Solutions to Problems in Condensed Matter Physics 2

(Textbook: Michael P. Marder, Condensed Matter Physics, Wiley, 2000)

Contents
Chapter
Chapter
Chapter
Chapter
Chapter
Chapter
Chapter
Chapter

page

6 and 7 (electrons)
11 (phonons)
12 (elasticity)
17 (transport phenomena)
22 and 23 (optical properties)
24 (ordering)
25 and 26 (magnetism)
27 (superconductivity)

1
6
8
10
16
21
26
33

(Niels Bohr Institute, February 2011)

Problems in Chapter 6 and 7 (electrons)

Solutions to the problems in Chapter 6

6.3 Pressure of a Fermi gas at zero temperature
The number of electrons N and the internal energy U , in the volume V , are

N =V

D()f ()d,

U =V

D()f ()d,

(1)

The Fermi distribution function f () and the density of states (per unit volume)
D() are

1
2m3
(2)
, D() = A , A = 2 3 .
f () = ()


e
+1
At zero temperature

1/3

2 kF2
, kF = 3 2 n
(3)
2m

Introducing the step function and D() = A in (1), we may determine U (0) and
an alternative expression for A:

F
3n
3/2
5/2
A d = 23 V AF A = 3/2 ,
U (0) = V 25 AF = 35 N F (4)
N =V
0
2F

f () = (F ),

F =

i.e. U F kF2 V 2/3 and the pressure at zero temperature is

(3 2 )2/3 n5/3
2
U (0)
2U
=
= 25 nF =
(5)
V
3V
5m
in sharp contrast to a classical gas, where the zero-temperature pressure is zero.
P =

6.4 Density of states in low dimensions

The density of states in k-space is Dk = 2/(2)d in d dimensions (6.34)

D() =

[dk]( k )

Dk ( k )dk =

2
(2)d

( k )dk

(1)

Introducing spherical coordinates in d dimensions and k = |k|, then

dk =

2 dk

2k dk

4k2 dk

, d=1
, d=2
, d=3

(2)

The factor 2 in the one-dimensional case appears because the one-dimensional wave
vector may assume both positive and negative values, whereas k 0 per denition.

2 k2
k = k =
2m

D() =

k=

2mk
,

dk =


1/2

2
m

2 ( k )
dk =

2 0
2
2 k


1/2



2mk

m
2
2 k

dk

2
k

(3)

2m 1/2
, d=1

m
,

2

d=2

2m3 1/2
,
2
3

d=3

2
dk =
( k )

(2)2 0


2
2 k


1/2

2m
m
2

4 2 k ( k )
dk =

3
2
(2)

1/2

Condensed Matter Physics 2

6.5 Fermi pancakes

A and Lz = d.
Thin layer of Ag: Lx = Ly = L = 106
The density of electrons is the same as the density of atoms (one conduction
A3 , according to the
electron per Ag atom), i.e. n = 5.86 1022 cm3 = 0.0586
properties given in the periodic table on the front page of Marder. The wave
function of the free electrons should vanish at the boundaries z = 0 and z = d.
This condition is fullled if assuming the one-electron wave function to be
(x, y, z) ei(kx x+ky y) sin (pqz) ,

q=

,
d

p = 1, 2, . . .

(1)

The electron states are characterized by the two-dimensional wave vector k =

(kx , ky , 0) and the band index p. The eigenenergies are (k2 = kx2 + ky2 ):
pk =



2  2

2 q 2
k + p2 q 2 10 =
2m
2m

(2)

At T = 0 (T  TF ), as assumed implicitly in the exercise, the occupied states are

all those with energies smaller than F . The number of electrons in the pth band,
Np , is zero if p0 > F . In the opposite case:
Np = L2

k
p
0

2
2
2 kp
2k
dk
=
L
(2)2
2

(3)

where kp is the Fermi wave number of the pth band, i.e. the largest value of k of
occupied states in the pth band, as determined by
pk



2  2
=
kp + p2 q 2 = F
2m

kp2

2mF
=
p2 q 2 = q 2

2

F
p2
10

(4)

1 , and
(a) In the rst case d = 4.1
A or q = /d = 0.76624 A
10 =

(1.054572 1027 0.76624 108 )2

2 q 2
=
eV = 2.2369 eV
2m
2 9.109389 1028 1.602177 1012

(5)

Assuming the Fermi energy to be smaller than the lowest energy of the (p = 2)
band, F < 20 = 410 , then only the (p = 1) states are occupied, in which case N1
has to be equal the total number N of free electrons:
N = L2 d n = N1 = L2

k12
2

k1 =

1
2d n = 1.22866
A

(6)

Introducing this result in (3) we get the Fermi energy and the band width W of
the occupied states:


F = 1k = 10 (k1 /q)2 + 1 = 7.99 eV,

1

W = F 10 = 5.75 eV

(7)

which is in accordance with our starting assumption of F < 20 = 8.948 eV. These
results may be compared with that obtained for bulk Ag:
F (bulk) = W (bulk) =

2  2 2/3
3 n
= 5.50 eV
2m

(8)

Problems in Chapter 6 and 7 (electrons)

(b) In the case of d = 8.2

A, q = /d = 0.38312
A1 and 10 = 0.5592 eV. In order
to determine F in this case, we need to include more bands than the lowest one.
The number of bands turns out to be 3, and


N = N1 + N2 + N3 2d n = q

F
12 +
10

F
22 +
10

F
32
10

(9)

Solving this equation with respect to F /10 , we get F /10 = 11.523, which is
larger than 32 but smaller than 42 , in accordance with the assumption that all
electrons are found in the three lowest bands. Hence the results are:
W = F 10 = 5.89 eV

F = 6.44 eV,
20

(10)

p
2

Band energy (eV)

16
12

8
4

k1

0
0.0

0.2

0.4
0.6
0.8
Wave number k [r.l.u.]

1.0

20

p
4

Band energy (eV)

16
3

12

2
1

k3

0
0.0

0.2

k2

k1

0.4
0.6
0.8
Wave number k [r.l.u]

1.0

Fermi energy and band width (eV)

12

10
8
6

Bulk value

2
0

4
8
12
16
Number of Ag layers (2d/a)

20

The two rst gures show the energy bands as functions of k in

the cases (a) and (b), where the
unit of k is the length of a reciprocal lattice vector 2/a [r.l.u.].
The last gure shows the Fermi
energy (and the band width) as a
function of the number of atomic
layers of Ag. The crystal structure of Ag is fcc with the lattice
parameter a = 4.09
A. This is
very nearly the thickness d assumed in (a), i.e. this case corresponds to a lm with two atomic
layers of Ag atoms.

Condensed Matter Physics 2

Solutions to the problems in Chapter 7

7.1 Normals to surfaces
r = (x1 , x2 , x3 ) = s(t) is the parametrization of a curve lying within the surface
dened by f (r ) = . Since f (s(t)) is a constant , the derivative of this function
is 0:
 f ds
ds
d
f (s(t)) =
= f
=0
(1)
dt
x
dt
dt

Because s(t) may be any arbitrary curve lying within the surface, the same is true
for the curve tangent ds(t)/dt, and (1) is only generally valid if f is normal to
the surface.
7.3 Van Hove singularities
(a) The problem becomes the same as the one considered in problem 7.1 if making
the replacements k r and  f (r), hence   is perpendicular to the
k nk
nk
energy surface dened by nk = .
(b) The energy is assumed to be nk = max k2 , and in the two-dimensional case

D() =

[dk]( nk ) =

1
=
2

0

2
(2)2

0

2k ( max + k2 ) dk
(1)

1
(
( max + k )d(k ) =
)
2 max
2

The density of states is zero if > max and 1/(2) when < max .
(c) In the three dimensional case, the result is

[dk]( nk ) =

D() =
=

1
2 2

0

2
(2)3

0

4k2 ( max + k2 ) dk

k ( max + k2 )d(k2 ) =

1 
max (max )
2 2

(2)

Solution to HSs problem 1

Heat capacity of a two-dimensional electron gas
GaAs/AlGaAs heterostructure (see Marder Section 19.5):
n = 1011 cm2 ,

2 k2
,
2m

m = 0.067me

(1)

The present situation corresponds to the case (a) of the previous problem 6.5 in
Marder, i.e. the gas is purely two-dimensional in the sense that only the (p = 1)
band
needs to be considered, and k is a two-dimensional vector with the length k =

2
kx + ky2 . The most important quantity is the Fermi energy, which is determined
by evaluating N at zero temperature:
n=

N
=
A

kkF

Dk dk =

2
(2)2

k
F
0

2k dk =

kF2
2

kF =

2n

(2a)

Problems in Chapter 6 and 7 (electrons)

The same result is obtained by using that, according to problem 6.4 or equation
(6.35), D() = m /
2 in the two-dimensional case:
n=

F
0

D() d =

m F

2

F =

n
2
m
1

A
Introducing the numbers, and using that kF = 1
F = 3.81 eV when the mass is me , then we get
kF = 0.793 10

A ,

or

kF =

2n

(2b)

corresponds to a Fermi energy

2

0.793 102
F =
3.81 eV = 3.58 meV
0.067

(3)

This Fermi energy corresponds to a Fermi temperature TF = F /kB = 41.5 K.

