Cornell University, Physics Department

PHYS-3341 Statistical Physics

Fall 2014
Prof. Itai Cohen

Solutions to Problem Set 9

David C. Tsang, Yariv Yanay

9.1

One Dimensional Paramagnet

(a) We have 1 = 1 , 2 = 1 /2 , 3 = 3 /1 2 = 3 /2 .
Induction: Assume i = i /i1 , for all i > 1. if this is true for all i n, then we must
have
n+1
n + 1
n+1
= Qn i =
.
n+1 = Qn
1 i=2 i1
n
i=1 i
QED.
(Since i = 1 we can also write: i = i1 i .
(b)
"

Z =

N 1
J X
exp
(
(1 2 . . . i )(1 . . . i + 1))
2kT i=1
1 ...N
"
#
N 1
X
J X
exp
(
(1 2 . . . i )2 i+1
2kT
1 ...N
i=1
#
"
N
X
J X
exp
i
2kT
1 ...N
i=2
"
#
N
X
J X
2
exp
i
2kT i=2
...

This is twice that of a partition functino of N 1 particles with B 0 = J/2B and

J 0 = 0.
Z = 2(eJ/2kT + eJ/2kT )N 1 = 2N coshN 1 (J/2kT )
(c) We have 1 i+p = i+1 . . . i+p since (i . . . i )2 = 1.

This gives us
J

1 i+p

PN

i+1 . . . i+p e 2kT j=2 j

=
P
J PN
j=2 j
2kT
1 ...N e
P
Qi+p
Jj /2kT
i+1 ...i+p
j=i+1 j e
= P
Qi+p
Jj /2kT
i+1 ...i+p
j=i+1 e
P

1 ...N

(eJ/2kT eJ/2kT )p
= tanhp (J/2kT )
J/2kT
J/2kT
p
(e
+e
)

This implies that

ln i i+p = p ln(tanh(J/2kT ))
which gives an exponential dependence of the spin-spin correlation function on the
distance p, where the correlation length is given by
= (ln[tanh(J/2kT )])1 =

1
ln[coth(J/2kT )]
2

9.2

Bosons in a Harmonic Trap

Solutions posted next week This problem is very similar to the Bose gas in a container, with
hamiltonian
1
p2
+ m 2 r2
(1)
H=
2m 2
(a) The energy of one particle in a single energy level nx , ny , nz is given by
3
3
n = ~ + (nx + ny + nz )~ = ~ + n~
2
2
where n = nx + ny + nz .
Let nr = the # of bosons in the rth energy level. The grand canonical partition
function is then
X
=
e(n1 E1 +n2 E2 +...) e(n1 +n2 +...)
n1 ,n2 ,...

!
e(E0 )n0

n0 =0

Y
nx ,ny ,nz =0

!
e(E1 )n1

n1 =0

!
e(E2 )n2

...

n2 =0

by the geometric series

1 e(n )

(b) We know
= 1
N

ln

From part (a) we have

ln =

1 e(n )

nx ,ny ,
nz =0

Therefore

X
1 X e(n )
1

N=
=
(
)
(
)
n
n
nx ,ny , 1 e
1
nx ,ny , e
nz =0

nz =0

(c) The chemical potential (N, T ) must be less than the ground state energy 0 for all
T , otherwise the lowest state would have negative occupation number! At very low
T , the chemical potential goes to 0 . Thus the largest number any excited state can
have is 1/(e(n 0 )/kT 1). As T 0 the number of particles in the excited states will
fall below N , and there the remaining particles must occupy the ground state. The
highest temperature at which the condensate exists is referred to as the Bose-Einstein
tranisition temperature and we shall denote it by Tc .
3

(d) In order to find the critical temperature Tc we examine the number of particles in the
excited state. To find an upper bound on the total number of bosons outside of the
groundstate we want to consider
ex = N
N0 =
N

X
nx ,ny ,
nz =0

1
e(n )

1
e(0 )

it will be useful to rewrite this as

ex =
N

X
n=1

g(n)
e(n ) 1

where the degeneracy in choosing n = nx + ny + nz is given by



m+2
(m + 2)(m + 1)
g(n) =
=
2
2
since the number of ways of divvying up m parcels of energy into 3 dimensions is
equivalent to choosing 2 dividers out of a list of m + 2 possible energy units and
dividers.
Now, we approximate the sum by an integral,
Z

(n + 2)(n + 1)
1
1 X (n + 2)(n + 1)

dn (~n+ 3 ~)

Nex =
3
(~n+
~)
2 n=1 e
2 n=1 e
2
2
1
1
"
#

2


Z
1 1
1
x
x
=
dx
+3
+2
2 ~ ~ ex+y 1
~
~
"
#
2


Z
1
x
x
1 1
dx
+ (3 2y)
+ (y 1)(y 2)
=
2 ~ ~ ex+y 1
~
~
where x = ~m, y = 32 ~ .
The integrals evaluate to
Z
1
dx x
= ln(1 e ) = ln(1/) + O()
e 1
Z
x
2
dx x
=
O()
e 1
6
Z
x
dx x
= 2(3) O() 2.4 O()
e 1

And so, if ~  1, y  1 (more on these assumptions later)

ex =
N

1
[(3) + O(~ + y)] .
(~)3
4

Bose Einstein condensation occurs when there is a macroscopic number of atoms in

the ground state, that is, when
ex < N

N
or
kTc =

~
1/3 .
N
1/3
((3))

