Solid State Physics Lecture Notes

Lecture Notes of
Solid State Physics
Carlo E. Bottani
a.y. 2017/18 - February 2018

1
Contents
1 Solid bodies 5
1.1 Order and symmetry . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.1 Simple Crystals: lattices and translational order . . . . 7
1.1.2 Complex crystals: lattice and basis . . . . . . . . . . . 13
1.2 Crystal binding . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.3 Tensorial observables . . . . . . . . . . . . . . . . . . . . . . . 15
2 Scattering theory 16
2.1 Elementary theory of elastic scattering . . . . . . . . . . . . . 17
2.2 Quantum particle scattering amplitude: Born approximation 19
2.3 Elastic scattering - Bragg law . . . . . . . . . . . . . . . . . . 22
2.4 X-Ray Diffraction . . . . . . . . . . . . . . . . . . . . . . . . . 24
3 Electronic States 29
3.1 Bloch’s Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Reduction to the first Brillouin zone . . . . . . . . . . . . . . . 35
3.3 Born-von Karman boundary conditions . . . . . . . . . . . . . 37
3.4 Free electron gas . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.4.1 Plane waves - Fermi energy - Density of states . . . . . 39
3.4.2 Chemical potential and electronic specific heat . . . . . 41
3.4.3 Quasi-particles and Fermi liquid theory (a qualitative
introduction) . . . . . . . . . . . . . . . . . . . . . . . 43
3.5 Band structure: nearly free electron approximation . . . . . . 44
3.5.1 Free electron diffraction . . . . . . . . . . . . . . . . . 44
3.5.2 Fermi surface and density of states . . . . . . . . . . . 46
3.6 Band structure of tightly bound electrons . . . . . . . . . . . . 47
3.6.1 LCAO Method . . . . . . . . . . . . . . . . . . . . . . 50
3.7 Electron dynamics in external fields . . . . . . . . . . . . . . . 51
3.7.1 Equivalent Hamiltonian. Effective mass theorem . . . . 52
3.7.2 Impurity levels . . . . . . . . . . . . . . . . . . . . . . 55
3.7.3 Semiclassical dynamics in a crystal . . . . . . . . . . . 55
3.7.4 Limitations in the effective mass description . . . . . . 58
3.7.5 Electric current - Electrons and holes . . . . . . . . . . 59
3.7.6 Excitons . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4 Adiabatic theorem and vibrational motions 64

2
5 Lattice dynamics 67
5.1 Lattice specific heat: Einstein model . . . . . . . . . . . . . . 67
5.2 Lattice dynamics: Born - Von Karman model . . . . . . . . . 69
5.3 Linear chain with two atoms per cell: acoustic modes and
optical modes . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5.4 Simple 1D crystal: acoustic phonons and elastic waves . . . . 78
5.5 Phonons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.6 Specific heat: Debye model . . . . . . . . . . . . . . . . . . . . 82
5.7 Polaritons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.8 Inelastic scattering . . . . . . . . . . . . . . . . . . . . . . . . 93
6 Optical properties 98
6.1 Kramers e Kronig relations . . . . . . . . . . . . . . . . . . . . 100
6.2 Probability per unit time of optical transitions . . . . . . . . . 103
6.3 Interband optical transitions . . . . . . . . . . . . . . . . . . . 105
6.3.1 Direct transitions . . . . . . . . . . . . . . . . . . . . . 105
6.3.2 Indirect transitions . . . . . . . . . . . . . . . . . . . . 108
6.4 Intraband transitions and plasmons . . . . . . . . . . . . . . . 111
7 Transport phenomena 112

7.1 Phenomenology . . . . . . . . . . . . . . . . . . . . . . . . . . 112
7.2 Kinetic theory . . . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.2.1 Appendix: Liouville theorem . . . . . . . . . . . . . . 118
7.3 Electrical conductivity . . . . . . . . . . . . . . . . . . . . . . 119
7.4 Electronic thermal conductivity . . . . . . . . . . . . . . . . . 120
7.5 Lattice thermal conductivity (insulators) . . . . . . . . . . . . 122
8 Appendix: Quantum approximate methods 124

8.1 Time-dependent perturbations . . . . . . . . . . . . . . . . . . 124
8.1.1 Harmonic perturbation . . . . . . . . . . . . . . . . . . 128
8.1.2 Photon-electron interactions: electric dipole selection
rules . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
8.2 Static perturbation theory . . . . . . . . . . . . . . . . . . . . 130
8.2.1 Adiabatic switching on . . . . . . . . . . . . . . . . . . 130
8.2.2 Degenerate energy levels . . . . . . . . . . . . . . . . . 132
8.3 Dirac delta function . . . . . . . . . . . . . . . . . . . . . . . . 133
8.4 Green functions . . . . . . . . . . . . . . . . . . . . . . . . . . 135
8.5 Selfconsistent mean field . . . . . . . . . . . . . . . . . . . . . 136
8.6 Semiclassical dynamics in a conservative force field . . . . . . 139
c
°2017-2018 Carlo E. Bottani Solid State Physics Lecture Notes 3
9 Appendix: Elasticity and Elastic Waves 142