1) T = 1 K is much smaller than the Fermi temperature and the heat capacity
may be determined by the leading order expression (6.77)
cV =

2
T 2
2
D(F )kB
T =
k k ,
3
6 TF F B

as

D(F ) =

m
kF2
kF2
=
=
(4)

2
2F
2kB TF

In the case of a sample with the area A = 1 cm2 the result is

1
(0.793106 cm1 )2 1.380661023 J/K = 1.1 1013 J/K
6 41.5
(5)

2
2
The explicit result is CV (el) = Am kB T /(3
). Hence, the small value of CV is
not due to the low electron density n but to the two-dimensionality of the system
and the small eective mass.
CV (el) = A cV = A

2) In order to estimate the phonon contribution to the heat capacity we shall use
the Debye model (Section 13.3.2 in Marder). The Debye temperature of GaAs is
D = 344 K (which value is not changed much when some of the Ga ions are
replaced by Al ions). Using (13.70), (13.75) in Marder (2.Ed.) and


0

4 4
x4 ex
dx
=
(ex 1)2
15

12 4
cV =
n k
5 A B

T
D

(T  D )

(6)

According to Table 2.5 (page 27) in Marder, GaAs has the zincblende structure (
diamond structure) with the lattice parameter a = 5.63
A. In this structure there
are 8 atoms per unit cell, and the density of atoms is nA = 8/a3 = 4.48 1022 cm3 .
[The mean atomic mass (periodic table) is (69.72 + 74.92)/2 = 72.3 u, implying a
mass density = 5.38 g cm3 ]. Assuming V = 1 cm3 and T = 1 K, the phonon
contribution becomes


CV (ph) =

12 4
1
4.48 1022 1.38066 1023
5
344

3

J/K = 3.55 106 J/K (7)

which is much larger than the electronic contribution. Utilizing that CV (el) T
and CV (ph) T 3 , the temperature T0 at which the two contributions are equal,
is determined by (T0 in K)
1.10 1013 T0 = 3.55 106 T03

T0 = 1.8 104 K

(8)

Condensed Matter Physics 2

A more fair comparison would be to consider a lm of thickness 100 m, i.e.

A = 1 cm2 and V = 102 cm3 , in which case T0 = 1.8 mK, a temperature within
an accessible range (however, the reduction of the size of the sample makes it more
dicult to determine the heat capacity).

Solution to the problem in Chapter 11

11.2 Zinc in copper
Copper is a monovalent fcc and zinc a divalent hcp (not fcc) metal. At a small
concentration of Zn++ , the crystal structure is fcc until the Fermi surface touches
some point on the edge of the Brillouin zone:
(a) In the nearly free electron approximation, the Fermi wave vector is (nearly)
determined as in the free electron case,

n=

|k|kF

Dk dk =

2
(2)3

k
F
0

4k2 dk =

1 3
k
3 2 F

kF = 3 2 n

1/3

(1)

Dening c to be the concentration of Zn++ ions and a to be the lattice parameter

of the fcc lattice, then the electron density is
4
4
n = 3 [(1 c) + 2c] = 3 (1 + c)
a
a

1/3

12 2
kF =
(1 + c)
a3

4.9109
(1 + c)1/3
a
(2)
(b) The primitive unit vectors of the face-centered cubic lattice are, (2.2):
a1 =

a
(1, 1, 0),
2

a2 =

a
(1, 0, 1),
2

a3 =

a
(0, 1, 1)
2

(3)

and the corresponding primitive vectors of the reciprocal lattice are

b = 2 a2 a3 = 2 (1, 1, 1),
1
a1 a2 a3
a

b = 2 (1, 1, 1),
2
a

b = 2 (1, 1, 1)
3
a

(4)
The boundaries of the 1. Brillouin zone are established by the planes perpendicular
to bi at the distance k1 = |bi |/2 from the origin. Hence the shortest distance from
the origin to the zone boundaries is

3
5.4414
|bi |
=
=
k1 =
2
a
a

(5)

kF (c = 0) is smaller than k1 , and kF (c) becomes equal to k1 at the zinc concentration c = c1

4.9109(1 + c1 )1/3 = 5.4414 c1 = 0.36
(6)
The experimental phase diagram (see Fig. 4 in Chapter 21 of Kittel) shows a
transition from fcc to bcc, for increasing c values, at about the Zn concentration
determined here.

Problems in Chapter 11 (phonons)

(c) The primitive unit vectors of the body-centered cubic lattice are, (2.3):
a1 =

a
(1, 1, 1),
2

a2 =

a
(1, 1, 1),
2

a3 =

a
(1, 1, 1)
2

(7)

and the corresponding primitive vectors of the reciprocal lattice are

b = 2 (1, 1, 0),
1
a

b = 2 (1, 0, 1),
2
a

b = 2 (0, 1, 1)
3
a

(8)

Notice, that the reciprocal of the bcc lattice is fcc, and equivalently, that the
reciprocal lattice of fcc is the bcc lattice.
In the case of bcc there are 2 atoms per cubic unit cell, hence
2
n = 3 (1 + c),
a

1/3

6 2
kF =
(1 + c)
a3

3.8979
(1 + c)1/3
a

(9)

The shortest distance from the origin to the boundaries of the 1. Brillouin zone is

2
4.4429
|bi |
=
=
k2 =
2
a
a

and k2 = kF

when

c = c2 = 0.48

(10)

Once again, this is about the right Zn concentration at which the Cu-Zn alloy
system shows a change of crystal structure (from bcc to a complex structure).
If the atomic density is assumed to stay constant at the transition fcc bcc, the
Fermi wave vector is the same on each side of the phase line, but the lattice parameter a is changed, abcc = (1/2)1/3 afcc = 0.794 afcc . The cubic lattice parameter
a also changes (increases) gradually with the Zn concentration, in between the
phase lines, since nCu = 1.33 nZn , however, all the possible variations of a have
no inuence on the arguments above. -brass is the alloy with c = 0.5, and it
shows an order-disordered phase transition at about 470 C, which we are going
to discuss later on.
Citation from Kittel: Why is there a connection between the electron concentrations at which a new phase appears and at which the Fermi surface makes contact
with the Brillouin zone? We recall that the energy bands split into two at the
region of contact on the zone boundary [Marder, Chapter 8, (8.24)]. If we add
more electrons to the alloy at this stage, they will have to be accommodated in
the upper band or in states of high energy near the zone corners of the lower band.
Both options are possible, and both involve an increase of energy. Therefore it
may be energetically favorable for the crystal structure to change to one which
can contain a Fermi surface of larger volume (more electrons) before contact is
made with the zone boundary. In this way H. Jones made plausible the sequence
of structures fcc, bcc, , hcp with increasing electron concentration.

Condensed Matter Physics 2

Solutions to the problems in Chapter 12

12.4 Elastic constants
(a) The application of a uniform external gas pressure (no shear stress) implies a
uniform dilations of an isotropic solid, where only exx = eyy = ezz are non-zero, or
e =

V
V V0
=
3V0
3V0

(1)

where V0 is the zero-pressure equilibrium volume. Introducing this in the free

energy expression for an isotropic solid (12.23)

F=

1
2

 

2

+ 2

e2

dr =

1
2 (V0

V
+ V )
V0

2

3 V
+ 2
9 V0

2 

(2)
the bulk modulus is found to be
B=V

2F
= + 23
V 2

(3)

(when V V0 ). This result is also obtained from the bulk modulus in the cubic
case (12.19), B = [c11 + 2c12 ]/3, by replacing the elastic constants with the Lame
constants of the isotropic solid, c12 = , c44 = , and c11 = c12 + 2c44 = + 2.
(b) According to (12.31), or Fig. 12.2, Youngs modulus Y is dened to be Y =
zz /ezz , when applying a uniform stress zz in the z direction. In the cubic case,
the diagonal elements of stress and strain tensors are related by, (12.14)-(12.17),

xx
exx
= ce ,
yy
yy
zz
ezz

c11
c = c12
c12

c12
c11
c12

c12
c12
c11

(4)

[the o-diagonal elements relations are = c44 e , where = 4, 5, 6. In the

case of = 4 this equation reads yz = 2c44 eyz ]. In order to determine ezz when
zz is non-zero, we need to determine the inverse matrix. The determinant of c is
D = c311 + 2c312 3c11 c212 = (c11 + 2c12 )(c11 c12 )2 , and the inverse matrix is found
to be

A
= B
B

B
A
B

B
B ,
A

A=

c11 + c12
,
(c11 + 2c12 )(c11 c12 )

B=

c12
A
c11 + c12

(5)

showing that ezz = Azz , hence

Y = A1 =

(c11 + 2c12 )(c11 c12 )

c11 + c12

(6)

[In order to complete the cubic case, then Poissons ratio (12.34) is B/A or
c12
, and the shear modulus is G = c44 ].
=
c11 + c12

Problems in Chapter 12 (elasticity)

12.5 Waves in cubic crystals

(a) Combining (12.14) and (12.16) we may write the free energy of a cubic solid as

F=

1
2

dr c11 e2xx + e2yy + e2zz + 2 c12 exx eyy + exx ezz + eyy ezz


+2 c44

e2xy

e2yx

e2xz

e2zx

e2yz

e2zy



1
2

dr

(1)
e

which shows that the stress-strain relations are

= c11 e + c12


=

e ,

= 2 c44 e (
= ),

u
u
+
r
r

(2)
Introducing these relations in the equations of motion, (12.25), we get

u =




2 u
2 u

u u
= c11
+
c12
+ c44
+
r2
r r
r r
r
=

(3)

(b) Introducing a plane-wave solution



u(r , t) = u0 ei(krt)

(4)

into these equations (3), we get

2 u0 = c11 k2 u0 c12

k k u0 c44

=


=

k2 u0 + k k u0

(5)

leading to the following matrix equation for determining u0 and



c11 k2

+ c44


=

k2

u0 + (c12 + c44 )


=

k k u0 = 0

(6)

(c) In the case of waves propagation along [100], the k-vector is k = (k, 0, 0) and
the matrix equation is diagonal:


c11 k2 2 u01 = 0,

c44 k2 2 u02 = 0,

c44 k2 2 u03 = 0

(7)

The sound velocities are



c11
=
,
k


c44
,
ct = =
k

cl =

u0  k

(longitudinal)

u0 k

(transverse)

(8)

i.e. one longitudinal and two degenerate transverse sound waves. Table 12.1 shows
that c11 = 165 GPa, c44 = 79.24 GPa, and c12 = 64 GPa in the case of silicon.
From the periodic table: the atomic density is n = 4.99 1022 cm3 and the atomic
mass is 28.09 u implying = 28.09 1.66054 1027 [kg] 4.99 1028 [m3 ] = 2328
kg/m3 . Introducing these numbers in (8), we get
cl = 8420 m/s ,

ct = 5830 m/s

(9)

Condensed Matter Physics 2

10

(d) Waves propagating along [111], in which case we assume k = (k/ 3)(1, 1, 1),
and the eigenvalue equation for determining = 2 is found to be

