Sie sind auf Seite 1von 106

Introduction to

Solid State Theory


Lecture Notes
Professor Dr. Stefan Kehrein
Insitute for Theoretical Physics
Physics Department
University Goettingen, Germany
Version February 11, 2012
Disclaimer
Please bear in mind that lecture notes cannot (and do not intend to)
replace a textbook.
Contents
1 Introduction and Motivation 1
1.1 High-energy physics vs. solid state physics . . . . . . . . . . . . . 1
1.2 Reminder: Quantum mechanics of more than one particle . . . . 2
1.3 Fundamental Hamiltonian and Born-Oppenheimer approximation 5
2 Homogeneous Electron Gas 10
2.1 Quantization conditions and ground state . . . . . . . . . . . . . 10
2.2 Thermal properties . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Homogeneous electron gas in a magnetic eld . . . . . . . . . . . 17
3 Crystal Structure 23
3.1 Crystal lattices . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Reciprocal lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
3.3 Classication of Bravais lattices and crystal structures . . . . . . 25
3.4 Scattering by crystals . . . . . . . . . . . . . . . . . . . . . . . . 27
4 Band Structure 36
4.1 Bloch theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2 Nearly free electron gas . . . . . . . . . . . . . . . . . . . . . . . 43
4.3 Tight-binding ansatz . . . . . . . . . . . . . . . . . . . . . . . . . 48
4.4 Real band structures and symmetry properties . . . . . . . . . . 54
5 Lattice Vibrations and Phonons 57
5.1 Classical theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
5.2 Quantum theory of the harmonic crystal . . . . . . . . . . . . . . 65
2
6 Electron Dynamics 74
6.1 Semiclassical approximation . . . . . . . . . . . . . . . . . . . . . 74
6.2 Dynamics in a magnetic eld . . . . . . . . . . . . . . . . . . . . 80
6.3 Electrical conductivity in metals . . . . . . . . . . . . . . . . . . 88
6.4 Thermal conductivity . . . . . . . . . . . . . . . . . . . . . . . . 95
7 Outlook: Many-Particle Eects 98
3
Chapter 1
Introduction and
Motivation
This is a 3rd year undergraduate course which only requires a good working
knowledge of quantum mechanics. It does not use or introduce methods of
many-body theory.
1.1 High-energy physics vs. solid state physics
The essential dierence between the goals of high-energy physics and solid state
physics can be viewed in the following manner:
In high-energy physics we know the correct model (the so called standard
model) to describe nature up to an energy scale of at least O(100Gev).
What we are looking for is the correct fundamental model up to the Planck
scale O(10
19
Gev) where we know that the standard model breaks down
and gravity needs to be quantized as well (in a way currently unknown).
One can also state this by saying that we know an eective low-energy
theory (well, of O(100Gev)) of an unknown grand unied theory that we
try to nd using experiments and theoretical considerations.
In solid state physics the fundamental Hamiltonian describing nuclei and
electrons with a Coulomb interaction is perfectly well known. What we
are interested in here is to nd the emergent behavior of many (meaning at
least of order 1000) constituents at low energies of order room temperature
O(20mev).
Notice that reductionism is not very helpful to understand the types of behavior
that we are interested in in this course. For example, it does not really help to
1
1 Introduction and Motivation
know what a block of iron is made of in order to understand ferromagnetism:
Ferromagnetism occurs when many constituents work together.
Also, while knowing the fundamental Hamiltonian of solid sate physics gives
us a recipe to solve for the behavior that we are interested in (namely put the
Hamiltonian for suciently many constituents on a computer and evaluate the
eigenenergies and eigenfunctions numerically), this is often only of limited use
for understanding the behavior that we are interested in. Due to the exponential
growth of the dimension of the Hilbert space with the number of particles, one
can nowadays solve problems with about 20 particles (spin-1/2), but 1000 parti-
cles are even in principle out of reach since dim H becomes of order the number
of particles in the universe.
The strategy in solid state theory is therefore to perform suitable systematic
approximations in our fundamental Hamiltonian that render the resulting model
solvable using analytical or numerical methods. A key role in this strategy
is played by symmetries, which allow us to simplify models without making
approximations. This strategy determines the structure of this course.
1.2 Reminder: Quantum mechanics of more than
one particle
Let us rst look at the case of N = 2 particles. The following fundamental
axiom of quantum mechanics sets the stage:
The total Hilbert space H of two particles (1 and 2) is the tenor product of their
individual Hilbert spaces,
H = H
1
H
2
. (1.1)
For indistinguishable fermions (e.g. electrons) we further need to restrict H
to only those states that are antisymmetric under exchange of the particles.
Likewise, for indistinguishable bosons (e.g. photons) we need to restrict H to
only those states that are symmetric under exchange of the particles.
So for two distinguishable particles with quantum numbers q (particle 1) and q

(particle 2) the total state is


[) = [q) [q

) (1.2)
But if these particles are indistinguishable bosons we have for q ,= q

[) =
1

2
([q) [q

) +[q

) [q)) (1.3)
and for q = q

[) = [q) [q) (1.4)


Introduction to Solid State Theory (Kehrein) 2 University of Goettingen
1 Introduction and Motivation
For indistinguishable fermions we nd
[) =
1

2
([q) [q

) [q

) [q)) (1.5)
Notice that this state does not exist when q = q

, which just expresses the Pauli


principle: Two indistinguishable fermions cannot occupy the same quantum
state.
Also notice that all the above states are normalized [) = 1 if we assume that
the single particle states are orthonormal, q[q

) =
qq
.
As an explicit example we consider two particles (position coordinates x
1
, x
2
)
in a one-dimensional harmonic oscillator. The Hamiltonian in position repre-
sentation is given by
H = H(1) +H(2) (1.6)
=

2
2m

2
x
2
1
+
m
2

2
x
2
1


2
2m

2
x
2
2
+
m
2

2
x
2
2
(1.7)
1. We consider the particles as distinguishable: Particle 1 has quantum num-
ber n, particle 2 has quantum number n

. One can easily verify that


[) = [n) [n

) is an eigenstate of the above H with eigenenergy


E = (n +n

+ 1) (1.8)
and wavefunction
(x
1
, x
2
) = (x
1
[ x
2
[) [) (1.9)
=
n
(x
1
)
n
(x
2
) (1.10)
Here
n
(x) is the well-known eigenfunction corresponding to eigenenergy
(n + 1/2) of one particle in a harmonic oscillator:

n
(x) =
_
m

_
1/4
1

2
n
n!
H
n
() e

2
/2
(1.11)
where H
n
() is the n-th order Hermite polynomial and
def
= x
_
m/.
2. We consider identical bosons with spin 0. Then the normalized eigenfunc-
tions are
for n ,= n

: (x
1
, x
2
) =
1

2
(
n
(x
1
)
n
(x
2
) +
n
(x
1
)
n
(x
2
))
for n ,= n : (x
1
, x
2
) =
n
(x
1
)
n
(x
2
) (1.12)
Introduction to Solid State Theory (Kehrein) 3 University of Goettingen
1 Introduction and Motivation
3. We consider electrons (spin-1/2). Then
x
1
, s
1
; x
2
, s
2
[) = (x
1
, s
1
; x
2
, s
2
) (1.13)
= (x
1
, x
2
) (s
1
, s
2
) (1.14)
where (x
1
, x
2
) is the orbital wavefunction and (s
1
, s
2
) is the spin wave-
function. From the antisymmetrization postulate we know that the total
wavefunction is antisymmetric under exchange of the particles
(x
1
, s
1
; x
2
, s
2
) = (x
2
, s
2
; x
1
, s
1
) (1.15)
and therefore we nd the following identity
(x
1
, s
1
; x
2
, s
2
) =
1
2
((x
1
, x
2
) (s
1
, s
2
) (x
2
, x
1
) (s
2
, s
1
))
=
1
4
_
((x
1
, x
2
) +(x
2
, x
1
)) ((s
1
, s
2
) (s
2
, s
1
))
+((x
1
, x
2
) +(x
2
, x
1
)) ((s
1
, s
2
) (s
2
, s
1
))
_
(1.16)
This means that the total wavefunction of two electrons can always be
written as a superposition of
a) symmetric orbital wavefunction, antisymmetric spin wavefunction
and
b) antisymmetric orbital wavefunction, symmetric spin wavefunction.
Notice that this cannot be generalized to N > 2.
a)
for n ,= n

: (x
1
, x
2
) =
1

2
(
n
(x
1
)
n
(x
2
) +
n
(x
1
)
n
(x
2
))
for n ,= n : (x
1
, x
2
) =
n
(x
1
)
n
(x
2
) (1.17)
with the spin wavefunction
=
1

2
([ ) [ ) [ ) [ )) (1.18)
that is total spin S = 0 (spin singlet).
b) (x
1
, x
2
) =
1

2
(
n
(x
1
)
n
(x
2
)
n
(x
1
)
n
(x
2
)) (1.19)
The possible spin wavefunctions are
= [ ) [ ) , [ ) [ ) ,
1

2
([ ) [ ) +[ ) [ )) (1.20)
that is total spin S = 1 and therefore threefold degenerate. Notice
that this state does not exist for n = n

, which again just exemplies


the Pauli principle. The Pauli principle turns out to be the key
ingredient for understanding the structure of matter.
Introduction to Solid State Theory (Kehrein) 4 University of Goettingen
1 Introduction and Motivation
We now generalize to N > 2 particles. The total Hilbert space is
H = H
1
. . . H
N
. (1.21)
For identical bosons the total state vector must be symmetric under exchange
of any two particles, for identical fermions the total state vector must be anti-
symmetric under exchange of any two particles.
In the sequel we will only discuss the case of identical fermions since this course
is mainly concerned with the electronic structure. The normalized basis of the
1-particle Hilbert space for the electrons is denoted by [), where can be some
multi-index (like [n, l, m, s) for the bound states of the hydrogen atom). One
can show that a normalized basis of the N-fermion Hilbert space is given by
[
1
, . . . ,
N
)
def
=
1

N!

S
N
(sgn) [
(1)
) [
(2)
) . . . [
(N)
) (1.22)
where the sum runs over all elements of the permutation group S
N
. The sign
of the permutation is dened as +1 for an even number of transpositions, and
-1 for an odd number of transpositions making up . One can show that the
many-particle wavefunction corresponding to these states can be written as a
determinant
(r
1
, s
1
; . . . ; r
N
, s
N
) = r
1
, s
1
; . . . ; r
N
, s
N
[
1
, . . . ,
N
) (1.23)
=
1

N!

1
(r
1
, s
1
)
1
(r
2
, s
2
) . . .
1
(r
N
, s
N
)

2
(r
1
, s
1
)
2
(r
2
, s
2
) . . .
2
(r
N
, s
N
)
.
.
.
.
.
.

N
(r
1
, s
1
)

N
(r
2
, s
2
) . . .

N
(r
N
, s
N
)

Here

(r, s) = r, s[) is the 1-particle wavefunction with respect to the basis.


The above determinant is called Slater determinant. From its structure one
can see immediately that the state [
1
, . . . ,
N
) only exists if all the quantum
numbers are dierent, otherwise two rows in the determinant are identical and it
vanishes. This is the general expression of the Pauli principle that two identical
fermions cannot be in the same quantum state, which follows from the general
antisymmetrization postulate.
1.3 Fundamental Hamiltonian and Born-Oppenheimer
approximation
For simplicity we consider a monoatomic solid consisting of N
n
nuclei of mass M,
nuclear charge Ze, and N
e
= Z N
n
electrons. The position coordinates of the
Introduction to Solid State Theory (Kehrein) 5 University of Goettingen
1 Introduction and Motivation
electrons will always be denoted with r
i
, the position coordinates of the nuclei
with R
i
.
1
Then we can write down the fundamental Hamiltonian that we need to solve:
H = H
e
+H
n
+H
en
(1.24)
Its various parts are
Electronic contribution:
H
e
= T
e
+V
ee
(1.25)
with the total kinetic energy of the electrons
T
e
=

2
2m
Ne

i=1

2
ri
(1.26)
and the Coulomb interaction energy of the electrons
V
ee
=
1
2

i=j
e
2
[r
i
r
j
[
(1.27)
Contribution from the nuclei:
H
n
= T
n
+V
nn
(1.28)
with the total kinetic energy of the nuclei
T
n
=

2
2m
Ne

i=1

2
Ri
(1.29)
and the Coulomb interaction energy of the nuclei
V
nn
=
1
2

i=j
(Ze)
2
[

R
i


R
j
[
(1.30)
Coulomb attraction between nuclei and electrons:
H
en
=

i=j
Ze
2
[r
i


R
j
[
(1.31)
Essentially there are only two approximations that we have made in writing
down our fundamental Hamiltonian:
1
This section closely follows lecture notes by Thomas Pruschke.
Introduction to Solid State Theory (Kehrein) 6 University of Goettingen
1 Introduction and Motivation
1. We have neglected relativistic eects. This is usually a very good approx-
imation because the relevant velocities in a solid are typically at most of
order c
vac
/100.
2. Related to this we have neglected all spin eects, especially spin-orbit
interactions. This is not always a good approximation, especially for heavy
nuclei. However, one can easily build in the spin-orbit interaction into the
Hamiltonian, but it makes the notation more cumbersome. That is why
we ignore it for the time being.
It is an empirical observation that (usually) in a solid the nuclei move only
slightly around their equilibrium positions

R
(0)
i
. We will give a theoretical rea-
son underlying this approximation further below (Born-Oppenheimer approxi-
mation). For now we want to plug it into our fundamental Hamiltonian (1.24)
because it motivates the outline of this course:
H = E
(0)
n
+T
e
+V
ee
+V
(0)
en
+H
ph
+H
eph
(1.32)
1. E
(0)
n
=
1
2

i=q
(Ze)
2
[

R
(0)
i


R
(0)
j
[
(1.33)
is the potential energy of the equilibrium conguration of the nuclei, that
is a constant without further importance for the dynamics.
2. V
(0)
en
=

i=j
Ze
2
[r
i


R
(0)
j
[
(1.34)
is the Coulomb interaction between the electrons and the nuclei at their
equilibrium positions. Notice that this is simply a potential term for the
electrons.
3. H
ph
= H
n
E
(0)
n
(1.35)
describes the dynamics of the nuclei around their equilibrium positions,
that is phonons.
4. H
eph
= H
en
V
(0)
en
(1.36)
describes the interplay of electron and lattice dynamics.
The outline of this course follows the various terms above, taking more and
more of them into account to arrive a more and more realistic picture of a solid:
1. Chapter 2, Homogeneous electron gas: We study only the kinetic energy
of the electrons, T
e
. This is a minimal model for the electrons in a solid
that already introduces important concepts like Fermi surface, etc.
2. Chapter 3, Crystal structure: The crystal structure determines E
(0)
n
.
Introduction to Solid State Theory (Kehrein) 7 University of Goettingen
1 Introduction and Motivation
3. Chapter 4, Band structure: We put the above pieces together and inves-
tigate the following terms in the fundamental Hamiltonian
H
0
= E
(0)
n
+T
e
+V
(0)
en
(1.37)
This is the most importance approximation for understanding solids and
explains a multitude of phenomena. A thorough understanding of (1.37)
is the key goal of this course.
4. Chapter 5, Lattice dynamics: H
ph
is very important for describing e.g.
the specic heat of solids.
5. Chapter 6, Transport: In addition to H
0
we need to study deviations from
the perfect crystal structure and scattering processes due to H
eph
.
6. Chapter 7, Magnetism: We will study magnetic phenomena based on a
very simplied treatment of V
ee
.
7. Chapter 8, Superconductivity: In most materials this is due to electron-
phonon scattering, H
eph
.
8. Chapter 9, Outlook: Quasiparticles and Fermi liquid theory: This gives a
rst glimpse at correlations eects due to V
ee
. A thorough investigation
of V
ee
is the content of the Masters course on solid state theory since it
requires many-body techniques.
Next we want to theoretically motivate the above treatment of the nuclei dynam-
ics as a (small) perturbation. This is a consequence of the Born-Oppenheimer
approximation, which is also one of the key approximations in quantum chem-
istry. The Born-Oppenheimer approximation starts from the observation that
for typical nuclei one has
m
M
= O(10
4
), the nuclei are much heavier than the
electrons. If we would set M = , then T
n
0 and every eigenfunction of the
full Hamiltonian can be written in the following way:
(r,

R) = (r) (

R) (1.38)
Here (

R) is a wave function describing the spatial structure of the nuclei


and (r) is an eigenfunction of
_
H
e
+
_
d

V
en
(r
i


R
j
) (

R)
_
(r) = E
e
(r) (1.39)
For
m
M
> 0 the ansatz (1.38) becomes an approximation, the so called Born-
Oppenheimer approximation. Another way of viewing (1.38) is the time scale
separation between fast electron dynamics and slow nuclei dynamics: therefore
the electrons move in a quasistatic background provided by the instantaneous
positions of the nuclei, which motivates the ansatz (1.38). The full problem
Introduction to Solid State Theory (Kehrein) 8 University of Goettingen
1 Introduction and Motivation
then amounts to a self-consistent determination of (r) and (

R) such that
the total energy becomes minimal. However, for the purposes of this course
we will take the position of the nuclei as given and strongly localized at the
positions

R
(0)
i
:
(

R) = (

R
1


R
(0)
1
) . . . (

R
Nn


R
(0)
Nn
) (1.40)
The remaining eigenvalue problem for the electrons is then eectively parametrized
by

R
(0)
i
.
A very rough estimate for the validity of the Born-Oppenheimer approximation
follows from the mass dependence of the kinetic energy in the ground state of
q harmonic oscillator,
T) =
1
4
with M
1/2
(1.41)
Therefore
T
n
)
T
e
)

_
m
M
(1.42)
which is typically still of order 10
2
.
2
Another frequently employed approximation is to only look at the dynamics of
the valence electrons, and to take into account the core electrons by working
with nuclei with a reduced charge and an eective potential due to the screening
from the core electrons. This is sometimes a very good approximation, but can
be problematic for electrons in d and f-shells where a clear separation of valence
and core electrons is not possible. We will say more on this topic in the chapter
on band structure.
2
It should be noted that correlation eects can lead to eective electron masses that are
much larger than the bare electron mass, for example in heavy fermion systems. In such cases
one needs to re-think the validity of the Born-Oppenheimer approximation.
Introduction to Solid State Theory (Kehrein) 9 University of Goettingen
Chapter 2
Homogeneous Electron Gas
The homogeneous electron gas is a minimal model for describing the electrons
in a solid. In (1.32) it only takes into account the kinetic energy of the electrons,
that is we are studying the Hamiltonian
H = T
e
=

2
2m
Ne

i=1

2
ri
(2.1)
Additionally, we need to make sure that the electrons in a given volume V
compensate the positive charge from the nuclei in V . Therefore the electron
density n = N
e
/V is taken as a given parameter in the homogeneous electron
gas. Notice that we can equally only consider the valence electrons, then N
e
is
just the number of valence electrons in V .
2.1 Quantization conditions and ground state
Assume that we already know the eigenfunctions of
_


2
2m

2
r
_

(r) =

(r) (2.2)
Then the Slater determinant
1

N!

1
(r
1
)
1
(r
2
) . . .
1
(r
N
)

2
(r
1
)
2
(r
2
) . . .
2
(r
N
)
.
.
.
.
.
.

N
(r
1
)

N
(r
2
) . . .

