

0 1000 2000
km / s

Maxwell’s argument: prob. distribution in x dir.: φ(vx )

rotational symmetry ⇒ same in y, z dir.


+ distribution cannot depend on direction
q 
⇒ p(~v) = p vx2 + vy2 + vz2 = p(v)

⇒ p(v) dvx dvy dvz = φ(vx ) φ(vy ) φ(vz ) dvx dvy dvz
2
It can be shown that the only solution is φ(vx ) = Ae−Bvx
2
⇒ p(v) dvx dvy dvz = 4πv 2 p(v)dv = 4πA3 v 2 e−Bv dv
m
equipartition of energy ⇒ B =
2kB T

History of kinetic theory

kinetic theory = understanding gas properties through atoms

1783 Bernoulli: pressure = bombardment of atoms


−vx
∆px = 2mvx
2L vx
next “hit” after ∆t = y
vx
L
∆px mvx2 x
⇒ F = =
∆t L
Nmhvx2 i Nmhv 2 i
all atoms: F = =
L 3L
F Nmhv 2 i Nmhv 2 i ! NkB T
⇒ pressure p = = = = (gas law)
A 3L3 3V V
1 3
⇒ mhv 2 i = kB T
2 2
B CLASSICAL GASES 30

1843: Waterson: rejected paper “Thoughts on Mental Functions”

1857: Clausius: rotating, vibrating modes

1859: Maxwell: distribution, gas viscosity independent of density

→ experiment, Cavensdish Lab

Boltzmann Eq.: dynamics in phase space → non-equilibrium

B.4 Diatomic gas


Molecules with 2 atoms ≈ 2 masses attached to spring

⇒ i) Rotation about 2 axes (ignore symmetry axis)

ii) vibration

⇒ Z1 = Ztrans Zrot Zvib

i) Rotation
1  
Lagrangian: Lrot = I θ̇2 + sin2 θ φ̇2
2
∂Lrot ∂Lrot
⇒ conjugate momenta: pθ = = I θ̇ , pφ = = I sin2 θ φ̇
∂ θ̇ ∂ φ̇
p2θ p2φ
⇒ Hamiltonian: Hrot = pθ θ̇ + pφ φ̇ − Lrot = +
2I 2I sin2 θ
Z Z r
1 −βHrot
−ax2 π
⇒ Zrot = 2
e dpθ dpφ dθ dφ e dx =
(2π~) a
Z π Z 2π
2πI 2I
= sin θ dθ dφ =
β(2π~)2 0 0 β~2
∂ ∂ 1
⇒ Erot = − ln Zrot = − [− ln β ± . . .] = = kB T
∂β ∂β β
B CLASSICAL GASES 31

⇒ gas with rotation: Z1 = Ztrans Zrot ∼ (kB T )5/2


Z1N 5
Z= ⇒ E = NkB T
N! 2
5
CV = NkB : 3 + 2 deg. of freedom
2

ii) vibration
ζ
harmonic oscillator with frequency ω

p2ζ 1
⇒ Hvib = + mω 2 ζ 2
2m 2
Z
1
⇒ Zvib = e−βHvib dpζ dζ
2π~
Z p2
1 ζ mω 2 ζ 2 1 kB T
= e−β 2m e−β 2 dpζ dζ = =
2π~ ~βω ~ω

⇒ Evib = kB T : 2 degrees of freedom! potential + motion

CV
7
diatomic gas: CV = NkB 7/2
vib
2
5/2
rot
Note: • I, ω dropped out
3/2
• at lower T : dof’s frozen out, e.g. H2 : tran
100 1000
QM effect; visible in hot gas! T/K

B.5 Interacting gas


Consider monatomic gas
N
ideal gas good for small
V
p N N2 N3
Virial expansion: = + B2 (T ) 2 + B3 (T ) 3 + . . .
kB T V V V
տ ր
virial coefficients
B CLASSICAL GASES 32

goal: get Bi (T ) from first principals, i.e. potential U(r) between atoms
1 p1 p2
2 features of U(r): • attractive 6
: dipols p1 , p2 ⇒ pot. energy 3
r r
p1
instantaneous dipol p1 ⇒ electr. field E ∼
r3
p1
⇒ induced dipol p2 ∼ E ∼
r3
“van der Waals interaction”

• strong repulsion from Pauli exclusion

 r 12  r 6
0 0
Lennard-Jones potential: U(r) ∼ − ; 12 = 6 · 2 chosen for convenience
r r
 U
∞ r < r0
hard-core potential: U(r) = 
 r0 6
−U0 r
r ≥ r0

hard-core easier! Take that... r0 r

B.5.1 Mayer f function and B2


Notation: ~r = particle position (instead of ~q)

N
X p~2i X
H= + U(rij ) ; rij ≡ |~ri − ~rj |
i=1
2m i>j
տ
count pairs once!

N
Z Y
1 1
⇒ Z(N, V, T ) = d3 pi d3 ri e−βH
N! (2π~)3N i=1
"Z # "Z #
1 1 Y P 2 Y P
= 3N
d3 pi e−β j p~j /(2m) · d3 ri e−β j<k U (rjk )
N! (2π~) i i
Z Y
1 P
= 3N d3 ri e−β j<k U (rjk )
λ N!
| i {z }
hard!
B CLASSICAL GASES 33

P X β2 X
1st try: Taylor: e−β j<k U (rjk )
= 1−β U(rjk ) + U(rjk ) U(rlm ) ± . . .
j<k
2 j<k, l<m, l>j

problem: for rij → 0: U(rij ) → ∞ not good expansion parameter

2nd try: Mayer f function: f (r) = e−βU (r) − 1

r → ∞ ⇒ f (r) → 0 , r → 0 ⇒ f (r) → −1

P Y
Def.: fij ≡ f (rij ) ⇒ e−β j<k U (rjk )
= (1 + fjk )
j<k

Z Y !
1 X X
⇒ Z(N, V, T ) = d 3 ri 1 + fjk + fjk flm + . . .
N! λ3N i j>k j>k, l>m, l>j

1st term: 1 → V N , ideal gas

2nd term: sum of terms like


Z Y Z Z
3 N −2 3 3 N −1
I12 ≡ d ri f12 = V d r1 d r2 f (r12 ) = V d3 r f (r)
i

where ~r = ~r1 − ~r2 , ~ = 1 (~r1 + ~r2 )


R
2
we were a bit sloppy with limits of integration

but this only matters near the boundary of the system

because f (r) has contributions only for r ≈ atomic distances

1 N2
We have N(N − 1) ≈ terms of type I12
2 2
 Z 
VN N2 3

(1 + ǫ)N ≈ 1 + Nǫ
⇒ Z(N, V, T ) = 1+ d r f (r) + . . .
N! λ3N 2V
 Z N
N 3
≈ Zideal 1 + d r f (r)
2V
 Z 
N 3
⇒ F = −kB T ln Z = Fideal − NkB T ln 1 + d r f (r)
2V
| {z }
→B2 (T )
B CLASSICAL GASES 34

What do we mean with low density?


Z
For Lennard-Jones or hard-core potential one can show: f (r)d3r ∼ r03

where r0 ∼ location of min. of potential


≈ atom size
The expansion is valid if higher-order terms are small
N 3 N 1
⇔ r ≪1 ⇒ ≪ 3
V 0 V r0
N 1
For liquids: atoms packed ⇒ ≈ 3
V r0
⇒ expansion good for gases at densities much below liquid state

B.5.2 Van der Waals equation of state


 Z 
∂F NkB T N 3
With ln(1+x) ≈ x in F ⇒ p = − = ... = 1− f (r)d r + . . .
∂V V 2V
Z
pV N
⇒ =1− f (r)d3 r ; f (r) = e−βU (r) − 1
NkB T 2V

Cases:

1) repulsion: U(r) > 0 ∀r and U(r → ∞) = 0 ⇒ f (r) < 0 ⇒ p > pideal

2) attraction: U(r) < 0 ⇒ . . . ⇒ p < pideal


Z Z r0 Z ∞ h i
3 3 3 +βU0 (r0 /r)6
3) hard-core: f (r)d r = −1d r + dr e −1
0 r0

6
 r 6
0
high T limit: βU0 ≪ 1 ⇒ eβU0 (r0 /r) ≈ 1 + βU0
r
Z Z r0 Z ∞
3 4πU0 r06
⇒ f (r)d r ≈ −4π r 2 dr + dr
0 kB T r0 r4
3
 
4πr0 U0
= −1
3 kB T
B CLASSICAL GASES 35

 
pV N a 2πr03U0 2πr03
⇒ =1− −b ; a= , b=
NkB T V kB T 3 3
  −1   
V N2 N N2 V
⇒ . . . ⇒ kB T = p+ 2a 1+ b ≈ p+ 2a −b
N V V V N
N
“van der Waals EOS”: valid at low , high T
V
NkB T N2
⇒p= −a 2
V − bN V

Comments: • a depends on U0 ⇒ attractive large r effects → smaller p

• b only function of r0 → hard-core repulsion → reduces V


4
• one atom blocks volume πr03
3
2 3
but b = πr0 Why?
3 r0 /2 r0 /2
st
1 atom has space V ,

2nd atom has space V − Ω , Ω = 2b

etc.