A
B
B
u01
0
B
A
B u02 = 0
u03
B
B
A
0
where

= 2 ,

A = 13 (c11 + 2c44 )k2 ,

The characteristic determinant is



B = 13 (c12 + c44 )k2

(10a)

(10b)

(A) (A )2 B 2 2B (A ) B 2 = (A+2B)(AB)2 = 0 (11)

Introducing the solution = A + 2B in (10a), we get u01 = u02 = u03 or u0  k,
whereas the two other degenerate solutions = A B imply u01 + u02 + u03 = 0 or
u0 k = 0. Hence the sound velocities are


c11 + 2c12 + 4c44

= 9340 m/s ,
cl = =
k
3


c11 c12 + c44

= 5080 m/s ,
ct = =
k
3

u0  k

(longitudinal)
(12)

u0 k

(transverse)

Solutions to the problems in Chapter 17

17.4 AC conductivity
In the presence of a time-dependent (uniform) electrical eld
 =E
 eit
E
0

(1)

we may use the general solution of the Boltzmann equation given by (17.24), before
the integration with respect to t is performed:
g=f

t

 0 e f (t )
dt e(t t)/ eit vk E

(2)

Considering only eects which are linear in the applied eld (the linearized solution
of the Boltzmann equation), then vk = vk (t ) and f (t ) within the integral may be
replaced by their time-independent equilibrium values at zero eld, and the time
integration may be performed straightforwardly
g=f

 e f eit = f
 e f
vk E
vk E
0
1 i

1 i

(3)

Introducing this expression into (17.43)-(17.44) we get the frequency-dependent

 0)
conductivity (valid in the limit of E
0
() = e2

[dk]

f
v v
1 i

(4)

or in the case of a cubic or an isotropic (free electron) system:

() =

ne2
ne2 1 + i

=
m 1 i
m 1 + ( )2

(5)

11

Problems in Chapter 17 (transport phenomena)

17.5 Current driven by thermal gradient

We shall consider a metal subject to a constant temperature gradient


T =

T
, 0, 0
x

k = k = 12 mvk2

and

(1)

According to (17.60), (17.62), and (17.68) the electrical current, in the case of
G = 0, is


j = L12 T
T

1
1 2
L12 = L(1) =
(k T )2  (F )
e
e 3 B

(2)

The assumption of an isotropic mass, (1), implies () to be diagonal, and according to (17.64) the diagonal element is
2

() = e

2
dkDk vk
( k ) = e2

dk D(k ) 13 vk2 ( k )

(3)

2 over all solid angles, at a constant |

k|, is 1/3 of the result
The integration of vk
2
2
2

deriving from Tr v = v and using v = 2k /m , we get

k

() = e2 D()

2
3m

() =


2e2 

D()
+

D
()
3m

(4)

This result is introduced in (2)



2
2

j = 1 (kB T )2 2e D(F ) 1 + F D (F )
e 3
3m
D(F )

T
T

(5)

2 T D( ), (6.77), we nally get

In terms of the heat capacity cV = ( 2 /3)kB
F

2e cV T
jx =
3m x

D (F )
1 + F
D(F )

where the last equality sign is valid only if D()

e cV T
m x

(6)

17.8 Hall eect elementary argument

 is along the x axis leading to a current j in

The Hall eect geometry: The applied eld E
this direction, i.e. an electron (hole) current in the minus (plus) x direction. The magnetic
 is in the minus y direction, when B
 is along z, leading
part of the Lorenz force F j B
to opposite signs of the resulting charge distributions in the electron and hole cases.

Condensed Matter Physics 2

12

(a) The equation of motion, when assuming the Drude model, is



 mv
 + v B
mv = e E
c

(1)

 = (0, 0, B), then we get v B

 =
in the case of electrons with charge e. Using B
(vy B, vx B, 0) and since the current, by geometry, is constrained to be along x,
the steady state is characterized by vx being constant and vy = 0. These conditions
imply


vx
v
mv y = e Ey B 0 = 0 Ey = x B
(2)
c
c
(b) The current is j = (jx , 0, 0) with jx = nevx , and the Hall coecient is

R=

Ey
1
1
v
=
= xB
,
B jx
c B(nevx )
nec

Ey = RB jx

(3)

The electric eld in the x direction may be determined from v x = 0, or

mv x = eEx

mvx
=0

Ex =

me
me jx
mj
j
vx =
= 2x = x

(ne)
ne

(4)

i.e. Ex is determined by the Drude resistivity 1 as in the case of B = 0.

17.9 Hall eect Boltzmann equation
The Boltzmann equation in the relaxation-time approximation is given by (17.17)
and (17.18)
dg
gf
g g  g
=
=
+ r
+k
(1)

dt
t
r

k
When the state is uniform in space and steady in time, this equation reduces to
k g = g f

k

(2)

The semiclassical equation of motion is



 + 1 v B
 ,

k = e E
c k

1 k

k
vk = r =
=

k
m

(3)

where the second equation expresses that the mass tensor is assumed to be isotropic
for simplicity. Like in Problem 17.4 we are only interested in the response (current)
which is linear in the electric eld. This means that g may be replaced by f in
 The linearized
products on the left hand side of (2) which already involve E.
version of the Boltzmann equation (2) is therefore
e
e  f
 (g f ) = g f

vk B
E




c

k
k

(4)

 f = 0 because f is parallel with v , as

when using that vk B
k
k
k
f
f k
f
=
=
vk



k
k k

(5)

13

Problems in Chapter 17 (transport phenomena)

 =
(a) The geometry is the same as applied in problem 17.8, hence we dene B


(0, 0, B) and assume the resulting E = (Ex , Ey , 0) to be perpendicular to B. In
this geometry, we guess that the solution has the form
g = f + a kx + b ky

(6)

Introducing this in eq. (4) and utilizing (5), we get



a kx + b ky
eB 
e
f 
E
k
+
E
k

a
k

b
k
=
x
x
y
y
y
x

m
m c

(7)

Since kx and ky are independent variables, this equation leads to two independent
conditions, which determine a and b to be
Ex c Ey
a=
1 + (c )2

e


m

f
,

Ey + c Ex
b=
1 + (c )2

where we have introduced the cyclotron frequency c =

e


m

(8)

eB
.
m c

The component of the current density j is, according to Marders eq. (17.43),
j = e
=

[dk] vk (gk fk ) = e

em

[dk] vk a kx + b ky

2
2
[dk] a vkx
x + b vky
y

(9)

where the last equality sign follows because the o-diagonal terms vanish, when
the mass tensor is assumed to be diagonal (isotropic). Using the same procedure
as in Marders eqs. (17.44)-(17.50), we have

em

e

2
[dk] vk

m

f
= e2

ne2
2 f
=
[dk] vk
0

(10)

and combining the three equations (8)-(10), we nally get

jx =

Ex c Ey
,
1 + (c )2 0

jy =

Ey + c Ex

1 + (c )2 0

(11)

 is assumed to be along the z axis parallel to the eld, we have

In the case where E
to add a term cz kz to the trial function g in (6). In this situation, the magnetic
eld does not contribute to the Boltzmann equation, and we get jz = 0 Ez . Hence,
for a system with an isotropic mass m , the total conductivity tensor is found to
be

1
c
0
0
eB

,
(12)
=
1
0
c =
c
2
1 + (c )
m c
0
0
1 + ( )2
c

when the magnetic eld B is applied along the z axis.

This result may also be expressed in terms of the Hall coecient R, where
c = RB0 ,

or

R=

1
nec

(13)

Condensed Matter Physics 2

14

 = j, is the inverse of the

The resistivity tensor , dened by the relation E
conductivity tensor and is particularly simple

1 1
=
c
0
0

c
1
0

0
1 1
0 =
0 RB
0
1
0

0 RB
1
0

0
0
1

(14)

which is in perfect agreement with the results derived from the Drude model (problem 17.8). The calculations in section 17.4.8 assume c 1, in which case the
diagonal elements of may be neglected in comparison with the o-diagonal ones,
xx = yy  0 and xy = yx  (c )1 .

Solution to HSs problem 3

Hall eect of a two-dimensional electron gas
A two-dimensional electron gas with an anisotropic dispersion
= a(kx2 + ky2 ) + b(kx4 + ky4 ),

a > 0,

b>0

(1)

Introducing the polar angle in the (kx , ky )-coordinate system of the reciprocal
lattice, the dispersion relation may be written
= a k2 + 14 b (3 + cos 4)k4 ,

kx = k cos ,

ky = k sin

(2)

reecting directly the four-fold, cubic symmetry of the dispersion.

1) The equation determining the constant energy contour is obtained by solving
(2) with respect to k2
2a
k = k () =
b(3 + cos 4)
2

1/2

b
1 + 2 (3 + cos 4)
a

(3)

In
 a2 , the square root may be expanded, and to second order
the case of b
( 1 + x = 1 + 12 x 18 x2 ) the result is


3b
k () 
1 2
a
4a
2

b2
cos 4
4a3

(4)

In the gure below (to the left) I show a constant energy contour, which diers
visible from a circle. It is obtained by a numerical evaluation of (3) (Mathematica
program) in the case of a = b = 1 and = 2 (assuming dimensionless quantities).
The thin line shows the average length of k(). The gure to the right show the
corresponding |k k | as a function of the angle . The gradient, and hence the
velocity vk = k k /
, is perpendicular to the constant energy contour. Notice
that |v | is smallest along the 11 directions, where |k()| has its maxima.
k

15

Problems in Chapter 17 (transport phenomena)

2) The Boltzmann equation in the relaxation-time approximation, (17.17)-(17.18),

dg
gf
g g  g
=
=
+ r
+k
dt
t
r

k

k g = g f

k

(5)

when considering the steady state of a uniform system. The elds are assumed to
 = (E, 0, 0) and B
 = (0, 0, B), and the semiclassical equation of motion is
be E


 + 1 v B
 ,

k = e E
c k

1 k
vk = r =

k

(6)

Introducing this in the right-hand part of (5) we get



 

e 
1


E + vk B


c
where

(g f ) f
+
k
k

gf

f
f k
f
=
=
vk
 k

k
k
k

(7)