N
(r
N
)

(2.3)
10
2 Homogeneous Electron Gas
is an eigenfunction of H = T
e
with the eigenenergy
E =
Ne

i=1

i
(2.4)
Therefore we can restrict ourselves to the solution of the 1-particle problem
(2.2). Such reductions to eective 1-particle problems will play a key role in
all future chapters and simplify the solution enormously, i.e. we do not need
many-body techniques. Notice that we have ignored the electron spin degree
of freedom in writing down the Slater determinant since this will just lead to a
double degeneracy of all energy levels.
In order to set the electron density n, we look at the homogeneous electron gas
in a cube with length L, V = L
3
. This leads to a number of possible boundary
conditions:
The wave function (r) vanishes outside the cube.
We identify opposite sides of the cube:
(x, y, z) = (x +L, y, z) = (x, y +L, z) = (x, y, z +L) (2.5)
These periodic boundary conditions are also called Born-von Karman
boundary conditions.
The dierences between these boundary conditions become negligible in the
bulk for suciently large volume V . For our purposes the Born-von Karman
boundary conditions turn out to be most convenient, therefore we will use them
in the sequel. However, if one is specically interested in boundary eects like
in two dimensional electron gases or topological materials the other boundary
conditions are more adequate (since they are more realistic).
The normalized eigenfunctions of (2.2) are just the plane waves

k
(r) =
1

V
e
i

kr
(2.6)
with eigenenergy

k
=

2
k
2
2m
(2.7)
Notice that the plane waves are also eigenfunctions of the momentum operator
i
r
with eigenvalue p =

k (corresponding to velocity v =

k/m).
The periodic boundary conditions lead to the quantization condition
e
ikx L
= e
iky L
= e
ikz L
= 1 (2.8)
Introduction to Solid State Theory (Kehrein) 11 University of Goettingen
2 Homogeneous Electron Gas
and therefore
k
x
=
2
L
n
x
, k
y
=
2
L
n
y
, k
z
=
2
L
n
z
(2.9)
with integers n
x
, n
y
, n
z
Z. These allowed k-values form a lattice with spacing
2/L. The number of k-values in a large volume in k-space is then to a very
good approximation given by

(2/L)
3
=
V
(2)
3
(2.10)
The ground state for N electrons follows from lling up all one-particle states
starting from k = 0 since the energy is proportional to

k
2
. The Pauli principle
needs to be respected (otherwise the Slater determinant vanishes), therefore
each discrete k-value can be lled with two electrons: one electron with spin up
and one electron with spin down. The occupied states in k-space then form a
sphere with radius k
F
(Fermi wavevector). Its volume is = (4/3) k
3
F
and the
corresponding number of electrons is
N
e
= 2
4
3
k
3
F
V
8
3
=
k
3
F
3
2
V (2.11)
The additional factor 2 counts the spin degeneracy. Hence the electron density
is
n =
N
e
V
=
k
3
F
3
2
(2.12)
Here n has to be interpreted as a given xed (and temperature-independent)
number ensuring overall charge neutrality of our solid, which therefore deter-
mines the Fermi wavevector k
F
.
The surface of the Fermi sphere plays a very important role in determining
the low-energy properties of solid. It is called Fermi surface and separates
the occupied states from the unoccupied states (in the ground state). Other
important quantities are the Fermi momentum p
F
= k
F
, the Fermi energy

F
=
2
k
2
F
/2m, the Fermi velocity v
F
= p
F
/m and the Fermi temperature
T
F
=
F
/k
B
. For example for Cu one has:

F
= 7.0 eV
T
F
= 8.16 10
4
K
k
F
= 13.6 nm
v
F
= 1.57 10
6
m
s
Introduction to Solid State Theory (Kehrein) 12 University of Goettingen
2 Homogeneous Electron Gas
2.2 Thermal properties
The (internal) energy of the homogeneous electron gas at zero temperature is
U = 2

k|<k
F
(

k) (2.13)
with (

k) =
2
k
2
/2m. The factor 2 originates from spin degeneracy. The energy
density is
u =
U
V
= 2
1
L
3
_
L
2
_
3
_
|

k|<k
F
d

k (

k) (2.14)
=
1
4
3
_
k
F
0
dk 4 k
2

2
k
2
2m
(2.15)
=

2
k
5
F
10
2
m
(2.16)
where we have used a continuum approximation for the discrete k-summation
(which holds with excellent accuracy for macroscopic volumes). We can also
write down the average energy per electron
U
N
e
=
U
V
V
N
e
=
3
5

F
=
3
5
k
B
T
F
(2.17)
From statistical physics one knows that the occupation probability for a state
with energy in a system of noninteracting fermions in the grand canonical
ensemble (with temperature T and chemical potential ) is given by the Fermi-
Dirac distribution
f() =
1
e
()/k
B
T
+ 1
(2.18)
Notice that this is consistent with our ground state when we identify with the
Fermi energy
F
(which holds to very good accuracy as we will see below): for
T 0 one has f() = 1 for <
F
, and f() = 0 for >
F
.
The internal energy at temperature T is then
u(T) =
_
d

k
4
3
(

k) f((

k)) (2.19)
Since we have to consider the electron density n as a given temperature-independent
number (depending on the material, of course), the chemical potential (n, T)
is implicitly determined by counting the number of electrons
1
n =
_
d

k
4
3
f((

k)) (2.20)
1
It is much more convenient to work in the grand canonical ensemble then enforcing a
given n in the canonical ensemble.
Introduction to Solid State Theory (Kehrein) 13 University of Goettingen
2 Homogeneous Electron Gas
Integrals like (2.19) and (2.20), where the integrand depends only on the energy,
appear very often in solid state physics and can be simplied by a change of
variables
_
d

k
4
3
F((

k)) =
1

2
_
dk k
2
F((k)) (2.21)
=
_

d () F() (2.22)
where the density of states () comes from the change of measure:
() =
k
2

2
1
_
d
dk
_ (2.23)
=
m

2
_
2m

2
(2.24)
One can rewrite this as
() =
_
3
2
n

F
_

F
for > 0
0 for < 0
(2.25)
The corresponding result for the two-dimensional electron gas will be worked
out in a problem set. The interpretation of the density of states is as follows:
() d =
1
V

_
Number of one-electron states in the energy interval [, +d]
_
(2.26)
The density of states at the Fermi energy is denoted by
2

F
= (
F
) =
3
2
n

F
=
mk
F

2
(2.27)
Notice that we count both spin directions when calculating the density of states,
sometimes one also nds other denitions in the literature.
We can now rewrite (2.19) and (2.20) in the following way:
u(T) =
_

d () f() (2.28)
n =
_

d () f() (2.29)
We will later see that these formulas can be generalized to arbitrary band struc-
tures, only the density of states () remembers the actual band structure.
2

F
plays a very important role in the sequel since typical excitation energies are generally
much smaller than the Fermi energy, therefore we can often replace () by
F
.
Introduction to Solid State Theory (Kehrein) 14 University of Goettingen
2 Homogeneous Electron Gas
In general these integrals cannot be done in closed form. However, because of
k
B
T for most purposes, the Fermi-Dirac distribution essentially only varies
from its zero temperature step form in a small interval of order k
B
T around the
chemical potential. One exploits this fact in the Sommerfeld expansion, which
will permit us to evaluate (2.28) and (2.29) in a systematic expansion in powers
of T/T
F
. The starting point of the Sommerfeld expansion is
_

d H() f() =
_

d K()
_

f()
d
_
(2.30)
where H() is an arbitrary function that vanishes for and
K()
def
=
_

H(

) (2.31)
The derivative on the right hand side of (2.30) is essentially only nonzero in
a small interval of order k
B
T around . This motivates performing a Taylor
expansion of K() around = :
K() = K() +

n=1
( )
n
n!
d
n
K()
d
n

=
(2.32)
Plugging this expansion into (2.30) yields
_

d H() f() = K()


_

d
_

f()
d
_
(2.33)
+

n=1
1
n!
d
n1
H()
d
n1

=
_

d ( )
n
_

f()
d
_
=
_

d H() +

2
6
(k
B
T)
2
H

() +O
_
_
k
B
T

_
4
_
(2.34)
where we have used that only even values of the summation index n contribute
due to symmetry reasons. Inserting this expansion in (2.28) and (2.29) yields
(up to terms in order O(T/T
F
)
4
)
u =
_

0
d () +

2
6
(k
B
T)
2
(

() +()) (2.35)
n =
_

0
d () +

2
6
(k
B
T)
2

() (2.36)
Obviously (T = 0) =
F
and (T)
F
T
2
, therefore we can approximate
up to terms in O(T/T
F
)
4
n =
_

F
0
d () + (
F
)
F
+

2
6
(k
B
T)
2

(
F
) (2.37)
Introduction to Solid State Theory (Kehrein) 15 University of Goettingen
2 Homogeneous Electron Gas
Since n =
_

F
0
d () one concludes
0 = (
F
)
F
+

2
6
(k
B
T)
2

(
F
) +O(T
4
) (2.38)
(T) =
F


2
6
(k
B
T)
2

(
F
)

F
(2.39)
=
F
_
1
1
3
_
k
B
T
2
F
_
2
+O
_
_
T
T
F
_
4
__
(2.40)
For typical temperatures the chemical potential only deviates very little from
the Fermi energy.
With this result we can write for the internal energy for given electron density n
u(T) =
_

F
0
d () (2.41)
+
F
_
(
F
)
F
+

2
6
(k
B
T)
2

(
F
)
_
+

2
6
(k
B
T)
2

F
+O(T
4
)
From (2.38) we know that the second term on the right hand side vanishes,
hence
u(T) = u
0
+

2
6
(k
B
T)
2

F
+O
_
_
T
T
F
_
4
_
(2.42)
The electronic contribution to the specic heat is therefore always linear in
temperature and directly proportional to the density of states at the Fermi
surface
c
V
=
_
u
T
_
n
=

2
3
k
2
B
T
F
(2.43)
Specically for the homogeneous electron gas in d = 3
c
V
=

2
2
k
B
T

F
nk
B
(2.44)
In chapter 5 we will nd another contribution T
3
to the specic heat of a
crystalline solid from lattice vibrations
c
V
= T +AT
3
(2.45)
For low temperatures T 0 the electronic contribution always dominates
c
V
T
= +AT
2
(2.46)
Introduction to Solid State Theory (Kehrein) 16 University of Goettingen
2 Homogeneous Electron Gas
is called Sommerfeld coecient and proportional to the electron density of
states at the Fermi energy.
2.3 Homogeneous electron gas in a magnetic eld
We consider a homogeneous magnetic eld in z-direction

B = (0, 0, B
0
) (2.47)
Remember that a classical particle with charge q moves uniformly in z-direction,
while its trajectory projected onto the x y-plane is a circle that the particle
transverses with the cyclotron frequency

c
=
[q[ B
0
mc
(2.48)
The quantum mechanical analysis for electrons starts from the Hamiltonian
H =
1
2m
_
p +
e
c

A
_
2

g
B

S

B (2.49)
with
B
= e/2mc and Lande factor g = 2. We choose a vector potential

A(r)
in Landau gauge

A(r) = (0, xB
0
, 0) (2.50)
Since [H, p
z
] = 0 the momentum in z-direction is still a good quantum number,
p
z
= k
z
. However, the other components of

def
= p +
e
c

A do not commute
[
x
,
y
] = i
e
c
B
0
(2.51)
therefore we can only pick one direction as another quantum number. We choose
the y-direction and have the following separation ansatz
(r) = e
i(kyy+kzz)
(x) (s) (2.52)
We plug this into the stationary Schrodinger equation and nd


2
2m

+
m
2
c
2
(x x
0
)
2
=
_
E

2
k
2
z
2m

1
2
g
B
B
0

_
(2.53)
= 1 correspond to spin wavefunction = [ ), [ ). Here
x
0
=
k
y
m
c
(2.54)
Introduction to Solid State Theory (Kehrein) 17 University of Goettingen
2 Homogeneous Electron Gas
hence the remaining problem in x-direction is a harmonic oscillator shifted by x
0
.
The eigenenergies are then given by
E(n, k
z
, ) =
c
_
n +
1 +
2
_
+

2
k
2
z
2m
(2.55)
with n = 0, 1, 2, . . .. These discrete levels are called Landau levels. The eigen-
functions are given by
(r) = e
i(kyy+kzz)

n
(x x
0
) (s) (2.56)
where
n
(x) is the eigenfunction of the harmonic oscillator corresponding to
quantum number n. Since the eigenenergies do not depend on k
y
, the Landau
levels are strongly degenerate. In order to calculate observable quantities, we
need the density of states, and therefore the degeneracy N
L
of the Landau levels.
We take periodic boundary conditions in y-direction leading to
k
y
=
2
L
y
n
y
(2.57)
with n
y
Z. In x-direction we use hard wall boundary conditions: Since the
wavefunctions in x-directions are exponentially localized around
x
0
=

m
c
2
L
y
n
y
(2.58)
we request 0 < x
0
L
x
leading to
0 < n
y

m
c
2
L
x
L
y
= N
L
(2.59)
The degeneracy of the Landau level (for xed k
z
, ) can also be expressed as
N
L
=

0
(2.60)
where = B
0
L
x
L
y
is the magnetic ux through the surface and the ux quan-
tum

0
def
=
2c
e
(2.61)
The degeneracy of the Landau level just counts the number of ux quanta
through the surface. This massive degeneracy also explains the asymmetry of
the eigenfunctions in x and y-directions: Suitable combinations of our eigen-
functions are plane waves in x-direction and harmonic oscillator eigenfunctions
in y-direction.
Introduction to Solid State Theory (Kehrein) 18 University of Goettingen
2 Homogeneous Electron Gas
In the next step we calculate the density of states
() =
1
V
N
L

n,kz,
( E(n, k
z
, )) (2.62)
=
N
L
V
L
z
2
_

dk
z

n,
( E(n, k
z
, )) (2.63)
=
m
c
4
2

n,
_
dk
z
(k
z
k
0
)
_

E(n, k
z
, )
k
z

k0
_
1
(2.64)
where = E(n, k
0
, ). The nal result takes the form
(, B) =
1
8
2
_
2m

2
_
3/2

n,

_

c
_
n +
1+
2
__
_

c
_
n +
1+
2
_
(2.65)
and displays characteristic square root singularities equally spaced at energies
0,
c
, 2
c
, 3
c
, . . . Since we have already seen that the density of states at the
Fermi energy plays an important role in physical observables, these singularities
lead to characteristic periodic eects as the magnetic eld is varied for xed
Fermi energy (since
c
B
0
).
The observable that we will look at in more detail is the magnetization. We
can nd the operator

M corresponding to magnetization by noticing that by
denition the interaction energy of magnetization and external magnetic eld is
M B (2.66)
therefore
H
B
=

M (2.67)
Once we know the ground state energy depending on the external magnetic eld
H(B) [
GS
(B)) = E
GS
(B) [
GS
(B)) (2.68)
we can deduce the expectation value of the magnetization via
E
GS
(B)
B
=

GS
(B)
B
[H(B)[
GS
(B)) +
GS
(B)[H(B)[

GS
(B)
B
)
+
GS
(B)[
H(B)
B
[
GS
(B)) (2.69)
= E
GS
(B)

B

GS
(B)[
GS
(B))
GS
(B)[

M[
GS
(B))
=
GS
(B)[

M[
GS
(B)) (2.70)
Introduction to Solid State Theory (Kehrein) 19 University of Goettingen
2 Homogeneous Electron Gas
since [
GS
(B)) is normalized,
GS
(B)[
GS
(B)) = 1 B. Magnetization den-
sity
m(B) =
1
V
E
GS
(B)
B
(2.71)
and magnetic susceptibility
(B) =
m(B)
B
(2.72)
therefore follow immediately once we know the ground state energy. This is
simple since we just need to ll all single particle states up to the chemical
potential
E
GS
(B) =
_

0
d (, B) (2.73)
One can easily verify to for typical magnetic elds one can take =
F
. The
explicit calculation of (2.73) is somehow lengthy, details can for example be
found in the book Quantum Theory of Magnetism by Nolting. The nal
result is (for
B
B
F
)
m(B) =
3
2
n

2
B
B
_
1
1
3
+F
osc
(B)
_
(2.74)
The rst term in the parantheses comes from Pauli spin paramagnetism, and the
second term from Landau-Peierls diamagnetism. This can be easily understood
in the context of the Lenz rule: the orbital motion of the electrons in the
magnetic eld generates a magnetic eld that opposes the external eld. The
third term is an oscillating contribution
F
osc
(B) =
1

_

F

B
B

=1

3/2
sin
_

4


F

B
B
_
(2.75)
At nonzero temperature one can derive (assuming k
B
T
F
)
F
osc
(B) =
k
B
T

B
B
_

F

B
B

=1

1/2
sin
_

4


F

B
B
_
sinh
_

2
k
B
T

B
B
_ (2.76)
Due to the sinh-term in the denominator this oscillating term vanishes exponen-
tially for larger temperatures, k
B
T
B
B. On the other hand, for suciently
low temperatures the oscillations as a function of the external magnetic eld are
observable experimentally and are called the de Haas-van Alphen eect. The
Introduction to Solid State Theory (Kehrein) 20 University of Goettingen
2 Homogeneous Electron Gas
period of this oscillations follows from
2 =

F

B
_
1
B
0

1
B
1
_
(2.77)

_
1
B
_
=
2
B

F
=
2e
c
1
A(k
F
)
(2.78)
where A(k
F
) = k
2
F
is the cross surface of the Fermi sphere perpendicular to
the magnetic eld.
Remarks:
1. The suggestive notation (2.78) already emphasizes that the de Haas-van
Alphen eect can be used to measure Fermi surfaces. We will say more
about this later on when we have introduced nontrivial band structures.
2. Not surprisingly the peaked structure of the density of states also shows up
in other quantities if one varies the magnetic eld. An important example
are the oscillations of the resistance as a function of the external magnetic
eld, the so called Shubnikov-de Haas oscillations.
3. If we can neglect F
osc
(B) (i.e. for suciently high temperatures) the free
Fermi gas is always paramagnetic according to (2.74). However, the dia-
magnetic term turns out to strongly depend on the band structure (since
it is related to orbital motion, as opposed to the Pauli spin term) and can
be much larger in systems with a nontrivial band structure. For this rea-
son e.g. many semiconductors are diamagnetic. We will say more about
this later on in this course.
4. In the literature one often nds a picture where the B = 0 Fermi sphere
is replaced by Landau cylinders for B ,= 0. This can be somehow con-
fusing since we have seen before that k
x
, k
y
, k
z
are not simultaneous good
quantum numbers in a magnetic eld. Hidden behind the picture of Lan-
dau cylinders is a semiclassical picture: a localized wave packet describing
an electron in position space will move on circles perpendicular to the
magnetic eld (for simplicity we take k
z
= 0) and has to fulll the Born-
Sommerfeld quantization conditions. In momentum space this translates
to motion around a Landau cylinder in a plane perpendicular to the mag-
netic eld. Therefore we have to think of k
x
, k
y
, k
z
as Fourier components
of the localized wave packet, and not as good quantum numbers.
Since the radius of the Landau cylinders is proportional to

B, they get
shifted out of the Fermi sphere for increasing magnetic eld. One can show
that all the Landau cylinders inside the Fermi sphere are occupied, hence
one has abrupt rearrangements of the electrons once a Landau cylinder
crosses the Fermi sphere. In the semiclassical picture these rearrangements
are responsible for the oscillations as a function of the external magnetic
Introduction to Solid State Theory (Kehrein) 21 University of Goettingen
2 Homogeneous Electron Gas
eld. Also notice that for typical magnetic elds (e.g. B = 1T correspond-
ing to
c
0.1meV ) there are a few 10000 Landau cylinders inside the
Fermi sphere.
Introduction to Solid State Theory (Kehrein) 22 University of Goettingen
Chapter 3
Crystal Structure
3.1 Crystal lattices
The key building block of crystal structures is the Bravais lattice. We give two
equivalent denitions:
1. A Bravais lattice is an innite set of points in space (or in the plane
for two-dimensional lattices) that looks the same seen from each of these
points.
2. A Bravais lattice is given by the points

R = n
1
a
1
+n
2
a
2
+n
3
a
3
(d = 3) (3.1)

R = n
1
a
1
+n
2
a
2
a
3
(d = 2) (3.2)
with n
i
Z. The so called primitive lattice vectors a
i
must be linearly
independent (i.e. not lie in a plane for d = 3) and are said to generate or
span the Bravais lattice.
Remarks:
1. Notice that the choice of a
i
is not unique.
2. A honeycomb lattice is not a Bravais lattice. One can see easily that it
violates denition 1 above, it does not appear identical from all points.
We will describe this situation below as a lattice with a (nontrivial) basis.
We dene the primitive unit cell as an object that covers the whole space without
overlaps or voids when tranlsated by vectors of the Bravais lattice. Notice that
the choice of the primitive unit cell is not unique:
23
3 Crystal Structure
1. If a
i
are the shortest possible primitive lattice vectors, then the points
r =

i
x
i
a
i
with x
i
[0, 1) dene the so called conventional unit cell. The
conventional unit cell is very convenient for the graphical representation
of lattices, but it does usually not display the full symmetry of the lattice.
Also it is not intuitive which lattice points belong to which unit cell.
2. The Wigner-Seitz cell is dened as all the points that are closer to a given
lattice point than to any other point of the lattice. It displays the full
symmetry of the lattice.
Now we can dene a general crystal structure, which is a Bravais lattice with
a basis. The crystal structure consists of identical copies of the same physical
object in the unit cell, which is translated by vectors of the Bravais lattice.
For example the honeycomb lattice is a lattice with a basis consisting of two
atoms (see problem sets). Other example are all crystals that are non monoatomic,
since then clearly the unit cell must contain dierent atoms.
3.2 Reciprocal lattice
The reciprocal lattice plays a fundamental role in the study of crystalline solids.
We can think of it as a generalization of the Fourier transform for functions
with the periodicity of the lattice. It will turn out to be the domain on which
the energy bands of a solid are dened.
The reciprocal lattice of a Bravais lattice is dened as the set of all points

K
that fulll
e
i

R
= 1

R Bravais lattice (3.3)


Remarks:
1. Plane waves with wavevectors

K from the reciprocal lattice have the pe-
riodicity of the Bravais lattice
e
i

K(r+

R)
= e
i

Kr
(3.4)
2. The reciprocal lattice is itself a Bravais lattice as one veries easily:
e
i(

K1+

K2)

R
= e
i

K1

R
e
i

K2

R
= 1 (3.5)
Conventionally the term Bravais lattice is used to refer to direct space.
3. The reciproal lattice of a lattice with a nontrivial basis always refers to
the underlying Bravais lattice.
Introduction to Solid State Theory (Kehrein) 24 University of Goettingen
3 Crystal Structure
The primitive lattice vectors of the reciprocal lattice can be constructed explic-
itly from the primitive lattice vectors a
i
in direct space:

