Homework 15 (Due Monday, October 28th)

1. Consider a two-dimensional ideal Bose gas (i.e., particles are on a surface).

(a) Show that
ln Q =

Agd
g2 (z) gd ln(1 z)
2

where Q is the grand canonical partition function, A is the area of the container (i.e. two-dimensional
volume), gd is the degeneracy factor for the energy levels and g2 is a Bose integral.
(b) Show that the number of particles in the excited states diverges as z 1 for T = 0 (i.e., no
Bose-Einstein condensation for a two dimensional boson gas).

2. (a) Show that for a 2-dimensional Fermi gas in an area A,

N
gd
gd
= 2 f1 (z) = 2 ln(1 + z)
A

and for a 2-dimensional Bose gas,

N
gd
gd
= 2 g1 (z) = 2 ln(1 z)
A

Note that there is no Bose-Einstein condensation in the 2-D Bose gas.

(b) Show that for a 2-dimensional Fermi gas in an area A,
U=

gd AkT
f2 (z)
2

and that the corresponding result for a 2-D Bose gas is,
U=

gd AkT
g2 (z)
2

For the next two problems read Section 7.4 of Pathria and Beale, which presents a more complete
treatment of phonons than what I presented in lecture. Note that they consider the more general case
where the density of states, g(), depends on the properties of the solids lattice; in lecture I only
discussed the Debye approximation (see eqn. (7.4.15).
3. Solve problem 7.25 in Pathria and Beale. You may nd it helpful to review the results in Section
3.8.
4. Using the result derived in problem 7.25, solve problem 7.26 in Pathria and Beale.

5. For the Ising model, mean eld theory tells us that the average value of s is found by solving the
transcendental equation,
s = tanh(sTc /T )
where Tc is the critical temperature. Solve this equation numerically and graph s versus T /Tc for
at least twenty values of temperature from T /Tc = 0.01 to 2. You can use any standard root nding
method (e.g., bisection, Newtons method) or a built-in routine (e.g., MATLABs fzero) to solve the
transcendental equation. Your graph should look something like Fig. 12.7 or Fig. 12.10 (for H = 0).

Solutions
1. (a) Using the fundamental denition of the grand canonical partition function for an ideal Bose gas,
ln Q

states

ln(1 zei )

i
states

= gd ln(1 z)

ln(1 zei )

i (=0)

For a two-dimensional system the sum converts to integrals as

states

i (=0)

gd
h2

dr

d
p=

gd
A(2)
h2

dp p
0

where the area A comes from the integration over position and the factor of (2) comes from the angle
integration of the momentum. Since = 12 mv 2 = p2 /2m,

2
2Agd
ln Q = gd ln(1 z)
dp p ln(1 zep /2m )
2
h
0
Using
ln(1 x) =
gives

xn
n
n=1

2
2Agd z n
ln Q = gd ln(1 z) +
dp p enp /2m
h2 n=1 n 0

The integral is now easy to do; specically,

dx xeax =
2

1
2a

so
ln Q

= gd ln(1 z) +

( )

2Agd z n m
h2 n=1 n n

= gd ln(1 z) +

2mkT Agd z n
h2
n2
n=1

which completes the derivation.

(b) Using
(
N

= z
=

(
ln Q = z

]
) [

Agd z n
gd ln(1 z) + 2
n=1 n2

Agd z n
gd z
+ 2
1z
n=1 n

gd z
Agd
+ 2 g1 (z)
1z

= Ngr + Nex
=

where Ngr and Nex are the number of particles in the the ground state and excited states, respectively.
But the Bose integral g1 may be written as
(
g1 (z) = ln(1 z) = ln

1
1z

Unlike g3/2 , which has a maximum value of g3/2 (1) 2.612, g1 as z 1.

