Introductory Solid-State Physics

Lecture notes

Alexander Tsirlin

Leipzig University

work in progress, use with caution

report any comments and errors to [email protected]

February 4, 2024

These lecture notes are released under the generic Creative Commons (CC-BY-SA) license. You
are free to disseminate and re-use the full document or any of its parts by providing attribution as
follows: Alexander Tsirlin, Leipzig University with a link to the homepage of this module.
Contents

1. Bravais lattice, or how to pack a crystal? . . . . . . . . . . . . . . . . 4

1.1. Bravais lattice and unit cell
1.2. Close packing and structures of metals
1.3. Electron microscopy
2. Symmetry as the guiding principle . . . . . . . . . . . . . . . . . . . 7
2.1. Symmetry operations
2.2. Translational symmetry: lattice centering
2.3. Rotational symmetry and birefringence
2.4. Reflection and inversion symmetry
2.5. Neumann’s principle
2.6. Rotation vs. periodicity
3. Systematics of crystals: symmetry groups . . . . . . . . . . . . . . . . 11
3.1. Point symmetry operations
3.2. Point groups and crystal classes. Polarity
3.3. Crystal systems
3.4. Open symmetry elements
3.5. Chirality
3.6. Circular birefringence and dichroism
3.7. Space groups
4. Unpacking the crystal structure. . . . . . . . . . . . . . . . . . . . . 17
4.1. Atomic coordinates and Wyckoff positions
4.2. Crystal structure
4.3. Bragg’s law
4.4. Scattering experiments and x-ray diffraction
4.5. Lattice planes and Miller indices
4.6. Crystal faces and crystal directions
5. Spaces of crystallography: Reciprocal lattice . . . . . . . . . . . . . . 22
5.1. Laue condition
5.2. Reciprocal lattice
5.3. Relation to Bragg’s law
5.4. Aperiodic crystals
6. Structure factor: All shades of diffraction . . . . . . . . . . . . . . . . 26
6.1. Structure factor
6.2. Atomic approximation
6.3. Extinction conditions
6.4. Crystal structure determination
6.5. Atomic form factors
6.6. Neutron diffraction
7. Bonding in crystals: ionic . . . . . . . . . . . . . . . . . . . . . . . . 31
7.1. Types of chemical bonding
7.2. Cohesive and lattice energies
7.3. Madelung constant
7.4. Born-Landé equation
7.5. Bulk modulus
7.6. Ionic radius and Pauling’s rule
7.7. Ionic crystals as close-packed structures

1
8. Bonding in crystals: covalent, metallic, and van der Waals. . . . . . . . 36
8.1. Covalent crystals
8.2. Metallic crystals
8.3. Polymorphism
8.4. van der Waals radius and Lennard-Jones potential
8.5. Lattice energy
8.6. Bulk modulus
9. Mechanical properties . . . . . . . . . . . . . . . . . . . . . . . . . 39
9.1. Hydrostatic pressure, Equation of state
9.2. Uniaxial pressure
9.3. Shear deformation
9.4. Stress and strain tensors
9.5. Elastic constants
9.6. Plastic deformation
10. Dielectric properties . . . . . . . . . . . . . . . . . . . . . . . . . . 44
10.1. Permittivity
10.2. Dielectric loss
10.3. Induced dipoles and polarizability
10.4. Local electric field
10.5. Clausius-Mossotti relation
10.6. Permanent dipoles, Debye relaxation
10.7. Cole-Cole plots
11. Phonons and sound . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
11.1. Underlying approximations
11.2. Monoatomic chain
11.3. Elastic waves and sound
11.4. Longitudinal and transverse phonons
11.5. Diatomic chain
11.6. Acoustic and optical phonons
12. Phonons and light . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
12.1. Refractive index and reflectivity
12.2. Fluctuating dipoles
12.3. LO vs. TO
12.4. Interaction with light, polaritons
12.5. Lyddane-Sachs-Teller relation
12.6. Infra-red active phonon modes
13. Phonons and the reciprocal lattice . . . . . . . . . . . . . . . . . . . 61
13.1. Brillouin zone
13.2. Dynamical matrix
A. Electrodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
A.1. Maxwell’s equations
A.2. Electromagnetic waves
A.3. Longitudinal waves
B. Thermodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
C. List of experimental techniques . . . . . . . . . . . . . . . . . . . . . 64

2
Introduction
Condensed matter represents the largest sub-field of physics. It is not without a reason, because
condensed matter, or colloquially cond-mat, covers many aspects of physics that we encounter
in our daily life. While different branches of condensed-matter research have many concepts
and experimental techniques in common, they are also highly diverse, especially when it comes
to research objects.
One generally distinguishes research on soft matter (polymers, gels, membranes, cellular
structures) from hard condensed matter or simply solid-state physics that addresses properties
of solid materials. This distinction is in fact rather subtle. It goes back to the different dynamics
of “soft” and “hard” systems. Soft matter shows pronounced dynamics at ambient conditions,
it does not have a “fixed” structure, which is integral to solids. However, soft-matter objects
may become crystalline and turn into solids, whereas solids may show certain features of a
liquid.
Solid-state materials are variable too. They can be crystalline or amorphous. Crystalline
solids feature long-range-ordered structures (crystal structures) that are typically periodic.1 By
contrast, amorphous solids lack long-range order in any form, although they still show some
organization on the short-range scale.
Traditionally, solid-state physics has been developed for crystalline materials. Some of its
findings can be extended to amorphous materials too, but many of the key concepts, such as
band structure, are properly defined for crystalline materials only. In fact, they strongly rely
on periodicity and would require some re-thinking even for aperiodic crystals.
The content of these lecture notes splits into three parts:
• Structure of crystals (Chapters 1 to 6). Here, we discuss how crystals are organized
and use symmetry for classification of crystals and their properties.
• Atoms in crystals (Chapters...). This part concentrates on lattice degrees of freedom,
namely, bonding between the atoms as well as atomic vibrations and motions, and how
all of this contributes to different crystal properties.
• Electrons in crystals (Chapters...) This part will mostly address metals and analyze
how electronic degrees of freedom (free or, more precisely, itinerant electrons) affect dif-
ferent physical properties.
Many of the crystals also have spin degrees of freedom, but we disregard those for the sake of
simplicity. Magnetic properties of solids are addressed in a separate module.

Hyperlinks are highlighted in blue color. They will usually direct you to useful web resources
or an additional information on the topic.

1
A few examples of aperiodic crystalline solids will be introduced in Chapters 2 and 5.

3
1. Bravais lattice, or how to pack a crystal?
1.1. Bravais lattice and unit cell
Periodic pattern of the crystal can be described by a Bravais lattice defined as a set of points

R = n1 a + n2 b + n3 c (1.1)

where n1 , n2 , n3 are integers and a, b, c are three vectors that do not lie in the same plane.
They are known as lattice translations or lattice vectors. The parallelepiped spun by these three
vectors is the repetition unit of the lattice or its unit cell with the lattice parameters (a, b, c)
and lattice angles (α, β, γ). It is customary to define the angle between b and c as α, the angle
between a and c as β, and the angle between a and b as γ.
Not every lattice is a Bravais lattice. Consider for example hexagonal (honeycomb) lattice,
one of the popular geometries in modern solid-state physics. This lattice is clearly periodic, but
it is not a Bravais lattice per se, because its lattice sites can not be described with Eq. (1.1).
Bravais lattice of the honeycomb lattice is constructed by connecting second neighbors, as
shown in Fig. 1.1. The unit cell then contains two sites of the parent honeycomb lattice, one
at the corner and one in the interior. This situation is known as a Bravais lattice with basis.
Here, basis is the group of atoms contained by the unit cell. Of course, it is also possible to
choose hexagon as the repetition unit of the honeycomb lattice. However, such a choice would
introduce a disparity with other lattices, such as square lattice where the repetition unit is
obviously a square. The advantage of Eq. (1.1) lies in the unified description of all periodic
structures, so it is not surprising that Bravais lattice became the cornerstone of solid-state
physics.

1.2. Close packing and structures of metals

A naive way of seeing crystals is arrays of atoms placed in the positions defined by Eq. (1.1).
The most dense packing is achieved when atoms touch each other. Consider atom as a sphere
and put such atoms onto a simple cubic lattice. Its lattice parameter should be twice the radius
of the sphere, a = 2r. With one sphere per unit cell, the volume occupied by the sphere is
Vsphere = 34 πr3 = πa3 /6. On the other hand, the unit-cell volume is Vcell = a3 . The ratio of these
two volumes is the packing fraction, p = Vsphere /Vcell , and amounts to only 52% for the simple

unit
cell

Figure 1.1: Left: Bravais lattice, lattice vectors, and unit cell. Right: honeycomb lattice is not a
Bravais lattice per se; its Bravais lattice is indicated by the red lines, with two atoms per unit cell.

4
Li Be

Na Mg Al

K Ca Sc Ti V Cr Mn Fe Co Ni Cu Zn Ga

Rb Sr Y Zr Nb Mo Tc Ru Rh Pd Ag Cd In Sn

Cs Ba La Hf Ta W Re Os Ir Pt Au Hg Tl Pb

Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu

face-centered cubic (fcc) hexagonal close-packed (hcp) other

close-packed (dhcp) body-centered cubic (bcc)

Figure 1.2: Structures of simple metals. dhcp is the double-hexagonal close-packing, ...ABACABAC...

cubic lattice. We can also introduce the coordination number, i.e., the number of neighboring
atoms, which is as low as 6 in this case. Simple cubic lattice does not allow dense structures.
Finding a dense packing of spheres (better known as close packing in solid-state physics) is
in fact a common mathematical problem. By solving it, or by using general intuition of putting
together spherical objects like billiard balls, we know that each sphere can be surrounded by
six other spheres, resulting in a close-packed honeycomb layer. The next layer should follow
the dips formed by the first one. Different possible stackings give rise to a series of close-packed
structures, including:
• ...ABABAB... (hcp or hexagonal close packing)
• ...ABCABC... (ccp or cubic close packing, or fcc = face-centered cubic structure)
These two structures are much denser than the simple cubic lattice, thanks to the coordination
number of 12 (six neighbors in each layer, plus three neighbors in each of the two adjacent
layers). They are in fact common for simple metals, which are well described by this packing
concept because only one type of atoms is present in the crystal and because denser packing
allows higher electron concentration, which is favorable for the metallic bonding. Fig. 1.2 shows
that more than half of simple metals adopt either hcp or fcc structures. Most of the remaining
metals form a bcc (body-centered cubic) structure, which is only slightly less dense than the
close-packed ones.

lattice type packing fraction coordination number

simple cubic 0.52 6
body-centered cubic (bcc) 0.68 8
hexagonal close-packed (hcp) 0.74 12
face-centered cubic (fcc) 0.74 12
The choice between the bcc, fcc, and hcp structures for a given metal is far from trivial
and depends on details of its band structure (Ch. ??). Nevertheless, already at this point we
can make a general prediction that bcc metals should transform into hcp or fcc structures on
compression. This happens indeed in alkaline metals and also in iron that looses its magnetism
upon transforming from bcc to hcp polymorph around 15 GPa (why denser structures are less
likely to be magnetic is a separate and rather non-trivial question that will only be covered in
the Advanced Solid-State Physics module).

5
1.3. Electron microscopy
Back in time researchers could only guess about lattice periodicity and packing. Nowadays it
is possible to visualize the crystal lattice and even see individual atoms with the direct imaging
done by electron microscopy. Electron microscopes are somewhat similar to common optical
microscopes, but they use electrons as radiation with the much shorter wavelength, and of
course they require a much more sophisticated system of electromagnetic lenses to direct and
focus the electron beams.
Electron microscopes come in two main varieties:
• transmission electron microscopes (TEM) collect electrons on a detector behind the sam-
ple, just like an optical microscope does with visible light
• scanning electron microscopes (SEM) collect secondary electrons that are produced when
electrons are scattered on the sample
SEM’s are more compact, because one does not need a long separation between the sample
and detector. They can also map a larger area by scanning the sample with the electron beam
(hence the name). Concurrently, they lack in resolution. A typical instrument of this type has
the resolution of about 10 nm and can be used for imaging microstructures but not individual
atoms. By contrast, modern TEM’s have the resolution of 1 Å. They can operate in different
imaging modes and highlight heavy or light atoms, or map out chemical composition and even
electronic states on the sub-nm scale.
Two points of concern when using electron microscopy are:
• small size of the probe; a region of only 20 − 30 nm is probed in a single image taken with
the atomic resolution. This problem is partially mitigated by STEM (scanning TEM),
but even in this case only a tiniest fraction of the sample volume can be studied
• sample damage caused by the strong electron beam and high vacuum
Electron microscopy is the method of choice for studying defects and microstructures. When
it comes to long-range-ordered crystals, electron microscopy plays a somewhat auxiliary role,
while main information about the crystal structure comes from diffraction methods (Ch. 6).

6
2. Symmetry as the guiding principle
2.1. Symmetry operations
Long-range order of a crystal reflects its underlying symmetry. Most generally, symmetry of an
object is a set of transformations that leave this object invariant. We will consider geometrical
transformations that can be represented as
   
Rxx Rxy Rxz tx
symmetry operation = R ⊕ t =  Ryx Ryy Ryz  ⊕  ty  (2.1)
Rzx Rzy Rzz tz

where R is a unitary transformation, such as rotation, and t is a translation in space by a given

vector.

2.2. Translational symmetry: lattice centering

The simplest translational symmetry of a crystal is its periodicity. From Eq. (1.1) we know
that a, b, c and their linear combinations are allowed translations, so all of them are symmetry
operations of a crystal, as long as this crystal is periodic. Some other translations may be
allowed too, leading to a classification of lattices by lattice centering into:
• Primitive lattice (P ) has only a, b, c and their linear combinations as allowed translations
• Face-centered lattice (F ) features additional translations t = 21 (a ± b), 21 (a ± c), 12 (b ± c)
(half of the face diagonal for each face)
• Body-centered lattice (I) features additional translations t = 21 (a ± b ± c) (half of the
body diagonals)
• Base-centered lattice (A, B, C) features additional translations by half of the face diagonal,
but only for two faces out of six. For example, the C-centered lattice allows t = 12 (a ± b)
• Rhombohedral lattice (R) features an additional translation by t = 13 (a + b + c) (one third
of the body diagonal)
One could then ask why these additional vectors t are needed. Should they exist, lattice
vectors can be re-defined as the shortest repetition vectors of the crystal, and every lattice will
be primitive. True indeed, but some rotational symmetry may be lost on the way, as shown in
Fig. 2.1. Therefore, in a centered lattice one distinguishes:
• Primitive cell, which is the repetition unit with the smallest volume defined by the shortest
translations ap , bp , cp

ap bp
cp
Figure 2.1: Conventional and primitive unit
cells of the body-centered cubic lattice. The
lattice angles of the primitive cell deviate
primitive from 90◦ , so it has a lower symmetry than
the conventional cell. Note that ap , bp , cp are
also shorter than a, b, c, hence the primitive
c conventional
cell has the twice smaller volume than the
conventional one.
b

7
• Conventional cell, which is the smallest repetition unit of the highest symmetry defined
by the lattice vectors a, b, c
In a primitive lattice, both unit cells coincide. In a centered lattice, the volume of the primitive
cell is twice (A, B, C, I), three times (R), or four times (F ) smaller than the volume of the
conventional unit cell.
From this point on, we will distinguish lattice vectors a, b, c that define the conventional
unit cell, from lattice translations t that include a, b, c along with other vectors allowed by the
lattice centering.
Lattice centering is an important part of describing crystal symmetry, but it may not have
any immediate implications. We do not expect crystals with face-centered lattices to be distinct
from the body-centered ones. However, in special cases, such as simple metals from Fig. 1.2,
lattice centering determines the packing fraction and directly affects density of the crystal.

2.3. Rotational symmetry and birefringence

The p-fold rotation axis is defined as the rotation by φ = 360◦ /p. Here, p must be integer
because applying the same operation several times should end up in the full 360◦ rotation.
Rotation axes are labeled with numbers: two-fold rotation axis is 2, four-rold rotation axis is
4, etc. For a clockwise rotation, the corresponding transformation matrix is
 
cos φ sin φ 0
R =  − sin φ cos φ 0  (2.2)
0 0 1

for a p-fold rotation axis parallel to z.

Rotation axes have some immediate ramifications. First, they impose constraints on lattice
parameters. Consider a four-fold rotation axis along c. This symmetry operation transforms a
into b and thus requires not only a = b, but also γ = 90◦ . Three of such axes render a crystal
cubic!
Another ramification concerns crystal properties. A crystal with one 4-fold rotation axis
shows distinct properties along this direction and in the plane perpendicular to it. This disparity
is pretty obvious if we turn the crystal and measure its property along different directions. In
fact, it can be seen even in a single experiment! Crystals refract light, and refractive index
depends on the light polarization. Light polarized along the symmetry axis (i.e., the light
with the electric field vector parallel to the symmetry axis of the crystal) and light polarized
perpendicular to the symmetry axis feature different refractive indices, n∥ and n⊥ , respectively.
A non-polarized light is split by such a crystal into two distinct beams. This effect is known as
birefringence (∆n = n∥ − n⊥ ).
Historically, birefringence has been the very first tool used to identify crystal symmetries. It
is still employed by geologists during field trips where advanced instruments are not available.
It is also used technologically for a quick monitoring of product quality. Amorphous solids are
normally isotropic and should not exhibit birefringence (for example, it’s not seen in ordinary
glass). However, strains lead to anisotropy and can be mapped out by shining a suitably
polarized light on the sample. You can read more about it here.

