Modul 2b-E1:

Introduction to Quantum Mechanics in Solid-State Physics

Tilmann Hickel

October 18, 2019

.
Contents

1 Experiments leading to the breakdown of classical physics 4

1.1 Blackbody radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.1 Wien’s law . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.1.2 Classical Rayleigh-Jeans approach . . . . . . . . . . . . . . . . . . . . 6
1.1.3 Planck’s derivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.2 The photoelectric effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 The Compton effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.4 Wave properties of particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
1.5 Electron diffraction by crystals . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.6 The Bohr atom . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 Double slit experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14

2 Wave functions in quantum mechanics 16

2.1 Probability interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2 Plane waves and wave packets . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.1 Wave packet with equal weights . . . . . . . . . . . . . . . . . . . . . . 17
2.2.2 The Gaussian wave packet . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.3 The group velocity of wave packets . . . . . . . . . . . . . . . . . . . . 18
2.3 Motivation of the Schrödinger equation . . . . . . . . . . . . . . . . . . . . . 19
2.4 Operators and wavefunctions . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

3 Solution of one-dimensional problems 22

3.1 The particle in a box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 The parity operator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 The quantum well . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.4 General considerations on quantum-well like potentials . . . . . . . . . . . . . 27
3.5 Scattering states for the quantum well . . . . . . . . . . . . . . . . . . . . . . 29
3.6 Tunneling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
3.6.1 The probability current . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.6.2 The potential barrier . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.7 The Kronig-Penney model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.7.1 The Bloch theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.7.2 Analytical solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.7.3 Limit of infinitely narrow barriers . . . . . . . . . . . . . . . . . . . . 35
3.8 Band formation in periodic pontentials . . . . . . . . . . . . . . . . . . . . . . 37

4 The harmonic oscillator 40

4.1 Sommerfeld’s polynomial method . . . . . . . . . . . . . . . . . . . . . . . . . 40
4.2 Observables and operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.1 Eigenstates and mixed states . . . . . . . . . . . . . . . . . . . . . . . 42
4.2.2 Adjoint operators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42

2
 4.3 Solving the harmonic oscillator with second quantization . . . . . . . . . . . . 43
4.3.1 Introduction of new operators . . . . . . . . . . . . . . . . . . . . . . . 43
4.3.2 The occupation number operator . . . . . . . . . . . . . . . . . . . . . 44
4.3.3 Derivation of wave functions . . . . . . . . . . . . . . . . . . . . . . . 45
4.4 Phonons and harmonic oscillator . . . . . . . . . . . . . . . . . . . . . . . . . 46
4.4.1 One-dim. vibrations with mono-atomic basis . . . . . . . . . . . . . . 46
4.4.2 One-dim. vibrations with two-atomic basis . . . . . . . . . . . . . . . 48
4.4.3 Generalization to phonon dispersion of crystals . . . . . . . . . . . . . 49
4.4.4 Quantization of phonons . . . . . . . . . . . . . . . . . . . . . . . . . . 50

5 The hydrogen atom 52

5.1 The Schrödinger equation for a central potential . . . . . . . . . . . . . . . . 52
5.2 The angular momentum operator . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.2.1 Raising and lowering operators . . . . . . . . . . . . . . . . . . . . . . 54
5.2.2 The spherical harmonics . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.3 The radial equation for the hydrogen atom . . . . . . . . . . . . . . . . . . . 56
5.4 Interpretation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
5.5 Discussion of eigenfunctions of the H atom . . . . . . . . . . . . . . . . . . . . 59

6 Magnetism in solids 61
6.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.1 Classification of magnetism . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.2 Bohr-van-Leeuwen theorem . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.3 The quantum-mechanical nature of spins . . . . . . . . . . . . . . . . 62
6.2 Magnetism of local moments . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2.1 Exchange interactions . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.2.2 The Heisenberg model . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
6.3 Mean-field theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
6.4 Magnetic excitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.4.1 Magnons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
6.4.2 Linear spin-wave theory . . . . . . . . . . . . . . . . . . . . . . . . . . 68
Chapter 1

Experiments leading to the

breakdown of classical physics

The introduction of quantum mechanics became necessary at the beginning of the last cen-
tury, since basic classical concepts turned out to be not fulfilled under certain experimental
conditions. On the one hand this is the concept of determinism, being one of the key
assumptions of classical physics. It basically says that the state of a systems at a certain
time defines its state at all upcoming times in future. The following aspects are related to
the concept of determinism:

• all physical properties of a system can in principle be measured exactly

• the measure of these properties (incl. energies) are continuous variables

• the time development corresponds to a well-defined motion in phase space (spanned by

position and momentum coordinates)

Within classical physics we further assume that there is a clear distinction between par-
ticles (atoms, electrons, solid bodies, etc.) and waves (electromagnetic waves, water waves,
etc.). Both kinds of systems can carry information and energy, but follow very different rules
for interactions.
These classical concepts reach their limits, if a measurement itself is modifying the con-
sidered system. This is unavoidable if the system size is becoming very small (mass ≈ 10−30
kg, length ≈ 10−10 m). This is the typical regime where quantum-mechanical effects set
in. Quantization is often discussed in terms of energies that are proportional to Planck’s
constant
h
h = 6.624 · 10−34 Js or h̄ = = 1.055 · 10−34 Js. (1.1)
2π
A typical relation, where Planck’s constant enters, is the Heisenberg uncertainty prin-
ciple
h̄
∆qi ∆pi ≥ , (1.2)
2
where ∆qi and ∆pi label the uncertainty in determining the position and momentum of
a certain degree of freedom of the system, respectively. It is important to stress at this
point that this uncertainty is not related to our ability to make precise measurements or to
statistical fluctuations. Instead, we are speaking about a principle property. Consequences
of this important relation will be discussed later.
In the following we will discuss three experiments, which demonstrate that the classical
concepts (continuous variables, distinction between particles and waves) breaks down.

4
Module 2b-E1 1.1. Blackbody radiation

Figure 1.1: Measured spectra of black-body radiation. The classical curve corresponds to the
model of Rayleigh and Jeans. [Taken from wikipedia.org]

1.1 Blackbody radiation

It is a common effect that metals start glowing at very high temperatures, i.e., they are
emitting visible light. However, even at lower temperatures every body is radiating, just
in another frequency range, which is usually heat radiation. A black box is an idealized
system with totally absorbing surfaces, i.e., it does not reflect any electromagnetic waves.
A very good realization of this situation is a box (enclosure) with a small hole. All waves
getting through the hole into the box are not directly reflected. They will, however, yield
an electromagnetic energy density in the box, which results into a radiation out of the small
hole. Such a radiation is called black body radiation.
Typical measured results are shown in Fig. 1.1. In the following we will theoretically
derive the spectrum of such a black-body radiation for different temperatures. We first need
to characterize the energy density of the electromagnetic field
1
ω= (E · D + H · B) . (1.3)
2
It is typically considered in terms of the spectral energy density
Z ∞
dω
ων = ⇐⇒ ω= ων dν. (1.4)
dν 0

From general considerations it is clear that ων is some function f (ν, T ) of the frequency ν
and the temperature T .

1.1.1 Wien’s law

The theoretical challenge was to determine the shape of the function f (ν, T ) mentioned above.
Wilhelm Wien was at least able to reduce one degree of freedom by writing
ν
 
3
ων ≡ f (ν, T ) = ν g , (1.5)
T
where, however, the one-dimensional function g is still completely undefined. Although
Wien’s law given in Eq. (1.5) can be understood exactly by using thermodynamic prin-
ciples, we skip the derivation here for time reasons. The law is in complete agreement with
experimental observations. Some consequences can be derived directly:

T. Hickel 5 WS 2019/20
Module 2b-E1 1.1. Blackbody radiation

• Stefan-Boltzmann law considers the total energy radiated by unit surface area of a
black body and per unit time :
Z ∞ Z ∞
ν
 
ω(T ) = ν 3g dν = T 4 x3 g (x) dx = αT 4 (1.6)
0 T 0

This energy is proportional to the fourth power of the temperature of this body.

• Wien’s displacement law considers the frequency of maximum intensity in the spec-
trum of a black-body:
" #
dων ν ν3 ν
  
2
=0 = 3ν g + g0 (1.7)
dν νmax T T T νmax
3
=⇒ 0 = g(x) + g0(x) (1.8)
x
νmax
=⇒ = x0 = const. (1.9)
T
Hence, the wave-length for which the intensity has a maximum is proportional to 1/T .

The remaining task is to find the functional dependence of g in Eq. (1.5). Again, Wien
made a suggestion based on empirical assumptions:
ν bν
   
gWien = a exp − = a exp [−βbν] . (1.10)
T kB T

Here, kB is the Boltzmann constant, β = kB1T and a, b are free fitting parameters. The ansatz
is not based on a rigid derivation and turns out to be in agreement with experiment only for
the high-frequency regime bν  kB T .

1.1.2 Classical Rayleigh-Jeans approach

Another approach was suggested by Rayleigh (1900). It is based on the equipartition
theorem, which states that in thermodynamic equilibrium each degree of freedom is connected
with the energy
1
E = kB T. (1.11)
2
This theorem is one of the fundamental laws of classical statistical mechanics. The application
of this theorem, therefore, requires the counting of all degrees of freedom related to the black-
body radiation. Coming back to the realization of the box with a hole, the resulting task is
the determination of the number of possible standing electromagnetic waves inside the box.
A standing electromagnetic wave is characterized by the fact that the electric fields dis-
appears at the walls yielding the condition (see also Fig. 1.2a)
y λ 1
a = ny · = ny · . (1.12)
2 2 cos α
Here the positive integer numbers ni for the three spatial dimensions are accordingly:
2a 2a 2a
nx = cos α, ny = cos β, nz = cos γ, or
λ λ λ
 2 
2a 2a 2 2aν 2
    
n2x + n2y + n2z = 2 2 2
cos α + cos β + cos γ = =
λ | {z } λ c
=1

Asking for the number N (ν) of all standing waves with a frequency ≤ ν is then identical
with counting the number of cubes with edge length 1 that fit into a sphere with radius 2aν
c

T. Hickel 6 WS 2019/20
Module 2b-E1 1.1. Blackbody radiation

a) b)

Figure 1.2: (a) Schematic representation of a standing electromagnetic wave with wavelength
λ in a box. (b) Counting the number of possible standing waves.

(sketched in Fig. 1.2b). This number can be determined by comparing the volume of a cube
(=1) with that of 18 (only positive ni allowed) of the sphere :
3
1 4π 2aν

N (ν) =
8 3 c
ν2
=⇒ dN (ν) = 4πa3 dν (1.13)
c3
ν2
=⇒ ων dν = 8π kB T dν (1.14)
c3
In the last step the energy per degree of freedom (1.11) has been used, having in mind that
each electromagnetic wave has two degrees of freedom (electric and magnetic component,
respectively) and two polarization directions. Comparing the result with Eq. (1.5) yields the
Rayleigh-Jeans spectral density
ν kB T
   
gRJ = 8π 3 . (1.15)
T c ν
This result describes the experiments correctly for small frequencies ν, but is also unable to
reproduce the maximum in the λ-dependent spectral density as shown in Fig. 1.1.
The failure of this approach becomes even more apparent when looking at the total energy
density of the black body
Z ∞ Z ∞
8π
ω= ων dν = kB T ν 2 dν = ∞. (1.16)
0 c3 0

The fact that this quantity diverges is known as ultraviolet catastrophe. Since the only as-
sumption entering this derivation is the equipartition theorem in Eq. (1.11), this fundamental
rule of classical statistics does apparently not apply for the black body radiation.

1.1.3 Planck’s derivation

In order to derive the correct spectral density of black-body radiation, Max Planck tried
to avoid the assumption of the equipartition theorem that each wave contributes the same
energy. Instead he considered the energies of different standing electromagnetic waves to be
due to the exchange of energy with oscillating atoms in the walls. The decisive hypothesis in

T. Hickel 7 WS 2019/20
Module 2b-E1 1.2. The photoelectric effect

his analysis is that all oscillators can only be in states with energies being an integer multiple
of an elementary energy quantum ε0 , i.e.

En = nε0 . (1.17)

In short: The energy is quantized!

Based on this hypothesis the average energy ε̄ per oscillator and hence (due to the cou-
pling) per standing wave can be derived. Using Boltzmann statistics the number of oscillators
with energy En is proportional to exp [−βnε0 ] with β = kB1T . Consequently, we have
P (∞ )
n nε0 exp [−βnε0 ] d X
ε̄ = P =− ln exp [−βnε0 ]
n exp [−βnε0 ] dβ n=0
d 1 ε0
 
= − ln = .
dβ 1 − exp [−βε0 ] exp [βε0 ] − 1

Using this energy instead of kB T for the step from Eq. (1.13) to (1.14) yields

8πν 2 ε0 ν 8πh 1
 
ων = 3 or gPlanck = 3
h i . (1.18)
c exp [βε0 ] − 1 T c exp h ν
kB T − 1

For the last step the comparison with Eq. (1.5) provided an important information about the
frequency dependence of the energy quantum:

ε0 = hν, (1.19)

one of the fundamental relations of quantum mechanics. The result in Eq. (1.18) reproduces
the experimental data for all frequencies and it has gWien and gRJ as limiting cases for high
and low frequencies, respectively.

1.2 The photoelectric effect

The photoelectric effect refers to experimental observations of Heinrich Hertz (1887) that
polished metal plates being irradiated by electromagnetic waves may emit electrons. This
effect alone is not very surprizing and can also be explained classically. Within this theory
the following considerations are necessary:

• The energy of the electromagnetic waves is proportional to their intensity.

• The electric field component of the waves forces oscillations of the electrons in the
metal.

• The electrons oscillate faster with increasing intensity and radiation time.

• The emission of an electron is possible as soon as enough energy has been absorbed.

Taking this classical argumentation line seriously, the following consequences need to be
derived:

• The higher the intensity of the radiation, the higher is the energy of the emitted electron.

• The emission of electrons starts with some time delay after the onset of the radiation,
at least for small intensities.

• There is no need for any dependence on the frequency/wavelength of the radiation.

T. Hickel 8 WS 2019/20
Module 2b-E1 1.3. The Compton effect

It was, therefore, quite surprising that completely different observations have been re-
ported by Heinrich Hertz for the photoelectric effect:
• The photoelectric effect happens only for certain wavelengths of the radiation, i.e. only
above a certain threshold frequency. Below this threshold also very high intensities do
not help to emit electrons.

• The energy of the electrons varies linearly with the frequency of the radiation (but does
not dependent on the radiation intensity).

• The photoelectric effect sets in without any time delay.

• The number of the emitted electrons is proportional to the intensity of the radiation.
It was only due to the genius interpretation of the photoelectric effect by Albert Einstein
that the discrepancy between the experimental observations and the classical interpretation
could be resolved. Einstein stated that
• Light and any other electromagnetic radiation consists of a collection of energy quanta
E = hν,

• There is a certain work function W to separate electrons from the metal. This quantity
is material dependent.

• The remaining energy of the electron that overcame the work function is kinetic
1
mv 2 = hν − W
2

• the electrons interact only with single energy quanta.

With these assumptions all experimental observation can be readily explained. However,
we have to accept that the energy is quantized in a similar way as in the black-body radiation
(compare Eq. (1.19)). More importantly, however, one has to accept that electromagnetic
radiation consists of small particles (carrying the energy quanta), which are called photons.
This is the onset of the particle-wave duality in quantum-mechanics.

1.3 The Compton effect

The Compton effect (1923/24) demonstrates the particle nature of radiation even more clearly.
The effect occurs during the scattering of radiation (short-wave X rays) with wave-length λ0
by thin metallic foils. The classical point of view on such a scattering event is:
• The electromagnetic wave excites an oscillation of the electrons in the material.

• The oscillating electrons (Hertz dipole) emit new electromagnetic waves.

• The frequency/wave length of the scattered wave is identical to that of the original
wave and independent of the scattering angle θ.

• Only the intensity of emitted radiation is θ-independent.

The experimental observations, however, were again in conflict with the classical assump-
tions:
• One peak in the scattered spectrum occurs at the same wave length λ0 as the incident
radiation. This peak is due to the interaction with the core electrons of the atoms,
which are rather immobile.

T. Hickel 9 WS 2019/20
Module 2b-E1 1.4. Wave properties of particles

• There is another peak for larger wavelengths λ = λ0 + ∆λ.

• The shift in wave lengths ∆λ depends on the scattering angle like

h
∆λ = (1 − cos θ) (1.20)
me c

Once more, these observations could only be explained with quantum-mechanical con-
cepts. The radiation is for this purpose not only considered to consist of energy quanta. The
photons carry in addition a momentum and are therefore able to perform an elastic collision
with the electrons. During this process the photon looses energy, resulting into an increase
of the wave-length.

Proof: The derivation of Eq. (1.20) is based on the classical mechanical description of a collision between
two particles. Since photons move with the velocity of light, c, concepts from the theory of relativity need to
be applied. Accordingly, the energy of a particle depends on the momentum p and the mass m as follows:
p
E= p2 c2 + m2 c4 . (1.21)
dE
Since the velocity v = dp
of a photon needs to be equal to c, its mass must vanish. Using the energy relation
E = hν, we conclude for the photon momentum p(ph) = hν
c
.
Momentum conservation, comparing (0) before and (1) after the collision, yields
(ph) (el) (ph) (el)
p0 + p0 = p1 + p1
h i2 h i2 h i2 h i2
(el) (ph) (ph) (ph) (ph) (ph) (ph)
=⇒ p1 = p0 − p1 = p0 + p1 − 2p0 · p1

h2  2
h i2
(el)
ν1 + ν02 − 2ν1 ν0 cos θ

p1 =
c2
At the same time energy conservation yields a similar relation:
(ph) (el) (ph) (el)
E0 + E0 = E1 + E1
r 2
(el)
hν0 + mel c2 = hν1 + p1 c2 + m2el c4
h i2
(el)
=⇒ c2 p1 = h2 (ν0 − ν1 )2 + 2hmel c2 (ν0 − ν1 )

Combining both conservation relations to

0 = 2hmel c2 (ν0 − ν1 ) − 2h2 ν0 ν1 (1 − cos θ)

and changing with c = νλ from frequency to wavelength:

ν0 − ν1 1 1 1 1
= − = (λ1 − λ0 ) = ∆λ
ν0 ν1 ν1 ν0 c c
yields the Compton shift given in Eq. (1.20). q.e.d.

1.4 Wave properties of particles

All the experiments described above demonstrate that effects classically associated with waves
can have particle properties. Hence, there exists a wave-particle duality. Even more fasci-
nating is probably the opposite phenomenon that classical particles behave like waves. This
concept is actually also reflected by the central equation of quantum mechanics, namely the
Schrödinger equation
HΨ = EΨ. (1.22)
Here, H is the symbol of the Hamilton operator, describing the development of a physical
system and the interaction of the involved particles. The total energy of the system is given
by the symbol E. Most important is, however, that the state of the system is described by a
wave function Ψ.

T. Hickel 10 WS 2019/20
Module 2b-E1 1.5. Electron diffraction by crystals

Accepting that particles behave like waves, a straight-forward question is about their
wavelength. For this purpose we return to the derivation of the Compton effect in Sec. 1.3.
There we found for the energy of a photon E = hν = pc, which in combination with c = νλ
results to
h
λ= . (1.23)
p
The assumption is that this relation applies not only to photons, but to all particles. The
suggestion that the wavelength is the quotient of Planck’s constant h and the momentum of
the particle p was first raised by De Broglie in 1923. Taking into account that Planck’s
constant h is a small quantity (see (1.1)), the wavelength is typically much more relevant on
an atomic than a macroscopic scale. Some consequences are discussed now.