2. (a) To formulate the grand canonical partition function for a 2-D ideal quantum gases we can start
with the general expression,
ln Q =

(
)
ln 1 + zei

i=1

where = +1 for fermions and 1. Well use this to nd N since, in general, the average number of
particles in the system is

(
N =z

)
ln Q

For a two-dimensional system the sum converts to integrals as

states

i (=0)

gd
h2

dr

d
p=

gd
A(2)
h2

dp p
0

where the area A comes from the integration over position and the factor of (2) comes from the angle
integration of the momentum. Since = 12 mv 2 = p2 /2m,

2
2Agd
ln Q =
dp p ln(1 + zep /2m )
h2
0
We wont worry about special treatment of the ground state since we know that theres no Bose-Einstein
condensation for a 2-D Bose gas. Using
ln(1 x) =
gives
ln Q =

xn
n
n=1

2
2Agd ()n z n
dp p enp /2m
2
h
n
0
n=1

Notice that this summation identity also tells us that f1 (z) = ln(1 + z) and g1 (z) = ln(1 z) from
the denitions of the Fermi and Bose integrals as summations.
The integral is now easy to do; specically,

2
1
dx xeax =
2a
0
so

( )

2Agd ()n z n m
2mkT Agd ()n+1 z n
ln Q =
=
h2 n=1
n
n
h2
n2
n=1
4


Since = h/ 2mkT , using the denitions of the Fermi and Bose integrals as innite series we get,

Agd f2 (z) for = +1

ln Q = 2
g (z) for = 1
2

Finally, using the recurrence relations, zf /z = f1 and zg /z = g1 , gives the desired result.
(b) To get the energy we just have to evaluate,
( )

U =
ln Q
z
Using the result from part (a),
2Agd
U=
3

d
d

f2 (z) for
g (z) for
2

= +1
= 1

Since
d
d
h
kT
=
()1/2 =

d
d
2
2m
we get the desired result.
3. From the results in Section 7.4 of Pathria and Beale,
(
)
(
)

h
CV (T ) =
=
g()d
T V,N
T V,N
exp(
h/kT ) 1
and since exp(h/kT ) 1 + h/kT for large T ,
(
)

h
CV (T ) =
g()d = k g()d
T V,N
h/kT
Combining these,

[(

h
=
k
h/kT ) 1
0
V,N exp(
]T =
[
h
=
kT
g()d
exp(
h/kT ) 1 T =0

{CV () CV (T )}dT

}
g()d

dT

Using the expansion about x = 0,

1 a
a
= + ...
1
x 2

eax
we have

{CV () CV (T )}dT =

1
h g()d
2

which is the zero-point energy (see discussion at the top of page 207) since the integration
the equivalent to the summation over the discrete vibration levels.

g()d is

Finally, the physical interpretation of this result is that if we add energy to a solid to raise its
temperature from 0 to the amount of energy required is

U =
CV (T )dT
0

For classical harmonic oscillators CVclassical = CV (T = ), which is a constant. So the shaded area in
the graph equals, U classical U quantum . It takes less energy to raise the temperature of the quantum
harmonic oscillators because they already have some energy, namely the zero point energy.
4. Using the Debye approximation for g(), we get,
(
)
D

1
1
9N 2
9
9
h g()d =

h
d = N hD = N kD
3
2
2
D
8
8
0
The mean value of any function f () over frequency is

f ()g()d
f () =
g()d
D
2 d
3
= D
= 0 D 2
4

d
0

so in the Debye approximation,

Finally,
1
1
3
3
h
= h = hD = kD
2
2
8
8

5. Your graph should look something like Fig. 1. I obtained this by solving the transcendental equation
using Newtons method (see MATLAB program below). For each value of T , the program takes 50
random initial guesses for the root and performs 10 iterations of Newtons method. The graph shows
the nal results (with the 50 initial guesses landing on one of the three roots).
% Ising - Program to solve mean field equation in Ising model
clear all;
Nplot = 30;

% Number of temperatures to plot

Tplot = linspace(0.01, 2., Nplot);

Ntry = 50;

% Number of initial values to try

Niter = 10;

% Number of iterations of Newtons method

for i=1:Nplot
T = Tplot(i);

% Value of T/T_c

fprintf(Doing T = %g\n,T);
for j=1:Ntry
6

s = 2*rand(1) - 1; % Try initial random value between -1 and +1

for k=1:Niter
f = tanh(s/T) - s;

% f(s) = 0

dfds = (1/T)*sech(s/T)^2 - 1;

% df/ds

s = s - f/dfds;

% Newton root finding

end
Splot(i,j) = s;
end
end
plot(Tplot,Splot,*);
xlabel(T/T_c); ylabel(\langle s \rangle);

0.8

0.6

0.4

0.2

0.2

0.4

0.6

0.8

0.2

0.4

0.6

0.8

1
T/T

1.2

1.4

1.6

1.8

Figure 1: Graph of solutions for s in the mean eld approximation for Ising model.