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

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

b
3
= 2
a
1
a
2
a
1
(a
2
a
3
)
(3.6)
One veries this immediately from

b
i
a
j
= 2
ij
.
The Wigner-Seitz cell of the reciprocal lattice is called the rst Brillouin zone.
Conventionally the term Wigner-Seitz cell refers to direct space, and rst Bril-
louin zone to reciprocal space. The rst Brillouin zone will turn out to be the
domain in which the energy bands of a solid are dened.
For example the reciprocal lattice of a simple cubic lattice with primitive lattice
vectors
a
1
= ae
x
, a
2
= ae
y
, a
3
= ae
z
(3.7)
is again a simple cubic lattice with primitive lattice vectors

b
1
=
2
a
e
x
,

b
2
=
2
a
e
y
,

b
3
=
2
a
e
z
(3.8)
However, usually the type of the reciprocal lattice is dierent from the direct
lattice: e.g. the reciprocal lattice of a body-centered cubic lattice is a face-
centered cubic lattice, and vice versa.
3.3 Classication of Bravais lattices and crystal
structures
So far we have only looked at the translation symmetry of a lattice. However,
there are additional symmetry operations like rotations that will turn out to be
of importance for understanding the band structure of solids.
We will rst investigate the classication of Bravais lattices, that is we have a
trivial basis.
Bravais lattices:
All rigid (= distance conserving) mappings of a Bravais lattice onto itself are
denoted the symmetry group or space group of the Bravais lattice. One can
show (nontrivial) that every element of can be written in one of the following
ways:
Introduction to Solid State Theory (Kehrein) 25 University of Goettingen
3 Crystal Structure
1. Translation by a Bravais lattice vector.
2. A symmetry operation which leaves at least one point of the lattice in-
variant: These are i) rotations, ii) reections on a plane and iii) inversion
at a point.
3. Succesive application of 1 and 2.
Symmetry operations of the Bravais lattice, which leave at least one point in-
variant, form a group, the so called point group T. T and the translation group
are subgroups of the space group (this is generally only true for lattices with
a trivial basis, see below).
One can show (nontrivial) that there are exactly seven dierent point groups of
Bravais lattices, which one calls crystal systems: cubic, tetragonal, orthorombic,
monoclinic, triclinic, trigonal, hexagonal. The following operations are possible
elements of the point group:
1. Rotation by an integer multiple of
2
n
around an axis (so called n-fold
rotation axis). One can verify (problem set) that only the values n =
2, 3, 4, 6 are possible in Bravais lattices.
2. Rotation-reection: A rotation by
2
n
followed by a reection at a plane
perpendicular to the axis of rotation.
3. Rotation-inversion: A rotation by
2
n
followed by inversion at a point on
the axis of rotation.
4. Reection at a plane.
5. Inversion at a point.
The largest point group is the one belonging to the cubic crystal system: it is
called the full octahedral group O
h
and has 48 elements. The smallest point
group is the one of the triclinic crystal system: it has only two elements, the
identity operation and inversion.
Taking into account the full space group including translations, one can classify
a total of 14 dierent Bravais lattices. In order to do this one needs to explain
what one means by dierent or same: Two Bravais lattices are of the same
kind if they can be continuously deformed into one another without violating
the point group symmetry anywhere along the deformation path.
For example two tetragonal lattices with dierent ratios c/a belong the same
kind of Bravais lattice by this denition, whereas a simple cubic and a body
centered cubic lattice are dierent (although both belong to the cubic crystal
symmetry class).
Crystal structures
We are now ready to look at the classication of Bravais lattices with a nontriv-
ial basis, that is general crystal structures. Since the basis need not have the
Introduction to Solid State Theory (Kehrein) 26 University of Goettingen
3 Crystal Structure
full symmetry of the point group of the underlying Bravais lattice, the overall
symmetry of the space group (rigid symmetry transformations of the crystal
structure) can be reduced in many dierent ways.
1
One nds a total of 230 dif-
ferent space groups or crystal structures in this manner. Likewise, there are
now 32 dierent point groups of a Bravais lattice with a basis, the so called
crystallographic point groups.
The elements of the space group can be expressed as either
1. g[a (so called Wigner notation) corresponding to the transformation
r

= g r +a (3.9)
where g is either rotation, reection or inversion, and a =

n
i
a
i
is the
translation by a Bravais lattice vector
2. a screw axis: The crystal structure remains invariant by translation through
a non-Bravais lattice vector followed by a rotation around the axis of this
vector
3. a glide plane: The crystal structure remains invariant by translation
through a non-Bravais lattice vector followed by reection at a plane that
contains this vector
The space groups where all elements have the structure g[a are called sym-
morphic space groups. There are a total of 73 symmorphic space groups and we
will usually simplify things by only considering such symmorphic space groups
(since then the point group is a subgroup of the space group, which makes it
easier to classify the band structure). For the remaining 197 non-symmorphic
space groups I refer to the advanced literature on this topic: non-symmorphic
space groups also contain screw axes or glide planes.
Notice that the existence of screw axes or glide planes symmetries requires
specic relations between the dimensions of the basis and the Bravais lattice
vectors. An important example is the monoatomic hexagonal closed packed
structure (hcp) that contains both a screw axis and a glide plane.
3.4 Scattering by crystals
We consider the problem of wave scattering by a sample. The three by far
most important examples in solid state physics are the scattering of electromag-
netic waves (x-ray scattering), of electrons or neutrons (matter waves). For the
time being we focus on matter wave scattering and briey review the relevant
concepts from quantum mechanics.
1
Imagine for example a cube with its two halves colored in dierent ways: this object no
longer displays the symmetry of the full octrahedal group.
Introduction to Solid State Theory (Kehrein) 27 University of Goettingen
3 Crystal Structure
Potential scattering in d = 3
Our starting point is the stationary Schrodinger equation
_


2
2m
+V (r)
_
(r) = E (r) (3.10)
We rewrite it as
_
+k
2
U(r)
_
(r) = 0 (3.11)
where E =

2
k
2
2m
is the kinetic energy of the incoming particle (we assume that
the potential falls o suciently quickly as r and U(r) =
2m

2
V (r).
As the boundary condition for the incoming wave we use (remember the d =
1 case from your quantum mechanics course) a plane wave with wave vector

k
i
, which we take in z-direction. We assume that the potential is localized
somewhere around the origin. Then asymptotically for large [r[ the solutions of
(3.11) take the following form
(r)
r
e
ikz
+f(, )
e
ikr
r
(3.12)
since the probability current needs to be conserved (
1
r
-decay of the scattered
waves). and are the spherical coordinates of r, and can therefore also be
thought of as the spherical coordinates of the wave vector

k
f
under which the
scattered particle is detected. The number of particles dn scattered into d
denes the dierential cross section
dn = F
i
(, ) d (3.13)
where F
i
is the ux of incoming particles. One can show easily
(, ) = [f(, )[
2
(3.14)
so the problem is reduced to calculating the scattering amplitude f(, ).
The most convenient way to do this is by using an equivalent integral equa-
tion formulation of (3.11). Let G(r) be a Greens function of the homogeneous
equation (we only need to specify it later)
(+k
2
) G(r) = (r) (3.15)
and
0
(r) be a solution of the homogeneous equation
(+k
2
)
0
(r) = 0 (3.16)
Introduction to Solid State Theory (Kehrein) 28 University of Goettingen
3 Crystal Structure
Then every (r) which fullls the integral equation
(r) =
0
(r) +
_
d
3
r

G(r r

) U(r

) (r

) (3.17)
is a solution of the stationary Schrodinger equation (3.11). One can verify this
easily by applying (
r
+ k
2
) on both sides of (3.17) and making use of (3.15)
and (3.16).
One can iterate the integral equation (3.17)
(r) =
0
(r) +
_
d
3
r

G(r r

) U(r

) (r

) (3.18)
=
0
(r) +
_
d
3
r

G(r r

) U(r

)
0
(r

) (3.19)
+ +
_
d
3
r

d
3
r

G(r r

) U(r

) G(r

) U(r

) (r

)
= . . .
which immediately sorts the various terms in powers of the (small) scattering
potential U(x). The rst Born approximation retains only the rst power in the
scattering potential and yields
(r) =
0
(r) +
_
d
3
r

G(r r

) U(r

)
0
(r

) +O(U
2
) (3.20)
Next we specically choose

0
(r) = e
i

kir
(3.21)
that is an incoming plane wave in direction

k
i
, and the Greens function
G(r) =
1
4
e
ik |r|
[r[
(3.22)
where k = [

k
i
[ (property (3.15) can be veried easily). The advantage of the
integral formulation now becomes apparent since (3.20) generates an eigensolu-
tion with the boundary condition set by
0
(r). Inserting everything into (3.20)
we arrive at
(r) = e
i

kir

1
4
_
d
3
r

e
ik |rr

|
[r r

[
U(r

) e
i

kir

(3.23)
In order to identify the scattering amplitudes we can analyze this expression for
large distances [r[. In this limit one veries immediately
[r r

[ r r r

+O
_
r

r
2
_
(3.24)
Introduction to Solid State Theory (Kehrein) 29 University of Goettingen
3 Crystal Structure
and hence
e
ik |rr

|
[r r

[
r

e
ikr
r
e
ik rr

(3.25)
=
e
ik|r|
r
e
i

k
f
r

(3.26)
plus terms that decay more quickly for large distances. Here we have introduced
the nal scattering wavevector via

k
f
def
= r [

k
i
[ (3.27)
Putting everything together
(r) = e
i

kir

1
4
e
ikr
r
_
d
3
r

U(r

) e
i(

k
f

ki)r

(3.28)
and by comparison with (3.12) we can read o the scattering amplitude
f(, ) ==
1
4
_
d
3
r

e
i(

k
f

ki)r

U(r

) (3.29)
and the dierential cross section
(, ) =
m
2
4
2

_
d
3
r

e
i

Kr

V (r

2
(3.30)
where

K
def
=

k
f

k
i
is the momentum transfer of the scattering process. Here
, enter via r = r(, ) from (3.27), i.e. they determine the direction (

k
f
)
of the scattered particle. The important observation is that in the rst Born
approximation the scattering amplitude is simply the Fourier transform of the
scattering potential at a wavevector given by the momentum transfer

K.
This has immediate consequences when we are considering scattering by a pe-
riodic structure. While we do not know exactly the form of the scattering
potential in a crystal (e.g. for neutron scattering), we can still say that it must
have the translation symmetry of the underlying Bravais lattice
V (r) = V (r +

R) (3.31)
for all Bravais lattice vectors

R. The scattering amplitude is therefore
f(

K) =
m
2
2
_
d
3
r e
i

Kr
V (r) (3.32)
=
m
2
2

R
_
xu.c.
d
3
xe
i

K(

R+x)
V (

R +x) (3.33)
Introduction to Solid State Theory (Kehrein) 30 University of Goettingen
3 Crystal Structure
where the sum runs over all lattice vectors

R and the integration over the unit
cell (u.c.). Because of (3.31) this can be written as
f(

K) =
m
2
2
_
_

R
e
i

R
_
_
_
xu.c.
d
3
xe
i

Kx
V (x) (3.34)
Apart from the Fourier transform of the scattering potential within the unit cell
(we will say more about this later), the characteristic feature for scattering by
a periodic potential is the sum

R
e
i

R
(3.35)
over all Bravais lattice vectors. Clearly, when the momentum transfer

K is a
reciprocal lattice vector, all terms in the sum are 1 because of (3.3). Hence

R
e
i

R
= number of unit cells in sample (3.36)
that is eectively innite. Likewise, when

K does not belong to the reciprocal
lattice one is summing up numbers distributed over the unit circle that eec-
tively average to zero.
We will take a small moment to check this statement in d = 1. From the
mathematical point of view it is the special case

n=
e
2ikn
=

m=
(k m) (3.37)
of the Poisson sum rule which has to be understood as an equality for distri-
butions: both sides need to be multiplied by suciently nice test function of k
and integrated over k. Notice that (3.37) is equivalent to the above statement:
When k is a reciprocal lattice vector (here: any integer number m) the sum over
the phases becomes a -function, otherwise it vanishes.
A nice proof of the Poisson sum rule uses the Dirac comb
C(x) =

n=
(x n) (3.38)
For convenience we use the following denitions of the Fourier transform

f(k)
def
=
_

dxe
2ikx
f(x) (3.39)
f(x) =
_

dk e
2ikx

f(k) (3.40)
Introduction to Solid State Theory (Kehrein) 31 University of Goettingen
3 Crystal Structure
One veries easily that the Dirac comb is identical with its Fourier transform

C(k) =

m=
(k m) (3.41)
for example by realizing that C(x) is a periodic function with period 1, that can
therefore be expressed as a Fourier series
C(x) =

m=
a
m
e
2imx
(3.42)
with some coecients a
m
. Therefore

C(k) =

m=
a
m
(k m) (3.43)
and applying this procedure another time leads to a
2
m
= 1, and hence a
m
= 1.
If one now looks at the convolution of an arbitrary function f(x) with the Dirac
comb one has
(f C)(x) =
_
dy C(y) f(x y) =

f(x n) (3.44)


(f C)(k) =
_
dxe
2ikx

f(x n) (3.45)
xx+n
=

n=
e
2ikn

f(k) (3.46)
From the convolution theorem we also know

(f C)(k) =

f(k)

C(k) =

f(k)

m=
(k m) (3.47)
and comparison of both expressions shows (3.37)

n=
e
2ikn
=

m=
(k m) (3.48)
The structure factor f(

K) from (3.34) can therefore only be nonzero if the


momentum transfer

K is a reciprocal lattice vector

K =

k
f

k
i
(3.49)
= h

b
1
+k

b
2
+l

b
3
(3.50)
This is called the Laue condition. Two remarks:
Introduction to Solid State Theory (Kehrein) 32 University of Goettingen
3 Crystal Structure
1. Here we are only considering elastic scattering, i.e. [

k
i
[ = [

k
f
[. Especially
in neutron scattering experiments one makes use of inelastic scattering for
studying the phonon dispersion as will be discussed later in Chapter 5.
2. One might wonder what happens to overall momentum conservation if

K ,= 0 in a scattering process. The answer lies in our simplied modeling


of the crystal as a rigid potential without kinetic energy for the nuclei. In
reality the momentum transfer

K contributes to the center of mass mo-
tion of the crystal (and is therefore completely negligible for macroscopic
samples).
For (3.50) one also writes

K = [h k l] with the so called Miller indices [h k l].
In order to understand the picture behind Miller indices, we rst introduce the
concept of a lattice plane: A lattice plane goes through at least three non-
collinear points of the Bravais lattice. A family of lattice planes are lattice
planes with a xed distance from one another. One can now easily show that
there is a one to one correspondence between families of lattice planes and
reciprocal lattice vectors: these lattice planes are perpendicular to

K and have
a separation d = 2/[

K[.
One can verify this immediately by constructing planes from the condition

K
r = 2 m, where m is some integer. Because of e
iKr
= 1 for all Bravais lattice
vectors this construction assigns each Bravais lattice point to one of the planes.
The Miller indices [h k l] therefore designate a family of lattice planes.
The Bragg scattering condition starts from the picture of specular reection from
such a family of lattice planes with distance d. If is the angle between incoming
ray and plane, then the path distance between two neighboring reected rays is
2d sin and constructive interference only occurs if
2d sin = n (3.51)
where n is an integer and the wavelength of the scattering wave. (3.51) is
called the Bragg condition and one can show (tutorial) that Bragg condition
(3.51) and Laue condition (3.50) are in fact equivalent.
Neutron scattering:
In the case of neutrons the potential V (x) describes the very short range nuclear
forces:
V (x) =

j in uc
b
j
(x x
j
) (3.52)
Here the sum runs over all the atoms in a unit cell. b
j
is the nuclear scattering
length of the atom located at x
j
. The structure factor is then
S(

K) =

j in uc
b
j
e
i

Kxj
(3.53)
Introduction to Solid State Theory (Kehrein) 33 University of Goettingen
3 Crystal Structure
x-ray scattering:
Photons are not described by the Schrodinger equation that we have used in
the derivation of the Laue condition. On the other hand, the Bragg derivation
employs the very simplied picture of geometrical optics. It is not trivial to
show that a more microscopic derivation still arrives at the same nal result for
the scattering of electromagnetic waves as for matter waves. This was rst done
by P. Ewald in the so called dynamic theory of x-ray scattering. In essence one
needs to model the generation and superposition of waves emitted by oscillating
dipoles, which in turn stem from the electrons driven by electromagnetic radi-
ation. The nal result is such that one can think of the scattering potential in
the above expression as being proportional to the density of electrons
V (x)

j in uc
Z
j
g
j
(x x
j
) (3.54)
Z
j
is the nuclear charge of ion j and g
j
is a short-range function that models
its electron cloud. However, for the wave length of x-rays this function is not
well described by a -function and the structure factor
S(

K) =

j in uc
Z
j
f
j
(

K) e
i

Kxj
(3.55)
contains the Fourier transform of g
j
.
Notice that the structure factor (3.53) or (3.55) can vanish for certain direc-
tions

K for a monoatomic basis that is commensurate with the Bravais lattice.
One calls these systematic absences since it implies that certain Bragg peaks are
missing.
The following reformulation of the Laue condition is useful for analyzing scat-
tering experiments: For elastic scattering one has k = k

, therefore
k

= [

k

K[ (3.56)
k
2
= k
2
2

k

K +K
2
(3.57)


k

K =
1
2
K (3.58)
This expresses the possibility for elastic scattering solely through the incoming
wave vector. The geometric meaning of (3.58) is that the tip of the incoming
wave vector

k must be located on a Bragg plane, that is a plane perpendicular to


a given reciprocal lattice vector

K and bisecting it. Bragg planes are not dense,
therefore in general the scattering condition is not fullled and experiments
make use of the following procedures to ensure a nonzero result:
1. Laue method: One uses a single crystal with a xed orientation, but the
incoming x-rays are not monochromatic. Therefore one can fulll the
scattering condition (3.58) for some k in the spectrum and nds Bragg
peaks.
Introduction to Solid State Theory (Kehrein) 34 University of Goettingen
3 Crystal Structure
2. Rotating crystal method: One uses a single crystal and monochromatic
radiation, but rotates the crystal.
3. Debye-Scherrer method: One uses monochromatic radiation but a poly-
crystalline sample. This can be thought of as an average over the rotating
crystal scattering picture and yields the Debye-Scherrer rings.
Scattering by amorphous materials or liquids does not lead to sharp Bragg
peaks, but the diraction spectrum often exhibits smooth peaks indicative of
preferred distances in the sample. Upon cooling and ordering these peaks be-
come increasingly sharp and turn into the crystal Bragg peaks.
Until 1982 it was the general consensus of the solid state community that the
only possibility for microscopic order is a periodic arrangement of some (possibly
large) unit cell. No counterexamples to this observation were known until in
1982 Dan Shechtman found sharp Bragg peaks with 5-fold symmetry in an
alloy of aluminum and manganese.
2
This rotation symmetry is forbidden in
a periodic system (as we have learned before) and the inescapable conclusion
is that this alloy is not periodic. Today we know that it has a quasiperiodic
structure, hence the name quasicrystal. Quasiperiodic designates structures
that are incommensurate slices of some higher-dimensional periodic structure.
A beautiful example is the Penrose tiling as a two dimensional slice through
a ve dimensional hypercube. These observations have led to a broadening of
the denition of crystal in 1991 by the International Union of Crystallography:
Any structure with an essentially discrete diraction spectrum is now called a
crystal.
2
Since then many more such systems have been found and D. Shechtman got awarded the
Nobel prize in 2011.
Introduction to Solid State Theory (Kehrein) 35 University of Goettingen
Chapter 4
Band Structure
In the fundamental Hamiltonian (1.32) of solid-state physics we now consider
the following parts
H
0
= E
(0)
n
+T
e
+V
(0)
en
(4.1)
We do not seriously address the question where the crystal structure underlying
E
(0)
n
and V
(0)
en
originates,
1
but use it as an input for our subsequent investigation.
Clearly H
0
therefore has the symmetry g[a of the underlying crystal structure.
H
0
is the minimum model for making any material specic statements in solid
state physics and therefore of fundamental importance. Also, it serves as a
starting point for investigating all other neglected terms using perturbation
theory, and its properties must therefore be very well understood. One property
that makes H
0
very appealing is that it can be thought of as a sum of one-
electron Hamiltonians, just like in the homogeneous electron gas. Hence it will
be sucient to solve a one-particle problem and we can again nd the many-
particle ground sate by constructing the Slater determinant like in Chapter 2.
In writing down (4.1) we have made the following three approximations:
1. We have neglected all perturbations of the periodic crystal structure (which
always occur in real materials) and surfaces.
2. We have neglected the lattice dynamics.
3. We have neglected the electron-electron interaction.
We will say more about these approximations in subsequent chapters. Especially
for transport properties it will be essential to lift the approximations 1 and 2.
2
1
This topic will be dealt with in the Masters course on Solid-State Theory.
2
Otherwise transport will be either innite or zero as will be discussed in Chapter 6.
36
4 Band Structure
4.1 Bloch theorem
We rst discuss the consequences of translation invariance on the spectrum
of the Hamiltonian. This leads to the most important theorem of solid state
theory, the so called Bloch theorem. The investigation of the point group part
of the space group is postponed until section 4.4 (and will also not be quite as
important).
The relevant one-particle Hamiltonian is
H =

2
2m
+U(r) (4.2)
where U(r) is the periodic lattice potential produced by the nuclei: U(r +

R) =
U(r) for all Bravais lattice vectors

R. Notice that we can ignore the constant
E
(0)
n
in (4.1).
The Bloch theorem now states the following regarding the eigenfunctions of
(4.2): All eigenfunctions can be written as a plane wave multiplied by a function
with the periodicity of the Bravais lattice

n,

k
(r) = e
i

kr
u
n,

k
(r) (4.3)
with
u
n,

k
(r +

R) = u
n,

k
(r) (4.4)
Here

k is a continuous quantum number (called crystal momentum) and n is an
additional discrete quantum number, more about this later. (4.3) and (4.4) are
also called Bloch eigenfunctions.
An immediate consequence of (4.3) and (4.4) is the observation that the wave-
function only acquires a phase
3
when one moves from one unit cell to another

n,

k
(r +

R) = e
i

n,

k
(r) (4.5)
Not surprisingly all observables that only depend on the modulus squared of
the wave function are identical in all unit cells as expected due to translation
symmetry. However, the additional phase factor in (4.5) will be of fundamental
importance for the spectrum and the dynamics.
Finally, let us mention that in the limit of a vanishing lattice potential the Bloch
theorem just reproduces the well known plane wave form of the eigenfunctions
for the homogeneous electron gas: u(r) turns out to be a constant in this case.
Because of the great importance of the Bloch theorem we will oer two proofs,
one based on (elegant) symmetry considerations and a second constructive proof
3
Strictly speaking we still need to show that

k is real as will be done later in this section.
Introduction to Solid State Theory (Kehrein) 37 University of Goettingen
4 Band Structure
(which will be the basis for latter calculations of the band structure for given
potentials U(r)).
Proof a (symmetry based):
We dene translation operators T

R
for all Bravais lattice vectors

R
T

R
f(r)
def
= f(r +

R) (4.6)
Obviously H from (4.2) commutes with all translation operators,

R [H, T

R
] = 0 (4.7)
Also one has
T

R
T

= T

R+

(4.8)
and therefore

R,

R

[T

R
, T

] = 0 (4.9)
We have therefore shown that H and all T

R
form a set of commuting observables.
Hence there exists a common eigenbasis
H = E (4.10)
T

R
= c(

R) (4.11)
Since the translation subgroup of the space group is abelian, we can easily work
out the eigenvalues c(