1.5 Electron diffraction by crystals

The reflection of waves by crystals is given by the Bragg condition. In a simplified derivation,
parallel lattice planes separated by a distance d are considered.1 If two waves that are
reflected by adjacent planes interfere constructively, a peak in the spectrum is observed. The
corresponding condition is:
2d sin θ = nλ, (1.24)
where θ is the angle of the incident beam with the planes and n is a positive integer number.
This simple relation already tells us that diffraction patterns are only visible if λ ≤ 2d. Hence,
not all waves can be used for crystals with lattice constants of a few Å. Typically wavelengths
are:

visible ligth −→ 390 − 750 nm

x rays −→ 0.01 − 10 nm
electrons accelerated by 1 V −→ 1.2 nm

One can see that the wavelength of electrons is in the right order of magnitude to make
spectroscopy of solids. Standard applications are nowadays transition electron microscopes
(TEM). A similar example, which makes directly use of Eq. (1.23), is neutron diffraction.

Derivation of Laue equations

Looking at the diffraction process more carefully, it is actually not due to mirror planes,
but due to individual atoms. The atoms are sitting on a crystal lattice, characterized by the
lattice vectors a1 , a2 and a3 . The position of every atom, Rn , is given by a linear combination
of these lattice vectors:
3
X
Rn = ni a i ni = 1, 2, 3, . . . , Ni (1.25)
i=1
The incident beam, that interacts with the lattice, can mathematically be described by
plane waves. The dependence of the amplitude on the position r and the time t is for such a
wave given by
A(r, t) = A0 ei(k·r−ωt) , (1.26)
where ω is called angular frequency and k wave vector. The latter determines the direction
of wave propagation, since the scalar product k · r corresponds to a projection of the position
vector r on the wave vector k. Its modulus k = |k| also determines the wave length of the
beam, since two points that are in beam direction separated by λ = 2π k have the same phase.
In the following we consider a snapshot at t = 0.
1
C. Kittel: Introduction to Solid State Physics (Wiley, New York, 1996) Chap. 2

T. Hickel 11 WS 2019/20
Module 2b-E1 1.5. Electron diffraction by crystals

Figure 1.3: Diffraction of an incident beam k by a crystal structure with atomic positions Rn .
The red star indicates the (far apart) observation point where the intensity of the diffracted
beam k0 is measured.

Figure 1.4: Visualization of the Ewald construction: The incident wavevector k should point
to a reciprocal lattice point (green ball). Then one draws a (three dimensional) sphere around
the termination point of the wave vector k with the radius |k|. A diffraction maximum is
observed in all those direction k0 , where there sphere intercepts a point of the reciprocal
lattice (red ball). Those points (green and red ball) should by definition of the reciprocal
lattice be connected by a reciprocal vector G.

T. Hickel 12 WS 2019/20
Module 2b-E1 1.5. Electron diffraction by crystals

The plane wave described by Eq. (1.26) gives at each lattice point Rn rise to a spherical
wave:
ikr0
0 ik·Rn e
A (r) = A0 e . (1.27)
r0
We place the centre of the coordinate system in the crystal and consider a point r at which
we want to observe the diffraction pattern. Therefore, the direction of r is the direction
of the diffracted beam with wave vector k0 = k rr . Here, |k| = k = |k0 | due to elastic
scattering. Then the distance r0 is in good approximation (considered point far enough)
given by r0 = r − k0 · Rn /k. Entering this into (1.27) yields a diffracted amplitude

eikr 0 n
A0 (r) = A0 0 n
ei(k−k )·R . (1.28)
r−k ·R
The term k0 · Rn is negligible in the denominator, but decisive in the phase. The prefac-
ikr
tor A0 e r is direction-independent, important for maxima and minima is the exponential
exp [i (k − k0 ) · Rn ]. Note, that this phase difference can also be obtained, if one assumes a
plane wave for the diffracted beam and considers the extra way part 1 of the beam has to
run as compared to part 2.
For the intensity of the diffracted beam the modulus square of the total amplitude, which
is a superposition of the effect of all individual atoms in the crystal, has to be considered. It
is given by2
2
h i 2
3 Ni 3 sin 1 Ni (k − k0 ) · ai
0 1 Y 0 1 2
ei(ni −1)(k−k )·ai
X Y
I (r) ∼ 2 = 2 h i (1.29)
r i=1 ni =1
r i=1 sin 1
(k − k0 ) · ai
2

The right part of Eq. (1.29) indicates the sinusoidal change of maxima in the diffraction
pattern. However, one sees more clearly from the left part of the equation that the intensity
becomes maximal, if all terms in the sum have their maximal value 1. One can conclude that
maxima in the diffraction patterns will occur, whenever the wave vector of the incident beam
k and the wave vector of the diffracted beam k0 fulfil the condition

ai k0 − k = 2πvi


vi = 1, 2, 3, . . . (1.30)

The equations on the right are called Laue equations.

The reciprocal lattice

The Laue equations given in Eq. (1.30) are a condition for the wave vector difference

G = k0 − k. (1.31)

In other words: All G vectors, for which ai · G = 2πvi form maxima in the diffraction
spectrum. Since the vi need to be integer numbers, these G form a discrete set of vectors.
This set is also called reciprocal lattice.
Also for this lattice it is possible to define basis vectors b1 , b2 and b3 . They are chosen
such that they themselves fulfil the Laue equations

bi · aj = 2πδij . (1.32)

It is easy to show that this condition is fulfilled by the following choice:

a2 × a3 a3 × a1 a1 × a2
b1 = 2π , b2 = 2π , b3 = 2π . (1.33)
a1 · (a2 × a3 ) a1 · (a2 × a3 ) a1 · (a2 × a3 )
Upcoming transformation makes use of geometric series and the relation 1 − eix = eix/2 e−ix/2 − eix/2 .
2
 

T. Hickel 13 WS 2019/20
Module 2b-E1 1.6. The Bohr atom

Based on this definition, a reciprocal lattice vector is given by a linear combination of the
basis vectors b1 , b2 and b3 :
G = v1 b1 + v2 b2 + v3 b3 (1.34)
That the reciprocal lattice of G vectors yields the diffraction pattern, e.g. of a neutron
scattering experiment can be most clearly seen in a Ewald construction (see 1.4).

1.6 The Bohr atom

There have been several assumptions on the market, how electrons are arranged within an
atom. The Rutherford planetary model assumed that negatively charged electrons travel
in circular or elliptic orbits about the nucleus. However, this should lead to a continuous
spectrum of atoms. Further, classical electrodynamics tells us that accelerated charges in an
electric field will yield a radiation and therefore a loss of energy. This is not consistent with
observations in experiment like the Balmer series (1885)

1 1 1
 
= Ry 2
− 2 . (1.35)
λ n m

Therefore, Niels Bohr (1913) came up with the following set of postulates:

1. The electrons in an atom belong to stationary states with discrete energies Ei and their
motion within these states happens without any energy radiation.

2. The transition between these states is related with the emission/absorption of electro-
magnetic waves with an unique frequency

hν = E1 − E2 with E1 > E2 (1.36)

3. A typical description of the quantization condition is given in terms of classical action

variables: I
J(E) = pi dqi = nh. (1.37)
ΓE

This situation becomes more clear, if the electrons are considered as waves. Then a stationary
state can be identified with a standing wave. Using the De Broglie wave length (1.23) in
Eq. (1.37) yields the quantization condition

2πr = nλ, (1.38)

which means that the circumvent of the orbit should be a integer multiply of the wavelength.
This interpretation allows one to understand the postulates of Bohr.

1.7 Double slit experiments

See also video under youtube:
http://www.youtube.com/watch?v=0ElSXo1HWS4

Consider a screen which does not allow the transmission of the radiation, but has two
slits S1 and S2 . On a detector plate behind this screen the incoming particles yield some
reaction to localize the position where the particles arrived. Several scenarios are possible:

T. Hickel 14 WS 2019/20
Module 2b-E1 1.7. Double slit experiments

a) Classical particles
It is possible that the particles are somehow influenced by the slits, however their effects are
independent. The result is the same if both slits are open together, or if both slits are open
one after the other for the same time. The intensity on the detector is given by:
(a) (a)
I (a) (x, y) = I1 (x, y) + I2 (x, y) (1.39)

b) Electromagnetic waves
There is already some diffraction pattern behind a single slit. It can be derived with Huygens
principle and yields intensity maxima at angles θ following the condition
1
 
n+ λ = d sin θ, (1.40)
2
with d being the thickness of the slit. If, however, both slits are open, the pattern is different
than just having the combined effect of two single slits. This is due to the interference term
in
(b) (b) (b)
I (b) (x, y) = I1 (x, y) + I2 (x, y) + I12 (x, y). (1.41)
The intensity is proportional to the modulus square of the amplitude sum

I (b) ∼ |E1 (r, t) + E2 (r, t)|2 6= |E1 (r, t)|2 + |E2 (r, t)|2 (1.42)

c) Light with very small intensity

Single photons are recognized by the detector. The position of the single photon is unpre-
dictable. However, the statistics over some time can be considered. The resulting interference
pattern corresponds to I (b) . Since the photons might have passed the two slits individually,
this means they must have interfered with themselves.

d) Single electrons
The situation is identical to the observation in c): Single electrons are seemingly randomly
distributed, the statistical pattern corresponds to I (b) . There is again a difference between
opening the slits one after the other and having them open at the same time. Therefore,
the same particle-wave duality as for light also applies for electrons. If the passage of the
electron through one of the slits is detected, the interference pattern looks again like that of
two independent slits.

Experimental confirmation
Tonomura has demonstrated the effect for electrons, using, however, not really a screen with
two slits, but a very thin filament which is positively charged. The motion of the electrons
is then bended, such that the region of electrons passing on the right and on the left side
overlap. An interference pattern has been observed.
Double slit experiments for neutrons with wavelength 18.5 Å have been performed by
Zeilinger in 1988. Further, diffraction experiments have been performed with molecular
beams of hydrogen and helium, or even fullerenes (C60 compounds).
All these experiments do not only qualitatively confirm the wave-like nature of particles,
the corresponding diffraction patterns also confirm that the wavelength of the particles is
correctly described by the de Broglie relation (1.23).

T. Hickel 15 WS 2019/20
Chapter 2

Wave functions in quantum

mechanics

2.1 Probability interpretation

It is already clear from the interpretation of the double-slit experiments that elementary
processes cannot be predicted, but that statistical arguments are necessary. This is best
met by considering material waves as probability waves. They are not really properties of
particles, but related to the statistics or particles. In single events, particles will behave like
classical objects.
Hence, the wave functions correspond to a probability. It also becomes apparent in the
above mentioned examples that the modulus squared should be the relevant quantity. This
becomes even mathematically clear, considering that not only real, but also complex wave
functions are possible. Therefore:

ρ(r, t)d3 r = |Ψ(r, t)|2 d3 r (2.1)

describes the probability of finding a particle at time t in the volume element d3 r.

The quantity ρ(r, t) in (2.1) is also called probability density, which should be normalized
and fulfils a continuity equation
∂
ρ(r, t) + ∇ · j(r, t) = 0. (2.2)
∂t
It is the only quantity that can really be measured, whereas the wave function Ψ(r, t) is not
directly accessible. Therefore, phase factors φ(r, t) in

Ψ(r, t) = |Ψ(r, t)| eiφ(r,t) (2.3)

are typically not important, unless some linear combination of wave functions

Ψ(r, t) = α1 Ψ1 (r, t) + α2 Ψ2 (r, t) (2.4)

is considered. In this case the probability becomes

|Ψ(r, t)|2 = α12 |Ψ1 (r, t)|2 + α22 |Ψ2 (r, t)|2 + [α1 α2 Ψ1 (r, t)Ψ∗2 (r, t) + h.c.] (2.5)

and everything depends on the product of the phase factors. Only if the two participating
waves are not coherent, the phase factors can be considered to be independent and average
out over time.

16
Module 2b-E1 2.2. Plane waves and wave packets

2.2 Plane waves and wave packets

There are two equivalent notations for a plane wave propagating in positive k-direction:

ψk (r, t) = A1 cos(kx − ωt) + A2 sin(kx − ωt) (2.6)

i(kx−ωt) −i(kx−ωt)
= Ae + Be , (2.7)

where x is the projection of r on the k direction. The missing dependence on y, z (perpen-

dicular directions to propagation direction) is responsible for the name plane wave. The
second version of these equations is often preferred in material science.
The wave length is directly related to the wave vector k
2π p
λ= =⇒ k= , (2.8)
k h̄
where the second relation makes use of the De Broglie wavelength (1.23). A similar relation
holds for the dependence of the frequency ν on the angular frequency ω:
1 ω E
ν= = =⇒ ω= , (2.9)
T 2π h̄
where E = hν has been used in this case. The dependence of ω on the wave vector k is called
dispersion relation. It can depend a lot on the medium through which the wave is travelling.
For light propagation in vacuum, we have

c = νλ =⇒ ω(k) = kc. (2.10)

In dispersive media the velocity is given by c/n with some diffraction constant, which depends
on the wave length λ.
The superposition of plane waves having different k vectors with corresponding amplitudes
A(k) is called wave packet
Z∞
ψ(x, t) = dkA(k)ei(kx−ωt) (2.11)
−∞

2.2.1 Wave packet with equal weights

Any plane wave has a well defined wave length of λ = 2πk and a corresponding momentum
of p = h̄k. Combining plane waves to packets, however, yields a mixture of wavelengths.
Consider a wave packet were all contributing plane waves have the same weight:
(
N −K ≤ k ≤ K
A(k) = (2.12)
0 elsewhere

This gives a spreading of the wave vectors over ∆k = 2K.

However, this uncertainty of the wave vector (and the momentum) brings clarity with
respect to the position of the described object. Note that for a single plane wave the proba-
bility of finding a particle at a certain position x is due to |ψ(x, t)|2 = 1 the same everywhere
in space. For the above defined wave packet we have instead
+K
N  iKx sin Kx
Z 
ψ(x, 0) = N dkeikx = e − e−iKx = 2N (2.13)
ix x
−K

With an error of less than 5 % the area below (sin x/x)2 is limited to the interval (−π, π).
Hence, the position x is given with an uncertainty of 2π/K, yielding ∆k∆x = 4π. The

T. Hickel 17 WS 2019/20
Module 2b-E1 2.2. Plane waves and wave packets

precise number on the right hand side depends of course on the definition of the width. More
generally, one has for Fourier integrals always
1 (2.8) h̄
∆k∆x > =⇒ ∆p∆x > (2.14)
2 2
which provides a link to the Heisenberg uncertainty relation.
Also for plane waves (2.14) is fulfilled, although ∆p = 0 in this case. However, the
complete uncertainty about the position can be expressed as ∆x = ∞.

2.2.2 The Gaussian wave packet

A special wave package is given by an amplitude that has the form of a Gaussian:
α 2
A(k) = e− 2 (k−k0 ) . (2.15)

For the probability of having a certain k vector in the packet the square of the function is
relevant. In other words, the uncertainty for the wave vector ∆k is given by the k − k0 for
which the A2 (k) is reduced to 1e of its maximum value. Taking the interval to both sides, we
√
have ∆k = 2/ α.
On the other hand the uncertainty for the x position of such a wave package can also be
calculated. Inserting (2.15) into (2.11) yields at t = 0:
x
+∞−i α
Z∞ r
x2 2π ik0 x − x2
Z
−α (k−k0 )2 ikx ik0 x − 2α −α q2
ψ(x, 0) = dke 2 e =e e dqe 2 = e e 2α (2.16)
x
α
−∞ −∞−i α

Therefore, also in real space the probability distribution has the shape of a Gaussian. Hence,
√
the uncertainty for x is given by ∆x = 2 α. Together we obtain ∆k∆x = 4, which fulfils
again (2.14).

2.2.3 The group velocity of wave packets

The motion of a wave packet as given in (2.11) is only defined, if the dispersion ω(k) is known.
We can make a Taylor expansion around the wave vector k0 where A(k) has its maximum:
!
∂ω 1 ∂2ω
 
ω(k) = ω(k0 ) + (k − k0 ) + (k − k0 )2 + ... (2.17)
∂k k=k0 2 ∂k 2 k=k0

The first two terms are typically most decisive and yield with q = k − k0 for the wave packet
in (2.11):
Z∞
ψ(x, t) = ei(k0 x−ω(k0 )t)
dq A(q + k0 )eiq(x−(∂ω/∂k)0 t) . (2.18)
−∞

The exponential in front describes the oscillating wave function with the major frequency
and wave length. The exponential in the integral describes the envelope and its motion.
Therefore, the so called group velocity is defined as
∂ω(k)
 
vg = . (2.19)
∂k k=k0

It describes the velocity with which information is transported. In the case of light propaga-
tion in vacuum (see Eq. (2.10) ), it
is identical with the speed of light c.
The next order β = ∂ 2 ω/∂k 2 0 in the Taylor expansion modifies the amplitude of the
wave packet. This is best demonstrated in the case of a Gaussian packet. In this case the

T. Hickel 18 WS 2019/20
Module 2b-E1 2.3. Motivation of the Schrödinger equation

1 2
additional terms eiqvg t and ei 2 q βt
can be incorporated into (2.16) by replacing α → α + iβt
and x → x − vg t. This yields
" #
2 2π α(x − vg t)2
|ψ(x, t)| = p 2 exp − (2.20)
α + β 2 t2 α2 + β 2 t2
√ p
resulting into a width of 2 α 1 + β 2 t2 /α2 . The latter systematically increases with time
and therefore corresponds to a spreading of the packet.

2.3 Motivation of the Schrödinger equation

We use the description of the wave function as a wave packet (2.11) and reformulate the
variables in terms of momentum and energy using (2.8) and (2.9). One obtains
Z∞
1
ψ(x, t) = √ dp φ(p)ei(px−Et)/h̄ (2.21)
2πh̄
−∞

If we now assume for the energy dispersion E = p2 /2m, the group velocity has the following
form
∂ω ∂E p
vg = = = . (2.22)
∂k ∂p m
This gives a consistent picture of a free particle and confirms for instance the relation
E = h̄ω in (2.9).
However, it is of course our desire to go beyond a free particle and to consider the influence
of a potential
p2
E= + V (x). (2.23)
2m
Including (2.23) into (2.21) would mean that V (x) only plays the role of an unimportant
phase factor. Therefore, it is more meaningful to identify a relation for (2.21), which can be
generalized to arbitrary potentials. For this purpose consider the following partial derivatives:
Z∞
∂ i 1
ψ(x, t) = − √ dp φ(p)E(p)ei(px−Et)/h̄ (2.24)
∂t h̄ 2πh̄
−∞
Z∞
∂2 i2 1
ψ(x, t) = √ dp φ(p)p2 ei(px−Et)/h̄ , (2.25)
∂x2 h̄2 2πh̄
−∞

which due to E(p) = p2 /2m both yield a factor of p2 for the free particle.
One can draw the following important conclusions. On the one hand the following corre-
spondences between physical quantities and derivatives seem to be reasonable:
∂ h̄ ∂
E −→ ih̄ and p −→ . (2.26)
∂t i ∂x
On the other hand, the wave function of the free particle (2.21) apparently fulfils the following
equation " #
h̄2 ∂ 2 ∂
− 2
ψ(x, t) = ih̄ ψ(x, t), (2.27)
2m ∂x ∂t
which is via (2.26) closely related E(p) = p2 /2m.
Having said this, it seems also to be clear, how a generalization according to (2.23) should
look like: " #
h̄2 ∂ 2 ∂
H ψ(x, t) := − + V (x) ψ(x, t) = ih̄ ψ(x, t). (2.28)
2m ∂x2 ∂t

T. Hickel 19 WS 2019/20
Module 2b-E1 2.4. Operators and wavefunctions

The Hamiltonian H in this equation is the operator of the total energy, i.e. the sum of kinetic
and potential energy, H = T + V . If both terms are considered in real space, then

p̂2 h̄2 ∂ 2
H= + V (x̂) = − + V (x). (2.29)
2m 2m ∂x2
Eq. (2.28) is the time-dependent Schrödinger equation. It becomes immediately clear
that plane waves (2.8) generally do not fulfil this equation. Still a superposition of plane waves
like in (2.21) is a possible solution. It can than be understood as a Fourier transformation of
the real-space solution.
In case that the potential V (x) shows no time dependence, the time-dependent Schrödinger
equation can be tackled with a separation ansatz:

ψ(x, t) = ψ̃(x) · ϕ(t). (2.30)

Since Eq. (2.28) needs to be fulfilled for all positions and all times, the problem then decouples:
∂
Hψ̃(x) ih̄ ∂t ϕ(t)
=W = . (2.31)
ψ̃(x) ϕ(t)

The time-dependent part (right-hand side) can be directly solved by setting

Wt
ϕ(t) = e−i h̄ , (2.32)

which also makes clear that W can be identified with the total energy E. The phase factor
vanishes if the modulus squared of the wave function is considered, as needed if we look at
the physically measurable probability interpretation. As a consequence of the transformation
above, we obtain the time-independent Schrödinger equation
" #
h̄2 ∂ 2
H ψ̃(x) := − + V (x) ψ̃(x) = E ψ̃(x). (2.33)
2m ∂x2

Both versions of the Schrödinger equation, (2.28) and (2.33), provide a clearer impression
of the nature of the Hamilitonian H, which was previously only characterized as “describing
the development of a physical system”. The time-dependent version (2.28) demonstrates best
that H defines the time evolution of the states. However, also the time-independent version
(2.33) involves with the spatial derivative a change of the wave function. The potential V (x),
on the other hand, is the result of an interaction with other particles.