8
2.4. Reflection and inversion symmetry
Eq. (2.2) covers only some of the unitary transformations. Two further important operations
are inversion center (labeled 1̄ or −1) and mirror plane (m),
   
−1 0 0 1 0 0
Rinversion =  0 −1 0  , Rreflection =  0 1 0  , (2.3)
0 0 −1 0 0 −1

where the latter describes a mirror plane perpendicular to z (m ⊥ c). These symmetry elements
can be combined with the rotation axes, namely, both can be present in the crystal at the same
time.
The reflection and inversion symmetries also have important implications. They determine
polarity and chirality of the crystal. We will get back to this in Ch. 3 after we learn how
individual symmetry elements build up a symmetry group.

2.5. Neumann’s principle

A general and very intuitive principle commonly attributed to Neumann is that properties of
a crystal should be invariant under its symmetry operations. Neumann’s principle is usually
invoked in its mathematical form, which states that any tensor property σ should follow

σ = R−1 σR. (2.4)

Consider linear-response properties, such as conductivity σ defined by Ohm’s law j = σ E

(j is electric current density and E is electric field) or permittivity ε defined by P = (ε − 1)ε0 E
(P is electric polarization). Both σ and ε are second-rank tensors because electric field applied
along one direction causes a response, polarization or current, along all three directions of the
crystal. We express this idea by writing

jα = σαβ Eβ (2.5)

and implying that conductivity has the general form of

 
σxx σxy σxz
σ =  σyx σyy σyz  (2.6)
σzx σzy σzz

with up to 9 independent components.

Such tensors are usually symmetric, σαβ = σβα . Their symmetry is rooted in some funda-
mental physical principles. For thermodynamic (equilibrium) properties it follows simply from
their definition via change of free energy in the applied field,

∂G
dG = −S dT + V dp − P dE ⇒ P = − . (2.7)
∂E T,p

Permittivity is then related to the second derivative of G with respect to E, and mixed deriva-
tives must be equal,
∂ 2G ∂ 2G
= ⇒ εαβ = εβα . (2.8)
∂Eα ∂Eβ ∂Eβ ∂Eα
A similar statement for transport properties is known as Onsager reciprocal relations. Without
going into details we only mention here that these relations need to be amended in the presence
of a magnetic field where conductivity tensor becomes antisymmetric (more on this in Ch. ??).

9
Symmetry of the tensor reduces the number of independent components from 9 to 6. Addi-
tional constraints can be derived from crystal symmetry using Eq. (2.4). For example, consider
the 4-fold rotation axis parallel to z. According to Eq. (2.2), it leads to the transformation
(x, y, z) → (y, −x, z) that converts σ into an equivalent tensor σ ′ ,
   
σyy −σyx σyz σxx σxy σxz
σ ′ =  −σxy σxx −σxz  vs. σ =  σyx σyy σyz  . (2.9)
σzy −σzx σzz σzx σzy σzz

The tensor components should be pairwise equal, so σyy = σxx , whereas −σyx = σxy . For a
symmetric tensor it means that all off-diagonal components vanish, leading to the final form of
 
σxx 0 0
σ =  0 σxx 0  (2.10)
0 0 σzz

for a crystal with the 4-fold symmetry axis.

Tensor form for a given crystal symmetry can be checked on the Bilbao server.

2.6. Rotation vs. periodicity

Different symmetry operations have been friends until now, but they can be foes too. Specifi-
cally, not every kind of rotational symmetry is compatible with periodicity of the lattice. One
can show that only 2-fold, 3-fold, 4-fold, and 6-fold rotations do not forbid periodicity in two
and three dimensions and can be thus present in periodic crystals. This statement is known as
crystallographic restriction theorem. Its mathematical proof can be found here.
Forbidden symmetries are not entirely impossible. For example, one may consider a crystal
with the 5-fold symmetry and, therefore, without periodicity. Such crystals have been dis-
covered in 1980’s and dubbed quasicrystals. They are typically metallic and combine several
chemical elements in rather weird proportions like Al65 Cu20 Fe15 . They serve as examples of
aperiodic crystals.

10
3. Systematics of crystals: symmetry groups
3.1. Point symmetry operations
Only a few local (point) symmetry operations R are compatible with periodicity. Every periodic
crystal can be then assigned to one or another symmetry group that comprises a subset of the
following symmetry elements:
• Rotation axes: 2, 3, 4, 6 (C2 , C3 , C4 , C6 )
• Inversion center: 1̄ (i)
• Mirror plane: m (σv , σh )
• Rotoinversion axes: 3̄, 4̄, 6̄
Here, we use labels in the international (Hermann-Mauguin) notation, which is common in
crystallography. The symbols in brackets are Schönflies notation favored by spectroscopists.
The last element, rotoinversion axes, has not been discussed yet and requires a further
comment. This symmetry element is a rotation followed by inversion,
 
− cos φ − sin φ 0
Rn̄ =  sin φ − cos φ 0  (3.1)
0 0 −1

The inversion center 1̄ is in fact the end member of this series, because it combines inversion
with the 1-fold rotation. Using Eqs. (2.2) and (2.3), one can verify that 2̄ = m. Three other
rotoinversion axes are independent symmetry elements, which are needed to describe special
situations, such as the symmetry of a tetrahedron (4̄).
No Schönflies symbols for rotoinversion axes exist, because a different symmetry element,
rotation-reflection axes (Sn ), is considered in this case. They are quite similar to n̄, but rotation
is followed by a mirror-plane reflection. This operation is also known as an improper rotation,
 
cos φ sin φ 0
RSn =  − sin φ cos φ 0  (3.2)
0 0 −1

assuming Sn ∥ z. Simple algebra shows that 3̄ = S6 , 6̄ = S3 , and 4̄ = S4 , whereas S2 = 1̄

and S1 = m, so we end up with exactly the same symmetry operations and an unnecessarily
confusing notation.

3.2. Point groups and crystal classes. Polarity

A combination of symmetry elements constitutes a symmetry group. Here, group is used in
mathematical sense, as a closed set of elements with an operation that transforms them into
each other. For example, a two-fold rotation axis and a mirror plane perpendicular to it generate
the symmetry group (Fig. 3.1):
2/m = { 1, 2, m, 1̄ } . (3.3)
Repeated application of the two-fold rotation leads to the unitary operation, 2 ⊗ 2 = 1. The
same is true for the mirror plane, m ⊗ m = 1. Finally, 2 followed by m results in the inversion
center 1̄.
Another example: a point group composed of three mutually orthogonal mirror planes
perpendicular to a, b, and c (Fig. 3.1). A consecutive application of ma and mb leads to
the coordinate transformation x, y, z −→ x̄, ȳ, z, so it is equivalent to a two-fold rotation axis

11
2/m 4mm mmm

Figure 3.1: Examples of point groups of symmetry.

along c. Each pair of planes generates a rotation axis, but we don’t need to list them all, and
the point group is labeled simply as mmm. It features eight symmetry elements,

mmm = { 1, 2a , 2b , 2c , ma , mb , mc , 1̄ } . (3.4)

Since one symmetry element can be obtained from a few others, it is customary to label the
symmetry group by its main generators, similar to the examples above. One usually chooses
rotation axis of the highest order and then symmetry elements perpendicular to it. Consider
the following examples:
• 4/m is the four-fold rotation along c and the mirror plane perpendicular to c.
• 4mm is the four-fold rotation along c, the mirror planes perpendicular to a and b, and
the mirror planes perpendicular to a ± b. The absence of / means that no mirror plane
perpendicular to c occurs. Moreover, the a and b directions are equivalent (orange planes
in Fig. 3.1), so they are mentioned only once, while the second m stands for the mirror
planes, which are perpendicular to the ab-diagonals (green).
There are altogether 32 point groups formed by the symmetry operations from Ch. 3.1.
The full list can be found on Wikipedia along with an explanation of the Schönflies symbols of
point groups that we will not consider here. Every point group defines its own crystal class, a
family of crystals with the similar shape that reflects the underlying symmetry of the crystal
structure.
The relation between the crystal class and crystal shape is most useful in mineralogy.
Synthetic crystals can be grown in different shapes, and they are often cut to a custom shape
as required by the experiments. The point symmetry remains important, though, because
it determines whether or not the crystal is polar. Polar direction is defined as the direction
that stays invariant under all symmetry operations of the system. Crystal or molecule with at
least one polar direction are capable of having dipole moment and electric polarization, which
is important for ferroelectric and pyroelectric properties. Inversion centers and rotoinversion
axes obviously forbid polarity. Examples of polar symmetries are 4mm and 3m, examples of
non-polar symmetries are 4/m and 3 2.

3.3. Crystal systems

32 points groups are hard to remember, and only crystallographers know them all (simply from
experience and not from hard learning), but it is useful to be familiar with seven crystal systems
that are distinguished by symmetry constraints imposed on the lattice parameters:

12
Cubic a = b = c, α = β = γ = 90◦ m3̄m, 4̄3m, 432, m3̄, 23
Tetragonal a = b ̸= c, α = β = γ = 90◦ 4, 4mm, 4/m, 4/mmm, 4̄, 422, 4̄2m
Orthorhombic a ̸= b ̸= c, α = β = γ = 90◦ 222, mm2, mmm
Hexagonal a = b ̸= c, α = β = 90◦ , γ = 120◦ 6, 6/m, 6mm, 6/mmm, 6̄, 622, 6̄m2
Trigonal a = b ̸= c, α = β = 90◦ , γ = 120◦ 3, 3m, 3̄, 3̄m, 32
Monoclinic a ̸= b ̸= c, α ̸= γ ̸= 90◦ , β = 90◦ 2, m, 2/m
Triclinic a ̸= b ̸= c, α ̸= β ̸= γ ̸= 90◦ 1, 1̄
Crystal system defines the form of tensor properties according to Neumann’s principle, as
explained in Ch. 2.5. The knowledge of the crystal system is also important for constructing
reciprocal lattice and analyzing diffraction experiments, as we will see in Ch. 4 and 5.
Crystal system can be combined with the lattice centering from Ch. 2.2. Back then, we
mentioned that centering only makes sense when the conventional unit cell has a higher sym-
metry than the primitive cell. Otherwise, lattice vectors could be re-defined without loss of
symmetry. Indeed, a centered triclinic lattice is redundant, because its unit cell can always be
reduced. On the other hand, both face and body-centering occur for cubic and some other crys-
tal systems where symmetry elements would be lost otherwise. Combining crystal systems and
lattice centering gives rise to 14 Bravais lattices that can be encountered in periodic crystals.

3.4. Open symmetry elements

Only point symmetry operations have been considered so far. Crystals also feature translational
symmetry, and they are allowed to have combined symmetry elements that contain both local
transformations and translations within one operation in the sense of Eq. (2.1). Such symmetry
elements are called open because they generate an infinite array out of a single atom.
Importantly, open symmetry elements can only occur in crystals, so they are always ac-
companied by pure translations (Ch. 2.2) and should be compatible with periodicity. Repeated
application of an open symmetry element leads to a simple shift of an atom without any reflec-
tion or rotation. This shift should match one of the lattice translations that have been defined
in Ch. 2.2.
Two types of open symmetry elements should be introduced: glide plane and screw axis.
Glide plane is a mirror reflection followed by a shift t′ . The label of the glide plane indicates
the shift direction:
• a, b, c involve the shift by half of the corresponding lattice vector; for example, t′ = a/2
in the case a
• n involves the shift by half of the face diagonal; for example, t′ = (a ± b)/2
• d involves the shift by one quarter of the face or body diagonal; for example, t′ = (a±b)/4;
it only occurs in the presence of lattice centering, because the shift by 2 t′ should be an
allowed lattice translation
The shift direction is always parallel to the plane. For example, an a glide plane can’t be
perpendicular to the lattice vector a.2 Otherwise, repeated application of the glide plane
would generate a new translational symmetry along a, which is incompatible with the lattice
translations.
Screw axis is a rotation followed by a shift of t′ along this axis. Here, the shift direction is
fixed, but the length of t′ should match the order of rotation. The screw axis np ∥ c involves
the rotation by 360◦ /n followed by the shift of t′ = np c. For example:
2
Note that we have to distinguish italicized symbols (a), which are used for the glide planes, from bold
symbols (a) used for vectors.

13
4

Figure 3.2: Helical chains formed

c by the 31 and 32 screw axes in el-
3 emental tellurium. The numbers 1-
4 2-3-4 show the consecutive action of
the symmetry operator. In the case
3 c
2 of 32 , the missing atoms are added
c through the regular lattice transla-
2 tion by c.
1 1
P 31 P32

• 31 is the 120◦ rotation followed by a shift of c/3

• 42 is the 90◦ rotation followed by a shift of c/2

3.5. Chirality
Screw axes impart their sense of rotation to the atomic arrangement and, therefore, to the
crystal as a whole. For example 31 and 32 produce similar helical chains with the counter-
clockwise and clockwise rotations, respectively (Fig. 3.2). It is an example of chirality.
More generally, chirality is defined as the property of an object not to match its mirror
image. Mirror-plane symmetry renders an object non-chiral because the object does not change
under this operation at all. On the other hand, any symmetry group that does not contain
mirror operations (nor it contains inversion centers and rotoinversion axes, which are equivalent
to the rotation-reflection axes) is chiral. Examples of chiral point groups are 2 2 2 and 3 2. Note
that some symmetry groups are polar and non-chiral or chiral and non-polar. Both polarity
and chirality are rooted in the crystal symmetry, but different aspects of the symmetry are
relevant in each case.
Chiral objects are interesting because they exist in (at least) two forms, known as left and
right enantiomers (for molecules) and enantiomorphs (for crystals). These forms are almost
indistinguishable, yet they may show drastically different properties. For example, many of
the aminoacids, carbohydrates, and other biologically relevant molecules are chiral. Only one
enantiomer is usually present in nature, and this choice of chirality has tremendous repercussions
for biological systems.

3.6. Circular birefringence and dichroism

Chirality has direct implications for optical properties. Previously (Ch. 2.3), we discussed
birefringence as the different refraction of light depending on its polarization relative to the
crystal directions. A counterpart of this effect is dichroism, the different absorption of light
depending on its polarization.3 Crystals with a suitable symmetry may change color on rotation
when illuminated by the linearly polarized light. This is linear dichroism (different absorption
of linearly polarized light) also known as pleochroism in the context of minerals.
A similar effect in chiral crystals is known as circual birefringence and circular dichroism.
Here, one uses left- and right- circularly polarized light and observes their different refraction
3
More generally, birefringence and dichroism describe anisotropy for the real and imaginary parts of the
refractive index, respectively. See also Ch. 12.1.

14
or absorption. Circular birefringence is the manifestation of an optical rotation. The polar-
ization direction turns when light goes through a chiral material. This has been the earliest
experimental tool for identifying chirality. It can be observed in very common systems, such as
crystalline sugar and sugar syrup.

3.7. Space groups

Full symmetry of the crystal is defined by its:
• lattice centering that contains all translations allowed in this crystal (Ch. 2.2)
• point symmetry elements (Ch. 3.1)
• open symmetry elements (Ch. 3.4)
The combination of the three makes a space group. Space groups are labeled by four symbols:
the first one indicates lattice centering, while the three others stand for symmetry elements
along three nonequivalent directions. For example, P nma means:
• primitive lattice
• n glide plane perpendicular to a
• mirror plane (m) perpendicular to b
• a glide plane perpendicular to c
Absent symmetry elements are usually not written. For example, P 2/m = P 1 2/m 1. On the
other hand, P 42 /m b c means
• 42 screw axis along c and a mirror plane (m) perpendicular to c
• b glide plane perpendicular to a (and, respectively, a glide plane perpendicular to b)
• c glide planes perpendicular to a ± b
(check again Ch. 3.2 for the choice of directions in the 4-fold symmetric case).
An exhaustive list of 230 space groups can be found on the Bilbao Crystallographic Server
and in International Tables for Crystallography.4 While 230 may look like a huge number, it
still means that the number of possible symmetries is finite, so it is possible to go through each
of them when necessary. Many of the interesting crystal properties become possible for selected
symmetries only. Then the classification of crystals by their space groups helps one to identify
those specific materials where the property of interest is likely to appear. This approach has
been common in many recent studies, for example here.5
The crystal class is determined by both point and open symmetry elements of the space
group, because open symmetry elements have the same effect on lattice symmetry as the point
ones. For example, P nma belongs to the mmm crystal class and orthorhombic crystal system.
P 42 /mbc belongs to the 4/mmm crystal class and tetragonal crystal system. Both space groups
are non-polar and non-chiral.
One tricky thing about space groups is that their different settings are usually possible. For
example, P mmn, P nmm, and P mnm are all the same space group with the different choices
of a, b, c. Likewise, P 2/a and P 2/c are the same space group, with the a and c axes swapped.
Such an ambiguity is unavoidable, but it has been mitigated by creating a catalog of space
groups in their standard settings, again at the Bilbao server and in related sources. Each space
group has got a unique number, which is often supplied in publications along with the space
4
Many of the universities, including Leipzig, do not have access to this book. Fortunately, symmetry diagrams
for many of the space groups can be also obtained free of charge from the The Fascination of Crystals and
Symmetry website by Frank Hoffmann.
5
Beware that you will need a quite advanced knowledge to understand the content of this paper.

15
group symbol. Crystalloghers have also done a great job in enforcing standard settings and
standard notations for the crystal symmetry. You can take a peek into their very systematic
and perfectly organized world by visiting the website of International Union of Crystallography
with its own dictionary, teaching pamphlets, and a lot more.