R). Clearly
c(

R) c(

) = c(

R +

R

) (4.12)
because of (4.8). If a
j
denote the three primitive Bravais lattice vectors we can
nd (in general complex) numbers x
j
such that
c(a
j
) = e
2ixj
(4.13)
For a general Bravais lattice vector

R = n
1
a
1
+n
2
a
2
+n
3
a
3
this implies
c(

R) = c(a
1
)
n1
c(a
2
)
n2
c(a
3
)
n3
(4.14)
= e
i

kR
(4.15)
with

k
def
= x
1

b
1
+x
2

b
2
+x
3

b
3
, where

b
i
are the reciprocal lattice vectors,

b
i
a
j
=
2
ij
.
Introduction to Solid State Theory (Kehrein) 38 University of Goettingen
4 Band Structure
Putting everything together we have shown
T

R
(r) = (r +

R) (4.16)
!
= c(

R) (r) = e
i

R
(r) (4.17)
Dening
u(r)
def
= e
i

kr
(r) (4.18)
we have therefore proven the Bloch theorem since u(r) has the periodicity of
the Bravais lattice
u(r +

R) = e
i

k(r+

R)
(r +

R) = e
i

k(r+

R)
e
i

kR
(r) (4.19)
= e
i

kr
(r) (4.20)
= u(r) (4.21)
Before looking at the constructive proof of the Bloch theorem we will address
the question which wave vectors

k are allowed in a large crystal. Like for the
homogeneous electron gas we will use Born-von Karman boundary conditions.
In order to do this we consider a large volume V which is commensurable with
the primitive unit cell, i.e. we take N
1
unit cells in direction a
1
, N
2
unit cells in
direction a
2
, etc. The total number of unit cells in V is N = N
1
N
2
N
3
and the
Born-von Karman boundary condition implies
(r +N
i
a
i
) = (r) (4.22)
for i = 1, 2, 3. From the Bloch theorem we now know

n,

k
(r +N
i
a
i
) = e
iNi

kai

n,

k
(r) (4.23)
!
=
n,

k
(r) (4.24)
Hence for

k = x
1

b
1
+x
2

b
2
+x
3

b
3
we nd
e
2iNixi
= 1 (4.25)
and therefore x
i
=
mi
Ni
, m
i
Z. The allowed values of

k (we will call these the
Born-von Karman set) are then

k =
3

i=1
m
i
N
i

b
i
, m
i
Z (4.26)
For later considerations we work out the volume k associated with each allowed
value of

k: this is simply
k =

b
1
N
1

b
2
N
2

b
3
N
3
_
=
1
N

b
1
(

b
2

b
3
) (4.27)
Introduction to Solid State Theory (Kehrein) 39 University of Goettingen
4 Band Structure
Now

b
1
(

b
2

b
3
) is just the volume of the unit cell in the reciprocal lattice, and
using the explicit construction (3.6) one immediately sees

b
1
(

b
2

b
3
) =
(2)
3
v
(4.28)
where v =
V
N
= a
1
(a
2
a
3
) is the volume of the primitive unit cell in direct
space. Hence
k =
(2)
3
V
(4.29)
like for the homogeneous electron gas.
Proof b (explicit construction):
Every wavefunction which fullls the Born-von Karman boundary condition can
be expanded as
(r) =

q
c
q
e
i qr
(4.30)
where q runs over the Born-von Karman set
q =
3

i=1
m
i
N
i

b
i
(4.31)
Likewise the periodic lattice potential can be expanded as a sum over all recip-
rocal lattice vectors
U(r) =

K
U

K
e
i

Kr
(4.32)
with
U

K
=
1
v
_
C
dr e
i

Kr
U(r) (4.33)
where the integration runs over the primitive unit cell (.
4
We now plug (4.30)
and (4.32) into the Schrodinger equation:
p
2
2m
(r) =

2
2m
q
2
c
q
e
i qr
(4.34)
4
This is simply a generalization of the usual Fourier series: Let f(r) be dened in [0, a] with
f(0) = f(a). Then f(r) =
P

n=
exp(2inr/a) fn with fn =
1
a
R
a
0
dr exp(2inr/a) f(r).
Introduction to Solid State Theory (Kehrein) 40 University of Goettingen
4 Band Structure
U (r) =
_
_

K
U

K
e
i

Kr
_
_
_
_

q
c
q
e
i qr
_
_
=

K, q

K
c
q

K
e
i q

r
(4.35)
where one can verify easily that q

= q +

K is also an element of the Born-
von Karman set. Comparing the coecients in the Schrodinger equation we
conclude
_

2
2m
q
2
E
_
c
q
+

c
q

= 0 (4.36)
for all q in the Born-von Karman set. Now one can always write
q =

k

K (4.37)
where

k lies in the rst Brillouin zone and



K is a reciprocal lattice vector. Then
we arrive at the key result
_

2
2m
(

k

K)
2
E
_
c

K
+

K
c

= 0 (4.38)
which must hold for all reciprocal lattice vectors

K. For any given

k in the rst
Brillouin zone one observes that (4.38) only couples coecients c

k
, c

K
, . . .
that dier by a reciprocal lattice vector. Hence the initial Schrodinger equation
decouples into N independent problems, one for each Born-von Karman value
in the rst Brillouin zone, and the solutions labelled by

k can be written as

k
(r) =

K
c

K
e
i(

K)r
(4.39)
= e
i

kr
u

k
(r) (4.40)
where we have dened
u(r)
def
=

K
c

K
e
i

Kr
(4.41)
Clearly u(r +

R) = u(r) due to the basic properties of reciprocal lattice vectors.
This completes our second proof of the Bloch theorem.
Remarks:
Introduction to Solid State Theory (Kehrein) 41 University of Goettingen
4 Band Structure
1. The good quantum number

k is called crystal momentum. Proof b demon-


strated that

k can be restricted to the rst Brillouin zone. Therefore the
rst Brillouin zone emerges as the denition range of the electronic band
structure in a crystal, which makes it one of the most important concepts
in solid state theory. Notice that

k is not the physical momentum, but
for the time being should be thought of as a good quantum number in
a periodic potential. We will say more about the physical meaning of
the crystal momentum when we study the dynamics of Bloch electrons in
chapter 6.
2. The role of the discrete quantum number n (band index) becomes clear
when we insert the Bloch form (4.3) of the eigenfunctions into the Schrodinger
equation. This leads to
H

k
u

k
(r) = E

k
u

k
(r) (4.42)
with periodic boundary conditions u

k
(r +

R) = u

k
(r) and
H

k
=

2
2m
(i

k)
2
+U(r) (4.43)
This constitutes a hermitean eigenvalue problem in the nite volume (
(primitive unit cell). Therefore we expect a discrete spectrum labelled by
a discrete quantum number n leading to
E
n,

k
= E
n
(

k) (4.44)
Upon variation of

k in the rst Brillouin zone the energy E
n
(

k) will vary
continuously for xed band index n. This yields the electronic band struc-
ture.
3. If one builds up a localized electron from eigenfunctions in band n with
wavevectors around

k, its group velocity is given by
v
n
(

k) =
1

k
E
n
(

k) (4.45)
according to quantum mechanics. It is quite remarkable that one can
construct such solutions with constant nonzero group velocity in spite of
the periodic scattering in the crystal. Notice that (4.45) also shows that for
a general dispersion relation (,= free electrons) there is no simple relation
between the crystal momentum and the physical momentum mv
n
(

k).
Introduction to Solid State Theory (Kehrein) 42 University of Goettingen
4 Band Structure
4.2 Nearly free electron gas
We now look more explicitly at the band structure which emerges for a weak pe-
riodic potential. Weak will turn out to mean small with respect to the kinetic
energy, and we can study this case in perturbation theory. The perturbation
consists of the Fourier components (4.32) U

K
of the periodic potential, where

K are the reciprocal lattice vectors. In the sequel we take


U

0
= 0 (4.46)
without loss of generality, since the average of the potential over the primitive
unit cell can always be incorporated into a constant in the Hamiltonian. Also
for a real potential one has for all

K
U

K
= U

K
(4.47)
In proof b of the Bloch theorem we have shown that for given

k in the rst
Brillouin zone the Fourier coecients of the Bloch eigenfunction obey
_

2
2m
(

k

K)
2
E
_
c

K
+

K
c

= 0 (4.48)
for all

K. The solution then gives
u

k
(r) =

K
c

K
e
i

Kr
(4.49)
Now the starting point of perturbation theory is the homogeneous electron gas
with U 0. Its eigenfunctions are plane waves (r) = e
i pr/
. For all p one can
nd a reciprocal lattice vector

K
0
such that
p = (

k

K
0
) (4.50)
with

k in the rst Brillouin zone. Hence u

k
(r) = e
i

K0r
and the starting point
of our perturbation expansion in U is:
c

K0
= 1 (4.51)

K ,=

K
0
c

K
= 0 (4.52)
E =

2
2m
(

k

K
0
)
2
(4.53)
If we plug this into (4.48) we nd to lowest order in U

K ,=

K
0
c

K
=
2mU

K0

2
[(

k

K)
2
(

k

K
0
)
2
]
(4.54)
Introduction to Solid State Theory (Kehrein) 43 University of Goettingen
4 Band Structure
This yields a consistent expansion in powers of U as long as the denominator is
not degenerate

2
2m

k

K)
2
(

k

K
0
)
2

[U[ (4.55)
and one shows easily
c

K0
= 1 +O(U
2
) (4.56)

K ,=

K
0
c

K
= O(U) (4.57)
E =

2
2m
(

k

K
0
)
2
+O(U
2
) (4.58)
Therefore in the non-degenerate case (4.55) eigenenergies and eigenfunctions
are only slightly changed from the free electron case. Notice that according to
(4.54) the periodic perturbation only couples coecients with the same good
quantum number

k. Hence a breakdown of non-degenerate perturbation theory
only occurs if there exists

K
1
,=

K
0
with
5
(

k

K
1
)
2
= (

k

K
0
)
2
(4.59)
With the wave vector q =

k

K
0
of the free electron we can rewrite this as
(q

K)
2
= q
2
(4.60)
where

K =

K
1


K
0
is a nonzero reciprocal lattice vector. The condition (4.60)
is identical to the Laue condition for scattering of a periodic potential (3.58).
We conclude that non-degenerate perturbation theory breaks down (however
weak the potential) when the wave vector q of the free electron lies on (near) a
Bragg plane
q =
1
2

K+

K (4.61)
This is intuitively obvious from the Laue picture: For incoming wavevectors
on a Bragg plane the periodic potential has a particularly strong eect due to
interference.
We now assume that q lies on exactly one Bragg plane.
6
This means that there
are exactly two degenerate states with the same quantum number

k and (4.48)
yields
_
E

2
2m
(

k

K
0
)
2
_
c

K0
= U

K1

K0
c

K1
_
E

2
2m
(

k

K
1
)
2
_
c

K1
= U

K0

K1
c

K0
(4.62)
5
The same observations hold for near degeneracy.
6
The situation of multiple Bragg planes can be discussed analogously, one just needs to
take more coecients into account.
Introduction to Solid State Theory (Kehrein) 44 University of Goettingen
4 Band Structure
where we have used U

0
= 0, plus terms in higher order in U. A nonzero solution
of this linear system of equations exists only if
det
_
E

2
2m
(

k

K
0
)
2
U

K1

K0
U

K1

K0
E

2
2m
(

k

K
1
)
2
_
= 0 (4.63)
This yields the two solutions
E =

2
2m
(

k

K
0
)
2
[U

K1

K0
[ (4.64)
where we have made explicit use of the degeneracy condition.
We observe that the periodic potential splits the dispersion relation in order U
if the wave vector of the free electron lies on a Bragg plane, which therefore
leads to band gaps in the dispersion relation (forbidden energy regions in the
rst Brillouin zone). These band gaps will turn out to be the key ingredient for
insulating behavior.
From (4.62) we can now also nd the structure of the Bloch eigenfunctions at
to these band gaps:
c

K0
= sgn(U

K1

K0
) c

K1
(4.65)
E.g. for U

K1

K0
> 0 (the other sign just leads to an exchange of + and -) one
nds:

=

2
2m
(

k

K
0
)
2
U

K1

K0
(4.66)

(r) = e
i

kr
_
e
i

K1r
e
i

K0r
_
(4.67)
[
i
(r)[
2
sin
2
_
1
2
(

K
1


K
0
) r
_
(4.68)

E
+
=

2
2m
(

k

K
0
)
2
+U

K1

K0
(4.69)

+
(r) = e
i

kr
_
e
i

K1r
+e
i

K0r
_
(4.70)
[
i
(r)[
2
cos
2
_
1
2
(

K
1


K
0
) r
_
(4.71)
One observes the characteristic change of the Bloch eigenfunctions at a band
gap: A node at the lattice points turns into a maximum, and vice versa.
Introduction to Solid State Theory (Kehrein) 45 University of Goettingen
4 Band Structure
We now consider the many-electron problem in the periodic potential. Just
as in the case of the homogeneous electron gas, a certain electron density n is
xed and determined by overall charge neutrality. The zero temperature ground
state is therefore given by a Slater determinant consisting of the 1-particle Bloch
eigenfunctions [n,

k, ), where =, is the spin: all such 1-particle states up to


an energy
F
contribute, E
n
(

k)
F
. The Fermi energy
F
itself is determined
by the electron density n.
We will later see (Chapter 6) that a Fermi energy
F
in a band implies metallic
behavior. Likewise,
F
in a band gap means that one has an insulator.
7
An
insulator with a band gap of less than approximately 2eV is also called a (intrin-
sic) semiconductor since thermal excitations into the unoccupied band become
possible.
The Fermi surface is dened as the surface of constant energy with E
n
(

k) =
F
.
In later chapters it will turn out to play a key role for describing the low-energy
behavior of materials. It can have a very complicated topology as we will see at
the end of this section. Notice that by denition a Fermi surface only exists in
metals.
A rst important property that we want to look at is the electron contribution
to the specic heat. The calculation runs along identical lines as for the homo-
geneous electron gas and one nds that the Sommerfeld coecient is given by
(2.45)
c
V
=

2
3
k
2
B
T
F
(4.72)
with the total density of states
F
at the Fermi energy: In general the Fermi
surface can lie in various bands and we need to sum over these contributions

F
=

n
(
F
) (4.73)
with the contribution of band n to the density of states dened as (compare
(2.26)

n
() =
_
1.BZ
d

k
4
3
( E
n
(

k)) (4.74)
The integration runs over the rst Brillouin zone.
Now one can learn a lot about
n
() by just looking at the dispersion relation.
One deduces this by rst dening S
n
() as the surface with E
n
(

k) = in the
rst Brillouin zone. The the volume between S
n
() and S
n
( +d) is just () d
() d =
_
Sn()
dS
4
3
k(

k) (4.75)
7
Strictly speaking, in an insulator
F
is dened as the zero temperature limit of the chemical
potential in the grand canonical ensemble. With this denition
F
turns out to lie in the middle
of band gaps.
Introduction to Solid State Theory (Kehrein) 46 University of Goettingen
4 Band Structure
where k(

k) is the length of a vector perpendicular to S


n
(E
n
(

k)) that connects


S
n
() and S
n
( +d). Since

E
n
(

k) is normal to the surface, we can also write


+d = +[

E
n
(

k)[ k(

k) (4.76)
k(

k) =
d
[

E
n
(

k)[
(4.77)
and we have the result

n
() =
_
Sn()
dS
4
3
1
[

E
n
(

k)[
(4.78)
Clearly, an insulator has no electronic contribution to the specic heat linear
in temperature (in fact it is exponentially suppressed). If

E
n
(

k) = 0 one
speaks of a van Hove-singularity that yields a particularly large contribution to
the density of states. In three dimensions a van Hove-singularity is integrable
and (
F
) remains nite. However, the higher order terms of the Sommerfeld
expansion need to be reconsidered since they now contain (singular) derivatives
of (). In one or two dimensions a van Hove-singularity at the Fermi energy
leads to a diverging
F
and a dierent low temperature behavior of the specic
heat (see tutorials).
Structure of Fermi surface:
We now investigate the structure of the Fermi surface in a nearly free electron
gas. One constructs it by rst drawing the simple Fermi surface (sphere) of the
free electron gas with the same density of electrons n. If this Fermi surface lies
completely in the rst Brillouin zone of the periodic potential, we are nearly done
with our construction: The only additional eect is the characteristic lowering
of the band energies at points closer to the boundary of the rst Brillouin zone
that we have derived in perturbation theory. Hence the Fermi surface of the free
electron gas at points that are close to the boundary moves even closer to the
boundary. Since the total volume enclosed by the Fermi surface has to remain
constant this implies that the Fermi surface at points that are further from the
boundary moves even further from the boundary.
If the Fermi surface of the free electron gas lies (at least partly) outside the rst
Brillouin zone the topology can become quite involved. Whenever it touches a
Bragg plane we have the shifts/opening of gaps from degenerate perturbation
theory. This e.g. often leads to characteristic neck-like structures of the Fermi
surface. In general the Fermi surface of the free electron gas will touch multiple
Bragg planes. In order to study the emerging structure we rst dene the
concept of a general Brillouin zone:
1. The rst Brillouin zone is the set of all points in

k-space that one can
reach from the origin without crossing a Bragg plane.
2. The second Brillouin zone is the set of all points that one can reach from
the rst Brillouin zone by crossing exactly one Bragg plane.
Introduction to Solid State Theory (Kehrein) 47 University of Goettingen
4 Band Structure
3. The n+1 Brillouin zone is the set of all points not contained in the n1
Brillouin zone that one can reach from the n Brillouin zone by crossing
exactly one Bragg plane.
One can show easily that all these Brillouin zones have the same volume (of the
rst Brillouin zone).
In the second step of our construction we now deform the free Fermi surface in
the characteristic manner prescribed by perturbation theory whenever it touches
a Bragg plane. Therefore the dierent Brillouin zones naturally correspond to
the dierent bands n.
In the last step of the construction one moves the contribution of the Fermi
surface in the n Brillouin zone back into the rst Brillouin zone by shifting with
suitable reciprocal lattice vectors.
8
This gives band n in the rst Brillouin zone.
Notice that in general the Fermi surface will consist of various bands, which can
make the topology very complicated.
4.3 Tight-binding ansatz
The tight-binding ansatz starts from the opposite limit than the nearly free
electron gas: In zeroth order one considers isolated atoms, whose overlap leads
to corrections in a suitable perturbative expansion around this limit.
We rst need to know the spectrum of the atomic Hamiltonian
9
H
at
=

2
2m
+U
at
(r) (4.79)
H
at

n
(r) = E
n

n
(r) (4.80)
The full crystal Hamiltonian is
H =

2
2m
+U(r) (4.81)
with the potential generated by a sum over all Bravais lattice vectors
U(r) =

R
U
at
(r

R) (4.82)
If we simply plug the atomic eigenfunctions into the full Hamiltonian we nd
H
n
(r) = H
at

n
(r) +U(r)
n
(r) (4.83)
8
One also often sees the so called repeated zone scheme where one draws the periodically
repated picture in reciprocal space.
9
More generally, this would be the Hamiltonian in a primitive unit cell.
Introduction to Solid State Theory (Kehrein) 48 University of Goettingen
4 Band Structure
where
U(r)
def
=

R=0
U
at
(r

R) (4.84)
can be thought of as the perturbation of the atomic potential in the unit cell
at the origin by the atomic potentials generated in the other unit cells. If the
atomic wavefunctions are suciently localized, they will mainly contribute in
the unit cell at the origin, where U(r) is small. In zeroth order one therefore
simply neglects the product U(r)
n
(r) and nds
H
n
(r) = E
n

n
(r) (4.85)
or likewise for all Bravais lattice vectors
H
n
(r

R) = E
n

n
(r

R) (4.86)
So we have found N degenerate eigenfunctions in zeroth order. Both for making
contact with the nearly free electron gas and for looking at corrections to this
result, it will be useful to bring these N eigenfunctions into Bloch form. We
dene for all

k from the Born-von Karman set in the rst Brillouin zone

k
(r)
def
=

R
e
i

n
(r

R) (4.87)
This has Bloch form because

k
(r +

R) =

e
i

n
(r +

R

R

) (4.88)
= e
i

e
i

k(

R)

n
(r (


R)) (4.89)
= e
i

k
(r) (4.90)
In zeroth order of the tight-binding approximation one therefore simply nds a
at band
E
n
(

k) = E
n
(4.91)
of N degenerate Bloch eigenstates with an eigenenergy given by the atomic
orbital.
Next we want to take the inuence of U(r) into account in leading order. For
the eigenfunctions of the full Hamiltonian we make the following ansatz

k
(r) =

R
e
i

R
(r

R) (4.92)
Introduction to Solid State Theory (Kehrein) 49 University of Goettingen
4 Band Structure
The same derivation (4.90) holds to verify that this ansatz is of Bloch form. Still,
one can wonder if all Bloch eigenfunctions can be written in the form (4.92).
The answer to this question is armative as will be seen in the subsection on
Wannier functions at the end of this section. So the only approximation comes
from making the additional ansatz
(r) =

n
b
n

n
(r) (4.93)
where we only want to take a small number of atomic orbitals into account.
In particular, one neglects scattering states in the decomposition (4.93) which
would be necessary to have a complete basis. Because of (4.93) the tight-binding
ansatz is also called LCAO method (linear combination of atomic orbitals).
Plugging this into the stationary Schrodinger equation with the full Hamiltonian
(4.81) yields
H

k
(r) =

R,n
e
i

R
b
n
_


2
2m
+U
at
(r

R) +U(r

R)
_

n
(r

R)
=

R,n
e
i

R
b
n
E
n

n
(r

R) (4.94)
+

R,n
e
i

R
b
n
U(r

R)
n
(r

R) (4.95)
!
= E(

k)

k
(r) (4.96)
= E(

k)

R,n
e
i

R
b
n

n
(r

R) (4.97)
One can rewrite this as

R,n
(E(

k)E
n
) e
i

R
b
n

n
(r

R) =

R,n
e
i

R
b
n
U(r

R)
n
(r

R)(4.98)
We multiply this equation by

m
(r) and integrate over r using the orthogonality
of the atomic orbitals
_
dr

m
(r)
n
(r) =
mn
(4.99)
which leads to
(E(

k) E
m
) b
m
=

R=0,n
(E(

k) E
n
) e
i

R
b
n
_
dr

m
(r)
n
(r

R)
+

n
b
n
_
dr

m
(r) U(r)
n
(r)
Introduction to Solid State Theory (Kehrein) 50 University of Goettingen
4 Band Structure
+

R=0,n
e
i

R
b
n
_
dr

m
(r) U(r

R)
n
(r

R)
(4.100)
where we have always separated the sums into terms

R = 0 and

R ,= 0. Now
all the terms on the right hand side of this equation are small: The rst and
the third term because the atomic orbitals are localized in a unit cell, and the
second term because U(r) is small in the unit cell at the origin. Hence we can
conclude that [(E(

k) E
m
) b
m
[ is small for all m. The only possible solution
is that the eigenenergy E(

k) lies close to the energy E


0
of some atomic orbital,
and that b
m
is only then not small when E
m
E
0
. In the rst nontrivial
order of the LCAO method one therefore takes only the terms with E
n
E
0
on the right hand side of (4.100) into account, since otherwise a small term b
n
is multiplied by another small integral.
If

m
is an s-orbital there are no degeneracies and (4.100) reduces to an
explicit equation for the energy of the s-band. We will study this situation
explicitly below.
For a p-orbital we have threefold degeneracy. Hence (4.100) is a linear
system of three equations, which yields E(

k) for the three p-bands.