2.4 Operators and wavefunctions

Mathematically, the time-dependent Schrödinger equation is a linear differential equation.
The fact that the Hamiltonian (2.29) includes derivatives makes clear that H is actually an
operator acting on the wave function. Also the potential V (x̂) can be considered as being an
operator. It results, however, just in a multiplication of V (x) with the wave function, provided
both are represented in position space. Some care is necessary if x̂ and p̂ are simultaneously
applied to ψ̃(x), since their order is important. More specifically

h̄ ∂ ∂  h̄ ∂ ∂
    
(x̂p̂ − p̂x̂)ψ̃(x) = x ψ̃(x) − xψ̃(x) = x −1−x ψ̃(x) (2.34)
i ∂x ∂x i ∂x ∂x
In short, one can define the so-called commutator of x̂ and p̂ as

[x̂, p̂]− = ih̄. (2.35)

T. Hickel 20 WS 2019/20
Module 2b-E1 2.4. Operators and wavefunctions

The fact that the operators x̂ and p̂ do not commute is closely related to the Heisenberg
uncertainty relation. It means that both quantities cannot be measured exactly at the same
time. This will become clearer later in the course.
After noting that H is an operator, it becomes furthermore apparent that Eq. (2.33) is
an eigenvalue equation. The eigenvalues are the energies E and the eigenfunctions are the
wave functions ψ̃(x).
A constant shift V0 in the one-dimensional potential is unimportant for the physics. Any
eigenfunction ψ̃(x) belonging to a potential V (x̂) is also an eigenfunction of the potential
V (x̂) + V0 . Only the eigenvalue changes from E to E + V0 . This will modify the phase factor
in the time dependence (2.32), but not the physical properties. The observation is consistent
with classical physics. Also there, the absolute energy is not important. What counts is the
gradient of the potential, which determines the forces acting on the particles.
Another general property of the wave functions that will become important in the con-
siderations of the next chapter is its continuity. This can be concluded from the following
modification of the 1d Schrödinger equation (2.33):

d2 2m
2
ψ(x, t) = 2 (V (x) − E) ψ(x). (2.36)
dx h̄
In this equation, the potential might be not continuous, containing finite size jumps. As
long as it is not diverging, however, the right hand side of (2.36) is always integrable. The
integration of the left side yields ψ 0 (x), which is therefore continuous. The same is then true
for the function ψ(x) itself. An exception in the argumentation occurs, if the jump in the
potential goes to ±∞. In this case, the continuity of ψ 0 (x) is not guaranteed, whereas it
should for physical reasons still be valid for ψ(x).

T. Hickel 21 WS 2019/20
Chapter 3

Solution of one-dimensional
problems

The aim of the module is to introduce quantum-mechanics in particular for solid-state physics.
Therefore, we will now start to work on scientific problems that are typical for solid-state
physics. One of the most prominent tasks in this field is the prediction of the electronic
band structure in the periodic potential of regularly arranged atoms in a crystal. The full
complexity of this problem is typically treated with density functional theory, one of the
modern ab initio techniques used in material science. For more details on this, please wait
for module 3a.
However, the fundamental understanding of such a band structure can already be achieved,
if a reduction to a single dimension is performed. Furthermore, we will start with a simplified
shape of this one-dimensional potential, increasing step by step the complexity.
Please note that apparently simple one-dimensional potentials can already be connected
to state-of-the-art research in materials science. To give an example, the field of quantum-
cascade lasers should be mentioned here. These lasers are based on multi-layer nanostructures
with a regular repetition of two different kind of semiconductor materials. Here, the required
difference of the materials is the hight of the valence band edge. If in addition an electric
field is applied, the potential has the shape as presented in Fig. 3.1. The special construction
of the potential makes it possible that the light emission of the electrons during changing the
energy levels in the potential wells happens coherently, giving rise to the laser effect.
Another problem with high relevance for solid-state physics is the vibration of atoms at
finite temperatures. In good approximation, the motion is characterized by a back-driving
force, which is proportional to the displacement of an atom out of its equilibrium position.
Such a behaviour gives rise to a quadratic potential and a harmonic oscillation. The potential
of an harmonic oscillator does not only occur in physics in the context of lattice vibrations,
but in various other fields. too. Due to its great importance, we will discuss this problem
rather extensively.

3.1 The particle in a box

The particle in the box is the simplest situation of a one-dimensional potential and the first
step towards an understanding of band structures. The corresponding potential is given as:

 ∞
 x<0
V (x) = 0 for 0<x<a (3.1)
 ∞

x>a

For all the regions, where the potential is ∞ we can immediately conclude that the wave
function vanishes. This is because otherwise an infinite left hand side of the Schrödinger

22
Module 2b-E1 3.1. The particle in a box

Figure 3.1: Position dependence of the potential for an electron in a quantum-cascade laser.
The regions with higher and lower potential correspond to two different semiconductors. The
overall slope is due to an external electric field. Electrons make transitions between bound
states (blue, red arrows) in the potential and tunnel from one potential well to neighbouring
well (grey arrows).

Figure 3.2: Potential and solutions for particle in a box

equation (2.33) needs to be identical to a finite right hand side, provided that the wave
function is still normalized. Therefore, the problem formulated in (3.1) reduces to the interval
(0, a) with the boundary conditions:
(
0 x<0
ψ(x) = . (3.2)
0 x>a

Within this interval, the differential equation to be solved is given by:

d2 ψ(x) 2mE
+ 2 ψ(x) = 0. (3.3)
dx2 h̄
In principle, two cases can be distinguished. In the case E < 0 we introduce the following
notation: s
d2 ψ(x) 2m|E|
− κ2 ψ(x) = 0 with κ = . (3.4)
dx2 h̄2
Two independent solutions of the 2nd order differential equation are eκx and e−κx , of which
only the linear combination of ψ(x) = sinh(kx) fulfils the left boundary condition in (3.2).
Since this solution is however a monotonous function, it will never fulfil the right boundary
condition in (3.2). Since the wave function needs to be continuous, there is therefore no
solution for this case.

T. Hickel 23 WS 2019/20
Module 2b-E1 3.1. The particle in a box

Therefore, we only consider the case E > 0 for which we introduce a similar notation
s
d2 ψ(x) 2mE
+ k 2 ψ(x) = 0 with k = . (3.5)
dx2 h̄2
Again the 2nd order differential equation has with eikx and e−ikx two independent solu-
tions. This time the linear combination ψ(x) = sin(kx) fulfills the left boundary condition
in (3.2). The decisive difference as compared to the first case is, however, that sin(kx) is
not a monotonous, but a periodic function. Therefore, also the second boundary condition
u(a) = 0 can be fulfilled, provided that
nπ h̄2 k 2 h̄2 π 2 2
k= ⇐⇒ En = = n (3.6)
a 2m 2ma2
for n = 1, 2, 3, . . .. The corresponding normalized eigenfunction is then given by
r
2 nπx
 
ψn (x) = sin . (3.7)
a a
We would like to discuss this solution in a bit more detail.
2 2
h̄ π
1. First of all let’s look at the eigenvalues. The lowest one is given by E1 = 2ma 2 and
is, therefore, non-zero. This is in contrast to classical physics, where the lowest energy
would just correspond to a particle at rest in the minimum of the potential, hence
E = 0.
2. We further notice that the eigenvalues (3.6) are quantized with a n2 dependence. After
all the considerations in Chap. 1 this is perhaps not surprising anymore. Still, it is
worth mentioning here, that the discreteness of the eigenstates is a direct consequence
of the boundary conditions of the problem.
3. Looking at the eigenfunction (3.7) one should note the orthogonality of wave functions
belonging to two different eigenstates:
Za
2  nπx   mπx 
hψn (x)|ψm (x)i = dx sin sin
a a a
0
Za     
1 (n − m)πx (n + m)πx
= cos − cos
a a a
0
sin(n − m)π sin(n + m)π
= −
(n − m)π (n + m)π
= δnm .

This is again a very general property for these kind of problems.

4. It is also clear that due to the relation to the kinetic energy, the expectation value hp2 i
is given by 2mEn and has therefore a finite value. In fact, the kinetic energy becomes
larger with increasing number of nodes in the solution. The latter can be explained by
noticing the relation to the curvature of the eigenfunction:
+∞ 2
dψ(x)
Z
hp2 i = h̄2 dx .
dx
−∞

5. The expectation value hpi of the momentum, on the other hand, vanishes for all states:
Za Za
d ih̄ d 2
hpi = −ih̄ dx ψ(x) ψ(x) = − dx ψ (x) = 0
dx 2 dx
0 0

T. Hickel 24 WS 2019/20
Module 2b-E1 3.2. The parity operator

Finally, we would like to investigate what happens, if the whole “particle in a box” problem
would have been considered for the symmetric situation that the wave function is non-zero in
the interval (−a/2, +a/2). The situation can be achieved by performing a shift of variables
x → x − a2 . The eigenvalue (3.6) are not affected by this procedure. This is a bit different for
the eigenfunctions (3.7):
r
2 nπx nπ
 
ψn (x) → sin − (3.8)
a a 2
r  r
2 nπx nπ 2 nπx nπ
      
= sin cos − cos sin (3.9)
a a 2 a a 2
r (
2 sin nπx


= a
for n = 2, 4, 6, . . . (3.10)
a cos nπx a for n = 1, 3, 5, . . .

Hence, we have now symmetric as well as antisymmetric functions as solutions. This concept
will be discussed more systematically in the next section.

3.2 The parity operator

The symmetrization of the particle-in-a-box problem has the logical consequence that also
the wave functions (3.10) become symmetric, i.e., they are either even or odd. We would like
to consider this property more systematically by introducing the parity operator. It is defined
such that the sign in the argument of the wave function to which it is applied changes:

Π̂ψ(x) = ψ(−x). (3.11)

Clearly, even and odd wave functions are eigenfunctions of this operator, since

Π̂ψ (+) (x) = ψ (+) (x) (3.12)

(−) (−)
Π̂ψ (x) = −ψ (x). (3.13)

and any wave function can be expressed as a sum of an even and an odd wave function
1  1 
ψ(x) = 1 + Π̂ ψ(x) + 1 − Π̂ ψ(x) (3.14)
2 2
1 1
= (ψ(x) + ψ(−x)) + (ψ(x) − ψ(−x)) . (3.15)
2 2
Furthermore, it is easy to show that “+1” and “-1” are the only eigenvalues of Π̂:

ψ(x) = Π̂ψ(−x) = Π̂2 ψ(x) = Π̂ (λψ(x)) = λ2 ψ(x), (3.16)

giving only the solutions λ = ±1 (eigenvalues of Hermitian operators are real).

Most important is, however, the time evolution of the symmetry character of a wave
function. According to the time-dependent Schrödinger equation (2.28) we have:
∂
 
Π̂Hψ(x, t) = Π̂ ih̄ ψ(x, t) . (3.17)
∂t
Provided that the Hamiltonian commutes with the parity operator, i.e.,
h i
Π̂, H = Π̂H − HΠ̂ = 0 (3.18)
−

the above equation can be changed to:

∂    
ih̄ 1 ± Π̂ ψ(x, t) = H 1 ± Π̂ ψ(x, t). (3.19)
∂t

T. Hickel 25 WS 2019/20
Module 2b-E1 3.3. The quantum well

Figure 3.3: Potential of a quantum well.

This means that the even and odd part of the wave function separately fulfils the Schrödinger
equation and that the wave function preserves its symmetry property over time. In other
words: whenever the condition (3.18) is fulfilled, it is completely sufficient to search for even
and odd wave functions as separate solutions of the Schrödinger equation. This knowledge
simplifies in many cases the calculations, as can be seen in the following example.

3.3 The quantum well

Starting from the particle-in-a-box problem (3.1), the quantum well

 0
 x < −a region A
V (x) = −V0 for −a < x < a ⇒ region B (3.20)

 0 x>a region C
is the next step towards a more realistic description of electrons in a crystal. It describes a
particle which is trapped in a certain region in space (for E < 0), but is able to escape from
this region if its energy is sufficiently high (E > 0). We focus the considerations here on the
bound states with E < 0.
With similar definitions as in (3.4) and (3.5), namely
s
d2 ψ(x) 2m|E|
region A & C: − κ2 ψ(x) = 0 with κ = (3.21)
dx2 h̄2
s
d2 ψ(x) 2m(V0 + E)
region B: + k 2 ψ(x) = 0 with k = . (3.22)
dx2 h̄2
we obtain the general solution

κx −κx x < −a (A)
 α+ e + α− e

ψ(x) = β+ eikx + β− e−ikx for −a< x <a (B) (3.23)
 γ eκx + γ e−κx

a< x (C)
+ −

Already due to the fact that a normalization of the wave function is necessary, an expo-
nential increase at ±∞ needs to ruled out. For this reason α− = γ+ = 0. Since the potential
(3.20) fulfills the symmetry condition (3.18), we can further assume that the wave function
is either symmetric or antisymmetric. We can therefore, just consider the region x > 0.
Knowing that the wave function and its first derivative need to be continuous, we have the
following boundary conditions:
 
β eika ± e−ika = αe−κa (3.24)
 
ikβ eika ∓ e−ika = −ακe−κa (3.25)

T. Hickel 26 WS 2019/20
Module 2b-E1 3.4. General considerations on quantum-well like potentials

Figure 3.4: Graphical solution of the system of equations (3.26) and (3.27). [Taken from W.
Nolting, Grundkurs Theoretische Physik, vol. 5-1, p. 266 (Springer, Berlin, 2009)]

The quotient of both equations gives:

(
k tan(ka) for symmetric case
κ= (3.26)
−k cot(ka) for antisymmetric case

However, this is not the only relation between k and κ. By definition (3.21) and (3.22) we
have in addition
2m
k 2 + κ2 = 2 V0 . (3.27)
h̄
This is the equation of a circle. In combination with Eq. (3.26) this leads to (several discrete)
solutions for k and κ. However, they cannot be derived analytically. Defining η = κa and
ξ = ka one can, instead, plot the dependencies as in Fig. 3.4.
The corresponding wave function is in the symmetric case (antisymmetric case is similar)
given by:  κx
 e −κa
 x < −a (A)
e
ψ(x) = α cos(ka) cos(kx) for − a < x < a (B) (3.28)
 −κx

e a< x (C)
Here, β in (3.25) is expressed by α and α can now be determined by ensuring a normalized
wave function.

3.4 General considerations on quantum-well like potentials

The potential wells we have considered so far are special due to the step-like behaviour of the
potential. The reason for this choice is the possibility to discuss the results in terms of an
analytical formula. However, also for a rather general shape of the potential well, as shown
in Fig. 3.5 a discussion of the qualitative behaviour of the wave function is possible.
Similarly to the regions A and C in the case of a step-like potential, a classically forbidden
region is characterized by the fact that the potential energy V (x)√is larger than the energy
of the particle E. In this case E − V (x) is negative and only κ = h̄2m |E − V (x)| is a real
p

number. In the same way as in Eq. (3.21) the consequence is that the second derivative of
the wave function has the same sign as the wave function itself: Whenever the wave function
becomes positive, it is convex. Whenever is becomes negative, it is concave. The nodes
(zero-points of wave function) are turning points of the wave function. In order to ensure the
normalizability of the wave function one has further to ensure that the wave function goes to
zero if x → ±∞. This convergence will happen monotonously and no intersection with the
zero line occurs, since it will result in a divergent behaviour.

T. Hickel 27 WS 2019/20
Module 2b-E1 3.4. General considerations on quantum-well like potentials

Figure 3.5: Dependence between value of the wave function and its curvature for regions in
space with classically allowed and forbidden energies, respectively.

Figure 3.6: Demonstration of the discretization of the energy in a general quantum-well like
one-dimensional potential.

The situation is vice versa in the classically

√
allowed region with E − V (x) > 0. In this
2m
p
case the differential equation with k = h̄ E − V (x) is similar to region B in Eq. (3.22).
Therefore, the second derivative has opposite sign as the wave function. A positive wave
function has a concave shape, and a negative wave function is convex. The nodes are again
turning points. Hence, the wave function is always curved such that it will “return” to the
zero axis, with a stronger curvature the larger the value of the wave function is. Consequently,
this results in an oscillating behaviour in the classically allowed region.
The situation of the three regions (classically forbidden, allowed, and again forbidden)
is also depicted in Fig. 3.6. This figure in addition demonstrates another aspect: From the
above discussion of the differential equations (3.21) and (3.22) the curvature of the wave
function is not only determined qualitatively, the potential (entering κ or k) determines also
the quantitative value of the curvature at any point. If we assume that the wave function
has been identified in the classically forbidden region at the left hand side, the potential fixes
the curvature for the x values close to the entrance into the classically allowed region. This
determines the continuation of the wave function a bit away from the entrance region and
automatically the curvature also for these x value. In this way, the wave function is fixed for
all further x values without any further adjustment. However, as visualized in Fig. 3.6, the
procedure does not yield in every case a physically convergence behaviour in the classically
forbidden region on the right hand side. Out of the three depicted scenarios, only case β works
out. The difference between cases α, β, and γ is only the assumed energy E. This determines
mainly the size of the classically allowed region, but also via κ and k the value of the curvature.
As a result, only a few energies give a solution for the Schrödinger equation. Hence, without

T. Hickel 28 WS 2019/20
Module 2b-E1 3.5. Scattering states for the quantum well

complicated mathematically consideration, the energy discretization for bound states can
already be understood from such simple considerations.