9.1 Stresses, strains and elastic constants . . . . . . . . . . . . . . 142
9.2 The acoustic waves and their phonons . . . . . . . . . . . . . . 146
c
Foreword
To make a proper use of these notes, students should carefully
read this foreword. The present text is the English translation
of an original Italian version used till the second semester of a.y.
2011/12. The translation was quickly accomplished to be ready
for the 2013/14 course of Solid State Physics and was far from
being completely satisfactory. A better edition is now available
for the 2017/18 course. Concerning the content, although I have
taken the occasion to introduce some minor changes, the nature
of the notes remains the same: a short and synthetic outline
of the theoretical part of the lectures on Solid State Physics I
gave at Politecnico di Milano in the last eight years. The con-
tents of more practical lectures (very specific examples, exercises
and description of experiments), which also belong to the course
program and are possible objects of the final exam, are almost
not included. Anyhow this notes, initially born as a personal
notebook of the teacher and not intended for student studying,
should be read only as an additional didactic aid and should
never be used in place of a good textbook of Solid State Physics,
which remains an indispensable tool for any student. For my lec-
tures I have taken inspiration from many famous textbooks like:
J.M. Ziman, Theory of solids (2nd edition), Cambridge University
Press (1995); C. Kittel, Introduction to Solid State Physics (8-
th edition), John Wiley &Sons (2004); N.W. Ashcroft and N.D.
Mermin, Solid State Physics, HRW International Editions (1981);
W.A. Harrison, Solid State Theory, Dover Publications (1979); F.
Bassani e U.M. Grassano, Fisica dello Stato Solido, Bollati Bor-
inghieri (2000). Students should choose one of them and study
it carefully. A further reason not to use these notes as a unique
reference text is their intrinsic degree of difficulty: many com-
ments, clarifications and examples I always make during lectures
are not reported here. Without them understanding very syn-
thetic written considerations may be rather hard. Yet I hope
these notes will be of some help for those students who mainly
like the very few fundamental principles upon which every phys-
ical theory is based: a dry skeleton of fundamental concepts. A
last comment: understanding these notes requires a good knowl-
edge of basic quantum mechanics even though the appendices
introduce some more advanced topics, e.g. approximate methods
like perturbation theory.
c
Figure 1: Crystals
1 Solid bodies
From a phenomenological and macroscopic point of view a solid body is
characterized by both a proper shape and a proper volume (while a liquid
owns only a proper volume and both gases and vapours neither). All solids
mechanically resist actions tending to varying their shape and/or their vol-
ume. The mechanical response, described by a stress-strain law1 , is reversible
(often linear) as long as the applied stresses do not exceed critical values:
the so called elastic regime. In this last condition shape and volume are
thermodynamic state variables like temperature. Beyond such critical stress
values solid bodies undergo irreversible processes, e.g. plastic deformation
and fracture. Crystalline solids become liquid at a characteristic tempera-
ture (melting point) with a first order phase transition. Disordered solids
instead exhibit different irreversible processes leading anyhow to a fluid state
beyond specific temperature values. Some solid bodies (for instance tung-
sten oxide  3 ) directly vaporize without passing through the liquid state:
this phase transition is called sublimation. From a microscopic (atomistic)
point of view solid bodies can be roughly classified as crystalline (monolithic
ordered atomic structures, see below), polycrystalline (bodies assembled by
many individual crystalline portions called grains, joined together through
disordered transition regions or grain boundaries) or amorphous (disordered
atomic structures). In the last twenty years, with the advent of nanotech-
1
More precisely one should speak of stress tensor and strain tensor. See Appendix on
elasticity and elastic waves
c
nology in the realm of materials science, also the terms nanocrystalline and
cluster assembled have been introduced, making the classification less sharp
and more complex. In principle the main goal of solid state physics should
be the forecast of the atomic structure of solid bodies (and thus of all their
properties) solely on the base of quantum mechanics and statistical thermo-
dynamics. This objective is impossible to be reached. The reason is that the
underlying quantum many body problem is far too difficult to be solved. In
the case of crystals the atomic structure is usually assumed as given (most
often based on experimental results like X-Ray diffraction) and then quan-
tum mechanics is used to compute the properties of that particular structure.
E.g. we are not able to understand from first principles why solid copper is
a face centered cubic crystal with a cubic cell of 3.61 Å edge but, assuming
this structure as true, we understand why copper is a noble metal, we can
compute its specific heat, electrical and thermal conductivities, optical re-
flectivity, thermal expansion and many other properties in good agreement
with experimental data. This is possible because, given the reference struc-
ture, many properties can be reconducted to elementary quantum excitations
that, to a first approximation, beahave as ensambles of independent parti-
cles which are born, move and die within the body. There exist two types of
elementary excitations: quasi-particles (quasi-electrons, holes, polarons) and
collective excitations (phonons, plasmons, polaritons, magnons). All these
excitations are deeply connected with the real particles like electrons and
nuclei which, in the normal energy range of interest for solid state physics,
are stable particles whose number is conserved. In the following we will deal
mainly with quasi-electrons (which will be simply called electrons), holes and
phonons.
1.1 Order and symmetry