For a d-orbital we likewise have ve bands, etc.
Another important approximation is to restrict the sums over

R to sums over
nearest neighbors (n.n.) since the overlap of the atomic orbitals decays rapidly
for larger separation.
For an s-band (4.100) becomes
E(

k) E
s
= (E(

k) E
s
)

R=0
e
i

R
(

R)

R=0
e
i

R
(

R) (4.101)
where we have dened
(

R) =
_
dr

s
(r)
s
(r

R) (4.102)
=
_
dr U(r) [
s
(r)[
2
(4.103)
(

R) =
_
dr

s
(r) U(r

R)
s
(r

R) (4.104)
This gives
E(

k) = E
s

R=0
e
i

R
(

R)
1 +

R=0
e
i

R
(

R)
(4.105)
Introduction to Solid State Theory (Kehrein) 51 University of Goettingen
4 Band Structure
The wave function for an s-orbital is real, also
s
(r) =
s
(r) and because of
inversion symmetry likewise U(r) = U(r). This yields (

R) = (

R) and
therefore

R=0
e
i

R
(

R)

n.n.
(

R) cos(

k

R) (4.106)
where we have restricted the sum to nearest neighbors. Since (

R) in the
denominator of (4.105) only gives a small correction, we have the following
result for the energy of an s-band in the rst nontrivial order of the tight-binding
approximation
E(

k) = E
s

n.n.
(

R) cos(

k

R) (4.107)
One makes the characteristic observation that the band width is set by the
overlap integral (4.104).
We now look explicitly at an s-band in an fcc-lattice. a is the size of the unit
cell. One sees immediately that the origin has 12 nearest neighbors with

R =
a
2
(1, 1, 0),
a
2
(1, 0, 1),
a
2
(0, 1, 1) (4.108)
which leads to 12 possible values of

k

R

k

R =
a
2
(k
i
k
j
) (4.109)
with i, j = x, y, i, j = x, z or i, j = y, z. Also U(r) has the full symmetry of
the cubic lattice and therefore (

R) = for all

R from (4.108). Then (4.107)
becomes
E
s
(

k) = E
s

_
2 cos(
a
2
(k
x
+k
y
)) + 2 cos(
a
2
(k
x
k
y
)) +. . .
_
= E
s

4
_
cos
_
ak
x
2
_
cos
_
ak
y
2
_
+ cos
_
ak
y
2
_
cos
_
ak
z
2
_
+cos
_
ak
x
2
_
cos
_
ak
z
2
_
_
(4.110)
with the overlap integral
=
_
dr

s
(x, y, z) U(x a/2, y a/2, z a/2)
s
(x a/2, y a/2, z a/2)
(4.111)
Introduction to Solid State Theory (Kehrein) 52 University of Goettingen
4 Band Structure
Wannier functions:
We will prove the following theorem: Every Bloch function for every band can
be written in the form

k
(r) =

R
e
i

n
(r

R) (4.112)

n
(r) is called the Wannier function for band n. Notice that we have implicitly
already used this theorem in writing down the ansatz (4.92).
The proof proceeds by considering
n

k
(r) as a function of

k for xed r. Since

k
(r) is periodic as a function of

k in the reciprocal lattice, we know that it
can be expanded in a Fourier series in the reciprocal reciprocal lattice, which is
of course just the direct lattice:

k
(r) =

R
f
n
(r,

R) e
i

k
(4.113)
where the Fourier coecients are given by
f
n
(r,

R) =
1
(

_
d

k e
i

k
(r) (4.114)
Here (

is the volume of the rst Brillouin zone. Also, the Fourier coecients
do not really depend both on r and

R since for every Bravais lattice vector

R
0
f
n
(r +

R
0
,

R +

R
0
) =
1
(

_
d

k e
i(

R+

R0)

k
(r +

R
0
) (4.115)
=
1
(

_
d

k e
i(

R+

R0)

k
e
i

R0

k
(r) (4.116)
= f
n
(r,

R) (4.117)
where we have explicitly used the Bloch form in the last step. Therefore

n
(r

R)
def
= f
n
(r,

R) (4.118)
and we have proven the above statement.
Remarks:
1. For narrow bands the Wannier functions are very similar to the atomic
orbitals.
2. For wider bands the Wannier functions are still somehow localized. They
are therefore a convenient starting point for investigations where local
properties are important, like in semiclassical transport theory or for dop-
ing in semiconductors. We will say more about this in later chapters.
Introduction to Solid State Theory (Kehrein) 53 University of Goettingen
4 Band Structure
3. Dierent from the atomic wave functions, the Wannier functions are or-
thogonal at dierent sites of the Bravais lattice
_
dr

n
(r

R)
m
(r

R

) =
nm

R,

(4.119)
This will be proven in a tutorial. (4.119) is a very convenient property for
explicit calculations.
4.4 Real band structures and symmetry prop-
erties
So far we have only made use of the periodic translation invariance as a sym-
metry property of the crystal. This has given us the Bloch form of the eigen-
functions. However, in general one also has a nontrivial point group and these
symmetries are reected in the band structure and to a certain extent determine
its general features.
In (4.43) we have already derived an eective

k-dependent Hamiltonian for
determining the Bloch eigenfunctions in a single unit cell
H

k
u

k
(r) = E(

k) u

k
(r) (4.120)
with periodic boundary conditions: u

k
(r +

R) = u

k
(r):
H

k
=

2
2m
(i

k)
2
+U(r) (4.121)
=

2
2m
+
_
U(r) +

2
2m

k
2
i

2
m

_
(4.122)
The term in the parantheses on the right hand side can be thought of as an
eective

k-dependent potential. The key question is now the interplay of the
point group symmetry of the crystal structure with this eective potential. For
simplicity, we will only consider symmorphic space groups, that is every element
g of the point group T is by itself a symmetry operation. We use the following
representation of T on the Hilbert space:
P
g
f(r)
def
= f(g
1
r) (4.123)
where g can be thought of as the matrix representing the point group operation
in position space. Clearly [P
g
, U(r)]=0 and both the constant in (4.122) and
the kinetic term commute with P
g
. The nontrivial part is how P
g
acts on the
single gradient term in the eective potential (4.122): One can show easily
P
g

r
i
=

j
g
ji

r
j
P
g
(4.124)
Introduction to Solid State Theory (Kehrein) 54 University of Goettingen
4 Band Structure
Hence
P
g
_
(

k

)f(r)
_
= (g

k) P
g
f(r) (4.125)
and therefore putting everything together:
P
g
H

k
= H
g

k
P
g
(4.126)
We now need to distinguish two cases:


k remains invariant under a certain point group operation, g

k =

k. Be-
cause of
[P
g
, H

k
] = 0 (4.127)
P
g
u

k
is again an eigenfunction (but not necessarily a new one) with the
same eigenenergy and crystal momentum.
For g

k ,=

k we conclude that P
g
u

k
is an eigenfunction of H
g

k
, therefore
we have degeneracies at dierent points of the Brillouin zone, E
n
(

k) =
E
n
(g

k).
For any given

k one denes the star of

k as the set of all vectors that are
generated by acting with the point group on

k. For a generic value of

k its star
will contain as many elements as there elements in the point group. However,
when

k lies on a symmetry line the number of elements in the star is reduced.
The subgroup of T which leaves

k invariant is called the small group of

k. The
extreme case is the -point (

k = 0) in the Brillouin zone where the small group


is obviously always identical with the full point group of the crystal structure.
10
While the case g

k ,=

k leads to a somehow trivial (expected) symmetry, the
elements of the small group allow nontrivial insights into degeneracies for given
values of the crystal momentum. In order to see this we rst distinguish be-
tween normal and accidental degeneracies: A normal degeneracy occurs when
all the degenerate eigenvectors can be connected by symmetry operations. By
denition this implies that one has an irreducible representation of the respec-
tive group. The situation with additional accidental degeneracies that cannot
be explained in this way will not concern us in the sequel.
Therefore all possible (non-accidental) degeneracies of energy bands are related
to irreducible representations of the respective small group of

k. By looking up
the possible dimensions of such irreducible representations
11
one can therefore
10
A similar situation can occur at edges of the Brillouin zone since momenta g

k which only
dier by reciprocal lattice vectors can be identied. Therefore the symmetries/degeneracies
in a band structure generically increase as one approaches the border of the Brillouin zone.
11
For example one can consult M. Tinkham, Group Theory and Quantum Mechanics, Ap-
pendix B.
Introduction to Solid State Theory (Kehrein) 55 University of Goettingen
4 Band Structure
say something about possible degeneracies along symmetry lines of the Bril-
louin zone. If one works harder one can even say something about how the
degeneracies split if one moves away from symmetry points.
Optical properties:
From the electronic band structure one can deduce some basic optical properties
of crystals. The most important process for the absorption of a photon is the
excitation of an electron from an occupied to an unoccupied state. Therefore
insulators with a band gap larger than 3.2eV appear transparent, only light
with UV frequencies or higher could possibly be absorbed by this process (there
are other weaker absorption processes like 2-photon processes, photon-phonon
scattering, etc.). Examples for such insulators are diamond, quartz and AlO.
Semiconductors will strongly absorb photons with an energy larger than the
direct bandgap, but less below. Hence they often appear colored. An example
is cadmium suld with a bandgap of 2.6ev, which therefore absorbs violet and
blue, and appears red.
The optical properties of metals are more complicated due to their large conduc-
tivity, which leads to reemission processes by the conduction electrons (therefore
they often appear shiny). This will be a topic for the Masters course in solid
state theory.
Introduction to Solid State Theory (Kehrein) 56 University of Goettingen
Chapter 5
Lattice Vibrations and
Phonons
In the fundamental Hamiltonian (1.32) of solid state theory
H = E
(0)
n
+T
e
+V
(0)
en
+H
ph
+H
eph
+V
ee
(5.1)
we have so far only considered the rst line. Specically, we have ignored the
dynamics of the lattice ions H
ph
in the second line and simply froze them at the
points specied by the Bravais lattice. This leads to some severe shortcomings:
1. We cannot explain the observed specic heat for temperature exceeding a
few K.
2. Without H
eph
we cannot explain the temperature dependent conductiv-
ity of metals.
3. Conventional superconductivity relies on H
eph
.
In this chapter we now set out to include the dynamics of the lattice ions.
5.1 Classical theory
We rst consider a classical theory, that is we think of the lattice ions as classical
particles. In the next section we will then quantize the theory, based on the
results of the classical treatment in this section.
We make two basic assumptions about the lattice dynamics:
57
5 Lattice Vibrations and Phonons
1. The average position of each lattice ions is still the (before) frozen Bravais
lattice (plus possibly basis) vector.
2. We only consider small deviations from equilibrium (in a sense to be made
more precise later).
With these assumptions we make the following ansatz for the position of ion

R:
r(

R) =

R +u(

R) (5.2)
where u(

R) is the deviation from equilibrium. The interaction potential of the


ions will only depend on the distance between two ions and can be expanded in
the small parameter u(

R)
U =
1
2

R,

(r(

R) r(

)) (5.3)
=
N
2

R
(

R)
+
1
2

R,

[u(

R) u(

)]

(

R

R

)
+
1
4

R,

_
[u(

R) u(

)]

_
2
(

R

R

)
+O(u
3
) (5.4)
The rst line on the rhs is just the constant contribution E
(0)
n
in the fundamental
Hamiltonian which plays no role in the dynamics. The second line vanishes since
by denition

R are the equilibrium positions: Therefore the force on a given ion

R exerted by all the other ions vanishes

R

R

) (5.5)
For the so called harmonic crystal we only take into account the quadratic term
in the third line of (5.4), especially we neglect all terms in O(u
3
) and higher.
Therefore we have the Hamilton function
H =

P(

R)
2
2M
+
1
2

R,

,
u

R) D

R

R

) u

) (5.6)
with
D

R

R

) =

R,

2
(r)
r

r
n
u


2
(r)
r

r
n
u

(5.7)
Introduction to Solid State Theory (Kehrein) 58 University of Goettingen
5 Lattice Vibrations and Phonons
as can be seen by expansion of the quadratic term in (5.4). Notice that speci-
cally the (unphysical) second derivative of the potential at zero separation can-
cels in (5.7).
a) One dimensional Bravais lattice
We take a one dimensional lattice with lattice constant a. The ions are located
around the equilibrium positions na with integer n. For simplicity we rst only
consider nearest neighbor forces

2
(r)
r
2

r=a
= K (5.8)
the derivative at all other positions therefore vanishes. This yields
D(na n

a) =
n,n
2K
n,n

1
K (5.9)
and the equation of motion
M u
m
= K (2u
m
u
m1
u
m+1
) (5.10)
where we write u
m
= u(ma). In order to solve this dierential equation we
make the ansatz
u
n
(t) e
i(knat)
(5.11)
with paramters k and that need to be specied. We will see below that the
real and imaginary part of (5.11) yield 2N linearly independent solutions of the
second order dierential equation (5.10) for N ions, which is therefore complete
and provides an a posteriori justication of the ansatz.
Since we are interested in bulk properties of large crystals (N 1) we are
justied to use Born-von Karman boundary conditions like when deriving the
electronic band structure, u
N+1
(t) u
1
(t), therefore we actually have a ring of
N ions. This boundary condition requires
e
ikna
= 1 (5.12)
and therefore
k =
2
a
n
N
(5.13)
with positive or negative integer n. Clearly, the solutions (5.11) remain invariant
under the shift k k +2/a and we can again restrict the wave vector k to the
rst Brillouin zone


a
k <

a
(5.14)
Introduction to Solid State Theory (Kehrein) 59 University of Goettingen
5 Lattice Vibrations and Phonons
This yields N allowed values of k in (5.13).
Next we insert the ansatz into the equations of motion
M
2
e
i(knat)
= K [2 e
ika
e
ika
] e
i(knat)
(5.15)

2
=
2K
M
[1 cos(ka)] (5.16)
This yields the dispersion relation
1
(k) = 2
_
K
M
[ sin(ka/2)[ (5.17)
and we have the required 2N solutions
u
n
(t)
_
cos(kna (k)t)
sin(kna (k)t)
(5.18)
These solutions describe propagating waves along the ring with phase velocity
(k)
k
and group velocity v
g
=
(k)
k
. One veries easily that the group velocity
vanishes at the boundary of the rst Brillouin zone since the dispersion relation
becomes at. For large wavelength (meaning: much larger that the lattice
constant a) one has [ka[ 1 and can expand in (5.17) leading to a linear
dispersion relation for small k:
(k) = a
_
K
M
[k[ (5.19)
Hence the group velocity becomes independent of frequency like for sound, v
g
=
a
_
K
M
, which is why this is called acoustical branch. Notice that one can show
these observations (vanishing of the group velocity at the zone boundary and
linear dispersion relation for large wavelength) generally also beyond nearest
neighbor interactions.
2
b) One dimensional Bravais lattice with basis
We consider specically the situation of a diatomic chain with identical ions.
Generalization will be discussed later, they show nothing new. The equilibrium
positions of the ions are na (deviation denoted by u
1,n
) and na + d (deviation
denoted by u
2,n
). Without loss of generality we can assume d < a/2. For nearest
neighbor forces one has two nonvanishing second derivatives of the potential: at
distance d, which we denote by K; and at distance a d, which we denote by
G. Physically we expect K > G. The equations of motion take the following
1
One can immediately verify that the negative root does not produce linearly independent
solutions.
2
As long as the potential falls of suciently fast.
Introduction to Solid State Theory (Kehrein) 60 University of Goettingen
5 Lattice Vibrations and Phonons
form
M u
1,n
= K [u
1,n
u
2,n
] G[u
1,n
u
2,n1
] (5.20)
M u
2,n
= K [u
2,n
u
1,n
] G[u
2,n
u
1,n+1
] (5.21)
We make the ansatz
u
1,n
(t) =
1
e
i(knat)
u
2,n
(t) =
2
e
i(knat)
(5.22)
which is again justied a posteriori since it leads to 4N linearly independent
solutions. k has to fulll the Born-von Karman boundary conditions like before.
Inserting the ansatz into the dierential equations yields
[M
2
(K +G)]
1
+ [K +Ge
ika
]
2
= 0
[K +Ge
ika
]
1
+ [M
2
(K +G)]
2
= 0 (5.23)
This is a linear system of equations for
1
,
2
that only permits nonzero solutions
when the determinant of the matrix vanishes:
[M
2
(K +G)]
2
+[K +Ge
ika
[
2
= 0 (5.24)
This gives the dispersion relation
=
1

M
_
K +G
_
K
2
+G
2
+ 2KGcos(ka) (5.25)
and the structure of the vectors

1
=
K +Ge
ika
[K +Ge
ika
[
(5.26)
In addition to the acoustical branch which is again linear for small k, one now
also nds the so called optical branch with (k = 0) ,= 0 (upper sign in the above
equation). This branch is called optical because in some crystals photons (which
have eectively k = 0) can couple to these excitations because (k = 0) is in
the optical range. One can also see that for large wavelength in the acoustical
branch the structure of the normal modes is
2
=
1
, hence the atoms in the
unit cell oscillate in phase. On the other hand, for the optical branch for large
wavelength one nds
2
=
1
, hence the atoms in the unit cell oscillate out
of phase. Both branches become at with vanishing group velocity at the zone
boundary.
c) Monoatomic three dimensional Bravais lattice
The harmonic approximation for the interaction of the lattice ions now reads

R) D

R

R

) u

) (5.27)
Introduction to Solid State Theory (Kehrein) 61 University of Goettingen
5 Lattice Vibrations and Phonons
with
D

R

R

) =

2
U
u

R) u

u=0
(5.28)
One can easily verify the following properties
1. D

R

R

) = D


R) because the derivatives can be exchanged.
2. D

) = D

R) because of inversion symmetry of the Bravais


lattice.
3.

R
D

R) = 0 because the interaction potential U must be invariant


under a translation u(

R) u(

R) +

d.
Thinking of D(

R

R

) as a 3x3 matrix we nd the following 3N equations of


motion
M

2
u(

R, t)
t
2
=

D(

R

R

) u(

, t) (5.29)
We make the following ansatz for the normal modes
u(

R, r) = e
i(

Rt)
(5.30)
where is the polarization. Similar to the discussion of the Born-von Karman
boundary conditions for electronic wave functions in a three dimensional lattice
in Sect. 4.1 we require
u(

R +N
i
a
i
, t) = u(

R, t) (5.31)
This leads to the same N = N
1
N
2
N
3
allowed values for

k like in (4.26)

k =
3

i=1
m
i
N
i

b
i
, m
i
Z (5.32)
that can be restricted to the rst Brillouin zone. Here

b
i
are the primitive
reciprocal lattice vectors. If we insert the ansatz into the equations of motion
we nd
M
2
= D(

k) (5.33)
with the so called dynamical matrix D(

k)
D(

k)
def
=

R
D(

R) e
i

R
(5.34)
Introduction to Solid State Theory (Kehrein) 62 University of Goettingen
5 Lattice Vibrations and Phonons
=
1
2

R
D(

R) (e
i

R
+e
i

R
)
=
1
2

R
D(

R) (e
i

R
+e
i

R
2)
= 2

R
D(

R) sin
2
(

k

R/2) (5.35)
where we have used the properties 1-3 from above. Clearly, D(

k) is a real
symmetric matrix with values symmetric in

k. Therefore for all

k we can nd
three orthogonal eigenvectors
D(

k)
s
(

k) =
s
(

k)
s
(

k) (5.36)

s
(

k)
s
(

k) =
ss
(5.37)
These describe three normal modes with polarization vector
s
(

k) and fre-
quency
3

s
(

k) =

s
(

k)
M
(5.38)
This is again a complete set of solutions, which a posteriori justies our ansatz.
The dispersion relation turns out to describe three acoustical branches. One
can verify this by expanding for small [

k

R[
D(

k) =
k
2
2

R
(

k

R)
2
D(

R) (5.39)
which leads to
s
(

k) = c
s
(

k) k for small k with the in general direction and


polarization dependent velocities c
s
(

k) which are just the square roots of the


eigenvalues of the matrix

1
2M

R
(

k

R)
2
D(

R) (5.40)
Similar to the electronic band structure one can nd degenerate
s
(

k)
along suitable symmetry lines in the rst Brillouin zone.
In the cubic crystal system one can always choose one polarization paral-
lel to

k (longitudinal branch) and two normal modes perpendicular to

k
(transverse branches).
3
Notice that one has s(

k) > 0 for stable equilibrium positions of the nuclei.