3.5 Scattering states for the quantum well

The considerations in Sec. 3.3 were focussed on the situation E < 0, resulting into bound
states for the electrons trapped by the quantum well (3.20)

 0
 x < −a region A
V (x) = −V0 for −a < x < a ⇒ region B (3.29)

 0 x>a region C
However, the situation E > 0 does also yield well defined solutions for the quantum well,
with the only difference that the such described electrons can travel from far left (x → −∞)
to far right (x → +∞) and are not localized next to the well. Still, they are influenced by
the presence of the well, an effect that is called “scattering”.
The general1 solution of the wave function for this problem is:

ik0 x + α e−ik0 x x < −a (A)
 e
 −
ψ(x) = β+ eikx + β− e−ikx for −a< x <a (B) (3.30)
 γ eik00 x

a< x (C)
+

The difference to (3.23) is twofold: (i) Since all regions are classically allowed for E > 0, we
have everywhere plane waves as solutions. (ii) We have set α+ = 1, which is our normalization
condition2 of the “incoming” wave and γ− = 0, which means that we have no wave “coming
in” from +∞.
If one extends this picture of an incoming particle from the left, then the remaining term
in part C describes the continuation of this particle after it has passed the region of the
quantum well. However, not the full particle will make such a “transmission”. The solution
(3.30) contains also a second part of the wave function in part A that is traveling in opposite
direction as compared to the incoming wave. One could speak about a partial reflection of
the wave. According to this picture, we can define a reflection and transmission coefficient as
k00
R = |α− |2
|γ+ |2 T = (3.31)
k0
The physical interpretation requires that R + T = 1. The need for the prefactor in the
transmission coefficient will become clear later. Hence, in order to determine the reflection
and transmission coefficients one only needs to calculate the pre-factors in front of the wave
functions in (3.30).
As usual they are determined from the boundary conditions. We have with y = k/k0 and y 0 = k/k00
0
γ+ eik0 a = β+ eika + β− e−ika


0
0
ik0 a 0 ika −ika (y 0 ± 1)γ+ eik0 a = 2y 0 β± e±ika
γ+ e = y β+ e − β− e
−ik0 a ik0 a
= β+ e−ika + β− eika


e + α− e
e−ik0 a − α− eik0 a = y β+ e−ika − β− eika
Adding the last two equations we obtain:
2e−ik0 a = (1 + y)β+ e−ika + (1 − y)β− eika
−i(k0 +k00 )a
⇒ 4y 0 e = (1 + y)(1 + y 0 )γ+ e−2ika − (1 − y)(1 − y 0 )γ+ e2ika
= γ+ [(1 + yy 0 )(−2i) sin(2ka) + (y + y 0 )2 cos(2ka)]
0 2 h
2 2
i
⇒ 4y 0 e−i(k0 +k0 )a = |γ+ |2 (2(1 + yy 0 ) sin(2ka)) + (2(y + y 0 ) cos(2ka))
−1
|γ+ |2 = 16y 02 4(y + y 0 )2 + 4(1 − y 2 )(1 − y 02 ) sin2 (2ka)

⇒
1
To make it even more general we artificially introduce a k00 with k00 = k0 .
2
Note that a plane wave cannot be normalized to a total probability of 1.

T. Hickel 29 WS 2019/20
Module 2b-E1 3.6. Tunneling

Figure 3.7: Energy dependence of the reflection and transmission coefficients for the scattering
states of a quantum well. [Taken from W. Nolting, Grundkurs Theoretische Physik, vol. 5-1,
p. 271 (Springer, Berlin, 2009)]

Substracting the same two equations we obtain:

2α− e−ik0 a = (1 − y)β+ e−ika + (1 + y)β− eika
−i(k0 +k00 )a
⇒ 4y 0 α− e = (1 + y 0 )(1 − y)γ+ e−2ika − (1 + y)(1 − y 0 )γ+ e2ika
= γ+ [(1 − yy 0 )(−2i) sin(2ka) − (y − y 0 )2 cos(2ka)]
h i
−2ik0 a 2 2 2
⇒ |α− |2 4ye = |γ+ |2 (2(1 − yy 0 ) sin(2ka)) + (2(y − y 0 ) cos(2ka))
16y 02 |α− |2 = |γ+ |2 4(y − y 0 )2 + 4(1 − y 2 )(1 − y 02 ) sin2 (2ka)
 
⇒

Therefore, with k00 = k0 and y 0 = y:

4y 2
T = |γ+ |2 = (3.32)
4y 2 + (1 − y 2 )2 sin2 (2ka)
(1 − y 2 )2 sin2 (2ka)
R = |α− |2 = 2 (3.33)
4y + (1 − y 2 )2 sin2 (2ka)
We can clearly see that R + T = 1 is indeed fulfilled. Furthermore,
√ wep observe that the
1 1
reflection and transmission coefficient depend via k0 = h̄ 2mE, k = h̄ 2m(E + V0 ), and
y = k/k0 on the ratio of the energy E as well as the parameters a and V0 of the potential
well. The dependence is plotted in Fig. 3.7.
We observe an oscillating behaviour, clearly indicating that the “incoming” wave realizes
the presence of the quantum well, even if its energy is above it. Classically, the transmission
coefficient should be always 1, quantum-mechanically this is only the case for very large
energies or if in (3.33) the so-called resonance condition
h̄2 π 2 2
sin(2ka) = 0 =⇒ 2ka = nπ =⇒ En(Res) = −V0 +
n (3.34)
8ma2
is fulfilled. The condition is connected to the particle in the box situation, as becomes obvious
when comparing with Eq. (3.6). In all other cases the reflection coefficient is non-zero and
hence the transmission is reduced. This is the “scattering” effect.

3.6 Tunneling
The concept of reflected and transmitted waves, as discussed in the previous section, becomes
in particular relevant in the context of tunneling, one of the quantum-mechanical phenomena

T. Hickel 30 WS 2019/20
Module 2b-E1 3.6. Tunneling

that cannot be understood classically. To describe this effect more quantitatively, we should
first introduce a new quantity, which is

3.6.1 The probability current

For this purpose we consider the time dependence of the probability:
∂ ∂ ∂ψ ∗ ∂ψ
ρ(x, t) = |ψ(x, t)|2 = ψ + ψ∗ (3.35)
∂t ∂t ! ∂t ∂t !
1 h̄2 ∂ 2ψ∗ 2
∗∂ ψ ∂ h̄ ∗ ∂ψ ∂ψ ∗
  
= − ψ + ψ = − ψ − ψ
ih̄ 2m ∂x2 ∂x2 ∂x 2im ∂x ∂x

where in the second step the Schrödinger equation (2.28) has been applied. If one has the
continuity equation
∂ ∂
ρ(x, t) + j(x, t) = 0. (3.36)
∂t ∂x
as for instance used in electromagnetism, in mind, the following definition of the probability
current is meaningful:
h̄ ∂ψ ∂ψ ∗
 
j(x, t) = ψ∗ − ψ . (3.37)
2im ∂x ∂x
The Eq. (3.36) just formulates mathematically that no probability can “get lost”. This is
because if we consider an interval (a, b) then
Z b Z b
d ∂
dxρ(x, t) = − dx j(x, t) = j(a, t) − j(b, t) (3.38)
dt a a ∂x
means that all probability change within this interval is equal to an effective flow of the
current in or out of the interval.
Given the definition (3.37) one can quickly derive the probability current of a plane wave
h̄k
ψ(x) = αeikx =⇒ j(x) = |α|2 . (3.39)
m
This definition allows a more formal interpretation for reflection and transmission coefficients
than used for the formulation of Eq. (3.31): The incoming wave corresponds to a probability
current j0 = h̄k
m from −∞. The other part of the wave function in region A, i.e. the reflected
0

corresponds to a current jr = −|α− |2 h̄k

m . The wave function in region C can in the same
0

h̄k00
spirit be interpreted as an transmitted wave with current jt = |γ+ |2 m . Accordingly, we can
now define the reflection and transmission coefficient as
jr jt k0
R= = |α− |2 T = = 0 |γ+ |2 , (3.40)
j0 j0 k0
where the physical interpretation requires again R + T = 1.

3.6.2 The potential barrier

For the discussion of the actual tunneling effect we consider the penetration of a particle
through a potential barrier, given by the potential

 0
 x < −a region A
V (x) = V0 for −a < x < a ⇒ region B (3.41)

 0 x>a region C.

The range of considered energies is this time 0 < E < V0 , which classically means that a
particle cannot come from the left to the right side. The quantum-mechanical tunneling is due

T. Hickel 31 WS 2019/20
Module 2b-E1 3.6. Tunneling

Figure 3.8: Potential of a potential barrier.

to our previous studies actually not surprising any more. We learned in the considerations of
Sec. 3.3 that the probability of a particle to exist in a classically forbidden region is not zero,
but just decays exponentially. Which amount of electrons can then actually penetrate through
a barrier just depends on its thickness. The probability becomes exponentially smaller the
larger the region is, through which it penetrates.
The formal solution of the corresponding Schrödinger equation can of course be derived
with a similar set of equations as in the case of scattering states of the quantum well. Much
faster, however, is the transfer of the translation of the previous solution by noting that the
only difference is the region B. This implies that the wave vector k in Eq. (3.30) is now
imaginary:
2m
k 2 = 2 (E − V0 ) = −κ2 =⇒ κ = ik (3.42)
h̄
This further implies that y = k/k0 we have defined for the previous
p solution is also imaginary
and it is more useful to define the real number z = κ/k0 = (V0 − E)/E and replace y by iz
in the solution (3.33) and (3.32). Consequently, we obtain
(1 + z 2 )2 sinh2 (2κa)
R = |α− |2 = (3.43)
4z 2 + (1 + z 2 )2 sinh2 (2κa)
4z 2
T = |γ+ |2 = 2 (3.44)
4z + (1 + z 2 )2 sinh2 (2κa)
The obtained expression for the tunneling probability is non-vanishing in all cases - an effect,
which can classically not be understood.
The limiting case towards low tunneling probabilities, which is relevant for most situations,
is given by
1q 1
κa = 2m(V0 − E)a  1, =⇒ sinh2 2κa ≈ e4κa (3.45)
h̄ 4
for which the transmission coefficient becomes
16k02 κ2 −4κa 16E(V0 − E) 4q
 
T (E) ≈ e = exp − 2m(V0 − E)a (3.46)
(k02 + κ2 )2 V02 h̄
The means that the tunneling probability reduces exponentially with the width 2a of the
barrier and the square root of the effective barrier height V0 − E.
For a proper description of tunneling processes, a rectangular shape of the potential is
not a proper description. Typically the potential V (x) will be a continuous function of the
spatial coordinate x. Using, however, an approximation of the general potential in terms of
several rectangular blocks (which becomes exact if the number of blocks goes to infinity), the
result (3.46) can be generalized. We first simplify the expression further by writing:
4q 16E(V0 − E) 4q
    
T (E) ≈ exp − 2m(V0 − E)a + ln ≈ exp − 2m(V0 − E)a (3.47)
h̄ V02 h̄

T. Hickel 32 WS 2019/20
Module 2b-E1 3.7. The Kronig-Penney model

Figure 3.9: Potential of the Kronig-Penney model. It consists of a periodic repetition of

quantum wells of width a and potential barriers of thickness b.

We further note that the tunneling probability through the combined set of subsequent rect-
angular potential barriers is given by the product of the probabilities to pass a single potential
barrier with thickness ∆x = 2a. We, therefore, obtain:
2q
Y  
T (E) ≈ exp − 2m(Vi − E)∆xi (3.48)
i
h̄
" #
2 Xq
= exp − 2m(Vi − E)∆xi (3.49)
h̄ i
Zb q
 
2
= exp − 2m(V (x) − E)dx . (3.50)
h̄
a
In the last step the limit of very narrow barriers have been considered in order to get the
transition from the sum to the integration. This limit is of course not completely consistent
with the assumption (3.45). Therefore, it should be mentioned here that the expression can
be derived more accurately within the WKB (Wentzel-Kramers-Brillouin) method. However,
for the purpose of this course the presented derivation should be sufficient.

3.7 The Kronig-Penney model

So far we have considered isolated potential wells and barriers and their discretized energy
states. Much more typical for a solid state is, however, a potential which repeats periodically,
due to the symmetric arrangement of the atoms in the crystal lattice. This periodicity gives
rise to a qualitatively new effect: the formation of energy bands, instead of discretized states.
We will again consider a simplified model such that the physical complexity of a real material
is reduced as much as possible, but ensuring that the band formation can still be observed.
As a consequence a rectangular potential of quantum wells with a width a, separated by
barriers of hight V0 and with a thickness b will be studied (see Fig. 3.9). This model is (a
slightly extended version of) the Kronig-Penney model.
We define with l = a + b the length of the periodicity interval. This periodicity should of
course also affect the wave function ψ(x) of the particle. We can directly assume |ψ(x+nl)|2 =
|ψ(x)|2 , since the probability as a physically measurable quantity should have the same
symmetry properties as the potential. Since the wave function is complex and can include a
phase factor this yields:
ψK (x + nl) = eiKnl ψ(x), with − π < Kl < +π. (3.51)
The wave vector K is characteristic for the wave function. Only for K = 0 we have ψ(x+nl) =
ψ(x), but other values of K in the given interval are also allowed. The condition (3.51) is
more generally expressed in solid state physics as

T. Hickel 33 WS 2019/20
Module 2b-E1 3.7. The Kronig-Penney model

3.7.1 The Bloch theorem

This theorem applies (in 3D) to potentials, which show a periodicity (translational symmetry)
of the form V (r+ai ) = V (r) for the three lattice vectors a1 , a1 and a3 . Then, one can provide
a Fourier series expansion of the potential like
X
V (r) = VG eiG·r , (3.52)
G

where the G’s are reciprocal lattice vectors as given by (1.34). We also transfer the wave
function into reciprocal space by writing it in the form of a wave packet
Z
ψ(r) = d3 K φK eiKr , (3.53)

without any constraints3 for K. The Schrödinger equation reads for these definitions
h̄2 K 2
Z Z Z
0
d3 K 0 φK0 eiK r = E
X
d3 K φK eiKr + VG eiG·r d3 K φK eiKr . (3.54)
2m G

If one performs in the second term a shift of variables K0 = K − G and compares the terms
in front of some eiKr on both sides, one obtains:
!
h̄2 K 2 X
− E φK + VG φK−G = 0. (3.55)
2m G

The decisive consequences is that the Schrödinger equation connects in the wave packet (3.53)
only this subset of plane waves, which differ by a reciprocal lattice vector G. Hence, one can
consider instead of (3.53) also the following kind of wave functions
!
X X
i(K+G)r iGr
ψK (r) = φK+G e = φK+G e eiKr = uK (r)eiKr . (3.56)
G G

The such defined function uK (r) automatically fulfils the same symmetry as the potential:
uK (r + ai ) = uK (r). Therefore, the wave function for a periodic potential can always be
expressed as a product of a phase factor and a function, which has the lattice symmetry:
ψK (r) = uK (r)eiKr . This is the Bloch theorem, which directly yields (3.51).

3.7.2 Analytical solution

We are again interested in bound energy states with 0 < E < V0 . With a similar notation as
for the quantum well in (3.21) and (3.22), i.e.
s s
2mE 2m(V0 − E)
k= and κ= (3.57)
h̄2 h̄2
we have the following ansatz for the wave function
(
βn+ eik(x−nl) + βn− e−ik(x−nl) for nl < x < a + nl (Bn )
ψ(x) = (3.58)
γn+ eκ(x−nl+b) + γn− e−κ(x−nl+b) nl − b < x < nl (Cn )

The x-values in the exponents are always chosen such that the interval effectively starts at 0.
This ensures that the phase factor in the Bloch theorem (3.51) directly affects the coefficients
of the wave function:
βn± = eiKnl β0± and γn± = eiKnl γ0± . (3.59)
3
Assuming Born-von-Karman boundary conditions, the P wave function must have the periodicity ψ(r +
N ai ) = ψ(r) for a very large L = |N ai | yielding K = 2π
L
n ai .
i i |ai |

T. Hickel 34 WS 2019/20
Module 2b-E1 3.7. The Kronig-Penney model

Therefore, only the four parameters β0− , β0+ , γ0+ , γ0− remain to be determined.
For this purpose the boundary conditions are of course again used:

ψ(x = 0) = β0+ + β0− = γ0+ eκb + γ0− e−κb

ψ 0 (x = 0) = ikβ0+ − ikβ0− = γ0+ κeκb − γ0− κe−κb
ψ(x = a) = β0+ eika + β0− e−ika = γ1+ + γ1− = γ0+ eiKl + γ0− eiKl
0
ψ (x = a) = ikβ0+ eika − ikβ0− e−ika = κγ1+ − κγ1− = γ0+ κeiKl − γ0− κeiKl

Using matrix notation the following set of equations has to be solved:

−eκb −e−κb
    
1 1 β0+ 0
 ik −ik −κeκb κe−κb    β0−  = 
   0 
 ika −ika iKl iKl  
 (3.60)
 e e −e −e γ 0+
  0 
ika −ika iKl iKl
ike −ike −κe κe γ 0− 0

In order to have a non-trivial solution for the four parameters, the determinant (det A) of this matrix
needs to be zero. The development of the determinant (first row) involves the following determinants:

−ik −κeκb κe−κb ik −ik −κeκb

D1 (k) = e−ika −eiKl −eiKl D3 (κ) = eika e−ika −eiKl
−ike−ika −κeiKl κeiKl ikeika −ike−ika −κeiKl

e−ika −eiKl
D2 (k, κ) = = e−ika eiKl (κ − ik) (3.61)
−ike−ika κeiKl
The following equations are relevant

det A = D1 (k) − D1 (−k) + e−κb D3 (κ) − eκb D3 (−κ)

D1 (k) = 2ikκe2iKl + κeκb D2 (k, κ) + κe−κb D2 (k, −κ)
e−κb D3 (κ) = 2ikκ + ike−κb D2 (k, −κ) + ike−κb D2 (−k, −κ)

These equations can be combined to

det A = eκb {(κ − ik)D2 (k, κ) − (κ + ik)D2 (−k, κ)}

e−κb {(κ + ik)D2 (k, −κ) − (κ − ik)D2 (−k, −κ)} + 4ikκ 1 + e2iKl


+
= eiKl eκb e−ika (κ − ik)2 − eiKl eκb eika (κ + ik)2
− eiKl e−κb e−ika (κ + ik)2 + eiKl e−κb eika (κ − ik)2 + 8ikκeiKl cos(Kl)
 
1 2
= 8ieiKl kκ (cos(Kl) − cos(ka) cosh(κb)) + k − κ2 sin(ka) sinh(κb)


2

Using the condition det A = 0 one obtains

κ2 − k 2
cos(Kl) = cos(ka) cosh(κb) + sin(ka) sinh(κb) (3.62)
2kκ
This equation provides a relation between the choice of K and the energy E of the particle
(included in k and κ). In principle, this relation can also be discussed. However, for our
needs it is sufficient to consider a simpler case in the following, which covers already most of
the relevant physics.

3.7.3 Limit of infinitely narrow barriers

In this limit we consider the width of the barriers to become infinitely small: b → 0. However,
the limit is performed such that bV0 = D = const., which implies for the potential: V0 → ∞.
As a consequence, we have the following simplifications:
2m
l → a, κ2 ∼ V0 → ∞, κ2 b → h̄2
D
√ 2m
κb ∼ V0 b → 0, cosh(κb) → 1, κ sinh(κb) → κ2 b → h̄2
D

T. Hickel 35 WS 2019/20
Module 2b-E1 3.7. The Kronig-Penney model

Figure 3.10: Visualization of the function given in Eq. (3.63) and the allowed and forbidden
regions for q = ka. [Taken from W. Nolting, Grundkurs Theoretische Physik, vol. 5-1, p.
295 (Springer, Berlin, 2009)]

Applying all these simplifications to (3.62) yields an equation that doesn’t contain b and κ
any longer:
mDa
cos(Ka) = cos(ka) + 2 sin(ka) (3.63)
h̄ ka
h̄2 k2
This equation is again a relation between the choice of K and the energy E = 2m . We
discuss it in the following.

1. Due to the properties of the cos function on the left hand side, the equation only has a
solution if the magnitude of the right hand side is smaller than 1. Consequently, there
are allowed and forbidden regions for q = ka. The former are called energy bands, the
latter are called energy gaps and are shaded in Fig. 3.10

2. A forbidden region starts if ka = nπ. This means that the upper edge of each energy
band is given by
h̄2 π 2 2
En = n , n = 1, 2, 3, . . . (3.64)
2ma2
These are exactly the energies (3.6) of a particle in a box. However, in this case not
only a discrete energy is allowed, but due to the different values of K a whole energy
band exists.

3. For the exact energy values En (K) Eq. (3.63) needs to be solved for the corresponding
energy band n. The energy dispersion is shown in Fig. 3.11 (left). In Fig. 3.11 (right)
it is shown that the width of the band gap increases with the product of the potential
barrier D and the lattice constant a. Hence, for large D and a the bands become again
more narrow and approach the discrete energies En . For fixed D and a the width of
the bands increases with the band index n.

4. In the opposite case of small D and a the second term in (3.63) looses importance. For
2 2
D = 0 we would have cos(Ka) = cos(ka) and hence, E h̄2m K
. For finite D this relation
is the better fulfilled the closer one is to the band edge ka = nπ.

