1 – Elements of Statistical ensemble – Microcanonical

ensemble
We consider a classical system composed of a large number of N molecules occupying a large
volume V:
N ≈ 1023 molecules
V ≈ 1023 molecular volumes

It is convenient to consider the limiting case:

N⟶∞
V⟶∞
N/V = v specific volume = finite number

Let us consider an isolated system = Energy is conserved

The state of the system is completely defined by the 3N coordinates q 1, q2,……., q3N and
3N momenta p1, p2,…….., p3N.
The dynamics of the system is determined by the hamiltonian

H ( q 1 , q 2 ,..., q 3 N ; p 1 , p 2 ,..., p 3 N )  H ( q , p )

Statistical Physics - Tatek Y. 1

Canonical equation of motion:

H (q, p ) H (q , p )
 q i and   p i Eq. 1
pi qi

with H = T + V kinetic energy + potential energy

1.1 Phase space of a classical system

The state of a given system at any time t may be defined by specifying the instantaneous
positions and momenta of all the particles constituting the system:

q 1 , q 2 ,..., q 3 N and p 1 , p 2 ,..., p3N

The set of coordinates (qi, pi) where i = 1, …, 3N may be regarded as a point in a space of 6N
dimensions: the phase space.
The phase point (qi, pi) represents a given state of the system.
qi and pi are functions of time t, their variation is determined by eq. (1).
As time changes (qi, pi) undergoes a continual change and curves out a trajectory in the phase
space.

Statistical Physics - Tatek Y. 2

 The trajectory must remain within a limited region of
p the phase space because:
i) the finite volume V of the system limits the values
of the qi and,
ii) the finite value of the energy E limits both qi and
pi.

In particular if the total energy E of the system has a

precise value, the corresponding trajectory will be
restricted to the hypersurface of the phase space:

q H (qi , pi )  E

If the energy lies anywhere between E and E + D, the corresponding trajectory will be restricted
to the hypershell defined by these limits.

Let us consider an ensemble of systems made of:

the given system + large number of mental copies of it
- The various members of the ensemble will be in different states
- In the phase space, the corresponding picture will consist of a swarm of representative points
all lying within the allowed region of the phase space.

Statistical Physics - Tatek Y. 3

 Let us define (q, p; t) = density function
p It gives the number of allowed states per unit volume of
the phase space for a given system.
≠0
Volume element of the phase space:

=0
p  d q 1 d q 2 ... d q 3 N d p 1 d p 2 ... d p 3 N
3N 3N
d qd

q
Number of representative points within the element of volume = (q, p; t) d3Nqd3Np
Ensemble average of a given physical quantity of the sytem f(q,p):

 f (q , p )  (q , p; t )d
3N 3N
qd p
 f  
  (q , p ; t )d
3N 3N
qd p

The integral extends all over the phase space, but only the populated regions of the phase space
( ≠ 0) really contribute.

Statistical Physics - Tatek Y. 4

 
Stationary ensemble, density of states does not depend explicitly on time:  0
t
⟶ for such case ⟨f⟩ is independent of time.

1.2 Motion of the swarm of representative points: Liouville’s theorem

Let us consider an arbitrary volume w in the relevant region of the phase space. Let  be the
surface enclosing w.
dw = number of states within dw

v  d w  total number of states in w

w
n̂
w The rate at which the total number of states changes is:


d  dw
t w

 d ≡ surface element

n̂  unit vector normal to d

Statistical Physics - Tatek Y. 5
The net rate at which the points flow out of w across  is given by:
 
  ( v . nˆ ) d  where v  ( q i , p
i)


v  velocity vector of the points

 (  .F )d V   ( F .nˆ ) d S
Divergence theorem:
V S

 
   ( v . nˆ ) d     .(  v ) d w
 w


3N
   
where  .(  v )      q i     p i  eq. 3
i 1  qi pi 
The total number of points in the phase space is conserved: the rate at which the number of
states changes + the net rate at which the points flow out of w = 0

 
  dw     .(  v ) d w
t w w Physics - Tatek Y.
Statistical 6
    
     .(  v )  d w  0 eq. 4
w  t 
  Equation of continuity for the swarm of
   .(  v )  0
representative points.
t
Using eq. 3 and eq. 4, we obtain:


3N
    3N
  q i i 
p
t
   q i  i  
p      0 eq. 5
i 1  qi pi  i 1  qi pi 

The last term of eq. 5 vanishes because of:

 q i  H (qi , pi )  H (qi , pi ) p
2 2
i
   
qi qipi piqi pi

Since (qi, pi; t), the remaining term in eq. 5 gives the total time derivative of :

d 
3N
   

t
   q i  i
p
dt i 1   Statistical
q i Physics - pi
Tatek Y.  7
Using eq. 1, we get

d 
3N
  H  H 

t
    
dt i 1  qi pi pi qi 
Poisson Bracket
 , H 
And finally
d  Liouville theorem
   , H  0    time constant
dt t eq. 6

The swarm of representative points moves in the phase space like an incompressible fluid.

Since we are interested by the equilibrium state, we restrict our consideration to ensembles
whose density function does not depend explicitly on time, i.e    0 and eq. 6 becomes:
t
 , H   0
3N
  H  H 
      0 Eq. 7
i 1   q i  p i Statistical
 Physics
pi  q
- TatekiY.  8
• One way to satisfy eq. 7 is to have (q, p) = constant
= = >  is uniformly distributed over the relevant region of the phase space, and the average
value of a physical quantity becomes:

 f (q, p ) dw   f (q, p ) dw
w w
 f   
  dw   dw
w w

1
  f    f (q, p ) dw
w w

Any member of the ensemble has equal probability to be in any state among the possible
states.
= = > equal probabilities ensemble = microcanonical ensemble

• A more general way to satisfy eq. 7 is by having a no explicit dependence of  on q and p, but
through the hamiltonian H(q, p):
 ( q , p )   ( H ( q , p )) Eq. 8

 H (q , p ) 
Example:  ( q , p ) ~ exp   
canonical ensemble
 Statistical kT
Physics - Tatek 
Y. 9
1.3 Microcanonical ensemble

The system is defined by the number of molecules N, the volume V and the energy which lies
between E – D/2 and E + D/2.
The representative points in the phase space can be anywhere within the hypershell defined by
the condition:
D D
E   H (q, p )  E 
2 2
The volume of the phase space within the shell is

 dw  
3N 3N
d qd p
shell shell

A microcanonical ensemble satisfies the following conditions:

D D
 ( q , p )  constant if E   H (q, p )  E 
2 2
  0 otherw ise

The number of states (or points) lying in a volume element dw inside the shell is:
d (E )   (q, p ) dw ~ dw

 (E )    ( q , p ) d w   ( q , p )w ~ w
Statistical Physics - Tatek Y. 10
w
The entropy is defined by
S ( E , V )  k log  ( E ) Eq. 9

where k is Boltzmann’s constant.

= = > The entropy is proportional to the logarithm of the number of states with energy E.

S(E, V) in eq. 9 should possess all the properties of the entropy function in thermodynamics:
i) S is an extensive quantity: If a system is composed of 2 subsystems:
Stot = S1 + S2 (show it)
ii) S should satisfy the second law of thermodynamics (prove it)

Volume of a “single state” (volume of a representative point) w  h

3N
0

w w h  Planck' s constant
w0   (E ) 
(E ) w0

 w 
 S  k log  
 
 w 0 

1.4 Examples of a microcanonical ensemble

i) Ideal gas of monoatomic particles

Statistical Physics - Tatek Y. 11
 N
1
H  
2
Hamiltonian pi (particles are not interacting)
2m i 1
D D
The gas is in a volume V, and has an energy E: E   H (q, p )  E 
2 2
The volume of its phase space is:
w   p   
3N 3N 3N 3N
d qd d q d p
shell V H

We can then express it as VN

w  V  (E )
N

 (E )  
3N
with d p
N 2
D pi D
E 
2
  2m
E 
2
i 1

D D
Let us define E '  E  and E '' E 
2 2

 (E )  
3N
We then have d p
N


' 2 ''

2 mE Statistical i  2 -mE
pPhysics Tatek Y. 12
i 1
(E) represents the volume of a 3N dimensional hypershell bounded by two hyperspheres of
radius:
R' 2 mE ' and R '' 2 mE ' '

Volume of the hyperspheres:

 2 mE ' 
3N /2
 (E ')   p  C  C
3N 3N
d N
R' N
N


2
2 mE '  pi
i 1

 2 mE ' '
3N /2
 (E '')   p  C  C
3N 3N
d N
R '' N
N


2
p i  2 mE ' '
i 1

where CN is given by
3N 3N

 2
 2

C N
   = gamma function
 3N   3N 
  1  !
 2   2 
Statistical Physics - Tatek Y. 13
For D << E we can make the following approximation
3N /2 3N /2
 D  /2  D 
 2 mE 
3N
 (E ')  C N  2 m ( E  )  C N  1  
 2   2E 

 3N D 
 2 mE 
3N /2
 C N  1  
 4 E 
Similarly,
 3N D 
 2 mE 
3N /2
 (E '')  C N 1  
 4 E 
We obtain therefore

 3N D 
 2 mE 
3N /2
 (E )   (E '')   (E ')  C N  
 2 E 
3N
3N

 2 3N
 3N D   2  mE  2  D 
  2 mE  2     
 3N   2 E   3 N   E 
 !   1 !
 2   2 

Statistical Physics - Tatek Y. 14

The volume of the relevant part of the phase space then becomes
3N

 D V  2  mE  2
N

w   
 E   3N 
  1 !
 2 
The number of states is
3N

w  D  V
N
 2  mE  2
(E )     3N
w0  E  h  3N 
  1 !
Entropy of the ideal gas  2 

 D V 3N  3N  
S ( E , V )  k  log  N log 3
 log  2  mE   log   1 !
 E h 2  2  
Since N is assumed to be large, let us use the Stirling’s formula:
log M !  M log M  M for large M
hence  
 3 
D  V  2  mE  2  3 Nk
S ( E , V )  k log  Nk log  3 3  
E h 2
  3N 2 
   
Statistical Physics - Tatek Y.  2 
  15
Finally after rearrangements
 
3

 D 3N  V  4  mE  2
S ( E , V )  k  log    Nk log    
 
3
 E 2  h  3N 
 
Let us consider a case where the energy of the molecules is less than E  0  H  E
3N

 2

 2 mE   2 mE 
3N /2 3N /2
 (E )  C N

 3N 
 !
3N  2 

N
and V 2

w  V  2 mE 
3N /2
 (E ) 
N
volume of the phase space
 3N 
 !
 2 
Number of states
 2  mE 
N
w
3N /2
 V 
(E )  3N
  3 
h  h   3N 
 !
 2 
Entropy 
3

V  4  mE  2 3 Nk
S ( E , V )  k log  ( E )  Nk log     
h  3N 
3
 2
Statistical Physics - Tatek Y.   16
Temperature
1  S  3 Nk 2E
     kT 
T   E V ,N
2 E 3N

Equation of state
P  S  Nk
     PV  NkT
T  V  E ,N V

ii) Harmonic oscillator

2
1 p
H (q, p )  
2
Hamiltonian (classical expression) kq
2 2m
k = spring constant; m = mass of the oscillating particle

Equations of motion

q  A cos w t    and p   m w A sin w t    Eq. 10

w = angular frequency = (k/m)1/2; A = amplitude

1
E  mw
2 2
Energy of the oscillator A
2 Statistical Physics - Tatek Y. 17
Phase space trajectory of the representative points (q, p)

Eliminate t in eq. 10:

2 2 2 2
q p q p
  1    1
m w
2 2 2 2
A A 2E 2 mE
mw
2

2 E
The trajectory is an ellipse with axis ~ E1/2 and area ~ E Area 
w
D D
If the oscillator energy is restricted, i.e. E   H (q, p )  E 
2 2
= = > the representative points are E+D/2
confined in the region
between the 2 ellipses