Introduction to Solid State Theory (Kehrein) 63 University of Goettingen
5 Lattice Vibrations and Phonons
d) Three dimensional Bravais lattice with basis
The three dimensional Bravais lattice with a nontrivial basis is a straightforward
generalization of the above case. For p ions in the basis one nds 3p normal
modes for every

k in the rst Brillouin zone. Three of these branches are acous-
tical, the remaining 3(p 1) are optical branches. As a side note it should be
mentioned that the polarization of these modes are now not necessarily orthog-
onal.
Specic heat of the classical harmonic crystal
In the canonical ensemble the internal energy density of a classical system is
given by
u =
1
V
_
d e
H
H
_
d e
H
(5.41)
where we integrate over the classical phase space with measure
d =

R,
du

R) dP

R) (5.42)
As usual one can rewrite (5.41) as the derivative of the logarithm of the partition
function
u =
1
V

ln
_
d e
H
(5.43)
In a harmonic crystal one can now make the substitutions
u

R)
def
=
1/2
u

R) ,

P

R)
def
=
1/2
P

R) (5.44)
leading to d

=
3N
d. With this one can rewrite the partition function
_
d e
H
=
_
d exp
_
_

_
_

P(

R)
2
2M
+U
(harm)
_
_
_
_
(5.45)
= e
E
(0)
n

3N
(5.46)

_
d

exp
_
_

P(

R)
2
2M

1
2

R) D

R

R

) u

)
_
_
One makes the important observation that the integral over d

produces a
temperature independent constant s. Therefore the energy density
u =
1
V

ln
_
e
E
(0)
n

3N
s
_
(5.47)
=
E
(0)
n
N
+ 3
N
V

1
(5.48)
Introduction to Solid State Theory (Kehrein) 64 University of Goettingen
5 Lattice Vibrations and Phonons
is completely independent of s. This yields a temperature and material proper-
ties independent specic heat
C
V
= 3N k
B
(5.49)
This is called the Dulong-Petit law. It correctly describes the high temperature
limit of real materials,
4
but is experimentally wrong for lower temperatures.
This will be corrected by the quantum mechanical solution in the next section.
5.2 Quantum theory of the harmonic crystal
We discuss the monoatomic three dimensional Bravais lattice. The relevant
quantum mechanical Hamiltonian is
H =

P(

R)
2
2M
+
1
2

R,

,
u

R) D

R

R

) u

) (5.50)
where P

R) and u

R) are 3N momentum and position operators. Similar


to the usual treatment of the harmonic oscillator we introduce creation and
annihilation operators
a

k,s
def
=
1

R
e
i

s
(

k)
_
_

M
s
(

k)
2
u(

R) +i

1
2M
s
(

k)

P(

R)
_
_
a

k,s
=
1

R
e
i

s
(

k)
_
_

M
s
(

k)
2
u(

R) i

1
2M
s
(

k)

P(

R)
_
_
(5.51)
where we have inserted dispersion relation and polarization vectors of the clas-
sical solution. These operators will later be identied with creation and annihi-
lation operators for phonons with quasimomentum

k and polarization
s
(

k).
From the canonical commutation relation of position and momentum operators
one can deduce the following bosonic commuation relations (as generalization of
the corresponding relation for the usual quantum mechanial harmonic oscillator)
[a

k,s
, a

,s
] =

k,


ss
(5.52)
[a

k,s
, a

,s
] = [a

k,s
, a

,s
] = 0 (5.53)
4
Except for additional corrections due to anharmonicities.
Introduction to Solid State Theory (Kehrein) 65 University of Goettingen
5 Lattice Vibrations and Phonons
One can also invert the above transformation and show
u(

R) =
1

k,s


2M
s
(

k)
(a

k,s
+a

k,s
)
s
(

k) e
i

R
(5.54)

P(

R) =
i

k,s

M
s
(

k)
2
(a

k,s
a

k,s
)
s
(

k) e
i

R
(5.55)
The sums over

k run over the rst Brillouin zone. Both these sets of equations
will be proven in a tutorial.
Next we substitute (5.54) and (5.55) into the Hamiltonian (5.50). First we look
at the kinetic term
1
2M

P(

R)
2
=
1
2M
1
N

k,

s,s

M
2
_

s
(

k)
s
(

)
(a

k,s
a

k,s
) (a

,s

,s

s
(

k)
s
(

) e
i(

k+

R
(5.56)
Using (3.36) we can perform the sum over

R

R
e
i(

k+

R
=
_
N if

k +

= reciprocal lattice vector


0 otherwise
(5.57)
Because

k and

k

are in the rst Brillouin zone this enforces



k

k. With this
one can also employ the orthogonality of the polarization vectors in the above
expression

s
(

k)
s
(

k) =
s
(

k)
s
(

k) =
ss
(5.58)
since
s
(

k) =
s
(

k) from inversion symmetry. The kinetic term then takes


the simple structure
1
2M

P(

R)
2
=

k,s

s
(

k) (a

k,s
a

k,s
) (a

k,s
a

k,s
) (5.59)
Along the same lines one simplies the harmonic potential term
1
2

R,

u
t
(

R) D(

R

R

) u(

) (5.60)
=
1
2
1
N

R,

k,

s,s

2M
1
_

s
(

k)
s
(

)
Introduction to Solid State Theory (Kehrein) 66 University of Goettingen
5 Lattice Vibrations and Phonons
(a

k,s
+a

k,s
) (a

,s

+a

,s

t
s
(

k) D(

R

R

)
s
(

) e
i

R+i

Using similar reasoning as for the kinetic term and employing translation in-
variance one shows

R,

D(

R

R

) e
i

R+i

= N

k,

D(

k) (5.61)
With this one uses the fact that the polarization vectors are eigenvectors of the
dynamical matrix

t
s
(

k) D(

k)
s
(

k) =
t
s
(

k)
s
(

k)
s
(

k)
=
s
(

k)
s
(

k)
s
(

k)
= M
2
s
(

k)
ss
(5.62)
where we have again used inversion symmetry. Hence the potential term has
the following structure
1
2

R,

u
t
(

R) D(

) u(

) =

4

k,s

s
(

k) (a

k,s
+a

k,s
) (a

k,s
+a

k,s
)(5.63)
The total Hamiltonian ist therefore
H =

4

k,s

s
(

k)
_
(a

k,s
a

k,s
) (a

k,s
a

k,s
)
+(a

k,s
+a

k,s
) (a

k,s
+a

k,s
)
_
(5.64)
=

k,s

s
(

k)
_
a

k,s
a

k,s
+
1
2
_
(5.65)
This is just a sum of 3N independent (commuting) harmonic oscillators with
frequencies given by the (classical) dispersion relation. The eigenenergies are
then
E =

k,s
_
n

k,s
+
1
2
_

s
(

k) (5.66)
with n

k,s
= 0, 1, 2, . . . One usually gives a particle interpretation to this quan-
tum number n

k,s
: It is the number of phonons of type s with wavevector

k. In
this sense phonons are emergent particles, that is they only exist in a crystal.
Emergent particles that describe excitations of a nontrivial quantum system
Introduction to Solid State Theory (Kehrein) 67 University of Goettingen
5 Lattice Vibrations and Phonons
(nontrivial meaning: not the vacuum) play an important role in modern con-
densed matter physics.
5
Specic heat:
Now we proceed to the calculation of the specic heat using the quantum theory.
In the partition function the classical measure d from (5.42) is replaced by a
trace over the Hilbert space H
Z = Tr e
H
(5.67)
Since H =

k,s
H

k,s
with commuting
H

k,s
=
s
(

k)
_
a

k,s
a

k,s
+
1
2
_
(5.68)
one can rewrite the partition function as
Z = Tr exp
_
_

k,s
H

k,s
_
_
(5.69)
= Tr

k,s
exp
_
H

k,s
_
=

k,s
Tr

k,s
exp
_
H

k,s
_
(5.70)
where the trace goes over the dierent sectors of the Hilbert space, H =

k,s
H

k,s
. Now for a single harmonic oscillator
Z
(harm)
(, ) = Tr exp
_
[(a

a + 1/2)]
_
(5.71)
=

n=0
e
(n+1/2)
= e
/2

n=0
_
e

n
=
e
/2
1 e

(5.72)
where we have used the eigenbasis of the harmonic oscillator in the trace and
the result for the geometric series. The internal energy for a harmonic oscillator
is then
U
(harm)
=
lnZ
(harm)

(5.73)
5
Such emergent particles can have properties completely dierent from elementary parti-
cles, like magnetic monopoles in spin ice, or anyons in fractional quantum Hall states.
Introduction to Solid State Theory (Kehrein) 68 University of Goettingen
5 Lattice Vibrations and Phonons
=

2
+
e

1 e

=
_
1
2
+n
BE
()
_
(5.74)
The rst term is the zero point energy and the second the Bose-Einstein distri-
bution
n
BE
(x) =
1
e
x
1
(5.75)
Going back to the crystal we have in (5.70)
Z =

k,s
Z
(harm)
(
s
(

k), ) (5.76)
lnZ =

k,s
lnZ
(harm)
(
s
(

k), ) (5.77)
Not surprisingly the internal energy of the crystal is just the sum of the internal
energies of the various harmonic oscillators
u =
1
V
lnZ

(5.78)
=
1
V

k,s
U
(harm)
(
s
(

k), )
=
1
V

k,s

s
(

k)
_
1
2
+n
BE
(
s
(

k))
_
(5.79)
The zero point energy just gives another temperatur independent contribution
like E
(0)
n
and plays no role for the specic heat
c
V
=
1
V

k,s

s
(

k)
e
s(

k)/k
B
T
1
(5.80)
=

s
_
1.BZ
d

k
(2)
3

s
(

k)
e
s(

k)/k
B
T
1
(5.81)
where we have used (2.10) and the

k-integration runs over the rst Brillouin
zone.
We analyze this expression in the limits of high and low temperatures. For high
temperatures such that k
B
T
s
(

k) for all

k, s we can expand the exponential


in the denominator to rst order
1
e
s(

k)/k
B
T
1

1
1 +
s
(

k)/k
B
T 1
=
k
B
T

s
(

k)
(5.82)
Introduction to Solid State Theory (Kehrein) 69 University of Goettingen
5 Lattice Vibrations and Phonons
which yields in (5.80
c
V
=
1
V

k,s
k
B
= 3
N
V
k
B
(5.83)
that is the Dulong-Petit result from our previous classical theory. At this point
it also becomes apparent where the classical theory fails: for phonon modes
with suciently high frequencies the quantization condition become important
unless the temperature is very high, k
B
T
s
(

k) for all

k, s.
In the low temperature limit one can also nd a closed result by making the
following three approximations:
1. We neglect all optical branches since for them
s
(

k) k
B
T in the low
temperature limit, therefore the exponential in the denominator of (5.81)
becomes very large.
2. In the low temperature limit only the vicinity of k = 0 in the acousti-
cal branches contributes due to the exponential. We can therefore safely
linearize

s
(

k) = c
s
(

k) k (5.84)
3. The integration in (5.81) can be taken to run over all values of

k.
Then we rewrite the

k-integral in (5.81) using spherical coordinates and the
substitution x = c
s
(

k) k
c
V
=

T

s
_
d

k
(2)
3
c
s
(

k) k
e
cs(

k) k/k
B
T
1
=

T

s
_
d
4
_

0
dx
2
2
x
2
[c
s
(

k)]
3
x/
e
x
1
=

T
_
(k
B
T)
4
(c)
3
3
2
2
_
dx
x
3
e
x
1
_
(5.85)
where we have introduced the average
c
3
def
=
1
3

s
_
d
4
1
[c
s
(

k)]
3
(5.86)
The x-integration in the above expression yields
4
/15, which produces the low
temperature result
c
V
=
2
2
5
k
B
_
k
B
T
c
_
3
(5.87)
Introduction to Solid State Theory (Kehrein) 70 University of Goettingen
5 Lattice Vibrations and Phonons
that is a characteristic T
3
-contribution from the phonons. The measured specic
heat of a crystal therefore takes the following form
c
exp
V
= T +
2
2
5
k
B
_
k
B
T
c
_
3
(5.88)
where the rst term is the electronic contribution (2.45) which always dominates
at suciently low temperatures.
For comparison with experiments it is also useful to know the Debye model,
which is basically just an approximation that linearizes the acoustical branches
using = c k and runs the

k-integration up to k
D
, which is determined by
requiring the correct total number of phonon modes

k|<k
D
!
= N (5.89)
Analogous to the calculation for the Fermi wavevector (2.12) this gives
N
V
=
k
3
D
6
2
(5.90)
Introducing the Debye frequency
D
and Debye temperature
D

D
def
= k
D
c (5.91)
k
B

D
=
D
(5.92)
one can evaluate the specic heat in the following manner
c
V
=

T
3c
2
2
_
k
D
0
dk
k
3
e
ck
1
= 9
N
V
k
B
_
T

D
_
3
_

D
/T
0
dx
x
4
e
x
(e
x
1)
2
(5.93)
Therefore in the Debye model the specic heat is a universal function of T/
D
.
Experimentally, one often uses the Debye temperature as a t parameter in the
low temperature expansion of (5.93)
c
V
= 234
N
V
k
B
_
T

D
_
3
(5.94)
The Deybe temperature has the interpretation that for temperatures T larger
than
D
all phonons are excited.
Experimental measurement of phonon dispersion relations
Introduction to Solid State Theory (Kehrein) 71 University of Goettingen
5 Lattice Vibrations and Phonons
The most important method for measuring the phonon dispersion relation is
neutron scattering. We have already discussed neutron scattering o a rigid
crystal in chapter 3.4 and found the following key result: If p is the momentum
of the neutron before and p

after scattering, then the scattering condition reads


p

p = reciprocal lattice vector (5.95)


We now generalize this to scattering o a crystal with lattice degrees of freedom.
Clearly this crystal can be in a dierent state after the scattering process. The
basic assumption underlying the following results is that this crystal can be
described as harmonic plus small anharmonic terms (so we cannot apply the
discussion below to e.g. liquids). The key observation is the periodic translation
invariance of the nuclei-neutron interaction
V
nn
=

R
w(r

R u(

R))
=

R
w(r (

R

R
0
) u(

R

R
0
))
=

R
w(r +

R
0


R u(

R

R
0
)) (5.96)
where r is the coordinate of the neutron and

R
0
an arbitrary Bravais lattice
vector. One therefore has a symmetry under the simultaneous transformation
r r +

R
0
and u(

R) u(

R

R
0
). This symmetry leads to a conservation law
for the crystal momentum operator (for details e.g. consult Ashcroft & Mermin,
appendix M) and one can show the following statement: Let the crystal be in
an eigenstate n

ks
before the scattering process. Then the crystal will be in a
superposition of eigenstates after the scattering process which fulll
p

p =

k,s

k n

ks
+ reciprocal lattice vector (5.97)
where n

ks
= n

ks
n

ks
. Relation (5.97) explains the interpretation of the
good quantum number

k as (crystal) momentum: It plays the same role in
conservation laws as usual momentum, but is only conserved up to reciprocal
lattice vectors (such momenta are absorbed by the lattice as a whole).
6
In addition to the crystal momentum conservation law one also has energy
conservation
E

E =

k,s

s
(

k) n

ks
(5.98)
6
Remember that phonons are not elementary particles but emergent particles.
Introduction to Solid State Theory (Kehrein) 72 University of Goettingen
5 Lattice Vibrations and Phonons
For the most important inelastic process, namely the excitation of a single
phonon, one can put these equations into the following form
E

= E
s
(

k) (5.99)
p

= p

k +

K (5.100)
E

=
p
2
2m
n
=
p
2
2m
n

s
_
p



K
_
=
p
2
2m
n

s
_
p

_
(5.101)
since the phonon dispersion relation is periodic in reciprocal space. In the
experiment p is given. The above equation therefore describes a surface in p

-
space where the scattering condition (for one phonon excitation) is fullled.
Since the detector species a given direction p

, this condition will in general


only be fullled for certain discrete values. The corresponding energy losses
of the neutron then correspond to the dispersion relation for the respective
momentum transfer p

p.
Introduction to Solid State Theory (Kehrein) 73 University of Goettingen
Chapter 6
Electron Dynamics
The starting point of our analysis are Bloch electrons, that is electrons in a
periodic crystal. We want to address questions like what is their behavior in an
external electrical eld (electrical conductivity), in an external magnetic eld,
etc.
6.1 Semiclassical approximation
Both free electrons and Bloch electrons are solutions of the resp. time-independent
Schrodinger equation. We contrast their properties in table 6.1.
The semiclassical model for electrons deals with Bloch electrons with well de-
ned momenta in the Brillouin zone, that is k a
1
. This automatially
implies that the corresponding wave packets are spread out over many primi-
tive unit cells in real space, r a. The second assumption is that external
elds (like electrical or magnetic elds) only vary very slowly on the scale set by
the spatial size of the wave packets. Under these conditions one can show the
validity of the semiclassical approximation
1
which treats average position r and
average crystal momentum p
cr
as classical variables in the classical Hamiltons
function
H(r, p
cr
) =
n
_
p
cr

_
+V
pot
(r) (6.1)
This yields the equations of motion

r =
1

k
H = v
(g)
n,

k
(6.2)
1
The proof of this assertion is complex and nontrivial. For example, only recently one has
realized that another condition needs to be met, namely that one has a topologically trivial
phase in the sense of a vanishing Cherns number.
74
6 Electron Dynamics
free electrons Bloch electrons
Quantum numbers

k = p/

k, n
(without spin) momentum crystal momentum, band index
Range all

k compatible with

k in the rst Brillouin zone compatible
Born-von Karman conditions with Born-von Karman conditions
Energy

k
=
2
k
2
/2m
n,

k
from the solution of the
periodic scattering problem.
General properties: periodic

n,

k+

K
=
n,

k
and gap at
the boundary of the Brillouin
zone (also higher order)
Group velocity v
(g)

k
=

k/m v
(g)
n,

k
=
1

k

n,

k
Wave function

k
(r) =
1

V
e
i

kr

n,

k
(r) = e
i

kr
u
n,

k
(r)
with periodicity u
n,

k
(r +

R) = u
n,

k
(r)

p
cr
=

r
V
pot
(r) (6.3)
From the point of view of the external classical elds one is therefore dealing
with a classical pointlike particle. The quantum mechanical description comes
in at the level of the scattering o the periodic crystal potential, which provides
this particle with a nontrivial dispersion relation
n
(

k).
In an electromagnetic eld the semiclassical approximation then leads to the
following behavior:
1. Equations of motion:

r =
1

k

n
(

k) (6.4)

k =
e

E(r, t) +
1
c

r

B(r, t)
_
(6.5)
2. There are no transitions between dierent bands, the band index n is
constant.
3. In equilibrium the occupation is determined by the Fermi-Dirac distribu-
tion
f(
n,

k
) =
1
e
(
n,

k
)
+ 1
(6.6)
Introduction to Solid State Theory (Kehrein) 75 University of Goettingen
6 Electron Dynamics
One obvious limitation of the semiclassical approximation becomes apparent in
the limit crystal potential U(r) 0 for xed homogeneous electrical eld. For
free electrons the kinetic energy will grow unbounded, while in the semiclassi-
cal model the kinetic energy remains bounded within one energy band. The
resolution of this contradiction is simply that for given external elds the band
gaps must not be too small, otherwise there will be transitions to other bands
violating rule 2 above. One can show the following criterion for the validity of
having no band transitions
e E a

c
_

(
n,

k
)
2

F
(6.7)
where
n,

k
is the band gap. In metals this condition is usually fullled (i.e.
there are no band transitions) since the largest electrical elds are of order
10
2
Vcm
1
. This gives
n,

k
10
5
eV, which is generically true. In insulators
one can apply much larger elds and thereby induce a dielectric breakdown,
which is due to band transitions. On the other hand, notice that in metals
in magnetic elds of order 1T band transitions become possible already for

n,

k
= O(10
2
eV) (leading to so called magnetic tunneling).
As expected the semiclassical approximation conserves the total energy of an
electron
E
tot
=
n
(

k(t)) e (r(t)) (6.8)


where (r) is the scalar potential corresponding to the electrical eld. Now
dE
tot
(t)
dt
=

n
k
i

k
i
e



r
i
= v
(g)
n
(

k(t)) (

k e

) (6.9)
= 0 (6.10)
since the term in parantheses is nothing but the equation of motion (6.5)

k = e

(6.11)
One important consequence of the semiclassical model is that lled bands are
inert, i.e. they do not contribute to transport. One can see this from the explicit
expression for the electrical current in the semiclassical approximation

j
e
= (e) v
(g)
n,

k
) (6.12)
=
e

2
_
occupied
d
3
k
(2)
3

k

n,

k
(6.13)
Introduction to Solid State Theory (Kehrein) 76 University of Goettingen
6 Electron Dynamics
where the integration goes over all occupied states. Now in a lled band this
region of integration is time-independent and always identical to the rst Bril-
louin zone. In order to show that the above integral vanishes we rst state a
small useful theorem: Let f(r) be a lattice periodic function. Then for all r

