Statistical Physics HS 14

Solution Sheet 2 Prof. M. Sigrist

Exercise 2.1 Classical Ideal Paramagnet

a) We define the magnetization M = nmH, with n = n+ − n− and N = n+ + n− , such that
n± = N 2±n . The (discrete) phase space area is the number of combinations of moments that
give the same magnetization, so Ω(n) = n+N!n!− ! . We then use the Stirling’s approximation
ln(N !) = N ln N − N + O(ln N ) and ignore terms of order log N .

log Ω(n) = log(N !) − log(n+ !) − log(n− !)

       
N +n N +n N −n N −n
≈ N (log N − 1) − log −1 − log −1 ,
2 2 2 2
1
where we neglected the term 2 log(π 2 (N 2 − n2 )). The entropy then reads:
N kB  n  n  n  n 
S = kB log Ω(n) = N kB log(2) − 1+ log 1 + + 1− log 1 −
2 N N N N
(1)

From the differential dS = (1/T )dU +(M/T )dH we can obtain the temperature as follows.
E
We insert n = − Hm to get S(E, H) and differentiate keeping n explicitly:
   
1 ∂S ∂n ∂S 1 ∂S
= = =−
T ∂E H ∂E H ∂n Hm ∂n
   
N kB 1  n 1  n 1 1 kB N +n
= log 1 + − log 1 − + − = log
2Hm N N N N N N 2Hm N −n
 
kB N Hm + E
=− log .
2Hm N Hm − E
Inverting the above equation yields E = −N Hm tanh(βmH), with β = 1/(kB T ). In order
to obtain the magnetization we first calculate the partial derivative
     
∂S ∂n ∂S E ∂S EkB N Hm + E
= =− 2 = log ,
∂H E ∂H E ∂n H m ∂n 2H 2 m N Hm − E
and then
 
∂S E
M =T =− = N m tanh βmH). (2)
∂H E H
The susceptibility reads:
N m2 β
 
∂M
χH = = . (3)
∂H cosh2 (βmH)
It is useful to study the magnetization and the susceptibility in the two regimes βmH 
,  1 (see Fig. 1). When βmH  1 (small field and/or large temperature limit) tanh x ≈
x − O(x3 ) and cosh x ≈ 1 + O(x2 ) such that the magnetization grows linearly in the field,
i.e., according to the Curie law of independent moments

M ≈ N m2 βH = χH with χH ≈ N m2 β = χ. (4)

1
Abbildung 1: Magnetization (2) as a function of magnetic field strength H scaled to kB T . The
dashed (solid) line is the asymptote at small (large) field (m = 1).

When βmH  1 (large field and/or small temperature limit) tanh x ≈ 1 − e−2x and the
magnetization tends to saturate, i.e.,
M ≈ mN (1 − 2e−2βmH ). (5)

It is also
interesting to
consider the heat capacity for constant external field H. From
dU = ∂U ∂U ∂M ∂M



∂T H dT + ∂H T dH and dM = ∂T H dT + ∂H T dH such that
h  ∂U  
∂M
 i h  ∂U  
∂M
 i
δQ = dU − δW = dU − HdM = −H dT + −H dH.
∂T H ∂T H ∂H T ∂T T
(6)
The heat capacity at fixed H is then given by:
N kB (βHm)2
   
∂U ∂M
CH = −H =2 , (7)
∂T H ∂T H cosh2 (βmH)
where we used ∂T f (β) = −kB β 2 ∂β f (β). Note that both the susceptibility and the heat
capacity are exponentially suppressed at low temperature ∼ T −α e−2Hm/kB T with α = 1, 2,
which is indicating a freezing of the degrees of freedom.
b) In order to determine the thermodynamics of the ideal paramagnet in the canonical en-
semble, we calculate the partition function:
N hX i 
Y N
Z= e−βHmσ = 2 cosh(βmH) = Zm
N
. (8)
i=1 σ=±

We can now easily calculate all the thermodynamic functions, e.g, the free energy:
1
F (T, H, N ) = − ln Z = −kB T N ln Zm (9)
β
and the internal energy
U (T, H, N ) = −∂β ln Z = −N mH tanh(βmH). (10)
From the free energy (9) we obtain the magnetization and the susceptibility, which are
equal to the ones obtained in the micro-canonical case (2), (3).