N −1  
Y N2 Ω
(V −mΩ) ≈ V N

whole configuration space: 1− + ... Ω≪V
m=0
2 V
 N

≈ V −N = (V − Nb)N
2
• Our method can only handle potentials
1
U(r) ∼ with n > 3 . Otherwise divergent integral of f (r)
rn
 
a
• 2nd virial term of van der Waals EOS: B2 (T ) = − −b
kB T
• Higher-order virial coefficients: e.g. cluster expansion,

cf. Sec. 2.5.3 in David Tong’s notes [2]

• More on van der Waals in Sec. E (Phase transitions)


C QUANTUM GASES 36

C Quantum Gases
gases where QM effects are important; includes light, phonons,. . .

C.1 Density of states


X Z
often convenient to → . . . dE ; requires density of states
n

ideal QM gas: no interaction


1 ~
model particles as plane waves ψ = √ eik~x
V
L

impose periodic boundary conditions


(the physics can be shown not to depend on the BCs)
L
2πni
⇒ ki = , ni ∈ Z
L
~2~k 2 4π 2 ~2 2 ∂
∧ E~n = = (n + n22 + n23 ) , k = |~k| , p̂ = −i~
2m 2mL2 1 ∂xi

X
one particle: Z1 = e−βE~n
~
n
s
2π~2 E~n ~2~k 2 λ2~n2
recall λ = ⇒ exponent =β ∼ 2
mkB T kB T 2m L

for macroscopic box: λ ≪ L ⇒ many ~n with E~n ≤ kB T


X Z L3
Z
4πV
Z
3 3
⇒ ≈ d n= dk= k 2 dk
(2π)3 (2π)3
~
n

~2 k 2 ~2 k
E= ⇒ dE = dk
2m m
Z r Z
X V 2mE m
⇒ ≈ 2 dE =: g(E)dE ,
2π ~2 ~2
~
n

 3/2
V 2m √
g(E) = 2 E
4π ~2
C QUANTUM GASES 37

Comments: • g(E)dE = # of states in energy interval [E, E + dE)

• looks classical, but useful for many QM systems!

C.1.1 Relativistic systems


√ ~2 k 2
all that changes is E = ~2 k 2 c2 + m2 c4 instead of
2m
~2 kc2
⇒ dE = dk
E

E dE E 2 − m2 c4
⇒ . . . ⇒ k 2 dk = 2 2
~c ~c
VE √ 2
⇒ ... ⇒ g(E) = E − m2 c4
2π 2 ~3 c3
V E2
for massless particles: g(E) = 2 3 3
2π ~ c

C.2 Photons: Blackbody Radiation


light = gas of photons

color of light at T = color of any object at T (equilibrium!)

if we ignore the atomic makeup (absorption lines, emission lines),

→ “blackbody radiation” because at T = 0 such a body is black

2πc
Photons: wavelength λ , frequency ω = = kc
λ
energy E = ~ω ; m=0

2 polarization states (transverse) ⇒ g(E) picks up extra factor of 2


V E2 V ω2
⇒ g(E) dE = dE = dω
π 2 ~ 3 c3 2 3
|π{zc }
=:g̃(ω)

g̃(ω) = # states for a single photon with frequency ∈ [ω, ω + dω)


C QUANTUM GASES 38

Note: photons are not conserved! Unlike atoms

→ one could work in GrCE with µ = 0 . In this case Φ = F

and our above relations of E, F, Φ to Z or Z become the same.

We’ll stick with CE notation.

partition function for photons in fixed state n with ωn :

Zωn = 1 + e−β~ωn + e−2β~ωn + . . . = “summing over all N”


1
=
1 − e−β~ωn

take all frequencies ⇒ the Z multiply ⇒ the ln Z add up


Z ∞ Z ∞
V 
⇒ ln Z = g̃(ω) ln Zω dω = − 2 3 ω 2 ln 1 − e−β~ω dω
0 π c 0

C.2.1 Planck distribution


Z ∞ Z ∞
∂ V~ ω3
energy: E = − ln Z = 2 3 dω = E(ω)dω
∂β π c 0 eβ~ω − 1 0

V~ ω3
⇒ E(ω) = = photon energy density in frequency space
π 2 c3 eβ~ω − 1

Comments: • as T decreases:

ωmax decreases, λmax increases

• ωmax = ζ kB~T ,

where 3 − ζ = 3e−ζ ⇒ ζ ≈ 2.822

“Wien’s displacement law”

→ color of object at T
C QUANTUM GASES 39

Z ∞

V (kB T )4 x3
total energy: E= 2 3 dx x = β~ω
π c ~3 ex − 1
|0 {z }
4
= ... = π15

E π 2 kB
4
⇒ = T4
V 15~3 c3

E c π 2 kB
4
J
energy flux from object = luminosity: L= = σT 4 ; σ= = 5.67 × 10 −8
V 4 60~3 c2 m2 s K
“Stefan-Boltzmann law” (cf. Reif [1] for details)

Comments: • factor c because it is flux


1
• 2 factors of : a) flux goes only away from object, not into it
2
θ
b) angular factor in integral

Z ∞

V kB T 2 −β~ω

pressure: F = −kB T ln Z = 2 3 ω ln 1 − e dω int. by parts
π c 0

Z ∞
V~ ω 3 e−β~ω
=− 2 3 dω
3π c 0 1 − e−β~ω
Z ∞
V~ 1 x3 V π2
=− 2 3 4 4 x
= − 3 3
(kB T )4
3π c β ~ 0 e −1 45~ c

∂F 1E 4σ 4
⇒ p= − = = T
∂V T 3V 3 c

1
Note: photon pressure = photon energy; important in cosmology!
3


∂F 16V σ 3 ∂E 16V σ 3
We also get: S=− = T , CV = = T
∂T V 3c ∂T V c
C QUANTUM GASES 40

C.2.2 Cosmic Microwave Background (CMB)

CMB = afterglow after big bang

measured by COBE, WMAP, PLANCK

CMB ≈ 2.725 K blackbody radiation (within 10−5)

photons have travelled for ∼ 13.7 billion years

the disagreement → exciting physics

C.2.3 The birth of QM


Consider the classical limit of Planck’s formula:

V~ ω3 V ~ ω3 V ω 2 kB T
x ≪ 1 ⇒ ex ≈ 1 + x
E(ω) = 2 3 β~ω ≈ 2 3 = 2 3
π c e −1 π c β~ω | π{zc }

= ERJ (ω)
“Rayleigh-Jeans”

Z ∞
Problem: ERJ (ω) dω diverges: “ultraviolet catastrophe”
0

QM: for ~ω ≫ kB T , the temperature cannot excite even one photon.

states “frozen out”

Planck used E = ~ω and Boltzmann statistics ⇒ first hint of QM


C QUANTUM GASES 41

C.3 Phonons
vibrations of crystal = sound waves

QM: electromagnetic waves 7→ photons

QM: sound waves 7→ phonons

phonon energy: E = ~ω = ~kcs (speed of sound)

L
differences from photons: • cs instead of c

• 3 polarization states; also longitudinal

• upper frequency limit:


2πcs
λ= < L ⇒ no shaking possible
ω
⇒ ω < ωD “Debye frequency”
 1/3
N
We expect ωD ∼ cs , but what is the prop. constant?
V
Debye: total number of atoms: N ; each atom has 3 directions of movement

⇒ 3N ways of moving
Z ωD
set 3N = g̃(ω) dω which is the number of states available for one phonon
0

Z ωD  1/3
3 V ω2 V ωD 3
6π 2 N
⇒ 3N = dω = ⇒ ωD = cs
0 2 π 2 c3s 2π 2 c3s V

~ωD
Def.: Debye temperature TD = = temperature at which highest-frequency
kB
states become excited

TD ≈ 100 K for lead, 2 000 K for diamond, ≈ 200 . . . 400 K for many materials
C QUANTUM GASES 42

phonons are not conserved → same game as for photons:


Z ωD
1
Zω = , ln Zphonon = g̃(ω) ln Zω dω
1 − e−β~ω 0
Z ωD Z ωD
g̃(ω) 3V ~ ω3
⇒E= ~ω β~ω dω = 2 3 dω x = β~ω
0 e −1 2π cs 0 eβ~ω − 1
Z
3V (kB T )4 TD /T x3
⇒E= dx
2π 2 (~cs )3 0 ex − 1

no analytic solution
Z ∞
x3 π4
Limits: 1) T ≪ TD ⇒ upper int. limit ≈ ∞: dx =
0 ex − 1 15
 3
∂E 2π 2 V kB
4
3 12π 4 T
⇒ CV = = T = NkB
∂T V 5(~cs )3 5 TD

2) T ≫ TD ⇒ Taylor expand integrand


Z TD /T Z TD /T  3
x3 2 1 TD
⇒ dx ≈ (x + . . .) dx = + ...
0 ex − 1 0 3 T
4 3
V kB T
⇒E∼T ∧ CV = 2 3 D3 = 3NkB (with N above)
2π ~ cs

Comments: • The high T behaviour known since early 1800s:

“Dulong-Petit law”

• Debye’s ωD reproduces the “3N”

• Debye’s new contribution was limit 1)


C QUANTUM GASES 43

C.4 The diatomic gas revisited


Recall Figure for CV : rotation, vibration frozen out at low T CV
7/2
vib

5/2
rot
1) Rotation
3/2
tran
p2θ p2φ
H= + 100 1000
2I 2I sin2 θ T/K

~2
QM ⇒ E levels: E= j(j + 1) ; j = 0, 1, 2, . . .
2I
degeneracy 2j + 1 (Lz quantum numbers)

X 2 j(j+1)/(2I)
⇒ Zrot = (2j + 1)e−β~
j=0

~2 β~2
a) T ≫ ⇒ ≪1
2IkB 2I
Z ∞
2 x(x+1)/(2I) 2I
⇒ Zrot ≈ (2x + 1)e−β~ dx = = Zclassic
0 β~2

~2
b) T ≪ ⇒ Zrot ≈ 1
2IkB
T insufficient to excite even the 1st state above ground level.