(8)

 and the linearized version of (7) is

This gradient is perpendicular to v B,
k

e
f
 (g f ) = g f
vk B
(9)
k


c

k
e  (g f )

, is of
Since (g f ) is going to scale with E, the term neglected, E


k
 = (E, 0, 0) and B
 = (0, 0, B) in (9), we nally get
second order in E. Inserting E
 v
eE
k

f
eB

=
v
vx
evx E


c y kx
ky

(g f )

(10)

3) With the assumption of g f = ky F (), we nd

vy

(g f )
(g f )
dF
dF
vx
= vy ky
vx F vx ky
kx
ky
d kx
d ky
dF
dF

v vx F vx ky

v = vx F
= vy ky
d x
d y

(11)

Condensed Matter Physics 2

16

When introducing this result in (10) and neglecting 1/ , we get

evx E

f
eB
[vx F ()]
=


c

F () =

cE f
B 

(12)

The current in the y direction is then

jy = e

e
cE
[dk]vy (g f ) =
B

e
cE
=
B
=

ecE
B

dkx

f
[dk]vy ky


k

ecE
f
ky
dky Dk
=
(
ky ) 
B

dkx dky Dk f =

ecE
B

dkx dky Dk ky

necE
1
[dk]f =
=
E,
B
RB

f
ky

R=

(13)
1
nec

when performing the y integration by parts, where Dk = 2/(2)2 is a constant and
the boundaries of the integral is the boundaries of the rst Brillouin zone. (13) is
the usual high-eld result for the o-diagonal conductivity yx = 1/(RB).

Solutions to the problems in Chapter 22 and 23

22.2 Faraday rotation
The combination of the semiclassical equation of motion and the Drude relaxation
time model leads to the following equation of motion for the electrons (when
k =
mv = mr)

e
 r
 e E
r = r B
m c
m

or

e
r

 = e E
r + + r B

m c
m

(1)

The (applied) magnetic eld is assumed along z. The electrical eld vector, due
to an incident light wave, is assumed to be circular polarized in the plane perpendicular to z
 = (0, 0, B),
 =E
 = E0 (
x i
y ) eit
(2)
B
E
If the light wave propagates in the positive z direction, i.e. E ei(qzt) , then the
plus (minus) sign in (2) corresponds to the left (right) circular polarization. In the
case of the right circular polarization, the end point of the E-vector is making a
right-handed screw line in the direction of propagation at a certain time t (see for
instance Griths, Introduction to electrodynamics, p. 374).
(a) Introducing the expressions for the elds in (1), the equation may be solved,
with respect to r , by assuming
x i
y ) eit
r = A (

(3)

Because (
x i
y ) z =
y i
x = i(
x i
y ), we get from (1)


eB
1
e
(i) + i A (
x i
y )eit = E0 (
x i
y )eit

m c
m

(4)

or


eE
1
eB
A = 0 (i) + i
m

m c

1

eE0

m 1 i ic

c =

eB
(5)
m c

17

Problems in Chapter 22 and 23 (optical properties)

where c is the cyclotron frequency (section 21.2). The current density and the
conductivity tensor are determined from
j = ner ,


j = E

(6)

 and dening j = j (
y ), then
The calculations show that r = (A /E0 )E
0 x i
+
+

+
+

 + E
 = E
 and j = E
+ + E
 = E
 . The
j = ++ E
+
+

current has the same circular polarization as the electric eld, i.e. the conductivity
tensor is diagonal with respect to the choice of sign, + = + = 0. The diagonal
components are (++ + and )
= ne

A
0
,
=
E0
1 i ic

0 =

ne2
m

(7)

(b) In the case of 1 and c , we get



0
i
0

1+
=
=
c
1 i ic
(i )

1

i
i

 0 1
c

(8)

(c) The dielectric constant, (20.14), is



p2
4i
i

 =1+
 1 2 1
c

(9)

where p is the plasma frequency introduced by (20.32) or (23.7)

p2 =

4ne2
40
=

(10)

The index of refraction, (20.18), is



n
= n + i =

p2
i

1 2 1
c ,
2

(11)

The Faraday rotation of the polarization vector per unit length along the sample
is then found to be
=

p2
p2 c



2 2 c =
n n+ =
2c
2c 2
2c 2

(12)

 is along the
Notice, that the sign of is determined by the sign of c , or B, i.e. if B
negative z direction then is negative. If the active carriers are holes rather than
electrons, we may perform exactly the same calculation except that e is replaced
by e in all expressions (m is then the positive hole mass), hence the sign of ,
like the Hall voltage, depends on the kind of carriers.
Derivation of the expression for the Faraday rotation angle: Assuming the incident
light wave to be linearly polarized along the x axis at z = 0, then we want to nd
the angle of rotation of the polarization vector in the xy plane as a function of z.
1

E(z)/E
x + i
y )ei(q+ zt) + 12 (
x i
y )ei(q zt)
0 = 2 (

(13)

Condensed Matter Physics 2

18


At z = 0, the polarization vector is parallel to x
, E(0)/E
eit . Introducing
0 = x
1
1
= 2 (q+ + q )z and = 2 (q+ q )z, or q+ z = + and q z = ,
then we get

E(z)/E
0 =

1
x
2 (

+ i
y )ei + 12 (
x i
y )ei ei(t)

(14)

The polarization vector at z is determined by the factor in the square bracket,

which is found to be
1
x + i
y )(cos
2 (

+ i sin ) + 12 (
x i
y )(cos i sin ) = x
cos y sin

This expression shows that the polarization vector at z is making the angle z =
= 12 (q+ q )z with the x axis. According to (20.18) q = n /c, and the
rotation angle per unit length, along the direction of light propagation, is
= 12 (q+ q ) =



n n+
2c

(15)

In the case where the carriers are electrons, then > 0, which sign corresponds to
a counter-clockwise or left-handed rotation of the polarization vector along z.
23.3 Helicon waves
(a) From the Maxwell equations (23.71) and (23.72) we nd


2
2


 = B = 1 E + 4 j = 1 E + 4 E
c E
t
c t2
c t
c t2
c t

(1)


Assuming
when utilizing the linear relation j() = ()E().
 =E
 0 ei(krt) ,
E

k = (0, 0, k),

k0

(2)

we get from the left- and right-hand sides of (1)



2
 + 4 (i)E
 = 1 (i)2 E
 =
c (ik) (ik) E
c
c
c

 =  E

E
c
(3)
when introducing the dielectric tensor , (20.14). Leaving out the common phase
factor, we nally get


k k E
 + k2  E
 = 0
(4)
0
0
0
4
1+i

 The conducting sample is assumed

(b) Symmetry analysis of the relation j = E:
to be isotropic (in the shape of a sphere, if surface eects are of importance). In the
 = Bz , the sample, or alternatively the coordinate
presence of a magnetic eld B
axes, may be rotated any angle around the z axis without changing the coordinate
 If
representation (the matrix equation) for the tensor relation between j and E.
the coordinate axes are rotated by an angle of 90 degrees around the z axis, then
(jx , jy , jz ) is renamed (jy , jx , jz ) and (Ex , Ey , Ez ) is replaced by (Ey , Ex , Ez ),
whereas the components determined by the sample are kept unchanged:

jx = xx Ex + xy Ey + xz Ez

j = E + E + E

jy = yx Ex + yy Ey + yz E z
z
zx x
zy y
zz z

jy = xx Ey + xy Ex + xz Ez

jx = yx Ey + yy Ex + yz Ez
jz = zx Ey + zy Ex + zz Ez

(5)

19

Problems in Chapter 22 and 23 (optical properties)

A comparison of the two sets of relations then shows that xx = yy , xy = yx ,

and zx = zy = xz = yz = 0.
[Extension of the symmetry arguments: At zero eld, the isotropic sample is insensitive to a rotation around any arbitrary direction and = xx , which result
shows that xy 0 when B 0. A reection of the system with respect to, for
 along z is reversed, whereas
instance, the yz plane implies that the axial vector B


it is the x components of the polar vectors j and E which change sign. Because of
this dierent behaviour of the polar and axial vectors it is straightforwardly shown
 is reversed, whereas nothing happens with the diagonal
that xy changes sign if B
elements. This symmetry element is in accordance with the general Onsager relation (B) = (B). Hence, in the case of an isotropic sample the diagonal
components may only involve terms quadratic (even powers) in B, whereas the
o-diagonal ones are linear (of odd powers) in B.]
These symmetry relations are consistent with the results obtained in the previous
problems 17.9 and 22.2, and they apply as well for the dielectric tensor , or

xx
 = xy
0

xy
xx
0

0
0
zz

(6)

 + k(k E
 ) = k2 E
 ), the
Introducing this in (4) in combination with k (k E
0
0
0
result is

0
k2 + xx k02
xy k02
0
Ex
2 +  k2
 k 2
Ey0 = 0
k
0
(7)
xy 0
xx 0
Ez0
0
0
zz k02
 , if the characteristic determinant vanwhich only has a non-zero solution for E
0
ishes, or if


2
(8)
zz (k/k0 )2 xx + 2xy = 0
We are going to utilize the solution of problem 22.2 in the following. Here is
 in the xy plane, or
dened by the vector relation j = E
jx + ijy =

 +
(Ex + iEy )

(9)

xx Ex + xy Ey + i(xy Ex + xx Ey ) = (xx ixy )(Ex + iEy )

showing that + = xx ixy and correspondingly, (9) in the solution of 22.2,

 = xx ixy

xx = 12 (+ +  ),

xy = 12 (+  )i

(10)

The latter relations reduce (8) to the diagonal form



zz (k/k0 )2 +

(k/k0 )2  = 0

(11)

(c) The situation is exactly the one considered in problem 22.2, and the result
found there is [combining (8) and (9) in the solution of 22.2]
p2
4i
0

=1+i
1 i ic
1 i ic
p2 1
p2
i
1
=
1

=+
=1+
i1 c

c

 = 1 +

(12)

