PH4211 Outline Solutions 2007

Question 1

a) S is the entropy. Ω is number of microstates in the macrostate. [2]

b) For an isolated system all microstates are equally probable. So probability of

observing a macrostate is proportional to the number of microstates corresponding to
it. State observed is the most probable one, so it is the state of greatest entropy. Upon
removing a constraint the system will evolve to a more probable state – it will evolve
to a state of greater (maximal) entropy. [2]

c) Ωa(Ea) is no of microstates in system a when it has energy Ea. And Ωb(Eb) is no of

microstates in system b when it has energy Eb. Total number of microstates of the
composite system is Ω = ΩaΩb. Let Et be the total energy of the composite system,
denote Ea by E and then Eb is Et – E, so that
  a ( Ea )b ( Eb )  a ( E)b ( Et  E) .
Must maximise this by varying E. So differentiate and set equal to zero
da db
b  a  0.
dE dE
or
d ln a d ln b

dE dE
but S = k lnΩ, so we get Ta = Tb, where
dE
T .
dS
Total derivatives here, since only heat energy is allowed to flow. [4]

Total energy of the composite system is fixed but energy can flow back and forth
between the two. Probability of a energy partition is given by Ω. It will be sharply
peaked at the equilibrium energy value. [1]

d) This is the square root of the average of the square of the deviations of E from its
mean value. It has the dimensions of energy and measures the extent of the deviations
from the mean. [2]

e)
E  
2
σ E2   E 2  2E E  E
2
E
[2]
 E2  2 E  E  E2  E
2 2 2

f)
1  E j / kT
Pj 
e
Z
is probability of Ej state. Here Z is the normalisation constant, known as the partition
function – since probabilities must sum to unity.

1
Mean value given by summing over probabilities
1 1
E   Pj E j   E j e j .
 E / kT
[2]
Z j Z j
g)
Starting from
1
E   E je j
 E / kT

Z j
we see that we could get at E 2 by differentiating with respect to temperature so that
another Ej comes down in the summation. It is simpler to get the Z on the left hand
side first.
E  e j   E je j .
 E / kT  E / kT

j j

We differentiate this with respect to temperature (at constant volume):

 E E 1
 e j  2  E j e j  2  E 2j e j .
 E / kT  E / kT  E / kT

T j kT j kT j
This is then divided by the partition function, to give
 E
2
E E2
  ,
T kT 2 kT 2
or
E
E 2  E  kT 2
2
.
T
This may be written as
E 2  E  kT 2 CV
2

so that
 E  kT 2CV .
Or do it in terms of β = 1/kT. Easier. [3]

h) Since CV is extensive, proportional to N, then fractional fluctuations

E N
 1 N .
E N
So fractional fluctuations go to zero in the thermodynamic limit. [2]

2
Question 2

(a) Classical expression for partition function – integrating over p and q. Exponent is
Hamiltonian / kT. Hamiltonian is total energy T + V, both expressed as functions of p
and q. The powers of h are to specify the size of the cells in phase space. The N!
accounts for indistinguishable particles – multiple occupancy is not considered. [3]

(b) Ideal gas partition function is

 pi2 
   kT
1  
3N 
Zid  e  i 2m 
d3 N p d3 N q
N !h
since here U = 0. But now can do integral over q, giving VN:
 pi2 
   kT
VN  
3N 
Zid  e  i 2m 
d3 N p
N !h
So Z then factorises as Z = ZidQN where
 U  qi , q j  kT
1
QN  N  e i j d3 N q .
V
The VN thus arises from the integration over q. [4]

(c) F = -kTlnZ and p = -F/V. So p = kTlnZ/V. Then

  ln Zid  ln QN 
p  kT   
 V T , N V T , N 
but first part gives ideal gas behaviour (or can show explicitly) so that
 N  ln QN 
p  kT   . [4]
V V T , N 
(d) Argument of exponential in partition function is either zero or infinity for q > σ or
q < σ. So there is no dependence on T. So partition function has no dependence on T.