16
4. Unpacking the crystal structure
4.1. Atomic coordinates and Wyckoff positions
Atomic positions in a crystal are defined as fractions of lattice translations,

r = xa + yb + zc, (4.1)

where x, y, z are fractional (crystallographic) coordinates. Their convenience in solid-state

physics context will become clear soon (Ch. 5). On the other hand, evaluation of interatomic
distances using x, y, z does become problematic when lattice angles deviate from 90◦ . Whereas
distances can be still calculated by hand, it is usually more practical to use dedicated software
like VESTA that not only draws crystal structures but also gauges distances between the atoms
and even measures bond angles. For atoms within the unit cell, 0 ≤ x, y, z < 1. Any other
atom can be obtained via lattice translations.
Let an atom be placed at a position (x, y, z). Without symmetry, this atom may feel rather
lonely, but a symmetry element will typically generate counterparts of an atom. For example,
inversion center at (0, 0, 0) creates a counterpart at (x̄, ȳ, z̄).6 A two-fold rotation axis, 2 ∥ a,
creates a counterpart at (x, ȳ, z̄) and so on. In this way, symmetry elements generate a Wyckoff
position.
0 1/2 1
y

m
0 Figure 4.1: Selected Wyckoff positions of the
1a
space group P mm2 (No. 25). 4i is the general
2f m position not lying on any symmetry element. 2f
4i and 1a are special positions with the lower mul-
tiplicity. In this figure, we used some standard
1/2
notation of the symmetry diagrams: solid lines
show the mirror planes, whereas gray pointed
ovals are the two-fold rotation axes. You can
find more of the symbols here.
1

x 2-fold rotation

These Wyckoff positions can be general or special. A general Wyckoff position does not lie
on any point symmetry element and shows the highest multiplicity allowed by the given space
group. A special Wyckoff position lies on at least one point symmetry element, which then
does not act on this atom, resulting in a lower multiplicity. For example, the general position
in the space group P mm2 has the multiplicity of 4; it is labeled 4i. By contrast, the position
(0, 0, z) on the two-fold rotation axis has the multiplicity of 1 (it is labeled 1a) because it stays
invariant under all symmetry operations (Fig. 4.1).
Wyckoff positions have two implications. First, they help us to make the description of
crystal structures more systematic. Lists of possible Wyckoff positions are available for every
space group (again, on the Bilbao server or in the International Tables for Crystallography).
Second, the distribution of atoms over different Wyckoff positions has ramifications for crystal
properties, such as the type of phonon modes and their response in infrared and Raman exper-
iments. We will briefly touch on this issue in Ch. 12. You can also learn more by trying the
SAM utility at the Bilbao server.
6
Crystallographers use bars to indicate minus sign in front of the coordinate.

17
4.2. Crystal structure
The structure of a given crystal is described by several elements:
• lattice parameters: a, b, c and α, β, γ
• symmetry (space group) symbol
• list of Wyckoff positions occupied by atoms, with the corresponding coordinates: xj , yj , zj
You will almost never see the full list of atoms in the unit cell. Only the Wyckoff positions (i.e.,
positions unrelated by symmetry) will be given. Other atoms can be generated using symmetry
elements.
Let’s address several beginners’ questions about crystal structures:
Where to find crystal structures? In databases that together contain over a million of
crystal structures that have been determined experimentally. You can use Crystallography
Open Database (free), Karlsruhe database or ICSD (commercial, inorganic compounds), and
Cambridge Structural Database or CSD (commercial, organic compounds). Each of these
databases contains the so-called cif-files that summarize structural information in a common
format.
How to read crystal structures? Experts can learn a lot by reading the cif-file as plain text.
However, most people will find it much easier to draw the crystal structure using software like
VESTA. Then you immediately see all atoms, their positions relative to the unit cell and lattice
translations, etc.
How to understand crystal structures? The answer to this question depends on the in-
formation that you seek to obtain. There are generally three ways of thinking about crystal
structures, sometimes they are called structural models:
• close packing, works for relatively simple compounds, such as monoatomic crystals dis-
cussed in Ch. 1. We will see some further applications in the ionic crystals too (Ch. 7).
• ball-and-stick means that you connect atoms, which are sufficiently close to each other,
and indicate chemical bonds. In this way, one finds connectivity of the crystal, whether
it is built of molecules, chains, layers... This approach is most fruitful in covalent and van
der Waals crystals (Ch. 8).
• polyhedral means that you connect an atom to its neighbors and treat the resulting unit
as a rigid body. Then you analyze connectivity of such polyhedra. This approach is
especially useful in more complex ionic crystals that can not be described as closed-packed
structures.

4.3. Bragg’s law

At first glance, there is not much more in a crystal than what we already considered: the atoms
and their exact positions in space that can be extracted from Wyckoff positions with the use of
lattice parameters and symmetry. However, we can also see this in a different light when shining
light (with a properly chosen photon energy) on a crystal. In fact, most of the information
about the crystal structure comes from scattering experiments where radiation with the short
wavelength (λ ∼ 1 Å in order to match the typical interatomic distances) is used. The simplest
model of this scattering is given by Bragg’s law.
Consider atomic planes separated by a distance d from each other (Fig. 4.2). The waves
reflected by two adjacent planes accumulate the path difference of 2d sin θ where θ is the angle
between the incident beam and the atomic plane. These waves interfere constructively when the
path difference equals an integer number of wavelengths (m), leading to the simple condition
2d sin θ = mλ (4.2)

18
Figure 4.2: Bragg’s law. Reflection of waves
θ θ from parallel atomic planes gives rise to a con-
structive interference when the path difference is
d θ path difference
equal to an integer number of wavelengths. Ex-
perimentally, the angle 2θ between the incident
and scattered waves is measured.

known as Bragg’s law. It is customary to choose m = 1. Higher-order interference maxima

may occur too, but we don’t need to take them into account. In Ch. 5.3 we will see that these
m ̸= 1 maxima are in fact due to the m = 1 scattering but from different planes.
Bragg’s law shows that by scattering waves such as x-rays on a crystal, one expects to see
a series of intensity maxima known as Bragg peaks or simply reflections. Experimentally, the
angle θ can not be measured, but the angle 2θ between the incident and scattered wave is easy
to monitor. The 2θ positions of the Bragg peaks correspond to the distances d that contain
information on the periodicity of the crystal. It is the basics of crystal structure determination,
although by no means a complete procedure yet, as we will see in the following.

4.4. Scattering experiments and x-ray diffraction

The setup described by Bragg’s law is an example of a scattering experiment. One generally
distinguishes
• elastic scattering where the scattered beam has the same energy as the incident beam
• inelastic scattering where some energy is exchanged between the incident beam and the
crystal
Elastic scattering is resolved in the angle 2θ. When Bragg peaks in the sense of Eq. (4.2)
are observed, this experiment is called diffraction. Inelastic scattering can be resolved in both
angle θ and energy transfer E . This experiment is in fact similar to spectroscopy, although
spectroscopy is a somewhat broader term that also includes experiments that are done at the
fixed scattering angle and resolved in energy, only.
Scattering experiments can be done with any kind of waves. We will see such examples later
in these lecture notes, but for now our favorite type will be x-rays because their wavelength is
matched to the typical interatomic distances and lattice parameters. Other types of radiation
can be neutrons and electrons, as we will see in Ch. 6.
X-ray diffraction, better known as XRD, is one of the most common experimental tools in
solid-state research. It is essentially the scattering of monochromatic x-rays (photons with the
single wavelength λ) on a crystalline sample. The scattered radiation is collected by a detector,
and the angle 2θ between the incident and scattered beams is measured. A diffraction pattern
with a series of Bragg peaks is recorded. The positions and intensities7 of the Bragg peaks can
be used for:
• identifying the sample and identifying different chemical compounds in mixed, multi-
phase samples; each crystal structure gives rise to a unique series of Bragg peaks that
serves as a fingerprint of this structure and underlying chemical compound.
• determining orientation of a crystal surface or a thin film
• resolving the crystal structure (more on this in Ch. 6)
7
The role of the intensities will become clear in Ch. 6

19
dhkl

c /l dhkl

a/h d100
b /k d200

Figure 4.3: Left: definition of Miller indices h, k, l. The leftmost plane goes through the origin. An
adjacent plane crosses the lattice vectors a, b, c at da = a/h, db = b/k, and dc = c/l. The distance
between the planes is dhkl . Middle and right: the (200) planes are similar to the (100) planes but
show a twice smaller spacing.

XRD is a much more versatile technique compared to electron microscopy (Ch. 1.3). It does not
destroy the sample and does not require high vacuum. It is compatible with different sample
environments and can be performed in a broad range of temperatures, pressures, electric and
magnetic fields.

4.5. Lattice planes and Miller indices

The concept of atomic or lattice planes used in Bragg’s law requires their proper classification.
Importantly, in the derivation of Bragg’s law one uses not a single lattice plane, but a series of
periodically spaced planes of the whole crystal. Consider now two of these planes, the one that
goes through origin and the plane adjacent to it. The latter crosses lattice vectors a, b, c at the
positions da , db , dc away from the origin (Fig. 4.3). Then Miller indices h, k, l for this family of
atomic planes are defined as

h = a/da , k = b/db , l = c/dc . (4.3)

The Miller index of zero means that the lattice plane never crosses the respective axis, i.e.,
it is parallel to this axis. Three faces of the unit cell will then have indices of (100), (010),
and (001). Increasing the Miller indices shortens the spacing between the planes. For example,
(200) are the same lattice planes as (100), but with the twice shorter spacing (Fig. 4.3).
The Miller indices must be integer to comply with periodicity of the lattice. Later we will
come to see that they also describe a position in the reciprocal space and can be arbitrary in
this sense. However, only integer values of h, k, l describe something that matches periodicity
of the crystal.
The Miller indices can be used to determine the d value in Bragg’s law. For a crystal with
α = β = γ = 90◦ ,
1 h2 k 2 l2
2
= + + . (4.4)
dhkl a2 b2 c2
We will prove this relation shortly (Ch. 5.3). For now let’s explain how the actual experiment
works. An XRD measurement returns a set of peak positions θi that can be re-calculated into
di ’s using Bragg’s law. Then one has to find the values of a, b, c, α, β, γ such that every di is
described by some integer values of h, k, l. This procedure is known as indexing of the Bragg
peaks. A successful indexing returns lattice parameters of the crystal.

20
4.6. Crystal faces and crystal directions
It should be clear from the above that crystal faces (and, consequently, surfaces) are labeled
with the same Miller indices h, k, l. Here, one considers a single plane only, so there will be no
difference between (100), (200), (300), and so on, hence the smallest indices are always used.
The notation of lattice planes and crystal faces should not be confused with the notation
of crystal directions, even though three integer numbers u, v, w are also involved in this case,

R[u v w] = u a + v b + w c. (4.5)

In simple cases, these vectors R are perpendicular to the lattice planes with the same indices.
For example, viewing a cubic crystal along the [100] direction means that you look at the (100)
face of the cube. However, this is not true in general. We will see very soon that the h, k, l
values define a vector in the reciprocal space, whereas Eq. (4.5) corresponds to a real-space
direction. Note also the different type of brackets used for the (lattice planes) vs. [crystal
directions].

21
5. Spaces of crystallography: Reciprocal lattice
5.1. Laue condition
Bragg’s law is so simple that it does not describe all features of the scattering process. It does
not account for the presence of different atoms, nor for the different reflection intensities, and it
treats atomic planes as simple mirrors that reflect the incident beam in one direction. A more
realistic description has been proposed by Laue who considered interference of waves scattered
on two different atoms inside the crystals.

k
n’
Figure 5.1: Laue condition of the construc-
n
tive interference. The waves scattered by two
atoms, which are separated by the lattice vec-
tor R, acquire a path difference that should
θ R be a multiple of the wavelength λ. The uni-
θ’ tary vectors n and n′ show the directions of
k and k′ , the propagation vectors of the in-
path cident and scattered waves, respectively.
difference
k’

Without loss of generality, we consider k and k′ , the propagation vectors of the incident
and scattered waves, respectively. If two atoms are separated by a vector R, the path difference
between the two scattered waves becomes (Fig. 5.1)

R cos θ + R cos θ′ = R (n′ − n) (5.1)

where n and n′ are unitary vectors along k and k′ , respectively. Using the definition of the
propagation vector, k = (2π/λ) n and k′ = (2π/λ) n′ , the interference condition can be written
as
λ
R(k′ − k) = mλ ⇒ R (k′ − k) = 2πm (5.2)
2π
with integer m. This equation is known as the Laue condition. It should hold for every lattice
vector R defined by Eq. (1.1), because a pair of atoms can be found for every such vector.

5.2. Reciprocal lattice

The vectors k′ − k satisfying the Laue condition form a reciprocal lattice of the crystal. One
defines the reciprocal lattice as a family of vectors G that fulfill

eiGR = 1 (5.3)

for every lattice vector R. It is common to say that R’s belong to the direct lattice of the
crystal in real space, whereas G’s occur in the reciprocal space.
Reciprocal lattice can be constructed explicitly by choosing the vectors

[b × c] [c × a] [a × b]
a∗ = 2π , b∗ = 2π , c∗ = 2π . (5.4)
a · [b × c] a · [b × c] a · [b × c]

Then it is easy to verify that

a∗ a = 2π, a∗ b = 0, a∗ c = 0 (5.5)

22
c (100) b Figure 5.2: Direct-lattice and reciprocal-
c*
lattice vectors for the orthogonal (right)
b*
and non-orthogonal (left) lattice vectors.
β Note that the reciprocal-lattice vectors are
a a not always parallel to those of the direct
a* a*
lattice, yet they are perpendicular to the
d10
0
corresponding lattice planes: for example,
a∗ is perpendicular to the (100) planes.

and so on, such that any vector

G = ha∗ + kb∗ + lc∗ (5.6)
with integer h, k, l satisfies the Laue condition, Eq. (5.2).
It is almost trivial to construct the reciprocal lattice when lattice vectors a, b, c are mutually
orthogonal. In this case, a · [b × c] = abc and [b × c] ∥ a. Then a∗ ∥ a and |a∗ | = 2π/a, hence
the name reciprocal lattice. Its directions are parallel to those of the direct lattice (Fig. 5.2),
whereas reciprocal-lattice parameters are inverse of the lattice parameters in real space, times
the pre-factor of 2π. This holds true for the orthorhombic, tetragonal, and cubic symmetries.
Other symmetries are more involved. Consider a monoclinic lattice with β ̸= 90◦ . Then,
a · [b × c] = abc sin β, and one finds
2π 2π 2π
|a∗ | = , |b∗ | = , |c∗ | = . (5.7)
a sin β b c sin β

Now, a∗ and c∗ are no longer parallel to a and c (Fig. 5.2). Instead, they form an angle of
β ∗ = 180◦ − β.
We should also note that a · [b × c] = V , the unit-cell volume in real space according to the
standard definition of the triple product. To obtain unit-cell volume in the reciprocal space,
one needs vector identities,

a1 · [a2 × a3 ] = a2 · [a3 × a1 ] and [a1 × [a2 × a3 ]] = a2 (a3 · a1 ) − a3 (a1 · a2 ), (5.8)

that return
2π 2π ∗ (2π)3
V ∗ = a∗ · [b∗ × c∗ ] = [b × c] · [b∗ × c∗ ] = b · [c∗ × [b × c]] = . (5.9)
V V V
This result will be needed in the future when we do summations over the reciprocal space.
At this point, reciprocal lattice may still look like a rather arbitrary mathematical con-
struction, but it is in fact undeniably real because it is observed in every scattering experiment:
intensity maxima appear at every or almost every (see Ch. 6.3) reciprocal-lattice site. We can
also say that each crystal lives two parallel lives. One is in real space and constitutes tangible
crystal properties. Another one is in the reciprocal space and incorporates intrinsic effects such
as atomic vibrations and electronic transitions. Only by grasping reciprocal-space phenomena
can one understand real-space properties of the crystal!

5.3. Relation to Bragg’s law

The Laue condition and reciprocal lattice have a direct connection to Bragg’s law. In fact,
every reciprocal-lattice vector defined by Eq. (5.6) corresponds to the lattice planes with the
Miller indices h, k, l. To verify this statement, choose the lattice planes (hkl) and define n as

23
b

Figure 5.3: Relation between the recipro-

n
G dhkl cal lattice and Bragg’s law. The vector G
chosen perpendicular to the lattice planes
is a reciprocal-lattice vector. The Miller
|G| = 2π/d indices h, k, l define the representation of
d b/k G in terms of a∗ , b∗ , c∗ .
φ
r=0 a
a /h

the unitary vector perpendicular to these planes. Now choose G = 2πn/dhkl and consider it as
a propagation vector of a wave with the wavelength of dhkl . This wave has the form eiGr . One
expects eiGr = 1 at r = 0. Since r = 0 corresponds to one of the lattice planes, the condition
eiGr = 1 should also hold on any other lattice plane because they are separated by an integer
number of wavelengths (Fig. 5.3). Each point R of the direct lattice belongs to one or another
lattice plane. Therefore, eiGR = 1 for every R, and the condition (5.3) is fulfilled. Then G
must be a reciprocal-lattice vector.
Now we have to prove that the reciprocal-lattice vector G = ha∗ + kb∗ + lc∗ corresponds to
the planes with the Miller indices h, k, l. Let this vector define some lattice planes, which are
perpendicular to it. We know that the corresponding interplane distance is given by |G| = 2π/d.
The distance between the origin and the nearest plane is a vector of the length d in the direction
of G (Fig. 5.3), namely, d = d G/|G| = d2 G/(2π). To determine the distance da that this
plane intersects on the a-axis, one has to compute
d d 2πa 2πa a
da = = = = ∗
= , (5.10)
cos φ d · a/(da) G·a ha · a h

which is equivalent to Eq. (4.3) that served as the definition of the Miller indices h, k, l.
We now realize that the lattice planes used in Bragg’s law are real-space manifestations of
the reciprocal lattice. Incident light sees crystal as an optical grating, with different gratings
identified by different families of the lattice planes. There need not be atoms on a given lattice
plane to produce a Bragg peak. The positions of these Bragg peaks define the reciprocal lattice
and directly convey lattice parameters of the crystal.
This statement also sheds light on Eq. (4.4) for dhkl that immediately follows from |G|2 =
h2 |a∗ |2 + k 2 |b∗ |2 + l2 |c∗ |2 as the vector length in the reciprocal space (assuming a, b, c are
mutually orthogonal). The calculation of dhkl for an arbitrary lattice also becomes straight-
forward, albeit tedious when non-90◦ angles have to be taken into account.