T. Hickel 36 WS 2019/20
Module 2b-E1 3.8. Band formation in periodic pontentials

Figure 3.11: Energy dispersion (left) and geometry dependent width of the band gap (right)
for the Kronig-Penney model. [Taken from W. Nolting, Grundkurs Theoretische Physik, vol.
5-1, p. 295 (Springer, Berlin, 2009)]

3.8 Band formation in periodic pontentials

The phenomenon of band formation can also be understood from another perspective. Here,
we consider an almost free particle in one dimension, which is affected by a periodic potential
V (x) = 2A cos(2πx/a). that has only a small amplitude A. Despite this small potential, the
wave functions have to fulfil the Bloch theorem (3.56). This means that they can always be
written in the form
ψK (r) = uK (r)eiKr . (3.65)
Here K is an arbitrary wave vector. The part of the function uK (r) must have the periodicity
of the lattice.
In the limit of a vanishing potential, the wave function needs to be a plane wave eikr .
This can only be fulfilled if uK (r) = eiGr . However, one still has the possibility to choose
the reciprocal lattice vector G. Each choice of G corresponds to a different band. The
energy of the band is for the electron just given by the second derivative of the wave function
h̄2
EG (K) = 2m (K + G)2 . We obtain the following solution for the different bands:

Figure 3.12: Energy dispersion of free electrons.

We can quickly realize that the yellow shaded area repeats periodically. Therefore, this
region has the special name first Brillouin zone (BZ). More generally (3.56), a wave function

T. Hickel 37 WS 2019/20
Module 2b-E1 3.8. Band formation in periodic pontentials

for a periodic potential can always be expressed in the form:

X 0 X 0
ψK (r) = φK+G0 ei(K+G )r = φK+G+G0 ei(K+G+G )r . (3.66)
G0 G0

Therefore, it doesn’t really matter if the origin of the coordinate system is chosen at K = 0
or at K = G. Both points are equivalent, since ψK (r) = ψK+G (r). That is the reason why
it is always sufficient to just consider the first BZ and to consider the images from the other
parts in K space within this interval − 12 G < K < 12 G.
We now want to take care of the potential. Since it is periodic, it needs to be expressed in
reciprocal space according to (3.52). Our special one-dimensional potential has the periodicity
V (x+a) = V (x), accordingly a reciprocal lattice vector has the form Gn = n 2π a . It is therefore
clear that
2π 2π 2π
Vn ein a x = Aei a x + Ae−i a x
X
V (x) = 2A cos(2πx/a) = (3.67)
n
or that V−1 = V1 = A and all the other parts of the potential vanish.
If we now use the Eq. (3.55) we can write
!
h̄2 K 2
− E φK + AφK− 2π + AφK+ 2π = 0. (3.68)
2m a a

Now, at the edge of the first boundary K = 21 G1 = πa . In principle the previous equation
connects φ πa with φ− πa and φ3 πa . As an approximation of this problem, we consider only
a two-component representation and ignore the last term. Hence, the two equations to be
solved are
!
h̄2 π 2
Aφ−π/a + − E φπ/a = 0
2ma2
!
h̄2 π 2
− E φ−π/a + Aφπ/a = 0
2ma2

Nontrivial solutions are obtained for

!2
h̄2 π 2 h̄2 π 2
−E − A2 = 0 =⇒ E= ±A (3.69)
2ma2 2ma2

Hence the potential yields a gap of the size 2A at the edge of the first BZ.
In order to determine also the slope and the curvature, we have to compare with points
that are slightly away from the edge of the first BZ: K1 = πa − ∆K. The second relevant
point in the two-component approximation is in this case K2 = K1 − 2π π
a = − a − ∆K and
the two relevant equations become:
!
h̄2 K12
AφK2 + − E φK 1 = 0
2m
!
h̄2 K22
− E φK2 + AφK1 = 0
2m

Its solution is a bit less straight forward. One first obtains

!
2 h̄2 K12 h̄2 K22 h̄2 K12 h̄2 K22
0 = E −E + + − A2
2m 2m 2m 2m
" #1/2
h̄2  2  h̄4  2 2
E = K1 + K22 ± K1 − K2
2
+ A2
4m 16m2

T. Hickel 38 WS 2019/20
Module 2b-E1 3.8. Band formation in periodic pontentials

! " #1/2
h̄2 π2 2 h̄4 π 2
= + (∆K) ± A 1 + (∆K)2
2m a2 A2 m2 a2
! " #
h̄2 π2 2 h̄4 π 2
≈ + (∆K) ± A 1 + (∆K)2
2m a2 2A2 m2 a2
!
h̄2 π 2 2 h̄
2
2 h̄2 π 2
= ± A + (∆K) 1 ±
2ma2 2m A 2ma2

As a result the slope vanishes and the curvature goes with C (1 ± 2EBZ /A), where C is the
curvature at the point K = 0 and EBZ is the energy at the edge of the first BZ. For example,
if A = 12 EBZ , then the curvature of the upper band is 5C and that lower band is −3C.

Figure 3.13: Modification of the band structure of a free particle due to the presence of a
small periodic potential.

T. Hickel 39 WS 2019/20
Chapter 4

The harmonic oscillator

The one-dimensional harmonic oscillator is one of the most fundamental physical problems
in quantum mechanics. Despite of its simplicity, however, it is relevant for many effects in
physics. The most interesting one from the point of view of materials science is probably
its relation to lattice vibrations. It is well understood that the atoms in a material oscillate
around their equilibrium positions due to the influence of temperature. This typically happens
in a coordinated way, giving rise to phonons (see Sec. 4.4 for more details). However, if one
thinks, for example, about a hydrogen atom in an iron matrix, its motion can (due to the
mass difference) be considered to be independent of motions of the other atoms.
If we call the displacement of this atom from the equilibrium position x, then the position
dependent force F (x) driving the atom back to the origin is in a very good approximation
proportional to x, i.e., F (x) = −kx. The potential corresponding to this situation is due
d
to F (x) = − dx V (x) given by V (x) = 12 kx2 . Hence, we have a parabolic potential.
p Ex-
pressing the spring constant k in terms of the frequency of the oscillation ω = k/m, the
corresponding Hamiltonian is given by

p2 1
H= + mω 2 x2 . (4.1)
2m 2
Despite the simplicity of the problem, its solution is non-trivial. Since a lot about typical
analytical formalisms in quantum-mechanics can be learned in this way, we provide here two
solutions.

4.1 Sommerfeld’s polynomial method

In the first approach we consider the time-independent Schrödinger equation belonging the
Hamiltonian in (4.1) as a differential equation for the wave-function. Hence, we actually solve
the following problem:

h̄2 d2 1
 
ψ(x) + E − mω 2 x2 ψ(x) = 0. (4.2)
2m dx2 2
We first make some modifications of this equation, which are just transformations of the
problem.
First of all, dimensionless parameters are introduced to simplify the notation1 . We obtain:

d2
r

2
 2E ωm
ψ(y) + η − y ψ(y) = 0 for η= and y = x (4.3)
dy 2 h̄ω h̄
1
In order to remove any factor in front of x, the pre-factors first need to be symmetrized. This is only
possible with the prefactor ωm
h̄
at both positions. The structure of the new dimensionless variables is a logical
consequence.

40
Module 2b-E1 4.1. Sommerfeld’s polynomial method

Secondly, some considerations about the behaviour of the wave function for large distances
(y → ±∞) should be made. We need to make sure that the wave function does not diverge
in this limit, since this hinders any attempt to normalize the total wave function to 1. In this
limit, we can definitely ignore the energy term η obtaining a simpler differential equation:
" #
d2 y2
ψ̃(y) − y 2 ψ̃(y) = 0 =⇒ ψ̃(y) ∼ exp − . (4.4)
dy 2 2

The solution on the right hand side is again only correct in the limit y → ±∞. Nevertheless,
these considerations make the following ansatz for the wave function reasonable:
" #
y2 X
ψ(y) = ν(y) exp − with ν(y) = αk y k . (4.5)
2 k

The polynomial ansatz for the second part of the wave function explains the name of the
whole method.
For further considerations, it is first necessary to express the previous differential equation
(4.3) now with (4.5) in terms of the polynomial part ν(y):
" #
d2 d
2
− 2y + (η − 1) ν(y) = 0 (4.6)
dy dy
X
=⇒ {αk+2 (k + 2)(k + 1) + αk (η − 1 − 2k)} y k = 0 (4.7)
k

In particular for the second step some straightforward transformations are necessary, which
we skip here for space reasons. Since a polynomial can only vanish identically, if its coefficients
are zero, we can directly conclude from (4.7) the following recursion relation for the coefficients
of our polynomial ν(y):
2k + 1 − η
αk+2 = αk . (4.8)
(k + 2)(k + 1)
The important message is now that the ansatz (4.5) doesn’t help to ensure a correct
behaviour for the wave function at y → ±∞ as long as the polynomial νk is not truncated.
In fact one can proof that the recursion relation (4.8) yields an increase of the polynomial
ν(y), which is as strong as the decrease of the exponential function in (4.5). The, therefore,
necessary truncation of the polynomial implies that one of the numerators in (4.8) needs to
be zero! This condition is the most important result of the derivation, since it is responsible
for the quantization of the allowed energy levels. One obtains:
1
 
η = 2k + 1 −→ En = h̄ω n + (4.9)
2
The result tells us that the energies of the harmonic oscillator are equidistant, with an energy
difference between two neighbouring states being always h̄ω. We can also see that the lowest
energy state of the harmonic oscillator is not 0 - as one would expect from classical physics -,
but 21 h̄ω. This result can also be understood in terms of the Heisenberg uncertainty principle.
It gives rise to effects like zero point vibrations in lattice vibrations.
With the derivation above also the complete information about the wave function is given.
We should add the information that the polynomials ν(y) are (apart from some normalization
factor) called Hermite polynomials and are well studied objects in mathematics. However,
we don’t want to spent here too much time to go into its properties, but rather would like to
present also an alternative solution of the same problem next.

T. Hickel 41 WS 2019/20
Module 2b-E1 4.2. Observables and operators

4.2 Observables and operators

Within the second approach to solve the harmonic oscillator, we make use of properties of
the involved operators. This needs first some background information about the nature of
operators in quantum mechanics and the relation to observables.

4.2.1 Eigenstates and mixed states

For the Hamilton operator we noticed before that it is related to a (typically quantized) energy
En of the system. This becomes particularly clear, if the time-independent Schrödinger
equation (2.33) is considered. It tells us that there is a certain wave function ψn (x) that
belongs to this energy. To make the connection between energy and wave function more
apparent, we denote the latter as ψn (x) ≡ |En i and formulate instead of (2.33):

H|En i = En |En i. (4.10)

This equation has the character of an eigenvalue equation, i.e., the application of H to the
state |En i reproduces this state and provides the corresponding energy as an eigenvalue.
However, even if we are confronted with a state |ψi that is not an eigenstate, it still a
justified to ask what the application of H to this state gives. In principle, one can always
expand the state |ψi in terms of eigenfunctions |En i of H:
X X
|ψi = αi |Ei i and obtains H|ψi = αi Ei |Ei i. (4.11)
i i

This means that |ψi is in a mixed state and the expansion is similar to a Fourier expansion.
If one defines in addition a scalar product of such states in terms of the following integral
Z∞
hξ | ψi = dx ξ ∗ (x)ψ(x) (4.12)
−∞

then the expectation value of H in this state is related to the probabilities that |ψi is one of
the eigenstates:
Z∞
dx ψ ∗ (x)Hψ(x) = hψ|H|ψi =
X
Ei |αi |2 (4.13)
−∞ i

Hence, the measurement of H lifts the uncertainty in which eigenstate the mixed state |ψi
really is. The operator H is on observable measuring the energy of the state.
Other observables, which are relevant for the harmonic oscillator are the position operator
∂
x̂ and the momentum operator p̂ = −ih̄ ∂x . For both operators a similar discussion as for H
can be done. Note in this context, that the eigenstates of x̂ and p̂ are not identical with the
eigenstates of H, since these operators do not commute. Hence, resolving the uncertainty
with respect to the eigenstates of H does not provide any information, for example, about
the position of the system.

4.2.2 Adjoint operators

For the upcoming considerations of the harmonic oscillator, it is now important to introduce
the concept of an adjoint2 operator A† to an operator A. It is defined for those states |ξi, for
¯ exists such that
which a state |ξi
¯
hξ|A|ψi = hξ|ψi defined as A† |ξi = |ξi
¯ (4.14)
2
often also called “Hermitian conjugate”

T. Hickel 42 WS 2019/20
Module 2b-E1 4.3. Solving the harmonic oscillator with second quantization

The definition is perhaps more clear with the integral notation of the scalar product:
Z∞ Z∞
∗ †
dxξ (x)A ψ(x) = dx (Aξ(x))∗ ψ(x) (4.15)
−∞ −∞

indicating that A† acts on the left state in the scalar product.

In case of operators that are observables, i.e. that measure a physical quantity, the eigen
and expectation values need to be real numbers. It can be easily shown that this is equivalent
to the statement that A† = A. Operators with such a property are also called Hermitian or
self-adjoint operators.
A lot of properties for adjoint and hermitian operators can be proven. One important
property is for example that the eigenstates of an hermitian operator are always orthogonal.
However, since this module is not intended to give a deep insight into theoretical physics,
we skip these general considerations. Instead we only discuss the applications necessary for
understanding solid state physics.

4.3 Solving the harmonic oscillator with second quantization

4.3.1 Introduction of new operators
We now present the second solution for the harmonic oscillator which is based on operator
algebra. Note again that the operator of the harmonic oscillator

p̂2 1
H= + mω 2 x̂2 (4.16)
2m 2
contains with x̂ and p̂ two operators, which fulfil the commutation relation [x̂, p̂]− = ih̄. The
idea of our new approach is, to consider instead a single operator â and its adjoint operator
aˆ+ , such that their commutation relation is similarly given by

[â, â+ ]− = 1. (4.17)

This new operator should be a linear combination x̂ and p̂ with two complex coefficients
c1 and c2 :
â = c1 x̂ + c2 p̂ =⇒ â+ = c∗1 x̂ + c∗2 p̂. (4.18)
Using (4.18) in (4.17) yields a first condition for the coefficients:

1 = ih̄(c1 c∗2 − c2 c∗1 ) (4.19)

where also [x̂, p̂]− = ih̄ has been used.

A second condition for the coefficients is given by the desire to have a resulting Hamilto-
nian in terms of â and â+ that is as simple as possible. For this purpose, we first express x̂
and p̂ in terms of the new operators. We do this by subtracting the two equations in (4.18)
such that the operator that we don’t need drops from the equations:3

c∗2 â − c2 â+
= ih̄ c∗2 â − c2 â+


x̂ = (4.20)
c∗2 c1 − c2 c∗1
c∗1 â − c1 â+
= −ih̄ c∗1 â − c1 â+


p̂ = ∗ ∗ (4.21)
c1 c2 − c1 c2
3
This requires a multiplication with the corresponding c-coefficients. The resulting prefactor in front of the
remaining operators is then written as a denominator.

T. Hickel 43 WS 2019/20
Module 2b-E1 4.3. Solving the harmonic oscillator with second quantization

In both cases the first condition (4.19) has been used for the second transformation. The
resulting Hamiltonian looks now like this:

h̄2 h ∗ 2 2 i
H = − c1 â + c21 (â+ )2 − |c1 |2 (2â+ â + 1)
2m
h̄2 mω 2 h ∗ 2 2 i
− c2 â + c22 (â+ )2 − |c2 |2 (2â+ â + 1) (4.22)
2
where the commutation relation (4.17) has been used in both terms. When we use as the
second condition for the coefficients:
1 2
c + mω 2 c22 = 0 (4.23)
m 1
then we ensure that the quadratic operator terms â2 and (â+ )2 vanish in the Hamiltonian.
Due the squares in (4.23), this condition can only be fulfilled by complex numbers. The
easiest choice is to assume that c1 is completely real and c2 is completely imaginary:
r
mω i
c1 = c2 = √ (4.24)
2h̄ 2h̄mω
With this choice the transformation equations (4.18) as well as (4.20) and (4.21) are:

1 √ p̂ 1 √ p̂
   
â = √ mωx̂ + i √ â+ = √ mωx̂ − i √ (4.25)
2h̄ mω 2h̄ mω
s s
h̄ h̄mω
x̂ = (â + â+ ) p̂ = −i (â − â+ ) (4.26)
2mω 2

The Hamiltonian becomes then particularly simple:

1
 
+
H = h̄ω â â + . (4.27)
2

4.3.2 The occupation number operator

The simplified Hamiltonian (4.27) contains in principle only one operator, defined as n̂ = â+ â.
This operator is called occupation number operator. The eigenvalue equation of this hermitian
operator is given as
n̂|ni = n|ni (4.28)
where n is first of all some real number. However, one can easily proof a set of properties of
the eigenvalues:

1. All eigenvalue on n̂ are not negative.

2. If |ni is an eigenstate than also â|ni and â+ |ni are eigenstates of n̂ with the eigenvalues
n − 1 and n + 1, respectively.

3. The eigenvalues of n̂ are not degenerate.

4. The smallest eigenvalue of n̂ is nmin = 0.

5. The spectrum of eigenvalues of n̂ has no upper limit.

6. There are no non-integer eigenvalues of n̂.

T. Hickel 44 WS 2019/20
Module 2b-E1 4.3. Solving the harmonic oscillator with second quantization

These properties can be proven step by step. We limit ourselves here to the proof of the
second property. It makes use of commutation relations involving the new operator n̂:

[n̂, â]− = [â+ , â]− â + â+ [â, â]− = −â (4.29)

+ + + + + +
[n̂, â ]− = [â , â ]− â + â [â, â ]− = â (4.30)
=⇒ n̂(â|ni) = [n̂, â]− |ni + ân̂|ni = −â|ni + ân|ni = (n − 1)(â|ni)
n̂(â+ |ni) = [n̂, â+ ]− |ni + â+ n̂|ni = â+ |ni + â+ n|ni = (n + 1)(â+ |ni)

The equations make clear, why â and â+ are usually called annihilation and creation operator,
respectively. We consider n̂ to count the number of energy quanta of the system and â / â+
yields to a state with one energy quantum less / more.
The properties for the eigenvalues of the operator n̂ can be used to make the following
statement on the eigenvalues of the harmonic oscillator:
1
 
En = h̄ω n + with n = 0, 1, 2, . . . (4.31)
2

This is the same result as obtained in (4.9) within Sommerfeld’s polynomial method.

4.3.3 Derivation of wave functions

To conclude the considerations of the second-quantization approach, we should also state
how the eigenfunctions of the Hamiltonian are obtained within this approach and that they
are identical with those of the q
previous approach. For this purpose we first use the same
dimensionless length unit y = mω h̄ ∂
h̄ x as in (4.3). Reminding in addition that p̂ = i ∂x we
obtain from (4.25) and (4.25) the following operator notations:

1 ∂ 1 ∂
   
â = √ y + â+ = √ y − . (4.32)
2 ∂y 2 ∂y

Now, we first define the ground state of the system |0i as the state which contains no
energy quanta. Accordingly, the application of the annihilation operator â to this state should
give zero. This provides direct access to the ground state wave function:
" #
∂ y2
 
â|0i = 0 =⇒ y+ ψ0 (x) = 0 =⇒ ψ0 (x) = c0 exp − (4.33)
∂y 2

with some normalization constant c0 .

