Chapter 1

Crystal structure

1.1 Crystal lattices

We will concentrate as stated in the introduction, on perfect crystals, i.e. on arrays
of atoms, where a given arrangement is repeated forming a periodic structure, in
principle over the whole space. Although crystals have in reallity a finite extension,
they posses of the order of 1023 atoms, such that to consider them infinite is a rather
good approximation. In doing so, we will disregard those phenomena taking place
close to the surface of the crytals, such that we will leave aside a big area of Solid
State Physics, namely Surface Physics. However, we will develop the ideas also
necessary to deal with that part of physics, so that those interested in that area,
can participate later in corresponding lectures offered as elective ones.
Another restriction due to the assumption of a periodic structure is that we will
not consider the effect of impurities and disorder. Impurities can be considered
as a perturbation of an ordered system, and as such, their treatment can be seen
as an extension of the lectures here, i.e. they could be the subject of a special
lecture. Disorder on the other hand, an important area of research in statistical as
well as in solid state physics, prompted the development of many new theoretical
techniques like e.g. supersymmetry, needed to study phenomena like localization or
glassy behavior. Unfortunately, we are not going to have time to deal with such
interesting subjects, but elective lectures are available in our faculty.
Once we stated what we are not going to deal with, let us come back to our
subject. A perfect crystal, that is a periodic structure posses a translation symmetry.
This means, that there is some group of atoms that can build up the whole crystal
by translations. The smallest of such groups is called the elementary cell. As shown
by the examples below (Fig. 1.1), the elementary cell is not uniquely determined.
We can also give a mathematical meaning by saying that translational symmetry
means that there exist vectors ai , i = 1, . . . , d, where d is the number of space
dimensions, such that the structure remains invariant under translations by any
vector which is the sum of integral multiples of these vectors. The vectors
X
Ri = nij aj , j = 1, . . . , d , (1.1)

5
6 K. Rodriguez Solid State Theory

with nij integers are called lattice vectors and define a Bravais lattice. The vectors
ai are called basis vectors.

~a2 ~a1
~a2 ~a1
~a2

~a1

Figure 1.1: Examples of different elementary cells.

Since as we said above, the determination of an elementary cell is not unique, we

can find an elementary cell by constructing a Wigner-Seitz cell, that is determined
by drawing the perpendicular bisector planes of the translation vectors from one
lattice point to its nearest neighbors (Fig. 1.2).

Figure 1.2: Wigner-Seitz cell.

A structure, where the atoms are sitting on the sites of a Bravais lattice, is called
a primitive lattice. In this case we have only one atom per elementary cell. In the
case, where there are several atoms per elementary cell, one speaks of a lattice with
a basis.

1.2 The reciprocal lattice

Let us suppose for definitness, that we are dealing with three dimensions, i.e. d = 3.
Then, the volume of the elementary cell is

Ω = a1 · (a2 × a3 ) . (1.2)
 Solid State Theory K. Rodriguez 7

Let us define now three new vectors as follows


b1 = (2π/Ω) a2 × a3  
 
 (
2π i, j, k = 1, 2, 3


b2 = (2π/Ω) a3 × a1 bαi = εijk εαβγ aβj aγk (1.3)


 Ω α, β, γ = x, y, z


b3 = (2π/Ω) a1 × a2



εijk and εαβγ are called Levi-Civita symbols. They are totally antisymmetric tensors
with the property


 1 for i = 1, j = 2, k = 3 or an even permutation of them
εijk = −1 for i, j, k an odd permutation of 123 (1.4)
0 for any two indices equal



Using the definition of the vectors bi and the properties of the Levi-Civita symbols,
it can be shown that
ai · bj = 2π δij , (1.5)
where δij is called the Kronecker delta with the property
(
1 for i = j
δij = (1.6)
0 else
The vectors bi can be seen as basis vectors of the reciprocal lattice, i.e. they can be
used to construct another Bravais lattices with vectors
3
X
G = k j bj , (1.7)
j=1

with kj integer. These vectors are called reciprocal lattice vectors. Recalling (1.1),
we find an important property of the reciprocal lattice vectors
X
G·R = ni kj ai · bj
i,j
X
= 2π ni ki = 2πM , (1.8)
i,j

with M some integer. This implies that

exp (iG · R) = 1 , (1.9)
a fact with important consequences for the next point.
Since the reciprocal lattice vectors form a Bravais lattice, we can construct a
Wigner-Seitz cell on this lattice. It is called the Brillouin-zone. Its volume can be
obtained in the same way as in real space, namely
ΩB = b1 · (b2 × b3 ) . (1.10)
Using (1.3), it can be shown that
(2π)3
ΩB = . (1.11)
Ω
8 K. Rodriguez Solid State Theory