Also explains why monatomic gas has no rot. degree of freedom:


~2
I tiny → T ≪ ⇒ “frozen out for almost all T ”
2IkB

2) Vibration
 
1
harmonic oscillator: En = ~ω n +
2
X X e−β~ω/2 1
e−β~ω(n+ 2 ) = e−β~ω/2
1
⇒ Zvib = e−β~nω = =
n n
1−e −β~ω
2 sinh β~ω
2
C QUANTUM GASES 44

~ω 1
a) T ≫ ⇒ Zvib ≈ = Zclass
kB β~ω


b) T ≪ ⇒ Zvib ≈ e−β~ω/2 = contribution from zero-point energy
kB
∂ ~ω
⇒ energy offset Evib = − ln Zvib = , but no contribution to CV
∂β 2

C.5 Bosons
QM: 2 types of particles: 1) bosons: integer spin: ψ(~r1 , ~r2 ) = ψ(~r2 , ~r1 )
1
2) fermions: 2
-integer spin: ψ(~r1 , ~r2 ) = −ψ(~r2 , ~r1 )

p, n, e fermions ⇒ odd # of p, n → fermions, e.g. He3

even # of p, n → bosons, e.g. He4


s
2π~2
thermal de Broglie λ =
mkB T
 1/3
V
T small ⇒ λ large; eventually ≈ particle separation
N
⇒ QM important

Here: only monatomic gases, non-interacting

C.5.1 Bose-Einstein (BE) distribution


Notation: state: |ri, # of particles in state |ri : nr

particles indistinguishable
⇒ system described by n1 ≡ # of particles in state |1i, n2 , n3 , . . .
X P X
CE: Z = e−β r nr Er ; sum over all states {nr } with nr = N ; Tricky!
{nr } r
C QUANTUM GASES 45

GrCE: chem. potential µ, N can fluctuate

any state can be occupied by any # of particles


X 1
⇒ Zr = e−βnr (Er −µ) = −β(E r −µ)
n
1−e
r

X X
“ # particles in state” = “ states of particles”
states particles

converges only if Er − µ > 0 . We fix E0 = 0 ⇒ BE gas needs µ < 0.


Y 1
⇒ Z=
r
1− e−β(Er −µ)

1 ∂ X 1 X 1
⇒ N= ln Z = β(Er −µ) − 1
=: hnr i ; hnr i =
β ∂µ r
e r
eβ(Er −µ) −1

BE distribution
TD limit ⇒ nr ≈ hnr i

cf. photons, phonons which are bosons!

Def.: “fugacity” ζ ≡ eβµ

BE gas: µ < 0 ⇒ 0 < ζ < 1

Ideal BE gas

~2 k 2
1 particle: E =
2m
 3/2
V 2m √
Recall: # of states in [E, E + de) for one particle: g(E)dE = 2 E dE
4π ~2
Z Z
X g(E)
→ N= hnr i = g(E) hnr i dE = dE = N(µ, T )
r
ζ −1eβE
−1

Comment: in practice N often fixed, e.g. # of He4 ; GrCE chosen for convenience!

⇒ we often need to invert N = N(µ, T ) to µ = µ(T, N) ;


cf. ideal gas in GrCE
Z
E g(E)
Energy: Etot = dE
ζ −1eβE−1
C QUANTUM GASES 46

Φ
pressure: p=−
V
1 1X  
⇒ pV = ln Z = − ln 1 − ζe−βEr
β β r
Z
1  
=− g(E) ln 1 − ζe−βE dE
β | {z } | {z }
u′ v

int. by parts, g(E) ∼ E 1/2


Z
2 E g(E) 2
⇒ pV = −1 βE
dE = Etot
3 ζ e −1 3
Z Z
g(E) E g(E)
Problem: we still need integrals dE, dE
ζ eβE − 1
−1 ζ −1eβE −1

C.5.2 QM gas at high T


Let us first consider the limit ζ = eβµ ≪ 1
Z  3/2 Z ∞
N 1 g(E) 1 2m E 1/2
⇒ = dE = dE
V V ζ −1eβE − 1 4π 2 ~2 0 ζ −1 eβE − 1
 3/2 Z ∞ −βE √

1 2m ζe E
= 2 2 −βE
dE x = βE
4π ~ 0 1 − ζe
 3/2 Z ∞

1 2m ζ √ −x −x

= 2 xe 1 + ζe + . . . dx solve integral with x = u2
4π ~2 β 3/2 0

 
N ζ ζ
⇒ ... ⇒ = 3 1 + √ + ... (∗)
V λ 2 2

What does ζ ≪ 1 mean?


λ3 N
Evidently ≪ 1 ⇒ λ ≪ particle distance ⇒ high T expansion
V

Note: appears surprising: T → ∞ ⇒ β → 0 , so ζ = eβµ should go to 1?


N ζ
No! Change T at constant N ⇒ µ changes: = const ≈ 3
V λ
⇒ ζ ∼ T −3/2 , µ → −∞
C QUANTUM GASES 47

High T EOS of BE gas

Z ∞  3/2 Z ∞
E 1 E g(E) 1 2m E 3/2
x = βE, ζ ≪ 1
= dE = dE
V V 0 ζ −1eβE − 1 4π 2 ~2 0 ζ −1eβE − 1
 3/2 Z ∞
1 2m ζ 3/2 −x −x

= 2 x e 1 + ζe + . . . dx
4π ~2 β 5/2 0
 
3ζ ζ
= ... = 3 1 + √ + ... (∗∗)
2λ β 4 2

Eliminate ζ in (∗) and (∗∗) for ζ ≪ 1 :


 
λ3 N 1 λ3 N
(∗) ⇒ ζ ≈ 1− √ + ...
V 2 2 V
  
E 3 N 1 λ3 N 1 λ3 N
⇒ = 1− √ + ... 1+ √ + ...
V 2 βV 2 2 V 4 2 V
 
! 2 λ3 N
⇒ pV = E = NkB T 1 − √ +...
3 4 2V
| {z }
= B2 (T )

= 2nd virial term; not from interactions but QM statitstics

BE statistics reduce pressure in the high T limit

C.5.3 Bose Einstein condensation


Z ∞   √
3 π
Def.: Gamma function: Γ(n) ≡ un−1 e−u du ; Γ =
0 2 2
Z ∞
1 xn−1
Def.: polylogarithms: gn (z) ≡ dx
Γ(n)
0 z −1 ex − 1
Z Z " ∞
#
1 zxn−1 e−x 1 X
⇒ gn (z) = dx = z xn−1 e−x z m e−mx dx
Γ(n) 1 − ze−x Γ(n) m=0

∞ Z ∞ Z
1 X m n−1 −mx 1 X zm
= z x e dx = un−1e−u du u = mx
Γ(n) m=1 Γ(n) m=1 mn
| {z }
=Γ(n)
C QUANTUM GASES 48


X zm
⇒ gn (z) = monotonically increasing with z
m=1
mn

X 1
Def.: Zeta function ζ̂(s) ≡
n=1
ns
 
3
⇒ gn (1) = ζ̂(n) ; g3/2 = ζ̂ ≈ 2.612
2
 3/2 Z ∞
N 1 2m x1/2 1
Recall: = 2 dx = g3/2 (ζ) (∗)
V 4π ~2 β 0 ζ −1 ex − 1 λ3
N
For fixed : decrease T ⇒ λ increases ⇒ g3/2 (ζ) must increase ⇒ ζ increases
V

Def.: Tc ≡ temperature where (∗) gives ζ = 1:


  2/3 s  1/3
2π~2 1 N 2π~2 V
Tc = ; λc ≡ = g3/2 (1)
kB m g3/2 (1) V kB Tc m N

Note: We shall see that (∗) is not correct, so we do not have ζ = 1 at T = Tc !

Problem: ζ must be < 1 . What happens at T ≤ Tc ?

According to (∗): λ increases ⇒ N decreases,

but particles don’t disappear through cooling!