Among solid bodies crystals are particularly important. Such bodies are
ideally characterized by long range order of atomic positions, which gives
crystals a unique property: translational symmetry. The term atomic po-
sition used in the crystallographic language should be understood, from a
more general point of view, as mean atomic position in conditions of ther-
modynamic equilibrium and not as instantaneous atomic position. In this
respect symmetry itself is a thermodynamic property of crystals. In real
crystals atoms vibrate around their stable mean positions (stable provided
the temperature is far from the melting point) due to thermal agitation. Fur-
thermore lattice defects (vacancies, interstitials, dislocations ecc), breaking
the translational symmetry, are unavoidably present in real crystals. Lattice
dynamics studies atomic vibrational motions. For didactic reasons we start
c
describing the properties of ideal crystals, free of both lattice defects and
thermal motions. First let us consider the so called simple crystals in which
atomic positions coincide with the points of a Bravais lattice.
1.1.1 Simple Crystals: lattices and translational order

A Bravais lattice is defined as the discrete infinite set of points (vectors)
given by the following formula:
n = 1 a1 + 2 a2 + 3 a3
1 2 3 being a triplet of integer numbers and a1  a2  a3 being the primi-
tive vectors (not all in the same plane) of the primitive cell. The primitive
cell, which contains only one lattice point, is the volume of space that, when
translated through all the vectors in a Bravais lattice, just fills all of space
without either overlapping itself or leaving voids. The vectors a are said
to generate or span the lattice. Given a Bravais lattice, the primitive vec-
tors are univocally identified (primitive unit cell ) provided the volume of
the cell is minimal and the three vectors connect the origin of the cell to
nearest neighbors (that is, to lattice points of minimal distance). Alterna-
tively the Wigner and Seitz method can be used. The lattice point taken
as origin is connected to all equivalent nearest neighbors generating a set of
segments. Then every segment is bisected by a normal plane. The set of
all such intersecting planes defines a polyhedron: the Wigner-Seitz cell, con-
taining only one Bravais lattice point, whose volume is identical to that of
the primitive unit cell. The advantage of using the Wigner-Seitz cell as the
primitive one is that all the symmetry properties of the Bravais lattice (also
the non-translational ones, see below) can be directly seen by inspection of it.
The figures also illustrate different kinds of cells: one instance of elementary
cell (minimal volume, but the connected points are not all nearest neigh-
bors) and one instance of conventional cell (non minimal volume; moreover
the one to one correspondence between lattice point n and cell is broken).
c
A lattice translation Tnn0 is the vector difference between two vectors n

0
and n and can thus be expressed by the same type of formula representing a
0 0
lattice point. If one assumes n = 0, Tn0 = Tn = n. Any two points r and
r within the crystal volume (in general neither coinciding with any lattice
point) are physically equivalent if they are connected by a lattice translation
0
r = r + Tn = r+1 a1 + 2 a2 + 3 a3
A physical observable of the crystal, e.g. the electric charge density, turns
out to be invariant with respect to any Tn :
0
(r ) = (r + Tn ) = (r)
c
0
 is then a multiply periodic function of position. In one dimension  = +
and (+) = (). Thus () is expressible as a Fourier series
X
+∞
2
() =     (1)
=−∞
whose Fourier coefficients are computed as

Z
1 2
 = ()−    (2)

0
2
The reciprocal lattice vectors  = 
 are then introduced and the series
can be written again as
X
+∞
() =    (3)
=−∞
The last is surely a periodic function because
2
  =   =    = 2() = 1 (4)
In three dimensions, for a lattice generated by periodic translation of a prim-
itive cell in the shape of a right-angled parallelepiped, the reciprocal lattice
vectors can be written in the simple form
2 2 2
g = u + u + u (5)
1 2 3
and the density can be expressed as
X
(r) =  g ·r (6)

Z
1
 = (r)−g ·r r (7)
 
Of course it turns out that
g ·T = 1 (8)
In the general case the following more complex definition of the primitive
vectors of the reciprocal lattice is needed:
a2 × a3
b1 = (9)
|a1 · a2 × a3 |
a3 × a1
b2 = (10)
|a1 · a2 × a3 |
a1 × a2
b3 = (11)
|a1 · a2 × a3 |
c
then one writes

g = 2(b1 + b2 + b3 ) (12)
being the volume of the primitive cell given by
( ) = |a1 · a2 × a3 | (13)
In this way it is always true that a · b =   and the (8) is authomatically

satisfied.
Fundamental theorems
1. Every reciprocal lattice vector g is perpendicular to a family of (di-

rect) lattice planes with Miller indices2 .
2. If the components of g have no common factors, then in general the

first neighbour planes distance  is related to g by
2
 = (14)
|g |
and, for cubic crystals (s.c., b.c.c., f.c.c.), with a conventional cubic cell
of edge , by

