1

Physics 7440, Solutions to Problem Set # 8

1. Ashcroft & Mermin 12.2

For both parts of this problem, the constant offset of the energy, and also the location of the minimum at k0 ,
have no effect. Therefore we work with the simpler form for the energy

~2
(k) = ki Mij1 kj . (1)
2
(a) For simplicity of notation, we define the matrix W by Wij = Mij1 . (So, then, M = W 1 .) W and M are
both real, symmetric matrices. Moreover they are positive definite (since otherwise k = 0 would not be
a minimum), which implies among other things that both the determinant and the diagonal elements are
positive. In general we have

Wxx Wxy Wxz
W = Wxy Wyy Wyz . (2)
Wxz Wyz Wzz

The z-axis is the axis along which we apply the magnetic field and is fixed. However, we can choose the x
and y axes any way we like (so long as they are perpendicular to the z axis), and it is possible to choose
them so that Wxy = 0. (This is because choosing the x and y axes this way corresponds to making an
orthogonal transformation to diagonalize the upper-left 2 2 block of W , which we can certainly do.) So
with this choice of axes we have a somewhat simpler form

Wxx 0 Wxz
W = 0 Wyy Wyz . (3)
Wxz Wyz Wzz

To proceed, we note that the electron orbits for a magnetic field along the z-direction are given by the
intersection of constant energy surfaces with surfaces of constant kz . We will show that such orbits form
ellipses, find their area as a function of energy and kz , and use this to get the cyclotron effective mass.
The surface of constant energy is specified by the equation

~2 
Wxx kx2 + Wyy ky2 + Wzz kz2 + 2Wxz kz kx + 2Wyz kz ky .

= (k) = (4)
2
If we also treat kz as a constant here, then the resulting equation for kx and ky specifies the orbit at kz
with energy . We rewrite this as follows:
2
Wzz kz2 = Wxx kx2 + Wyy ky2 + 2Wxz kz kx + 2Wyz kz ky . (5)
~2
By completing the square twice on the right-hand side, we have
2 2 2 2
2 2
 Wxz kz 2  Wyz kz 2 Wxz kz Wyz kz
2
Wzz kz = Wxx kx + + Wyy ky + . (6)
~ Wxx Wyy Wxx Wyy

Rearranging terms, we have

h 2 2 2 2
2 2 Wxz kz
Wyz kz i  Wxz kz 2  Wyz kz 2
+ kz + Wzz = Wxx kx + + Wyy ky + . (7)
~2 Wxx Wyy Wxx Wyy

Now, we recognize this as the equation specifying an ellipse for the variables kx and ky . We are only
interested in the area of this ellipse, so we can drop the constant terms inside the parentheses on the
right-hand side, since these just shift the center of the ellipse. So we work with the equation
h W 2 k2 2 2
2 xz z Wyz kz i
2
+ kz2 + Wzz = Wxx kx2 + Wyy ky2 . (8)
~ Wxx Wyy
 2

If, by diving out by the left-hand side, we put this equation into the form
1 = kx2 /a2 + ky2 /b2 , (9)
then the area of the ellipse is A(, kz ) = ab. After some algebra we find
 2 2 2 2
h 2
2 Wxz kz
Wyz kz i
A(, kz ) = p + kz + Wzz . (10)
Wxx Wyy ~2 Wxx Wyy
Using this, and Ashcroft & Mermin Eq. (12.44), we get the simple result
~2 A 1
m = = p . (11)
2 Wxx Wyy
Now we just have to express this in the desired form. Using standard formulas for the inverse of a 3 3
matrix, and applying them to W , we have
1 Wxx Wyy
Mzz = Wzz = = Wxx Wyy det M , (12)
det W
so that
Mzz
Wxx Wyy = , (13)
det M
and therefore
r
det M
m = . (14)
Mzz
(b) The specific heat depends on the band structure only through the density of states at the Fermi energy,
according to
2 2
cv = k T D(F ). (15)
3 B
We need to calculate D(F ) for the given dispersion. To simplify the calculation, we are free to choose axes
to that M , and hence M 1 is diagonal, and
~2 X ki2
(k) = . (16)
2 i=x,y,z mi

The density of states is

d3 k 
Z X
2
ki2 /mi

D() = (~ /2) (17)
4 3 i
Z 3
d q 
(~2 /2)q 2

= mx my mz 3
(18)
4
Z 3
d q 
(~2 /2)q 2 .

= det M 3
(19)
4

In the second line above, we have changed variables to qi = ki / mi . The q-integral above is exactly
D(F ) as evaluated for free electrons in Ashcroft & Mermin chapter 2, with the mass set to unity (m 1).
Therefore looking at A & M Eq. (2.61), we have
r
1 2F
D(F ) = det M 2 2 . (20)
~ ~2
We want to write this in the form
r
m 2m F
D(F ) = 2 2 , (21)
~ ~2
where m is the specific heat effective mass. Equating the above two equations, we see that
m = (det M )1/3 . (22)
 3

2. Ashcroft & Mermin 12.4

(a) We imagine applying an electric field E, which is felt in the same way by every band. Therefore the current
density of the nth band is

jn = 1 E. (23)

The total current density is then

X hX i
j= jn = (1
n ) E. (24)
n n

Then we have for the electric field in terms of the current density:
hX i1
E= (1
n ) j (25)
n