_
unit cell
d
3
r

r
f(r +r

) = 0 (6.14)
This is simply a consequence of
I(r

)
def
=
_
unit cell
d
3
r f(r +r

) (6.15)
being independent of r

due to the lattice periodicity of f. Hence

r
I(r

) = 0 (6.16)
=
_
unit cell
d
3
r

r
f(r +r

)
=
_
unit cell
d
3
r

r
f(r +r

) (6.17)
Especially
_
unit cell
d
3
r

r
f(r) = 0 (6.18)
Completely analogous one makes use of the periodicity of the dispersion relation
in reciprocal space,
n
(

k +

K) =
n
(

k), and nds


_
1. BZ
d
3
k
(2)
3

k

n,

k
= 0 (6.19)


j
e
= 0 (6.20)
for lled bands: Electrons in lled bands do not contribute to transport. Elec-
trical conductivity comes from only partially lled bands, hence the terminology
conduction electrons and conduction band. Also notice that the above observa-
tion provides an a posteriori justication for our previous denition of insulators
in Chapter 4.2.
Electron vs. hole conductivity
The above theorem states completely generally (for all llings)
0 = e
_
1. BZ
d
3
k
4
3
v
(g)
n,

k
= e
_
occupied
d
3
k
4
3
v
(g)
n,

k
+ (e)
_
unoccupied
d
3
k
4
3
v
(g)
n,

k
(6.21)
Introduction to Solid State Theory (Kehrein) 77 University of Goettingen
6 Electron Dynamics
therefore

j
e
= e
_
occupied
d
3
k
4
3
v
(g)
n,

k
(6.22)
= +e
_
unoccupied
d
3
k
4
3
v
(g)
n,

k
(6.23)
From this simple equality one can deduce that the electrical current can either
be interpreted as a current of electrons (occupied states) with charge -e, or
completely equivalently as a current of holes (unoccupied states) with charge +e.
The interpretation as a hole current is particularly useful for nearly lled bands,
like the valence band in a doped semiconductor: there the current contribution
of the valence band is set by the concentration of the holes, and not by the total
number of electrons in the valence band.
Bloch oscillations
Consider a homogeneous static electrical eld and no magnetic eld. The solu-
tion of the semiclassical equation of motion

k = e

E (6.24)


k(t) =

k(0)
e

E

t (6.25)
describes a crystal momentum that grows linearly in time. Of course, once it
reaches the boundary of the rst Brillouin zone its value needs to be identied
with the value on the other side of the Brillouin zone. Hence the solution (6.25)
runs through the rst Brillouin zone periodically. Now, for the group velocity
this implies oscillatory behavior as a function of time. Remember that the group
velocity vanishes at the boundary of the Brillouin zone, which just corresponds
to the point where the crystal momentum jumps
2
. An electron in the rst
band initially at rest will rst accelerate as expected in the electrical eld, then
its velocity becomes smaller again until it reaches the boundary of the rst
Brillouin zone. After which the acceleration is opposite to what is expected in
this electrical eld, before changing sign and returning to the initial condition.
This at rst sight counterintuitive behavior again emphasizes the dierence be-
tween crystal momentum and physical momentum. The elementary particle
electron experiences a force due to both external elds and periodic crystal
potential, which aects the physical momentum of the electron. However, we
have incorporated the periodic lattice potential into properties of a new emer-
gent particle, namely the Bloch electron. Bloch electrons only see the external
elds, the periodic lattice potential is incorporated into their dispersion relation.
Hence the crystal momentum has no immediate connection to the instantaneous
physical momentum.
2
More accurately: is periodically identied
Introduction to Solid State Theory (Kehrein) 78 University of Goettingen
6 Electron Dynamics
In the situation described above the electrical eld drives the electron into states
closer and closer to the Bragg plane at the boundary of the Brillouin zone,
therefore the eect of the periodic lattice potential becomes more and more
important until we have coherent reection (like for e.g. neutron scattering in
chapter 3.4.).
3
A remarkable consequence of the above result is the prediction that a dc-eld
should produce an ac-current in a partially lled band. Such so called Bloch
oscillations were actually observed, but only recently in ultraclean semiconduc-
tors or cold atomic gases. In most materials there are scattering processes due
to imperfections of the periodic lattice, or due to phonons, which permit Bloch
electrons to only travel a tiny fraction in the Brillouin zone according to (6.25)
before being scattered. Hence without such imperfections there would be no
dc-conductors. And, as we will see later on, without such imperfections conduc-
tivities would also diverge.
Eective mass:
In many semiconductors electrical transport is nearly exclusively due to elec-
trons and holes in the vicinity of symmetry points

k

in the Brillouin zone (e.g.


the -point). At such symmetry points the dispersion relation will be either a
maximum or a minimum and one can use a quadratic approximation with good
accuracy:

n
(

k) =
n
(

)

2
2
M
1
ij
(

) k
i
k
j
(6.26)
The upper signs corresponds to a maximum (hole conduction), the lower sign to
a minimum (electron conduction). The matrix M is called eective mass tensor
4
and is determined by the inverse curvature of the band at

k

. The equations of
motion (6.4) and (6.5) become very simple in this quadratic approximation

r = M
1

k (6.27)


r = M
1

k
= M
1
(e)(

E +
1
c

r

B) (6.28)
M

r = e
_

E +
1
c

r

B
_
(6.29)
This is just the expected equation of motion for holes (charge +e) or electrons
(charge -e) with mass M due to the Lorentz force.
3
The physical momentum transfer proportional to a reciprocal lattice vector is picked up
by the center of mass motion of the crystal.
4
In isotropic materials it becomes a direction-independent eective mass.
Introduction to Solid State Theory (Kehrein) 79 University of Goettingen
6 Electron Dynamics
6.2 Dynamics in a magnetic eld
We rst consider a homogeneous time-independent magnetic eld

B and a van-
ishing electrical eld

E. The semiclassical equations of motion (6.4) and (6.5)
then take the following form

r = v
(g)
(

k) =
1

k)

k
(6.30)

k =
e
c
v
(g)
(

k)

B (6.31)
We immediately read o two conserved quantities:
1. The energy of the electrons is conserved
d
dt
(

k(t)) =
e
c
v
(g)
(

k) (v
(g)
(

k)

B) = 0 (6.32)
2. The component of the crystal momentum parallel to the magnetic eld
k

k

B is conserved
d
dt
k

=
e
c

B (v
(g)
(

k)

B) = 0 (6.33)
These conserved quantities determine the geometry of the trajectories, which
we will rst consider in k-space: The electrons move on curves generated by
the cut of the plane perpendicular to the magnetic eld through k

with sur-
faces of constant energy
n

k
. The orientation of the trajectory follows from the
observation that v
(g)
(

k) points from lower to higher energies. Therefore closed


orbits around energies, which are smaller than the energy on the orbit (electron
orbits) are circled in the opposite direction from orbits around energies, which
are higher (hole orbits). Also notice that there can be open orbits if the cut
extends from one end of the Brillouin zone to the other (the respective

k-vectors
being identical up to reciprocal lattice vectors.)
From this we can construct the trajectories in real space, which we will later
need for discussing transport properties. First we look at the projection of the
real space trajectory on the plane perpendicular to the magnetic eld
r

def
= r

B(

B r) (6.34)
One derives

k =
e
c

B (v
(g)


B)
=
eB
c
_
v
(g)


B(v
(g)


B)
_
=
eB
c

(6.35)
Introduction to Solid State Theory (Kehrein) 80 University of Goettingen
6 Electron Dynamics
and therefore
r

(t) r

(0) =
c
eB

B (

k(t)

k(0))
=
c
eB

(t) (6.36)
The real space trajectory in the plane perpendicular to

B is therefore simply
the k-space trajectory rotated by 90

around the

B-axis. Clearly, closed orbits
in k-space map to closed orbits in real space. For open orbits in k-space the
situation is only slightly more complicated: (6.46) is valid from one boundary
of the Brillouin zone to another, and then has to be integrated again with a new
starting point r

(0). Therefore the open orbit in k-space becomes an unbounded


(=open) orbit in real space.
The real space motion in the direction parallel to the magnetic eld follows
simply by integration
r

(t) = r

(0) +
_
t
0
dt

v
(g)

(t), k

) (6.37)
Notice that this motion is in general not uniform for a generic dispersion relation
where the group velocity depends on

k

(t). Also notice that we obviously


recover the expected circular motion of free electrons since surfaces of constant
energy for free electrons are simply spheres (hence the cut with a plane gives a
circle).
Cyclotron frequency
We want to derive the period for a closed orbit. Let us rst calculate the time
it takes to transverse the orbit between

k
1
and

k
2
t
2
t
1
=
_
t2
t1
dt =
_
k2

k1
d

k
[

k[
=
c
eB
_
k2

k1
d

k
[v
(g)

k)[
=

2
c
eB
_
k2

k1
d

k
1

k
_

(6.38)
Let

(

k) be the vector in the plane of the orbit which is perpendicular to the


orbit in the point

k and which points to the neighboring orbit with energy +


in the same plane:
=

k) =
_

k
_

k)
=

k) (6.39)
Introduction to Solid State Theory (Kehrein) 81 University of Goettingen
6 Electron Dynamics
Inserting this into (6.38) yields
t
2
t
1
=

2
c
eB
1

_
k2

k1
d

k (

k) (6.40)
The integral on the right hand side is just the area between the neighboring
orbits. Taking the magnetic eld in z-direction we therefore nd the period for
closed orbits
T(, k
z
) =

2
c
eB
A(, k
z
)

(6.41)
Here A(, k
z
) is the area enclosed by the orbit with energy and crystal momen-
tum k
z
in the direction of the magnetic eld. The derivative on the right hand
side therefore provides information on the rate with which this area changes
as a function of energy, which can yield very useful information on the band
structure.
For comparison with the free electron result (2.48)
T =
2

c
=
2mc
eB
(6.42)
one denes the eective cyclotron mass
m

(, k
z
) =

2
2
A(, k
z
)

(6.43)
Generically this eective mass is not identical to e.g. the denition via the
specic heat. However, close to maxima or minima of the band structure where
we have introduced the eective mass tensor M in (6.26) one can work out a
simple relation between M and m

(, k
z
), see problem set 10. For this reason
the measurement of the cyclotron frequency is a very important experimental
technique to determine eective electron and hole masses in semiconductors.
Crossed

E- and

B-elds
We consider homogeneous and time-independent electrical and magnetic elds.
Let us rst determine the relation between orbits in k-space and real space.
Analogous to (6.35)

k =
e
c

B (v
(g)


B) e

B

E
=
eB
c

e

B

E (6.44)
which yields

=
c
eB

k +v
d
(6.45)
r

(t) r

(0) =
c
eB

B
_

k(t)

k(0)
_
+v
d
t (6.46)
Introduction to Solid State Theory (Kehrein) 82 University of Goettingen
6 Electron Dynamics
with the drift velocity dened as
v
d
def
= c
E
B
(

E

B) (6.47)
The trajectory in real space in the plane perpedicular to the magnetic eld is
therefore the superposition of the trajectory in k-space, rotated by 90

around

B,
and a homogeneous drift v
d
.
For the particularly important case of perpendicular magnetic and electrial elds
one can also nd the trajectory in k-space easily: One denes a shifted band
structure
(

k)
def
= (

k)

k v
d
(6.48)
With this denition one nds

e
c


B =
e
c


B +
e
c
v
d


B
=
e
c
v
(g)


B +e
E
B
(

E

B)

B
=
e
c
v
(g)


B e

E
= e
_

E +
1
c
v
(g)


B
_
=

k (6.49)
Hence the equation of motion

k =
e
c


B (6.50)
is identical to the equation of motion in a pure magnetic eld (6.31), but with
the shifted band structure (

k). So the orbits in k-space are cuts of planes


perpendicular to the magnetic eld with surfaces of constant energy . Again
k

and
n
(

k) are constants of motion.


Translating this back into real space one arrives at the following picture: Let us
take the magnetic eld in z-direction and the electrical eld in the x y-plane.
Then as before one has a non-uniform motion in z-direction, and the rotated
orbits with a superimposed drift v
d
in the xy-plane. If one has closed orbits in
k-space the real space picture resembles a circular motion drifting in the plane.
For open orbits in k-space the real space orbits are also open with an additional
bending in the direction v
d
.
Hall eect
We now discuss the Hall eect, that is transport in crossed electrical and mag-
netic elds. The relaxation time is dened such that dt/ is the probability
Introduction to Solid State Theory (Kehrein) 83 University of Goettingen
6 Electron Dynamics
for scattering of an electron from lattice imperfections, phonons, etc., in the
time interval dt. We will say much more about such relaxation eects in the
next section, for now we only need a very preliminary understanding. Typical
values for the relaxation time in good metals are of order 10
14
s. The sub-
sequent discussion assumes large magnetic elds in the sense
c
1, that
is closed orbits are traversed many times between collisions. Notice that for
typical experimental conditions in metals
[

k v
d
[
e
2
a
eEa

c
= O(10
4
eV ) (6.51)
therefore we can take (

k) = (

k) in very good approximation.


Let us rst discuss the situation that all occupied or unoccupied states that
contribute to transport lead to closed orbits. The electrical current density in
the semiclassical picture is

j = nev
ave
(6.52)
where v
ave
is the average velocity of the current carrying particles between col-
lisions. We consider an experimental setup where magnetic eld and electrical
eld are perpendicular, therefore also magnetic eld and current will be perpen-
dicular. From (6.46) we know for the trajectory in position space perpendicular
to

B
r

() r

(0) =
c
eB

B
_

k()

k(0)
_
+v
d

v
ave
=
r

() r

(0)

=
c
eB

k()

k(0)

+v
d
(6.53)
Since

k(t) remains bounded in k-space for closed orbits (remember: the trajec-
tory follows from a cut with the dispersion relation) the rst term on the right
hand side is eventually negligible for suciently large . Hence in the limit

c

v
ave
= v
d
(6.54)


j =
nec
B
(

E

B) (6.55)
for electrons (occupied states). Likewise for holes (unoccupied states)

j =
n
h
ec
B
(

E

B) (6.56)
The interpretation of the above results is that for closed orbits the Lorentz force
is so ecient that on average the electrons cannot gain energy in the applied
electrical eld, therefore the drift velocity dominates.
Introduction to Solid State Theory (Kehrein) 84 University of Goettingen
6 Electron Dynamics
A typical experimental setup for the Hall eect is a two dimensional bar with a
magnetic eld (z-direction) perpendicular to it. Except for an initial transient
the current must run through the bar from one end to the other (x-direction)
since charge accumulates at the sides of the bar. In the stationary state this
charge accumulation must result in an electrical eld

E = Ee
y
in y-direction
since according to (6.55) only then

j = j e
x
. The Hall resistance is dened as
the voltage dierence in y-direction divided by the total current
R
H
def
=
U
y
I
x
=
E
y
j
x
(6.57)
=
_
_
_

B
enc
electrons
B
en
h
c
holes
(6.58)
Some remarks:
In real materials there is also a (small) electrical eld in x-direction since
< .
Notice that the Hall current has the opposite sign for holes, which cannot
be understood in the Drude model.
For large magnetic elds there are additional quantization eects not cap-
tured in the semiclassical picture so far, most notably the Quantum Hall
eect. We will say more about this below.
The above picture changes when also open orbits contribute to transport.
For open orbits one no longer has a circular motion superimposed with
a drift perpendicular to

E and

B, instead the electrons can gain energy
in the electrical eld. Technically,

k()

k(0) in (6.53) does not remain


bounded but becomes proportional to . Hence this yields an additional
contribution to v
ave
in the direction of the open orbit. Among other eects
one no longer nds R
H
B, for more details consult the literature.
Quantization eects
In section 2.3 we had solved the Schrodinger equation for free electrons in a
magnetic eld and found a spectrum consisting of discrete Landau levels. The
obvious question arises where the Landau levels are in our discussion of Bloch
electrons in a magnetic eld. Now strictly speaking this shows the breakdown
of the semiclassical approximation in a (suciently strong) magnetic eld. The
reason for this is that the circular motion generated by a magnetic eld leads
to interference eects (or equivalently: Bohr-Sommerfeld quantization in phase
space) which needs to be taken into account. The usual way to proceed is to ex-
tend the semiclassical model by imposing a quantization condition on the Bloch
electrons (more precisely: the Bloch electron wave packets). In a pure magnetic
Introduction to Solid State Theory (Kehrein) 85 University of Goettingen
6 Electron Dynamics
eld one can do this by either employing Bohr-Sommerfeld quantization, or the
Bohr correspondence principle in the following form

+1
(k
z
)

(k
z
) =
h
T(

(k
z
), k
z
)
(6.59)
where the left hand side is the energy dierence of two neighboring permitted
energy levels and the right hand side contains the classical period T(

(k
z
), k
z
)
of the orbit from (6.41). This correspondence limit becomes valid (usually, there
are a number of caveats) in the limit of large quantum numbers . From (6.41)
we know
T(, k
z
) =

2
c
eB
A(, k
z
)

(6.60)
(
+1
(k
z
)

(k
z
))
A(, k
z
)

(
+1
(k
z
)

(k
z
))
A(
+1
) A(

+1
(k
z
)

(k
z
)
=
2eB
c
(6.61)
where we have made use of the fact that the area A(, k
z
) enclosed by the orbits
typically varies on the energy scale by the Fermi energy, whereas [
+1
(k
z
)

(k
z
)[ = O(
c
) = O(10
4

F
). Hence
A = A(
+1
, k
z
) A(

, k
z
) =
2eB
c
(6.62)
or
A(

, k
z
) = ( +) A (6.63)
with some constant .
We have already seen that many low temperature properties of metals are es-
sentially determined by the density of states at the Fermi energy. One expects
strong peaks in the density of states whenever the above quantization condition
is fullled for many levels at the same time.
5
This will be the case for so called
extremal orbits where
A(, k
z
)
k
z

z
= 0 A
e
= A(, k

z
) (6.64)
5
Notice that a full quantum mechanical calculation like for the free electron gas shows that
the density of states obtains square root singularities at the points where the quantization
condition is fullled, see (2.65).
Introduction to Solid State Theory (Kehrein) 86 University of Goettingen
6 Electron Dynamics
Thus we expect peaks in the density of states at the Fermi energy for
( +) A = A
e
(
F
, k

z
) (6.65)
= ( +)
2eB
c
(6.66)
Upon varying the magnetic eld this yields peaks for

_
1
B
_
=
1
B
+1

1
B

=
2e
c
1
A
e
(
F
, k

z
)
(6.67)
If there are multiple values of k

z
where (6.64) is fullled one observes multiple
oscillation frequencies
_
1
B
_
. An important example are the Haas-van Alphen
oscillations of the magnetic susceptibility already discussed in section 2.3: Haas-
van Alphen measurements provide information on the extremal orbits of the
Fermi surface, and varying the direction of the magnetic eld can therefore be
used to map the Fermi surface. Also notice that the oscillation frequency (2.78)

_
1
B
_
=
2e
c
1
k
2
F
(6.68)
for free electrons is consistent with the above results since the Fermi sphere has
only one extremal orbit with an area A(k
F
) = k
2
F
.
6
Hofstadter buttery
A full quantum mechanical treatment of the interplay of periodic crystal po-
tential with a magnetic eld is possible in certain situations. Such a treatment
shows interesting features not captured by the above semiclassical quantization
(in the limit of very large magnetic elds). This was rst demonstrated by
D. Hofstadter (Phys. Rev. B 14, 2239 (1976)), who discussed a two-dimensional
tight-binding square lattice (lattice constant a). The solution of the Schrodinger
equation yields a band structure depending on the number of ux quanta per
unit cell
=
a
2
B
2c/e
(6.69)
with the following structure:
For rational =
p
q
one nds q bands.
For irrational the specture is an uncountable set of measure zero (a
Cantor set).
Notice that typically = 1 corresponds to magnetic elds of order 10
5
T, and
a slight broadening for realistic small yields a spectrum that shows the
expected Landau level structure.
6
In this context it is also worthwile remembering the Bohr-van Leeuwen theorem (see
problem set 10) which shows all magnetic eld eects in equilibrium are quantum mechanical
and do not exist in a classical theory.
Introduction to Solid State Theory (Kehrein) 87 University of Goettingen
6 Electron Dynamics
6.3 Electrical conductivity in metals
Bloch electrons (that is wave packets made of Bloch eigenfunctions) propagate
through the crystal without dissipation, as opposed to the Drude model. This
contradicts the experimental observation that every induced current decays as
a function of time if it is not driven by an external electrical eld. The most
important scattering mechanisms leading to a decay of the current, and therefore
contributing to a nonvanishing resistance, are the following:
1. Scattering from impurities and lattice defects, that is deviations from the
perfect lattice periodicity. These eects are nearly temperature indepen-
dent and therefore dominate the resistivity for low temperatures T 0.
This is denoted residual resistivity.
2. Scattering from deviations from perfect lattice periodicity due to lattice
vibrations, that is electron-phonon scattering. This is the most important
temperature dependent contribution to the resistivity and dominates at
higher temperatures.
3. Electron-electron scattering is relatively unimportant as compared to the
above contributions since crystal momentum conservation only allows for
a less ecient decay channel (since the current is related to the group
velocity).
Boltzmann equation
The key theoretical tool in transport theory is the Boltzmann equation (or
kinetic equation). It was rst introduced by Boltzmann to model dilute gases:
The distribution function f(t, x, p) describes the particle density at the phase
space point (x, p) at time t. Its time evolution follows from the Boltzmann
equation (here: without external forces)
f(t, x, p)
t
+
_
p
m