 = √ (15)
 +  2 + 2
2
3. The volume of the primitive cell of the reciprocal lattice can be com-
puted as
8 3 8 3
( ) = = (16)
( ) |a1 · a2 × a3 |
4. The direct lattice is the reciprocal of its reciprocal lattice. For the
primitive cell of the reciprocal lattice the most common definition is
the Wigner-Seitz one: the corresponding cell is called first Brillouin
zone. The zone center is called Γ point.
Besides from its translational symmetry each Bravais lattice is charac-

terized by its point symmetry. The point group is the set of symmetry op-
erations which leave a lattice point (taken as origin) fixed. A point group
contains rotations  of an angle 2 (with  = 2 3 4 6) about crystal
2
(hkl) denotes a plane that intercepts the three points 1 , 2 , and 3 , or some
multiple thereof. That is, the Miller indices are proportional to the inverses of the inter-
cepts of the plane, in the basis of the lattice vectors. If one of the indices is zero, it means
that the planes do not intersect that axis (the intercept is "at infinity").
c
Figure 2:
axes passing through the fixed point, the inversion  (r → −r) and the roto-
inversions (equivalent to reflections across crystal planes passing through
the fixed point) consisting of a rotation  followed by the inversion  (in
the group algebra the corrisponding operation is the product   of the two
distinct operations). All Bravais lattices own the inversion symmetry with
respect to any lattice point. The space group of the lattice contains both the
symmetry translations and the symmetry operations belonging to the point
group. Groups theory shows that (in three dimensions) do exist only four-
teen (14) different Bravais lattices (result obtained by the French physicist
Auguste Bravais in the year 1845) each charatterized by a specific primitive
cell (a specific triplet of vectors a1  a2  a3 ). The 14 Bravais lattices distribute
themselves among the seven (7) crystal systems (syngonies), illustrated in
the figure, for there are only 7 point groups compatible with translational
symmetry.
c
Figure 3:
c
1.1.2 Complex crystals: lattice and basis
Figure 4:
In the most general case a crystal can be conceptually built decorating

every primitive cell of a Bravais lattice with a specific set of  atoms called
the basis of the crystal. The basis is then the elementary building block
which, periodically translated along the space directions defined by vectors
a1  a2  a3 , reproduces the whole (in principle infinite) crystal structure. The
possible point symmetry (compatible with the translational one) in the pres-
ence of a basis is defined by the thirtytwo (32) crystallographic point groups.
The overall set of symmetry translations and of non-translational symmetry
operations3 which leave the crystal invariant can be further classified among
two hundred and thirty (230) space groups. In the following we often shall
refer only to simple crystals, unless the discussed physics does require the
presence of a basis.
1.2 Crystal binding

A further different classification of crystals hinges on the type of (chemi-
cal/physical) binding responsible for their cohesion. Thinking a solid body
as a complex many body system made of "valence" electrons and "lattice"
ions (the ions in turn being made of both nuclei and "core" electrons), we
give the following definition of cohesion energy: the energy needed to dis-
assemble the crystal into a set of neutral atoms positioned each other an
3
Combining the symmetry of a Bravais lattice with that of the basis, new symmetry
operations can arise such glide planes and screw axes.
c
infinite distance apart (without loosing any original particle). Now think-
ing to the reverse process (at zero temperature), starting from the disas-
sembled state, the formation of the stable final bound state of the crystal,
occupying a finite volume, must correspond to an energy gain (lowering of
the total energy) equal to (defining) the cohesion energy. The main acting
forces are of electrostatic origin (repulsive between charges of same sign:
electron-electron, nucleus-nucleus; attractive between charges of opposite
sign: electron-nucleus). Magnetic forces, associated with elementary mag-
netic moments, connected with electronic and nuclear spins, are not equally
important, with some exceptions. Gravitational forces are negligible (except
for oriented crystal growth). Yet, only in the case of ionic solids (see below)
the role of electrostatic interactions is fully understandable in terms of clas-
sical physics. To understand both crystal binding and properties in general a
quantum approach is needed. Within this quantum description key items are
the overlapping of atomic orbitals (to create extended states), the exchange
interactions (related to exchange symmetry of the electrons wave function
and thus, in a single particle mean field approximation, to Pauli exclusion
principle) and, in some cases, the zero point energy (a consequence of Heisen-
berg uncertainty principle). An essential role is played by the charge spatial
distribution depending on individual presence probability densities of valence
electrons and of lattice ions. In this respect metal crystals and ionic crystals
constitute two different extreme cases. In the former case the wave functions
of valence electrons fill all crystal volume with a weak periodic inhomogene-
ity of the probability density (quasi-plane wave states) within simple lattices
with the highest packing index, mainly face centered cubic (f.c.c.), e.g. ,
, ,  and hexagonal closed packed (h.c.p.), e.g.  , , , but also
body centered cubic (b.c.c.), e.g.  , ,  . Here lattice ions are treated,
to a first approximation, as hard spheres embodied in an almost uniform va-
lence electrons cloud. In the latter case (perfect ionic binding) the crystal is
a lattice of positive and negative ions (e.g. the salt + · − ) and extended
(Bloch) states for valence electrons do not exist, being all electrons in local-
ized atomic states. In the case of single element insulating crystals, e.g. 
diamond or semiconducting, e.g. Si and , binding is covalent. Extended
states of valence electrons exhibit a strong interionic localization (bonds)
from which strong bonding anisotropy follows in rather open structures with
atomic tetrahedral coordination based on 3 atomic orbitals hybridization.
III-V semiconductors, e.g. , or II-VI, e.g. , show a situation inter-
mediate between covalent and ionic binding (there are extended states with
strong bonding directionality and partial charge accumulation- of opposite
sign- on pairs of nearest lattice ions). At very low temperature noble gases
crystallize in a lattice of spherical neutral atoms bonded together because of
c
the weak interaction between instantaneous atomic dipoles, fluctuating with