Condensed Matter Physics 2

20

or nally
 = 1

p2 1 (c /
)
,

1 (c /
)2

c =

eB
,
m c

40
4ne2
=

p2 =

(13)

which result agrees with the expression for xx = (+ +  )/2 given by (23.82).
(d) The atomic density (front page) of aluminium metal is 6.02 1022 cm3 . According to Table 6.1, the number of conduction electrons per atom is 3, i.e. the
electron density in Al3+ -metal is n = 18.06 1022 cm3 . The cyclic plasma frequency is p = 5.64(15) 1015 rad/s, when n = 1022 cm3 and m = m, see (23.3).
[Incidentally, the unit of a cyclic frequency is not Hz but rad/s or just s1 . The
unit Hz (hertz) is reserved for the case of cycles per second, = /(2).] In the
case of Al-metal,
p = 24 1015 s1 ,

p = 3.8 1015 Hz,

p = 16 eV

(14)

The cyclotron frequency may be expressed in terms of the fundamental magnetic

quantity, the Bohr magneton:
B =

e

= 5.788383 109 eV/G
2mc

c =

m
eB
= 2 B B
m c
m

(15)

Assuming m = m and B = 1 T = 10 kG = 104 G, we get

c = 2.8 1010 Hz,

c = 0.12 meV,

c = 18 1010 s1

(16)

The numbers show that c  105 p  p . In the case of c , the fraction in

the expression for  is of the order of 1010 , and the adding of one to this fraction
has no eect at all.
(e) Assuming  p we get from equation (12)
 =

p2
( c + i )

p2
,
( c )

c 1

(17)

In most cases  < 0, which implies a purely imaginary refraction index n

=

 and therefore no light propagation. However, there is one situation, where

electromagnetic waves with  p are capable to propagate through the sample,
which one is determined by the solution + , when < c (or  if the carriers
are holes). Hence, when placing a metal like aluminium in a strong magnetic
eld, it is possible for low-frequency light to propagate through the metal, but low
temperature and high purity is demanded in order to secure that the damping is
small, c 1.
The dispersion relation for these helicon waves is determined by (11) and (17):
 2

p2
(c )

c (ck)2
 2c (ck)2 ,
2
2
p + (ck)
p

p
c
(18)
showing that the dispersion is not linear but quadratic in k in the long wavelength
limit. The name of the waves is a derivative of helix and refers to the circular polarization of the waves (+ corresponds to left circular polarization) the
polarization vector of the helicon wave is forming a helix along the z direction.
2

k =

k02 +

k

21

Problems in Chapter 24 (ordering)

The helicon resonance is related to the Hall eect, and the result may be written
in a simple form in terms of the Hall coecient R = (nec)1 :
=

B|R|
(ck)2
4

(cgs-Gaussian units)

(19)

In the experiments of, for instance, R. Bowers, C. Legendy and F. Rose, Phys.
Rev. Lett. 7, 339 (1961), wavelengths of the order of 1 cm are considered. In the
case of Al-metal, = 1 cm corresponds to a helicon frequency of = 1.77 Hz at
B = 10 kG. In sodium (n = 2.54 1022 cm3 ) this frequency is 12.3 Hz. Bowers et
al. observed a standing wave resonance in a Na sample of length l = 4 mm at 32
Hz and 10 kG. The resonance condition is l = /2, and our simple model predicts
/2 to be about 3.1 mm at the same frequency and eld.

Solutions to the problems in Chapter 24

24.4 Mean eld theory
The antiferromagnetic two-dimensional Ising model on a square lattice. In this
lattice, the number of nearest neighbours is z = 4. The Hamiltonian is supposed
to be

H=


R


R

J R
 R




R

HB R
 =


1
J R
HB R
 R


,
2  


RR

J < 0 (1)

where H = E in the usual representation of R

 = 1. Because J is negative, the
 and R

ground state is not the ferro- but the antiferromagnet. In the sums, R
are nearest neighbours, and the antiferromagnetic ground state is assumed to be
determined by the mean eld values
(R
 , R

) =

 sublattice A
R
 sublattice B
R

( , ),
( , ),

(2)

The two sublattices of the square lattice: A

shown by lled and B by open circles. The g belongs
ure shows the case, where the site at R
to the A-sublattice, in which case R
 = .
  , then
Its four nearest neighbours at positions R
all belong to the B sublattice and R

 = .

(a) In the mean-eld (MF) approximation we get, when inserting (2) in (1),
MF

E=
=



1 
J R
R
 R

 + R

R
  R
 R

 HB

2  



RR

R


R

J R

R
 HB

 

N/2


RA



R

R
 +

zJ + B H R


1
2 N zJ

 

N/2


RB

(3)


1
zJ + B H R
 + 2 N zJ

Condensed Matter Physics 2

22
!

Because of the MF-approximation, the total energy E = R

 ER
 is the sum of the
individual contributions of each site, and the partition function of the total system
Z=


all

"


R


all

ER


"

ER


R

ZR


(4)

is reduced so to become the product of the partition functions ZR

 for the individual
 belongs to the sublattice A
sites, see for instance (6.40). For the sites, where R
(), we get, when omitting the constant energy term 12 zJ ,


Z =

e [(zJ +B H)] =

=1

e (zJ +B H ) = 2 cosh zJ + B H

(5)

=1

which implies that

R
  = =



1 
e (zJ +B H ) = tanh (zJ + B H)
Z =1

(6)

Notice that the omission of a constant energy term does not aect the calculation
of a thermal expectation value. Z and are determined equivalently


Z = 2 cosh zJ + B H ,

= tanh (zJ + B H)

(7)

and the (total) free energy is

F=



N
kB T ln Z + ln Z + N 12 zJ =
2

# 


$ N
N
kB T ln 2 cosh (zJ + B H) + ln 2 cosh (zJ + B H) + zJ
2
2
(8)
The constant energy term neglecting in the partition functions adds to F.

(b) The self-consistent equations determining and are derived above.

(c) In the paramagnetic phase, T > TN , and vanish in the limit H 0.
Expanding tanh to leading order, tanh x  x, we get from (6) and (7)


= tanh (zJ + B H)  (zJ + B H)

= tanh (zJ + B H)  (zJ + B H)

= =

B H
1 zJ

(9)

The expectation values and are equal and are proportional to the eld. The
magnetization is M = (N/V )B ( + )/2 and the susceptibility is dened as
= M/H in the limit of zero eld, and we nd
=

2B
N
2B
N

,
V kB T zJ
V kB (T )

z|J |
= TN
kB

(10)

The paramagnetic Curie, or CurieWeiss, temperature is dened by (24.41),

(T )1 . In the present system is negative (J is negative), and || is, as
we shall see, the same as the Neel temperature, TN , below which the system is
antiferromagnetically ordered at zero eld.

23

Problems in Chapter 24 (ordering)

(d) The Neel temperature is the temperature below which the self-consistent equations for and have a non-zero solution at zero eld. The solution is obtained
by assuming = , and H = 0, in which case we get


= tanh (z|J |)( ) = tanh(z|J | ) = tanh( )

(11)

This equation is the same as considered in the case of the ferromagnet, (24.56), and
it has a non-zero solution, when the coecient 1. The Neel temperature TN
is the temperature at which = z|J | = 1, or TN = z|J |/kB . The paramagnetic
susceptibility, which diverges at the transition, is obtained by applying a eld with
the same symmetry as the ordered phase. The application of a staggered eld,
where HR
 = +H at the A sites and H at the B sites determines straightforwardly
(using = ) the staggered susceptibility to be

stag =

N
N B ( )
2B
=
V
2H
V kB (T TN )

(12)

which diverges at TN .
(e) Spin-ip transition: In the ordered phase, the application of a small eld along
z, in the up direction, does not change = 1 or = 1. However, if the eld
becomes suciently large, the spins antiparallel to the eld are going to ip so to
become parallel to the eld. This rst-order spin-ip transition happens, when the
Zeeman-energy gain is able to compensate for the loss of exchange energy.
Introducing dimensionless quantities, then


+ h 
,
= tanh
t

+ h 
= tanh
,
t

h=

B H
,
z|J |

t=

T
TN

(13)

In the limit of t 0, then the introduction of = 1 in the second equation

implies = 1 if 1 > h > 0 and = 1 if h > 1. The rst equation then predicts
= 1 in both cases, hence assuring that this is the self-consistent solution. Using
ln[2 cosh x] = ln[ex + ex ] ln e|x| = |x| for x , the free energy is found to
be determined by


t
F
+ h
=
f=
ln 2 cosh
N z|J |
2
t
t

)(



+ h
+ ln 2 cosh
t

+ h )) + h ))
1
+)
) 2 =
t
t

(

12

12 ,
0<h<1
h + 12 ,
h>1

(14)
Hence, the free energies of the antiferromagnetic phase f = 1/2 and of the spinipped phase f = h + 1/2 are equal at the transition at h = 1. At a nite
temperature the spin-ip transition is smeared out due to thermal eects. The
gure below shows the numerical calculated magnetization, ( + )/2, as a
function of the eld h at various values of temperature t.

Condensed Matter Physics 2

24

1.2
t
0
0.2
0.4
0.6
0.8
1.0

Averaged spin

1.0
0.8
0.6
0.4
0.2
0.0
0.0

0.4

0.8
1.2
Magnetic field (h)

1.6

24.6 Superlattices
Model of Cu3 Au In the fcc lattice, the sites in the
centers of the six faces are named As and the 8 corners
are named Bs as indicated on the gure by, respectively, black and white spheres. There is 8/8 = 1
Bs site and 6/2 = 3 As sites per unit cell.
The distance between nearest neighbours is a/ 2, and the
coordination number is z = 12. The kind of nearest
neighbours of the two kind of sites are

As -site: 8 As sites and 4 Bs sites.