5.4. Aperiodic crystals

We can also see reciprocal lattice as a Fourier transform of the direct lattice. Bragg peaks at
the reciprocal-lattice points indicate periodicity of the crystal in real space. This statement
is in fact much more general, because aperiodic crystals also show characteristic diffraction
patterns with a series of Bragg peaks. Such peaks manifest the underlying long-range order of
the aperiodic crystal. The difference from periodic crystals is that the reciprocal lattice spun
by a∗ , b∗ , and c∗ is no longer sufficient to describe the Bragg peaks.
Many of the aperiodic crystals can be seen as modulated structures. Their diffraction pat-
terns are described by a 3D reciprocal lattice plus one or more additional vectors ti known as
modulation vectors. Mathematically, these ti ’s can be still decomposed into reciprocal-lattice

24
vectors,
ti = p1 a∗ + p2 b∗ + p3 c∗ . (5.11)
When p1 , p2 , p3 are simple fractions like 21 or 15 , the structure is called commensurately modulated.
It is nothing but a periodic crystal with the larger unit cell. For example, p1 = 12 would mean
that a∗ should be twice shorter, hence the lattice parameter a should be twice longer: some
element of the structure develops a twice longer periodicity than the rest of the crystal and
requires a two-fold expansion of the unit cell. By contrast, a structure with random values of
p1 , p2 , p3 is incommensurately modulated.
Special mathematical formalism has been developed for modulated structures. It is based
on a proper (periodic) lattice defined in a 3 + n-dimensional space known as superspace where
modulation vectors ti are accommodated along additional, artificial dimensions, in order to
restore periodicity of the system. For example, quasicrystals can be described as periodic
structures in the 5D or 6D reciprocal space. This space is, of course, unphysical, but its
3D projection forms the physical reciprocal space where diffraction pattern is observed. The
introduction of additional dimensions may seem bizarre at first glance, but it becomes more
palatable if one considers that crystallographic restriction theorem forbids 5-fold symmetry
only in 3D space. In higher dimensions, five-fold rotations may be compatible with periodicity.
Incommensurate modulations may have different origin depending on the chemical nature
of the crystal. Sometimes it is related to deformations or rotations of structural units that
occupy a certain position in space, but vary with a different periodicity compared to the rest
of the lattice. Other examples of aperiodic crystals are framework structures with channels
where atoms in channels have their own periodicity compared to the framework. Aperiodic
crystals are inconspicuous but abundant. They may occur in elemental solids, including Bi and
Te under pressure, or in such a common material as Na2 CO3 , better known as washing soda.
A short memorial article can serve as a good introduction into aperiodic crystals.

25
6. Structure factor: All shades of diffraction
6.1. Structure factor
Reciprocal lattice serves as a connection between positions of Bragg peaks and lattice parame-
ters of the crystal. Similarly, structure factor connects intensities of Bragg peaks to the atomic
coordinates.
To calculate intensity of a Bragg peak, we go back to Laue’s representation of the scattering
process and take the scattering mechanism into account. Atoms are not point-like objects, and
atoms are not scatterers per se. Instead, one should consider periodic scattering density ρ(r)
that satisfies ρ(r + R) = ρ(r) for every lattice vector R.
Every small volume element dr scatters the incident wave. The phase of the scattered wave
depends on the position of dr inside the crystal. Indeed, from Laue’s argument (Ch. 5.1) the
path difference for a given position r relative to the origin is
λ λ
r(k − k′ ) = Gr (6.1)
2π 2π
and the corresponding phase shift is ∆φ = G r where G is a reciprocal-lattice vector. The
scattered wave is obtained by integrating such phase shifts over the unit cell, taking into
account ρ(r) as the concentration of scatterers in a given position r,
Z
i(k′ r̃−ωt) ′
e · ρ(r) eiGr dr = ei(k r̃−ωt) F (G) (6.2)

where the newly introduced structure factor F (G) is the amplitude of the scattered wave that
travels away from the crystal into some arbitrary position r̃. Intensity is proportional8 to the
squared amplitude, Z
I(G) ∼ |F (G)|2 , F (G) = ρ(r) eiGr dr. (6.3)
UC

Considering the periodicity of ρ(r), it is sufficient to integrate over the unit cell (UC) when
calculating F (G), because other unit cells will simply lead to an additional pre-factor, which
should be the same for every reflection.9
It is instructive to compare the structure factor, Eq. (6.3), with the Fourier decomposition
of the scattering density,
Z
X
iGr 1
ρ(r) = ρG e , ρG = ρ(r) eiGr dr, (6.4)
G
V

where we restricted the summation to the reciprocal-lattice vectors G because ρ(r) is periodic
in direct space. In fact, structure factor is merely the Fourier component of the scattering
density, up to a pre-factor.

6.2. Atomic approximation

Strictly speaking, it is not necessary to think of atoms when we calculate F (G), but life becomes
easier when we do, namely, when the scattering density ρ(r) is represented as a superposition
8
Experimentally, intensity depends on the exposure time and can be defined up to a scale factor, only.
9
An adverse consequence of this multiplication is that F (G) and I(G) go to infinity in thermodynamic
limit (for an infinite crystal). This is the drawback of the kinematic theory that neglects multiple scattering,
although, as a matter of fact, a scattered wave should scatter again, and again, and again when the crystal is
infinite. Such a multiple scattering is taken into account in the dynamic theory of diffraction that we will not
consider here because for many practical purposes the kinematic theory suffices.

26
5
1 r’ Figure 6.1: Scattering density of the crys-
r₁ 3 tal, ρ(r), is represented as a superposi-
r tion of the scattering densities of individ-
ual atoms, ρj (r′ ), with the vector r de-
2 composed into r1 + r′ .
4 6

of scattering densities ρj (r′ ) from individual atoms. Consider j = 1, . . . N atoms located at the
positions r1 , . . . rN within the unit cell. By introducing the volume element dr′ within an atom,
we represent r = rj + r′ (Fig. 6.1) and re-write the structure factor as a sum over individual
atoms,
Z N Z N
iGr′
X X
iGr iGrj ′ ′
F (G) = ρ(r) e dr = e ρj (r ) e dr = fj (G) eiGrj . (6.5)
j=1 atom j=1

Here, we introduced the atomic form factor fj (G) that describes the scattering of the wave by
a given atom j in the direction of G.
To a first approximation, any given atom scatters waves in the same way regardless of the
chemical environment of this atom. Then the same set of atomic form factors can be used for
all crystals, and the variation of the Bragg peak intensity with G arises from two ingredients:
• material-specific: positions of atoms rj within the unit cell
• generic: angular dependence of the scattering by a given atom, as defined by fj (G),
regardless of the exact nature of the crystal

6.3. Extinction conditions

The structure factor defined by Eq. (6.3) reflects the arrangement of atoms inside the crystal.
Experimental structure factors can be used to determine atomic positions, as we will further
discuss in Ch. 6.4. Interestingly, using only a simple maths one can draw some conclusions
on the symmetry of the crystal when lattice centering (Ch. 2.2) or open symmetry elements
(Ch. 3.4) connect different atoms therein.
i) Lattice centering. Consider a body-centered crystal structure with the lattice translation
of t = 12 (a + b + c). Then an atom located at (x, y, z) has a counterpart at (x + 12 , y + 21 , z + 12 ).
Using the standard representation G = ha∗ + kb∗ + lc∗ , one finds

G rj = (ha∗ + kb∗ + lc∗ )(xj a + yj b + zj c) = 2π(hxj + kyj + lzj ) (6.6)

(note an immediate advantage of fractional atomic coordinates in conjunction with the recip-
rocal lattice!) We can now write the structure factor as
N N/2 h i
1 1 1
fj (G) e2πi(hxj +kyj +lzj ) + e2πi[h(xj + 2 )+k(yj + 2 )+l(zj + 2 )]
X X
iGrj
F (G) = fj (G) e =
j=1 j=1
N/2
X
fj (G) e2πi(hxj +kyj +lzj ) 1 + eπi(h+k+l)

= (6.7)
j=1

The expression in square brackets can be either 0 or 2 depending on whether h + k + l is odd

or even, respectively. We thus see that lattice centering defines a special rule for the reflection

27
intensities. Only reflections with h + k + l = 2n (even) are observed. This is the reflection
condition for a body-centered crystal. The systematic absence of reflections is known as an
extinction, so we could also introduce an extinction condition h + k + l = 2n + 1 (odd).
ii) Open symmetry elements. Glide planes and screw axes cause their own extinctions.
Consider the glide plane a ⊥ c. It leads to the transformation (x, y, z) → (x + 21 , y, z̄). Then
the structure factor becomes
N/2 h i
2πi[h(xj + 21 )+kyj −lzj ]
X
2πi(hxj +kyj +lzj )
F (G) = fj (G) e +e . (6.8)
j=1

This expression does not look very spectacular, but it becomes simpler in the special case of
l = 0. Then,
N/2
X
fj (hk0) e2πi(hxj +kyj ) 1 + eπih

Fhk0 = (6.9)
j=1

and we expect the reflection condition hk0, h = 2n for the a-glide plane perpendicular to c.
Reflection conditions for a given space group can be obtained from the HKLCOND utility
at the Bilbao Server. By analyzing extinctions, crystal symmetry can be determined or, more
precisely, narrowed down to a list of possible space groups.
More generally, one can see the reflection (extinction) conditions as a manifestation of the
periodicity. The reciprocal lattice defined using three lattice vectors, Eq. (5.4), corresponds to
the situation when no additional translations are allowed. Crystals with lattice centering have
a shorter periodicity indicated by the primitive cell, so their reciprocal lattices should be more
sparse. This effect is achieved by removing some of the reciprocal-lattice points via extinctions.
For example, reciprocal lattice of a bcc lattice is an fcc lattice with the periodicity of 4π/a.

6.4. Crystal structure determination

Crystal structures are determined from diffraction experiments that involve an accurate mea-
surement of both position and intensity for every Bragg peak. The structure solution includes
three consecutive steps:
• indexing of the Bragg peaks, finding lattice parameters from the peak positions
• analyzing extinctions in order to determine the crystal symmetry10
• finding a structural model that gives the best fit to the experimental Bragg peak intensities
The principal advantage of the diffraction methods is that many Bragg peaks can be mea-
sured in order to determine atomic positions with a very high accuracy. A typical diffraction
experiment covers thousands of Bragg peaks, and interatomic distances are determined with
an uncertainty of less than 0.01 Å. Such a high spatial resolution is only possible with the
reciprocal-space technique (measurement of Bragg peaks in the reciprocal space) and can not
be matched by the direct-space imaging with electron microscopy (Ch. 1.3).

6.5. Atomic form factors

From Eq. (6.5), atomic form factor is defined as the Fourier-transformed scattering density ρ(r)
of a single atom, Z
f (q) = ρ(r) eiqr dr (6.10)

10
More precisely, lattice centering can be determined, and a subset of possible space groups defined. Space
groups that contain neither lattice centering nor open symmetry elements do not have any reflection conditions
and can not be distinguished in this way.

28
ρ(r) f(q)

neutron
(nuclear)
electron density

x-ray

nuclear spin density

neutron
(magnetic)
r q

Figure 6.2: Scattering densities (left) and the resulting atomic form factors (right) for the x-ray and
neutron scattering. Different densities are not drawn to the scale. Spin density is typically associated
with the d- and f -atomic orbitals that have nodes at r = 0.

where q is a vector in the reciprocal space (the vectors q are no longer restricted to the
reciprocal-lattice vectors G, because atom is not periodic). Spherical symmetry of an atom
implies that f depends only on |q| and not on the direction of q. It is then customary to
represent |q| as 2π/d similar to Ch. 5.3 and consider 1/d ∼ sin θ/λ from Bragg’s law. Therefore,
atomic form factors are tabulated as a function of sin θ/λ, measured in Å−1 . They can be found
in the International Tables for Crystallography and on the web.
The scattering density in Eq. (6.10) depends on the nature of the wave being scattered.
Several cases are of special importance (Fig. 6.2):
• x-rays are scattered by electrons, so ρ(r) is electron density that gradually decreases
away from the atom. Then f (q) is also a slowly decreasing function. Larger values of
|q| correspond to smaller values of dhkl that, in turn, correspond to the higher angles θ.
Therefore, one sees lower intensities in XRD patterns at high angles
• neutrons are scattered by the atomic nuclei, which are very small. Then, ρ(r) looks
more like a δ-function, while its Fourier transform is a constant. In a neutron diffraction
experiment, reflection intensities only weakly depend on the angle θ.
• neutrons are also scattered by the spin density, namely, by valence electrons with an
unpaired spin. This is the primary experimental tool for studying magnetic order in
crystals. The corresponding magnetic form factor decreases even faster than the x-ray
one because only valence electrons are involved, so magnetic Bragg peaks are observed at
low angles only
• electrons are scattered by other electrons, so their f (q) is similar to that of x-rays. The
intensities decrease with the angle. However, electrons are impatient and will usually
scatter several times as they travel through the crystal. Intensities of Bragg peaks in
electron diffraction are, therefore, strongly affected by multiple scattering.

6.6. Neutron diffraction

Neutrons can be an alternative to x-rays in the diffaction experiments. The main advantage of
neutrons is their sensitivity to light elements. Since x-rays are scattered by electrons, the x-ray
atomic form factor scales with the number of electrons, i.e., the atomic number Z. Then the
quadratic scaling of the reflection intensity, I ∼ Z 2 , leads to a very sharp difference between
the scattering from heavy and light elements. Ultimately, hydrogen atoms scatter so weakly
that they become hard to locate using x-rays.
Neutrons are scattered by the atomic nuclei with no systematic dependence on Z, so they
can be used to locate light elements in the crystal structure. Another advantage is the sensitivity

29
of neutrons to magnetic order, the feature that x-rays lack.11
Neutron diffraction experiments have major downsides too. Neutrons can not be produced
in the lab, they are obtained either from a nuclear reactor or from a so-called spallation source.
Both are huge installations, and only a handful of neutron research facilities exist around the
world. Neutron beams are hard to focus, so large instrumentation is needed, and sample volume
is usually much larger than in the case of x-rays. Tiny objects like thin films are hard to study
using neutrons, and spatial resolution is much lower.

11
Unless experiments are performed at the x-ray energy that matches the absorption edge, see resonant x-ray
scattering.

30
7. Bonding in crystals: ionic
7.1. Types of chemical bonding
Many of the crystal properties can be understood from the perspective of bonding between the
atoms. Four main types of chemical bonding are usually distinguished:
• ionic, due to electrostatic forces between charged units (ions)
• covalent, due to an overlap of atomic orbitals
• metallic, due to shared (itinerant) electrons
• van der Waals, due to dipole-dipole and other weak interactions, such as London disper-
sion force (between induced dipoles)
One often identifies a separate group of hydrogen bonds, which occur between atoms with
small, fractional charges, such as H and O in water and organic molecules. These bonds are
somewhat stronger than the typical van der Waals bonds, yet for our purpose they fall into
the same group of weak bonding. We will also discuss molecular crystals where finite units
(molecules) formed by covalent bonds are held together by van der Waals bonds.

7.2. Cohesive and lattice energies

Stability of the crystal is characterized by its energy. In this context, two closely related energies
can be defined. Cohesive energy is the energy required to break the crystal into individual
atoms. Lattice energy is the energy required to split the crystal into its constituents.12 The
cohesive and lattice energies are equal for metallic crystals and for simple covalent crystals like
diamond. However, they are very different in molecular crystals where molecules formed by
covalent bonds are held together by van der Waals bonds. In this case, only lattice energy is the
relevant energy scale because it shows the energy cost of breaking the crystal into individual
molecules (think of sublimation of iodine into a gas of I2 molecules). A somewhat similar
situation occurs in ionic crystals. Compare

AB(s) + E coh −→ A(g) + B(g) ,

−
AB(s) + E lat −→ A+
(g) + B(g)

where the subscripts (s) and (g) denote the solid and gas states, respectively.
Lattice energies are obtained from calorimetry experiments where the amount of heat re-
leased or absorbed in a given process is measured. Sublimation enthalpy of a metallic, covalent,
or molecular crystal is a direct measure of its lattice energy.13 More often, though, the quan-
tity measured in the experiment is the formation enthalpy ∆f H, which is neither cohesive nor
lattice energy of the crystal:

Na(s) + 12 Cl2 (g) −→ NaCl(s) + ∆f H

This formation enthalpy can be related to the lattice energy by considering all intermediate
processes (Fig. ??) and adding up relevant energies: sublimation energy of the Na metal,
ionization potential of the Na atom, bond dissociation energy of the Cl2 molecule, and electron
12
Older textbooks, including the one by Ashcroft and Mermin define cohesion energy simply as lattice energy
and describe the formation of crystal from its constituents as cohesion.
13
In thermochemistry, enthalpies are commonly used instead of energies because measurements are performed
at a constant pressure rather than constant volume. Then, strictly speaking, lattice enthalpy is obtained. This
difference is usually unimportant, since solids show only a minor difference between energy and enthalpy owing
to the small thermal expansion, see also Ch. ??.

31
E tot

Figure 7.1: Crystal energy E tot as a func-

tion of the interatomic distance r. Lattice
r₀ parameters define the equilibrium value
r
r0 , while the depth of the energy minimum
is the lattice energy E lat .
E lat

affinity of the Cl atom. All these quantities can be measured in separate experiments, and
eventually lattice energy of NaCl can be obtained from ∆f H of NaCl. This method is known
as the Born-Haber cycle. Thermochemistry data can be found in multiple sources, such as
NIST Chemistry WebBook and CRC Handbook of Chemistry and Physics.
It is also helpful to represent crystal energy as a function of an effective interatomic distance,
E (r), as shown in Fig. 7.1. Then the position of the energy minimum yields the equilibrium
distance r0 and, consequently, the lattice parameter of the crystal. The depth of the minimum
is the lattice energy E lat .