E-D/2
Volume of the phase space

 D   D 
2  E   2  E  
 2   2  2 D
   ... d q d p 
w

w

w
D D Statistical Physics - Tatek Y. 18
E   H ( q , p ) E 
2 2
In quantum mechanics, the energy level of an oscillator is given by:
 1 
En  n  w n  0 ,1 , 2 ,...
 2 

For a given level n, the trajectory of the representative points is as follow:

p
n=4
n=3
n=2
q
n=1

n=0

Area of the phase space between two consecutive trajectories: D  E n  1  E n   w

   2    h = volume of a single state for one oscillator
D
Number of states for E  D   w :  
w
For N oscillators, the volume of a single state is equal to hN
Statistical Physics - Tatek Y. 19
 2 – Canonical ensemble and grand canonical ensemble

Canonical ensemble = system in thermal equilibrium with a larger sytem

large sytem = heat reservoir

Nr particles, H(qr, pr), Er

subsystem
N particles, H(q, p), E

Nr >> N, Er >> E

Total energy: E0 < Er + E < Er + 2D

Volume occupied by the reservoir in the phase space = wr(Er)

Number of possible states within the volume = r(Er)

The probability of finding the subsystem in a state within dq dp regardless of the state of the
reservoir:
d    ( q , p )d q d p ~ r ( E r )d q d p
Statistical Physics - Tatek Y. 20
   ( q , p ) ~ r ( E 0  E ) (Er  E0  E )

 S r ( E 0  E ) ~ k ln  r ( E 0  E ) Entropy of the reservoir

S r (E r )
Since E  E 0 Sr (E0  E )  Sr (E0 )  E
E r
Er  E0

E
Sr (E0  E )  Sr (E0 )  T = temperature of the reservoir
T

 Sr (E0 )   E 
  r ( E 0  E )  exp exp   
 
 k   kT 
First factor is independent of E ==> constant for the subsystem

 E   H (q, p ) 
  ( q , p ) ~ exp     exp    canonical ensemble
 kT   kT 
density of states for
the subsystem Statistical Physics - Tatek Y. 21
Total number of states for the subsystem:
3N 3N
d qd p 1
Q N
(V , T )   3N
exp   H (q, p )  
N !h kT
Because the particles are identical Eq. 1
QN(V,T) = partition function

The partition function can be expressed in terms of the Helmoltz free energy A(V,T)
Q N (V , T )  exp    A (V ,T ) Eq. 2
In order to justify the above identity, we need to prove that:
a) A is an extensive quantity
b) A is related to the internal energy U ≡⟨H⟩ and the entropy S = (-∂A/∂T)V by the
thermodynamic relation A = U – TS
Proof
a) If we have 2 systems with N1 and N2 particles and N = N1 + N2
  ( A1  A 2 )  A
QN  QN QN  e  e  A  A1  A 2
1 2

b)  A 
A  U  TS  A   H   TS   H   T  
 T V

 A 
 H   A  T   Statistical Physics - Tatek Y. 22
 T V
Then using eq. 1 and 2 we get

 d
3N
qd
3N
p exp   A (V ,T )  H (q, p )  N !h
3N


  d
3N
qd
3N
p exp   A (V ,T )  H (q, p )  0


  A ( V ,T )  H ( q , p ) 
  A  
 d        0
3N 3N
qd pe  A ( V , T ) H ( q , p )
   V 

 A 
 A (V , T )  U (V , T )  T    0
  T V

which proves the relation A  U  TS

Other thermodynamic functions

 A   A 
P    ; S    ; G  A  PV
 V T  T V

 A     A  (A /T )
U  A T   T  
2
    
  T V ,N  T  T V ,N   (1 / T ) V ,N
Statistical Physics - Tatek Y. 23
Specific heat at constant volume
 U    A 
2

CV     T  
 
  T V  T
2
,N V ,N

2.1 Application to the ideal gas

1  H (q, p )
Partition function: Q N (V , T )  3N  e dw
N !h
dw  d
3N 3N
where qd
N
1
Hamiltonian of the ideal gas H ( q , p )  
2
pi
2m i 1
Partition function 

2 N


 d 
1 pi
(V , T )  
2m 3 3 3
Q N 3N
e qid pi
N !h i 1

Integration over space coordinates yields VN

Integration over momentum coordinates:

 
V
N
 
N
pi
2

(V , T )  3N    e 2m 
3 3
QN d pi
N !h  
0 i 
Statistical Physics
1  Y.
- Tatek  24
 
Differential of a vector of constant magnitude p i  d p i  4 p i d p i
3 2
d
N
  p 
2
N
V
 Q N (V , T )   e ( 4 p d p ) 
2 mkT 2
3N
N !h  
0 
 1/ 2
 x
2 1  
Using
 x dx 
2
e  
4 
3
0
we obtain  p
2

( 4 p d p )   2  mkT 
3/2


2 mkT 2
e
0

which gives V 1 
N

 2  mkT 
3/2
QN (V , T ) 
 3 
N!h 

Helmoltz free energy


  N h
3
 

A ( N , V , T )  NkT  log    1 
 2  mkT 
3/2

 V  


Statistical Physics - Tatek Y. 25

Thermodynamics of the ideal gas
 A   N h
3

Chemical potential      kT log  
 2  mkT 
3/2
  N V ,T V 

 A  NkT
Pressure P    
  V  N ,T V

 V  2  mkT   5
3/2
 A   
Entropy S     Nk  log    
 T V
3
,N 
  N h  2


   A  3
Internal energy U   T 2
     A  TS  NkT
 T  T   N ,T
2

Remark: The partition function of a system of N particles can be written in terms of the partition
function for a single particle
N 3
Q1 V
Q N
 where Q1  3
 2  mkT  2
N! h

This is possible because the particles areStatistical

not interacting.
Physics - Tatek Y. 26
2.2 Energy fluctuation in the canonical ensemble

Mean square fluctuation in the canonical ensemble   U  H  

2

Definition of the average energy

 H  (q , p )d q dp
U 
  (q , p )d q d p
When applied to the canonical ensemble
 H

U 
 He dqdp
H
 e dqdp

After rearranging the equation, we obtain:

 U  H e
 H
dqdp  0

We then divide both sides of the equation by the partition function as defined in eq. 2

 U  H e
  A ( V ,T )  H ( q , p ) 
dqdp  0

Statistical Physics - Tatek Y. 27

We differentiate the equation with respect to :
 U  
 U  H    A e   A ( V ,T )  H ( q , p )  d q d p
  A (V ,T )  H ( q , p ) 
  e  H  0
   

A A
 A  H    A  H  T  U  H
 T
We finally obtain
 H
U 
2
U

 e  H dqdp
 0
H

e dqdp

U
  U  H 
2
   0

The mean square fluctuation of energy is then:

U U
 U  H 
2
  H   H     kT
2 2 2

 T

C V   D H 
2
 H   H   kT 
2 2 2

Statistical Physics - Tatek Y. 28

The relative root-mean-square fluctuation is
1

 D H 
2 2
 kT CV N 2
1
 ~ ~ 1
H  U N 2
N
For N - - - > ∞, the relative RMS fluctuation of H is negligible.
This implies that a system in the canonical ensemble has an energy almost equal to the mean
energy = = => canonical ensembles are equivalent to microcanonical ensembles.

2.3 A system of harmonic oscillators

Two important applications:

i) Black-body radiation
ii) Lattice vibration
1 1
H (qi , pi )  mw qi 
2 2 2
pi
Classical hamiltonian of the 2 2m
ith oscillator:
i  1 , 2 ,...., N

Partition function for a single one-dimensional oscillator

  1 1  dqdp
Q1 (  )   exp     mw 
2 2 2
q p 
  2 2m  h
Statistical Physics - Tatek Y. 29
The previous integral is known 
1

 x
2 1   2
e dx  
2  

0

Therefore, we get 1 1

1  2  2
 2 m 2 1
Q1 (  )  
 
 
 
 
h  mw    w
2
 

The partition function for N oscillators is thus

  w 
N
Q N (  )  Q1 
N

w
Helmoltz free energy A   kT log Q N  NkT log
kT
w
Chemical potential   kT log
kT
A
Pressure P    0
V

A  kT 
Entropy S    NkT log 1

T w Y. 
 Physics- Tatek
Statistical 30
Average energy
U  A  TS  NkT

 U   U 
Heat capacity C P
    C V
    Nk
 T  P   T V

Quantum mechanical case

Energy level of a one-dimensional harmonic oscillator

 1 
 n
 n  w n  0 ,1 , 2
 2 

 
    n  1 / 2  w
Partition function Q1 ( )   e n
  e
n0 n0
geometric series of
  w
ratio e
1
  w
2
e 1 1
 Q1 (  )    w
 1 1

1 e  w   w  1 
e 2
 e 2 2 sinh   w 
Statistical Physics - Tatek Y.
 2  31
For N oscillators  N N
  1 
1  e 
  w N
  w
Q N
(  )   2 sinh   w   e 2

  2 
Helmoltz free energy
  1  1 
A  NkT log  2 sinh    w    NkT    w  log 1  e
 w

 
  2  2 

 w
 N

 kT log 1  e 
 w

 2 
 A  A
Chemical potential     
  N  V ,T N

 A 
Pressure P     0
 V  N ,T
Entropy

 A  1  1    1  
S     Nk    w coth    w   log  2 sin    w   
 T  N ,V 2  2    2  

  w   w 
 Nk
  w
 log 1  e  
 - Tatek Y.
e 1 
Statistical Physics 32
Internal energy    A  1  1 
U  T     N  w coth   w 
2

 T  T   N ,V
2  2 

1 w 
 N  w   w
 
2 e  1

Specific heats
 U   U 
C P
    CV   
 T  P   T V
2 w
 1   1  e
  w  cosech    w   Nk   w 
2
 Nk 
e  1
w 2
 2   2 

2.4 Grand canonical ensemble

Grand canonical ensemble = a system for which the number of particles is not fixed, the
average number of particles is determined by conditions
external to the system.

The density of states in the phase space is given by (q, p, N).

Statistical Physics - Tatek Y. 33
In order to find (q, p, N), let us consider the following system:
Small system : N particles and volume V

Large system = heat and particles reservoir:

Nr particles and volume Vr

Subsystem:
Particles : N’ = Nr – N
Volume: V’ = Vr - V

Total hamiltonian H ( q , p , N )  H ( q , p , N )  H ( q ' , p ' , N ' )  H

r r int

Partition function of the whole system interaction term between

the 2 subsystems
dqdp  H (qr , pr ,N ≈ neglibible

)
Q N
(V , T )  3N
e r

r r
h N r!

We can express the total partition function in terms of the 2 subsytems variables
N r
1 N r!  H ( q , p , N )  H ( q ', p ', N ' ) 
Q N (V , T ) 
r
h
3N r
N r!
 dpdp'  N ! N '!
V
dq 
V '
d q 'e
N 0

number of ways of getting N out of Nr

Statistical Physics - Tatek Y. 34
The expression can be further written as
N r
1  H (q, p,N )
Q N (V , T ) 
r
 h
3N
N!
 d p  d qe
V
N 0

1   H ( q ', p ', N ' )


h
3N '
N '!
 d p '
V '
d q 'e

The relative probability (q, p, N) that there are N particles in V with coordinates (q, p) is
equal to
 H (q, p,N )
1 e 1   H ( q ', p ', N ' )
 (q, p, N )  3N 3N '  d p ' d q 'e
V '
Q N (V , T ) h N! h N '!
r

and fulfilling the normalization condition

N r

  d qd p  ( q , p, N )  1
N 0

We can therefore express  as

 H (q, p,N )
Q N ' (V ' , T ) e
 (q, p, N )  3N
Q N (V , T ) h N!
r

Statistical Physics - Tatek Y. 35

We can express the partition function in terms of the Helmoltz free energy

Q N
 exp   A ( N r
,V r , T ) and Q N '
 exp   A ( N ' , V ' , T )
r

Q N ' (V ' , T )
  exp    A ( N r
 N ,V r  V , T )  A ( N r ,V r , T ) Eq. 1
Q N (V r , T )
r

Since Nr >> N and Vr >> V, we can use the approximation

 A   A 
A(N  N ,V r  V , T )  A ( N r ,V r , T )  N  V 
r  
 N ' N ' N r
 V 'V ' V r

 A ( N r , V r , T )  N   VP

 A(N r
 N , V r  V , T )  A ( N r , V r , T )   N   VP

Let us define the fugacity: z  e 

From eq. 1, we get

  N   VP   H (q, p,N )
e e
 (q, p, N )  3N
h N!