zero average (van der Waals forces). Even in this case there are no extended
stationary electronic states. Lastly we find layered crystals composed by
stacking individual atomic planes (quasi 2D crystals), with strong covalent
in-plane bonds and weak interplanar van der Waals interactions.  graphite
belongs to this class. Graphite has a hexagonal Bravais lattice with a basis
of four carbon atoms. In plane bonding is due to overlapping of 22 hybrid
atomic orbitals (with a trigonal coordination) while charge fluctuations in
the cloud of 2 electrons, originating from atomic orbitals perpendicular to
covalent graphenic planes, provides weak van der Waals interactions between
pairs of such planes. Thus graphite is a solid lubricant due to the easy glide
between adjacent planes. 2 orbitals give rise to extended interplanar quasi-
free electronic states responsible for electrical conductivity and most optical
properties of this crystal. Graphite can be conceived as a sort of molecular
crystal, being the graphenic planes the composing molecules. Recently sin-
gle graphenic planes have been isolated as true 2D crystals (graphene) with
exotic optoelectronic properties deriving from electronic bands with linear
dispersion associated with the so called massless Dirac fermions.
1.3 Tensorial observables

The properties of a given crystal depend from both its structure (including
defects) and chemical composition. The physical, chemical... observables
can be classified as scalar (rank 0 tensors), vectorial (rank 1 tensors) and
tensorial (rank ≥ 2 tensors). Without pretending any generality, we examine
by means of an example the role of symmetry in determining the crystal
properties. Here we consider only spatially homogeneous observables (the
local quantities are averaged over a primitive cell). In particolar we examine
dielectric susceptibility χ representing the response of an insulating crystal
to an external electrostatic field E. Introducing the polarization vector P
(mean electric dipole moment per unit volume) it turns out that (for not too
intense fields and excluding ferroelectric crystals4 )
 = ◦   (17)
In this formula, as in all tensorial equations, the sumPfrom 1 to 3 over all

pairs of repeated indexes is understood (in our case =13 ) and ◦ is the
vacuum dielectric constant. χ = is a rank 2 tensor. In isotropic solids the
4
ferroelectric crystals can own a permanent polarization P e (as a function of tempera-
ture) indipendently from the application of an external field. In this case the law must be
modified as  = ◦   + e
c
tensor is diagonal:  =   . In this case the nine components of χ collapse

into a unique scalar  and the vector P is always parallel to the electric field
E. In general there is a variable degree of crystalline anisotropy. Let us now
consider the effect of a coordinates transformation, for instance a rotation.
After this geometric operation the new coordinates (e.g. of a lattice point)
can be expressed in terms of the old ones as:
0
 =   (18)
³ ´
where the matrix  = cos d
0
  is made by the cosines of the angles
between each new coordinate axis and each old one. Then linear algebra
gives us the tensor transformation formula:
0
 =    (19)
If the considered rotation is a symmetry operation of the crystal (that is,

0
it belongs to the crystal space group) then χ = χ and the constraints 5
 −    = 0 lower the number of independent components of tensor
χ. The number of independent tensor components can be reduced further
by other physical laws not directly related with the space symmetry of the
crystal. For example, the stress tensor   is symmetric   =   (it has
only 6 independent compenents) because of the requirements of static equi-
librium with respect to rigid rotations. The strain tensor, related to the
displacement vector u by definition as  = (12)(  +   ), is
symmetric too by its definition itself. It follows that the elastic constants
tensor  (a 4 rank tensor), expressing the elastic linear response of a crys-
tal in the generalized Hooke’s law   =   , consists of no more than 36
independent components (in front of a total number of components 34 = 81).
Different considerations of energetic character lower that number further to
21. Every residual reduction is caused by symmetry, down to cubic lattices
(s.c., b.c.c. and f.c.c.) described by only 3 independent elastic constants and
to the limiting case of isotropic solids (e.g. polycrystalline bodies) described
by only 2 independent elastic constants. A complete description of the role
of symmetry on tensor observables requires the use of group theory.
2 Scattering theory
To ascertain the static (average) structure of solids elastic scattering of probe
particles is utilized. Using undulatory terms one can say that the structural
0
5
These conditions can be written in compact form as χ = aχa−1 = χ, introducing
the matrix a−1 , the inverse of matrix a.
c
information is extractable from diffraction of the wave associated with the