7.3. Madelung constant

We will now concentrate on the ionic crystals and try to calculate E (r). To this end, we assume
that the interatomic forces are Coulomb in nature and determine the electrostatic potential at
a given site i,
e X zj
Vi = (7.1)
4πε0 j rij
where zj is the ionic charge and rij is the distance between ions. This summation can be
replaced with
e αi
Vi = (7.2)
4πε0 r
where r stands for the nearest-neighbor distance and αi is the Madelung constant for the
site i. This Madelung constant is defined for a given structure type and ionic charge and does
not depend on the exact lattice parameter. Getting the actual values of αi is far from trivial
because Coulomb interactions are long-range, and the series defined by Eq. (7.1) is conditionally
convergent.14 One practical approach to this problem is the Ewald summation.
Coulomb energy of the crystal can be then written as
X X e2 zi αi
1 1
E Coul = 2
ezi Vi = 2
(7.3)
i i
4πε0 r

with the pre-factor 21 to avoid double-counting. This summation will include several terms
according to the number of different atomic positions in the crystal. In a simple AB crystal,
symmetry requires that α+ = α− = α. Using z+ = −z− = z, one finds
αz 2 e2
E Coul = − (7.4)
4πε0 r
14
Some instructive examples of the conditional convergence of electrostatic energy in ionic crystals can be
found here and here.

32
This energy does not have a minimum, and indeed an ionic crystal with only Coulomb
forces should collapse. An energy minimum in the vein of Fig. 7.1 appears when Coulomb
energy is augmented by a repulsive energy,
αz 2 e2 Crep
E tot (r) = − + m (7.5)
4πε0 r r
where Crep = const and m is also a constant that takes values between 6 and 10 depending on
the ions. This repulsive potential is empirical in nature and mimics the intuitive understanding
that different atoms cannot penetrate into each other. Beyond this common intuition, the
repulsion goes back to the Pauli exclusion principle, as explained here.

7.4. Born-Landé equation

Equilibrium distance r0 corresponds to the energy minimum of E tot (r) from Eq. (7.5),
dE tot αz 2 e2 Crep m−1 mCrep αz 2 e2 m−1
=0 ⇒ − m = 0 ⇒ r 0 = 4πε0 ⇒ C rep = r (7.6)
dr 4πε0 r02 r0m+1 αz 2 e2 4πε0 m 0
It can be used to calculate lattice energy,
αz 2 e2 αz 2 e2 1 αz 2 e2

1
E lat = −E tot (r0 ) = − = 1− , (7.7)
4πε0 r0 4πε0 r0 m 4πε0 r0 m
resulting in the Born-Landé equation for ionic crystals.
This equation has some immediate implications:
• lattice energy of an ionic crystal increases with the ionic charge; therefore, oxides (z = 2)
usually have much higher melting points than halides (z = 1)
• lattice energy of an ionic crystal decreases with the interatomic distance (r0 ); therefore,
melting points decrease from NaF to NaI and from NaF to CsF and correlate with the
lattice parameter
Typical lattice energies of ionic crystals are in the range of 5 − 10 eV/f.u. Born-Landé equation
can also be used to calculate lattice energy for the known interatomic distance, which is usually
determined by XRD. The unknown parameter m enters the energy as 1/m and changes the re-
sult only marginally. This parameter can be determined when another experimental observable,
such as bulk modulus, is available.

7.5. Bulk modulus

The isothermal bulk modulus is defined as

∂p
B = −V (7.8)
∂V T

It shows the change in the crystal volume under pressure (more on this in Ch. 9.1). The
bulk modulus can be obtained from E tot if one considers thermodynamic definition of pressure,
p = −(∂E /∂V )T . The bulk modulus is essentially the second derivative of E tot with respect
to V . In contrast to lattice energy, Eq. (7.7), the exact expression for the bulk modulus depends
on the structure type, which defines the relation between V and r in the crystal.
Let’s choose rocksalt-type structure (Ch. 7.6) as an example. The distance r is the separa-
tion between the cation and anion, r = a/2. Then, V = a3 /4 = 2r3 (volume per formula unit)
and
d 1 d d dE tot r d 1 dE tot
= 2 ⇒ B=V = . (7.9)
dV 6r dr dV dV 18 dr r2 dr

33
This value should be taken at r = r0 where dE tot /dr = 0. Then, the equation simplifies to

1 d2E tot m − 1 αz 2 e2
B= = (7.10)
18 r0 dr2 r=r0 18 4πε0 r04

where we used Eq. (7.6) to express Crep via r0 .

Clearly, bulk modulus shows the same trends as lattice energy. Oxides are less compressible
than halides, that is, they feature higher bulk moduli. Compressibility of an ionic crystal
increases upon increasing its lattice parameter. By measuring both B and r0 , the value of
m and, eventually, the exact lattice energy can be determined. Typical bulk moduli of ionic
crystals are in the range of 10 − 200 GPa.

7.6. Ionic radius and Pauling’s rule

Whereas the Born-Landé equation, Eq. (7.7), offers a simple explanation for changes in lattice
energy, melting points, and bulk moduli across different ionic crystals, it fails to answer one
fundamental question: which structure type is chosen by a given ionic material? The Madelung
constants for different structure types can be compared, but how to deal with the distance r0 ?
Typical values of r0 can be obtained from the so-called ionic radii that have been estimated
for every possible ion of every chemical element by analyzing the statistics of interatomic
distances in thousands of crystal structures determined experimentally. One assumes r0 ≃
r+ + r− and finds the optimal values of r+ and r− that fit the experimental r0 in different
materials. Different sets of ionic radii exist, the most common one being the system by Shannon
and Prewitt quoted in multiple handbooks, for example here. Because the ionic radii are
determined from the experimental data, they take different values depending not only on the
ionic charge but also on the coordination number. Increasing the charge and/or the coordination
number reduces ionic radius of a cation. The opposite is true for anions, which are generally
larger than cations because adding an electron requires extra space.
Ionic radii are often sufficient to analyze structure types of different ionic crystals. Pauling’s
first rule15 postulates that the coordination number of a cation in an ionic crystal is determined
by the ratio r+ /r− . This rule goes back to a simple geometrical argument that anions should not
come too close to each other. Therefore, smaller cations require lower coordination numbers.
The threshold values of r+ /r− have been derived as follows:

CN polyhedron r+ /r− larger than

3 triangle 0.155
4 tetrahedron 0.225
6 octahedron 0.414
8 cube 0.732
The most common coordination numbers are 4 and 6. Representative structure types are
zinc blende and rocksalt shown in Fig. 7.2. Very large cations like Cs+ may form CsCl-type
structures with the cubic coordination.
Pauling’s consideration of the ionic radii inspires polyhedral description of ionic crystals.
Anions form the polyhedron around a cation. By defining these polyhedra, the connectivity of
the structure can be analyzed.
15
Four other Pauling’s rules are even more empirical and can be found here.

34
Zn
Cl Cs
S
Na Cl

zinc blende (CN = 4) rocksalt (CN = 6) CsCl-type (CN = 8)

Figure 7.2: Common structure types of ionic crystals: zinc blende (fcc lattice), rocksalt (fcc lattice),
and CsCl-type (primitive cubic lattice).

7.7. Ionic crystals as close-packed structures

Structures of ionic crystals can be also analyzed from the perspective of close packing. Anions
form close-packed layers because they are usually bigger than cations. Then cations fill octahe-
dral and tetrahedral voids between these close-packed layers. There is one octahedral and two
tetrahedral voids per anion.
Let anions form the cubic close packing:
• filling octahedral voids yields the rocksalt structure (NaCl)
• filling 21 of the tetrahedral voids yields the zinc blende structure (ZnS, CdTe)
• filling all tetrahedral voids yields the fluorite structure (CaF2 ). Note that in fluorite
cations form the close-packed structure, whereas anions occupy the tetrahedral voids. The
situation is reversed in antifluorite (Na2 O) where anions form the close-packed structure
and anions are in the voids.
• filling octahedral and 12 of the tetrahedral voids yields the spinel structure (MgAl2 O4 )
A similar construction is possible for the hexagonal close packing of anions too:
• filling octahedral voids yields the NiAs structure (MnS)
• filling 12 of the tetrahedral voids yields the wurtzite structure (ZnS, AgI)
This simple principle describes many of the common structure types of ionic compounds and
also elucidates their hexagonal or fcc symmetries. However, it is not always easy to understand
which structure type is preferred by a given compound. One compound may form different
structures as in the case of ZnS. This is an example of polymorphism.

35
8. Bonding in crystals: covalent, metallic, and van der Waals
8.1. Covalent crystals
The formation of covalent bonds can be probed using the covalent radius.16 When rAB ≤
rA +rB , a covalent bond occurs between the atoms A and B. Shorter distances indicate stronger
covalent bonds. This is well known from carbon in organic molecules where single (1.54 Å),
double (1.34 Å), and triple (1.20 Å) C–C bonds can be distinguished by their typical length and
chemical environment. Covalent bonds in solids are not always identifiable as single, double,
or triple, but the same trend holds.
Complex nature of the covalent bonds does not allow a generic description in the same vein
as Eq. (7.5). We can only say that shorter covalent bonds should lead to the increased lattice
energy and higher bulk modulus. The typical lattice energies of covalent crystals are on the
order of 10 eV/atom and the bulk moduli are in the range of 50 − 200 GPa.

8.2. Metallic crystals

There is not much to say about metallic crystals at this stage. We will address them in more
detail in Ch. ??. For now let’s mention without any explicit derivation that lattice energy of a
metallic crystal is given by a rather non-intuitive expression

3 ℏ2 2
E lat = (3π 2 ne ) 3 , (8.1)
10 me
so it depends on the electron concentration ne and decreases with increasing the lattice param-
eter.
Metals can have lattice energies as low as 1 eV/atom and feature bulk moduli of several GPa
only. Some of the elemental metals have melting points close to room temperature (cesium,
gallium), whereas mercury is even a liquid. Nevertheless, other metals like tungsten are char-
acterized by very high melting points and low compressibilities, exceeding those of the typical
ionic crystals. It all depends on the electron concentration.

8.3. Polymorphism
In elemental solids, the type of the crystal structure will be usually determined by chemical
bonding. Metals tend to adopt simple and dense structures (see Ch. 1.2), non-metals will
often develop similar structures but with the packing of weakly bonded atoms (noble gases)
or diatomic molecules (hydrogen, oxygen). Elements spanning the boundary between metals
and non-metals show the most diverse behavior, as they can form different types of covalently-
bonded structures and simultaneously appear in the metallic form. Tin is a celebrated example.
It forms metallic crystals of white tin (body-centered tetragonal structure) that abruptly trans-
form into non-metallic gray tin (diamond structure) on cooling. These two forms of tin are
called allotropes. Allotropes are also known for carbon, sulphur, phosphorous.
Binary and more complex compounds will usually stick to one type of bonding and differ
only in their structural details, although these details are by no means unimportant. They can
have major implications for observed properties. One usually distinguishes:
• polymorphs as different structural forms of the same chemical compounds. They are
labeled with Greek letters: α-Sn, β-Sn, etc. Strictly speaking, allotropes can be also
considered as polymorphs, as we just did for tin, and so all researchers do, especially
16
The term atomic radius is somewhat more loosely defined, but most often it implies the covalent radius.

36
when they deal with high-pressure structures of elemental solids. The somewhat old-
fashioned word “allotrope” is basically reserved for different forms of elemental solids
observed at ambient pressure.
• enantiomorphs as left and right forms of a chiral structure (see Ch. 3.5)
• polytypes as different structures arising from different stacking sequences. Polytypes are
common in layered structures like graphite, but they could also occur in structures derived
from close-packed layers, as in Ch. 7.7. Labels of polytypes include the number of layers
per unit cell and the indication of symmetry (cubic, hexagonal, rhombohedral). For
example, zinc blende and wurtzite are the 3C- and 2H- polytypes of ZnS. Such polytypes
are often observed in binary semiconductors.

8.4. van der Waals radius and Lennard-Jones potential

Coming to van der Waals crystals, one usually defines the van der Waals radius as an effective
radius of an atom or molecule. Such radii appeared due to van der Waals who proposed an
equation of state for non-ideal gas,
a
p + 2 (V − b) = RT (8.2)
V
with the volume correction b that subtracts volumes of individual molecules from the total
volume of the system.17 Ironically, van der Waals never worked on the problem of chemical
bonding (let alone on solids), but his concept of an effective radius proved very useful to identify
main interactions between weakly bonded atoms or molecules. An appreciable van der Waals
bonding occurs when the distance between atoms is less than the sum of the corresponding van
der Waals radii. Any further contacts will of course lead to some minute bonding too, but it is
even weaker than the aforementioned one and, to a first approximation, negligible.
The interactions in van der Waals crystals are usually described by an effective potential,
A B
V (rij ) = − 6
+ 12
(8.3)
rij rij
where A and B are constants, and rij is the distance between atoms. The first (attractive)
term is due to London dispersion force between dipoles. The second (repulsive) term is an
empirical potential akin to the one we used in Eq. (7.5). The power of 12 is chosen for reasons
of mathematical convenience because this effective potential can be recast into the form
" 12 #
6
σ σ
V (rij ) = 4ϵ − + (8.4)
rij rij
1
known as Lennard-Jones potential18 with σ = (B /A ) 6 and ϵ = A 2 /4B .

8.5. Lattice energy

We will now repeat the same steps as in Ch. 7 to determine the equilibrium distance r0 and
the corresponding lattice energy. Total energy of a van der Waals crystal is determined by the
summation of Eq. (8.4),
X σ 6 σ 12
1
E tot (r) = 2 V (rij ) = 2ϵ −A6 + A12 (8.5)
i,j
r r
17
The constant a defines additional pressure caused by interactions between the molecules.
18
We use σ and ϵ following the standard notation of the Lennard-Jones potential. They should not be confused
with the stress and strain appearing in Ch. 9.

37
where r is the nearest-neighbor interatomic distance and we introduced lattice sums A6 and
A12 . They are similar in their meaning to the Madelung constant (Ch. 7.3), yet much easier to
calculate because the series converge very fast. For an fcc crystal, A6 = 14.45 and A12 = 12.13.
Setting the derivative of E tot to zero yields the equilibrium distance,
" 6 12 # 1
dE tot 6A6 σ 12A12 σ 2A12 6
= 0 ⇒ 2ϵ − = 0 ⇒ r0 = σ (8.6)
dr r0 r0 r0 r0 A6

that solely depends on the σ parameter of the potential. Then,

" 6 12 #
σ σ A2
E lat = −E tot (r0 ) = −2ϵ −A6 + A12 =ϵ 6 , (8.7)
r0 r0 2A12

so lattice energy solely depends on ϵ. This is certainly more elegant than the ionic case.
An interesting feature of van der Waals crystals is that they become more stable on in-
creasing the lattice parameter. This is because A increases with the atomic radius, as the
atoms become more polarizable. The B value should increase too to ensure the increase in
1
σ = (B /A ) 6 and the lattice parameter, but ϵ = A 2 /4B increases concurrently, because it con-
tains A 2 . Larger atoms and molecules develop stronger van der Waals bonds. For example,
melting points increase across the family of noble gases from neon to xenon, and likewise in-
crease across halogens from F2 to I2 . On the absolute scale, lattice energies of van der Waals
crystal remain quite low, though, typically in the range of 10 − 200 meV/atom.

8.6. Bulk modulus

To calculate the bulk modulus, we repeat the procedure from Ch. 7.5. We√ will assume the fcc
but now the atoms are at the lattice sites, so r = a/ 2. Then volume per
structure again, √
atom is V = r3 / 2 and √
d 2 d
= 2 , (8.8)
dV 3r dr
hence √ 5
d2E tot 2 d2E tot

4ϵ A6 2
B=V = = 3 A12 . (8.9)
dV 2 9 r0 dr2 r=r0 σ A12

There is now the ratio of ϵ and σ 3 , so the trend may be less intuitive than in the case of
energy, but nevertheless bulk modulus of van der Waals crystals will typically increase with
increasing the lattice parameter. Iodine is less compressible than solid bromine. The absolute
values of the bulk modulus are typically below 10 GPa and even below 1 GPa for some of the
lighter atoms and molecules.

38
9. Mechanical properties
Starting from this chapter, we will go through different crystal properties and discuss the main
parameters that describe them. We will also see how these properties are related to microscopic
aspects of the crystals, especially the atoms and their chemical bonding that constitute lattice
degrees of freedom.

9.1. Hydrostatic pressure, Equation of state

Hydrostatic pressure implies that the same force acts on the sample from all directions. It is
the type of pressure experienced by a sample immersed into water or another liquid. However,
even liquids develop some pressure gradients. The most uniform pressure can be achieved by
placing the crystal into helium gas. Helium solidifies at 11.5 GPa at room temperature, but it is
a van der Waals solid with a very low bulk modulus, so it ensures almost hydrostatic (so-called
quasi-hydrostatic) pressure conditions up to at least 20 − 25 GPa.
Pressure renders chemical bonds more stiff, because atoms become closer to each other.
Therefore, the bulk modulus increases on compression. One empirical (but very reasonable)
approximation is linear pressure dependence of the bulk modulus,
B(p) = B0 + B0′ p (9.1)
where B0 is the bulk modulus at ambient pressure and B0′ = const is pressure derivative of the
bulk modulus. Using this pressure dependence in Eq. (7.8) gives rise to a differential equation,
dp dp dV
−V = B0 + B0′ p ⇒ ′
=− (9.2)
dV B0 + B0 p V
that can be integrated to obtain the relation between p and V ,
" ′ #
B
B0 V0 0
p(V ) = ′ −1 . (9.3)
B0 V
It is the Murnaghan equation of state that describes hydrostatic compression of crystals at low
pressures V /V0 ≥ 0.9 where Eq. (9.1) holds. This equation of state describes the crystal at a
constant temperature.
At higher pressures, a quadratic term should be added to Eq. (9.1), leading to the second-
order Murnaghan equation of state. However, it is more common to use the so-called Birch-
Murnaghan equations of state obtained by expanding free energy in powers of the strain (the
step-by-step derivation can be found here). This equation can be second-, third-, or even higher
order depending on how many terms in the expansion are retained. For the reference, we quote
here the third-order Birch-Murnaghan equation of state, which is commonly used in geoscience,
" 7 5 # " 23 !#
3B0 V0 3 V0 3 3 ′ V0
p(V ) = − 1 + (B0 − 4) −1 . (9.4)
2 V V 4 V

V
Murnaghan
3!"-order Birch-Murnaghan

Figure 9.1: Pressure dependence of volume

upon hydrostatic compression, according to
the equations of state, (9.3) and (9.4).
B = const

39
Fx

ΔLx /2 ΔLy Fy

Lx ΔLy /2 Lx

Ax

Figure 9.2: Hydrostatic compression (left), uniaxial pressure (middle), and shear deformation (right).
Uniaxial strain is defined as ϵx = ∆Lx /Lx , whereas shear strain is defined as ϵxy = tan φ ≃ φ.