2
Exercise 2.2 Classical Ideal Lattice Gas
Micro-canonical case: We need to calculate the number of microscopic realizations that yields
the same energy

E = NA EA + NB EB (11)

with NA(B) the number of particles on the sites with energy EA(B) . Note, that NA ≤ N1 , and

N1 = NA + NB . (12)

At zero temperature, as N1 < N2 we expect that when EA < EB , NA = N1 with zero entropy,
as there is only one configuration possible (all atoms sitting on the NA sites). On the other
hand, when EA > EB , it will be favourable for the system to occupy the B sites, i.e., NA = 0
and since N2 > N1 the number of allowed configurations will be larger than 1 and the residual
entropy non zero. More quantitatively,
N1 ! N2 ! N1 ! (N − N1 )!
Ω = ΩA ΩB = = ,
(N1 − NA )!NA ! (N2 − NB )!NB ! (N1 − NA )!NA ! (N + NA − 2N1 )!(N1 − NA )!
(13)
where we used N = N1 + N2 and N1 = NA + NB . We immediately see that when NA = N1 ,
NB = 0 and Ω = 1, while when NA = 0 Ω = ΩB > 1. The entropy reads S = kB ln Ω, with

ln Ω = N1 ln N1 + 2N1 − 2(N1 − NA ) ln(N1 − NA ) − NA ln NA + (N − N1 ) ln(N − N1 )−
 (14)
− (N + NA − 2N1 ) ln(N + NA − 2N1 ) .
In order to obtain the temperature, we write
1 ∂S ∂NA ∂S 1 ∂S
= = = , (15)
T ∂E ∂E ∂NA EA − EB ∂NA
such that
1 kB  
= 2 ln(N1 − NA ) − ln(N + NA − 2N1 ) − ln NA (16)
T EA − EB
which implies
(N1 − NA )2
= e−β(EB −EA ) . (17)
NA (N + NA − 2N1 )
At zero T , when EA < EB the R. H. S. of the equation above is zero and this can only be
satisfied when N1 = NA . On the other hand, when EA > EB the R. H. S. diverges, implying
NA = 0, as N + NA − 2N1 = 0 yields NB = N2 which is not possible. Therefore, the distribution
NA (EA ) goes from N1 when EA < EB to zero when EA > EB in a step-like fashion at zero
temperature. For larger T , the step is smeared out over an energy interval ∼ kB T .
Grand-canonical case: It is straightforward to write the partition function in the following way:

Z = (1 + e−β(EA −µ) )N1 (1 + e−β(EB −µ) )N2 (18)

as each one among the N1,2 sites can be either occupied or empty. The quantity µ fixes the
particle number and results from the shifting of the Hamiltonian H = HA + HB → H =
HA + HB − µ(NA + NB ) = H̃A + H̃B . Therefore, the internal energy reads:

e−β ẼA e−β ẼB !

U = −∂β ln Z = N1 ẼA + N2 ẼB = NA ẼA + NB ẼB , (19)
1+ e−β ẼA 1+ e−β ẼB

3
where ẼA = EA − µ and ẼB = EB − µ are the shifted energies according to the shifted Hamil-
tonian. From the last equivalence, we obtain

NA = (N1 − NA )e−β ẼA and NB = (N2 − NB )e−β ẼB (20)

such that
NB (N1 − NA ) (N1 − NA )2
= = e−β(ẼB −ẼA ) = e−β(EB −EA ) , (21)
NA (N2 − NB ) NA (N + NA − 2N1 )
as in the micro-canonical case discussed above. Alternatively, one could obtain the Boltzmann
factors (20) by maximizing the phase space count (13) with respect to NA with the energy
(11) and particle number (12) constraints imposed through Lagrange multipliers. The multi-
plier β fixing the total energy can be shown to be equal to the inverse temperature via the
thermodynamic relation
1 dS ∂S ∂NA ∂S ∂NB
= = + = kB β. (22)
T dE ∂NA ∂E ∂NB ∂E