Statistical Physics - Tatek Y. 36

 N
z   VP  H (q, p,N )
  (q, p, N )  3N
e Eq. 2
h N!
Thermodynamic functions


Q ( z ,V , T )  
N
Grand partition function z Q N
(V , T )
N 0

For general case where 0 ≤ N < ∞

Let us integrate eq. (2) on both sides:

  N
z   VP 
   
 H (q, p,N )
 (q , p , N )d q d p  3N
e dqdp  1
n 1 N 1 h N!

  H (q, p,N ) 
  PV e   PV
   dqdp    1
N N
z e 3N
z e Q N
N 1 h N! N 0


  PV   PV
 e   e Q ( z ,V , T )  1
N
z Q N
N 0

PV
  log Q ( z , V , T )
kT
Statistical Physics - Tatek Y. 37
Average number of particles 


N
Nz Q N
N 0
 log Q ( z , V , T )
N  N   
 z
z Eq. 3

N
z Q N
N 0

Eq. (3) is justified by: 



N 1
Q ( z ,V , T ) Nz Q N
 log Q ( z , V , T ) z N 0
  
z Q ( z ,V , T )

N
z Q N
N 0
Internal energy: 


N
Hz Q N
N 0
 log Q ( z , V , T )
U  H   
 


N
z Q N
N 0

Specific heat
 U     log Q ( z , V , T ) 
CV       
  T V T   

Statistical Physics - Tatek Y. 38

Entropy T
CV
S   dT
T
and A  U  TS
0

2.5 Density fluctuation in the grand canonical ensemble

Mean square fluctuation:

   log Q ( z , V , T )  Prove this
 D N 
2
  N   N   z
2 2
z
  relation
z  z 

Using the following identities

d z  d e 
 PV
z d and log Q ( z , V , T ) 
kT
we obtain
    PV   z   z   PV  
N   N   z  
2 2
z    
 z   z  kT   z     z     kT  

  P
2 2
1  PV 
 N   N     kTV
2 2

  
2 2 2
 kT 

Statistical Physics - Tatek Y. 39

Let us express the fluctuation in terms of other quantities

The Helmoltz free energy is an extensive quantity:

V
A ( N , V , T )  Na ( v ) v 
N
Chemical potential
 A  a (v )
     a ( v )  N
 N V N
Since we have
V dv
N   dN  V 2
v v
a (v ) a (v )
2
V v
   a (v )   a (v )  v
v V v v

Pressure

 A  a (v ) a (v )
P     N   because dV  N dv
 V  N V v

Statistical Physics - Tatek Y. 40

Let us use the following identity:

 a (v ) a (v )  a (v )  a (v )
2 2

   v  v
v v v v v
2 2

 a (v )
2


P P / v v
2 1
   
 a (v )
2
  / v v
 v
v
2

We have therefore
 P  1 v
2
 1  1 1 1 1
       
   v  v  v  a (v ) / v v P / v
2 2 3 2 2 3

And finally
 P T
2

N   N   kTV  N kT V  Nv
2 2
and

2
v

1
with T  = isothermal compressibility
v   P /  v 

Statistical Physics - Tatek Y. 41

Small density fluctuation case

The probability that a system in the grand canonical ensemble has N particles is proportional to:

W (N )  z ( V , T )  exp   N  A ( N , V , T ) 
N
Q N

When the density fluctuations are small, W(N) is peaked around N  N

In the case of small fluctuations A ( N , V , T )  A ( N , V , T ) and we can calculate A

from Q
A ( N , V , T )  kT N log z  kT log Q ( z , V , T )
A
(Q  z QN  z
N N
e )

2.6 Examples

Classical ideal gas

We can express the partition function
Q 1 
N
Partition function Q (V , T )  this way because the particles are not
N
N! interacting.

N! comes from the fact that the particles are indistinguishable. Indistinguishability is due to the
fact that they are identical and non-localized.
Statistical Physics - Tatek Y. 42
Partition function for a single particle
1 H (q, p )
Q 1 (V , T ) 
h
3  e d q d p  Vf ( T )

where f(T) is only a function of temperature.

The grand canonical partition function is then

 
 zVf (T ) 
N

Q ( z ,V , T )   z
N
Q N (V , T )    exp  zVf (T ) 
N 0 N 0 N!

Pressure kT kT
P  log Q ( z , V , T )  zVf ( T )  zkTf ( T ) Eq. 1
V V
Number of particles

N  z log Q ( z , V , T )   zVf ( T ) Eq. 2
z
Average energy
    f (T ) d f (T )
U   log Q ( z , V , T )    zV  zVkT
2
Eq. 3
   z ,V  dT
Specific heat
 U 
2
d f (T ) d f (T )
    zVkT
2
CV  2 zVkT
  T V
2
dT
Statistical Physics - Tatek Y.
dT 43
Helmoltz free energy
A  N   PV   zVf ( T )  zkTf ( T )V  kT log zVf ( T )  zkTf (T )

 A  NkT log z  zVkTf (T )

U  A 1  d f (T ) 
Entropy S   zVkT
2
 NkT log z  zVkTf (T )
 
T T  dT 

 d f (T ) 
 S   Nk log z  zVk T  f (T )
 
 dT 
Eliminating f(T) in eq (1) and eq (2) yields the equation of state

PV  NkT
U can be expressed differently by eliminating z in eq (2) and eq (3)
2
NkT d f (T )
U 
f (T ) dT
Specific heat
 U   f ' ' (T ) f ' (T ) 
2
f ' (T )
    2 NkT  NkT 
2
CV  
  T V
2
f (T )  f ( T ) f ( T ) 


Nk
2
2 Tf (T ) f ' (T )  T
2
f (T ) f ' ' (T )  f ' (T )
2

f (T ) Statistical Physics - Tatek Y. 44
In simple cases n
f (T ) ~ T
which gives n 1
nT
U  NkT  n ( NkT )
2
n
T

CV 
Nk
2n
2 nT 2n
 n ( n  1)T
2n
 n T
2 2
 nNk
T

For non-relativistic gas: n = 3/2 ; for relativistic gas : n = 3

System of independent localized particles (e.g. harmonic oscillators)

Particles are distinguishable, and the partition function is

Q N (V , T )  Q 1 (V , T ) 
N

Because of the localized nature of the particles, Q1(V,T) is independent of the volume occupied
by the system:
Q 1 (V , T )   ( T )
Grand partition function:
 N
1
Q ( z ,V , T )    z  (T )  
1  z  (T ) Note: Q0 = 1
N 0

Statistical Physics - Tatek Y. 45

For Q(z, V, T) to be convergent, z(T) < 1

Pressure kT kT
P  log Q ( z , V , T )   log 1  z  ( T ) 
V V
Number of particles
  z  (T )
N  z log Q ( z ,V , T )   z log 1  z  ( T )   Eq. 4
z z 1  z  (T )
Energy
 z  ' (T )
U  kT
2
log Q ( z , V , T )   kT
2 Eq. 5
T 1  z  (T )
Free energy
A  N   PV  NkT log z  kT log 1  z  ( T )  Eq. 6

Entropy

U  A kTz  ' ( T )
S    Nk log z  k log 1  z  ( T )  Eq. 7
T 1  z  (T )

Statistical Physics - Tatek Y. 46

From eq. (4) we get:
N 1  z  ( T )   z  ( T )

 ( N  1) z  (T )  N
N 1 1
 z  (T )    1 for N  1
N 1 1 N
1
N
1
 1  z  (T ) 
N
U
U  kT z  ' (T ) N   kT z  ' (T )
2 2
Using eq. (5)
N
Since N 1
z  (T )   1 z 
N 1  (T )
U  ' (T )
  kT
2

N  (T )
Using eq. (6)
A   NkT log  ( T )  kT log N

A  log N 
   kT log  ( T )  o  
N  N 
Statistical Physics - Tatek Y. 47
Using eq. 7
 ' (T )
S  NkT  Nk log  ( T )  k log N
 (T )
S  ' (T )  log N 
  T  log  ( T )  o  
Nk  (T )  N 

Note: Quantum mechanical harmonic oscillators (1D)

1
  w  
 ( T )   2 sinh   
  kT  
Classical harmonic oscillators
kT
 (T ) 
w

Application: Solid-vapor equilibrium

Two phases in equilibrium are contained in gas phase

a volume V at temperature T.
The phases are free to exchange particles
solid phase

Statistical Physics - Tatek Y. 48

The equilibrium is attained when the phases have equal chemical potential, hence equal fugacity
 g
  s
   zg  zs

N g
Fugacity of the gas (cf. eq. (2)): zg 
V g f (T )
Ng = number of particles of the gas; Vg = volume occupied by the gas
Typically Vg ≈ V
1
Fugacity of the solid: zs 
 (T )

The equality of fugacities yields

N 1 N f (T )
zg  zs 
g
 
g
 Equilibrium particle density
V g f (T )  (T ) Vg  (T ) in the vapor phase

If the density in the vapor phase is low and the temperature of the system is high:

N g f (T )
P  kT  kT
Vg  (T )

Statistical Physics - Tatek Y. 49

For monoatomic vapor
 2  mkT 
3/2

f (T )  3
h
If the solid can be approximated by a set of 3D harmonic oscillators
3
  w  
 ( T )   2 sinh   
  kT  

An atom in the solid phase is more stable than an atom in the gas phase. Therefore, a certain
energy  is required to take an atom from the solid phase to the gaseous phase = =>
Latent heat of vaporization
f(T) has to be modified accordingly:

 2  mkT 
3/2
  / kT
f (T )  3
e
h

The pressure of the gas becomes

3
3 
 2  mkT  2   w   
P  kT  2
  2 sinh    e
kT

 h    kT  

Statistical Physics - Tatek Y. 50

Condition for the formation of a solid phase: the density of the gas should be smaller to the
density if all molecules are in the gaseous phase

f (T ) N g N
 
 (T ) V V

Or equivalently, below a characteristic temperature Tc

T  Tc

The critical temperature at which the transition occurs is given by

f (T c ) N

 (T c ) V

Statistical Physics - Tatek Y. 51

 3 – Quantum Statistical Mechanics
3.1 Quantum mechanical ensemble theory: the density matrix

- Let us consider an ensemble of N identical systems (N >> 1)

- The systems have the same hamiltonian denoted by the operator Hˆ . 
- The states of each system in the ensemble is totally determined by its wave function  ( r , t ).

- Let  k ( r , t ) the normalized wave function describing the kth system of the ensemble.
-The time variation of  k ( t ) is solution of Schrodinger equation:

Hˆ  ( t )  ih 
k k
(t )

- ( t ) may be written as a superposition of a complete set of orthonormal functions  n :

k

 (t )   a n ( t ) n
k k

d = volume element of the coordinate space

with a nk ( t ) 
  n ( t ) d 
*

- a nk ( t )  probability amplitudes for the various system of the ensemble to be in various

states n.
2
 probability that the system is in state n at time t.
k
- a n (t )
Statistical Physics - Tatek Y. 52
The probability is normalized:
2

  1
k
a n (t )
n

The density operator ˆ ( t )

N

 a 
1
 nm ( t ) 
k k*
m
(t ) a n (t )
N k 1

( t )  ensemble average of  a m ( t ) a n ( t ) 
k k*
- 
nm
2
- diagonal elements nn(t) = ensemble average of a n ( t )
k

- hence   nn  1
n

d ˆ
- We can show that ˆ satisfies i 
 Hˆ , ˆ  commutator
dt
which is the quantum mechanical version of Liouville’s theorem.
d ˆ
- A system in equilibrium must have  0  Hˆ , ˆ   0
dt
Statistical Physics - Tatek Y. 53
- In order to satisfy the latter condition, we need to have:

i) ˆ  ˆ ( Hˆ )  ˆ and Hˆ commute , and

ii) Hˆ must be independent of time (explicitly).