The other eigenfunctions are obtained by applying a certain power of the creation operator
to the ground state. For this purpose we should add the information about the correct
prefactor: √
â+ |ni = cn |n + 1i with cn = n + 1 (4.34)
The value of the prefactor is determined by making use of the normalization condition of all
eigenstates |ni. The norm is always obtained by the scalar product of a state with itself. We
get from (4.34):

|cn |2 hn + 1|n + 1i = hn|ââ+ |ni = hn|â+ â + 1|ni = (n + 1)hn|ni (4.35)

where the commutation relation (4.17) has also been used. The rest follows from hn|ni =
hn + 1|n + 1i = 1. The relation (4.34) is directly extended to:
n
1 1 ∂


n
|ni = √ â+ |0i or similarly ψn (x) = √ y− ψ0 (x). (4.36)
n! n
2 n! ∂y

T. Hickel 45 WS 2019/20
Module 2b-E1 4.4. Phonons and harmonic oscillator

This provides already the full construction rule to determine all eigenfunctions of the harmonic
oscillator.
In order to make the agreement with the result from the Sommerfeld’s polynomial method
more clear, we introduce here similarly to the seperation ansatz (4.5) the Hermite polynomials
by writing:
" #
y2
r
mω 1
ψn (y) = 4
√ exp − Hn (y) with (4.37)
h̄π 2n n! 2
" # n " #
y2 ∂ y2
Hn (y) = exp y− exp − (4.38)
2 ∂y 2

As a consequence, the addition of the two equations (both similar ot (4.34))

1

∂

√
â+ ψn (y) = √ y− ψn (y) = n + 1ψn+1 (y) (4.39)
2 ∂y
1 ∂ √
 
âψn (y) = √ y+ ψn (y) = nψn−1 (y) (4.40)
2 ∂y

becomes
√ √ √
2yψn (y) = n + 1ψn+1 (y) + nψn−1 (y) (4.41)
2yHn (y) = Hn+1 (y) + 2nHn−1 (y) (4.42)

We can also use (4.40) directly to obtain

∂
Hn (y) = 2nHn−1 (y) (4.43)
∂y

and the subsequent application of (4.43), (4.42), and again (4.43) yields:

∂2 ∂ ∂ ∂
2
Hn (y) = 2n Hn−1 (y) = 2Hn (y) + 2y Hn (y) − Hn+1 (y) (4.44)
∂y ∂y ∂y ∂y
∂
= 2y Hn (y) − 2nHn (y) (4.45)
∂y

This is exactly the differential equation (4.6) we have previously derived for the Sommerfeld
polynomial method, which completes the proof of the equivalence of both approaches.

4.4 Phonons and harmonic oscillator

As mentioned in the introduction of the chapter, lattice vibrations are the most relevant
application of the harmonic oscillator in solid state physics. The quantization unit is in this
case called “phonon”. A phonon with a certain wave vector k corresponds to a concerted
wave-like motion of atoms in a crystal. It can be generally shown that lattice vibrations are
in harmonic approximation always a superposition of these wave-like motions. We will first
explain this concept for simplified one-dimensional Hamiltonians, before coming to a more
complete description as they are used in materials science.

4.4.1 One-dim. vibrations with mono-atomic basis

A harmonic oscillation of an atom s implies that a displacement us from its equilibrium
position gives rise to a back-driving force Fs that is proportional to this displacement. More
precisely, it will actually depend on the distance to all other atoms, taking into account that

T. Hickel 46 WS 2019/20
Module 2b-E1 4.4. Phonons and harmonic oscillator

a crystal is characterized by a strong binding between the atoms. Reducing this interaction
only to nearest neighbour atoms in one dimension, this relation is given by
d2 us
Fs = M = C (us+1 − us ) + C (us−1 − us ) . (4.46)
dt2
Here, the constant C is the spring constant of the atomic motion. Furthermore, it has been
used that according to Newton’s law, the force is proportional to the acceleration of the atom
with the mass M as proportionality constant.
Similar to the electronic problem, we solve also in this case the linear differential equation
(4.46) by a separation ansatz:
u(rs , t) = us (t) = ūs eiωt = ūeiωt eisaK . (4.47)
Here, the position of the rs = s·a is given as multiplies of the lattice constant a. This position
needs to be distinguished from its displacement out of the equilibrium position, which is only
in the case of longitudional vibrations happening in the direction of the considered atomic
chain. Furthermore, Eq. (4.47) introduces the angular frequency ω and the wave vector K.
First, making only use of the time dependence of the wave function, the set of coupled
differential equations resulting from (4.46):
−M ω 2 ūs = C (ūs−1 − 2ūs + ūs+1 ) (4.48)
can be written as a tri-diagonal matrix equation:
.. .. ..
    
. . .
2− M
 2
   

 −1 Cω −1 0 0 0   ūs−1

 
  0 

 ···
 0 −1 2− M
Cω
2
−1 0 0 ···   ūs

=
  0 
 (4.49)

 0 0 −1 2− M
Cω
2
−1 0   ūs+1

 
  0 

.. .. ..
. . .
Second, using the full solution (4.47) one ensures that the matrix equation has a non-trivial
solution, since then all the individual equations become equivalent (independent of s):
M 2 Ka
ω = −e−iKa + 2 − e+iKa = 2 (1 − cos Ka) = 4 sin2 (4.50)
C 2
This means one obtains an energy dispersion of
s
4C 1
ω(K) = sin Ka . (4.51)
M 2
It is clear that this dispersion has the periodicity of the sin function. This means if one
considers instead of K the wave vector K 0 = K + G with a reciprocal lattice vector G = 2π a
one gets the same energy value. This periodicity can also be seen in the wave functions. The
ansatz (4.47) with K and K 0 = K + 2π a needs to be already equivalent, because in both cases
iKa
us+1 /us = e . This equivalence can also be seen if comparing the wave length in both
cases: With λK being the wave length belonging to the wave vector K we have:
n
λK = na =⇒ λK 0 = a. (4.52)
n+1
This means for wave K the atoms at position A and A + na vibrate in phase. For the wave
K 0 in principle also all points with distance m n+1 n
a from A are also in phase. However, in
most of the cases there are no atoms at these points. Therefore, also in this case the first
atom that is in phase with atom A is the atom at the position na. Therefore, the waves K
and K 0 finally describe the same wave.
This symmetry is the reason why one concentrates only on those wave vectors that lie
in the interval − πa , πa . This interval is called the first Brillouin zone of a one-dimensional


lattice.

T. Hickel 47 WS 2019/20
Module 2b-E1 4.4. Phonons and harmonic oscillator

4.4.2 One-dim. vibrations with two-atomic basis

If the crystal structure has a two-atom basis then the assumption that all atoms vibrate with
the same amplitude ū – as used for the separation ansatz (4.47) – is not justified anymore.
Instead we have to change (4.46) to a set of two differential equations (assuming atom vs is
sitting right to atom us ):

d2 us
M1 = C (vs − us ) + C (vs−1 − us ) (4.53)
dt2
d2 vs
M2 2 = C (us+1 − vs ) + C (us − vs ) (4.54)
dt
The generalization for the wave function (4.47) becomes in this case

us (t) = ūeiωt eisaK and vs (t) = v̄eiωt eisaK . (4.55)

The matrix equation (4.49) contains then alternating masses M1 and M2 and reduces for the
full solution (4.55) to the following set of equations:
   
M1 2
− 1 + e−iKa
! !
2− C ω  ū 0
 
M2 2
 = . (4.56)
− 1 + e+iKa 2− v̄ 0
C ω

A non-trivial solution is only obtained, if the determinant of the coefficient matrix van-
ishes, yielding the equation:

0 = M1 M2 ω 4 − 2C(M1 + M2 )ω 2 + 2C 2 (1 − cos Ka) or (4.57)

s 2
1 1 1 1 2
 
ω2 = C + ±C + − (1 − cos Ka) (4.58)
M1 M2 M 1 M2 M1 M2
In the case of very small Ka this yields
1 1 1 1 1 1
   
2
ω = C + ±C + ∓C (Ka)2 yielding
M1 M 2 M1 M2 M1 + M2 2
1 1
 
ω2 = 2C + (optical branch) or (4.59)
M1 M2
C 1
ω2 = K 2 a2 (acoustic branch) (4.60)
2 M1 + M2
whereas for the edge of the Brillouin zone K = ± πa one obtains
s 2
1 1 1 1 4
 
2
ω = C + ±C + − yielding
M1 M2 M1 M2 M1 M 2
2C
ω2 = (optical branch if M2 > M1 ) (4.61)
M1
2C
ω2 = (acoustic branch if M2 > M1 ) (4.62)
M2
Entering (4.59) into (4.56) one notices that for the optical phonons ū/v̄ = −M2 /M1 , which
means that these atoms vibrate against each others. In case of two oppositely charged atoms,
such a phonon mode easily interacts with the electric field of a light wave. Therefore, this
phonon branch is called “optical”. On the other hand, if one enters (4.60) into (4.56) one
obtains ū = v̄, the characteristic of an acoustic wave.
Furthermore, one can convince oneself graphically that for M1 = M2 , the solution for
the two-atom basis becomes identical with that of the mono-atomic basis if the size of the
Brillouin zone is adjusted accordingly.

T. Hickel 48 WS 2019/20
Module 2b-E1 4.4. Phonons and harmonic oscillator

4.4.3 Generalization to phonon dispersion of crystals

In the treatment of real crystal structures it is of course not sufficient to limit the considera-
tions to nearest-neighbour interactions in one dimension only. In this case the force on single
atoms do actually depend on the position of all other atoms. The corresponding equation of
motion – generalizing Eq. (4.46) – is in this case:

d2 ui X
Fi = Mi 2
=− Wij uj , (4.63)
dt j

where the sum is performed not only over all atoms, but at the same time over all three
lattice dimensions. The matrix Wij denotes the Hesse matrix of the system. It is defined as
" # " #
∂ ∂2
Wij = − Fi ({RI }) = E BOS ({RI }) (4.64)
∂Rj {R0I }
∂Ri ∂Rj 0 {R0I }

From the second definition it is clear that the Hesse matrix actually stems from a Taylor
expansion of the so-called Born-Oppenheimer energy surface of the atomic motion
1
E0BOS ({RI }) = E0BOS ({R0I }) + ui Wij uj + . . . (4.65)
2
The linear term in this Taylor expansion disappears, if the ground state E0 of this surface is
considered. The Hesse matrix itself is closely related to the dynamical matrix of the system,
which is given by:
1
Dij = p Wij . (4.66)
M i Mj
The solution of the differential equation (4.63) is again performed via the same kind of
separation ansatz as in the one-dimensional case before
1
ui (t) = √ ūl (q)eiqRn eiωt (4.67)
Ml
To be precise, the indes i stands here for a combination of the index n for a unit cell within
the supercell and an index l counting the atom and direction within the unit cell.
Entering the ansatz into (4.63) one obtains the equation
1
Dnl,n0 l0 eiq(Rn0 −Rn ) ūl0 (q)
XX
−ω 2 ūl (q) = − (4.68)
l0 n0
M l Ml 0

where j similarly to i has the parts n0 and l0 . Introducing the Fourier transformation of the
dynamical matrix one obtains the following eigenvalue equation:
(x) X (x)
−ωx2 ūl (q) = − D̃l,l0 (q)ūl0 (q). (4.69)
l0

Hence, if displacements along the eigenvectors ū(x) (q) are performed, then the system oscil-
lates with the phonon frequency ωx (q). In order to determine the phonon spectrum, the only
task to do is, therefore, to determine all eigenvalues of the Fourier transformed dynamical
matrix. This is similar to solving matrix equations as (4.49) above.

T. Hickel 49 WS 2019/20
Module 2b-E1 4.4. Phonons and harmonic oscillator

4.4.4 Quantization of phonons

In order to understand now the quantization of phonons, one should first note the relation of
an equation of motion such as (4.46) or (4.63) to the Hamilton operator. We make this here by
using the relation that has already been introduced at the beginning of this chapter, namely
that the linear force of the harmonic oscillation is due to a quadratic potential. Therefore, it
is clear that the Hamiltonian (in the one-dimensional case with mono-atomic basis) should
actually have the form
X 1 1

2 2
H= p̂ + C(q̂s+1 − q̂s ) . (4.70)
s 2M s 2
Inspired by the solution for the displacements (4.47) we transform the position operators
q̂s according to

q̂s = N −1/2
X
Q̂K eisaK with (4.71)
K
= N −1/2 q̂s e−isaK
X
Q̂K (4.72)
s

We use here Born-von-Karman boundary conditions, which means that the wave vectors are
discrete according to K = 2π Nna for some large N . In parallel, the momentum ps should also
be transformed. This needs to be done in such a way, that PK and QK are afterwards canon-
ical variables, i.e. fulfill the commutation relations of position and momentum operators.
This is fulfilled if

p̂s = N −1/2 P̂K e−isaK

X
with (4.73)
K
= N −1/2
X
P̂K p̂s eisaK (4.74)
s

is chosen.
We test the commutation relation of the new operators:
0
[Q̂K , P̂K 0 ]− = N −1 [q̂r , p̂s ]− e−i(Kr−K s)a
XX
(4.75)
r s
−1 −i(K−K 0 )ra 0
= ih̄N −1 e−i2π(n−n )r/N
X X
= ih̄N e (4.76)
r r
= ih̄δK,K 0 (4.77)
Furthermore, we insert the transformations into the Hamiltonian:
X XXX 0
p̂2s = N −1 P̂K P̂K 0 e−i(K+K )sa

s s K K0
XX X
= P̂K P̂K 0 δ−K,K 0 = P̂K P̂−K
K K0 K
X XXX 0
 0

(q̂s+1 − q̂s )2 = N −1 Q̂K Q̂K 0 e−iKsa e−iKsa − 1 e−iK sa e−iK sa − 1


s s K K0
X
= 2 Q̂K Q̂−K (1 − cos Ka)
K

Using the definition s

2C √
ωK = 1 − cos Ka (4.78)
M
one there obtains the Hamiltonian
X 1

2
H= P̂K P̂−K + M ωK Q̂K Q̂−K (4.79)
K
2

T. Hickel 50 WS 2019/20
Module 2b-E1 4.4. Phonons and harmonic oscillator

which has almost exactly the form of the harmonic oscillator.

From here, one can also get back to the equation of motion, using the general notion that
the time dependence of an operator is given by
d h i
ih̄ Â(t) = Â, H . (4.80)
dt −

Hence,
d 1
ih̄ Q̂K = [Q̂K , H] = ih̄ P̂−K (4.81)
dt M
d2 d 1 2
ih̄ 2 Q̂K = [ Q̂K , H] = ih̄ [P̂−K , H] = ih̄ωK Q̂K (4.82)
dt dt M
what corresponds to the equation of motion.
From the form of the Hamiltonian in (4.79) one can immediately conclude, that the
eigenvalues are given by
1
 
K = h̄ωK + nK . (4.83)
2

T. Hickel 51 WS 2019/20
Chapter 5

The hydrogen atom

So far we have only considered one-dimensional problems. This helped mainly to understand
phenomena like energy quantization and tunneling, relevant for solid-state physics. However,
realistic materials can typically only be considered as a three-dimensional system. One of
the main tasks in solid-state physics is the derivation of the energy related to a system of
atoms that are situated on a 3dim crystal lattice. Moreover, even individual atoms are 3dim
systems. It is known that the eigenstates of individual atoms can be associated with certain
orbitals. It might be surprising, though, that the exact wave function of the electrons, i.e.,
the exact eigenstates of the corresponding Hamiltonian, cannot be derived accurately. This
is mainly due to the complexity related to the electron-electron interaction, being present
even in a single atom. The only exception is the hydrogen atom, which only includes a single
electron. For this situation an analytical solution exists. Although this solution is not trivial
we should at least sketch it here, since a lot about the structure of atoms and the periodic
table of elements can be learned from this derivation.

5.1 The Schrödinger equation for a central potential

Most generally the Hamiltonian of the H atom is the Hamiltonian of a single electron in a 3d
potential, hence !
p2
Hψ(r) = + V (r) ψ(r) = Eψ(r) (5.1)
2µ
Here, the momentum operator is a three-dimensional object that can be written as1
!
h̄ ∂ ∂ ∂ ∂2 ∂2 ∂2
 
2 2
p= , , =⇒ p = −h̄ + + = −h̄2 ∆xyz (5.2)
i ∂x ∂y ∂z ∂x2 ∂y 2 ∂z 2
Nevertheless, the situation simplifies for the hydrogen atom such that the (Coulomb)
potential is a central potential:
Ze2 1
V (r) = V (r) = − . (5.3)
4πε0 r
Hence, the potential does not anymore depend on three variables x, y, z, but on a single
variable r only. To make use of this symmetry, it is highly recommended to use the right
coordinate system, namely that of spherical coordinates:
   
x r sin θ cos φ
 y  =  r sin θ sin φ  (5.4)
   
z r cos θ
1
The mass is written with the symbol µ, here, since m is reserved for the z-component of the angular
momentum (see below).

52
Module 2b-E1 5.2. The angular momentum operator

The potential becomes simple in these coordinates, the problem is the kinetic energy part of
the Hamiltonian. One can derive that the Laplace operator ∆ looks in spherical coordinates
like this:
1 1
∆rθφ = − 2 p2r + ∆θφ or p2 = p2r + 2 L2 (5.5)
h̄ r
Here, pr is the radial momentum, which is defined as
!
h̄ ∂ 1 ∂2 2 ∂ 1 ∂ ∂
   
pr = + =⇒ p2r = −h̄ 2
+ = −h̄2 r2 (5.6)
i ∂r r ∂r2 r ∂r r2 ∂r ∂r

Further, ∆θφ is an operator which acts only on the θ and φ parameters and L is the angular
momentum operator. The latter is defined by the cross product of position and momentum:
 
ypz − zpy !
2 2 ∂2 ∂ 1 ∂2
L = r × p =  zpx − xpz  =⇒ L = −h̄ + cot θ + (5.7)
 
∂θ2 ∂θ sin2 θ ∂φ2
xpy − ypx

and describes the rotation of a particle2

The Schrödinger equation (5.1) can therefore be written as
" ! #
h̄2 ∂2 2 ∂ L2
Hψ(r) = − + + + V (r) ψ(r) = Eψ(r) (5.8)
2µ ∂r2 r ∂r 2µr2

It is important to note that due to the structure of L, it commutes with the Hamiltonian of
the system:
[H, L]− = 0 (5.9)
This means that the eigenfunctions of H are also eigenfunctions of L. Before solving the
full hydrogen problem, we should therefore first think about some properties of the angular
momentum operator L.

5.2 The angular momentum operator

Similarly to the harmonic oscillators discussed before, there are two ways to deal with the
angular momentum operator: Either one applies the included derivatives to the wave function
and solves the resulting differential equation. Or one makes use of some operator properties
to simplify the equations such that the solution can at the end be given directly. We use the
second option.
Instead of considering a three-dimensional operator L it might be helpful to consider just
the components Lx , Ly and Lz given in (5.7). A common set of eigenfunctions can only be
expected, if these components commute with each other. This is not the case:

[Ly , Lz ]− = ih̄Lx 

[Lz , Lx ]− = ih̄Ly =⇒ L × L = ih̄L (5.10)

[Lx , Ly ]− = ih̄Lz 

However, it is an important consequence of these commutation relations that

[Lz , L2 ]− = [Lz , L2x ]− + [Lz , L2y ]− = 0. (5.11)

2
It is only constant for a constant rotation and vanishes if the particle moves straight through the centre
of the coordinate system.