particle used as probe. For X-rays the probe particle is a photon and the
wave is of electromagnetic nature. When electrons or neutrons are employed
as probes, the wave is their quantum wave function. In the case of crystals,
from the measurement of the intensity of the scattered wave as a function
of scattering angle it is possible to reconstruct both the Bravais lattice and
the atomic positions of the basis. In the following mainly the quantum scat-
tering of non-zero rest mass particles, like electrons and neutrons, will be
treated. Because of different interactions (electrons interact with internal
charges through electromagnetic forces and have a small penetration power;
neutrons interact only with nuclei through strong forces at very short dis-
tances and also sense nuclear spins) electrons are suitable to explore surface
crystallography or that of thin films, while neutrons give us bulk information
and are very sensitive probes of light elements solids, in particular if solids
contain hydrogen and its isotopes. Even X-Rays can provide bulk informa-
tion. For X-Rays the interaction takes place mainly with valence and core
electrons. In this case, beyond elastic scattering, also Compton scattering is
relevant.
2.1 Elementary theory of elastic scattering

Some general theoretical quantities involved in the explanation of elastic
scattering/diffraction experiments off crystals can be introduced in the fol-
lowing semi-phenomenological way. Let us consider a crystal hit by a mono-
chromatic plane wave. In the case of an electromagnetic wave we could
write its incident electric field as E◦ exp(k · r). Let us ussume that the
pointlike atoms of the crystal, positioned at the points n of its Bravais
lattice, react to the incident field each emitting in turn a spherical wave
E◦ exp(k · n)n exp( |r − n| ) |r − n| whose amplitude is proportional to
the incident field in n and to an atomic scattering amplitude n . Here
 = |k | = 2, being  the wavelength of the incident wave. We define the
scattered wave vector as k = r |r| = r in such a way that |k | = |k |
(elastic scattering). The angle between k and k is the scattering angle .
We set our detector of scattered waves atq the point r very far from the (finite)
¡  ¢2 p
crystal in such a way that  |r − n| =  1 − 2 r·n
+ 
'  1 − 2 r·n

'
¡ r·n
¢ √
 1 −  '  − k · n, where we have used 1 −  ' 1 − 2 valid
if  ¿ 1 (far field approximation). Then the total waveX amplitude at
the detector can be written as: E(r) ≈ E◦ exp(k · r) + E◦ exp(k ·
n
n)n exp() exp(−k · n)Now, if the incident beam is collimated and if
c
the angle  is not too small, in the simplest

Xcase of identical atoms (n = )
we can write: E(r) ≈ (E◦  exp()) exp [− (k − k ) · n], or, intro-
n X
ducing the transferred wave vector Q = k −k , E(r) ≈ (E◦  exp()) exp (−Q · n).
n
The measured intensity of the scattered wave is thus proportional to
¯ ¯2
¯X ¯
2 ¯ ¯
(Q) ∝ |E(r)| ∝ ¯ exp (−Q · n)¯
¯ n ¯
which is a function of the scattering angle as |Q| = 2 sin(2)In

X the case
of a cubic lattice n = u +  u +  u , the sum exp (−Q · n) is
n
X
−1
the product of three unidimensional sums of the type exp (−  ) =
 =0
1−exp(− )
X
∞
1−exp(− )
(a partial sum of the geometric series exp (−  ) =
 =0
1
1−exp(− )
). In the crystal there are  3 primitive cells. We can trans-
sin( 2)
form this last result, without changing its squared modulus, as sin( 2)
 In
2 ( 2) sin2 ( 2) sin2 ( 2)
this way (Q) ∝ sin   
sin2 ( 2) sin2 ( 2) sin2 ( 2)
. The study of the function
sin2 ( 2)
sin2 ( 2)
shows that,
I(x) 100
75
50
25
0
-15 -10 -5 0 5 10 15
x
  =   = 10
for big enough N, it is almost zero everywhere except where  = (2) ,
c
being  any integer (positive or negative, including zero). There the function
(an extremely sharp maximum) is equal to  2  Remembering the expression
for the reciprocal lattice vectors of a cubic lattice, we conclude that the con-
dition to have maxima in the scattering intensity (constructive interference)
is Q = g for any triplet  also corresponding to a family of crystal planes
of Miller indeces . In the next subsection this selection rule is further ex-
amined in a more rigorous way for particle scattering and shown to be fully
general.
2.2 Quantum particle scattering amplitude: Born ap-

proximation
Let us consider a steady flow of fast particles of mass , initially free and all
with the same kinetic energy ~2 |k |2 2, which interact with a crystal ideally
without thermal motions. This may be a beam of electrons or neutrons.
The stationary state of a single particle that interacts with the crystal is
determined by the mean potential energy field  (r) which has the same
symmetry of the direct lattice, and the wave function of a scattered particle
is governed by the time-independent Schroedinger equation.
~2 2
− ∇ (r)+ (r)(r) =(r) (20)
2
If
2
|k |2 = |k|2 = = 2 (21)
~2
the equation can be written as:
¡ 2 2¢ 2(r)
∇ + (r) = (r) (22)
~2
The initial state of incoming particles is described by the plane wave
 (r) =k ·r (23)
where
p = ~k = ~k (24)
is the linear momentum of the particles and k the incident wave vector. In
general an incident probability current density vector is collectively associated
to states like this
~ ~k
j = (∗ (r)∇ (r) −  (r)∇∗ (r)) = (25)
2 
c
The plane wave satisfies the homogeneous equation (empty lattice):