- If the basis functions n are eigenfunctions of Ĥ, we have:

H mn
 E n  mn  mn   n  mn , Hˆ and ˆ diagonal
- But in general ˆ is symmetric, i.e. nm = mn

Average value or expectation value of a physical quantity G

- Corresponding operator Gˆ always hermitian, i.e. Gˆ  Gˆ *t

1
N
 
G      a n a m G nm 
k* k

N k 1  n ,m 
with
ˆ
G nm   nG md
*

Statistical Physics - Tatek Y. 54

  G     mn G nm   ˆ Gˆ  mm

 Tr ˆ Gˆ 
n ,m m

diagonal matrix because defined in

the space of eigenfunctions n

-If k is not normalized, G  


Tr ˆ Gˆ 
Tr  ˆ 

3.2 Ensembles in quantum statistical mechanics

a) Microcanonical ensemble

The microcanonical ensemble is defined by:

- Systems with a fixed number of particles N in a fixed volume V and energy E±D.
- The microstates are equiprobable and their number is (N, V, E).
The density matrix expressed in the eigenfunctions space is diagonal:
 mn   nm

 1 D D 
and  E   En  E  
n    2 2 

0 otherwise 
Statistical Physics - Tatek Y. 55
where En are the eigenvalues of Ĥ

Entropy: S  k log 

b) Canonical ensemble

The integration over the phase space is replaced by a sum over all the states:
1
N !h
3N  dqdp  
n
 En
e
The density matrix is:  mn   n  0 ,1 , 2 ,...
 
,
mn   Hˆ
Tr e

1  En
where   and e  Boltzmann factor
kT

1
(V , T )  Tr  ˆ   E
Partition function:   n

 
Q e
N   Hˆ
Tr e n

  Hˆ
e
Density operator is independent of the representation ˆ 
Tr e    Hˆ

Statistical Physics - Tatek Y. 56
The ensemble average of G is:

Tr Gˆ e 
  Hˆ

 G   Tr ˆ Gˆ  1
Tr Gˆ e 
  Hˆ

Tr e 
  Hˆ
Q N (V , T )

c) Grand canonical ensemble

  ( Hˆ   nˆ ) 1   ( Hˆ   nˆ )
The density operator is: ˆ ~ e  e
Q (  ,V , T )
nˆ  number operator w ith eigenvalue s equal to 0 ,1, 2 ,...

The grand partition function is:



Q (  ,V , T )   e
 N
Q N
(V , T )  Tr e    ( Hˆ   nˆ )

N 0

Ensemble average of a physical quantity G




N
Gz Q (V , T )
 
N
1 ( Hˆ   nˆ )
Tr Gˆ e
 N 0
G   
Q (  ,V , T )

N
z Q N
(V , T )
N 0
Statistical Physics - Tatek Y. 57
3.3 Example

An electron in a magnetic field

1
Single electron with an intrinsic spin  ˆ and a magnetic moment B
2
e
ˆ  Pauli spin operator; B 
 mc
The applied field is B  B zˆ

The spin can have two possible orientations, up and down with respect to the field.

Hamiltonian: Hˆ    B

ˆ . B   B
B ˆ z

Pauli matrices

0 1 0  i 1 0 
ˆ x   ;
 ˆ y
  ;
 ˆ z   

1 0  i 0  0  1

Note: we have chosen a representation that makesˆ z diagonal.

Statistical Physics - Tatek Y. 58

Density matrix   Hˆ 
e B 
B
e 1 0
ˆ    
Tr e    Hˆ
 e
 B B
 e
  B B 
 0 e
  B B 

Expected value (ensemble average) of z
 B   B
e B
 e B

 z
  Tr  ˆ ˆ z    B   B
 tanh   B
B 
e B
 e B

3.4 The ideal gas: microcanonical ensemble

N non interacting identical particles

2
pˆ i
N

Hamiltonian: Hˆ   pˆ i  momentum operator  i 

i 1 2m
There are two types of N identical systems in nature: Bose systems and Fermi systems
The systems are characterized by the behavior of their wavefunctions under any interchange of
coordinate pairs.

Let n(q,p,s) be the eigenfunctions of Ĥ

-Bose systems:  n (q, p, s)   n ( p, q, s)

symmetric
- Fermi systems:  n ( q , p , s )   n ( p , q , s )
antisymmetric
Statistical Physics - Tatek Y. 59
Particles belonging to Bose systems are bosons.
Particles belonging to Fermi systems are fermions.

Boltzmann systems are systems of particles whose eigenfunctions are all eigenfunctions of Ĥ.
Boltzmann systems include Bose and Fermi systems + more

At high temperature, Bose and Fermi systems -----> Boltzmann systems

Let us compute the number of states (E) of the 3 systems for an energy eigenvalue between
E and E + D.
For simplification, we will consider spinless particles.

Energy of a single particle = level

2
p  
 
p
 , where p  p and p  momentum eigenvalue of the
2m the single particle

 2  
p  n
L
1
w ith n(n x , n y , n z ) and n   0 ,  1 ,  2 ,..
L  V 3

  x, y, z
Statistical Physics - Tatek Y. 60
The state of an ideal system is given by n p   set of occupation numbers

n  

p
number of particles having momentum p

Total energy of the system: E   

 p n p
p

Total number of particles: N  


n p
p

Occupation number for bosons, fermions (spinless) and Boltzmann gas:

 0 ,1, 2 ,... bosons 
n p   
 0 ,1 fermions 

Boltzmann gas:

n 
N!
n p  0 ,1 , 2 ,... but  specifies states of the system

p
n p !

p

In order to find (E), let us consider V ⟶ ∞ ===> p form a continuum.

Statistical Physics - Tatek Y. 61

Let us divide the energy spectrum into groups of levels p containing respectively
g1, g2, g3,…. levels:

g4
Each group is called a cell

g3 The ith group has gi levels and i average energy

Occupation number of the ith cell = ni

g2
ni   n 
p
levels inside ith cell

W n i   number of states of the system corresponding

g1 to the set of occupation numbers n i  .
p

The number of states is then

(E )   W n i  w ith E    ini
n i  i

N   ni
i

wi ≡ number of ways to assign ni particles in the ith cell.

Statistical Physics - Tatek Y. 62
For Bose and Fermi gas, interchanging particles in different cells does not lead to a new state.

 W n i    wi treat each cell as separate entity

i

For Boltzmann gas, interchanging particles in different cells leads to a new state = = => we
consider the N particles together.

Bose gas

Each level can be occupied by any number of particles

gi-1 ith cell: gi levels or subcells and gi-1 partitions

ith cell 2 Number of ways to assign particles

1
n i  g i  1 !
wi 
n i !  g i  1 !
(i-1)th cell
Number of states

n i  g i  1 !
W n i    wi  
i i n i !  g i  1 !
Statistical Physics - Tatek Y. 63
Fermi gas

The number of particles in each of the gi levels is 0 or 1 = = > wi = number of ways of choosing
ni things out of gi.
g i!
wi 
n i !  g i  n i !
The number of states is then
g i!
W n i    wi  
i i n i !  g i  n i !

Boltzmann gas

The N particles are placed into cells, the ith cell having ni particles.
N!
- The number of ways of placing the particles =
 (ni!)
- The ith cell has gi levels. i
n
= = > The number of ways for the ni particles to occupy the gi levels = g i i

- Among the ni particle in the ith cell, the first one can occupy these levels gi ways.

Statistical Physics - Tatek Y. 64

= = > The total number of ways to obtain
ni
gi
n i   N !
i ni!
ni
gi
- We divide by N! to obtain W n i   
i ni!

Entropy
S  k log  ( E ) w ith (E )   W n  i
n i 

Very difficult summation!

Let us approximate (E) by W n i  where n i  is the set of occupation numbers that
maximizes W n i  with the two constraints:
E   n i i
and N   ni
i i

And the entropy becomes: S  k log W n i 

In order to calculate W n i  , we use Lagrange multipliers method.

Statistical Physics - Tatek Y. 65

- Case of Fermi gas
g i!
log W n i    log
i n i ! ( g i  n i )!

The value of ni that maximizes W n  by taking the constraints into consideration satisfies
i
the following equality:

 
  log W n i     ni    n i i   0
 i i  ni  ni

where  is an arbitrary variation of ni and  and  are Lagrange multipliers.

 g i! 
   log   ni    n i i   0
 i n i ! ( g i  n i )! i i  ni  ni

Let us use the Stirling’s approximation for the first term. And after some calculation we get:

   n i
log n i   n i log( g i  n i )   n i   i  n i n i  ni
 0
i

 ( g i  ni ) 
  n i  log
ni
     i   0
i   ni  ni

For any variation

Statistical Physics - Tatek Y. 66
Since ni are independent variations:
( g i  ni )
log      i
 0
ni
( g i  ni )      
 log  e i
 g i  ni  nie i

ni
gi
The average occupation number for a cell is then ni    
e i
1

ni 1
Average level occupation number n p     
1
p
gi e

We use the conditions E  

 pn p and N  np

p p

 1
To find  and :    and  
kT kT

1
We finally obtain n p
 
1
1
p
z e


with z being the fugacity z  e
Statistical Physics - Tatek Y. 67
Similarly, we can do the same calculations for Bose and Boltzmann systems. The results are
1
Bose system: n p
 
1
1
p
z e

 
Boltzmann: n p  ze
p

And the expressions of the entropy for the three systems are

  gi ni  
  n
 i log( 1  )  g i
log( 1  ) Bose 
 i  ni gi  
S 
  gi ni  

 log W n i    n
 i log(  1 )  g i
log( 1  )  Fermi 
k  i  n i
g i  
  gi  
  n i log( ) Boltzmann 

 i  n i  


Note: We used Stirling’s approximation + gi >> 1 to obtain the above identities

Statistical Physics - Tatek Y. 68

The previous relations can be expressed differently

   i  log z 

 gi 1  i
 log 1  ze    i

Bose 

 z e 1 
 i 
   i  log z   i  
 
S
  g i  log 1  ze Fermi 
 1  i 
k  i  z e 1  
 z  g ie 
  i
  i  log z   Boltzmann 
 i 
 
ni   ni    n i 
2 2 2
The above relations are valid if

Boltzmann gas

Number of particles

     zV p
2 zV
N   ni  z  g ie  z e   d p 4 p e 
i p 2 /2m


3 3
 h
i i p 0
1

 2  
2 2
where thermal wavelength or de Broglie wavelength of a particle
    of mass m and energy kT.
 mkT 

Statistical Physics - Tatek Y. 69

     
Energy E   n i  z  g i i e  z 
i p

i

p
e

i i p


 p   p2
2
zV 3
  d p 4 p  e 
2 /2m
NkT
3  
h 0  2m  2

   log z    E  N log z
Entropy S   

 z e
p

p
k 
p

 
3

  
2
3  N 2  2

 N  N log  
   
2 V  mkT 
 

3.5 Ideal gases in other ensembles

 
g n e E n 
Partition function Q N (V , T )  
p

p
n p 

where E n p     p n p and N   n p
 
p p

Statistical Physics - Tatek Y. 70

Number of states corresponding to {np}
 1 Bose and Fermi 
 1 
g n p   
N! 1
  Boltzmann 
 N!  
n ! 
p  
n !

p 
 p p 

Boltzmann gas

Partition function  n    n 1 1
 e 
 
0 0
e 1     N
Q N
  

n0! n1!
...  