T. Hickel 53 WS 2019/20
Module 2b-E1 5.2. The angular momentum operator

This relation holds of course also for any other component of L, but we consider here only
Lz . The consequence is that Lz , L2 and also any Hamiltonian with central potential (like for
the hydrogen atom) have a common set of eigenfunctions.3
We denote these eigenfunctions as Yl,m (θ, φ) = |l, mi, where l and m stand for the eigen-
values of angular momentum operator:

L2 |l, mi = h̄2 l(l + 1)|l, mi (5.12)

Lz |l, mi = h̄m|l, mi (5.13)

We do this first without any statement on what the wave function or its eigenvalues really
are. The only thing we can already clarify now is that due to hl, m|L2x |l, mi ≥ 0 and similarly
for the other components, we can also assume hl, m|L2 |l, mi ≥ 0 and hence l ≥ 0. 4

5.2.1 Raising and lowering operators

We define the new operators
L± = Lx ± iLy (5.14)
for which the following set of properties can be proven

[L+ , L− ]− = 2h̄Lz (5.15)

[Lz , L± ]− = ±h̄L± (5.16)
h i
L2 , L± = 0 (5.17)
−
2
L = L+ L− + L2z − h̄Lz (5.18)

We note that not only |l, mi is an eigenvalue of L2 and Lz , but also L± |l, mi:

L2 L± |l, mi = L± L2 |l, mi = h̄2 l(l + 1)L± |l, mi (5.19)

Lz L± |l, mi = (L± Lz ± h̄L± )|l, mi = h̄(m ± 1)L± |l, mi (5.20)

The second relation give L± the name as raising and lowering operators, respectively.
To see more clearly how L± acts on a wave function, we define

L± |l, mi = C± (l, m)|l, m ± 1i (5.21)

Using

|C+ (l, m)|2 hl, m + 1|l, m + 1i = hl, m|L− L+ |l, mi =

= hl, m|L2 − L2z − h̄Lz |l, mi = h̄2 l(l + 1) − m2 − m = h̄2 [(l − m)(l + m + 1)]
 

we obtain
q q
C± (l, m) = h̄ l(l + 1) − m(m ± 1) = h̄ (l ∓ m)(l ± m + 1) ≥ 0 (5.22)

We can conclude l(l + 1) ≥ m(m + 1) and l(l + 1) ≥ m(m − 1) and since l ≥ 0 therefore
obtain the important inequality
−l ≤ m ≤ l (5.23)
3
Note, that we could have also constructed the set of eigenfunctions such that they are common for Lx
and L2 , which is due to (5.10) not the same. In other words, not any eigenfunction of L2 needs to be
an eigenfunction of Lz , but one can always choose the eigenfunctions such that they are common for both
operators.
4
In principle l ≤ −1 would also be possible, but this can be circumvented by introducing l0 = −(l + 1) ≥ 0
and noting that l0 (l0 + 1) = l(l + 1).

T. Hickel 54 WS 2019/20
Module 2b-E1 5.2. The angular momentum operator

and obtain from (5.22)

L− |l, mmin i = 0 =⇒ mmin = −l (5.24)

L+ |l, mmax i = 0 =⇒ mmax = l (5.25)

Consequently, (2l + 1) must be an integer5 , and m can have the values

m = −l, −l + 1, −l + 2, . . . , l − 2, l − 1, l (5.26)

The case of l being a half-integer leads to the concept of spins. For the angular momentum,
it follows from Eq. (5.28) below that only integer values of l are allowed.

5.2.2 The spherical harmonics

So far we have only resolved the eigenvalues of Eqs. (5.12) and (5.13). We still need to say
something about the eigenfunctions. For this we need the representation of the new operators
in terms of spherical coordinates
h̄ ∂ ∂ ∂
 
Lz = , L± = h̄e±iφ ± + i cot θ (5.27)
i ∂φ ∂θ ∂φ
The simple structure of Lz gives direct access to the φ dependence of the wave function:

Lz Yl,m (θ, φ) = h̄mYl,m (θ, φ) =⇒ Yl,m (θ, φ) = Fl,m (θ)eimφ (5.28)

Due to the geometrically required identity Yl,m (θ, φ + 2π) = Yl,m (θ, φ) it directly follows that
m and with (5.26) therefore also l need to be integer numbers. For the θ dependence of the
wave function, we make use of Eq. (5.25) and derive
∂
 
L+ Yl,l (θ, φ) = 0 =⇒ − l cot θ Fl,l (θ) = 0 (5.29)
∂θ
=⇒ Fl,l (θ) = (sin θ)l (5.30)

All the other wave function are then obtained by applying L− to Yl,m (θ, φ). This results to
s
m 2l + 1 (l − m)! m
Yl,m (θ, φ) = (−1) P (cos θ)eimφ (5.31)
4π (l + m)! l
l−m
(l + m)! (1 − u2 )−m/2 d

Plm (u) = (−1)l+m (1 − u2 )l (5.32)
(l − m)! 2l l! du
Where Yl,m (θ, φ) are called spherical harmonics and Plm (u) are the associated Lagendre poly-
nomials. Explicitely the first spherical harmonics are given by:
1
Y0,0 = √ (5.33)
4π
r
3 iφ
Y1,1 = − e sin θ (5.34)
r 8π
3
Y1,0 = cos θ (5.35)
r 4π
15 2iφ 2
Y2,2 = e sin θ (5.36)
r32π
15 iφ
Y2,1 = − e sin θ cos θ (5.37)
r 8π 
5 
Y2,0 = − 3 cos2 θ − 1 (5.38)
16π
5
The mmin and mmax are connected by an integer number of steps.

T. Hickel 55 WS 2019/20
Module 2b-E1 5.3. The radial equation for the hydrogen atom

Further functions are obtained from the relation Yl,−m = (−1)m Yl,m ∗ , which shall not be

proven here. The set of spherical harmonics yields a complete basis for the space of spherical
angles, i.e., the angle dependence of each wave function can be expanded in terms of the
functions Yl,m (θ, φ).

5.3 The radial equation for the hydrogen atom

We can now return to the Schödinger equation (5.8) of the hydrogen atom. Since Yl,m (θ, φ)
is an eigenfunction of L2 it is also an eigenfunction of the H. However, to resolve the full
wave function ψ(r), its r dependence also needs to be determined. The latter can also depend
on l since the application of L2 within (5.8) brings an h̄2 l(l + 1) term into the Hamiltonian.
Therefore, the following separation ansatz is justified:

ψ(r) = Rn,l (r)Yl,m (θ, φ). (5.39)

Resolving the previously determined angle dependence, the following radial equation needs
to be solved:
" ! #
h̄2 ∂2 2 ∂ h̄2 l(l + 1)
− + + + V (r) Rn,l (r) = En,l Rn,l (r). (5.40)
2µ ∂r2 r ∂r 2µr2

First of all, looking at (5.6) and with a good guess the equation can be simplified by
writing
!
u(r) ∂2 2 ∂ 1 ∂ ∂ u(r) 1 ∂ 2 u(r)
 
R(r) = =⇒ 2
+ R(r) = 2 r2 = (5.41)
r ∂r r ∂r r ∂r ∂r r r ∂r2

which yields " #

∂2 l(l + 1) 2µ
2
− − 2 (V (r) − E) u(r) = 0. (5.42)
∂r r2 h̄
We would again like to go to dimensionless variables. For this purpose a typical length
scale for the Coulomb potential (5.3), considered here, should be obtained. We do this by
noting6 that rpr is of the order of h̄. Since at the same time p2r /µ is of the same order as
V (r) we have as a typical radius

h̄2 e2 1 4πε0 h̄2 r

= ⇐⇒ aB = . =⇒ ρ=Z (5.43)
µa2B 4πε0 aB e2 µ aB

The such defined value for the typical radius aB is also called Bohr radius. We further ask
about a typical energy of the system. We use again the kinetic energy as an indicator for this
and write s
h̄2 e4 µ 1 E
ER = 2 = 2 =⇒ η = − (5.44)
2µaB 2h̄ (4πε0 )2 Z ER
The such defined value for the typical radius ER is also called Rydberg energy. The definition
already takes into account that for bound states, in which we are interested here, we have
an energy E < 0. Based on these variables (5.42) transforms to the following differential
equation " #
∂2 l(l + 1) 2 2
− + − η u(ρ) = 0. (5.45)
∂ρ2 ρ2 ρ
6
This can be concluded from the Heisenberg uncertainty relation, from the commutation relation between
both variables, or from the operator definition of pr .

T. Hickel 56 WS 2019/20
Module 2b-E1 5.3. The radial equation for the hydrogen atom

The equation is solved by studying some limiting cases first. On the one hand, we can con-
sider the limit ρ → ∞, for which the wave function should vanish. The relevant, approximate
equation in this case is:
" #
∂2
− η 2 u(ρ) = 0 =⇒ u(ρ) ∼ e−ηρ . (5.46)
∂ρ2

On the other hand, we have in the limit ρ → 0 a relevant, approximate equation which can
also be solved directly7 :
" #
∂2 l(l + 1)
2
− u(ρ) = 0 =⇒ u(ρ) ∼ ρl+1 (5.47)
∂ρ ρ2

Therefore, the following ansatz for the wave function seems to be reasonable:

u(ρ) = e−ηρ ρl+1 P (ρ) =⇒ R(ρ) = e−ηρ ρl P (ρ) (5.48)

Here, P (ρ) describes some remaining ρ-dependence. This yields the following derivatives:
 
∂ l+1
u(ρ) = − η u(ρ) + e−ηρ ρl+1 P 0 (ρ)
∂ρ ρ
2
 2  
ηρ −(l+1) ∂ l+1 l+1 l+1
e ρ u(ρ) = − 2 P (ρ) + − η P (ρ) + − η P 0 (ρ)
∂ρ2 ρ ρ ρ
 
l+1
+ − η P 0 (ρ) + P 00 (ρ)
ρ
 
ηρ −(l+1) 00 l+1
e ρ [. . .] u(ρ) = P (ρ) + 2 − η P 0 (ρ)
ρ
" 2 #
l+1 l + 1 l(l + 1) 2 2
+ −η − 2 − + − η P (ρ)
ρ ρ ρ2 ρ
 
l+1 2
0 = P 00 (ρ) + 2 − η P 0 (ρ) + [1 − η(l + 1)] P (ρ)
ρ ρ
(5.49)

We note that for the last step of the derivation the cancellation of several terms in the
prefactor of P (ρ) is used. This cancellation is pure coincidence and a particularity of the
Coulomb potential.
The problem is again solved with the Sommerfeld polynomial method. For this we assume
?
X
P (ρ) = αk ρk (5.50)
k=0
?
X
αk k(k − 1)ρk−2 + 2k(l + 1)ρk−2 − 2kηρk−1 + 2 [1 − η(l + 1)] ρk−1
 
=⇒ 0 =
k=0
?
X
= [αk+1 (k + 1)k + 2αk+1 (k + 1)(l + 1) + 2αk [1 − η(k + l + 1)]] ρk−1
k=0

The polynomial on the right hand side of Eq. (5.49) can only vanish for all ρ, if all coefficients
are zero. This yields to the following recursion relation of the coefficients αk in (5.49):

η(k + l + 1) − 1
αk+1 = 2 αk (5.51)
(k + 1)(k + 2l + 2)
7
In principle there exists with u(ρ) ∼ ρ−l another possible solution, for which however it can be shown that
the divergence at ρ = 0 is not physical. For instance it would imply that the operator pr is not hermitean for
this state.

T. Hickel 57 WS 2019/20
Module 2b-E1 5.4. Interpretation

In order to ensure that the complete solution (5.48) can be normalized, we have to ensure that
that the polynomials has not infinitely many terms. Instead a truncation of the polynomial
is needed. This yields the condition for the numerator in (5.51):
1 1
η= = with n = l + 1, l + 2, l + 3, . . . (5.52)
k+l+1 n
We use Eq. (5.44) to return to proper energy values and obtain:
Z 2 ER
En = − with n = 1, 2, 3, . . . and l = 0, 1, 2, . . . , n − 2, n − 1 (5.53)
n2

5.4 Interpretation
The equation (5.53) is the central result for the hydrogen atom. It states that the energies of
this system are quantized and proportional to n−2 for a single quantum number n, resulting
from the solution of the radial equation.
The lowest energy is for n = Z = 1 the Rydberg energy defined in Eq. (5.44). This lowest
energy shell of an atom is typically called K-shell. For the upcoming quantum numbers
n = 2, 3, 4, . . . one speaks about the L-, M-, O-shell, respectively. Transitions of electrons
between these shells are related to electromagnetic (light) absorption or emission, given by
the energy difference
1 1
 
hνn,m = ER − , (5.54)
n2 m2
which can be seen as peaks in a spectral analysis. Historically the corresponding series of
peaks for the different n values are named after the physicists Lyman, Balmer, Paschen, and
Brackett.
The remarkable observation is that the quantum numbers l and m from the analysis of
the angular momentum part, do not enter the energy expression (5.53) although l is part of
the radial equation. This is mainly due to the particular structure of the potential, which
leads to the cancellation effects reported for Eq. (5.49) above. Still n and l in (5.53) are
not completely independent. The relation can either be expressed in the form n ≥ l + 1 or
equivalently as l ≤ n − 1. The second version gives for a fixed n value a constraint for the l
values. The different l values correspond to the possible orbitals for a fixed shell n. For each
orbital there exist different states corresponding to the 2l + 1 possible m values. Altogether
one has
n value l value possible m values name of orbital
1 0 0 1s
2 0 0 2s
2 1 1, 0, -1 2px , 2py , 2px
3 0 0 3s
3 1 1, 0, -1 3px , 3py , 3px
3 2 2, 1, 0, -1, -2 3d → 5 states
.. .. .. ..
. . . .

From these rules for n, l, m in principle the full structure of the periodic system of elements
can be understood, if the following two aspects are additionally taken into account:
• For each of the electron states with quantum numbers n, l, m there are two possible
spin states, typically called spin-up and spin-down.
• A fully characterized quantum state (identified by the quantum numbers n, l, m and
the spin) can only be occupied by one electron in a certain system. This rule is called
Pauli exclusion prinicple.

T. Hickel 58 WS 2019/20
Module 2b-E1 5.5. Discussion of eigenfunctions of the H atom

The number of quantum states for the same energy (5.53) is also called degeneracy. It is
given by the following formula:
n−1
X
gs = 2 (2l + 1) = 2n2 . (5.55)
l=0
However, in most realistic system this degeneracy is slightly lifted, giving rise to a so-called
fine-structure or hyperfine-structure. This is due to several physical processes. On the one
hand, there is an interaction of the spin-momentum of a singular electron with its own angular
momentum. Furthermore, the electron spin can also lead to an interaction with the spin of
the nucleus. On the other hand, there is an interaction of the different electrons within one
atom, modifying the effective potential of an individual electron and in particular lifting in
this way the degeneracy with respect to l.

5.5 Discussion of eigenfunctions of the H atom

As already stated in (5.39) the complete wave functions corresponding to the eigenstates of
the hydrogen atom are given by
ψ(r) = Rn,l (r)Yl,m (θ, φ). (5.56)
For a certain set of l, m values the angular part of this wave function has already be pro-
vided explicitly. The structure of the radial part of the wave function is basically given by
Eqs. (5.48), (5.50) and (5.51). Applying the recursion (5.51) several times, one obtains the
following explicit formula for the coefficients in (5.50):
 k
k 2 (2l + 1)!(n − (l + 1))!
αk = α0 (−1) (5.57)
n k!(k + 2l + 1)!(n − (l + k + 1))!
Due to this structure, the polynomial part of the radial wave function can be expressed in
terms of so-called associated Laguerre polynomials. However, we refrain here from go more
into the mathematical features of these polynomials.
It is more instructive to look at some examples. For the situation l = n − 1 the radial
part has the following form:
3/2 s n−1
2Z 1 −(Zr)/(naB ) 2Zr
 
Rn,n−1 (r) = e (5.58)
naB (2n)! naB
One can clearly see here the competition between the term rn−1 yielding an increase of
the wave function with increasing r close to the nucleus and the term e−r/aB yielding an
exponential decrease far away from the nucleus.
Even more explicitly, some of the radial wave functions with small quantum numbers are:
 3/2
Z
R10 (r) = 2 e−Zr/aB (5.59)
aB
 3/2  
Z Zr
R20 (r) = 2 1− e−Zr/2aB (5.60)
2aB 2aB
 3/2
1 Z Zr −Zr/2aB
R21 (r) = √ e (5.61)
3 2aB aB
3/2 
2Zr 2(Zr)2
 
Z
R30 (r) = 2 1− − e−Zr/3aB (5.62)
3aB 3aB 27a2B
√  3/2  
4 2 Z Zr Zr
R31 (r) = 1− e−Zr/3aB (5.63)
3 3aB aB 6aB
√  3/2  2
2 2 Z Zr
R32 (r) = √ e−Zr/3aB (5.64)
27 5 3aB aB

T. Hickel 59 WS 2019/20
Module 2b-E1 5.5. Discussion of eigenfunctions of the H atom

Figure 5.1: Visualization of the radial probability distribution of an electron in a H-atom as

function of the distance from the centre of the nucleus. [Taken from W. Nolting, Grundkurs
Theoretische Physik, vol. 5-2, p. 123 (Springer, Berlin, 2009)]

An interesting quantity in the discussion of these wave functions is the maximum proba-
bility for the distance r of the electron from the centre of the nucleus. It is not just given by
|Rn,l |2 . Instead, one needs to take care that the change to spherical coordinates is related to
a functional determinant r2 sin θ for integrations. It gives the size of typical segment, when
splitting the volume into parts drdθdφ. Accordingly the probability of the an electron to
have the distance r is given by

Zπ Z2π
2
wnl (r)dr = r dr sin θdθ dφ|ψnlm (r)|2 = r2 dr|Rn,l |2 . (5.65)
0 0

For typical quantum numbers n, l the latter is visualized in Fig. 5.1. We note that the
maximum radius is for l = n − 1 given by n2 /aB /Z. This means that the 1s states of the H
atom has its maximum probability for the position of the electron at a distance aB from the
centre of the nucleus. This explains the relevance of the Bohr radius aB introduced earlier.
For the visualization of the angular probability distribution

2l + 1 (l − m)! m
ŵlm (θ, φ) = |P (cos θ)|2 (5.66)
4π (l + m)! l

one usually uses pole-figures where the size of (5.66) is plotted as the length of the radius
vectors. This yields to the typical figures of p- and d- orbitals, which can be found in the
text books discussing atomic orbitals.

T. Hickel 60 WS 2019/20
Chapter 6

Magnetism in solids

6.1 Motivation
Magnetism is one of the most interesting phenomena in solid state physics. On the one hand,
magnetic properties can directly been used in applications. Advanced materials can make
use of

• (giant) magneto-resistance effect for storage applications,

• magnetocaloric effects for cooling of devices, or

• magnetic shape-memory effects for sensors and actuators.

On the other hand, magnetism indirectly influences the behaviour of materials. The fact that
iron shows a body-centred cubic (bcc) structure in the ground state, for example, is only to a
stabilization effect caused by magnetism. Assuming iron to be non-magnetic, would yield the
face-centred cubic (fcc) structure to be the most stable one of iron as can be seen in Fig. 6.1.

6.1.1 Classification of magnetism

There is a large variety of magnetic phenomena, which can be classified into

• diamagnetism, occuring if a magnetic field induces magnetic dipoles, which (according

to Lenz’s law) are oriented antiparallel to the field (in almost all materials, in particular
superconductors)

• paramagnetism, occuring if previously disordered magnetic moments are aligned by a

magnetic field (effect for example of transition metals, system with quasi-free electrons
– Pauli paramagnetism)

• collective magnetic ordering, such as ferromagnetism, ferrimagnetism or antiferro-

magnetism (several metals with localized or itinerant magnetic moments).

6.1.2 Bohr-van-Leeuwen theorem

It can be shown rigorously that magnetism is solely a quantum effect. The corresponding the-
orem, which states that in a purely classical description no dia-, para- or collective magnetism
exists, is the Bohr-van-Leeuwen theorem.
Proof : The definition of the magnetization of a system of identical magnetic moments
m is given as
N ∂H
M = hmi with m=− . (6.1)
V ∂B0

61
Module 2b-E1 6.1. Motivation

Figure 6.1: Total energy vs volume curve for two crystallographic (fcc, bcc) and three mag-
netic structures (non- (nm), ferro- (fm), antiferro- (afm, afmd) magnetic) of single crystalline
iron. The example shown here reveals that the T = 0K thermodynamic ground state of bulk
iron is the ferromagnetic bcc structure.