¡ 2 2¢
∇ +  (r) =0 (26)
0 ¡ ¢
The Green function (r r ) of operator ∇2 +2 is, by definition, the solu-
tion of the non homogeneous equation
¡ 2 2¢ 0 0
∇ + (r r ) = (r − r ) (27)
0
where (r) is the Dirac delta function. Using (r r ) the more complex
equation ¡ 2 2¢
∇ + (r) = (r) (28)
can be solved by means of a superposition integral:
Z
0 0 0
 (r) = (r)+ (r r )(r )r (29)
The explicit form of the Green function turns out to be (see e.g. Davydov,
Quantum Mechanics, Pergamon Press):
 
 0
r−r 
0 
(r r ) = − 0 (30)
4 |r − r |
The initial Schroedinger equation transforms then into the integral equation:
 
Z  0
r−r 
  0 0 0
 (r) = (r)− 0  (r )  (r )r (31)
2~2 |r − r |
¯ 0¯
If ¯r ¯   1 (far field approximation) the spherical wave emitted from
0
point r within the scattering volume can be approximated by the product of
a spherical wave emitted from the origin (always taken within the scattering
volume) and of a plane wave emitted in the direction of wave vector k =
r  
 0
r−r 
  −k ·r0
0 ≈  (32)
|r − r | 
being k = r the scattered wave vector 6 . The angle between k and k
is called scattering angle. The direction of k can be given by the angles 
¯ ¯ r ³ 0 ´2 q ³ ´
¯ 0¯ 0 0 0 0
 ¯r − r ¯ =  1 − 2 r·r
6
 + 
 '  1 − 2 r·r
 '  1 −
r·r
 '  − k · r
√
where we have used 1 −  ' 1 − 2 valid if  ¿ 1
c
and  of spherical coordinates and the infinitesimal solid angle about this
direction is Ω = sin . The radial probability current density vector
µ ¶
~ ∗  (r) ∗ (r)
j (r) =  (r) −  (r) (33)
2  
is associated to the scattered particle beam. The integral equation can be

written again as
k ·r 
 (r) = +  (34)

Seen from afar, the set of scattering centers (contained within the scatter-
ing volume  ), when hit by a monochromatic plane wave, generates a
modulated spherical wave characterized by the scattering amplitude
Z
 −k ·r
0 0 0 0
 = − 2
  (r ) (r )r (35)
2~

Within Born approximation potential energy is considered as a perturbation

with respect to kinetic energy. Consequently under the integration symbol
0 0
one writes  (r ) ≈k ·r . In this way multiple scattering is neglected and
the first order scattering amplitude is obtained as:
Z
 0 0 0
 = − 2
 (r )−Q·r r = (36)
2~


= −  k + Q|(r)|k 
2~2
where
Q = k − k (37)
is the transferred wave vector. Within this approximation the scat-
tering amplitude is proportional to the Fourier transform of index
Q of interaction potential energy. Perturbations produced by the corre-
sponding force field generate transitions between initial states |k  and final
states |k + Q = |k  both described by plane waves (far from the scat-
tering volume). The quantities  k |(r)|k  are called matrix elements of
perturbation operator (r).
The differential cross section (Ω) = (number of scattered particles per
unit time within the solid angle Ω)/(flux of incident particles) = ( 2 Ω )
turns out to be:

(Ω) = | |2 Ω (38)

c
2.3 Elastic scattering - Bragg law

In case of elastic scattering initial kinetic energy is conserved and thus |k| =
|k|  Furthermore, in crystals the potential energy owns the lattice periodicity
and can be represented as the Fourier series
X
(r) =  g ·r (39)

g being the reciprocal lattice vectors, with Fourier coefficients

Z
1
 = (r)−g ·r r (40)
 
Then the scattering amplitude can be written as:

Z X
 0
g ·r −Q·r
0 0
 = − 2
   r =
2~
 
Z
 X (g −Q)·r
0 0
= − 2
   r (41)
2~ 

If the scattering volume is much bigger than |Q|−3 and also, as it is generally
true in most scattering experiments, than  , and if the far field approxi-
mation is still valid, in practice the integration can be extended over all space
( → ∞), so obtaining
4 2  X
 = −  (Q − g ) (42)
~2 
From this last result it is seen that the scattering amplitude is different from
zero (in the real case, it is maximum) if and only if Q = g . Formula (42)
has a simple physical meaning thanks to the properties of reciprocal and
direct lattices. A particular family of direct lattice planes owns as normal
vectors the reciprocal lattice vectors labelled by the triplet of least integers
() (identical to crystallographic Miller indices of that family of planes) and
by all tripletys {} that are obtained from () multiplying every index
by the same integer . E.g. in a simple cubic lattice: (110),(220),(330),.....
Vectors g220 ,g330 ,..... share the same direction with vector g110 and for the
moduli we have: |g220 | = 2 |g110 |, |g330 | = 3 |g110 |,.....√ The first neighbour
planes distance of this family is 110 = 2 |g110 | =  2 and represents the
space period of the Bravais lattice along direction  110  perpendicular
c
to all considered planes. Let us consider the necessary condition to have