N!
e 0
 e 1
 ...
n 0 , n 1 , n 2 ,..  
 ni  N
n
   n 
    xi i
k
In order to make the above calculation,  xi    
we used the multinomial theorem  i  k 1 , k 2 ,... k
 1 , k 2
,...  i
 k i n


  
p
2 V
V   Q1     d p 4 p 
p 2 /2m
as e e

3

p 0

N
1 V
Q 1 (V ,T )
N
 Q N
(V , T )  
N !
3N
N!
Statistical Physics - Tatek Y. 71
 
Grand partition function  zV 
Q ( z ,V , T )   (V , T )  exp
N
z Q N  3 
N 0   
Bose and Fermi gas

QN(V, T) can not be easily evaluated because of the constraint

  n p p

Q N (V , T )   N   n
p
e with p
n p  p

Instead, let us evaluated Q:

    n p p    
e 
  n

    
p
Q ( z ,V , T )   
N N
z z 
p p
e
N 0  n p   N 0  n p  p 

  ze   ze  
  n0   n1
 0 1
...
n 0 , n 1 , n 2 ,..

 ni  N

  
 ze   ze 
  n0   n1
  0
 
1
 ...
 n0   n1 

Statistical Physics - Tatek Y. 72

Bose case: np = 0, 1, 2, … Fermi case: np = 0, 1

 1 
 
 p 1  ze   p
Bose


 Q ( z ,V , T )   
  1  ze   p
 Fermi 

 p 

Equations of state


 log 1  ze    p
 Bose 
PV  p 
 log Q ( z , V , T )   
kT  log 1  ze
 
 p
 Fermi 
 p 

Average number of particles

 
 ze
p

  
Bose 
  p 1  ze p

N  z log Q ( z , V , T )     p  Eq. 1
z  ze
 Fermi 
 p 1  ze   p

 
Statistical Physics - Tatek Y. 73
Average occupation number <np>
1
   n p p
1 
n p      
N
log Q ( z , V , T )
p
z n pe
Q ( z ,V , T ) N 0 n p    p

  p
ze
  
Bose and Fermi
p
1  ze

Equation (1) says: N   n p 

p

The above results are equivalent to those in the microcanonical ensemble.

Ideal gases

When the volume gets large, the levels V

V      
3
form a continuum: 3
d p
p h

i) Ideal Fermi gas

PV
 log Q ( z , V , T )   
log 1  ze
  p

kT p

 
2
 p
V 
  d p 4  p log  1  ze 
2 2m

h
3
 
0 
Statistical Physics - Tatek Y.
 74
The equations of state are then

  
2

4
p
P 
   d pp log  1  ze 
2 2m

 kT h
3
 
0   v = V/N = Specific volume
  
N 1 4 1
 
   d pp
2

 
3 2
p
V v h 0 1

 z e 2m
1 

The above equations can be written in the following form

P 1 1 1
 f5 (z) and  f 3
(z)
 
3 3
kT 2
v 2
1

 2  
2 2

where    

is the thermal wavelength, and
 mkT 

 1
  l 1

f5 (z) 
1
 x
2
log 1  ze  x
2

d x   5
2
 0 l 1 2
l
 1
 l 1

l
z
f3 (z)  z
z
f5 (z)   3
2 2 l 1 2
l
Statistical Physics - Tatek Y. 75
ii) Ideal Bose gas

The equation of state has a term in log 1  ze    p


For z - - -> 1, the term corresponding to p = 0 diverges.
We need to split terms corresponding to p = 0 and replace the rest by integrals.

   
2

4
p
P  1
    d pp log  1  ze   log 1  z  
2 2m

 kT h
3
  V 
0  
  
N 1 4 1 1 z
 
   d pp 
2

  p  0
2
V 1 z
3 p
V v h 0 1

 z e 2m
1 

The above equations of state can be written the following way:

 P 1 1 
 g (z)  log 1  z 
 5 

3
 kT 2
V 
 
1 1 1 z
  g 3
(z)  
  V 1 z 
3
 v 2 

Statistical Physics - Tatek Y. 76

where the two g functions are
 

 d x
l
4 2 z

x
(z)    log 1  ze 
2
g 5
x 5
2
 0 l 1 2
l


l
z
g 3
(z)  z
z
g 5
(z)   3
2 2 l 1 2
l

z 
Note that   n 0  is the average occupation number for the level p  0
1 z

Internal energy 1
   n p p

U ( z ,V , T )     pe
N p
z
Q ( z ,V , T ) N 0 n p 


  log Q ( z , V , T )


PV
Since log Q ( z , V , T ) 
kT
Statistical Physics - Tatek Y. 77
  3 kT  =P
f5 (z) Fermi
 
2 
3
1  
U ( z ,V , T )   2

V 3 kT
 g 5
(z) Bose  = P if log (1 –z) / V is
 2  
3
 2  neglected

This allow us to find a simple relation between the energy P and V

3
U  PV
2

Statistical Physics - Tatek Y. 78

 4 – Bose Systems
4.1 Bose-Einstein condensation

Equations of state
 P 1 1 
 g (z)  log 1  z  Eq. 1
 5 

3
 kT 2
V 
 
1 1 1 z
  g 3
(z)   Eq. 2
  V 1 z 
3
 v 2 

In order to study eq. (1), we need to find z as a function of T and v in eq. (2). To do so, we need
to know the properties of g3/2(z).
To satisfy eq. (2), z has to be 0 ≤ z ≤ 1. For this choice of values range
2 3
z z
g 3
(z)  z  3
 3
 ... is bounded.
2 2
2 32
At z = 1, g’3/2(z) diverges, but g3/2(z) is finite:
1 1  3 
g 3
(1 )  z  3
 3
 ...      2 . 612 ... Riemann zeta function
2 2
 2 
2 32
Statistical Physics - Tatek Y. 79
Graphical representation of g3/2(z) z
Since  n 0   , eq. (2) may be written
g3/2(z) as follow: 1 z

n0  
3

   g 3 (z)
3
2.612
V v 2

Which means that

n0  
3

  0  g 3( 1 )
3
when
V v 2
The last relations imply that a finite number of

particles occupy the level p  0
1 z
This phenomenon is known as Bose-Einstein
condensation


3

The condition  g 3 ( 1 ) defines a subspace of the thermodynamic P-v-T space of the

v 2
ideal Bose gas and corresponds to the transition region of the B-E condensation. The subspace is
separated from the rest of the P-v-T space by a 2D surface of equation

3

 g 3( 1 ) Eq. 3
v 2
Statistical Physics - Tatek Y. 80
Let us denote by <ne> the average number of particles in the excited levels:
ne  1 1
ne   N  n0    g 3 (z)  g 3( 1 )
 
3 3
V 2 2

The maximum number of particles in the excited levels is for z = 1.

Therefore, the limiting value of <ne> is given by
3

ne  1  mkT  2
lim  g 3(1 )  g 3( 1 ) Eq. 4
 
  2 
3 2
z1 V 2  2

V 
If N  g 3 ( 1 ) , the rest of the particles will be pushed to the ground state p  0

3
2

and
n0  N ne  1  V  Bose-Einstein
  lim   N  g 3( 1 )
V V V 
V  
3  condensation
2 

The critical temperature Tc for which we have B-E condensation is then

3

N 1  mkT  2
Eq. 5
 g 3( 1 ) 
c
g 3( 1 )
 
  2 
3 2
V 2  2

Statistical Physics - Tatek Y. 81

Using eq. (4) and eq. (5), we get 3

ne   T  2

  
 
N  Tc 
Below the critical temperature Tc, the system may be looked upon a mixture of two phases:
3

 T 2
i) a normal phase consisting of  n e   N   particles distributed over the excited

states ( p  0 ), and T
 c 

ii) a condensed phase consisting of  n 0   N   n e  particles accumulated in the


ground state ( p  0 ).
Behavior of the condensed phase near the
critical temperature
1
n0  N  ne 
ne  
n0  N N
N
3
N
 T  2

 1  
 
 T c 

T 3 Tc  T

1 T cPhysics - Tatek Y. 2 Tc
Statistical 82
Variation of z with T
3
1 v
Because ~ T 2 , it is simpler to look at the variation of z with

3

3

v 1
For 0   g 3
(z) (cf. eq. (2)) which corresponds to 0  T  T c , z ≈ 1 because

3
2

z n0  1
n0   0 and n0    z   1
1 z n0   1 n0 

For a given v, eq. (3) defines a critical temperature Tc:

2 
2
/ m
 c  vg (1 )  kT c 
3
3 2
2
  3

 vg 3
(1 ) 
 2 

Tc is the temperature at which the thermal wavelength ( is of the same order of magnitude
as the average interparticle separation.

3

For a given temperature T, eq. (3) defines a critical volume vc: vc 

The region of condensation is the region in which g 3
(1 )
T < Tc or v < vc 2

Statistical Physics - Tatek Y. 83

 
3
v 1
For  g 3
(1 ), z  1 and is determined by g (z)   g (1 )

3 3 3
2
2
v 2
 l
v z
For 3  1 , g 3
( z )  1 implies z  1 because g 3
(z)   3
 2 2 l 1 2
1
l
 v 
Therefore g 3
(z)  z  z   3 
2   

Dependence of z on T curve
z

1
1
 v 
z   3 
  

1 v
g (1 ) 
3
3
2 Statistical Physics - Tatek Y. 84
(P-T) diagram

For T < Tc and z = 1, let us use the equations of state to obtain:

5
kT g (1 )  1 . 341
P (T )  g 5
(1 ) ~ T 2 5


3 2
2

P(T) is independent of v implying for the system an infinite compressibility.

We use the expression of Tc to find the value of the pressure at the transition point:
3

 2 m  2 5

P (T c )   2
  kT c  2 g 5
(1 )
   2

g 5
(1 )
N  N 
 2
kT c  0 . 5134  kT c 
g 3
(1 ) V  V 
2

For T > Tc, the pressure becomes

g 5
(z)
N
P (T ) 
2
kT
g 3
(z) V
2
Statistical Physics - Tatek Y. 85
 PV 2 U
 transition line ~ T5/2
NkT c
3 NkT c

2
B-E ~ T

1
gas phase

1 2 Tc

(P-v) diagram (isotherm)


3 3

Critical volume vc  ~ T 2

g 3
(1 )
2

kT
For v < vc P0  g 5
(1 ) independent of v

3
2

Statistical Physics - Tatek Y. 86

The region of the mixed phase is marked by a boundary line of equation:
5 2
g 5
(1 )

  const.
3 2
P0 v c
2 m
5

  3

g 3
(1 ) 
 2 

 const.
5/3
Pv

P 02
T 2  T1

P01

v
vc2 v c1

Statistical Physics - Tatek Y. 87

4.2 The blackbody radiation

We consider the equilibrium properties of electromagnetic radiation enclosed in a volume V at

temperature T = blackbody cavity.

It can be experimentally produced by heating a cavity of any given material to a temperature T.

The hamiltonian of a free electromagnetic field is equal to the sum of hamiltonians of harmonic
oscillators at different frequencies.

In quantum theory, a harmonic oscillator at frequency w can only have energies

1
 n
 (n  ) w , n  0 ,1 , 2 ,...
2
A state of the free electromagnetic field is given by the number n for each oscillators = number
n of photons for each frequency.

We will consider linearly polarized photons.

  w
A photon of frequency w has energy   w , momentum   k , k 
  c
 
polarization vector   ,   1, k .  0

Statistical Physics - Tatek Y. 88

Such photon corresponds to a plane wave of electromagnetic radiation whose electric field is
   i  k . r  w t 
E (r , t )   e

Periodic boundary condition in a cube of volume V = L 3, gives


 2 n 
k  n  (n x , n y , n z ) with n   0 ,  1 ,  2 ,...   x, y, z
L
V
4 k d k
2
The allowed number of momentum values between k and k + dk is
2  
3

The total energy is 

E n k ,     
 w n k , 
k ,


where n  
k ,
 is the set of occupation numbers of photons of momentum k and polarization 

and w  c k , n k ,   0 ,1 , 2 ,...