What’s wrong?
 3/2 Z
X V 2m √
Answer: we used ≈ 2 E dE
k
4π ~2
R P
The ground state E = 0 doesn’t contribute to but should be in

⇒ The “missing” particles are in the ground state:


1 1 ζ
hn0 i = = =
eβ(E0 −µ) −1 ζ −1 −1 1−ζ
1
For most ζ ∈ (0, 1) , hn0 i is small, but for ζ ≈ 1 − , hn0 i ≈ N
N
V ζ
⇒ we must correct (∗) : N= 3
g3/2 (ζ) +
λ 1−ζ
C QUANTUM GASES 49

T < Tc ⇒ ζ close to 1
V
⇒ N≈ g3/2 (1) + hn0 i
λ3
 3  3/2
n0 V λc T
⇒ =1− g3/2 (1) = 1 − =1−
N Nλ3 λ Tc
⇒ at T < Tc a macroscopic number of particles is in the ground state

“BE condensation”

First experimental BE cond. in 1995:

Rb, Na, Li , N ∼ 104 . . . 107 atoms , Tc ∼ 10−7 K

→ 2001 Nobel Prize

Peak: Ground state in momentum space

Z
2 2 E g(E) kB T V
EOS: Recall pV = Etot = dE = g5/2 (ζ)
3 3 ζ −1 eβE − 1 λ3
Comment: The ground-state contribution can be shown to be negligible
kB T
T < Tc ⇒ ζ ≈ 1 ⇒ p = g5/2 (1)
λ3
N
Note: p ∼ T 5/2 and p independent of particle density
V

C.5.4 Heat capacity: A first look at phase transitions


Let us consider CV near Tc
Etot 3 3 kB T
Recall: = p= 3
g5/2 (ζ) ∼ T 5/2
V 2 2 λ
CV 1 ∂Etot 15 kB 3 kB T dg5/2 dζ
⇒ = = g 5/2 (ζ) +
V V ∂T 4 λ3 2 λ3 dζ dT
C QUANTUM GASES 50

Case 1: T < Tc
dζ 15 V kB
ζ≈1 ⇒ ≈ 0 and CV ≈ g5/2 (1)
dT 4 λ3

Case 2: T & Tc

X ζm dgn 1
gn (ζ) = ⇒ = gn−1 (ζ)
m=1
mn dζ ζ

dg5/2 dg5/2
⇒ = ζ −1 g3/2 (ζ) ; as T → Tc : → g3/2 (1)
dζ dζ

We still need
dT
Nλ3
For T > Tc , (∗) is still valid: = g3/2 (ζ)
V
 n0 
−6 6
say ζ = 1 − 10 ⇒ n0 ≈ 10 ⇒ ζ ≈ 1 ∧ ≈0
N
Z ∞
dg3/2 1 1 1 x−1/2
= g1/2 = dx diverges for ζ → 1
dζ ζ ζ Γ(1/2) 0 ζ −1ex − 1
Z ǫ
1 1 x−1/2 x
e ≈1+x
= −1
dx + finite
ζ Γ(1/2) 0 ζ (1 + x) − 1
Z ǫ r
1 x−1/2 x
= dx + . . . u=
Γ(1/2) 0 (1 − ζ) + x 1−ζ
Z ǫ
2 1 1
=√ du + . . .
1 − ζ Γ(1/2) 0 1 + u2
p
⇒ Series expansion: g3/2 (ζ) = g3/2 (1) + A 1 − ζ + . . .

Nλ3 p
⇒ = g3/2 (ζ) ≈ g3/2 (1) + A 1 − ζ
V
 2
1 Nλ3
⇒ ζ ≈ 1 − 2 g3/2 (1) −
A V
 2/3  3/2
2π~2 1 N Tc λ3 N 1
Recall: Tc = ⇒ =
kB m g3/2 (1) V T V g3/2 (1)
C QUANTUM GASES 51

"  3/2 #2  2
[g3/2 (1)]2 Tc T − Tc
⇒ ζ ≈1− 1− ≈1−B T = Tc (1 + ǫ)
A2 T Tc

15 kB V dζ
So for T & Tc : CV = 3
g5/2 (ζ) + b̃ ; b̃ finite > 0
4 λ dT
15 kB V T − Tc
⇒ CV = 3
g5/2 (ζ) − b with b > 0
4 λ Tc

Comments: • The first term smoothly goes over to the result of Case 1) for T < Tc

• The second term goes to zero as T → Tc , but with finite slope

⇒ CV has discontinuous deriv.


 3/2
ζ n0 T
• Recall: hn0 i = , for T < Tc : =1−
1−ζ N Tc
 −1  −1
1 1 1
⇒ ζ = 1+ = 1+ for T < Tc
n0 N [1 − (T /Tc )3/2 ]
dζ 1
⇒ ∼ for T < Tc
dT N CV

whereas = O(1) for T > Tc
dT
High T limit
⇒ true discontinuity and phase transition

only in TD limit

Tc T

Superfluid Helium-4

He4 : bosons

superfluid transition at t = 2.17 K

“λ transition”

superfluidity results from interaction of particles

not the same, but related to non-interacting BE cond.


C QUANTUM GASES 52

C.6 Fermions
non-interacting fermions: e− , in metals, He3 , white dwarfs, neutron stars, . . .

fermions: integer + 21 spin, ψ(~r1 , ~r2 ) = −ψ(~r2 , ~r1 ) ⇒ Pauli exclusion
X
GrCE: Zr = e−βn(Er −µ) = 1 + e−β(Er −µ) : state occupied or not
n=0,1
Y
Z= Zr
r

1 ∂ X 1 X
⇒ N = hNi = ln Z = = hnr i
β ∂µ r
eβ(Er −µ) + 1 r

1
⇒ hnr i = “Fermi-dirac” (FD) disttibution
eβ(Er −µ) +1
µ can be positive or negative; unlike BE!

C.6.1 Ideal Fermi gas


~2 k 2
E= , one particle
2m
degeneracy for spin s : gs = 2s + 1 ; e.g. e− : gs = 2
 3/2
gs V 2m
⇒ g(E) = E 1/2
4π 2 ~2
Z Z
g(E) Eg(E)
N= −1 βE
dE , Etot = dE ,
ζ e +1 ζ eβE + 1
−1

Z
1 1  2
pV = −Φ = ln Z = g(E) ln 1 + ζe−βE dE = Etot int by parts
β β 3
 λ3 N 
high T ⇒ Sec.
. . .5.2 ⇒ pv = NkB T 1+ √ +...
4 2g V
| {z s }
=B2 (T )

virial coeff. B2 (T ) > 0 : QM statistics → increase p


C QUANTUM GASES 53

C.6.2 Degenerate Fermi gas and the Fermi surface


(
1 1 for E < µ
T → 0 ⇒ FD distr. simple : β(E−µ)

e +1 0 for E > µ

⇒ Each Fermion falls to the lowest available state

Def.: Fermi energy EF ≡ µ(T = 0) at fixed N

= energy limit of occupied states at T = 0

EF
Def.: Fermi temperature TF ≡
kB
∼ 104 K for e− in metal ; ∼ 107 K in white dwarfs

p
Momentum space: ~kF = 2mEF kz
ky
All states with |~k| ≤ kF filled

Fermis “sea” or “sphere“ with “surface” |~k| = kF


kx

What is EF (N) ?
Z ∞ Z EF  3/2
g(E) gs V 2m 3/2
T →0 ⇒ N = −1 βE
dE = g(E)dE = 2 2
EF
0 ζ e +1 0 6π ~
 2/3
~2 6π 2 N
⇒ EF =
2m gs V
Z EF
3
⇒ E = hEi = E g(E) dE = NEF
0 5
2
⇒ pV = NEF “degeneracy pressure” 6= 0 even at T = 0
5
important for compact stars
C QUANTUM GASES 54

C.6.3 Fermi gas at low T


Z ∞ Z ∞  2/3
g(E) E g(E) gs V 2m
Recall: N = dE , Etot = dE , g(E) = E 1/2
0 ζ −1eβE + 1 0 ζ −1eβE + 1 4π 2 ~2

Rigorous treatment for small T tricky because n(E) discontinuous at T = 0

Non-rigorous discussion: 1) At small T : n(E) kB T


<--->
the FD distr. only changes near EF
T=0
T>0

2) We assume =0
dT T =0

EF E
dN dµ
Claim: = 0 at T = 0 if = 0 at T = 0.
dT dT
Z ∞ Z ∞  
dN d g(E) d 1
Proof: = dE = g(E) dE
dT dT 0 eβ(E−µ) + 1 0 dT eβ(E−µ) + 1
Z ∞  
∂ 1
≈ g(EF ) dE
0 ∂T eβ(E−µ) + 1
↑ տ
d
1) dT
(FD) ≈ 0 2) no inner deriv. of µ ;
except E ≈ EF µ = EF

Z ∞  
dN E − EF 1
⇒ ≈ g(EF ) 2 2 dE ≈ 0 
dT 0 kB T 4 cosh [β(E − EF )/2]
| {z } | {z }
odd in E−EF even in E−EF

Z ∞  
∂E ∂ 1
Heat capacity: CV = = E g(E) dE
∂T N,V 0 ∂T eβ(E−µ) + 1
3
Taylor: E g(E) ∼ E 3/2 ⇒ E g(E) ≈ EF g(EF ) + g(EF ) (E − EF )
2
Z ∞   
3 ∂ 1
⇒ CV = EF g(EF ) + g(EF ) (E − EF )
0 | {z } |2 {z } ∂T eβ(E−EF ) + 1
even | {z }
odd
odd in (E − EF )
cf. above
C QUANTUM GASES 55

Z ∞

3 x2
⇒ . . . ⇒ CV ≈ g(EF ) T 2 x dx
x = β(E − EF )
2 −∞ 4 cosh 2

Z ∞ Z ∞
Here we extended the integral from . . . dx to . . . dx because
−βEF −∞

contributions far from E = EF are negligible!

⇒ CV ∼ T g(EF )

Interpretation: at low T only particles within ∆E ≈ kB T of EF take part in the physics

Each of these picks up an energy ∼ kB T

There are ∼ g(EF ) kB T such particles ⇒ E ≈ const + g(EF ) (kB T )2

⇒ CV ∼ g(EF ) T
3/2
Recall: N ∼ EF ∼ EF g(EF ) ∼ TF g(EF )
T
⇒ CV ∼ NkB
TF

Comment: More rigorous treatment: Sommerfeld expansion ; cf. Sec. 3.6.4 in D. Tong [2]
π2 T
Lengthier calculation ⇒ CV = NkB
2 TF

Heat capacity of metals

Recall: phonon contribution at low T : CV ∼ T 3

now: FD gas = e− contribution: CV ∼ T

⇒ CV = γT +αT 3

TD3
One can show that the 2 contributions are ≈ equal for T 2 ∼
50 TF
TD3
Typically: TD ∼ 102 K , TF ∼ 104 K ⇒ = O(1) K
50 TF

Note: Surprising that e− are well described by ideal FD gas.