Bs -site: 12 As sites.
The correct solution to the equations (24.62) and (24.63) is not (24.64), but
f (R
 , R

) =

BB AA
2AB AA BB
(R
R
 + R

)
 R


4
4

(24.64 )

when leaving out the constant term C1 = (2AB + AA + BB )/4. This modication
aects the denition of the eective parameters in (24.67), which should read
B H =

BB AA
z,
2

J=

2AB AA BB

4

(24.67 )

(a) In the present system we have a 3:1 mixture of A and B atoms. The eective
Hamiltonian equals the energy minus the chemical potential times the dierence
between the numbers of B and A atoms, is (leaving out the constant contribution)

H=E



R

R
 = J


R


R

R
 R

B H



R

R


(1)

We are going to use the same choice as Marder, that R

 = 1 or 1 signies a site
occupied by, respectively, a B or an A atom. With this choice, the eective energy
parameters are those given by (24.67 ) above. In the mean-eld approximation
MF

H = J

 
R


R

R
 R

 + R

R
  R
 R

 B H



R

R
 =



R

HR


(2)

25

Problems in Chapter 24 (ordering)

where
 
R


R

 
1 
R
R
R
 R

+R

R
 =
 R

 + R

R
 =

2 


RR

N


3N

s
A



(nn)
R

B
 12 A
  +
s

s
B

R



A
8 A

  + 4 B
 
s

(3)


(B
 ), if R = As

Here N is the number of unit cells (or Bs sites), and R

 = A
s
s
 s ) is the position of an As (Bs ) site. The (grand) partition function of the total
(B
system is the product of the individual partition functions for each site, because
!
H= R
 HR
 , and in the case of an As site (omitting the constant energy term)


ZAs =

[J (8   +4   )+B H ]
As

Bs

(4)

=1

implying that
A
 =
s

[J (8   +4   )+B H ]
1 
As
Bs
e
ZAs =1

= tanh 8 J A
  + 4 J B
  + B H
s

and equivalently

(5)

B
  = tanh 12 J A
  + B H
s

(6)

(b) R
 = +1 for the B atoms and the number of these atoms is equal the number
of unit cells N , whereas the number of A atoms is 3N . These conditions imply

N 3N = 2N


N
3N

R
 = 
B
A

s +
 s = N B
 s  + 3N A
 s


R

s
B

B
  = 2 3 A
 
s

s
A

(7)
(c) The chemical potential, or eectively B H, has to be adjusted so that the two
equations (5) and (6) are in accordance with the relation B
 s  = 2 3 A
 s .
The eective Zeeman term may be eliminated:


1
tanh1 A
  tanh B
  = 8 J A
  + 4 J B
  + B H

s

12 J A
  + B H = 4J B
  A
  = 8J 1 + 2A
 
s

(8)

Introducing an eective order parameter Q and J = |J |, we nally get

1

f (Q) tanh

Q+1
2

+ tanh

3Q 1
2

= 8|J |Q,

Q = 1 2A
  (9)
s

In the disordered phase both the As and the Bs sites are occupied by an A atom
3
1
1
with the probability 34 or A
 s  = B
 s  = (1) 4 + (+1) 4 = 2 . This result is in
accordance with (7) and Q = 1 2A
  = 0. In the completely ordered phase at
s

Condensed Matter Physics 2

26

zero temperature, all A (B) atoms are placed on the As (Bs ) sites implying that
A
  = 1 and B
  = 1 and hence Q = 1.
s

The numerical solution of (9) is discussed

in a Mathematica program. The system
orders at a rst-order transition, when
f (Q) = Q f  (Q) [the line y = Q with
= f  (Qc ) is parallel to the tangent
of f (Q) at Q = Qc , and this line just
touches f (Q) at Q = Qc , when f (Qc ) =
Qc . This then becomes a solution to
(9) if choosing = 8|J|]. The solution
is Q = Qc = 0.3455 and dening the effective temperature scale t = (8|J|)1
then Q jumps from zero to Qc at the
temperature t = tc = 1/f  (Qc ) = 0.4143.
The calculated variation of Q as a function of t is shown in the gure.

Solutions to the problems in Chapter 25 and 26

25.2 Classical electrons in a magnetic eld
The spin degree-of-freedom does not occur in classical physics. The elementary
particles do neither exist, but accepting their existence, they only interact with
the electromagnetic elds because of their charges, and with the gravitational eld
because of their masses. The classical Hamiltonian H of a many-body system (see
for instance Goldstein) is a function of the independent canonical variables (ri , pi )
of all the particles. The canonical impulse pi , independent of ri , is dened in terms
of the Lagrangian L = L(ri , r i , ) as pi = L/r i (or, in terms of the Hamiltonian
ri ).
theory, as the generator of an innitesimal translation ri ri + 
(a) If the external constraints (elds) are independent of time, the Hamiltonian
is equal the total (kinetic plus potential) energy of the system. In this case the
classical partition function is
Z = e F =

1
hn

dr1 dp1 drn dpn e H(r1 ,p1 ,,rn ,pn )

(1)

 only contributes
(b) In the presence of a time-independent electromagnetic eld, E
to the potential energy as determined by the scalar potential. The magnetic eld
 only aects the velocities of the particles via the vector potential A
 = A(
 r ),
B


where B = A. For the ith particle with mass mi and charge qi , the canonical
impulse is
q 
pi = mir i + i A
c

2

1
q 
Ti = 12 mi (r i )2 =
pi i A
2mi
c

(2)

[The magnetic eld needs not to be uniform in space, but only to be constant in
time the case of a spatial uniform eld is discussed in the next problem 25.4].

27

Problems in Chapter 25 and 26 (magnetism)

(c) The canonical impulse pi appears in the Hamiltonian only via the kinetic energy
Ti . Assuming H in (1) to be the Hamiltonian function of the system, when the
magnetic eld is zero (but including the potential energy contributions of the
electric eld), then the introduction of the magnetic eld implies that


H(r1 ,p1 , ,rn ,pn ) H r1 ,p1


q1 
q 
A(r1 ), ,rn ,pn n A(
rn )
c
c

(3)

The partition function Z in the presence of the magnetic eld is then



q
1
 r ),,r ,p qn A(
 r )
H r1 ,p1 c1 A(
n n
n
1
c
d
r
d
p

d
r
d
p
e
1 1
n n
hn


1




= n dr1 dp1  drn dpn  e H(r1 ,p1 ,,rn ,pn ) = Z(0)
h

 =
Z(B)

(4)

The integration with respect to the canonical impulse pi may be performed before
q 
ri ),
the ri -integrations. Introducing the following change of variables pi  = pi ci A(


then dpi = dpi , since A(ri ) is independent of pi . This transformation of variables
 in the partition
therefore leads to the same integral as in (1). The elimination of B
function means, for instance, that the magnetic susceptibility of a classical equilibrium system is zero. This result is called the Bohrvan Leeuwen theorem. Notice,
that it is important for the argumentation that the system is in equilibrium (the
basis for the use of the partition function).
25.4 Quantum electrons in a magnetic eld
 is
The Hamiltonian of an electron with charge e in a magnetic eld B


1
H=
p +
2m

e
A
c

2

(1)

The choice of the Landau gauge

 = (0, B x, 0)
A

 =A
 = (0, 0, B) ,
B

 =0
A

(2)

 in the z direction.
i.e. this vector potential corresponds to a spatial uniform eld B
 is zero, the two quantum mechanical operators A
 and p
Since the divergence of A
commute, p = (
/i) in the r -representation. Introducing this choice of gauge in
(1), we get



2
1
eB
2
2
x + pz
H=
p + py +
(3)
2m x
c
The wave function of the electron is assumed to be
(r ) = u(x) ei(yky +zkz )

(4)

and the eigenvalue equation is

i(yky +zkz )

H = e

2

eB
1
x
p2x +
ky +
2m
c

+ (
kz )

u(x)

(5)

When the common phase factor is deleted, this eigenvalue equation is reduced to
one, which only involves a determination of u(x), and the equation may be written:

p2x
H u(x) =
+ 1 m c2 (x x0 )2 + z u(x)
2m 2

(6)

Condensed Matter Physics 2

where
c =

eB
,
mc

x0 =

ky
,
mc

28

z =

2 kz2
2m

(7)

Equation (6) is the eigenvalue equation for the one-dimensional harmonic oscillator,
displaced to be centered around x = x0 and shifted in energy by the constant (with
respect to x) energy z . Hence, the nal eigenstates are classied by (ky , kz ) and
the occupation number , and the energy eigenvalues are


1
2

,ky ,kz = +

c +

2 kz2
,
2m

= 0, 1, 2,

(8)

This is the result (25.48) utilized in the analysis of the Landau diamagnetism.

26.2 Ferromagnetic ground state

We are going to discuss the ferromagnetic ground state of the Heisenberg model
for an arbitrary integer/half-integer value of S. The Heisenberg Hamiltonian is

H=


ll


 S
= 1
Jll
S
l
l
2


l=l

 S


Jll
S
l
l

(1)

according to (26.21) in Marder. The possibility of Jll

= 0 may be excluded, since
such a term would contribute only by a constant term Jll S(S + 1) to the energy.
 fulll the usual spin-commutator relations
The components of S
l
[Slx , Sl
y ] = i ll
Slz ,

[Sly , Sl
z ] = i ll
Slx ,

(a) Dening the following operator:

Sz =

[Slz , Sl
x ] = i ll
Sly

(2)
(3)

Slz


in (1) is
 S
then the commutator of this operator with a single term S
l
l
 S
] = [ S , S
 ]S

+ S [ S , S

]
[ Sz , S
l
l
z
l
l
l
z
l
(4)

= [Slz , Slx ] Sl
x + [Slz , Sly ] Sl
y + Slx [Sl
z , Sl
x ] + Sly [Sl
z , Sl
y ]
= i Sly Sl
x i Slx Sl
y + i Slx Sl
y i Sly Sl
x = 0

showing that [ Sz , H ] = 0. The Sz -representation for each site is dened by

*
!
*
Slz |ml  = ml |ml , where ml = S, S+1, . . . , S, and Sz l |ml  = ( l ml ) l |ml .
Most of these eigenstates are highly degenerated except for the two singlets,
*
*
*
l |ml = S  =
l | + S  or
l | S . Because Sz and H commute the two
non-degenerate eigenstates of Sz are also eigenstates to H, and

"

|ml = S  = 12 N J (0)S 2

"

J (k) =

|ml = S ,


l

Jll
eik(rl rl
) (5)



 and using S2 = (S

 +S
 )2 = S
2 + S
 2 + 2S
 S

 +S
(b) Dening S = S












R
R
R
R
R

| S2 | =

 S |
S(S + 1) + S(S + 1) + 2|S



R
R
S(S + 1),
S = 0, 1, , 2S

(6)

 | attains its maximum value, when S = 2S, and

 S
which shows that |S



R
R


|} = 1 [2S(2S + 1) 2S(S + 1)] = S 2

 S
Max{|S


2
R
R

(7)

29

Problems in Chapter 25 and 26 (magnetism)

(c) The eigenvalue of H in (5) for the fully polarized states requires that each pair
of operators in (1) contributes by their maximum. Hence, if all exchange constants
Jll
0, then the eigenvalue in (5) is the lowest possible value, i.e. the two fully
polarized states are the two degenerate ground states.
The lowest exited states are linear combination of the states, where all ml = S
except that ml = S 1 for one particular site. Naming the product of these states
for |l, where l is the site at which ml = S 1, then it is straightforward to show
that the eigenstates of H constructed from these states are the Fourier transforms
of |l. These excited states are characterized by the wave vector k of the Fourier
transformation, and the energy dierence between one excited k-state and the
ground state is the spin wave energy, S[J (0) J (k)]. Hence, these lowest excited
states, the one-magnon states, are exact solutions to the Heisenberg Hamiltonian
but this is no longer the case if proceeding to the two-magnon states, where the
eigenvalue of Sz is N S 2.