E n   
Q  
k ,
Partition function e
n    
k ,

Note: the number N of photons is not defined.

Statistical Physics - Tatek Y. 89

  
 
w n  
k ,

1
  w n
 Q     
k ,
e e   w
n   
 
k , n0
 
k ,
1 e
k ,

 log Q    log 1  e
 
   w
  2  log 1  e

   w

k , k

because we have 2 values of the

polarization for a given k
Average occupation number for photons of momentum k:
1  log Q 2 1
  w n 
k

 n k   
  ( w )

e
 w
1

Q


n k e k

Internal energy: 
U  

log Q  
 w  n k  Eq. 6
k

 
 1 / 3
  c 2 n V
Pressure: log Q   2  log 1  e

k

 2 n
In eq. (7), we used the fact that: w  c k  c
L
1  1
Therefore P 
 V
log Q 
3V k

 w  n k  Eq. 7

Statistical Physics - Tatek Y. 90

 1
By comparing eq. (6) and eq. (7), we get: PV  U
3
Let us calculate U for V ⟶∞
 
w
3
2V  ck V
U   d k 4 k   dw
2

2 
  ck  w
 c
3
1 1
2 3
0
e 0
e

Internal energy per volume:


w
3
U 
V
  d w u (w , T ) where u (w , T ) 
 c
2 3
e
 w
1
0

  kT 
2 4
U
 
 w 
3
V 15

1 U 4
2 4 3
k T
Specific heat per unit volume: C  
15   c 
V 3
V T

Amount of energy radiated by the blackbody cavity per second per unit area of the opening:
d c
I (w , T )  c  u ( w , T ) cos   u (w , T )
4 4

Statistical Physics - Tatek Y. 91

Integration over the frequencies yields:


2 4
k
I (T )  c  d w I (w , T )   T  
4
, 3 3
60  c
Stefan’s law Stefan’s constant

Radiation intensity as a function of frequency for different temperatures

I(w, T)

T1 < T2 < T3

T3
T2
T1
w
w1 w2 w3

Statistical Physics - Tatek Y. 92

 5 – Fermi Systems
5.1 Equations of state of an ideal Fermi gas
1

 2  
2
 
2
P 1 V
 f ( z ) v  ,    

 5 

3
 kT 2
 N  mkT 
  
1 1 2
  f3 (z)  4 x

 v 
3
2


f3 (z) 

 dx z
1
e
x
2

1
2 0

Case of small values of z

2 3 4
z z z
f3 (z)  z     ... power series expansion
3 3 3
2 2
2 32 4 2

Case of large z – asymptotic expansion (Sommerfeld method)

 
z  e and   log z 
kT
Let us make the following change of variable
1
y  x  dy  2 xdx  x dx 
2 2
y dy
2
Statistical Physics - Tatek Y. 93
The function f3/2(z) then becomes
 2 
4 x 2 y
f3 (z) 

 dx e
x
2

1


 dy e
y 
1
2 0 0

Let us calculate the above integral by parts

y 
1 e
u( y)   u '( y)  
e 
y  2
e 1 y 
1
3
2
v'( y)  y  v( y)  y 2

3

2 y
 f 3
(z) 

 dy e
y 
1
2 0 =0

 
3 3
2  2 y 
 4 y  4 y e
   dy
3   1 e  1
y  2
e 3  y 

  0
0

3
 2 y 
4 y e
  dy
e  1
2
3  0
y 

Statistical Physics - Tatek Y. 94

Let us expand y3/2 in a Taylor series about :
3 3 1 1
3 3 
y  y 
2
y 2
 2
  2
  2
  ...
2 8
3 1 1
3 3 
   t    .... t  y 
2 2 2 2
t with
2 8

 
t 3 1 1
4 e 3 3 
    .... 
 f3 (z)   dt  t 
2 2 2 2
t
 
e 
2
2
3  
t
1  2 8 
  

Since
    
  


 
t 3 1 1
4 e 3 3 
    .... 
 f3 (z)   dt  t 
2 2 2 2
t
 
e 
2
2
3  
t
1  2 8 


The integral  is of the order of e-, therefore negligible.


Therefore, the integral becomes

 
3 1 1
4 3 3 
f3 (z)   I 0 2
 I 1 2
 I 2 2
 .... 
 
2 3   2 8 
Statistical Physics - Tatek Y. 95
 
where 4 t e
n t

In   dt
e 
2
3  
t
1
t
e
is an even function of t = = = > In = 0 for odd n.
e 
2
1
t

d 1
For n = 0 I0  2 dt  1
e 
2
1
t
0
dt

For n > 0 I n  ( n  1 )! ( 2 n ) 1  2 
1 n
(n) Riemann Zeta function
  
2 4 6

 (2)  ;  (4)  ;  (6 ) 
6 90 945
  
3 2 1
4
  log z    log z

 f3 (z)  2 2  ... 
2 3   8 
f 3
(z)
2
3
4
  log z 2
3

3 
As f 3 ( z ) ~ T
2
, for fixed v, v 3 
2
 z
z increases when T decreases. z
Statistical Physics - Tatek Y. 96
z
High temperatures and low densities 
3

 1
v
In this case, the average interparticle separation is much larger than the thermal wavelength
2
    
3 2 3 3
z 1
  and
1/3
v  z   ...  z      ...
3/2 3/2  
v 2 v 2  v 

3

For T   ,   0 and z 
v

Average occupation number

  p
ze   Boltzmann gas
n    ze
p

p  
1  ze
p

≈1
Equation of state

v   
2 3
Pv z 1
  z   ...   1  ... Virial expansion
3  
 
5/2 5/2
kT 2  2 v

Statistical Physics - Tatek Y. 97

Low temperatures and high densities 
3

 1
v
In this case, the average interparticle separation is much smaller than the thermal wavelength

 
1/3
v = = => quantum effects become predominant.
3/2
 1  2  
3 2
4
 log z
3/2
For T  0 ,  f3 (z)    
 
v 2
v  mkT  3 
2/3
 6 
2 2


z  e F
with F = chemical potential at absolute 0 = Fermi energy:  F
 



2m  v 
Average occupation number
  p
ze 1 1
n p
   
 
  ( 
1 )
1  ze 1 1
p p p F
z e e

 (  )
If  p
  F
, e
p F
 0 for T  0 (    ) and n p   1

And the average occupation number becomes

1 ( p
  )
F
n p T 0   
0 (  F )
 p 
Statistical Physics - Tatek Y. 98
Pauli exclusion principle: no two particles can be in the same state, hence <np> = 1 or 0

In the ground state of the system, particles occupy lowest possible levels and fill the levels up
to F.

In momentum space, particles fill a sphere of radius pF. The surface of the sphere is called
Fermi surface.

General form of F

Let us consider a g-fold degenerate single particles energy levels:

g = 2s + 1 for particles of spin s.

There are N states with the energy below the Fermi energy
2
pF N  4 3  g 1
     p F  
 2  
F 3
2m V  3 v

2/3
 6 
2 2

    
F  
2 m  gv 

Statistical Physics - Tatek Y. 99

Thermodynamic functions for low temperatures and high densities
 
2
  kT 
2

Chemical potential   kT   kT log z   F 1  




 ...  Eq. 1
 12   F  
 
   
3 2 1
4
  log z  2   log z  2  ...  

we used f 3 ( z )  to obtain eq. (1).
2 3   8  v

Let us define the Fermi temperature T F 
F

k
For T << TF, the gas is degenerate implying that TF is the degeneracy temperature.
1
Average occupation number:  n p    
where  is given by eq. (1)
1
p
e
n p

kT

T=0
Fig. 1

T>0
2
p
 p

 F
2m
Statistical Physics - Tatek Y. 100
 
Internal energy V 4
U    n p    n p  p dp
4
p 3
 h 2m
p 0

After partial integration, we get

   
   V
5 6 p
V p p e
  n p   
U   dp   dp
4 20  m  e 
2
p
2 2 2 2 3  
m 0
5   0
p
1
with u(p) = p5/5 and v(p) = <np> Eq. 2

From figure 1, we note that  n p  is sharply peaked at p = pF:
p


 n p 
p  function for T  0

T  0

p
pF
Statistical Physics - Tatek Y. 101
Integral (2) can be evaluated by expanding p6 about p = pF:
 pF  6 p  pF   15 p 
6
 pF  ...
6 6 4
p pF

and by replacing  by its expression in eq. (1):

 
2
5  kT 
2
3
U  N 1     ... 
F  
5 

12  F  

The first term corresponds to the ground state energy of the Fermi gas E0 which can also be
evaluated as follow:
pF
 p  2  gV
2 2
p V
E0    g  4  p dp   
2 5
p

3   3 F
p  pF 2m h 0  2m  5 mh
1/3
 3 
But since p F  h  
 , we have
 4  gv 

2  gNv   3 2
2
3 3 pF 3
E0  

h pF 
 N  N F
 4  gv
3
5 mh  10 m 5

Statistical Physics - Tatek Y. 102

Specific heat CV
U
2
3 10 k T Nk
CV   N  
2
F
T 
2
5 12 F
3

2
CV kT
  2
Nk 2  F

At low temperatures C V ~ T
3
For T   CV  kT
2  F

Pressure
 
2
2  5  kT 
2
2 U
P  
F
1     ... 
 
3 V 5 v 

12  F  


5.2 Electron gas in metals

Let us consider sodium metal.

Since electrons in a metal are not really free, they are given an effective mass m’ < m.

Statistical Physics - Tatek Y. 103

Fermi energy 2 2/3
h  3N  g = 2 = 2s + 1
   
F
m '  8 V  m’ = effective mass of the electron in the gas.

ne = number of conducting electrons per atom

Electron density: N  n e n a
3 na = number of atoms per unit cell
V a
a = lattice constant (cell length)
For sodium we have: n e  1 , na  2, a  4 . 29 angstrom, m '  0 . 98 m e

Fermi energy and Fermi temperature are then:

 F  Na  5 . 03  10
 12
Erg  3 . 14 eV


T F  Na  3 . 64  10
4
K 
F

k
(TF)Na is higher than room temperature (~ 300 K) = = => conducting electrons in Na are highly
degenerate.
  kT    T 
2 2

Specific heat C V  Nk    Nk  
   
2 
 F  2  TF  ~ 10-2
3
CV is much smaller than the value obtained for classical gas ( C V  Nk ).
2
Statistical Physics - Tatek Y. 104
5.3 The Ising model

In some metals (Fe, Ni) a finite fraction of the spins of the atoms becomes spontaneously
polarized in the same direction creating a macroscopic magnetic field.

This takes place below a critical temperature known as Curie temperature.

The Ising model is a simple model of ferromagnetism.

In the Ising model, we have an array of N fixed points called lattice sites that form a n-
dimensional periodic lattice (n = 1, 2, 3).

The lattice structure may be cubic or hexagonal.

Each site has a spin variable Si (i = 1, 2, …., N) which can take the values +1 or -1 (spin up or
down) ===> Ising spins = atomic magnetic moments .

The energy of the system in the configuration specified by {Si} is

N

E S i      ij
SiS j
 H  Si <ij> = nearest neighbor pair of spins
 ij  i 1

applied external field

Statistical Physics - Tatek Y. 105
As <ij> is identical to <ji>, there are N/2 terms in the summation.
4 (2D square lattice)
= 6 (3D square lattice)
8 (3D body-centered cubic lattice)

Let us consider isotropic interaction, i.e. ij = . The energy becomes

N

E S i      SiS j
 H  Si
 ij  i 1

 > 0 = = = > ferromagnetism,  < 0 = = = > antiferromagnetism

Partition function
N
 E S i 
Q (H ,T )    e
i Si

Since we have to possible values for the spins, Si = +1, -1, there are 2N terms in the summation.