Exercise 2.3 Classical Ideal Gas in a Harmonic Trap

a) In the micro-canonical ensemble, the connection to thermodynamics is provided through
the phase space volume
Z
Φ(E) = ΛN dp dq, (23)
H(p,q)≤E

with
N  2 
X p~
H(p, q) = i
+ a~qi2 . (24)
2m
i=1
√ √
We perform the rescaling, Pi = pi / 2m and Qi = qi a, such that the Hamiltonian is
simplified
N h
X i
H(P, Q) = P~i2 + Q
~ i2 (25)
i=1

as well as the phase space integral

 3N/2 Z
2m
Φ(E) = ΛN dP dQ, (26)
a H(P,Q)≤E

since dp = N
Q 3
QN 3
i=1 d pi and dq = i=1 d qi . We then need to calculate, as already explained
in the lecture, the volume of a sphere in 6N dimensional space, i.e.,
 3N/2
2m
Φ(E) = ΛN C6N E 3N , (27)
a

with
π n/2
Cn = . (28)
Γ( n2 + 1)

4
 In order to obtain the thermodynamics, we can then refer to the lecture, with the substi-
tutions 2m → 2m/a and E 3N/2 → E 3N , i.e., the entropy
h 1 r 2m πE 3 i
S(E, N ) = N kB ln + 4N kB . (29)
N a 3N h
Inverting the above relation we find:

3N 4/3 h
r
a h S i
U (S, N ) = E = exp − 4/3 (30)
π 2m 3N kB

such that the equation of state can be obtained as follows,

 
∂U U
T = = → U = 3N kB T, (31)
∂S N 3N kB

which expresses the equipartition law.

b) Within the canonical ensemble, we need again to calculate the partition function. It reads:
Z "N Z #" N Z #
−βp2i /2m qi2
Y Y
−βH(p,q) 3 3 −βa~
Z = ΛN dp dq e = ΛN d pi e d qi e =
i=1 i=1
Z N Z N
2 +y 2 +z 2 )/2m 2 +y 2 +z 2 )
= ΛN dxdydz e−β(x dxdydz e−aβ(x = (32)
 3N/2
2m
= ΛN [πkB T ]3N .
a

We can then obtain all thermodynamic functions, as in Ex. 1, e.g., the free energy:

1 h 1 r 2m πk T 3 i
B
F (T, N ) = − ln Z = −N kB T ln − N kB T (33)
β N a h
and the caloric equation of state,

U (T, N ) = −∂β ln Z = 3N kB T. (34)

c) In the grand-canonical ensemble, the thermodynamics for fixed chemical potential µ and
varying particle number N is given through the grand partition function,
∞
X
Z= z N ZN , (35)
N =0

with the fugacity z = exp(βµ), while ZN is the partition function of the corresponding
canonical ensemble (32), i.e., with given N . We then obtain:
#3N
∞ ∞ r 2m πk T 3 N
"r " #
X 1 βµN 2m πkB T X 1 βµ B
Z= e = e =
N! a h N! a h
N =0 N =0
" # (36)
r 2m πk T 3
B
= exp eβµ .
a h

5
We are then able to calculate all thermodynamic functions, i.e., the grand potential
r
1 βµ 2m π
3
Ω(T, V, µ) = −pV = − ln Z = −e (kB T )4 . (37)
β a h
In order to define compressibility, we exploit the Gibbs-Duhem relation

G(T, p, N ) = µN → SdT − V dp + N dµ = 0, (38)

where G is the Gibbs free energy. We can then write:

   
S ∂µ ∂p
dµ = vdp − dT → =v , (39)
N ∂v T ∂v T

where v = V /N . One then obtains

N 2 ∂µ
     
∂µ ∂N ∂µ
= =− (40)
∂v T ∂v ∂N T V ∂N T

while for the R.H.S. of (39)

     
∂p ∂V ∂p ∂p
= =N (41)
∂v T ∂v ∂V T ∂V T

such that one can conclude:

N 2 ∂µ
       
∂p ∂µ ∂p
− =N → N = −vV . (42)
V ∂N T ∂V T ∂N T ∂V T

According to the definition of isothermal compressibility,

 
1 ∂V
κT = − , (43)
V ∂p T

that quantifies the normalized reduction in volume when changing the pressure at fixed
temperature, we obtain:
 
v ∂N
κT = . (44)
N ∂µ T

In the grand-canonical ensemble, we replace N with hN i, which can be calculated as

follows:

hN i = z∂z ln Z = ln Z = −Ωβ. (45)

Therefore,
 
v ∂z ∂hN i v ∂z
κT = = = vβ. (46)
hN i ∂µ ∂z T z ∂µ