26.4 Diagonalizing spin waves

The result derived in (26.62) may be written in a reduced form:
=
h
=





1  +
1 
H + N z|J |S 2 =
2a ak + a+ a+ + ak ak cos(k )
k
k k
2z|J |S
2z 


k

a+ ak
k

1
2

a+ a+
k k

1
Bk =
cos(k )
z 

+ ak ak Bk ,

(1)

The operators are dened by (26.51) in terms of al and its hermitean conjugate
a+
l , which are the annihilation and creation operators, also named Bose operators,
of a harmonic oscillator at the site l. These operators obey the Bose commutator
+ +
relations [al , a+
l
] = ll
and [al , al
] = [al , al
] = 0, which imply
1   ikrl + ik
rl

1 
1  i(kk
)rl



al e
, al
e
=
ll
eikrl ik rl
=
e
k
N ll

N ll

N l
(2)
or that the Fourier transforms of the Bose operators, diagonal with respect to the
site index, are Bose operators in reciprocal space
[ak , a+
] =

[ak , a+
] = kk
,

k

[ak , ak
] = 0 ,

[a+ , a+
] = 0
k

(3)

which are diagonal with respect to the wave vector index (notice that the wave
vector k is assumed to be a discrete variable). The Hamiltonian (1) is not diagonal in the number representations of the harmonic oscillators, since it involves
a+ a+ + ak ak . One way to solve an eigenvalue problem is to make a canonical
k k
transformation of the Hamiltonian to one for which the eigenvalue equations have
been solved. In the present case, we want to nd a new Bose operator  , in
k
is
only depends on the number operators + , in which case h
terms of which h
k

k

diagonal. This procedure is called the Bogoliubov transformation:

ak = uk k + vk + ,
k

a+ = uk + + vk k = uk + + vk k

k

(4)

Condensed Matter Physics 2

30

The last equation indicates that uk and vk are assumed to be real functions. This
is a valid simplication in the present case, but it is not a general requirement.
The demand that the new operators should also be Bose operators implies
[ak , a+
] = [uk k + vk + , uk
+
+ vk
k
] = uk uk
kk
vk vk
kk
= kk
(5)
k

and, similarly, [a , a
] = u v
 
v u
 
= 0. Hence, the fulllment of the

k k
k k kk
k k k k
Bose commutator relations for k requires
u2k vk2 = 1 ,

uk vk vk uk = 0

(6)

Inserting (4) in (1) we get

=
h

 

uk + + vk k



uk +
k

#

k

uk k + vk + + 12 Bk



+ vk k + uk k +

uk + + vk k

vk +
k



uk k +

vk +
k

 $

(7)

Collecting corresponding terms and replacing k with k in some of the terms

(utilizing Bk = Bk and k k = k k ), then (7) is reduced to
=
h



2
+
u2k + Bk uk vk + k + v
k + Bk uk vk k 
k

k

$

1
2

+ 2uk vk + Bk uk uk + vk vk

+ +
k k

(8)



+ k k

In order to diagonalize this Hamiltonian we have to search for values of uk and
vk , where the last term in (8) vanishes and the requirements of (6) are fullled.
In order to simplify the search, we make the ansatz u  = u , which is consistent
k
k
with the dierent conditions only if vk = vk . In this case the equations are


2uk vk = Bk u2k + vk2 ,

u2k vk2 = 1

(9)

Introducing v2 = u2 1 in the squared version of the rst equation, we get an
k
k
equation of second degree in u2k . Determining the solution of this equation subject
to the condition v2 > 0 or u2 > 1, we get
k

u2k =

1 + 

2

vk2 =

1 

2

2uk vk =

B

k =

1 B2

(10)

According to the denition in (1), |B | 1 implying 0  1. This is the same

k
k
solution as given by equations (26.63) and (26.64) in Marder, since uk = cosh (k )
and vk = sinh (k ) imply u2k vk2 = 1 and tanh (2k ) = 2uk vk /(u2k + vk2 ) = Bk .
Introducing the solution (10) in (8) we nally get
=
h

 
1
2

k

k + 1 + k +
k

H = N z|J |S 2 z|J |S

k 1 k + =

 


k

Ek = 2z|J |S k ,

1
2

1 k +


k

1
2

k

Ek + k
k

+ k  = nk = [e Ek 1]1

k

k 1 + k + k

(11)

31

Problems in Chapter 25 and 26 (magnetism)

The antiferromagnetic Neel state, where the spins of the two sublattices are either
+S or S is not an eigenstate of the Hamiltonian. The second term of H, discussed by Marder, shows that the ground state energy is smaller than that derived
from the Neel state. Additionally, the magnitude |Sz (l)| is smaller than S at
zero temperature. This zero-point reduction of the antiferromagnetic moment
is determined by the site average of |0|Sz (l)|0| = S 0|a+
l al |0, where |0 is the
product ground state of the N independent harmonic oscillators, k |0 = 0,
S =

1 
1 
0|a+
0|a+ ak |0 =
l al |0 =
k
N l
N 
k


) ,
1  2
1  1 k
1  + )) 
)
0) uk + + vk k uk k + vk + )0 =
vk =
k

k
N 
N 
N  2k
k

(13)

which number is calculated [P.W. Anderson, Phys. Rev. 86, 694 (1952)] to be 0.078
in the case of a simple cubic lattice (z = 6).

Solution to HSs problem 4

Localized atoms: heat capacity and magnetic susceptibility
1) In the case of non-interacting atoms, the total energy is the sum of the con!
tributions from each site E = l El . This implies that the partition function is
the product of the partition functions determined for each site (see also the note
Magnetic energy and domains), and when the atoms are identical:

Z=

N
"

Zl = Z N ,

Zl = Z =

ei = e + 1 + e

(1)

l=1

The free energy is



F = kB T ln Z = N kB T ln Z = N kB T ln e + 1 + e

(2)

The internal energy is U = F + T S, where the entropy is

F
1 Z
= N kB ln Z + N kB T

T
Z T
1 Z
1  i
1 Z
= N
=N
e
U = F + T S = N kB T 2
Z T
Z
Z i i

S=

(3)

Introducing the population factor for the ith level (the probability that this level
is occupied):
pi
!

ei
Z

U =N


i

i pi = N

e e
e + 1 + e

(4)

Notice, that i pi = 1. In terms of the population factors, the thermal expectation

is A
= ! i|A|ip
value of any single site operator A
i
i , and the entropy is S =
!
N kB i pi ln pi .

Condensed Matter Physics 2

32

2) In the limit kB T or  1, we have from (1)

Z  1 + + 12 ()2 + 1 + 1 + 12 ()2 = 3 + ()2


F  N kB T ln 3 + ()2 = N kB T ln 3 1 + 13 ()2 = N kB T ln 3

S=

F
N 2
,
= N kB ln 3
T
3kB T 2

U = F + TS =

2N 2
3kB T

N 2
3kB T

(5)
The heat capacity may be determined either as


2N 2
U
=
C=
T
3kB T 2

S
N 2
= T (2)
C=T
T
3kB T 3

or

2N 2
3kB T 2

(6)

3) The derivative of F with respect to is

1 Z
2N
F
= N kB T
= N kB T (p1 + 0 p2 p3 ) = N (p3 p1 ) 
(7)

Z
3kB T
The last equality is most simply obtained from F derived in (5). In the presence of
a magnetic eld H, applied along the z axis, the Hamiltonian is specied in terms
of the Zeeman interaction

H=

Hl = gB H

Slz ,

S=1

(8)

The Hamiltonian of one of the sites is diagonal in the Sz -representation and the
eigenenergies are

|Sz = 1

|S = 0
z
|Sz = 1

i = gB H =
i = 0
i = gB H =

(i = 3)
(i = 2)
(i = 1)

(9)

Hence the Hamiltonian (8) leads to the case considered when gB H = , and the
derivative of F with respect to is proportional to the magnetization, or
)

(gB ) F
N 2 (gB )2
1 F
=

H,
M =
V H
V
V 3kB T

M ))
N 2 (gB )2
=
=
H )H0
V 3kB T
(10)
The susceptibility is calculated in the limit of H or 0, hence the expression
derived is valid at all T . Notice, that M may also be obtained from
M V = gB

Slz  = N gB

i|Sz |i pi = N gB (p1 p3 )

(11)

The result for in (10) is Curies law in the case of S = 1, see (25.31)-(25.32).
4) Dening x = the heat capacity derived from U in (4) is
U
U
2
C=
=
=
N
T
kB T 2 x
kB T 2 x
=N

2 e + e + 4
kB T 2 (e + 1 + e )2

ex ex
ex + 1 + ex
N

=N

2 /(kB T )
e
kB T 2

2 ex + ex + 4
kB T 2 (ex + 1 + ex )2
for T 0
(12)

33

Problems in Chapter 27 (superconductivity)