1.2.1 Fourier decomposition of periodic functions

Let us recall without demonstration, what a Fourier transform is. For a mathemat-
ical rigorous discussion, see e.g. Mathematics for Physicists by P. Dennery and A.
Krzywicki, Dover 1995, ch. III, sec. 11.    
Given a function f (x) defined in the interval [− a2 , a2 ], such that f a2 = f − a2 ,
one has the Fourier expansion
∞
1 X 2π
 
f (x) = √ fm exp im x ,
a −∞ a
Z a
1 2π
 
2
fm = √ dx f (x) exp −im x , (1.12)
a − a2 a

where m is integer. It is here easy to see that the function f (x) is a function with
period a, i.e. f (x) = f (x + a). At this point we could introduce a quantity

2π
g = m , (1.13)
a
with the obvious property g na = Ma, with n and M integers, so that g is the
one-dimensional version of a reciprocal lattice vector. With such a notation, the
generalization to more than one dimension is straightforward.
∞
1 X
f (x) = √ fg exp (ig · x) ,
Ω −∞
1
Z
fg = √ dx f (x) exp (−ig · x) , (1.14)
Ω Ω

where we have to require now that f (x) = f (x + R). But this is equivalent to
require that exp (ig · R) = 1. Recalling (1.9), we see that g = G, as we already
expected.
A special periodic function that will be usefull later is the following
X
f (x) = δ (x + R) , (1.15)
R

where δ (x) is the Dirac delta-function (recall that this is actually a generalized
function, or distribution) with the following properties:
Z
d3 x δ (x) = 1 ,
Z
d3 x f (x) δ (x − x0 ) = f (x0 ) . (1.16)

The function (1.15) is obviously translational invariant since it contains all lattice
vectors, and therefore, it will not be affected by a translation by any R. Using
 Solid State Theory K. Rodriguez 9

(1.14), we can now calculate the coefficients of the Fourier expansion

1
Z
fG = √ d3 x f (x) exp (−iG · x)
Ω Ω
1 X
Z
= √ d3 x δ (x + R) exp (−iG · x)
Ω R Ω
1
= √ . (1.17)
Ω
Inserting this result in the Fourier expansion as shown in (1.14), we finally have
X X
Ω δ (x + R) = exp (iG · x) . (1.18)
R G
This equality will be useful in the next section.

1.2.2 Periodic boundary conditions

Although the fact that we are dealing with a periodic crystal will imply for many
quantities that they have the same period as the crystal, like e.g. the electronic
density, there may be other quantities that will not have this periodicity. One
could imagine departures from the equilibrium electronic density caused by the
interaction with an electromagnetic wave, or departures of the atoms from their
equilibrium position due to the excitation with an acoustic wave. In these cases,
the waves induced in the solid by the external perturbation should essentially have
a wavelength of the order of those of the perturbation. For electromagnetic waves
(light) this will be of the order of several thousand Angstroms and for acoustic wave
even much longer. This means that we have to treat the solid as a whole and in the
spirit me mentioned at the beginning of this chapter, the infinite solid. However,
dealing with infinities have to be made with care. We therefore introduce periodic
boundary conditions.
We assume that the system we are dealing with consists of a periodic repetition
of what happens in a cell of linear size Ni ai , i = 1, 2, 3, with Ni  1. We set the
condition that all physical quantities fulfill the condition
f (R + Ni ai ) = f (R) , (1.19)
for i = 1, 2, 3. At the end of the calculation we take the limit Ni → ∞, in order to
reach the thermodynamic limit. Since again we are dealing with a periodic function,
we can perform a Fourier transformation as in (1.14). Let us call the corresponding
wavevector k. The condition (1.19) will imply now
exp (ik · Ni ai ) = 1 . (1.20)
As in Sec. 1.2.1, we can set k equal to a reciprocal lattice vector, however we have
now many more vectors also satisfying the condition above, namely
n1 n2 n3
k = b1 + b2 + b3 . (1.21)
N1 N2 N3
10 K. Rodriguez Solid State Theory
 