Coulomb interaction? Explained by Landau’s Fermi-liquid theory . . .


C QUANTUM GASES 56

C.6.4 White Dwarfs and the Chandrasekhar limit

When stars exhaust their fuel (H, He, . . . ): T → 0 , degeneracy pressure ; “White Dwarfs”
3 GM 2
constant density approximation ⇒ Egrav =− ; G = Newton’s constant
5 R
minimize Egrav + Ekin ⇒ . . . ⇒ R ∼ M −1/3 (example sheet)

Note: WDs shrink when mass is added!


 2/3
~2 6π 2 N
⇒ EF = increases
2m gs V
⇒ gas becomes relativistic

ultrarelativistic regime with gs = 2:


 
V 2 m2 c4
g(E) = 2 3 3 E − + ... E≫m
π ~ c 2
Z EF  
V 1 4 m2 c4 2
⇒ Ekin = E g(E) dE = 2 3 3 E − EF + . . .
0 π ~c 4 F 4
Z EF  
V 1 3 m2 c4
N= g(E) dE = 2 3 3 E − EF + . . .
0 π ~c 3 F 2
4
White Dwarf mass, volume: M = N mp , V = πR3 , mp = proton mass ,
3
eliminate EF in Ekin , N to leading order
"  1/3 #
3~c 9πM 4 3 1
⇒ . . . ⇒ Egrav + Ekin = 4
− GM 2 + O(R)
4 4mp 5 R

1
Case 1: leading term > 0
R
dEtot 1
⇒ = 0 has a solution: term balances R term
dR R
⇒ star settles into new equilibrium

1
Case 2: leading term r < 0 ⇒ no equilibrium ⇒ neutron star or black hole
R
 3/2
~c 1
This happens if M > Mc ∼ ≈ 1.5 M⊙
G m2p
C QUANTUM GASES 57

C.6.5 Pauli paramagnetism (not lectured)


~
Consider e− gas in magnetic field B

⇒ 2 effects: ~
1) Coupling of spin to B
~
2) Lorentz force ~v × B

Here 1)

e− has “spin up”, s = +1, or “spin down”, s = −1


|e|~
⇒ Espin = µB s B ; µB ≡ “Bohr magneton”
2mc
Z ∞
1 xn−1 !
Def.: fn (ζ) ≡ −1 x
= −gn (−ζ) cf. Sec. C.5.3
Γ(n) 0 ζ e +1

The two spin states now have different energies


 3/2 Z ∞ 1/2
N↑ 1 2m Ekin 1
⇒ = 2 2 β(E +µ B−µ)
dEkin = 3 f3/2 (ζe+βµB B )
V 4π ~ 0 e kin B +1 λ
 3/2 Z ∞ 1/2
N↓ 1 2m Ekin 1
= 2 dEkin = f (ζe−βµB B )
3 3/2
V 4π ~2 0 eβ(Ekin −µB B−µ) + 1 λ

∂E
Def.: Magnetization M ≡
∂B
one e− : Espin = s µB B
µB V   
⇒ M = µB (N↑ − N↓ ) = 3
f3/2 ζeβµB B − f3/2 ζe−βµB B
λ

High T limit: One shows (as for bosons in Sec. C.5.2): ζ → 0


Z ∞ Z ∞ n−1
1 xn−1 ζ x
⇒ fn (ζ) = dx ≈ dx = ζ
Γ(n) 0 ζ −1ex + 1 Γ(n) 0 ex
| {z }
=Γ(n)

2µB V ζ
⇒ M≈ sinh (βµB B)
λ3
2V ζ
likewise: N ≡ N↑ + N↓ ≈ cosh (βµB B)
λ3
⇒ M ≈ µB N tanh (βµB B) = classical result!
C QUANTUM GASES 58

∂M
Def.: magnetic susceptibility χ =
∂B
For small B : tanh (βµB B) ≈ βµB B
Nµ2B
⇒ χ(B = 0) = “Curie’s law”
kB T

(ln ζ)n
Low T limit: One can show for βµ ≫ 1: fn (ζ) =
Γ(n + 1)
 3/2
µB V 2m  3/2 3/2

⇒ M = µB (N↑ − N↓ ) = (E F + µ B B) − (EF − µ B B)
6π 2 ~2
 3/2
µ2B V 2m 1/2
µB B ≪ EF ⇒ M ≈ 2 2
EF B
2π ~
 3/2
gs V 2m √
density of states: g(E) = E ; gs = 2
4π 2 ~2

⇒ M ≈ µ2B g(EF )B
∂M
⇒ χ= = µ2B g(EF ) ∼ const
∂B
Interpretation: The e− deep below the Fermi surface cannot flip their spin because those
states are already occupied. Only e− at the Fermi surface can flip the state.
These e− have density of states g(EF ).

Note: χ > 0 . Such materials are “paramagnetic”

µ2B !
Effect 2) [Lorentz force] ⇒ . . . ⇒ M = − g(EF )B < 0 ; “diamagnetism”; cf. [2].
3
D CLASSICAL THERMODYNAMICS 59

D Classical Thermodynamics
Macroscopic description of systems without regard to microscopic constituents

widely applicable: Black Holes, biological systems, engineering,. . .

D.1 Temperature and the 0th law


Def.: “insulated system”: no influence from outside, enclosed in adiabatic walls

“diathermal system”: enclosed in non-moving walls, heat flow possible

“equilibrium”: no change in time

defining quantities: for now: pressure p, volume V

(sometimes more, e.g. magnetic field, magnetization, . . . )

0th law: If two systems A, B are in equilibrium with a third system C

⇒ A, B are in equilibrium with each other. “transitivity”

Suppose A is in state (p1 , V1 ) and C in (p3 , V3 )

A, C in equilibrium ⇒ given p1 , V1 , p3 fixes V3 = fAC (p1 , V1 ; p3 )

B, C in equilibrium ⇒ V3 = fBC (p2 , V2 ; p3 )

⇒ fAC (p1 , V1 ; p3 ) = fBC (p2 , V2 ; p3 )

Fix p3 at some value and define ΘA (p1 , V1 ) ≡ fAC (p1 , V1 ; p3 )

ΘB (p1 , V1 ) ≡ fBC (p2 , V2 ; p3 )

0th law: A, B are in equilibrium ⇔ ΘA (p1 , V1 ) = ΘB (p2 , V2 )

Def.: Temperature T ≡ Θ(p, V ) “Equation of State”


p
We could choose temperature = Θ(p, V ) or so, but ∃ canonical choice:

Carnot cycle; cf. below


pV
Reference choice for now: ideal gas T =
NkB
D CLASSICAL THERMODYNAMICS 60

D.2 The 1st law


1st law: The amount of work required to change an otherwise isolated system

from state 1 to state 2 is independent of how the work is done.

⇒ There exists a function E(p, V ) , energy, and ∆E = W

Heat: If the system is not otherwise isolated, ∆E 6= W

A change resulting exclusively from T differences is called heat Q

⇒ ∆E = Q + W

Note: We cannot write E = Q + W ; neither Q, W are functions of state.


∂E ∂E
dE = dp + dV is a total derivative.
∂p ∂V
not possible for Q, W

we write dE = dQ
¯ + dW
¯

E.g. specific way to do work: “squeeze” ⇒ dW


¯ = −pdV

meaning of d¯: There exists no function W (p, V ) such that “dW = −pdV ”

Def.: Quasistatic process: a process of E transfer where the system is always

effectively in equilibrium; view as “slow” change

vary system quasistatically from A to B p


Z
I B
⇒ dE = E(p2 , V2 ) − E(p1 , V1 ) ,
R
but W = − pdV depends on path
A II

V
D CLASSICAL THERMODYNAMICS 61

D.3 The 2nd law


Reversible processes

Def.: Reversible process ≡ quasi-static process that can be run backwards: “no friction”
I
For a roundtrip: dE = 0 p
p2
I
But in general: p dV 6= 0
I I
st
1 law ⇒ dQ
¯ = p dV p
1

V1 V2 V

It would be great to run this circle such that heat is transformed into work...

2nd law: Kelvin: No process is possible whose sole effect is to extract heat from

a hot reservoir and convert it entirely into work.

Clausius: No process is possible whose sole effect is the transfer of heat

from a colder to a hotter body.

Keywords: “sole effect”! E.g. fridge uses work to cool a colder system.

Comment: Kelvin ⇔ Clausius !


hot
Imagine a machine in violation of Kelvin’s form.
Q Qh
Use that to drive a fridge. W
not Kelvin fridge
The compound machine violates Clausius form.
Qc
One similarly finds “Kelvin” ⇒ “Clausius”
cold
D CLASSICAL THERMODYNAMICS 62

D.3.1 The Carnot cycle


I I
Recall: for a reversible cycle: ¯ =−
dQ dW
¯

Can we use this to contradict Kelvin?

No! It must do something else: deposit heat into a cold reservoir

Consider Carnot cycle

A) isothermal expansion TH = const , QH into system

B) adiabatic expansion Q = 0 , T, p decrease

C) isothermal contraction TC = const < Th , QC out of system

D) adiabatic contraction Q = 0 , T, p increase

net heat absorbed: QH − QC = W done by the system

W QC
Def.: efficiency η ≡ =1−
QH QH
Kelvin forbids η = 1 ⇔ QC = 0

Carnot’s theorem: Of all engines operating between a hot reservoir H and

a cold one C, a reversible one is the most efficient.