This equation contains the same three parameters as Eq. (9.3), but entails a different underlying
approximation because no linear pressure dependence of B is assumed.
All equations of state show that volume decreases with pressure and develops a positive
curvature, as shown in Fig. 9.1, because crystals harden on compression. Direct measurement of
this V (p) curve using high-pressure XRD (i.e., x-ray diffraction on a crystal placed into a pres-
sure cell) is the main experimental tool for studying compressibility of solids and determining
their bulk moduli.

9.2. Uniaxial pressure

In the absence of pressure medium like liquid or gas that could transmit the applied pressure
into all directions and make it hydrostatic, one deals with uniaxial pressure. Such pressure is
often called stress, σx = Fx /Ax where Fx is the applied force and Ax is the area of the sample
surface perpendicular to which this force is applied (Fig. 9.2, middle). Uniaxial stress creates
strain, ϵx = ∆Lx /Lx , defined as the relative length change of the sample. Stress has units of
pressure, whereas strain is a dimensionless quantity.
While it is difficult to expand the crystal hydrostatically, uniaxial pressure could either
compress or stretch the crystal along a certain direction. One thus distinguishes between the
compressive (ϵ > 0) and tensile (ϵ < 0) strains. Practically, uniaxial pressure can be created
simply by squeezing the crystal between the two plates. However, more and more often piezo-
devices are used instead. A suitably shaped crystal is attached to a piezoelectric stage that can
create strains up to 1.5 − 2%, both compressive and tensile.
Stress and strain are related by the familiar Hooke’s law,

σx = Y ϵx (9.5)

where Y is Young’s modulus.

Uniaxial pressure changes not only the sample length, but also its aspect ratio. Squeezing
the crystal along one direction will typically force it to expand along the perpendicular direction,
and the other way around. This effect is gauged by the Poisson’s ratio, ν = −ϵy /ϵx . Most of the
solid-state materials are characterized by ν > 0. However, it is also possible to design materials
with ν < 0 known as auxetic. They play an important role in engineering, as explained here.

40
9.3. Shear deformation
Force can be applied not only perpendicular to the sample surface, but also parallel to it, leading
to a so-called shear deformation that changes sample shape without changing its volume. Such
shear stress τxy = Fy /Ax is defined in the same way as the uniaxial stress, whereas shear strain
ϵxy = ∆Ly /Lx = tanφ is the relative displacement of the sample surface caused by Fy . For low
displacements, tanφ ≃ φ, and shear strain is measured simply as angle (Fig. 9.2, right).
The same linear relation between stress and strain holds in this case,

τxy = G ϵxy , (9.6)

with the shear modulus G that complements B and EY in describing elastic properties of
materials.
Up to this point we never referred to periodicity of the crystals. In fact, all of the above
applies to any solid, be it crystalline or amorphous. In engineering, one often considers an
isotropic medium, namely, a medium that shows the same behavior along all the directions.
Such an isotropic medium is characterized by

EY = 3B(1 − 2ν) = 2G(1 + ν). (9.7)

Its elastic properties are then fully described by only two parameters. This is the case for a
typical amorphous material like glass. Note however that Eq. (9.7) does not hold for crystals,
because even cubic crystals are not isotropic. Their [100], [110], and [111] directions are distinct,
as they are not related by any symmetry.

9.4. Stress and strain tensors

9.5. Elastic constants
General relation between stress and strain is set by the elastic constants Cijkl ,

σij = Cijkl ϵkl . (9.8)

They have the same units of pressure (Pa) as all the elastic moduli. Elastic constants describe
mechanical response of anisotropic media, including crystals. Technically, Cijkl is a fourth-
rank tensor (elasticity tensor) with 81 components, but symmetries of the stress and strain
tensors allow several simplifications. Indeed, with only six independent components in σ and
ϵ, respectively, it becomes convenient to introduce the aliases,
xx yy zz xy xz yz
1 2 3 4 5 6
and consider 36 elastic constants from C11 to C66 .
Elastic constants can be also defined in a thermodynamic fashion as second derivatives of
the elastic energy (energy acquired by the crystal due to its deformation),
1X
E = E0 + Cijkl ϵij ϵkl . (9.9)
2 ij,kl

Mixed second derivatives are equal, so Cijkl = Cklij (see also Ch. 2.5), thus reducing the number
of independent elastic constants to 21. Symmetry constrains them further. For example,

41
elasticity tensor of a cubic crystal takes the form
 
C11 C12 C12 0 0 0
 C12 C11 C12 0 0 0 
 
 C12 C12 C11 0 0 0 
C=  0
 (9.10)
 0 0 C44 0 0 

 0 0 0 0 C44 0 
0 0 0 0 0 C44

with only three independent parameters.

Inversion of the elasticity tensor yields the compliance tensor,

ϵij = Sijkl σkl . (9.11)

In Ch. 7 and 8, we saw that bulk modulus depends on the chemical bonding in crystals.
Stronger bonds give rise to less compressible crystals. The same would be true for all the elastic
constants. They are determined by the type of the crystal structure and the nature of chemical
bonds.
Experimentally, elastic constants are determined from resonant ultrasound spectroscopy.
Single crystal of the material is placed between two plates. Mechanical vibration is induced in
one of these plates and transmitted to the opposite plate that shows vibrations at those frequen-
cies where external modulation resonates with own frequencies of the crystal. These frequencies
depend on the elastic constants of the material, but also on the shape and dimensions of the
crystal. With the sufficient number of the resonant frequencies, all the 21 elastic constants can
be determined and further measured as a function of temperature or other external parameters.

9.6. Plastic deformation

Instead of the ultrasound spectroscopy, engineers use mechanical tests where stress is applied
to the sample and strain is measured, so that a stress-strain curve is obtained (Fig. 9.3). The
initial, linear part of this curve yields one or another elastic modulus from Ch. 9.2 and 9.3.
Such measurements will typically extend beyond the linear part of the curve until the regime of
plastic deformation is reached. Whereas elastic deformation is reversible, plastic deformation
remains in the sample after the stress is removed.
σ
elas!c Figure 9.3: Stress-strain curve with the re-
b
c gions of elastic (reversible) and plastic (irre-
versible) deformation. Point a is the yield
point, the onset of the plastic deformation.
a
Point b shows the ultimate strength of the ma-
terial. Point c indicates the fracture (break-
Hooke’s law plas!c ing point).
"

Plastic deformation is characterized by a nonlinear regime of the stress-strain curve that

starts at the so-called yield point and usually reaches a plateau where sample length keeps
increasing without any additional stress applied. The material starts to “flow”, similar to the
heavily loaded plastic bag carried from the supermarket. This stretching region is followed by
another increase (strain hardening) until the material reaches its point of ultimate strength.
When this point is reached, a neck forms spontaneously even if no additional stress is applied,
and eventually the sample breaks apart.

42
Materials characterized by a broad region of the plastic deformation are called ductile, in
contrast to brittle materials that break close to the yield point. Ductility is usually defined
with respect to tensile strain. The robustness of a material toward plastic deformation upon
compressive or shear strain is called malleability. It is of crucial importance for metals and
shows whether they can be rolled and bent. Finally, hardness is defined as the robustness of a
material against indentation.
Microscopically, plastic deformations are related to the propagation of dislocations, linear
defects that hinder periodicity of the crystal. The motion of dislocations shifts parts of the
crystal relative to each other and increases the crystal length. Dislocations are especially
abundant in metallic crystals where itinerant electrons are responsible for the chemical bonding,
and relative positions of the atoms are less important. Therefore, metals, in contrast to the
covalent and ionic crystals, are amenable to plastic deformations. On the other hand, alloys like
steel where iron atoms are interspersed with carbon atoms, are less ductile because impurity
atoms block the motion of dislocations.
Importantly, only elastic properties are intrinsic properties of the crystal, i.e., they are fully
determined by the structure of the crystal, its constituent atoms and chemical bonds. On the
other hand, ductility and other properties related to the plastic deformation are rooted in the
microstructure, a combination of impurities and defects that affect the motion of dislocations.

43
10. Dielectric properties
10.1. Permittivity
Response to the external electric field E is described by permittivity (ε) of the crystal. This
parameter shows the onset of the electric polarization P and of the corresponding displacement
field D,
D = εε0 E = ε0 E + P, (10.1)
as local dipoles build up in the crystal and/or become aligned with the applied field.
Polarization is defined as the dipole moment per unit of volume, P = nd µd where nd is the
dipole concentration and µd is the moment of a single dipole. P has the units of C/m2 and
also serves as a measure of charge per area, but one should remember that this charge is always
compensated by the same charge of the opposite sign sitting somewhere else in the crystal.
It is also customary to define dielectric susceptibility,
P
χe = ⇒ ε = 1 + χe . (10.2)
ε0 E
Both ε and χe are 3 × 3 tensors19 that reduce to scalars in cubic crystals.
Both dielectric susceptibility and permittivity are frequency-dependent complex numbers.
One typically separates them into the real and imaginary parts,

ε(ω) = ε′ + iε′′ and χe (ω) = χ′e + iχ′′e (10.3)

that show, respectively, the polarization in-phase and out-of-phase with the oscillating electric
field, E = E0 e−iωt . Indeed,

P = χ′e E0 e−iωt + χ′′e E0 ie−iωt = (ε′ − 1)E0 e−iωt + ε′′ E0 ie−iωt . (10.4)

At low frequencies, ε′ approaches static permittivity εst , whereas ε′′ vanishes. At high
frequencies, ε′ approaches 1 (vacuum permittivity), because electric field oscillates too fast to
induce any polarization, and ε′′ again vanishes.

10.2. Dielectric loss

Complex permittivity shows the energy loss (heating) of a dielectric in oscillating electric fields.
Consider Maxwell’s equation (A.4) that in the absence of the magnetic field can be represented
as,
∂D
j=− = −ε0 ω (iε′ − ε′′ ) E = j ⊥ + j ∥ . (10.5)
∂t
No electric current flows through the dielectric, but there is an instantaneous local current
generated by dipoles that move in order to get aligned with the applied field. This “dipole cur-
rent” comprises two components, which are perpendicular and parallel to the field, respectively.
Joule heating is given by E · j, so only ε′′ (j ∥ ) is responsible for the heating.
The imaginary permittivity (ε′′ ) is often called dielectric loss. One also defines the loss
tangent,
ε′′ |j ∥ |
tan δ = ′ = (10.6)
ε |j ⊥ |
with the loss angle δ. The loss tangent and loss angle show how much heat is produced by a
dielectric placed into an oscillating electric field. Materials with low dielectric loss are required
19
In the general case, χe is the derivative of P with respect to E.

44
in power applications to avoid heating by the ac-field. On the other hand, commercial microwave
ovens are tuned to the frequency that renders the large dielectric loss in water. Therefore, food
can be microwaved and heated, whereas its container remains cold.

10.3. Induced dipoles and polarizability

Local dipoles in dielectrics can be divided into two groups, permanent and induced. Permanent
dipoles occur in polar molecules and crystals. They carry some dipole moment that can be
turned by the applied electric field, resulting in orientation polarization. By contrast, induced
dipoles arise from the polarization of those atoms and molecules that would remain nonpolar
in the absence of the applied field. One prominent example of this mechanism is the ionic
polarization due to an optical phonon (Ch. 12).
Induced dipole moments are proportional to the electric field,

µd = α Elocal (10.7)

where α is polarizability (again, tensor property that in lucky cases can be reduced to a scalar).
This definition is quite tricky because it involves local electric field near the atom/molecule,
which is generally different from the applied electric field. We will deal with this problem
shortly. For now let’s notice that polarizability has rather non-intuitive SI units of C m2 /V or
F m2 . It becomes much more palatable in electrostatic CGS units [ε0 = 1/(4π)] where α is
measured in cm3 and corresponds to an effective volume of an atom/molecule.

10.4. Local electric field

Local electric field from Eq. (10.7) includes three contributions (Fig. 10.1),

Elocal = Eext + Edepol + EL . (10.8)

One is the external field Eext , the second one is the depolarizing field Edepol arising from charges
that accumulate on the surface, and the third one is the Lorentz field EL due to charges in the
interior of the dielectric. The first two terms, E = Eex + Edepol , build up the electric field used
in Maxwell’s equations.
One good thing about Edepol and EL is that they are proportional to polarization, because
both fields arise from charges inside the dielectric and on its surface. The depolarizing field
depends on the sample shape,
P
Edepol = −f , (10.9)
ε0
where f is the depolarization factor, basically a form factor. Its calculation is a rather tedious
exercise in electrostatics that, reportedly, Landau gave to his students as one problem of the
“theory minimum”. Less ambitious students should be alright with the simple knowledge that
flat sample features f = 1 because a lot of charge is accumulated on its surfaces, and it acts as
a capacitor of some sort. By contrast, needle-like sample features f = 0. Spherical sample is
somewhat in between with its f = 13 , as we will see shortly.
Lorentz field arises from charges in the interior of the dielectric. Envisage a small spherical
hole that has been cut out inside the sample. The charge accumulated on this new surface
produces electric field in the center of the sphere and acts opposite to the depolarizing field.
We introduce spherical coordinates with the polar angle θ that gauges the angle with respect
to E. We then slice the sphere perpendicular to the direction of E, such that each slice is a ring
with the radius R sin θ and thickness R dθ (Fig. 10.1, right). The polarization is maximum in

45
++++++++++++++ ++++++++++++++ +++++
dq
- - - - - -- - - - - - - - - -
---- Rsinθ -
--- -
- - - - - - - - - - ----- - -- --
R dθ
Eext - -
- -
-
-
- -
- θ R
δEL
EL
Edepol + +
+ +
++ +
++

-------------- ----------

Figure 10.1: Left: local electric field is obtained as a superposition of the external field Eext , depolar-
izing field Edepol produced by charges on the surface, and Lorentz field EL produced by charges in the
interior of the dielectric. Right: calculation of the Lorentz field by slicing the sphere and evaluating
the electric field δEL created by each slice in its center.

the zenith and minimum at θ = π/2, hence for a given slice P (θ) = −P cos θ, and the overall
charge on the slice is polarization times area, namely,
δq = −P cos θ × 2πR sin θ × Rdθ = −2πR2 P sin θ cos θ dθ. (10.10)

Electric field imposed by such a ring is

1 δq
δEL = − .
4πε0 R2
The components perpendicular to E cancel, whereas those parallel to E contribute to the overall
field created by the ring. Therefore, we should take the projection of δEL , which adds the factor
of cos θ, resulting in
∥ 1 δq P
δEL = − 2
cos θ = cos2 θ sin θ dθ (10.11)
4πε0 R 2ε0
for the electric field produced by the ring. We are left to integrate this over θ,
Z π Z 1 1
P 2 P 2 P cos3 θ P
EL = cos θ sin θ dθ = cos θ d(cos θ) = = . (10.12)
2ε0 0 2ε0 −1 2ε0 3 −1 3 ε0
1
This calculation also elucidates the depolarization factor of f = 3
for the sphere.

10.5. Clausius-Mossotti relation

Induced dipole moments depend on the local field that, in turn, depends on the polarization
created by these dipoles. Therefore, microscopic polarizability of atoms/molecules is directly
related to the macroscopic permittivity of the sample. Consider Elocal = E + EL and use it in
the definition of the polarization. One one hand,

P
P = nd µd = nd αElocal = nd α E + . (10.13)
3 ε0
On the other hand,
P = χe ε0 E = (ε − 1)ε0 E. (10.14)
By evaluating E and canceling P, one finds
ε−1 χe nd α
= = , (10.15)
ε+2 3 + χe 3ε0

46
ε

ε’
E(t) εst

P(t) -t/τ ε’’

~e
Pst ε

8
t ω
1/τ

Figure 10.2: Debye model of relaxation. Left: an abrupt change in the electric field E gives rise to an
exponential change of the polarization with the single relaxation time τ . Right: real and imaginary
parts of the permittivity as a function of frequency.

the Clausius-Mossotti relation, also known as the Lorentz-Lorenz equation. Simple at first
glance, it has been very significant historically, because it offered the very first experimental
tool for probing molecular property (polarizability) using a macroscopic measurement (permit-
tivity).