Are the parameters we assumed to be small actually small? We took ~  1; we

3 . Thats very small. We
just found that at the critical temperature c ~ 1/N
0 = y1 or
also know that the number of particles in the ground state is given by N
e 1
y = ln(1 + N10 ) N10 . This is a small number even if the condensate has 10 particles,
, so its even smaller.
but at this point it has a number on the order of N
(e) What happens in lower dimensions? The calculation is similar, except that g(n) changes.
1/2 .
In two dimensions this simply leads to kTc2D = ~3 N
Things are more complicated in one dimension. Here, g(n) = 1 and the integral is
Z
1
1 ln(1/~)
1 1
1D

dx x+y

Nex =
2 ~ ~ e
1
2 ~
from which we find
kTc1D 2~

N
.
ln N

This is a condensation temperature, and we can calculate it for any finite system. But
what happens as we increase the number of particles in the system? If we simply take
, the term N
/ ln N
still increases. However, we should increase not just the
N
particle number but all extrinsic sizes, like the volume. In particular, if we write the
Hamiltonian as
p2
1
p2
1  x 2
H=
+ m 2 x2 =
+ V
2m 2
2m 2
L
we can rewrite
kTc3D
kTc2D
kTc1D

p
 1/3
~ V /m N
=
,
((3)1/3 L
p
 1/2
~ V /m N

=
,
L
3
p
 
2~ V /m N
=
.

L
ln N
5

In two or more dimensions, the critical temperature remains constant. In one dimension, the critical temperature drops logarithmically as we increase the size of the
, L
, N
/L
= const., it goes
system, and in the thermodynamic limit, with N
to zero. So we say that a harmonic oscillator is only condensed at zero temperature in
one dimension.
The calculation is very similar for an unconfined gas, except that the correct quantum
numbers to use are the momenta, k, and the energies are given by  k 2 . We get
ex =
N

kmax

kmin

d k 
e k 1
D

max

min

 2 1
d 
.
e 1

Compare this to what we had above:

In three dimensions we have  in the numerator, and we will have regular condensation.
In two dimensions we have 1 in the numerator, like we had for a one-dimensional
trap above. This means we only have a BEC at zero temperature.
In one dimension, things are even worse - we have 1/ in the numerator. This
means there is no condensation in the thermodynamic limit!
2

9.3

Velocity Distribution of a Fermi Gas

(a) Clearly vx = 0 since each particle is as likely to be moving forward as backwards.

We do not have a velocity distribution but we have calculated the energy dependent
density of states g(E). From class and from Reif, we know for a Fermi gas g(E) E 1/2 .
We also know
3
1
(2)
E = m(vx2 + vy2 + vz2 ) = mvx2 .
2
2

Thus we have vx2 = 2E/3m.

Let g(E) = CE 1/2 where C is a constant to be determined by the normalization
condition
Z
Z
2
(3)
1=
g(e)dE =
CE 1/2 dE = C3/2
3
0
0
since only the levels up to energy are filled at T = 0. Hence C = 32 3/2 .

Now to calculate E:
Z mu
Z
3 3/2 3/2

E dE
E =
g(E)EdE =
0
0 2


3
2
3
3/2
5/2
=
mu
mu
= .
2
5
5
Thus we have
vx2 =

2
.
5m

(4)

(b) If the temperature were not zero, but still much smaller than the Germi temperature,
vx2 would increase. This can be understood with the following picture:

For T 0, electrons move from low energy states to higher energy states. The temperature dependence of this increase in barE can be estimated as follows
E [g()(kT )](kT ) T 2
7

(5)

where the term in square brackets is approximately the number of energy levels that
shift, which are shifted by the amount (kT ).
The exact calculation is more difficult (see Reif p.396-397), which gives the result
Vm
1 2
(kT )2 2 2
E =
N 3
2 ~
With vx2 =

2
E
3m


1/3
2N
3
.
V

(6)

the increase in vx2 .

2

9.4

Electron Gas
Reif 9.17: Consider an ideal gas of N electrons in a volume V at absolute zero.
(a) Calculate the total mean energy E of this gas.
(b) Express E in terms of the Fermi energy .
(c) Show that E is properly an extensive quantity, but that for a fixed volume V ,
E is not proportional to the number N of particles in the container. How do
you account for this last result despite the fact that there is no interaction
potential between the particles?

(a) We have
3/2
V 1/2
2m

E =
()d
where () =
2
~
2
0
where weve included the spin in the density of states. Carrying out hte integral we
see


3/2
3/2
Z
2m
2m
V
V 2 5/2
3/2

E=
 d =

2
2
2
~
2 0
~
2 2 5
Giving us a chemical potential
Z

(T = 0) =

~2
(3 2 N/V )2/3
2m

(7)

and the average energy

5/3
~2 V
2
E =
3
N/V
2m 5 2
(b) Rewriting we see that
3
E = N
5
which matches with our result from problem 10.3.
(c) If we start adding electrons to the box without changing the volume we are foricing the
electrons to fill up the lowest energy states, without expanding the number of these
energy states. Thus we expect that the energy should not scale proportionally with
number, if the volume isnt also scaled. Indeed we see that if we scale both volume
and number by we get

5/3
~2 V
2

3
N/V
= E
E(N, V ) =
2m 5 2

5/3
~2 V

5/3
E(N,
V)=
3 2 N/V
= E
2
2m 5
which clearly is not linear.

(8)
(9)

2
9

9.5

Two Particles

(a) Distinguishable Particles Possible states:

2
0

3

0
3
0
3


1
0

3
0

0
3

3

Z = e0 + e2 + e6 + 2e + 2e3 + 2e4

(b) Fermions are indistinguishable and do not want to be in the same energy level. Possible
states:
2

3
3

1
0
0


Z = e + e3 + e4
(c) Bosons are indistinguishable but can be in the same energy level. Possible states:
2
0

3
3
3

1
0
0
0

3

Z = e0 + e + e2 + e3 + e4 + e6

10