⇒ all reversible engines have the same η = η(TH , TC )


D CLASSICAL THERMODYNAMICS 63

Proof: Consider an irreversible engine “Ivor” and hot


Q’H QH
use it to drive a reversed Carnot W
Ivor Carnot
reversed
⇒ Q′H − QH extracted from H,
Q’C QC
Q′C − QC = Q′H − QH deposited to C cold

Clausius ⇒ Q′H ≥ QH
Q′C Q′H − Q′C QH − QC QH − QC
⇒ ηIvor =1− ′ = ′
= ′
≤ = ηCarnot
QH QH QH QH

Suppose Ivor were reversible. Then likewise we’d show ηIvor ≥ ηCarnot

⇒ A reversible Ivor has the same η as Carnot. The only variables are TH , TC

⇒ η = η(TH , TC )

D.3.2 Thermodynamic temperature scale and ideal gas


0th Law ⇒ ∃ a function Θ(p, V ) whose equality implies thermal equilibrium

What shall we choose? Θ, Θ, . . . ?

Consider 2 Carnot engines A, B


T1> T2
⇒ Q2 = Q1 [1 − η(T1 , T2 )] Q1
A W12
∧ Q3 = Q2 [1 − η(T2 , T3 )] = Q1 [1 − η(T1 , T2 )] [11 − η(T2 , T3 )]
Q2
T2> T3
Now regard AB ≡ A + B as a compount Carnot engine Q2
⇒ Q3 = Q1 [1 − η(T1 , T3 )] B W23
Q3
⇒ 1 − η(T1 , T3 ) = [1 − η(T1 , T2 )] [1 − η(T2 , T3 )] T3
f (T2 )
T2 cancels on rhs. ⇒ 1 − η(T1 , T2 ) =
f (T1 )

T2
Def.: Thermodynamic temperature: Choose T such that η = 1 −
T1
D CLASSICAL THERMODYNAMICS 64

Carnot cycle for ideal gas

pV
ideal gas: T =
kB T
3
energy E = NkB T
2

Note: Without statistical physics (19th century) these are empirical

A → B : isothermal expansion: dT = 0 ⇒ dE = 0 ⇒ dQ ¯ = −dW


Z B Z B Z B Z B
NkB TH VB
⇒ QH = dQ
¯ = −dW = pdV = dV = NkB TH ln
A A A A V VA

! 3
B → C : adiabatic expansion: dQ
¯ = 0 ⇒ dE = −pdV = NkB dT
2
NkB T 3 dV 3 dT
pV = NkB T ⇒ − dV = NkB dT ⇒ − =
V 2 V 2 T
3 p
⇒ − ln V = ln T + const ⇒ p ∼ V −5/3 AX
2
⇒ T V 2/3 = const
X B
2/3
2/3 2/3 TC VB
⇒ TH VB = TC VC ⇒ = 2/3
TH VC X
D X C
V

VD VC
C → D : isothermal compression: QC = −NkB TC ln = NkB TC ln
VC VD
Note: QC > 0 ; heat given away by engine
2/3 2/3
2/3 2/3 TC V ! V
D → A : adiabatic compression: TC VD = TH VA ⇒ = A2/3 = B2/3
TH VD VC

QC TC ln(VC /VD ) TC
Balance: η = 1 − =1− =1−
QH TH ln(VB /VA ) TH

⇒ Thermodynamic T = ideal gas T


D CLASSICAL THERMODYNAMICS 65

D.3.3 Entropy
Notation: Count Q as heat absorbed by the system

Q < 0 ⇒ system releases heat

Set T1 ≡ TH , T2 ≡ TC ⇒ Q1 = QH , Q2 = −QC
2
QC TC X Qi
Carnot cycle ⇒ = ⇒ for any Carnot cycle =0
QH TH i=1
Ti

p
Subdivide Carnot cycle: X A

QAB QCD
Normal cycle ABCD: ⇒ + =0
TH TC
E
X
QGF QEB
Mini cycle EBGF : ⇒ + =0 B
TF G TH X

FX
Bizarre cycle AEF GCD: ⇒ QAB = QAE + QEB X X
G
D C
X
QF G = −QGF V
QAE QF G QCD QAB − QEB QGF QCD
⇒ + + = − + =0
TH TF G TC TH TF G TC

We can approximate any reversible cycle through such subdivisions!


I
dQ
¯
⇒ for any reversible cycle =0
T
p
Z B
dQ
¯ I B
⇒ for any two states A, B: is pathindependent,
A T
as long as the process is reversible
A II

Def.: Fix some reference state O

For any state A with p, V we define the entropy


Z A
dQ
¯
S(p, V ) ≡ as given by a reversible path
O T
D CLASSICAL THERMODYNAMICS 66

dQ
¯
Comment: dS = ⇒ 1st law dE = T dS − p dV
T
⇒ This entropy is the same as that defined in Sec. A

Irreversibility

Q′C QC Q′H − Q′C QH − QC


ηirr = 1 − ′
≤ ηCar = 1 − ⇒ ′

QH QH QH QH

Consider a reversible and an irreversible machine doing the same work:


1 1
W = Q′H − Q′C = QH − QC ′
≤ ⇒
QH QH
   
Q′H Q′C QH QC ′ 1 1 ′ 1 1
⇒ − = − +(QH − QH ) − = (QH − QH ) − ≤0
TH TC TH TC TH TC | {z } TH TC
| {z } ≥0 | {z }
=0 <0
Q′1 Q′2
⇒ + ≤0
T1 T2
I
dQ
¯
Subdivision of Carnot cycle ⇒ ≤ 0 for any path ; “Clausius inequality”
T

Consider two states A, B, path I irrev., path II rev. p


I B
I Z Z
dQ
¯ dQ
¯ dQ
¯
⇒ = − ≤0
T I T II T

Z A II
dQ
¯ dQ
¯
⇒ ≤ S(B) − S(A) ; ≤ dS for irrev. process
I T T
V
Suppose, path I is adiabatic ⇒ dQ
¯ = 0 ⇒ S(B) ≥ S(A)

If path I is also rev. ⇒ S(B) = S(A)

Isolated systems can only evolve to equal or higher entropy → time arrow
D CLASSICAL THERMODYNAMICS 67

D.4 Thermodynamic potentials: Free Energy, Enthalpy


We have many thermodynamic variables: p, V, T, E, S, . . .

We can choose any two to describe the system. Which? Depends . . .

E.g. energy is best expressed as E(S, V ): dE = T dS − pdV

Free energy

good if T = const

F = E − TS ⇒ dF = −SdT − pdV “Legendre trafo”



∂F ∂F
⇒ = −S , = −p
∂T V ∂V T

What’s “free” about F ?

Consider an isothermal process ⇒ dF = −pdV


Z B
⇒ F (B) − F (A) = −p dV = −“work done by system”
A

⇒ F = measure of work free to be done at constant T

Gibbs Free Energy

Def.: Gibbs free energy G ≡ E + pV − T S

Consider system S with fixed pressure, and a reservoir R such that

Vtot = VR + VS = const

⇒ Stot (Etot , Vtot ) = SR (Etot − ES , Vtot − VS ) + SS (ES⊣ , VS )



∂SR ∂SR ∂S p
≈ SR (Etot , Vtot ) − ES − VS + SS (ES , VS ) =
∂Etot ∂Vtot ∂V E T

ES + p VS − T SS
= SR −
T
Stot max if Gibbs free energy G = E + pV − T S = F + pV is min.

dG = −SdT + V dp

with N: dF = −SdT − pdV + µdN, dG = −SdT + V dp + µdN


D CLASSICAL THERMODYNAMICS 68

Comment: G = G(T, p, N) ; only N is extensive ⇒ G(p, T, N) = µ(p, T ) N


∂G G
µ= =
∂N N

Enthalphy

Def.: Enthalpy H ≡ E + pV ⇒ dH = T dS + V dp

E, F, G, H are “thermodynamic potentials”

D.4.1 Maxwell’s relations


Regard E = E(S, V ) ; 1st law: dE = T dS − p dV

∂E ∂E
⇒ =T , = −p
∂S V ∂V S

∂2E ∂2E
Second derivatives commute ⇒ =
∂S ∂V ∂V ∂S

∂p ∂T
⇒ − = “Maxwell Relations”
∂S V ∂V S

mathematically trivial, physically non-obvious!

Play the same game for F : dF = −S dT − p dV



∂S ∂p
⇒ =
∂V T ∂T V

∂S ∂V ∂T ∂V
and for G, H:
∂p T
=−
∂T p
,
∂p S
=
∂S p
→ 4 Maxwell relations

To remember: cross-multiplication always → T S , pV ; 4 ways to construct Eqs.