10.6. Permanent dipoles, Debye relaxation

Let’s now consider permanent dipoles. To a first approximation, they do not change their
dipole moment, yet they turn in order to get aligned with the field. This rotation takes time
and causes a delay between P and E, hence the imaginary part of ε. To analyze this delay,
consider first a simple situation when external electric field is changed abruptly. We assume
exponential nature of the relaxation with the characteristic relaxation time τ (Fig. 10.2), such
that
dP P (t) − Pst
P (t) = Pst 1 + e−t/τ

⇒ =− (10.16)
dt τ
where Pst = ε0 χ0 E is the static polarization reached at t → ∞ assuming that the field does
not change.
Using the standard ansatz P (t) = P0 e−iωt , we find

ε0 χ0 E0 P χ0
P0 = ⇒ χe = = , (10.17)
1 − iωτ ε0 E 1 − iωτ
the Debye relaxation. It is a standard form of a response function for a system with the
single relaxation time. We will see it again when we come to discuss ac-conductivity of metals
(Ch. ??). For now, let’s get more familiar with the expressions. Using χ = χ′ + iχ′′ , one finds
χ0 χ0 ωτ
χ′e (ω) = , χ′′e (ω) = , (10.18)
1 + ω2τ 2 1 + ω2τ 2
where χ0 is static susceptibility.
We can also use Eq. (10.2) to calculate permittivity,

εst − 1 (εst − 1) ωτ
ε′ (ω) = 1 + , ε′′ (ω) = , (10.19)
1 + ω2τ 2 1 + ω2τ 2
with the static permittivity εst = 1 + χ0 . The imaginary component ε′′ (ω) has a peak-like
structure with the maximum at ωτ = 1, i.e., at a characteristic relaxation frequency determined
by the relaxation time τ (Fig. 10.2). It is the target frequency if we want to maximize dielectric

47
ε’’

Davidson-Cole
Debye Figure 10.3: Dielectric relaxation rep-
resented by the Cole-Cole plot. De-
bye relaxation (single value of τ ) man-
ifests itself by a semi-circle. The Cole-
Cole and Davidson-Cole models de-
Cole-Cole scribe dielectrics with multiple relax-
ation times.
ε’
ε εst
8

loss and heat the sample by microwave radiation.20 The real part ε′ equals static permittivity
at ω = 0 and decays to 1 at ω → ∞.
Debye relaxation occurs at frequencies determined by the value of τ . Liquids feature re-
laxation times on the order of picoseconds and the ε′′ (ω) peak in the GHz range. In solids,
dipoles are more constrained, so their relaxation time increases to milliseconds, whereas the
peak shifts into the kHz range. All of these frequencies are anyway much lower than typical
phonon frequencies, which lie in the THz range. In fact, Debye relaxation is never the only pro-
cess induced by the oscillating electric field, and ε′ (ω) never reaches 1 at high frequencies (see
also Ch. 12). This aspect is taken into account by the modified expression where we introduced
the high-frequency limit ε∞ ,
εst − ε∞
ε(ω) = ε∞ + . (10.20)
1 + iωτ
Here, “high frequency” means the frequency well above the frequency ω = 1/τ of the Debye
relaxation.

10.7. Cole-Cole plots

In practical dielectric measurements, one seeks to determine the relaxation time(s) of the sys-
tem. To this end, Cole-Cole plots of ε′′ vs. ε′ with ω as an implicit parameter are commonly
used. Debye relaxation gives rise to the semicircular Cole-Cole plot, because all points de-
fined by Eq. (10.20) lie on the semi-circle centered at (εst + ε∞ )/2 and having the radius of
(εst − ε∞ )/2, see Fig. 10.3.
The model of Debye relaxation works well for liquids like water or ethanol. In more complex
cases, and especially in mixtures of liquids, multiple relaxation times may be present, and the
semi-circle becomes deformed. Some common approximations in this case are the Cole-Cole
equation,
εst − ε∞
ε̂(ω) = ε∞ + , (10.21)
1 + (iωτ )1−α
and the Davidson-Cole equation,
εst − ε∞
ε̂(ω) = ε∞ + . (10.22)
(1 + iωτ )β

Needless to say, α and β can be used at the same time, thus leaving even more flexibility. The
Cole-Cole equation renders the ε′′ vs. ε′ graph non-circular, while keeping the symmetry with
respect to the mid-point at (εst − ε∞ )/2. In contrast, the Davidson-Cole relaxation leads to a
semi-circle stretched on one side and suppressed on the other (Fig. 10.3). The Cole-Cole plots
20
Commercial microwave ovens operate at 2.4 GHz, which is somewhat away from the peak of ε′′ (ω) for water,
in order to avoid overheating.

48
are widely used for analyzing relaxation behavior based on frequency-dependent measurements,
not only in dielectrics but also for example in spin glass.

49
11. Phonons and sound
11.1. Underlying approximations
Atomic motion determines many, if not all, properties of a crystal. Molecules show characteristic
vibrations (normal modes) that occur at special frequencies. Crystals feature collective atomic
vibrations too, but their nature is somewhat more involved than in the case of molecules. To
understand these vibrations in the simplest possible way, we will take advantage of several
approximations:
• classical, namely, we will treat atomic motion using classical mechanics and obtain dis-
placement waves that describe collective atomic motion in the crystal. Quantum mechan-
ics requires this motion to be quantized. The corresponding quantum is called phonon,
but we usually do not need to solve the respective problem on the quantum level. It is
sufficient to quantize the waves obtained classically, as we will consciously do in Ch. ??.
• harmonic, namely, elastic energy is quadratic in the displacement, or in other words
any displacement creates a restoring force, which is proportional to the displacement.
Harmonic approximation works well most of the time, although in Ch. ?? we will see the
need to go beyond it.
• adiabatic, namely, atomic (nuclear) and electronic degrees of freedom are separated from
each other. This is possible because nuclei are much heavier than electrons, so electrons
move a lot faster. One can then think of electrons as creating a potential energy landscape
for the nuclear motion. The typical time scale of electronic and nuclear motion is 10−15 s
and 10−12 s, respectively. This approximation is also known as the Born-Oppenheimer
approximation, especially in the context of molecules. It goes back to the very general
adiabatic theorem of quantum mechanics. The adiabatic approximation fails when two
different electronic states have similar energies, and nuclei no longer know in which po-
tential energy landscape to move. Such cases are fairly rare in the ground state,21 yet
they become more common when excited electronic states are considered, for example,
when crystal is hit by a laser that drives an electronic excitation of some sort.

11.2. Monoatomic chain

Let’s consider the simplest possible case, a chain of atoms with equal spacings a. We will use the
index p to label atoms along this chain, and introduce the atomic displacements up (Fig. 11.1).
Any displacement changes the interatomic distance, and we will assume that chemical bonds
behave as springs with the stiffness k . The overall elastic energy is then
1X
E elastic = k (up+1 − up )2 . (11.1)
2 p

Atomic displacements can be determined from the equations of motion,

d2 u p ∂E elastic
m 2
=− = −k (2up − up+1 − up−1 ). (11.2)
dt ∂up

where m is the atomic mass. Instead of trying a brute force solution of these coupled differential
equations, we will choose an ansatz,

up (t) = ei(qpa−ωt) (11.3)

21
But by all means not impossible, see the Jahn-Teller effect.

50
ω

ave
p -1 p p +1

w
s!c
k

ela
a up-1 up

q = π/a
q
-π/a π/a

Figure 11.1: Atomic displacements up in the monoatomic chain. The right panel shows the dispersion
relations for the displacement wave (orange) and elastic wave (blue). The light-blue arrows indicate
the atomic displacements at q = π/a.

that essentially describes oscillations with the frequency ω that propagate along the chain with
the propagation vector q (we consider the 1D situation, so the scalar suffices). The first part,
eiqpa , is basically a phase shift between the oscillations of different atoms.
Using Eq. (11.3) in Eq. (11.2), one finds −mω 2 = −k (2 − eiqa − e−iqa ) and
r
k k qa
ω 2 (q) = 2 [1 − cos(qa)] ⇒ ω(q) = 2 × sin (11.4)
m m 2
where we used the absolute value because frequency must be positive. This is a dispersion
relation for the displacement wave (phonon) in a monoatomic chain. One immediately notices
that ω(q) is symmetric with respect to q = 0 and shows periodicity of 2π/a (Fig. 11.1) that
intriguingly matches periodicity of the reciprocal lattice (Ch. 5.2). Full implications of this
fact will become clear in Ch. 13. For now we will stay away from it and analyze the dispersion
relation. By definition, the value of q shows the phase shift between the displacements of
neighboring atoms. When this phase shift is small, springs stretch very little, and the wave
energy (hence frequency) is low. It increases with increasing q. At q = π/a, the phase shift
becomes π, so adjacent atoms move opposite to each other, thus forming a standing wave
(Fig. 11.1).
At low q, one can use sin(qa/2) ≃ qa/2 and write the linear dispersion relation
r
k
ω(q) = a ×q (11.5)
m
that corresponds to the group velocity of
r
dω k
vph = =a (11.6)
dq m
This is the speed of sound, as we will see shortly.

11.3. Elastic waves and sound

Sound is an elastic wave that propagates in any compressible medium. Consider a bar with the
cross-section A subject to a deformation that leads to the displacement ux (x) of the volume
element A dx parallel to the bar (Fig. 11.2). Time dependence of the displacement is described
by the usual equation of motion,

∂ 2 ux X
m = F = F (x + dx) − F (x) (11.7)
∂t2

51
ux

F(x) F(x+ dx) A

Figure 11.2: Compressional and shear elas-
y tic waves. The forces F (x) and F (x + dx)
x x + dx
compressional cause the displacements ux and uy along the
x
uy
bar and perpendicular to the bar, respec-
tively.
F(x)
F(x+ dx)

x x + dx shear

where forces act on both sides of the volume element. By introducing volume density ρ = m/V ,
one can write
∂ 2 ux ∂ 2 ux 1 ∂F ∂ςxx
(ρA dx) 2
= F (x + dx) − F (x) ⇒ ρ 2
= = (11.8)
∂t ∂t A ∂x ∂x
Hooke’s law relates stress and strain, ςxx = Y ϵxx via the Young’s modulus Y , whereas strain
can be represented as ϵxx = ∂ux /∂x, so overall

∂ 2 ux Y ∂ 2 ux
= × . (11.9)
∂t2 ρ ∂x2

This is the standard wave equation solved by our earlier ansatz, Eq. (11.3), but with the
dispersion relation s
Y
ω 2 = vs2 q 2 where vs = . (11.10)
ρ
The speed vs resembles our earlier result, Eq. (11.6) obtained in the q → 0 limit, and serves as
the speed of sound.
An important difference between the compressional elastic wave in a continuous medium, as
we considered here, and the displacement wave in a crystal (Ch. 11.2) is that linear dispersion
relation, ω = vs q, breaks down in the latter at higher q’s (Fig. 11.1). This happens because
crystal is discrete, so it can’t be treated as a continuous medium when period of the displacement
wave becomes comparable to the interatomic distance or, in other words, q approaches π/a.
However, the continuum approximation is perfectly justified at low q’s where period of the
displacement wave is large compared to the interatomic distance. Continuum approximations
are widely used in solid-state physics for describing mesoscopic phenomena, for example long-
period spin textures and magnetic skyrmions.
We can also envisage a wave of displacements uy , which are perpendicular to the bar
(Fig. 11.2). The solution is similar to the previous case and involves the shear stress, τxy = F/A,
as well as the shear strain, ϵxy = ∂uy /∂x, related by τxy = G ϵxy with the shear modulus G.
The resulting wave equation,
∂ 2 uy G ∂ 2 uy
= × , (11.11)
∂t2 ρ ∂x2
p
describes a shear wave propagating with the velocity of G/ρ.
Compressional wave propagates in any elastic medium, whereas shear waves only propagate
in solids because gases and liquids feature G = 0 (their shape can be changed at no energy
cost). This is the main reason why sound is a compressional wave. Shear wave can be thought
as sound too, but it’s not audible because our hearing mechanism involves sound transmission
through air.

52
11.4. Longitudinal and transverse phonons
The phonon described in Ch. 11.2 is called acoustic because q → 0 part of its dispersion
relation describes sound propagation in crystals. A similar acoustic phonon with displacements
perpendicular to the propagation direction exists too. Different names can be used to describe
these two types of motion depending on the context,
acoustic phonon longitudinal transverse
elastic wave compressional shear
seismic wave p-wave s-wave
Speeds of the corresponding waves are given by
s s
B + 43 G G
vLA = , vTA = (11.12)
ρ ρ

which are, for example, velocities of the primary (p) and secondary (s) seismic waves. Since
vLA > vTA , p-wave always arrives earlier than s-wave. This time delay measured by a seismome-
ter gauges the distance to the epicentre of an earthquake. Likewise, the data on the propagation
of p-waves and s-waves provides information on the internal structure of Earth. The existence
of the liquid outer core and inner solid core is inferred from the refraction of p-waves and from
the fact that s-waves do not reach points on the opposite side of Earth, because shear waves
do not propagate through liquid.
In Eq. (11.12), vTA is equal to the velocity of the shear wave as determined in Ch. 11.3.
However, vLA is different from the velocity of the compressional wave in a thin bar, because
Poisson’s ratio should be taken into account in a 3D solid.
In liquid and gas, G = 0 and speed of sound is described by the Newton-Laplace equation,
s
B
v= (11.13)
ρ

11.5. Diatomic chain

Let us now extend this analysis to a diatomic chain with alternating atoms and equal separa-
tions, such that all bonds have the same stiffness k . We will proceed similar to the previous
case and consider atomic displacements propagating along the chain, but now the two atoms
have different mass and should thus show different displacement amplitudes,

up (t) = A1 ei(qpa−ωt) , vp (t) = A2 ei(qpa−ωt) . (11.14)

The equations of motion become

d2 up
m1 = −k (up − vp ) − k (up − vp−1 ),
dt2
d2 vp
m2 2 = −k (vp − up ) − k (vp − up+1 ).
dt
Our ansatz for up and vp returns a system of two linear equations for the amplitudes A1 and
A2 , (
−m1 ω 2 A1 = −k A1 + k A2 − k A1 + k A2 e−iqa
−m2 ω 2 A2 = −k A2 + k A1 − k A2 + k A1 eiqa

53
ω

k k k k
ω+
m₁ m₂ op!cal
up vp
acous!c

ω-
a

q
-π/a π/a

Figure 11.3: Left: atomic displacements in the diatomic chain comprising atoms with the masses m1
and m2 connected by the springs with the same stiffness k . Right: dispersion relation for the two
phonon modes, acoustic and optical.

or (
(m1 ω 2 − 2k )A1 + k (1 + e−iqa )A2 = 0
(11.15)
k (1 + eiqa )A1 + (m2 ω 2 − 2k )A2 = 0
The solution for A1 and A2 exists when the determinant is zero, namely,

m1 ω 2 − 2k k (1 + e−iqa )
=0 (11.16)
k (1 + eiqa ) m2 ω 2 − 2k
This condition returns a quadratic equation for ω 2 ,
m1 m2 ω 4 − 2k (m1 + m2 ) ω 2 + 2k 2 [1 − cos(qa)] = 0, (11.17)
and r
k 1 4 qa
ω 2 (q) = ± k − sin2 (11.18)
m m 2 m1 m2 2
where we introduced the reduced mass,
1 1 1
= + (11.19)
m m1 m2

11.6. Acoustic and optical phonons

There are now two dispersion relations that define two phonon branches. The ω− (q) branch has
zero frequency at q = 0. It is called acoustic because it reduces to an elastic wave in the low-q
limit and describes sound propagation. The ω+ (q) branch is optical, because it can p interact
with light, as we will see in Ch. 12. Optical phonon has the finite frequency of 2k /m at
q = 0.
The main intrinsic difference between the acoustic and optical phonons lies in the amplitudes
A1 and A2 . For an optical branch at q = 0, Eqs. (11.15) yield A1 /A2 = −m2 /m1 . The atoms
of different type oscillate out-of-phase. On the other hand, ω− = 0 of the acoustic mode yields
A1 /A2 = 1, such that atoms of different type oscillate in-phase near q = 0.
p p
At q = π/a, one finds ω+ = 2k /m2 and ω− = 2k /m1 when m1 > m2 . Then,
k (1 + e−iqa ) 2
−k

A1 A1 m2 ω−
=− 2
→ 0, =− → ∞. (11.20)
A2 + m1 ω+ − 2k A2 − k (1 + eiqa )
It means that only the light atoms (m2 ) oscillate in the optical mode, whereas only the heavy
atoms (m1 ) oscillate in the acoustic mode. A nice visualization of the modes at arbitrary
q-values and for different m2 /m1 ratios can be found here.

54
12. Phonons and light
Whereas acoustic phonons are responsible for propagation of sound, optical phonons interact
with light. To understand this interaction, we will first review the meaning of optical constants
and then derive optical reflectivity of an ionic crystal due to phonons.

12.1. Refractive index and reflectivity

Optical response of a material (solid, liquid, or gas) is determined by its refractive index. The
refractive index n appears in the dispersion relation of an electromagnetic wave (Ch. A.2),

ε ω 2 = c2 k 2 with ε = n2 (12.1)

where ε is permittivity, k is the length of the propagation vector, and c = 3 × 108 m/s is
speed of light in vacuum. Since ε is generally a complex number, n is a complex number too,
n = n′ + in′′ .22 Using k = (n/c) ω = n k0 , one writes the oscillating electric field as
′′ k ′
E = E0 ei(kr−ωt) = E0 ei(nk0 r−ωt) = E0 e−n 0r
ei(n k0 r−ωt) (12.2)

where k0 is the propagation vector in vacuum (ε = 1). It is then clear that n′ , real part of the
refractive index, describes refraction, namely, the change in speed of light inside the material.
In contrast, the imaginary part n′′ describes absorption, the attenuation of the wave amplitude
inside the material.

x incident (Ei )
Figure 12.1: Propagation of light at
the crystal surface. The incident beam
splits into the reflected and transmit-
ted beams. Their intensities are deter-
transmitted (Et ) mined by the complex refractive index
of the crystal, n = n′ + in′′ .
reflected (Er )

Absorption measurements require a suitably chosen sample thickness, such that the fraction
of the absorbed intensity is neither too large nor too small. It is often more convenient to
measure reflectivity, which is defined as the intensity ratio of the incident and reflected light.
In the case of normal incidence, all of the incident, reflected, and transmitted waves travel along
the z direction perpendicular to the surface (Fig. 12.1),

incident : Ei = Ei0 ei(k0 z−ωt)

transmitted : Et = Et0 ei(nk0 z−ωt)
reflected : Er = Er0 ei(−k0 z−ωt)

At the interface, which we choose as z = 0, electric field should change continuously, so the
amplitudes fulfill the condition Et0 = Ei0 + Er0 . A similar condition should hold for the
magnetic field too, ωB = k × E according to Eq. (A.8). Therefore,

nEt0 = Ei0 − Er0 . (12.3)

22
We use primes in these lecture notes, but it is also customary to write n = n1 + in2 and ε = ε1 + iε2 .

55
It is then easy to eliminate Et0 and calculate reflectivity,23
|Er0 |2 |1 − n|2
R= = . (12.4)
|Ei0 |2 |1 + n|2
In Ch. 12.4, we will see that R values close to 1 are obtained when n′′ is large, whereas
′
n vanishes. Then any light transmitted through the interface is immediately absorbed, so the
sample is non-transparent: no light can go through. In a more general situation, part of the
transmitted amplitude, Et0 , is then lost due to absorption (A), whereas transmittance of the
whole sample is T = 1 − R − A.