Since by definition E = j, we have the desired result,

hX i1
= (1
n ) . (26)
n

(b) In this case

 
1 R1 H
1 = , (27)
R1 H 1

and similarly for 2 . We have

 
1 1 R1 H
1
1 = 2 , (28)
1 + R12 H 2 R1 H 1

and
 
1 2 R2 H
1
2 = 2 . (29)
2 + R22 H 2 R2 H 2

Adding these and taking the inverse, we have after some algebra
= [11 + 1
2 ]
1
(30)
 2 2 2 2 2 2 2 2 2 2 2 2

1 1 (2 + R2 H ) + 2 (1 + R1 H ) R1 H(2 + R2 H ) R2 H(1 + R1 H )
=
(1 + 2 )2 + H 2 (R1 + R2 )2 R1 H(22 + R22 H 2 ) + R2 H(21 + R12 H 2 ) 1 (22 + R22 H 2 ) + 2 (21 + R12 H 2 )
Writing this as
 
RH
= , (31)
RH

we have the desired results:

1 2 (1 + 2 ) + (1 R22 + 2 R12 )H 2
= , (32)
(1 + 2 )2 + H 2 (R1 + R2 )2

and
R1 22 + R2 21 + R1 R2 (R1 + R2 )H 2
R= (33)
(1 + 2 )2 + H 2 (R1 + R2 )2

(c) For simplicity, assume band 1 is an electron band (all occupied orbits closed), and band 2 is a hole band
(all empty orbits closed). Band 1 has a density ne of electrons and band 2 a density nh of holes. We focus
on this case because it allows us to consider the case of a compensated metal (nef f = ne nh = 0). We
could easily repeat the analysis for 2 electron bands and 2 hole bands.
 4

In the high field limit, R1 = 1/ne ec, and R2 = 1/nhec. Therefore, in the high field limit,
1 1 1  1 ne nh 1 neff
R1 + R2 = + = = . (34)
ec ne nh ec ne nh ec ne nh
Because this is nonzero, only the terms proportional to H 2 in the numerator and denominator of R will
contribute in the high-field limit, and we have
R1 R2
R (35)
R1 + R2
1 1 ec ne nh 1
= = . (36)
ne ec nh ec neff neff ec
Now we consider the case of a compensated metal (neff = 0), and work out the high-field behavior of both
and R. Now R1 + R2 = 0, so the Hall coefficient is field-independent:
R1 22 + R2 21
R= . (37)
(1 + 2 )2
Notice that this can no longer be simply related to the densities in the two bands. Indeed, even the sign of
R is completely determined by which of 1 and 2 is larger, which has nothing to do with carrier densities.
As for the resistivity, the H 2 -term in the denominator drops out because R1 + R2 = 0, but this is not the
case for the term in the numerator. The H 2 -term in the numerator then dominates in the large-field limit,
and we have
(1 R22 + 2 R12 )H 2
, (38)
(1 + 2 )2
which increases proportional to H 2 for large fields.
3. Ashcroft & Mermin 12.6
Let TR be the operator that translates a wavefunction by the Bravais lattice vector R. Acting on a general
1
wavefunction (r), we have TR (r) = (r+R). Moreover, on the position operator r, we have TR rTR = r+R.
The latter statement can be verified by considering
TR r(r) = TR [r(r)] = (r + R)(r + R). (39)
However, we can also calculate this by writing
1 1
TR r(r) = [TR rTR ]TR (r) = [TR rTR ](r + R). (40)
1
The two expressions are only consistent provided TR rTR = r + R.
Now lets consider the problem at hand. Were told that (r, 0) is a linear combination of Bloch wavefunctions,
all with the same crystal momentum k. This means that
TR (r, 0) = eikR (r, 0). (41)
We want to show that

TR (r, t) = eik R (r, t), (42)
where k = k eEt/~. This is what we want to show, because this statement is true if and only if (r, t) is a
linear combination of Bloch wavefunctions, all with crystal momentum k .
To show this consider the following expression:

TR (r, t) = TR eiHt/~ (r, 0) = [TR eiHt/~ TR

1
]TR (r, 0) = eikR [TR eiHt/~ TR
1
](r, 0). (43)
We need to compute the term in square brackets. First, note that

1 ~2 2
TR HTR = + U (r + R) + eE (r + R) (44)
2m
~2 2
= + U (r) + eE r + eE R (45)
2m
= H + eE R. (46)
 5

(In the second line we used the fact that U (r) is a periodic potential.) Next, I claim that, for any operator A,
 
TR eA TR
1
= exp [TR ATR 1
] . (47)

This can be proved by using the Taylor expansion of the matrix exponential:

hX 1 n i 1
TR eA TR
1
= TR A TR (48)
n=0
n!

X 1
= TR An TR
1
(49)
n=0
n!

X 1 1 n
= [TR ATR ] (50)
n=0
n!
 
1
= exp TR ATR . (51)

The crucial step is the 2nd to last line. It is easiest to illustrate by considering a particular term, for example

TR A2 TR
1 1
= TR ATR 1
TR ATR 1 2
= [TR ATR ] . (52)

Combining these results, we have

TR eiHt/~ TR
1
= eiHt/~ietER/~ = eietER/~ eiHt/~ . (53)

Going back to Eq. (43), we have

TR (r, t) = eikR eietER/~ eiHt/~ (r, 0) = eik R (r, t), (54)

as desired.