∂T ∂S ∂p ∂V
keep the conjugate variable constant: , , ,
∂ . S ∂ . T ∂ . V ∂ . p

minus signs: not obvious . . . , sorry


D CLASSICAL THERMODYNAMICS 69

Heat capacity revisited

Taking further derivatives of the Maxwell eqs. , we can derive

further useful relations → example sheet



∂S ∂S
These include for CV = T , Cp = T :
∂T V ∂T p

∂CV ∂ 2 p ∂Cp ∂ 2 V
=T , = −T
∂V T ∂T 2 V ∂p T ∂T 2 p

∂V ∂p
Cp − CV = T
∂T p ∂T V

For ideal gas: Cp − CV = NkB

Cp > CV because more heat needed to increase T at constant

pressure; some goes into work

D.5 The 3rd law


We have only considered changes in entropy so far; no reference S0

3rd law: “Nernst’s postulate”: lim S(T ) = 0


T →0

S
can be relaxed to: → 0 for T → 0 , N → ∞
N
⇒ the ground state entropy must not grow too much with N
Z
CV
Comment: In Sec. A.2.2, we saw: ∆S = dT
T
integral must converge ⇒ CV ∼ T n with n > 0
3
ideal gas: CV = NkB ∼ T 0 violates 3rd law
2
⇒ The 3rd law tells us that the low T world is QM, not classical
E PHASE TRANSITIONS 70

E Phase transitions
Abrupt, discontinuous changes in systems; cf. BE condensate

E.1 Liquid-gas transition p


T = 8/9 TC
kB T a V
Recall van der Waals EOS: p = − 2 ; v= T = TC
v−b v N T = 10/9 TC
a
3 shapes: T > TC : ≈ ideal gas; ignore 2 term
v
8a
T = TC : kB TC = → inflection point
27b
b v
a
T < TC : kB T ≈ realized in range v > b ⇒ local min, max
v

p
At T < TC : for some pressures: > 1 volume possible

dp
VB : > 0 ⇒ unstable
dv T
A B C
expand system → p increases X X X

squeeze system → p decreases


b v
VA : v & b : atoms are as close as they can be → liquid

van der Waals is not supposed to work for

liquids, but let’s ignore that for now

VC : v ≫ b : gas

E.1.1 Phase equilibrium


What’s going on between A and C? Co-existence of gas, liquid

→ chemical equilibrium µliq = µgas


G
Recall: Gibbs free energy G: µ =
N
E PHASE TRANSITIONS 71

Maxwell construction

Clearly, between A, B, p(v) does not behave like van der Waals

At A) µ = µliq

At C) µ = µgas

We now assume that µliq = µgas and from A to C, µ does not change.

What does this imply for p(v)


p
∂µ
along isotherm: dµ = dp
∂p T

∂µ 1 ∂G 1
= = V
∂p T N ∂p N,T N A B C
pliq X X X
Z p
V (p̃, T )
⇒ µ(p, T ) = µliq + dp̃
pliq N b v
V
from A to C we don’t want µ to change. >0
N
⇒ the only way to avoid a change in µ is dp = 0

⇒ horizontal line from A to C in pv diagram

But what is pliq ?

Maxwell: Assume the van der Waals EOS were correct.


Z p
V (p̃, T ) !
⇒ µ(p, T ) = µliq + dp̃ = µgas
|{z} pliq N |{z}
=µA =µC

Z C Z B Z C
!
⇒ V dp = V dp + V dp = 0
A A B

⇒ The shaded areas AB and BC must be equal

⇒ The coexistence isotherm between A and C is p = const = pliq such that AAB = ABC

As we move from A to C, more and more liquid becomes gas.


E PHASE TRANSITIONS 72

E.1.2 The Clausius-Clapeyron Equation


p
Consider the liquid-gas transition in the p − T plane.
liquid
Say, we sit in the gas phase at T < TC very close to the liquid edge. critical
point
Then increase p just a bit.
gas
→ liquid ; V shrinks discontinuously.

→ phase transition TC T

µliq = µgas ⇒ Gliq = Ggas

how does G change along the line of phase transition?


!
dGliq = −Sliq dT + Vliqdp = dGgas = −Sgas dT + Vgas dp
dp Sgas − Sliq
⇒ =
dT Vgas − Vliq

Def.: latent heat L ≡ (Sgas − Sliq ) T


dp L
⇒ Clausius-Clapeyron Eq.: =
dT T (Vgas − Vliq)

Def.: For an nth order phase transition, the nth deriv. of the thermodynamic potential

(typically F or G) is discontinuous.

∂G ∂F
Here: V = , S=− discontinuous
∂p ∂T
⇒ liquid-gas phase transition is 1st order.

As T → TC : Vliq → Vgas .

One can show that at TC we have a 2nd order Ph.Tr.


p
At T > TC no distinction between gas, liquid.

In general, phase diagrams also include a solid phase. solid

liquid

gas

TC T
E PHASE TRANSITIONS 73

E.1.3 The critical point


Critical point: inflection point in p = p(V )
∂p ∂2p 8a
⇒ = 2 = 0 ⇒ . . . ⇒ kB TC =
∂v ∂v 27b

Now consider vdW EOS: pv 3 − (pb + kB T )v 2 + av − ab = 0

T < TC ⇒ 3 roots

T > TC ⇒ 1 root

T = TC all 3 roots coincide

⇒ pC (v − vC )3 = 0
8a a
Compare coeffs. ⇒ kB TC = , vC = 3b , pC =
27b 27b2
The law of corresponding states

vC 27 (kB TC )2
Invert the last relations: b = ; a=
3 64 pC
T v p
Def.: reduced variables: T̄ ≡ , v̄ ≡ , p̄ ≡
TC vC pC
8 T̄ 3
⇒ vdW: p̄ = 1 −
3 v̄ − 3
v̄ 2

pC , TC , vC only depend on 2 vars.: a, b

⇒ eliminate a, b
pC vC 3
⇒ compressibility ratio = should hold for all gases
kB TC 8
pC vC
Experiment: = 0.28 . . . 0.3 . OK, given that vdW is not really good for liquids.
kB TC

Coexistence curve for various gases in reduced

variables is nearly universal: Ne, O2 , CO, CH4 ,. . .

Chemical makeup appears irrelevant → “universal” behaviour


E PHASE TRANSITIONS 74

Critical exponents

1) How does vgas − vliq behave at the critical point?

8 T̄ 3 8T̄ 3 8T̄ 3
vdW: p̄ = 1 − 2
= − 2 = − 2
3 v̄ − 3
v̄ 3v̄liq − 1 v̄liq 3v̄gas − 1 v̄gas

(3v̄liq − 1) (3v̄gas − 1) (v̄liq + v̄gas )


⇒ T̄ = 2 v̄ 2
(∗)
8v̄gas liq

Critical point ⇒ v̄gas , v̄liq → 1 ⇒ T̄ → 1


ǫ ǫ
Expand (∗) in ǫ = v̄gas − v̄liq ⇒ v̄gas = 1 + , v̄liq = 1 −
2 2
1
⇒ . . . ⇒ T̄ ≈ 1 − (v̄gas − v̄liq )2
16
⇒ vgas − vliq ∼ (TC − T )1/2

2) How does v change with p along a critical isotherm?

There exists a unique function p = p(v, TC )


∂p ∂2p
Furthermore = 2 = 0 at critical point
∂v ∂v
Taylor expansion ⇒ p − pC ∼ (v − vC )3


1 ∂v
3) How does the compressibility κ ≡ − change as T → TC from T > TC ?
v ∂p T

∂p
At the critical point: =0
∂v TC

∂p
Taylor expand in T − TC ⇒ = −α(T − TC ) + . . . α = const
∂v T,v=vC

1
⇒κ∼
T − TC

How do these agree with experiment? How good is our vdW based model?
E PHASE TRANSITIONS 75

Answer: Experimental results do not depend on the type of gas. Good!

But the exact scaling is different: vgas − vliq ∼ (TC − T ) ≈0.32

p − pC ∼ (v − vC ) ≈4.8
1
κ∼
(T − TC ) ≈1.2
The exponents are called “critical exponents”

Conclusion: The vdW EOS does a good job qualitatively, but is quantitatively inaccurate.
∆N 2 1
One can show that for T → TC : ∼κ∼ , i.e. diverges.
N T − TC
We work with averaged quantities! Becomes inaccurate for large fluctuations

→ New physical areas...

E.2 The Ising model


N sites in a d-dimensional lattice: each has spin up: ↑, si = +1, or

spin down: ↓, si = −1

N
X
magn. field B ⇒ EB = −B si ; ↑ has lower energy ⇒ ↑ is “favored”
i=1

In contrast to our 2-state spin system from Sec. A.2.3, we use here interaction:
X
EI ≡ −J si sj
hiji

hiji = summation over “nearest-neighbour pairs”

Number of nearest neighbours: q ; d = 1 ⇒ q = 2, d = 2 ⇒ q = 4, d ⇒ q = 2d

We consider J > 0 ⇒ neighbours prefer to be aligned.


  
X X X X
CE: Z = e−βE[si] = exp β J si sj + B si 
{si } {si } hiji i

1 X 1 ∂
Def.: magnetization: m ≡ hsi i = ln Z
N i Nβ ∂B
E PHASE TRANSITIONS 76

E.2.1 Mean-field theory


m is the average spin per particle

⇒ si sj = [(si − m) + m] [(sj − m) + m]

= (si − m)(sj − m) + m(sj − m) + m(si − m) + m2

Mean field theory: Fluctuations in the particle spin are small when summed over hiji .

This is a weaker assumption than assuming small h(si − m)2 i !


X  X
⇒ E = EB + EI = −J m(si + sj ) − m2 − B si
hiji i

1 X
= JNq m2 − (Jqm +B) si
|2 {z } |{z}
(2) i
(1)

(1) Each particle has q nearest neighbours: Nq pairs, but every one is counted twice!
X Nq
⇒ The sum has pairs – try for small N . . .
2
hiji

For periodic boundary conditions this is exact.

For non-periodic BCs, a good approximation for large N.

Nq
(2) Again, we have pairs in the sum.
2
Particles i and j appear equally in the sum si + sj
1
⇒ one sum over i and the factor 2
cancels the factor 2 from the sum si + sj .