The gure to the left shows the reduced heat capacity c = C/(N kB ) as a function
of the reduced temperature t = kB T / the thin lines are the results for c(t) in
the high- and low-temperature limits, (6) and (12). This characteristic behaviour
of the heat capacity is called a Schottky anomaly. The number of levels involved in
the anomaly for an arbitrary value of S is (2S + 1), i.e. 3 levels in the present case
of S = 1. This number may be determined from a heat capacity experiment by
evaluating the entropy in the high temperature limit, since S = N kB ln(2S + 1)
for a general value of S. A measurement of C = T ( S/T ) determines S, and

S=

T
C
0

dT

s =

S
=
N kB

Heat capcity
c

c
0

dt = ln(2S + 1)

(13)

Entropy
s

0.6

0.5

0.8

0.4

0.6

0.3

0.4

0.2

0.2

0.1
1

Solutions to the problems in Chapter 27

27.1 Superconducting sphere
The free-energy density of a type-I superconductor is determined by the critical
 = 0 throughout the volume of the type-I superconductor, when
eld Hc . Since B
surface eects are neglected (L 0 compared to the dimensions of the sample)
i = H
 i + 4 M
 = 0
B


 = 1 H
M
4 i

(1)

Assuming the sample to be a thin needle along the direction of the applied magnetic
 0. In this case,
 = N M
 (along z), the demagnetization eld H
eld H
0
z
d
 =H
 and the magnetization is M
 =
the magnetic eld within the sample is H
0
i
 /(4). According to eq. (2.6) in the note Magnetic energy and domains, the
H
0
magnetic energy density is then,
1
F = GM =
V M

sample

dr
V



 H
 = 1
M
0
4

H
0
0

2
  dH
  = H0
H
0
0
8

(2)

 is the independent
[This energy density corresponds to G in Marder because H
0
variable, see problem 27.2]. The total energy density is the sum of GM and the
eld-independent condensation-energy density, GS . Since the two contributions
just outbalance each other at the critical eld H0 = Hc , we have a measure for GS ,

G = GS + GM ,

GS =

Hc2
8

(3)

Condensed Matter Physics 2

34

If the sample has instead the shape of a sphere, then (1) and (3) still apply, but
 =H
 + DH
 =H
 +H
 =H
 N M
 ,
H
i
0
d
0
z
0
i

Nz
1
=
4
3

(sphere)

(4)

implying
i =
H


H
0
,
1D

G = GS + GM =

H02
Hc2
+
8
8(1 D)

(5)

The internal eld Hi is larger than the applied eld and becomes equal the critical
one, when the applied eld is H0 = (1 D)Hc = 23 Hc . This condition is not
necessarily critical, since the
global free energy of the superconducting sphere is still
negative (as long as H0 < 1 D Hc ). A detail discussion of the intermediate state
of the superconductor, when (1 D)Hc < H0 < Hc , may be found in Landau and
Lifshits, Electrodynamics of continuous media (Volume 8, 2nd Edition, page 189).
In the intermediate range of the applied eld, local thermodynamic instabilities
are going to destroy superconductivity in parts of the sphere such that the eective
D is reduced suciently to sustain superconductivity in the remaining parts. The
assumption that the superconducting domain takes the form of a prolate ellipsoid
with radius a < R and constant height 2R (the largest volume for a certain value of
D) is the most obvious one, but this conguration turns out to be thermodynamical
unstable (the eld in the normal part of the sphere is smaller than Hc near the
end points of the ellipsoid). Nevertheless, domains of superconductivity are going
to be present as long as H0 < Hc .

27.2 Energy of normal-superconducting interfaces

(a) To start with we consider the limit of 0 or L  . Then B = 0 for x > 0

and according to (27.41) the wave function, which vanishes at x = 0 and saturates
at 0 at large x, is


x
(x) = 0 (x) = 0 tanh
2

(1)

The free energy density, the GinzburgLandau energy is, (27.27),

F=

20

L
dx
0

2 4
1
2 0

2
+
( )2
2m

(2)

35

Problems in Chapter 27 (superconductivity)

Here 20 = / is the equilibrium value at large x, and the change in energy per
unit area, in comparison with the homogeneous case, is
LF =

20

dx ( 1) +

2
4
1
2 0 (

2
1) +
( )2
2m

(3)

Introducing Hc2 = 42 /, (27.35), and 2 =
2 /(2m ), (27.31), then (3) may

be written
LF =

Hc2
4

 

0

dx

1
2

1 2

2

+ 2 ( )2 =

Hc2
4

0

dx 1 2

2

(4)

using  = (1 2 )/( 2), (27.40), in the last step. Applying this relation ones
more, we get



Hc2 
Hc2 2 1 
Hc2 2 2
2

2
(5)
dx 1
2 =
1 d =
LF =
4 0
4
4
3
0
which energy per unit area is positive and in agreement with (27.52) [G = F
 = 0].
since B
(b) Next we consider the opposite limit of  L , in which case (x)  0 and
the gradient  may be neglected. In this case the free energy density is

F = FS + FM ,

FM =

dr
V

B2
4e2
A2 20
+
8
2m c2

dr
8V

A2
B + 2
L

(6)

where the rst term is the zero-eld contribution of the homogeneous superconductor, (2) when  = 0, and FM is the magnetic energy density in (27.27), when
2L = m c2 /(16e2 20 ), (27.32), is inserted. The applied eld is assumed to be
 =H
 in the normal metal [this conparallel to the interface, i.e. along z, and B
0
 to be along the
guration leads to the largest possible energy gain]. Assuming A




y direction, A = (0, A(x), 0), then B = A = (0, 0, A (x)) for x > 0. According
to (27.9), Bz = B(0) ex/L for x > 0, where B(0) = H0 . Hence
A (x) = B(x) = H0 ex/L
or

A(x) = L H0 ex/L ( 0 for x )

[A(x)/L ]2 = B 2 (x) = H02 e2x/L

(7)

introducing this in (6), we get (for the eld energy within the superconductor)
L FM =

L
0

dx

2B 2 (x)
=
8

0

dx

H02 2x/L
H2
= 0 L
e
4
8

(8)

The magnetic free energy may also be calculated in a direct fashion

FM

1
=
4

dr
V



 B
 = 1
H
4

H0
dx
0


x/L

He

dH =

H02 L
8 L

(9)

 = ex/L H,
 and that the H-eld

when using B
within the superconductor is
 is
the same as the applied eld in the normal metal. In this free energy density B
the independent variable. The thermodynamical potential which is at a minimum
 is the
in the equilibrium state, is the one where the applied magnetic eld H

Condensed Matter Physics 2

36

independent variable. The relevant thermodynamic potential to be minimized is

obtained by a Legendre transformation:

= FM





1
1
dr  
dr
 H

H B
B
=
4
V
4
V


1 dx 2 L /x
H 2 L
H 2 L

= FM 0
H0 e
= 0
4 0 L
4 L
8 L

GM = FM

(10)

The magnetization in the superconductor is along the z axis and is determined

 =H
 + 4 M
 to be M = M (x) = (1 ex/L )H /(4). The allowance
from B
0

of the magnetic B-eld to penetrate into the superconductor gives rise to the
 = 0 (or
energy density gain determined by (10) in comparison with the case of B
L = 0). Based on eq. (2.6) in the note Magnetic energy and domains, this gain
of magnetic energy density may be calculated in a direct fashion as




dr
 H
 + dr
 H

M
M
V
V
 0
B=

L

H

L




0
H
dx
H
dx H0
=
dH 1 ex/L
dH

4
4
0 L 0
0 L 0





 H
2
H
dx H0
L
=
= 0
= GM
dH ex/L
L
4
8
L
0
0

1
FM =
V

(11)

 = N M
 is neglected in all the calculations, i.e.
[The demagnetization eld H
d
the sample is assumed to be a thin rod along the eld, see problem 27.1]. In case
the applied
eld is equal to the critical eld, H0 = Hc , the interface is stable if,
roughly, 2 2 Hc2 /(12) L Hc2 /(8) < 0 or if > 4 2/3. This value is reduced
when including interference between the two contributions, i.e. that L becomes
larger close to the surface where || is reduced.
27.3 Diraction eects in Josephson junctions
The gure shows a weak link of normal metal (dashed lines) between two
parts of a superconductor. The mag is applied in the z direcnetic eld H
 = 0 within the superconduction. B
 = H
 in the normal
tors, whereas B
metal. The magnetic ux through the
weak link is = AB = AH, where the
area is A = dx dy .

(a) With the choice of the Landau gauge, the vector potential within the normal
metal is

 = (0, x H, 0),
A

 = (0, 0, H) = B,

A

 =0
A

(1)

(see also problem 25.4).

(b) The path integral, calculated at a constant x, between the two superconductors
1 and 2 is

2

dy
 =
ds A
x H dy = Hdy x
(2)
1

37

Problems in Chapter 27 (superconductivity)

The current density in the y direction between the two superconductors, (27.68a),
is independent of z (when staying within the cross-section of the normal metal):


2e
j(x) = j0 sin 2 1 +

c

2
1

 = j sin + 4 Hd x
ds A
0
2
1
y
0

(3)

The magnetic ux quantum is here dened, (25.51), 0 = hc/e, however, in much

of the literature on superconductivity, the ux quantum is dened as the one
applying for Cooper-pairs, i.e. 0 = hc/e . The total current is

d
z

d
x

dx

4Hdy
J=
dz
dx j(x) = dz j0
cos 2 1 +
x
0
0
0


(
4
0
= (dz dx )j0
cos(2 1 ) cos 2 1 +
4
0
0
4Hdy

(4)

when introducing the total magnetic ux = dx dy H. Dening J0 = dz dx j0 and

using the general cosine relation: cos(u v) cos(u + v) = 2 sin u sin v, we nally
get




2
0
2
sin 2 1 +
sin
(5)
J = J0
2
0
0
(c) At zero voltage the phases of the wave functions in the superconductors 1 and
2, 1 and 2 , are constant in time. The result (5) clearly shows that the maximum
current at a certain eld and zero voltage is
)

Jmax

)

)
2 ))
= J0 0 ))sin
2
0 )

(6)