You are on page 1of 5

Physics 141B, Spring 2010

Problem Set #9 Solutions


David Strubbe

1. (Kittel 11.4)
a) Let the energies of the two states be 0 and kB . Then
U=

e/T
1
0e/T + kB e/T
=
k

= kB
B
/T
/T
1+e
1+e
1 + e/T

(1)

The specific heat is


T2 e/T
U
CV =
= kB
2 = kB
T
1 + e/T

2

e/T

b) If T , then the argument of the exponential

1, and e/T 1.

2T

CV kB

2

1
= kB
(1 + 1)2

1 + e/T

2

2

(2)

(3)

2. (Kittel 11.5)
a) Kittel has confused notation in this and the next problem. N is actually the
number of electrons, not the concentration, or else the units are manifestly
incorrect in the expressions to be proven. Let n be the concentration.
For a total concentration n (both spins), the Fermi wavevector is given by
kF = 3 2 n

1/3

(4)

However, if you remove a spin degeneracy of 2 from the derivation, you get
for each spin
kF = 6 2 n

1/3

(5)

The total energy of each band is given by



Z k
Z k  2 2
F
F
V
V
~ k

2
E =
(k) 4k dk = 2
B k 2 dk
2 0
2m
(2)3 0
 2 5

kF
Z k  2 4
F
V
1
~ k
~ k
V
2
3
Bk dk = 2
Bk
= 2
2 0
2m
2 10m 3
0
#
"

5
2

~ kF
V
1
3
= 2
B kF
2
10m
3
#
"

5/3

~2 6 2 n
V
1
= 2
B 6 2 n
2
10m
3
#
"

5/3

~2 3 2 n (1 )
V
1
2
= 2
B 3 n (1 )
2
10m
3


2/3 2
1
1 ~2
5/3
2
3 n
3 n (1 ) nB (1 )
=V
10 2 2m
2


1
3
5/3
nF (1 ) nB (1 )
=V
10
2
1
3
= N F (1 )5/3 N B (1 )
10
2
1
5/3
= E0 (1 ) N B (1 )
2

(6)
(7)
(8)
(9)
(10)
(11)
(12)
(13)
(14)

b)
Etot = E + + E
1
1
= E0 (1 + )5/3 N B (1 + ) + E0 (1 )5/3 + N B (1 )
2
2 

= E0 (1 + )5/3 + (1 )5/3 N B

To find the minimum, we differentiate with respect to :




5
5
Etot
2/3
2/3
N B
= E0
(1 + ) (1 )

3
3
i
5 h
= E0 (1 + )2/3 (1 )2/3 N B
3

(15)
(16)
(17)

(18)
(19)

Make the approximation 1:




Etot
2
5
2
E0 1 + 1 + N B

3
3
3
 
5
20
4
= E0
N B = E0 N B
3
3
9

(20)
(21)

The condition for an extremum is that the derivative is zero, satisfied when
=

9 N B
20 E0

(22)

We can confirm it is really a minimum by checking that the second derivative


is positive.
2 Etot
20
E0 > 0
2

9
The total magnetic moment is given by



1
1
+

MV = N N =
N (1 + ) N (1 + ) = N
2
2
2
2
2
2
9 N B
9N B
9N B 10
3N 2 B
= N
=
=
=
20 E0
20E0
20
3N F
2F

(23)

(24)
(25)

The actual magnetization is


M=

3n2 B
2F

(26)

3. (Kittel 11.6)
a) If the number of electrons is N , the number of pairs of electrons is 21 N (N 1)
1 2
2 N for N 1. Then the exchange energy for each spin is

2

1
1
1
1
2

V = V
N (1 ) = V N 2 (1 )2
(27)
Ex = N
2
2
2
8

b) The total energy is now


h
i
h
i 1
Etot = E0 (1 + )5/3 + (1 )5/3 V N 2 (1 + )2 + (1 )2 N B
8
(28)
h
i 1

= E0 (1 + )5/3 + (1 )5/3 V N 2 1 + 2 N B
4
(29)
3

The derivative with respect to is


i
Etot
5 h
= E0 (1 + )2/3 (1 )2/3

3
20
E0
9

1
V N 2 N B
2
1
V N 2 N B
2

(30)
(31)

The extremum is at
=

N B
21 V N 2

(32)

20
9 E0

The magnetization is
M=

N 2 B
3N 2 B
N 2 2 B
=
=
1
1
20
2
2
2F 32 V N
9 E0 2 V N
3 F 2 V N

(33)

c) For the extremum to be a minimum, we must have the second derivative positive. It is actually a maximum, and hence unstable, if the second derivative
is negative.
2 Etot
20
1
E0 V N 2
2

9
2

(34)

It is negative if
20
E0 <
9
40E0
V >
9N 2

1
V N2
2
4F
=
3N

(35)
(36)

4. (Kittel 11.7)
a) The total energy is

U=

2sinh kB T
e/kB T + e/kB T

=
= tanh

/k
T
/k
T
B
kB T
e B +e
2cosh k T

(37)

The specific heat (at constant ) is






U

2
2
C=
sech2
=
= kB
sech
2
T
kB T
kB T
kB T
kB T
4

(38)

b) The average specific heat is


1
C =
0

0
0

1
C () d =
0

kB

1
0 kB T 2

kB T
Z 0

2

sech2

d
kB T

(39)

2 sech2

d
kB T

(40)

In the limit kB T 0 , the argument 0 /kB T 1, and


sech

1
1

1
=
/k T = e/kB T
= /k T

/k
T
B
B
kB T
e
+e
e B
cosh kB T

(41)

Then the average becomes


C

1
0 kB T 2

2 e2/kB T d

(42)

By two rounds of integration by parts, this is



1
1
1 2 2

C
kB T 20 e20 /kB T kB
T 0 e20 /kB T
0 kB T 2
2
2

2T
1
1 3 3 kB
1 3 3  20 /kB T
k
T
=
T
1
kB T e
4
0 kB T 2 4 B
40

(43)
(44)

Another valid approach is to non-dimensionalize the integral via x = /kB T


into
2 T Z 0 /kB T
2T Z
kB
kB
k2 T 2
2
2

C=
x sech x dx
x2 sech2 x dx = B
(45)
0 0
0 0
0 12
where kB T 0 means the upper limit of integration x0 1. The integrand is very small for large arguments, so we add only a negligible contribution by integrating to infinity, removing all dependence on T from the
integral.
Note that the approaches give different prefactors (both presumably incorrect), so we are only looking at the dependence on T .

You might also like