Then
p N  ln QN
 
kT V V N
and the rhs is independent of temperature. This means that the only free variables for
the rhs are V, N. The lhs is intensive; V/N is the only intensive combination.
So p/kT = fn(N/V) [3]

pV V  ln QN
(e)  1 .
NkT N V N
Clearly  ln QN V depends on σ and can depend on N/V. Overall we require that
V
 ln Q V be dimensionless so σ and N/V must combine as Nσ3/V.
N

-- Virial expansion, Padé approximants, molecular dynamics. [3]

(f) No transition to a liquid since the fluid can’t be self-bound as there is no attractive
part to the interaction. But could be a solid phase; molecular dynamics indicate so. [3]

3
Question 3

(a) Order parameter – measure of the order in the system, zero in disordered phase.
Order parameter varies discontinuously in a first-order transition and continuously in
a second-order transition. [3]

(b) Order parameter is polarisation P. Odd terms in the free energy expansion vanish
if the lattice has inversion symmetry (or because P is a vector while F is a scalar).
Expansion terminated at 6th order. Need the 6th term because the 4th term may be
negative, while we require the system to be stable / φ bounded. [3]

(c)
f (φ)
f (φ) h h
s

t c
φ
φ

a l

h: high temp. t: transition temp. c: critical temp. l: low temp.

First order Second order [4]

(d) If F4 coefficient is negative then we need the F6 term (with a positive coefficient)
so that φ is bounded. The sextic curve is the first-order form above.
If F4 coefficient is positive then since φ is assumed small the F6 term can be neglected
and the system will be stable. And the quartic curve is the second-order form above.
[3]

(e)
f (φ)

t
φ
Δφ Transition state

At the transition temperature the three minima are level. Write:

F  F2 2  F4 4  F6 6
The transition point is characterised by the conditions
bg
F  0 and
dF
d
0

or

4
 F2 2  F4 4  F6 6  0
2 F2  4 F4 3  6 F6 5  0.
We know we have the   0 solution; this can be factored out. The others are found
from solving the simultaneous equations
F2  F4 2  F6 4  0
F2  2 F4 2  3F6 4  0.
The solution to these is
F4
2  
2 F6
or
F4

2 F6
as we know that F4 is negative.

This gives the discontinuity in the order parameter at the transition since  will jump
from zero to this value at the transition. Thus we can write
F4
  .
2 F6

  0 as F4  0 as expected. [4]

(f) F2 vanishes at the critical point since the free energy is then a pure quartic. Lineat
temperature dependence is simplest and F2 must be +ve in the high-temp disordered
phase.
F  a T  Tc   2  F4 4  F6 6 .
Entropy:
F
S ,
T
so that
F
S   0  a 2 .
T
The latent heat is
L  Ttr S
 aTtr  2 .
So that latent heat as
F4
L  aTtr .
2 F6
L  0 as F4  0 and transition becomes 2nd order. [3]

5
Question 4

(a) i Order parameter is magnetisation M.

ii Order parameter is non-conserved
iii Rotational symmetry broken
iv Continuous symmetry broken
v Transition is 2nd order. [5]

(b) B = Bext + b , so that

M  M B b
 tanh  0 ext 
M0  N kT 
but b = λM and we want the spontaneous magnetisation in the absence of the external
Bext. Then
M  M M 
 tanh  0 
M0  N kT 
which can be written as
M  M Tc 
 tanh  
M0  M0 T 
where
Tc   M 02 Nk .
Tc is the transition temperature / Curie temperature. [9]

(c)
1.0

M/M0

0.5

0.0
0.0 0.5 1.0 T/Tc

M is continuous at the Tc so the transition is 2nd order. [2]

(d)   NkTc M 02 so b  NkTc M M 02 . Put M M 0 ~ 1 .

Then b ~ NkTc M 0 . M 0  2 N B - here μB is Bohr magneton. So
kTc 1.4 1023 1043
b~   788T .
2B 2  9.27 1024
Dipole field is many orders of magnitude less than this. Explanation is Heisenberg
exchange interaction. [4]

6
Question 5

Essay question.

Must cover microscopic and macroscopic description of states. Very large number of
microstates compatible with a macrostate necessitates the adoption of a statistical
approach.

Microcanonical – Isolated system – Boltzmann entropy S = klnΩ – all microstates

equally likely.

Canonical – thermal contact with a reservoir (T replaces E as proper variable) – Gibbs

entropy S = –kΣjPjlnPj – probability of microstate  exp – Ej/kT.

Can obtain canonical from microcanonical by regarding (canonical) system plus

reservoir as isolated (microcanonical).

Partition function Z = Σj exp – Ej/kT. Helmholtz free energy F = –kTlnZ.

F = E – TS, so dF = –SdT – pdV + μdN. From this we obtain S, p, μ from
S  F T , p  F V ,   F N .

So we can obtain all thermodynamic properties from the partition function – and that
just needs knowledge of the microscopic energy levels.