Wavevectors with ni = 0, ±1, . . . , ± N2i − 1 , N2i , belong to the first Brillouin zone.
As we pass to the thermodynamic limit, we will increase the density of wavevectors
inside the first Brillouin zone, and since the number of states inside it, is equal to the
number of degrees of freedom in the system, it will contain all the relevant physical
information as long as the orginal periodicity is preserved.
At this point we can consider again a Fourier decomposition of a function f (x),
as prescribed in (1.14),
1 X
f (x) = √ fq exp (iq · x) , (1.22)
V all q

where by ‘all q’ we mean all the q-vectors compatible with the periodic boundary
conditions, and V is the volume of the system with V = ΩN, and N = N1 N2 N3 .
This sum can be splitted as follows
X X X
= , (1.23)
all q G q ∈ BZ
such that (1.22) can be written as
1 X X
(1.22) = √ exp (iq · x) fq +G exp (iG · x) . (1.24)
V q ∈ BZ G
Let us assume that the Fourier coefficients do not depend on the reciprocal lattice
vectors and see which are the consequences of that. In this case we have
1 X X
(1.24) = √ fq exp (iq · x) exp (iG · x) . (1.25)
V q ∈ BZ G
Next recall the formula (1.18), that leads to
Ω X X
(1.25) = √ δ (x + R) fq exp (iq · x) . (1.26)
V R q ∈ BZ
This tells us, that the assumption we made implies that f (x) is a function only
defined on the lattice sites:
√ X
f (x) = Ω f (R) δ (x + R) , (1.27)
R
with
1 X
f (R) = √ fq exp (−iq · R) . (1.28)
N q ∈ BZ

We can understand this fact physically, since a wavevector G corresponds to a wave-

length shorter than a, the lattice constant. Then, if we find that the Fourier coeffi-
cients do not depend on G’s, this means that we have no information on distances
 Solid State Theory K. Rodriguez 11

shorter than a. Functions defined only on lattice sites are commonly encountered
in models in solid state physics, as we will see in the course of the lecture (tight
binding electrons). Using (1.28), we can obtain also the coefficients of the Fourier
expansion.
s
Ω X
Z
fq +G = f (R) dx δ (x + R) exp [−i (q + G) · x]
V V
R
1 X
,→ fq = √ f (R) exp (iq · R) . (1.29)
N R

We see that in fact, the coefficients do not depend on the reciprocal lattice vectors.
The last formula together with (1.28) give the transformation laws for functions on
the lattice with periodic boundary conditions.
Let us close this section by discussing several formulae that are usefull in dealing
with periodic systems:
Z
d3 x exp [i (G − G0 ) · x] = Ω δG,G0 , (1.30)
Z Ω

d3 k exp [i (R − R0 ) · k] = ΩB δR,R0 , (1.31)

ΩB
X X
ΩB δ (k + G) = exp (i k · R) . (1.32)
G R

It is easy to see that (1.30) is fulfilled in one dimension, and after seeing this, the
generalization to higher dimensions is straightforward. Once (1.30) is understood,
(1.31) results obvious, since its evaluation requires the same steps. Of course, the
integral in (1.31) is understood in the thermodynamic limit, where there is a dense
set of points inside the Brillouin zone. Finally, (1.32) can be obtained in the same
way as (1.18) by exchanging the roles of k → x and of R ↔ G.

1.3 Symmetries and types of Bravais lattices

Apart from translational symmetry there are other structural symmetries that trans-
form a crystal into itself. As for the translation, it is said that these symmetry
transformations build a group. A group is defined as follows:
A set G is a group if

a) there is an operation (‘product’) that for any two elements T, T 0 ∈ G, assigns

another element T 00 ∈ G,

b) ∀ T , T 0 , and T 00 ∈ G, it holds

(T T 0 ) T 00 = T (T 0 T 00 )
12 K. Rodriguez Solid State Theory

c) ∃ E ∈ G, such that
T E = ET = T

d) ∀ T ∈ G, ∃ T −1 ∈ G, such that
T T −1 = T −1 T = E

The kind of symmetries that can appear in a crystal are:

• Point group: all operations that transform the lattice into itself and leave a
given point invariant. These operations are rotations, reflections, inversions
(no translations).

a) b)

c) d)

Figure 1.3: Examples of point symmetry operations on a cube a): b) Rotations by

90o , c) rotations by 120o , d) reflexion planes.

• Space group: all operations that transform the lattice into itself (point group
operations + translations).