Thermodynamic functions

Helmoltz free energy A ( H , T )   kT log Q ( H , T )

Statistical Physics - Tatek Y. 106

 
N
 A 
Magnetization M (H ,T )       Si 
 H  kT  i 1

M(0, T) = spontaneous magnetization Ensemble average

If M(0, T) is non-zero, the system is ferromagnetic or antiferromagnetic.

Case of non-interacting particles subject to a field H ( = 0)

Energy of a spin E   H

= magnetic moment

Partition function for a single particle

External Parallel Antiparallel
field H spin spin   H  H  H 
Q1  e  e  2 cosh  
 kT 
Partition function for N particles. Since they are not interacting, we get:

N
  H  
Q N  Q1   2 cosh   
N

  kT  

Statistical Physics - Tatek Y. 107

   H 
Helmoltz free energy A   NkT log  2 cosh  
  kT 

Entropy A    H  H  H 
S     Nk  log  2 cosh    tanh  
T    kT   kT  kT  

Internal energy  H 
U  A  TS   N  H tanh  
 kT 


 A   H 
Total magnetic
Magnetization
M      N   N  tanh   moment
 H  kT   kT 

2
 H 
Specific heat Nk  
U  kT 
C  
T 2  H 
cosh  
 kT 

Statistical Physics - Tatek Y. 108

Case of interacting particles with magnetic moment 0

Mean-field approach: each particle is subject to the external field plus an average field due to
to the other (neighbors) particles.

H eff
 H  H ' H '  qM

external field field due to magnetization

molecular field
neighbor spins
constant

Each particle in the system has an average magnetic moment in the direction of H eff:

 0H  N number of atoms per

  0
eff
tanh   n   unit volume
 
 kT  V

Magnetization:  0 
M  n  n 0 tanh  H  qM  Eq. 3
 kT 
H’ = qM is determined in a self-consistent manner

  number of neigbouring spins

    
  
 
 average number of upward and downward spins
Statistical Physics - Tatek Y. 109
Let us evaluate the molecular field constant
qM   0  
  
  
 n M
  
  
      where M 
 n 0
0 n 0 M 

M 
 0 H '    
  
     0 qM  q 
M 
0M 

Equation (3) can be then written as

M  0H  M 
 tanh   
 
M   kT kTM  

The spontaneous magnetization Ms(T) in the absence of an external field (H = 0) is determined

by the equations:

   M
X  tanh  X , X 
 kT  M 

Statistical Physics - Tatek Y. 110

 M s
(T )
M  X  tanh x

T
X  x
Tc

This means that the intersection of the two graphs

kT T 
X  tanh x, X  x  x (T c  )
 Tc k

M (T )
gives x(T) and so X ( T )  s

M 

M s
 saturation magnetizat ion

M 
 magnetizat ion at 0 K
Statistical Physics - Tatek Y. 111
 6 – Phase Transitions and Critical Phenomena

A thermodynamic system can exist in a number of different phases whose macroscopic behavior
can differ dramatically.

Generally, systems become more ordered as the temperature is lowered because forces of
cohesion overcome thermal motion and atoms rearrange themselves into more ordered state.

Phase changes (or transitions) occur abruptly at some critical temperatures.

Phase transitions are characterized by the tendency of a large number of microscopic

constituents of the system to interact with one another in a strong cooperative fashion.

This cooperative behavior (or correlation) is macroscopically significant at critical temperatures

(Tc).

The situation in which particles interaction is important is very different from the models
considered so far for which interactions between particles are neglected:
ideal gases, solids (modeled as non interacting harmonic oscillators), Bose
condensation, blackbody radiation, paramagnetism

Statistical Physics - Tatek Y. 112

Examples of phase transitions
- condensation of gases
- melting of solids
- phase separation in mixtures and solutions
- ferromagnetism and antiferromagnetism
- order-disorder transition in alloys
- superfluid transition from liquid He I to liquid He II
- transition from normal to superconducting material

Most phase transitions can be classified into 2 categories: first-order phase transitions and 2nd
order (or continuous) phase transitions.

Order parameters are often associated with phase transitions and give the “level of ordering” of
the phases.

Several phase transitions can be described microscopically using simple class of lattice models.

Other theoretical approaches are: mean-field theory and renormalization group theory.

Statistical Physics - Tatek Y. 113

6.1 First and second order phase transitions

In order to characterize phase transitions in a general

  way, let us consider a system for which we
have the Gibbs free energy: G ( N , T , P , H , E ,...)

For a first-order phase transition, one of the first derivatives of the Gibbs free energy with
respect to the external parameter is discontinuous:

 G   G   G 
S    , V    , M   
  T  N , P ,...   P  N , T ,...   H  N , T ,...

This discontinuity produces a divergence in the higher derivatives, like the specific heat C P, the
compressibility , the susceptibility , or the expansion coefficient 

 S    G  1  V  1   G 
2 2

C p
 T    T 



       



 T  P  V  P T V  P
2 2
 T P T

1  V  1   G 
2
 M    G 
2
     


        
V  T  P V  P T 
  H T 
2
 H H

Statistical Physics - Tatek Y. 114

Case of discontinuous entropy S
G S CP
∞ P = const.
P = const. P = const.

Phase 2 Phase 2 Phase 2

Phase 1 Phase 1 Phase 1

T T T
Tc Tc Tc
Gibbs free energy, the entropy and the specific heat as a function of T for a first-order phase
transition (e.g. liquid-gas transition)

First-order phase transitions are associated with latent heat, which implies that transitions such
as liquid-vapor, solid-liquid, solid-gas are 1st order phase transitions.

For a phase transition of second order, the first derivatives of the Gibbs free energy are
continuous.

Second derivatives like CP, , … are discontinuous or divergent.

Statistical Physics - Tatek Y. 115

Case of continuous entropy
G S CP
∞ P = const.
P = const. P = const.

Phase 2

Phase 1 Phase 2 Phase 1

Phase 1 Phase 2

T T T
Tc Tc Tc
The Gibbs free energy, the entropy and the heat capacity as a function of T for a 2 nd order phase
transition

6.2 A lattice model of phase transition: the Ising model

It is a simple model which can describe phase transitions in ferromagnets, gas-liquid, liquid
mixtures, binary alloys, etc.

Only nearest-neighbors interactions are included in the interaction energy.

We consider an external magnetic field B directed in the z positive direction.

Statistical Physics - Tatek Y. 116
Hamiltonian of the system
H  i
 J   i j
 B   i
n .n i

where  i
  1 , J = coupling constant,  = magnetic moment

For J > 0, it is energetically favorable for neighboring spins to be aligned = = = > for low enough
temperature, existence of cooperative phenomenon called spontaneous magnetization which
involves long range correlations between spins.

Partition function

Q N
(B ,T )    ...  exp   H  i

 1  2  N

 
   ...  exp   J   i j
  B   i
 1  2  N  n .n i 

Helmoltz free energy A ( B , T )   kT log Q ( B , T )

N

 A  
Internal energy U ( B , T )   T 2
   kT
2
log Q (B ,T )
N
T  T  T

Statistical Physics - Tatek Y. 117

Specific heat U 
2
A
C (B ,T )   T
2

T T
2

Net magnetization
H 1   log Q N   A 
M (B,T )     i
 
B
 
  B
   
i T  B T

M ( 0 , T )  spontaneous magnetization

If M ( 0 , T )  0 below a critical temperature Tc, the system is called ferromagnetic under Tc

and paramagnetic above Tc.

Evaluation of the partition function

We have a system made of N spins. Let us introduce the following quantities:

N+ = total number of up spins

N- = total number of down spins
N++ = total number of nearest-neighbor up-up pairs
N-- = total number of nearest-neighbor down-down pairs
N+- = total number of nearest-neighbor pairs with opposite spins

Statistical Physics - Tatek Y. 118

We then have N 
 N 
 N Eq. 1

If q is the coordination number of the lattice we get the following identities:

qN 
 2N 
 N 
, qN 
 2N 
 N 
Eq. 2
Using eq. (1) and (2) we get the following relations:
N 
 N  N 

N 
 qN 
 2N 

1
N 
 qN  qN 
 N 
2

The total number of nearest-neighbor pairs of all types is:

1
N 
 N 
 N 
 qN
2
The hamiltonian of the system in terms of N+ and N++ is

H N
(N 
,N 
)   J N 
 N 
 N 
  B N   N 

 1 
 J qN  2 qN 
 4N     B 2 N 
 N 
 2 
Statistical Physics - Tatek Y. 119
Let us define by gN(N+, N++) the number of distinct ways in which the N spins of the lattice can
be arranged for a given set (N+, N++).

The partition function can be expressed in terms of this quantity:

Q N (B,T )   g N
(N 
,N 
) exp  H N
(N 
,N 
)
N  ,N 

 1 
N  qJ   B  N
2  qJ  B N  4  JN
 
 2 
 e e  g N
(N 
,N 
)e 

N  0 N 

The main problem is to find the combinatorial form of gN(N+, N++) for various lattices.

6.3 The lattice gas and the binary alloy

The Ising model can be further extended to include other systems such as the lattice gas and
the binary alloy

(i) Lattice gas

Let us consider a system of Na atoms which can occupy only discrete positions in space.

Each lattice site can be occupied by a maximum of 1 atom.

Statistical Physics - Tatek Y. 120
Therefore, the state of a site can be described by
1 if it is occupied
0 if it is empty

Moreover we assume that the structure has a coordination number q and the interaction energy
between 2 nearest-neighbor occupied sites is –0.

The configurational energy of the system is then E = - 0 Naa, where Naa is the total number of
nearest-neighbor pairs.

Let us define by gN(Na, Naa) the number of ways of distributing Na atoms in N sites so that we
have Naa pairs.

Partition function of the system



N
Q N (N ,T )  g N
(N a
,N aa
)e 0 aa

a
N aa

N can be considered as the volume V of the system

Grand partition function

N


N
Q (z, N ,T )  z a
Q N (N ,T )
a
N a 0

Statistical Physics - Tatek Y. 121

Pressure of the gas
N
PV PN  PN 
 
N N
  log Q ( z , N , T )  e  z a
g N
(N a
,N aa
)e 0 aa

kT kT N a 0 N aa
Mean number of particles  Nz   P 
N a
 z log Q ( z , N , T )   
z kT   z  T
Average volume per particle 1 N  P 
z
 
a
 
v N kT   z  T

Correspondence between the lattice gas and the ferromagnet

Lattice gas ferromagnet Note

The canonical ensemble
N ,N  N N ,N  N
a a   of the ferromagnet
corresponds to the grand
 4J
0
canonical ensemble of the
z exp  2   qJ   B  lattice gas.

 A 1  Also we used:
P    qJ   B 
 N  M   N 
2

 N 
 1  N  1  M 
  2 N N 
N a
   

  1   
 2  N  
N  v  N   
Statistical Physics - Tatek Y. 122
(ii) Binary alloy

The lattice structure consists of 2 types of atoms, 1 and 2, whose respective numbers are N 1
and N2

There are therefore three types of nearest-neighbor pairs: N11, N22, N12, and 11, 22, 12 are
their corresponding interaction energies.

Total configurational energy: E   11 N 11   22

N 22
  12 N 12

We have also the following relations

N  N1  N 1
2 N 22
 qN  qN 1
 N 11
qN 1
 2 N 11  N 12 ===> 2
N 12  qN  2 N 11
qN 2
 2N 22
 N 12 1

Therefore, the energy can be expressed as

 1 
E   11 N 11   22  qN  qN 1
 N 11    12  qN 1
 2 N 11 
 2 
1
 q 22
N  q   12   22
N 1    11   22
 2  12  N 11
2
Statistical Physics - Tatek Y. 123
Correspondence between binary allow and lattice gas:
Lattice gas binary alloy
N a
,N  N a
N1, N  N 2
 N 2

  0
 11   22
 2  12 
1
A A  q 22
N  q  12   22
N 1
2

Correspondence with ferromagnets:  11    J,  12  J

22

6.4 The Ising model in the zeroth approximation

The average energy of an atom of the system is determined by the average degree of order
prevailing in the system rather than by the configurations of the neighboring atoms. This is
called the zeroth approximation.