12.2. Fluctuating dipoles

We now return to the phonons and consider the diatomic chain (Ch. 11.5) where the two atoms
not only have different masses but also carry the charges of +qe and −qe , respectively. At q = 0,
the optical mode involves opposite displacements of these atoms, such that a fluctuating dipole
moment µd (t) is induced by the phonon. Our strategy now will be calculating this dipole
moment and the corresponding permittivity ε to obtain the refractive index n and analyze
optical response of the crystal. We will do this using a very simple – and, in all honesty,
oversimplified – model before drawing an argument why the predictions of this model remain
valid in the more general case.
Light features an oscillating electric field, E ∼ e−iωt , that interacts with the charges. Then
the force acting on the atoms should be augmented by the Columb force, and the equations of
motion from Ch. 11.5 should be revised as
2
m1 ddtu2p = −2k (up − vp ) + qe Elocal
(

d2 vp
m2 dt2
= −2k (vp − up ) − qe Elocal
The phonon is considered at q → 0, so all atoms of a given type undergo the same displacement,
and we end up with only one differential equation for l = up − vp , the dipole length. Indeed,
dividing the first equation by m1 and the second equation by m2 returns the single equation
for l,
d2 l d2 l

1 1 1 1
= − k + l + + q e Elocal ⇒ m + m ω02 l = qe Elocal . (12.5)
dt2 m1 m2 m1 m2 dt2
p
This is a standard equation for a driven harmonic oscillator with ω0 = 2k /m , the frequency
of the optical phonon at q = 0 in the absence of any electric charges (as in Ch. 12).
Electric field is local in the sense of Ch. 10.4. It includes the internal field due to light, as well
as Lorenz and depolarization fields, which are proportional to polarization. The polarization
oscillates too, following oscillations of the dipole moment µd = qe l. Therefore, it makes sense
to choose the oscillating form of Elocal = E0 e−iωt and search for the oscillating solution, l(t) =
A e−iωt . The result is,
qe /m qe2 /m
A(ω) = E0 ⇒ µ d (t) = q e l(t) = Elocal (t), (12.6)
ω02 − ω 2 ω02 − ω 2
and polarizability of the crystal due to an optical phonon becomes
qe2 /m
α(ω) = 2 . (12.7)
ω0 − ω 2
23
This expression is derived for the case of normal incidence. More complex Fresnel equations are required
for an arbitrary incidence angle. In particular, polarization of light changes upon the non-90◦ reflection, which
is the cornerstone of the optical method called ellipsometry.

56
Electric field drives oscillations and creates resonance when its frequency matches the frequency
of the system, which is the frequency of the optical phonon at q → 0.

12.3. LO vs. TO
We should now use this polarizability α to calculate optical parameters from Ch. 12.1. Before
doing that, let’s compare electric polarizations created by the TO and LO phonons. These
two types of phonons differ by the direction of their displacements relative to the propagation
direction q. The displacements take the form u = u0 ei(qr−ωt) where q is the chain direction,
so we expect P = P0 ei(qr−ωt) with P ∥ q (LO) and P ⊥ q (TO).
In the absence of an external electric field, Eq. (10.1) becomes
D = ε0 Edepol + P, (12.8)
and we expect the same wave-like form of D = D0 ei(qr−ωt) and Edelop = E0 ei(qr−ωt) because
both of them arise from the same atomic displacements. Additionally, one of Maxwell’s equa-
tions requires divD = 0 ⇒ D · q = 0. For the LO phonon, this condition holds with D = 0
only, hence Edepol = −P/ε0 corresponding to the depolarization factor f = 1. On the other
hand, the TO phonon entails D ⊥ q for an arbitrary D and satisfies the above condition, but an-
other Maxwell’s equation, rot E = 0, requires q × E = 0, which is possible with Edepol = 0 only
because Edepol ⊥ q in this case. Therefore, the TO phonon corresponds to the depolarization
factor f = 0. Altogether,
P P 2P P
LO : Elocal = − =− , TO : Elocal = (12.9)
3 ε0 ε0 3 ε0 3 ε0
where P (t) is electric polarization created by the fluctuating dipoles µd (t).
Using these expressions in the definition of the electric polarization, P = nd µd = nd αElocal ,
one finds simple relations for the LO and TO phonon frequencies
nd qe2 −2P 2 nd qe2
P = 2 2
⇒ ωTO = ω02 − (12.10)
3m ε0 ω0 − ωLO 3m ε0
nd qe2 P 2 2nd qe2
P = 2 2
⇒ ωLO = ω02 + (12.11)
3m ε0 ω0 − ωTO 3m ε0
We thus conclude that ωLO ≥ ωTO . The difference between these frequencies gauges ionicity
p of
the crystal. We basically see that atomic charges modify the phonon frequency ω0 = 2k /m
that has been obtained for neutral atoms. The LO and TO phonons shift charges in different
ways and create different local polarizations. The LO phonon separates the charges and creates
internal electric fields that counteract the atomic displacements. Therefore, higher energy is
required for the LO-type atomic motion.
The oversimplified nature of this model is not to be overlooked, though. We assumed the
same stiffness k for the longitudinal and transverse atomic motion, which is of course far from
realistic in the light of the differences between the compressional and shear waves (Ch. 11.4).
Nevertheless, we will see that even this primitive model does lead to a qualitatively correct
physical picture, which justifies the exaggerated approximations involved.

12.4. Interaction with light, polaritons

Expressions (12.10) and (12.11) for the phonon frequencies lead to a convenient simplification
of the permittivity, which is obtained via the Clausius-Mosotti relation, Eq. (10.15),
1 + 2nd α/(3ε0 ) ω2 − ω2
ε(ω) = = LO
2
(12.12)
1 − nd α/(3ε0 ) ωTO − ω2

57
ε
ω R

1
εst
ε ωLO
8

ω ωTO
ωTO ωLO

k 0 ω
ωTO ωLO

Figure 12.2: Interaction of optical phonon with light. Left: frequency dependence of the permittivity.
Middle: polariton dispersion caused by the interaction of light with the phonon. Right: frequency
dependence of the reflectivity with the Reststrahlen band at ωTO < ω < ωLO .

The conjectured frequency ω0 disappears, and only the tangible frequencies, ωTO and ωLO ,
remain. Permittivity diverges at ω = ωTO and shows a zero crossing at ω = ωLO (Fig. 12.2,
left).
This form of the permittivity modifies an electromagnetic wave at frequencies near ωTO
and ωLO . Indeed, by using Eq. (12.12) in the dispersion relation, Eq. (12.1), one finds that the
standard linear dispersion ω ∼ k does not hold in the vicinity of the phonon frequency, and no
real solution exists for ω(k) at ωTO ≤ ω < ωLO . It means that photons can not propagate in the
crystal within this frequency range. The characteristic dispersion shown in Fig. 12.2 (middle)
is often ascribed to a polariton, an electromagnetic wave strongly coupled to an intrinsic and
dipole-carrying excitation of the crystal.
The polariton formation causes the crystal to reflect electromagnetic waves. Indeed, at
ω < ωTO and at ω > ωLO , one finds ε > 0, such that n = n′ + in′′ is a real number with
n′ ̸= 0 and n′′ = 0. Crystal refracts light, as we know from NaCl, quartz, and many other
ionic crystals. On the other hand, at ωTO < ω < ωLO , negative ε renders n a purely imaginary
number with n′ = 0 and n′′ ̸= 0.24 The resulting reflectivity is

|1 − in′′ |2
R= = 1, (12.13)
|1 + in′′ |2

so ionic crystals reflect light in the frequency range between ωTO and ωLO (Fig. 12.2, right).
These so-called Reststrahlen bands are observed, for example, in ionic crystals in the infrared
range (ω = 0.3 − 400 THz) typical for optical phonons. A similar mechanism underlies the
response of polar molecules that absorb infrared radiation at frequencies of their vibrational
modes.

12.5. Lyddane-Sachs-Teller relation

2 2
Eq. (12.12) returns the low-frequency (static) limit of εst = ωLO /ωTO and the high-frequency
limit of 1. Practically, however, more than one optical phonon is usually present in the crystal,
so this high-frequency limit of one phonon mode will be the low-frequency limit of another
one with a higher energy. Moreover, beyond all phonon frequencies one still expects electronic
excitations at energies of E = 100 − 104 eV. Therefore, it is customary to introduce ε∞ ̸= 1 and
24
This distinction between the purely real and purely imaginary n is, of course, a drawback of our naive
approximation. The purely real ε(ω) arises from the purely real α(ω) obtained in Ch. 12.2 from the equations
of motion that did not include any damping term. Realistically, damping creates a phase shift between µd (t)
and Elocal (t), thus rendering all optical parameters complex numbers.

58
write permittivity as25
2
ωLO − ω2
ε(ω) = ε∞ 2
, (12.14)
ωTO − ω2
resulting in
εst ω2
= LO
2
, (12.15)
ε∞ ωTO
the Lyddane-Sachs-Teller relation.

ε Figure 12.3: Schematic representation

of different contributions to the per-
εst mittivity. Re-orientation of permanent
ε’
dipoles at low frequencies (Fig. 10.2,
permanent dipoles
right) is followed by the ionic polar-
ization due to phonons and eventu-
ionic polarization
ally by the electronic polarization due
electronic polarization
to induced dipoles within individual
ε’’ atoms. Electronic and ionic relaxations
ω (s )
-1
give rise to wiggles, which are broad-
106 109 1012 1015
ened versions of the divergence of ε(ω)
radio microwave infrared UV
shown in Fig. 12.2 (left).

An interesting outcome of this analysis is that static permittivity of a crystal is determined

by its phonon frequencies. Even a nonpolar crystal without any permanent dipoles may show
εst well above 1.0 because charges inside the crystal move as a result of the atomic motion. A
crystal with permanent dipoles shows an additional contribution to εst due to these pre-existing
dipoles. Then, εst of the Lyddane-Sachs-Teller relation is ε∞ of the Debye relaxation (Ch. 10.6).
Different contributions to the permittivity are sketched in Fig. 12.3.

12.6. Infra-red active phonon modes

While the appealing form of the permittivity, Eq. (12.12), is the result of several simplifications
introduced in our model (linear chain, same spring constant for the longitudinal and transverse
displacements), the qualitative behavior displayed in Fig. 12.2 is quite general. We can see
this by re-iterating the arguments from Ch. 12.3 but now considering a displacement wave in
the crystal triggered by the oscillating electric field of light. Assume that the electric field
E = E0 ei(kr−ωt) generates the polarization P = P0 ei(kr−ωt) and the displacement field D =
D0 ei(kr−ωt) . According to Maxwell’s equations,

div D = 0 and rot E = 0 ⇒ k · D = 0 and k × E = 0. (12.16)

The former condition requires D = 0 or k ⊥ D (hence k ⊥ E), while the latter requires E = 0
or k ∥ E (hence k ∥ D).
Consider the TO phonon. In this case, D ⊥ k, so E must be zero for the second condition
to be fulfilled. Then, ε → ∞ because D = εε0 E, so ε(ω) diverges at ωTO . On the other hand,
the LO phonon implies E∥k, so D must be zero, which is only possible at ε = 0. Therefore,
a zero crossing of ε(ω) should occur at ωLO . The pre-condition for these arguments is our
assumption that D ∥ E or, basically, the diagonal form of the permittivity tensor expected in
crystals with the sufficiently high symmetry (Ch. 2.5).
25
This equation can be obtained rigorously by augmenting the ionic polarizability in Eq. (12.7) with the
atomic polarizability α0 as the origin of ε∞ ̸= 1. Unfortunately, this small amendment leads to a far more
tedious algebra than what we did here, so this exercise is recommended for nerds only.

59
The situation becomes more complex in crystals with lower symmetry. Indeed, even dis-
tinguishing between longitudinal and transverse phonons is not possible in this case, because
atomic displacements are not constrained to be parallel or perpendicular to the propagation
vector of the phonon (see also Ch. 13). However, what matters for our analysis is not q but
k, the propagation direction of light. One can think of splitting atomic displacements of an
arbitrary phonon mode into two components. The component perpendicular to k leads to a
divergence of ε(ω) at ω = ωTO , whereas the component parallel to k results in the zero crossing
at some effective frequency ω = ωLO .
It may seem that almost every phonon mode of a non-cubic crystal should show up in ε(ω)
and in the optical (infrared) spectrum. Not quite. No coupling to light occurs when phonon
mode does not generate any dipole moment. This may happen in a covalent crystal or when the
phonon preserves inversion symmetry.26 What ultimately matters is the formation of a dipole
moment (i.e., atomic displacements) parallel to E of light. The coupling between this dipole
moment and oscillating electric field renders phonon-mode IR-active. Polarized light is often
used to single out modes with a given direction of the atomic displacements.

26
Such modes are called even (gerade), in contrast to odd (ungerade) modes that break the inversion symmetry.
For a given crystal, the distribution of modes into odd and even can be analyzed on the basis of crystal symmetry
and occupied Wyckoff positions, for example using the SAM utility at the Bilbao server.

60
13. Phonons and the reciprocal lattice
13.1. Brillouin zone
13.2. Dynamical matrix
...

61
A. Electrodynamics
A.1. Maxwell’s equations
For a non-magnetic solid one writes the four Maxwell’s equations in the form

div D = ρe , (A.1)
div B = 0, (A.2)
∂B
rot E + = 0, (A.3)
∂t
∂D
rot H − = j, (A.4)
∂t
where ρe is the charge density and j is electric current density. The displacement field D = εε0 E
is defined via the permittivity ε, while the two magnetic fields are related by B = µ0 H.
One can also derive the continuity equation by taking divergence of Eq. (A.4) and using the
vector identity div (rot H) = 0,

∂ ∂ρe
div (rot H) − div D = div j ⇒ div j + = 0. (A.5)
∂t ∂t
It shows that any change in the charge density is caused by the flux of the electric current, and
vice versa.
The above equations are written in the SI units where ε0 = 8.85 × 1012 C/(V m) and
µ0 = 4π × 10−7 V s/(A m) are the vacuum permittivity and permeability, respectively. It is not
advisable to use CGS units unless you really have to, because Maxwell’s equations will contain
additional pre-factors in this case.

A.2. Electromagnetic waves

A solution to Maxwell’s equations can be found in the form of plane waves,

E = E0 ei(kr−ωt) , B = B0 ei(kr−ωt) , (A.6)

where k is the propagation vector and amplitudes E0 , H0 can be complex to allow for a phase
shift between E and B. By replacing Eqs. (A.6) into the Maxwell’s equations, one finds

εε0 (k · E) = 0 µ0 (k · H) = 0 (A.7)
k × E = ωB k × H + εε0 ω E = j/i. (A.8)

The current j can be eliminated using Ohm’s law, j = σE, which transforms Eq. (A.8) into

k × H = −(εε0 ω + iσ) E. (A.9)

The most natural solution to Eqs. (A.7)–(A.8) is the transverse wave with k ⊥ E and k ⊥ B,
thus satisfying Eqs. (A.7). From the remaining two equations one concludes that E ⊥ B. This
way, E and B oscillate along two different (orthogonal) directions, whereas the wave propagates
along the third one. It is the standard electromagnetic wave (light) as we know it.
To explore the behavior of such transverse waves, we use the vector relation

k × [k × E] = k(k · E) − (k · k)E = −k2 E

62
(k · E = 0 since the two vectors are orthogonal) and apply it to Eqs. (A.8). Then we obtain

−k2 E = −µ0 ω(εε0 ω + iσ) E

and arrive at the dispersion relation

k 2 − εε0 µ0 ω 2 − iµ0 σ ω = 0 (A.10)

that is usually written in a more concise linear form,

c2 k 2 = ε̂(ω) ω 2 , (A.11)
√
with c = 1/ ε0 µ0 = 3 × 108 m/s, speed of light in vacuum, and complex permittivity

iσ
ε(ω) = ε + . (A.12)
ε0 ω
It is what we call permittivity throughout these lecture notes. The practical advantage of
the complex permittivity is the simple and convenient linear dispersion relation, k ∼ ω. The
conceptual physical reason for merging permittivity and conductivity into a single parameter
is that both of them describe system’s response to the electric field, but charge displacements
(ε) have a phase shift of π/2 with respect to charge velocities (σ). Ch. ?? offers some further
insights into this issue.

A.3. Longitudinal waves

Another solution of Eqs. (A.7)–(A.8) exists when ε = 0. Then the first equation of Eq. (A.7)
is fulfilled for any k and E. Specifically, a non-magnetic solution (B = 0) becomes possible
when k ∥ E, with no other restrictions imposed on k, whereas frequency is determined by the
ε = 0 condition. This solution is a longitudinal wave, where electric field oscillates along the
propagation direction. It does not occur in vacuum where ε = 1, but it may occur in medium
when ε is tuned to zero. This is one way to see the LO phonon (Ch. 12.3) as well as plasmons
in metals (Ch. ??).

63
B. Thermodynamics

C. List of experimental techniques

64