1
Comments: • The JNqm2 term in E is merely a constant factor in Z
2
⇒ no effect on physics

• We now have a non-interacting system with Beff = B + Jqm

!N
X X X
− 21 βJN qm2
P
Z=e ... eβBeff i si = eβBeff si
s1 sN si =±1

1 2 N 1 2
⇒ Z = e− 2 βJN qm e−βBeff + eβBeff = e− 2 βJN qm 2N coshN βBeff
E PHASE TRANSITIONS 77

1 ∂
⇒m= ln Z = tanh(βB + βJqm) ; implicit equation for m
Nβ ∂B

1) B = 0: ⇒ m = tanh(βJqm)
1
tanh x ≈ x − x3 ⇒ slope of tanh(βJqm) at m = 0: βJq
3 m
a) βJq < 1 ⇒ only one solution: m = 0
tanh
at high T (small β), temperature randomizes
m
the system ⇒ no average magnetization

b) βJq > 1 ⇒ 3 solutions: m = 0, ± m0


m
m = 0 can be shown to be unstable.
tanh
at low T , interactions win out and align
the spins (either up or down) m

T → 0 : β → ∞ ⇒ tanh becomes Heaviside function

⇒ m0 = ±1

The critical T seperating a), b) is: kB TC = Jq

Note: Magnetization turns off at finite T = TC

2) B 6= 0: a) β → 0 : m = tanh [β(B + Jqm)] ≈ β(B + Jqm)


B
⇒m≈
kB T
Note: now m smoothly decreases to 0 at infinite T

b) low T : m asymptotes to ±1 as T → 0 , but the sign of B

determines the sign of m : sign(m) = sign(B);

the other solution can be shown to be metastable


E PHASE TRANSITIONS 78

Summary m B=0
+1
B = 0: • phase transition at T = TC

• can be shown to be of 2nd order


TC T
as we vary T at B = 0

-1

B 6= 0: • no phase transition as we vary T

• but if we vary B at fixed T < TC , we have m


+1 B>0
a phase transition: m swaps sign
1 ∂ 1 ∂
• m= ln Z = − F
Nβ ∂B N ∂B T

⇒ 1st order phase transition


B<0
-1

E.2.2 Critical exponents


Let’s compare the 1st order phase transition of the Ising model with the liquid-gas one.

Ising: Fix T < TC , vary B from positive to negative (or vice versa)

⇒ 1st order transition: magnetization m jumps

liq.-gas: Fix T < TC , vary pressure across the liquidization value

⇒ 1st order transition: v jumps from vgas to vliq

Critical exponents; cf. gas-liquid


E PHASE TRANSITIONS 79

1) m at B = 0 as T → TC
1
Recall: m = tanh(βB + βJqm) = tanh(βJqm) ; tanh x ≈ x − x3
3
T = TC ⇒ βJq = 1

T . TC ⇒ βJq = 1 + ǫ
(1 + ǫ)3 3 1 + 3ǫ 3
⇒ m = tanh [(1 + ǫ)m] ≈ (1 + ǫ)m − m ≈ (1 + ǫ)m − m
3 3
 
1 + 3ǫ 2 1 √
⇒1≈1+ǫ− m ⇒ + ǫ m2 ≈ ǫ ⇒ m ≈ ± 3ǫ
3 3
Jq Jq Jq 1 Jq
ǫ= − = (TC − T ) ≈ (TC − T )
kB T kB TC kB T TC kB T 2

⇒ m ∼ ±(TC − T )1/2

dm
⇒ ∼ (TC − T )−1/2 → ∞ at TC
dT
cf. vgas − vliq ∼ (TC − T )1/2

2) Fix T = TC , how does m vary as B → 0 ?


 
B
At T = TC : βJq = 1 ⇒ m = tanh +m
Jq
Note: For simplicity we assume now that m grows less than linearly with B

We shall see that this is true.


Expand tanh for small B, m
 3
B 1 B B 1
⇒m≈ +m− +m ≈ + m − m3 + . . .
Jq 3 Jq Jq 3
1 B
⇒ m3 ≈ ⇒ m ∼ B 1/3
3 Jq

cf. vgas − vliq ∼ (p − pC )1/3 along isotherm


E PHASE TRANSITIONS 80


∂m
3) Def.: magnetic susceptibility χ ≡ N ; cf. gas compressibility
∂B T
Fix B = 0; how does χ change as T → TC from T & TC ?
 
Nβ Jq
m = tanh (βB + β Jqm) ⇒ χ = 1+ χ at B = 0
cosh2 (β Jqm) N

T → TC ⇒ β Jq → 1 and m → 0 ⇒ cosh2 (β Jqm) → 1


   
Jq 1 1
⇒ χ = Nβ 1 + χ ⇒ Nβ ≈ χ(1 − β Jq) ∼ χ −
N TC T
1
⇒ . . . ⇒ χ ∼ (T − TC )−1 ; cf. κ ∼ for gas
T − TC

How good are our mean-field predictions? Depends on dims. of lattice...

d = 1 : Wildly wrong: Analytic solution of d = 1 Ising model

⇒ no phase transition; cf. Sec. 5.2.4 in [2]

1 1
d = 2: m0 ∼ (TC − T )α : αan = αmf =
8 2
m ∼ B 1/δ : δan = 15 δmf = 3
7
χ ∼ (T − TC )−γ : γan = γmf = 1
4
d = 3: Numerics ⇒ α ≈ 0.32 , δ ≈ 0.48 , γ ≈ 1.2

Same as van der Waals! Both are wrong in the same way.

Memory of microphysics has been lost...

d = ∞: Mean field theory turns out to be exact.

Explanation: Fluctuations spoil mean-field theory approximation.

small d ⇒ few neighbors ⇒ high fluctuations


E PHASE TRANSITIONS 81

E.3 Landau Theory


Unified way to look at phase transitions: Microphysics arbitrary.

Key variable: Free energy F ; we defined it for equilibrium configurations.

Now take its definition for any configs.

1 1 N
e.g.: Ising model: F = − ln Z = JNqm2 − ln (2 cosh βBeff )
β 2 β
Note: F is a function of m

Equilibrium: F can be shown to be minimal


∂F
⇒ = 0 ⇒ m = tanh(βBeff ) ; cf. Sec. E.2.1
∂m
In Landau theory, m is called an order parameter:

m 6= 0 ⇒ order ; m = 0 ⇒ randomness

E.g. gas-liquid transition: m = vgas − vliq

E.3.1 Second order phase transitions


Here we consider systems with symmetry under m → −m

⇒ Expansion of F in m has only even powers: F (T ; m) = F0 (T ) + a(T ) m2 + b(T ) m4 + . . .


1 1 1
B = 0 Ising model: Use cosh x ≈ 1 + x2 + x4 , ln(1 + y) ≈ y − y 2 , Taylor in m, B
2 4! 2
 
NJq Nβ 3 J 4 q 4 4
⇒ . . . ⇒ F (T ; m) = −NkB T ln 2 + (1 − Jqβ) m2 + m + ...
2 12
∂F
Equilibrium: = 0. Solutions depend on signs of a(T ), b(T ).
∂m
a(T) > 0
F
We assume b(T ) > 0 ; otherwise we’d need m6 terms. a(T) < 0

Consider F (m): a(T ) > 0 ⇒ F has only one extremum.

a(T ) < 0 ⇒ F has three extrema.

-m0 m0
Ising model: a(T ) > 0 ⇔ T > TC ⇒ 1 equilibrium: m = 0 m
a(T ) < 0 ⇔ T < TC ⇒ 2 equilibria: m = ±m0
E PHASE TRANSITIONS 82

If a(T ) is smooth ⇒ Equilibrium changes smoothly from m = 0 to m 6= 0 at T < TC .

⇒ 2nd order phase transition at T = TC where a(TC ) = 0

Once we know the equilibrium value m, plug this into F (T ; m)



∂F ∂S
→ F (T ) → all physical quantities: S = − , CV = T , ...
∂T ∂T V

Critical exponents: assume that near TC : b(T ) ≈ b0 , a(T ) ≈ a0 (T − TC )


r
a0
⇒ . . . ⇒ m0 ≈ ± (TC − T )1/2 for T < TC
2b0

Comments: Landau theory predicts m0 ∼ (TC − T )1/2 for all dims. d of the Ising model

At T < TC the system must choose between m0 , −m0 → symmetry breaking

E.3.2 First order phase transitions


Now consider systems where F also has odd powers of m:

F (T ; m) = F0 (T ) + α(T )m + a(T )m2 + γ(T )m3 + b(T )m4 + . . .

Example: B 6= 0 Ising model: Taylor expansion in m, B:


JNq 2 N N
F (T ; m) = −NkB T ln 2 + m − (B + Jqm)2 + (B + Jqm)4 + . . .
2 2kB T 12(kB T )3
F
We again assume b(T ) > 0 for all T B<0
B>0

Low T : either one min or two min, one max

When α(T ), γ(T ) change sign meta stable

⇒ true ground state changes from m < 0 to m > 0 m


true groundstate

High T : double-well potential is lost


F
⇒ single min in F shifted from m = 0 ; e.g.:

m
REFERENCES 83

References
[1] F. Reif. Fundamentals of statistical and thermal physics. McGraw-Hill, Inc., 1967.
[2] David Tong’s lecture notes on Statistical Physics:
http://www.damtp.cam.ac.uk/user/dt281/statphys.html.
[3] Mehran Kardar’s Lecture Notes:
http://web.mit.edu/kardar/www/teaching/.