Let us define a long-range order parameter L:

1 N  N 2N  N N
     
L  i
   2 1 Eq. 3
N i N N N

Statistical Physics - Tatek Y. 124

L is bounded between 2 values  1  L  1

and we have L   1 - complete order

L  0 - complete disorder
From eq. (3), N+ and N- can be expressed in terms of L

N N
N 
 1  L , N 
 1  L 
2 2
Magnetization
M  N   N 
  N L
The above relation means that L is a direct measure of the magnetization.

Hamiltonian to avoid double

H  i
 J   i j
 B   i counting
n .n i

 1 
 J q  
From eq. (3) i

  i
 NL  2  i
i
And we get the following relation
1 1
   L  N    
N i L
Statistical Physics - Tatek Y. 125
The total configurational energy is then

 qJ L  NL
1
E    (  B ) NL
2
The average or expectation value of E is
1 1
E  U     BN L   qJN    BN 
2 2
qJN L
2 2
The average energy difference per spin between the up and down state is
DU
D    qJ  ( D  )   B ( D  )
N
difference between up
 qJ  and down spins
 2     B  for D  2
  
qJ
  internal (molecular) field of Weiss


Relative values of the equilibrium numbers N 

and N 

N   D   2 (B  B ') 
 exp     exp    Bolztmann distribution
N   kT   kT 

Statistical Physics - Tatek Y. 126

where qJ   M 
B'  qJ  
  N
2

Moreover, using eq. (3) we get
N 
 N 
N 
 N 
N 
1 L
L    
N N 
 N 
N 
1 L
and
1 L  2 ( qJ L   B ) 
 exp   
1 L  kT 

and finally
qJ L   B 1 1 L 1
 log  tanh L
kT 2 1 L

Condition for a spontaneous magnetization at zero external field (B = 0)

 qJ L 0 
L0  tanh   Eq. 4
 kT 
The above equation may be solved graphically

Statistical Physics - Tatek Y. 127

  Tc 
Critical temperature:
tanh  L 0 
 T  qJ
Tc 
1 k
T  Tc

T  Tc

 L0
L0
L0

L 0 ( T ) is determined by the point of intersection of

i. the straight line y  L 0 ( T ) and

-1  qJ L 0 
ii. the curve y  tanh  
 kT 
The solution L 0 ( T )  0 is here for any T.

  qJ L 0 
L   1
A non-zero solution exists if d qJ d
 tanh      0
dL0   kT  L0  0
kT dL0

Statistical Physics - Tatek Y. 128

The latter condition implies a constraint on the temperature
qJ
T   T c  critical temperature below which the system can acquire a spontaneous
k magnetization.

Tc = Curie temperature that marks a transition from ferromagnetic to paramagnetic state.

Approximated expressions of L 0 ( T )

Near the critical temperature Tc

Near Tc, the spontaneous magnetization takes small values L 0 ( T )  0 and we can use
binomial expansion in eq. (4):
3
x
tanh x  x 
3
3 3
 Tc  1  Tc  Tc 1 2  Tc 
 L 0 (T )   L 0    L0   1   L0  
 T  3  T  T 3  T 
1/ 2 ≈ 1 because Tc ≳ T
Therefore, we finally obtain   T 

L 0 (T )   3 1  
 
  Tc 
Statistical Physics - Tatek Y. 129
Near T = 0 K

When T  0 , L 0 (T )  1

We then have the following asymptotic relationship for T/Tc << 1

 2Tc 
L 0 ( T )  1  2 exp    Eq. 5
 T 

L 0 (T )
1

T
0 1 Tc

Statistical Physics - Tatek Y. 130

Field free (B = 0) configurational energy and specific heat

The energy is 1
(T )  
2
U 0
qJN L 0
2

The specific heat takes then the following form

U
2
(T ) dL0 Nk L 0
C 0 (T )    qJN L 0 
0
2
Eq. 6
T dT T 
 
 T c  T

1  L0
2
Tc
For T  T c U 0
(T )  C 0 (T )  0

Value of the specific heat at Tc as approached from below

Nk  3 x 3 T
C 0 ( T c  )  lim  Nk (x  1 )
1  x
2
x 0 2 Tc
 (1  x )
1  3x

This means that the specific heat displays a discontinuity at the transition point.

Statistical Physics - Tatek Y. 131

Value of C0(T) around T = 0

Let us insert equation (5) in equation (6)

2
  2Tc  
Nk  1  2 exp     Let us note
  T 
C 0 (T )   2Tc 
2
X  2 exp   
T 
   T 
 T c  T
2

  2Tc   Tc
1   1  2 exp     C 0 (T )
  T 
Nk
The previous relation then becomes
2
3
 Tc 
C 0 ( T )  Nk   1  X 
2
1  1  X 
2
 2
 T 
≈ 2X
2
 Tc   2Tc 
 C 0 ( T )  4 Nk   exp   
 T   T 

and lim C 0 ( T )  0 T
T 0
0 1 Tc
Statistical Physics - Tatek Y. 132
Configurational entropy of the system at Tc
Tc
C 0 (T ) d T
S 0 (T c )   T
0

dL0
But since C 0 ( T )   qJN L 0
dT
Tc
dL0
The entropy becomes S 0 ( T c )   qJN  L0
T
0
And we use the following relation
qJ L 0 1 k 1
L 0  tanh   tanh L0
kT T qJ L 0
To evaluate the integral
1
1
S 0 ( T c )  Nk  (tanh L 0 )d L 0  Nk log 2  S 0 ( k )
0

Let  be the total number of states

S 0 ( k )  Nk log     2
N

This means that all the microstates are equally likely.

Statistical Physics - Tatek Y. 133

Magnetic susceptibility of the system

 M   L  N 1  L (B,T )
2 2

 ( B )     N    
 B T  B  k 
T  Tc 1  L ( B , T )
2

To establish the above relation we used
 qJ L   B  qJ
L  tanh   and Tc 
 kT  k
Let us evaluate the approximated expressions of the magnetic susceptibility

(i) For L  1 , two cases:

- T > Tc or T ≳ Tc and B ⟶ 0
N
2

 0 (B)  (T  T c )
1 Curie-Weiss law
k
- T ≲ Tc and B ⟶ 0
N
1/ 2 2
 T   0 (B)  (T  T c )
1
We use L 0 ( T )   3 (1  ) to obtain
  2k
 Tc 
(ii) For T ⟶ 0
4N
2
 2Tc 
 0 (T )  exp    and lim  0
(T )  0
T  0
kT  T 
Statistical Physics - Tatek Y. 134
Relation between L and B at T = Tc

 qJ L   B  1 qJ L   B
L  tanh    tanh L 
 kT  kT

We use the approximate expression of L at T = Tc and obtain

1/3
B  3B 
3
L Tc
L   L   L   
 
3 Tc kT c  kT c 

6.5 Ising model in the first approximation

It is also called the Bethe approximation: a 0 spin has q nearest neighbors. The interaction
between 0 and its q nearest neighbors is taken into account exactly. The interaction of these
neighbor spins with the other spins is taken into account through a mean molecular field B’.

The corresponding Hamiltonian is

q q

H q 1
  B 0
  ( B  B ' )  j
 J   0 j
j 1 j 1

interaction between
interaction between neighbor spins / (external field interaction between
central spin/ external field + molecular field) central spin/ neighbor spins
Statistical Physics - Tatek Y. 135
B is the external field
B’ is determined by the condition of self consistency :  0
  j

Partition function


 1  q q

Z   exp  kT  B
 0
  ( B  B ' )  j
 J   0 j


 0 , j   1 
  j 1 j 1 

 q q
 Eq. 7
  exp    0
 (   ' )   j
    0 j 
 0 , j   1  j 1 j 1 

B B ' J
with   ,  ' ,  
kT kT kT

Equation (7) can be written as the sum of 2 terms related to  0

 1

Z  Z 
 Z 

with
 q

Z 
  exp     (    '   )   j 
 j  1  j 1 
Statistical Physics - Tatek Y. 136
Let us write x     '   and let us consider Z+:


 q

Z 
 e  exp  x   j 
 j  1  j 1 
There is 1 way to have q spin up and 0 spin down

There are q ways to have q  1 spins up and 1 spin down

….. …. …

q!
There are ways to have q  i spins up and i spins down
( q  i )! i !

Therefore Z+ can be expressed as a binomial expansion

q

Z 
 e

e qx
 qe
( q 1) x

 ...  e


q!
e
(qi) x

i0 ( q  i )! i !

e   2 cosh x
  x q  q
 e  e  e
x

which finally gives 

2  
q
Z 
 e cosh   ' 

Statistical Physics - Tatek Y. 137

Average value of the central spin
(1 )  Z  (  1)  Z Z  Z
 0

 

 

Z Z
Average value of any of the neighbor spins
q
1
 j

q
  j
j 1

1 1     
q q q

   
 
 exp    (   ' )       0 
q Z  j 1
j  0 j j  
   0  1  j 1 j 1   

1  1 Z 
  
q  Z  '

Using the last equality, we obtain

 j
 Z  tanh    '   Z 
tanh    '  

The average values of the spins are equal, therefore

 0
  j
 Z 
1  tanh    '    Z 
1  tanh    '  
Statistical Physics - Tatek Y. 138
Next, we replace Z+ and Z- by their respective expressions in the previous relation

 2 cosh    1   
q
e   '  tanh   ' 

 2 cosh    1   
q
 e   '  tanh   ' 

The above relation can be written as follow

 2 cosh    2 cosh    
 q 1
e   '    '  sinh   ' 

 2 cosh    2 cosh    
 q 1
 e   '    '  sinh   ' 

 2 exp     '    2 exp    '  

q 1
And we finally obtain
2 '  cosh    '  
e    Eq. 8
 cosh    '   
kT
Equation (8) allows to find ’ which in turn gives B '   '

Spontaneous magnetization at zero external field
B
It occurs when    0
kT
Statistical Physics - Tatek Y. 139
 q 1
Equation (8) becomes
2 '  cosh  '  
e   
 cosh  '   

And ’ is equal to q 1  cosh  '  

 ' log   Eq. 9
2  cosh  '   

In the absence of spin-spin interaction J = 0 so  = 0 yielding ’ = 0

It exists a range of values of  for which ’ = 0: 0     c

  ' 0

In order to find c, let us expand equation (9) around ’ = 0:

  ' 
3

 '  ( q  1 ) tanh    '  sech   ... 

2
Eq. 10
 3 
For any value of , we still have the solution ’ = 0

Condition for a non-zero solution

  ' 
2

Using equation (10) we have 1  ( q  1 ) tanh  1  sech 2   ... 

 3 
 '
2
which gives 1
sech   1  0
2
Eq. 11
3 ( q  1 ) tanh 
Statistical Physics - Tatek Y. 140
From equation (11), we obtain

1  1  1  q 
( q  1 ) tanh   1     c
 tanh 

 
 log  

 q  1  2  q  2 
Critical temperature
2J 2J 1
 c
  T  Tc  Curie temperature of
kT c k  q  the system
log  
 q  2 
Using equation (11), we can determine ’ for T ≲ Tc
1/2
  1 
 '   3 cosh  c  1  
2

  ( q  1 ) tanh   
1/2
  ( q  1 ) tanh   1  
  3 cosh  c  
2

  ( q  1 ) tanh  
≳1
 ( q  1 ) tanh   1  
1/2
 3 cosh 
2
c

1/2
 J T 
3 ( q  1 )(    c )
1/2
   3 ( q  1) (1  )
 kT c Tc 
Statistical Physics - Tatek Y. 141
Order parameter L  

Z 
1
Z 
sinh 2  2 '
L  
Z 
1
cosh 2  2  '   exp  2 
Z 

Statistical Physics - Tatek Y. 142