Here, V is the volume, N the number of magnetic particles, B0 an external magnetic field
and H the classical Hamilton function (the total energy). The latter is given as

1 3N
Xe
H= (pi + eAi )2 + H1 (x1 , . . . , xNe ) (6.2)
2m i=1

Here, it is assumed that the only particles carrying a magentic moment are the Ne electrons
per ion, which interact with the external field. Using the definition of the vector potential
that B0 = curlA and the well known Lorentz force FL = e(v × B0 ) the modification of the
kinetic energy by the the vector field becomes more reasonable. The proper derivation is
however beyond the scope of this course.
Classical thermodynamics tells us that
1
Z Z
hmi = ... dx1 · · · dxNe dp1 · · · dpNe m e−βH (6.3)
ZZ Z
with Z = ... dx1 · · · dxNe dp1 · · · dpNe e−βH (6.4)

being the partition function. Based on (6.1) it is easy to show that

1 ∂Z
hmi = . (6.5)
βZ ∂B0
Using the Hamiltonian (6.2) in (6.4) one can easily convince oneself by shifting the integration
variables that the result for the partition function does not depend on the actual choice of the
vector potential A. Hence, the derivative (6.5) must also vanish. As a consequence (using
(6.1)) the total magnetization is zero, i.e., there is no long-range magnetic order possible if
classical assumptions are used.

6.1.3 The quantum-mechanical nature of spins

The existence of a finite magnetization of a material requires the presence of magnetic mo-
ments. Decisive is the magnetic moment carried by the electrons in the material, which is

T. Hickel 62 WS 2019/20
Module 2b-E1 6.2. Magnetism of local moments

called spin. The fact that an electron has a spin, can be rigorously derived by generalizing
the Schrödinger equation relativistically to the Dirac equation. From these derivations, which
we don’t want to perform here, one can also understand that the spin quantum number of
an electron is 1/2. This corresponds to a magnetic moment of 1 µB (Bohr magneton).
Representing a magnetic moment, the spins of an electron are three-dimensional objects
with x−, y−, and z−coordinates. It is important to note that from a mathematical point
of view, spins do not behave like vectors, but rather like so-called “spinors”. One of the
differences is, that a 360◦ rotation transforms a spinor into its negative, and it takes a
rotation of 720◦ for a spinor to be reproduced. The components of an electron spin S = 12 h̄σ
are therefore typically represented by so-called Pauli matrices:
! ! !
0 1 0 −i 1 0
σx = , σy = , σz = . (6.6)
1 0 i 0 0 −1
It is also possible that several electrons form a combined magnetic moment, which cor-
responds to a larger spin quantum number S. In any case, the spin operators by definition
need to fulfil the commutations relations of angular momentum operators (compare (5.10)):
[Siy , Sjz ]− = ih̄Six δij



[Siz , Sjx ]− = ih̄Siy δij =⇒ Si × Sj = ih̄δij Si . (6.7)
[Six , Sjy ]− = ih̄Siz δij



Here i, j indicate site indices and δij ensures that the commutation relations only apply to
spinor components corresponding to the same site. The relations (6.7) establish the decisive
difference between a spinor and a classical vector, which becomes only unimportant in the
limit S → ∞.
The implication of the quantum nature of spins is a discrete set of eigenvalues, in the
same way as for any other angular momentum operator. In particular, for the z-component
1
of the single electron spin, represented by ! 2 h̄σz , the eigenvalue
!
are + 12 h̄ and − 12 h̄. The
1 0
corresponding eigenvectors are given by and , corresponding to what is typically
0 1
called a spin-up and a spin-down electron. One can furthermore derive that the ladder
operators
! !
h̄ 0 2 h̄ 0 0
S + = S x + iS y = , S − = S x − iS y = , (6.8)
2 0 0 2 2 0
transform one of the eigenstate into the other.

6.2 Magnetism of local moments

The above mentioned combination of electronic moments σ to a single magnetic moment Si
requires that these electrons are all permanently attributed to the same lattice site i. This
case is called local moment magnetism. In the opposite case the conduction electrons that
carry the spin are quasi-free in the crystal, what is called itinerant magnetism (or sometimes
band-magnetism). It is important to note that completely different theoretical concepts have
to be employed in these two cases. We focus here on the local magnetism, since it is the more
common situation in materials science.

6.2.1 Exchange interactions

The possibility of a collective order of the local magnetic moments requires some kind of
interaction mechanism. Mathematically this is typically expressed by the term
Ji,j Si · Sj (6.9)

T. Hickel 63 WS 2019/20
Module 2b-E1 6.2. Magnetism of local moments

in the Hamiltonian. Here, Ji,j is the interaction energy between the localized spins Si and Sj
at two different lattice sites. We know from the Bohr-van-Leeuwen theorem that this inter-
action is not of classical origin. Instead, it can be derived from considerations of quantum-
statistics, which implies symmetry conditions for a many-particle wave function. According
to these general concepts, the exchange of two particles should (apart from a sign) not change
the wave function. If one factorizes the wave function |Ψi into an orbital part |Φi and a spin
part |σi, both parts need to by symmetric or antisymmetric in such a way that the total func-
tion is antisymmetric with respect to exchanges. For the spin-part this can be realized by a
triplet (symmetric) or a singlet (antisymmetric), and also the orbital part allows a symmetric
or an antisymmetric superposition of single-particle states:
1
|σisym = | ⇑⇑i or | ⇓⇓i or √ (| ⇑⇓i + | ⇓⇑i) (6.10)
2
1
|σiasym = √ (| ⇑⇓i − | ⇓⇑i) (6.11)
2
1
|Φisym/asym = √ (|φA (1)i|φB (2)i ± |φA (2)i|φA (1)i) . (6.12)
2
Due to Coulomb interaction |Φisym and |Φiasym describe states with different energies (bond-
ing vs. antibonding state). The coupling of the symmetry conditions, on the other hand,
implies that also the different spin configurations (singlet vs. triplet) correspond to different
energies. In this way one can understand that the principle of exchange together with the
Coulomb-interaction implies a coupling of magnetic moments. Such an indirect coupling of
spins is therefore also called exchange interaction.
Interpreting the interaction energy classically instead, would actually mean that the mag-
netic moment µB of an atom sitting at site i is influenced by the magnetic field Bex caused by
all other atoms. Equalizing the corresponding magnetic energy µB Bex (ordering) with ther-
modynamic energy kB TC at the critical temperature (disordering), yields for iron an exchange
field of about 1500 Tesla. This is unrealistically high.

6.2.2 The Heisenberg model

The Hamiltonian of the Heisenberg model is typically given in real space as:
X µB X
H=− Ji,j Si · Sj − g B0 Siz (6.13)
i,j
h̄ i

For the exchange integrals Ji,j the symmetry conditions Ji,j = Jj,i holds. Furthermore,
Ji,i = 0 by convention. The components (Six , Siy , Siz ) of the spin operators Si used here show
the commutation relations (6.7). Furthermore, the ladder operators are again defined as:
1 +  1  + 
Si± = Six ± iSiy =⇒ Six = Si + Si− , Siy = Si − Si− (6.14)
2 2i
Using these relations, one can rewrite
1 +   1  
Si · Sj Si + Si− Sj+ + Sj− −
= Si+ − Si− Sj+ − Sj− + Siz Sjz
4 4
1 + − − +

z z
= S S + S S + Si Sj (6.15)
2 i j i j  µB
Ji,j Si+ Sj− + Siz Sjz − g
X X
⇒ H = − B0 Siz (6.16)
i,j
h̄ i

In the last step we used Ji,j = Jj,i and Ji,i = 0.

T. Hickel 64 WS 2019/20
Module 2b-E1 6.2. Magnetism of local moments

For the Fourier transformation the following rules are used:

1 X
J(k) = Ji,j eik(Ri −Rj ) (6.17)
N i,j
1 X
Ji,j = J(k)e−ik(Ri −Rj ) (6.18)
N k
Siα e−ikRi
X
S α (k) = (6.19)
i
1 X α
Siα = S (k)eikRi (6.20)
N k

for (α = x, y, z, +, −). Since the exponential factor is the same for α = + and α = −, it
follows that
+
S + (k) = S − (−k). (6.21)
From the commutation relations in real space one can directly (just by inserting the
Fourier transformations and removing one sum due to the δij in the real space) conclude
about corresponding commutations in k-space:

[Siz , Sj± ]− = ±h̄δij Si± ⇒ [S z (k1 ), S ± (k2 )]− = ±h̄S ± (k1 + k2 ) (6.22)
[Si+ , Sj− ]− = 2h̄δij Siz ⇒ [S + (k1 ), S − (k2 )]− = 2h̄S z (k1 + k2 ) (6.23)

The Fourier transformation of the Heisenberg Hamiltonian is given by

1 X µB
J(k) S + (k)S − (−k) + S z (k)S z (−k) − g B0 S z (0).

H=− (6.24)
N k h̄

Again this follows directly from entering the Fourier transformations (6.18) and (6.20) into
the Hamiltonian (6.16). The only additional step to be done is to use the relation
1 X ikRi
e = δ(k, 0), (6.25)
N i

which removes in the first part of the Hamiltonian two of the three summations in k-space.
Provided the exchange integrals Ji,j are positive, it is immediately clear that the ground
state |FMi of this Hamiltonian in real as well as in reciprocal space is the state with saturated
magnetic moments. For this state we have:

Siz |FMi = h̄S|FMi −→ S z (k)|FMi = h̄N S|FMiδ(k, 0) (6.26)

Si+ |FMi =0 −→ +
S (k)|FMi = 0 (6.27)

Hence:
1 X 1 X
J(k)S + (k)S − (−k)|FMi = − J(k) S − (−k)S + (k) + 2h̄S z (0) |FMi
 
−
N N
k k
= −2N h̄2 SJii |FMi = 0
1 X 1
− J(k)S z (k)S z (−k)|FMi = −h̄N S J(0)S z (0)|FMi = −N J0 h̄2 S 2 |FMi
N N
k

This means that the ground state of the Heisenberg Hamiltonian has the energy

EFM = −N J0 h̄2 S 2 − N gµB B0 S, (6.28)

P
where J0 = J(0) = i Ji,j .

T. Hickel 65 WS 2019/20
Module 2b-E1 6.3. Mean-field theory

6.3 Mean-field theory

The Heisenberg model in the real-space notation 6.16 describes two processes: On the one
hand, there is a coordinated change of the z component of the spins on two sites i and j via
the term Si+ Sj− . Whenever it decreases somewhere, it must increase somewhere else. In the
ferromagnetic ground state this term vanishes, since a further increase of spin is not possible.
On the other hand, each spin on site i feels the presence of the other spins j via the term
Siz Sjz in a similar manner as if the other spin would create a magnetic field. In mean-field
theory, this idea is taken more seriously, by expressing the net effect of all atoms j by one
effective exchange field.
 Mathematically,
 the mean-field theory results in a neglect of fluctuations of the form
 − hÂi B̂ − hB̂i . For the terms of the Heisenberg model this means:

Siz Sjz = Siz hSjz i + hSiz iSjz − hSiz ihSjz i = Siz hS z i + hS z iSjz − hS z i2
Si+ Sj− = Si+ hSj− i + hSi+ iSj− − hSi+ ihSj− i = 0

We assume in the first equation a system with translational symmetry, i.e. a ferromagnet.
The Hamiltonian becomes in this case
µB
 X X
HMF = N J0 hS z i2 − 2J0 hS z i + g B0 Siz = Ẽ − B̃ Siz , (6.29)
h̄ i i

where the first term in the bracket plays the role of the exchange field.
If the eigenvalues von Siz are denoted h̄mS , then the single-particle partition sum
+S
sinh(βh̄B̃(2S + 1))
Z1 = e−β Ẽ e−βh̄B̃mS = e−β Ẽ
X
(6.30)
mS =−S sinh(βh̄B̃)

can be used to determine the magnetization of the system as

1N ∂
h̄hS z i = ln Z1
β V ∂ B̃
N
 
= SgµB BS 2βJ0 ShS z i + βSgµB B0 (6.31)
V
2S + 1 2S + 1 1 x
   
with BS (x) = coth x − − coth . (6.32)
2S 2S 2S 2S

The expression BS (x) is called Brillouin function and describes in good approximation the
dependence of the magnetization of a paramagnetic system in a magnetic field. This is an
equation that describes hS z i as a function of hS z i, which needs to be solved self-consistently.
Such a solution is analytically in general not possible. Of particular interest is, however,
the special case of vanishing magnetization hS z i → 0 without an external magnetic field, i.e.
B0 = 0. In this limit, the Brillouin function function can be Taylor expanded

S+1 S + 1 2S 2 + 2S + 1 2
BS (x) ≈ x− x (6.33)
3S 3S 30S 2
and only the linear term matters. This is the same as requiring that both sides of Eq. (6.31)
have the same slope at the criticial temperature to get an solution. The temperature at
which this happens is called Curie temperature TC . In the mean-filed approximation it is
hence given by
2
kB TC = h̄2 J0 S(S + 1). (6.34)
3

T. Hickel 66 WS 2019/20
Module 2b-E1 6.4. Magnetic excitations

6.4 Magnetic excitations

6.4.1 Magnons
In the next step we consider the state

S − (k)|FMi. (6.35)

Making use of the Fourier transformation (6.19) it is clear that this state is a superposition
of all the state Si− |FMi. These kind of states correspond to the situation that out of the N
spins in a system, which are all oriented parallel in the ground state, only one spin has a
z-component which is reduced by 1 (or which is flipped to the opposite direction if S = 1/2).
One can easily convince oneself that due to the exchange interaction Ji,j such a single spin
flip Si− |FMi is not an eigenstate of the Hamiltonian (6.13). In contrast to this, the combined
state expressed in reciprocal space by (6.35) is an eigenstate of the Hamiltonian (6.24). This
will be shown in the following.
For this purpose one further commutator is needed:
1 X
[H, S − (k)]− J(p) [S + (p), S − (k)]− S − (−p) + S z (p)[S z (−p), S − (k)]−

= −
N p
1
+[S z (p), S − (k)]− S z (−p) − gµB B0 [S z (0), S − (k)]−
h̄
1 X
J(p) 2h̄S z (k + p)S − (−p) − h̄S z (p)S − (k − p)−

= −
N p
−h̄S − (k + p)S z (−p) + gµB B0 S − (k)
1 X
J(p) −2h̄2 S − (k) + 2h̄S − (−p)S z (k + p) + h̄2 S − (k)

= −
N p
−h̄S − (k − p)S z (p) − h̄S − (k + p)S z (−p) + gµB B0 S − (k)
h̄ X
J(p) 2S − (−p)S z (k + p) − S − (k − p)S z (p)

= −
N p
−S − (k + p)S z (−p) + gµB B0 S − (k)
1
P
using Jii = N p J(p) = 0 in the last step. As a result the application of the commutator to
the ground state |FMi is given as
 
[H, S − (k)]− |FMi = gµB B0 + 2h̄2 S (J0 − J(k) S − (k)|FMi .


(6.36)
| {z }
h̄ω(k)

Here, also J(k) = J(−k) has been used.

The statement that S − (k)|FMi is an eigenstate of the Hamiltonian follows now easily:

H S − (k)|FMi = S − (k)H|FMi + [H, S − (k)]− |FMi

= (EFM + h̄ω(k)) S − (k)|FMi

(6.37)

If we assume that |FMi is normalized, then

hFM|S + (−k)S − (k)|FMi = hFM| 2h̄S z (0) + S − (k)S + (−k) |FMi = 2h̄2 N S


(6.38)

We call the normalized state

1
|ki = √ S − (k)|FMi (6.39)
h̄ 2SN
a one magnon-state. It is the lowest excited state of the Heisenberg model. Since it corre-
sponds to a specific wave vector k, this state describes a wave-like distortion of the magnetic

T. Hickel 67 WS 2019/20
Module 2b-E1 6.4. Magnetic excitations

Figure 6.2: Spin-wave spectrum of one-magnon excitations for different wave vectors k in
bcc Fe. Experimental data obtain with spin-resolved neutron scattering are compared with
quantum-mechanical calculations in the framework of density functional theory (DFT).

moments in real space (also called spin wave). The physical situation becomes more clear, if
one calculates the expectation value
1
 
hk|Siz |ki = h̄ S − . (6.40)
N
Since this value is independent of i and k, the spin deviation of 1h̄ is uniformly distributed
over all lattice sites.
The excitation energy related to this process is given by h̄ω(k) and is dependent on the
wave vector k. One therefore obtains a magnon dispersion, which depends on the exchange
integrals and is therefore characteristic for each metal (see Fig. 6.2). The other way around:
The measurement of the magnon spectrum (e.g., with spin-polarized neutron scattering)
provides information on the exchange integral J(k). This is a standard procedure in materials
design to obtain the parameters of the Heisenberg model, which is then used for deriving
thermodynamic properties related to magnetism.

6.4.2 Linear spin-wave theory

A magnon is a quasiparticle, which describes the excitation of a spin wave. It is therefore
a very similar concept as the phonon studied in Sec. 4.4. Looking at Fig. 6.2, however, we
observe already one difference as compared to phonons. While their dispersion starts linearly
at the Γ point (see for instance Eq. (4.51)), we have a quadratic onset in the case of magnons.
This can also be shown analytically: The energy of a magnon has been determined in (6.37)
to be proportional to J(k). Using the Fourier transformation (6.17) and a Taylor expansion
of the exponential one obtains
1 X
J(k) = Ji,j eik(Ri −Rj ) (6.41)
N i,j
1 X 1 X
= J0 + k Ji,j (Ri − Rj ) + k2 Ji,j (Ri − Rj )2 + . . . (6.42)
N i,j 2N i,j

Here, the linear second term vanishes due to Ji,j = Jj,i and the sum in the quadratic third
term can be summarized to the so-called spin-wave stiffness constant D:
h̄ω(q) = gµB B0 + Dq2 + . . . , (6.43)

T. Hickel 68 WS 2019/20
Module 2b-E1 6.4. Magnetic excitations

which demonstrates the quadratic dependence. (The usage of q instead of k is a convention

for magnons and phonons.)
The excitation of magnons is a perturbation of the ferromagnetically saturated ground
state and therefore particularly relevant at low temperatures. In this regime, we can sim-
ilarly to the phonons in (4.28) define an occupation number operator and a corresponding
Hamiltonian:
X
H = EFM + h̄ω(q)n̂q with energies (6.44)
q
X
Enq = EFM + h̄ω(q)nq . (6.45)
q

In order to determine the average occupation number of the magnons, we first need to cal-
culate the partition function as:
∞
1
Z = e−βEFM e−βh̄ω(q)nq = e−βEFM
YX Y
(6.46)
q nq q 1− e−βh̄ω(q)

and obtains in a second step

1 ∂ 1
hn̂q i = − ln Z = βh̄ω(q) . (6.47)
β ∂(h̄ω(q)) e −1

This is the Bose-Einstein distribution function, indicating the magnons have statistical the
character of bosons.
The average magnon occupation number is for small temperatures directly related to the
magnetization of the described material. The latter is obtained from the free energy (or the
grand canonical ensamble)
h i
ln 1 − e−βh̄ω(q)
X
F (T, B0 ) = −kB T ln Z = EFM (B0 ) + kB T (6.48)
q

by the definition

1 ∂F (T, B0 ) N kB T X βgµB e−βh̄ω(q)

 
M (T, B0 ) = − = gµB S −
V ∂B0 T V V q 1 − e−βh̄ω(q)
!
1 X
= M0 1− hn̂q i . (6.49)
NS q

The sum is replaced by an integral, the geometric series is re-introduced and furthermore the
relation (6.43) is applied to obtain:
∞
V
Z
2
e−nβgµB B0 e−nβDq
X X
hn̂q i = d3 q
q (2π)3 n=1
∞ Z ∞
V 2
e−nβgµB B0 q 2 e−nβDq
X
= dq
2π 2 n=1 0
∞ −nβgµB B0
kB T 3/2 X e
 
= V (6.50)
4πD n=1
n3/2

In this way we have derived the famous Bloch T 3/2 law, which states that the magnetization of
the Heisenberg Hamiltonian decreases at low temperatures with T 3/2 . This result is rigorous
in the limit of low temperatures and fully agrees with experimental observations.

T. Hickel 69 WS 2019/20