There are seven possible point groups and fourteen Bravais lattices that are enu-
merated in the following.
 Solid State Theory 13 K. Rodriguez

1. Cubic: here belong all lattices with the symmetry group of a cube. There
are 3 Bravais lattices with non-equivalent space groups: simple cubic (sc),
body-centered cubic (bcc), and face-centered cubic (fcc).

a) b) c)

Figure 1.4: Cubic Bravais lattices: a) simple cubic, b) body centered cubic, c)
face-centered cubic.

2. Tetragonal: here belong all lattices with the symmetry group of a rectangular
prism. There are 2 Bravais lattices with non-equivalent space groups: simple
tetragonal, and centered tetragonal

Figure 1.5: Simple tetragonal Bravais lattices.

3. Orthorombic: here belong all lattices with the symmetry group of an object
with perpendicular faces and with all three perpendicular sides with unequal
length. There are 4 Bravais lattices with non-equivalent space groups: simple
othorombic, base-centered orthorombic, body centered orthorombic, and face-
14 K. Rodriguez Solid State Theory

centered orthorombic.

Figure 1.6: Orthorombic Bravais lattices.

4. Monoclinic: here belong all lattices with the symmetry group of an object with
a base made of a parallelogram and two faces perpendicular to it. There are
2 Bravais lattices with non-equivalent space groups: simple monoclinic and
centered monoclinic.

Figure 1.7: Monoclinic Bravais lattices.

5. Triclinic: here belong all lattices with the symmetry group of an object with
all three faces forming a parallelogram with sides of unequal length. This is
the Bravais lattice with minimum symmetry, so that there is only 1 type of
Solid State Theory 15 K. Rodriguez

this kind.

Figure 1.8: Triclinic Bravais lattices.

6. Trigonal: here belong all lattices with the symmetry group of an object ob-
tained from a cube by stretching it along a body diagonal, generating a body
with sides of equal length making equal angles with one another. This type is
also called rhombohedral and there is only 1 such type of Bravais lattice.

Figure 1.9: Trigonal Bravais lattices.

7. Hexagonal: here belong all lattices with the symmetry group of a right prism
16 K. Rodriguez Solid State Theory

with a regular hexagon as base. There is only 1 such type of Bravais lattice.

Figure 1.10: Hexagonal Bravais lattices.

This is an enumeration that should illustrate the different types of lattices but
we will not go further into more details, since this is a well established topic and
further lectures on crystallography may be attended by those interested in them. An
interesting topic related to structure and still a subject of research is concerned with
the fact that, no crystal structures are possible with n-fold rotational symmetry for
n 6= 2, 3, 4, and 6. The first point to notice is that, given an n-fold rotation axis,
there is allway a set of planes where atoms form a lattice. This can be seen case
by case in each of the Bravais latices discussed above. In general we can argue that
given the rotation axis, we can allway find a plane with at least one atom. However,
if translational symmetry is present, it should be possible to translate the axis by
some amount to an equivalent position, and the same should happen to the atom
found. Therefore, there should be in general a two-dimensional lattice associated
with a rotation axis. Next, we assume that we found two nearest neighbor atoms,
and choose units of length such that this distance is unity (Fig. 1.11). We perform

B’ a’

A B
Figure 1.11: Atoms A and B are assumed nearest neighbors. a0 gives the distance
between B and B’ after a rotation by 2π/n.

now a rotation by 2π/n, from the line joining both atoms around an axis passing
 Solid State Theory K. Rodriguez 17

through one of them. Then, the distance between the new position and the atom
nearest neighbor to the one where the axis rotation was passing through is
π
a0 = 2 sin (1.33)
n
Since by assumption, a0 > 1, we obtain a contradiction for n ≥ 7, as sin(π/n) < 1/2
for n ≥ 7. On the other hand, for n = 5, we can perform the rotation around one
point in one direction and around the other point in the other direction (Fig. 1.12),
such that the distance between the new points is

a’’
B’ A’

A B
Figure 1.12: Atoms A and B are assumed nearest neighbors. a00 gives the distance
between A’ and B’ after a rotation by 2π/n arround A and B in different directions.

2π
a00 = 1 − 2 cos , (1.34)
5
reaching again a contradition. Or put in other words, it is not possible to make
a perfect covering of the plane with regular figures other than triangles, squares,
or hexagones. Experiments show that there are ordered systems (although not
perfectly periodic) with 5- and 12-fold rotation symmetry. These systems are known
as quasicrystals.
18 K. Rodriguez Solid State Theory