_
f(t, x, p) = I[f](t, x, p) (6.70)
The gradient term on the left hand side just describes a streaming term for par-
ticles entering and leaving the phase space region around (x, p). The functional
I[f] is the so called Stossterm, modeling collisions of gas particles. For example
for local elastic 2-particle collisions one writes
I[f](t, x, p) =
_
d p
2
d p
3
d p
4
W(p, p
2
; p
3
, p
4
) (6.71)
(( p) +( p
2
) ( p
3
) ( p
4
)) ( p + p
2
p
3
p
4
)
[f(t, x
3
, p
3
) f(t, x
4
, p
4
) f(t, x, p) f(t, x
2
, p
2
)]
W( p
1
, p
2
; p
3
, p
4
) = W( p
3
, p
4
; p
1
, p
2
) describes the scattering of particles with
momenta p
3
and p
4
to momenta p
1
and p
2
, and vice versa. The terms in square
brackets are gain and loss terms due to such scattering processes. Some remarks:
Introduction to Solid State Theory (Kehrein) 88 University of Goettingen
6 Electron Dynamics
In spite of microscopic reversibility W( p
1
, p
2
; p
3
, p
4
) = W( p
3
, p
4
; p
1
, p
2
)
Boltzmann showed that the entropy
S[f] = H[f] =
_
dxd p f(t, x, p) ln f(t, x, p) (6.72)
can only increase as a function of time (Boltzmanns H-theorem)
dS[f]
dt
0 (6.73)
There is a whole body of literature devoted to understanding how irre-
versibility enters in the derivation of the Boltzmann equation. The key
observation is that the derivation relies on the assumption that the par-
ticles are uncorrelated before the collision (while they certainly become
correlated after the collision). For dilute gases this assumption seems
plausible since it is unlikely for particles to scatter repeatedly. Under
certain conditions this can even be established with mathematical rigor.
One can show (problem set 10) that the only xed point/equilibrium for
an isotropic system is a Maxwell distribution
dS[f]
dt
= 0 f(x, v) T
3/2
e
v
2
/k
B
T
(6.74)
Likewise one can show that the only xed point (i.e. no entropy produc-
tion) for a non-isotropic system with v) = 0 is a local Maxwell distribution
f(x, v) T(x)
3/2
e
v
2
/k
B
T(x)
(6.75)
Boltzmann equation for electrons
Applying the Boltzmann equation to quantum transport involves some addi-
tional approximations. In particular one is neglecting interference eects, i.e.
one is eectively relying on a semiclassical picture. The validity of this picture
needs to be veried for a specic setup. However, it is key to take into account
the exchange statistics of quantum particles. For example for the distribution
function for electrons n(t, x,

k) in phase space (probability of nding a semi-


classical electron in a volume
3
around (x,

k)) one has the constraint from the


Pauli principle: 0 n(t, x,

k) 1. The collision term describing scattering from


impurities and/or phonons takes the form
I[n] =
_
d
3
k

(2)
3
[W
k

k
n(k

) (1 n(k)) W
kk
n(k) (1 n(k

))] (6.76)
where the gain and loss terms incorporate the condition that the scattering has
to go into empty states. The Boltzmann equation for electrons experiencing the
Lorentz force

F(r,

k) = e
_

E(r) +
1
c
v
(g)
(

k)

B(r)
_
(6.77)
Introduction to Solid State Theory (Kehrein) 89 University of Goettingen
6 Electron Dynamics
then reads
n
t
+v
(g)
(

k)

r
n +

F(r,

k)
1

k
n = I[n](r,

k) (6.78)
Notice that the terms on the left hand side just model the solution of the
semiclassical equations of motion
n(r,

k, t) = n
_
r v
(g)
(

k) dt,

k
1

F(r) dt, t dt
_
(6.79)
The Stossterm can be split up in (elastic) scattering from lattice impurities V
imp
and electron-phonon scattering W
ph
W(

k) = V
imp
(

k) +W
ph
(

k) (6.80)
Assuming an equilibrium distribution of the phonons with a temperature prole
T(r) one can show by analyzing the scattering matrix elements
7
W
ph
(

k) e
(

k)/k
B
T(r)
= W
ph
(

k,

) e
(

)/k
B
T(r)
(6.81)
Elastic scattering from (diluate) random impurities with concentration n
imp
described by a potential U
sc
(r) for an impurity at the origin yields
V
imp
(

k) =
2

n
imp
((

k) (

)) [k[U
sc
[k

)[
2
(6.82)
with the matrix element evaluated between Bloch eigenfunctions
k[U
sc
[k

) =
_
dr

nk
(r) U
sc
(r)
nk
(r) (6.83)
Clearly V
imp
(

k) = V
imp
(

k,

).
Eq. (6.81) yields a unique xed point for the Stossterm (6.76) given by the local
equilibrium distribution
8
n
(0)
(r,

k) =
1
e
((

k)(r))/k
B
T(r)
+ 1
(6.84)
Therefore it makes sense to linearize the functional Stossterm around this dis-
tribution
n(r,

k) = n
(0)
(r,

k) +n(r,

k) (6.85)
7
For details see Madelung, Solid-State Theory, Ch. 4.2.
8
Strictly speaking the chemical potential prole (r) must also be determined by coupling
to suitable baths to make this unique.
Introduction to Solid State Theory (Kehrein) 90 University of Goettingen
6 Electron Dynamics
Inserting this in (6.76) and keeping only terms linear in n yields after some
straightforward algebra
I[n] =
_
d
3
k

(2)
3
_
V
imp
(k

, k) (n(k

) n(k)) (6.86)
+
_
V
ph
(k

, k) n(k

) V
ph
(k, k

) n(k)
_
_
where we have dened
V
ph
(k

, k) = W
ph
(k

, k)
1 n
(0)
(r, k)
1 n
(0)
(r, k

)
(6.87)
Putting everything together with
V (k

, k)
def
= V
imp
(k

, k) +V
ph
(k

, k) (6.88)
yields
I[n](r, k) =
_
d
3
k

(2)
3
(V (k

, k) n(r, k

) V (k, k

) n(r, k)) (6.89)


This structure of the Stossterm motivates the so called relaxation time approx-
imation
I[n](r,

k) =
n(r,

k)
(

k)
=
n(r,

k) n
(0)
(r,

k)
(

k)
(6.90)
with the relaxation time

1
(k)
def
=
_
d
3
k

(2)
3
V (k, k

) (6.91)
Under certain conditions one can explicitly show the equivalence of (6.89) and
(6.90) with a slightly dierent denition of the relaxation time:
Elastic processes (

) = (

k) ( V (k

, k) = V (k, k

)), which is approxi-


mately also fullled for scattering from acoustic phonons.
Spherical energy dispersion (

k) = ([

k[)
The distribution function n(r,

k) is always of local equilibrium form (6.84),


albeit with arbitrary temperature and chemical potential proles and pos-
sibly also a shift

k

k + q with some q. One can show that this con-
dition self-consistently remains fullled in the dynamics generated by the
Boltzmann equation if one starts with a local equilibrium distribution (for
suciently short relaxation times).
Introduction to Solid State Theory (Kehrein) 91 University of Goettingen
6 Electron Dynamics
Details of this derivation can be found in Madelung, Ch. 4.2. At this point it
should be mentioned that the relaxation time approximation (6.90) is employed
quite generally to describe a Stossterm in the Boltzmann equation, even if the
above conditions are not fullled. Essentially the idea is to model exponential
relaxation to some equilibrium, which is strictly enforced in the limit of van-
ishing relaxation time 0. Via this reasoning one also (often using intuitive
arguments) identies the equilibrium/xed point distribution n
(0)
(r,

k).
Armed with this knowledge we can now solve the Boltzmann equation (6.78) in
the relaxation time approximation for small starting from e.g. n(r,

k, t = 0) =
n
(0)
(r,

k):
9
n(r,

k, t) = n
(0)
(r,

k) (6.92)
+
_
t
0
dt

e
(tt

)/(

k)
_

_
v
(g)
(

k)
_
e

E(t

)

(r)
(

k) (r)
T(r)

T(r)
_
Here f() is the Fermi-Dirac distribution
f() =
1
e
()/T
+ 1
(6.93)
One can verify the solution (6.92) by explicit insertion into (6.78), which shows
that corrections are higher order in . Also notice that the magnetic eld does
not appear in (6.92) since v
(g)
[v
(g)


B] = 0.
Introducing
(

k, t)
def
=
_
t
0
dt

e
(tt

)/(

k)
= (

k)
_
1 e
t/(

k)
_
(6.94)
one can rewrite (6.92) for time-independent elds in leading order in as
n(r,

k, t) = n
(0)
_
r (

k, t) v
(g)
(

k),

k
(

k, t)

(e)

E
_
(6.95)
One sees that a stationary state is approached exponentially fast
n
()
(r,

k) = lim
t
n(r,

k, t) = n
(0)
_
r (

k) v
(g)
(

k),

k +
(

k)

e

E
_
(6.96)
9
The asymptotic time-invariant distribution n(r,

k, t = ) is independent of the initial


value as can be veried easily.
Introduction to Solid State Theory (Kehrein) 92 University of Goettingen
6 Electron Dynamics
Notice that the Boltzmann equation is essential to have a time-invariant distri-
bution function
n(t)
t
= 0 (6.97)
if there are external electrical elds or a non-uniform temperature/chemical
potential. We had already seen that when discussing Bloch oscillations, likewise
a non-uniform temperature or chemical potential also requires some scattering
process to reach a stationary state.
dc-conductivity
We calculate the dc-conductivity, that is

E time-independent and uniform, B =
0 and (r) = , T(r) = T. From (6.96) we read of the stationary state
n
()
(

k) =
1
e
((

k+ e

E/))/k
B
T
+ 1
(6.98)
The electrical eld eectively shifts the Fermi surface by e

E/. This amounts
to an electrical current density

j = e
_
d

k
4
3
v
(g)
(

k) n
()
(

k)
= e
_
d

k
4
3
v
(g)
(

k) n
(0)
(

k)
+e
2
_
d

k
4
3
v
(g)
(

k) (

k)
_

k)

E v
(g)
(

k) (6.99)
The rst integral vanishes since it is a product of an antisymmetric with a
symmetric function. From the second integral we read of the conductivity tensor

j =

E (6.100)
namely

ij
= e
2
_
d

k
4
3
(

k) v
(g)
i
(

k) v
(g)
j
(

k)
_

k)
(6.101)
If more than one band contributes to transport we need to additionally sum over
the various bands. Notice that the derivative of the Fermi function in (6.101) is
only nonvanishing for energies within k
B
T of the Fermi energy
F
. Hence the
contribution of lled bands vanishes as already shown before.
From the fact that the derivative of the Fermi function only leads to contribu-
tions in the vicinity of the Fermi surface one can also verify that (6.101) can be
approximated in order (k
B
T/
F
)
2
by its T = 0 value. Therefore the relaxation
Introduction to Solid State Theory (Kehrein) 93 University of Goettingen
6 Electron Dynamics
time can be evaluated at the Fermi surface and taken out of the integral. Also
from the chain rule
v
(g)
j
(

k)
_

k)
=
1

k
j
f((

k)) (6.102)
This allows integration by parts

ij
= e
2
(
F
)
_
d

k
4
3
f((

k))
1

k
j
v
(g)
i
(

k) (6.103)
= e
2
(
F
)
_
occ.
d

k
4
3
M
1
ij
(

k) (6.104)
where the integral runs over all occupied levels. M
ij
is the eective mass tensor
already introduced for semiconductors in (6.26)
M
1
ij
(

k)
def
=
1

2
(

k)
k
j
k
j
(6.105)
Notice that the relaxation time (
F
) will in general still have a strong temper-
ature dependence.
Some additional remarks:
1. The dc-conductivity vanishes in the limit of very short relaxation time as
intuitively expected.
2. In a general crystal structure the conductivity tensor is not diagonal, i.e.
an electrical eld can induce a current which is not parallel to it. However,
for cubic crystals one nds
ij
=
ij
since clearly
xx
=
yy
=
zz
for
symmetry reasons and all o diagonal matrix elements must vanish (if
a eld in x-direction would induce a current in y-direction, that current
would actually vanish since symmetry makes both y-directions equivalent).
3. Since the eective mass (6.105) is the derivative of a periodic function, its
integral over the Brillouin zone vanishes similar to the discussion following
(6.14). Hence we can alternatively express the conductivity as an integral
over the unoccupied states

ij
= e
2
(
F
)
_
unocc.
d

k
4
3
M
1
ij
(

k) (6.106)
thereby again showing the equivalence of particle and hole picture.
4. For free electrons the expression for the conductivity takes the Drude form
M
1
ij
=

ij
m

ij
=
ij
ne
2

m
(6.107)
Introduction to Solid State Theory (Kehrein) 94 University of Goettingen
6 Electron Dynamics
It is actually surprising that the Drude picture gives a reasonable an-
swer.
ac-conductivity
See problem set 11.
6.4 Thermal conductivity
We are interested in the thermal current

j
(q)
transported by electrons, which is
the most important contribution in metals under normal conditions. Because
of Q = T dS one has (consider a small volume)

j
(q)
= T

j
(s)
(6.108)
with the entropy current

j
(s)
. Then because of T dS = dU dN one concludes
T

j
(s)
=

j
()

j
(n)
(6.109)
with the energy current

j
()
=
_
d
3
k
4
3
(

k) v
(g)
(

k) n(

k) (6.110)
and particle current

j
(n)
=
_
d
3
k
4
3
v
(g)
(

k) n(

k) (6.111)


j
(q)
=
_
d
3
k
4
3
_
(

k)
_
v
(g)
(

k) n(

k) (6.112)
We already calculated the stationary distribution function for

B = 0 and time
independent elds (6.92)
n
()
(r,

k, t) = n
(0)
(

k)+(

k)
_

_
v
(g)
(

k)
_
e

c +
(

k)
T
(

r
T)
_
(6.113)
where we have dened

c
def
=

E +

e
(6.114)
One can read of that electrical and thermal current are related to the external
elds via four 3x3 matrices

j = L
11

c +L
12
(

T) (6.115)

j
(q)
= L
21

c +L
22
(

T) (6.116)
Introduction to Solid State Theory (Kehrein) 95 University of Goettingen
6 Electron Dynamics
This is a typical linear response result. We dene
L
(m)
ij
= e
2
_
d
3
k
4
3
_

_
(

k) v
(g)
i
(

k) v
(g)
j
(

k) ((

k) )
m
(6.117)
for m = 0, 1, 2. Then the linear response coecients can be expressed as
L
11
= L
(0)
(6.118)
L
21
=
1
e
L
(1)
= T L
12
(6.119)
L
22
=
1
e
2
T
L
(2)
(6.120)
Assuming a relaxation rate that only depends on energy we can dene

ij
()
def
= e
2
()
_
d
3
k
4
3
( (

k)) v
(g)
i
(

k) v
(g)
j
(

k) (6.121)
L
(m)
=
_
d
_

_
( )
m
() (6.122)
Employing the Sommerfeld expansion (2.45) one derives
L
(m)
=
_
d f()

[( )
m
()]
=
_

[( )
m
()]
+

2
6
(k
B
T)
2

2

[( )
m
()]
+O
_
(T/)
4
_
(6.123)
One can read of
L
(0)
= () (6.124)
L
(1)
=

2
6
(k
B
T)
2
2

() (6.125)
L
(0)
=

2
6
(k
B
T)
2
2() (6.126)
plus terms that are smaller by the ratio (T/
F
)
2
. This gives the key result for
the linear response coecients
L
11
= (
F
) = (6.127)
L
21
= T L
12
=

2
3e
(k
B
T)
2

(
F
) (6.128)
L
22
=

2
3
k
2
B
T
e
2
(6.129)
Introduction to Solid State Theory (Kehrein) 96 University of Goettingen
6 Electron Dynamics
where is just the (electrical) conductivity tensor (6.101). The thermal conduc-
tivity tensor K relates a temperature gradient to the induced thermal current

j
(q)
= K (

T) (6.130)
under the condition that the electrical current vanishes

j = 0. This implies from


(6.115)

c = (L
11
)
1
L
12
(

T) (6.131)
K = L
22
L
21
(L
11
)
1
L
12
(6.132)
Since the conductivity only varies on the energy scale
F
, the second term in
the expression for K is of order (T/
F
)
2
and can be neglected. K = L
22
is just
the Wiedemann-Franz law
K =

2
3
_
k
B
e
_
2
T (6.133)
relating the electrical conductivity tensor to the thermal conductivity tensor.
At this point it is worth emphasizing that we have used the relaxation time
approximation in writing down the stationary solution, especially implying the
condition that the electrons only scatter elastically. This is not strictly true
for scattering o phonons, which can lead to deviations from the Wiedemann-
Franz law (since inelastic scattering aects thermal and electrical currents dif-
ferently). Also notice that the other linear response coecients derived above
contain information about other interesting physical eects like the Seebeck ef-
fect: A temperature gradient induces an electrical eld for vanishing current,

E = Q

T.
Introduction to Solid State Theory (Kehrein) 97 University of Goettingen
Chapter 7
Outlook: Many-Particle
Eects
The most problematic approximation in our discussion of solid state systems so
far is neglecting the Coulomb repulsion between electrons V
ee
in the fundamental
Hamiltonian (1.32). In this chapter we will give a brief outlook on what happens
when we go beyond this independent electron approximation. Essentially, this
serves as a motivation for the Masters course Advanced Solid State Theory.
Neglecting the lattice degrees of freedom, we are now interested in the Hamil-
tonian
H =
N

i=1
_


2
2m

2
i
+U
ion
(r
i
)
_
+
1
2

i=j
e
2
[r
i
r
j
[

= E (7.1)
for the N-particle wave function (r
1
, s
1
; . . . ; r
N
, s
N
). Our previous treatment
of many-body Hamiltonians was based on the fact that these Hamiltonians
consisted of commuting terms acting only on one of the electrons. This allowed
us to solve a single particle problem and construct the many-body eigenfunction
as a Slater determinant. This approach is not applicable for (7.1) anymore
because of the Coulomb repulsion. Still, one can set the goal to nd the best
ground state wave function (meaning: lowest ground state energy) among the
states that can be expressed as a single Slater determinant
(r
1
, s
1
; . . . ; r
N
, s
N
) (7.2)
=
1

N!

1
(r
1
, s
1
)
1
(r
2
, s
2
) . . .
1
(r
N
, s
N
)

2
(r
1
, s
1
)
2
(r
2
, s
2
) . . .
2
(r
N
, s
N
)
.
.
.
.
.
.

N
(r
1
, s
1
)
N
(r
2
, s
2
) . . .
N
(r
N
, s
N
)

98
7 Outlook: Many-Particle Eects
In problem set 11 it was shown that this variational problem leads to a set of
nonlinear equations for the optimal 1-particle basis
i
(r)


2
2m

i
(r) + U
ion
(r)
i
(r) +
_
_

j
_
dr

[
j
(r

)[
2
e
2
[r r

[
_
_

i
(r)

j
_
d
3
r

e
2
[r r

si,sj

j
(r

)
i
(r

)
j
(r) (7.3)
=
i

i
(r)
such that the ground state energy E
gs
=

N
i=1

i
is minimal. The above set
of equations are called Hartree-Fock equations. The rst Coulomb interaction
term just describes the interaction of an individual electron with the average eld
(mean eld) generated by the other electrons. The second Coulomb interaction
term is called exchange interaction and due to the Pauli principle. It complicates
the solution of the Hartree-Fock equations considerably since it is a genuine
integral operator.
Jellium model
In order to focus our investigation on the new eects due to the electron-electron
interaction, we neglect the periodic lattice potential in (7.3). More accurately,
we choose U
ion
(r) = ne
2
such that the lattice potential just compensates the
negative charge of the electrons in a nite system (essentially, we are smearing
out the positive charge of the ions). This ensures overall charge neutrality as is
necessary. In the Hartree-Fock equations (7.3) the lattice potential then exactly
cancels the average eld generated by the electrons and we are left with only the
exchange term (one can justify this a posteriori after verifying that plane waves
are eigensolutions for the electron wave functions, see below). This denes the
so called Jellium model with its Hartree-Fock equations


2
2m

i
(r)

j
_
d
3
r

e
2
[r r

si,sj

j
(r

)
i
(r

)
j
(r) =
i

i
(r)(7.4)
The Jellium model will be at the center of the Advanced Solid State Theory
course. Its Hartree-Fock equations (7.4) can be solved easily with plane waves

j
(r) =
e
i

kjr

V
spin wave function (7.5)
yielding eigenenergies
(

k) =

2
k
2
2m

1
V

k<k
F
4e
2
[

k
F
[
2
(7.6)
Introduction to Solid State Theory (Kehrein) 99 University of Goettingen
7 Outlook: Many-Particle Eects
The second term here is just from the Fourier transformation of the Coulomb
potential
e
2
[r r

[
= 4e
2
1
V

q
e
i q(rr

)
q
2
(7.7)
One can write the single particle eigenenergies in terms of a dimensionless func-
tion F(x)
F(x)
def
=
1
2
+
1 x
2
4x
ln

1 +x
1 x

(7.8)
(

k) =

2
k
2
2m

2e
2

k
F
F
_
k
k
F
_
(7.9)
The ground state energy in the Hartree-Fock approximation is given by lling
all levels with two electrons
E
gs
= 2

k<k
F
(k) (7.10)
and after some algebra one nds
E
gs
N
=
e
2
2a
0
_
3
5
(k
F
a
0
)
2

3
2
(k
F
a
0
)
_
(7.11)
Ry
_
2.21
(r
s
/a
0
)
2

0.946
(r
s
/a
0
)
_
(7.12)
The length r
s
is the radius of a sphere that contains one electron charge for a
given density of electrons n
r
s
def
=
_
3
4n
_
1/3
(7.13)
In metals its ratio to the Bohr radius takes typical values
rs
a0
= 2 . . . 6. The
rst term in (7.11) is just the kinetic energy of the homogeneous electron gas
already derived in (2.17). Notice that the second contribution coming from the
exchange interaction is therefore typically of the same order as the rst term,
and by no means negligible.
An even more disturbing feature is the observation that for the Hartree-Fock
result (7.9) the derivative

k
diverges at the Fermi energy. Since this derivative
is the group velocity, which enters in all transport quantities, this result makes
clearly no sense. This problem can be traced back to the long range charac-
ter of the Coulomb interaction (7.7) and the ansatz with just a single Slater
determinant.
Introduction to Solid State Theory (Kehrein) 100 University of Goettingen
7 Outlook: Many-Particle Eects
What becomes apparent using many-body techniques is that the ground state
cannot be well approximated by such a single Slater determinant, and improve-
ments that take this into account lead to the important physical eect of screen-
ing, which makes the Coulomb repulsion eectively short range.
Remarkably, the resulting theory can again be well described in a single particle
picture, however, for quasiparticles and not the physical electrons. One eect
of the Coulomb interaction is then just an interaction-induced change of the
dispersion relation (typically in strongly correlated materials the eective mass
becomes larger due to interactions). The formal justication of these statements
is the key result of Landaus Fermi liquid theory in the Advanced Solid State
Theory course.
Introduction to Solid State Theory (Kehrein) 101 University of Goettingen

Das könnte Ihnen auch gefallen