maximum scattering amplitude:
Q = k − k = g (43)
and write it as
|k − g |2 = |k |2 (44)
that is
|k |2 − 2k · g + |g |2 = |k |2 (45)
For elastic scattering and introducing the Bragg angle  as the complement
to 900 of the angle between k and g , one gets
|g | − 2 |k | sin  = 0 (46)
This law tells us that maximum scattering is obtained when the projection
of particle wave vector onto the vector g equals half modulus of g (see
Umklapp below). For  = 2 |g | and  = 2 |k | is the de Broglie
wavelength of incident particles, one eventually obtains
2 sin  =  (47)
the well known Bragg law (which gives the angular positions of Bragg peaks,
being the Bragg angle one half of scattering angle: consider the vector dia-
gram of equation k − k = g ). To arrive at (42) we used the result:
Z 0 0
(g −Q)·r r = (2)3 (g − Q) (48)
easily obtainable in 1D as
Z +∞
  = 2() (49)
−∞
considering that (see Appendix)

Z +∞ Z +
1 1
  = lim   =
2 −∞ →∞ 2 −
sin()
= lim = () (50)
→∞ 
The form sin()


is anyhow useful to describe the interaction with small crys-
tals and is the basis for the experimental determination of crystalline do-
mains size analyzing the shape of diffraction peaks. If a family of lattice
c
planes {} satisfies Bragg law g − Q = 0, the peak intensity (area) is
proportional to the squared modulus of  and dipends on the matter dis-
tribution (distribution of scattering centers) within a single primitive cell,
that is on the basis containing  atoms. In the case of neutron scattering,
due to the extremely small interaction range,Pin (40) one can write (being
the integral limited to only one cell) (r) = =1  (r − r ). The r are
the positions of the  basis nuclei of primitive cell and all coordinates refer
to the local origin of the cell itself. This form of interaction energy between
probe particles and crystal atoms is called Fermi pseudo-potential. The 
are the scattering amplitudes of nuclei. Then it turns out:
Z X
1
 =  (r − r )−g ·r r =
  =1
1 X
=  −g ·r (51)
 =1
The quantity X
 =  −g ·r (52)
=1
is called the geometric structure factor. The intensity of a single Bragg peak
() is then proportional to the squared modulus of  . Sometimes, due
to a particular basis structure generating distructive interference,  may
be null and the corresponding Bragg peak disappears (see below). In the
case of both X-Rays and electrons  is given by the same formula but
with electronic quantities  (Q = g ) substituting the nuclear constants  .
The  (Q) is an individual atomic property mainly related to core electrons.
Quantities  (Q) are called atomic form factors (see below).
2.4 X-Ray Diffraction

For X-Ray elastic scattering (XRD: X-Ray Diffraction ) most consider-
ations already introduced for particle scattering remain valid. In particular
similar formulae for both scattering amplitude and differential cross section
are obtained, provided the role of  (r) is played by the electronic charge
density (r). In fact the wave equation, in the monochromatic case, for the
electric field propagating in a medium of index of refraction (r) can be
c
reconducted to the vector Helmholtz equation7

µ 2 2
¶
2  
∇+ 2 E(r) = 0 (53)

Taking into account that (above plasma frequency, see eq. 428), for X-Rays
we can write
(r)
2 = 1 − (54)
2 0
one gets the equation
µ 2
¶
2  (r)
∇ + 2 E(r) = E(r) (55)
 2 0
with 2 2 = 2 . This equation is quite similar to the scalar equation 22. In
our analogy the role of  is then played by  :
Z
1
 = (r)−g ·r r (56)
 
if X
(r) =  (r − r ) (57)
=1
(in the assumption, not applicable to covalent solids, that the electron
charge density is the sum of the electron densities of each atom  ):
Z X
1
 ∝  (r − r )−g ·r r =
  =1
Z X
1
=  (r − r )−g ·(r−r ) −g ·r r =
  =1
X Z
1 −g ·r
=   (r)−g ·r r =
 =1 
1 X
= −g ·r  (g ) (58)
 =1
where
7
The propagation of an electromagnetic wave in a mean with refractive index  is de-
2 2
scribed by equation ∇2 E = 2 E
2 , where E is the electric field. When E = E(r) exp(−ω),
the Helmholtz equation is obtained.

c
Z
 (Q) =  (r)−Q·r r (59)

once that the diffraction condition is set Q = g .

In case of X ray diffraction we define the structure factor as:
X
 = −g ·r  (g ) (60)
=1
where  depends on the nature of the pth atom.
Monoatomic lattice with a basis:

X
 =  (g ) −g ·r = (g ) (61)
=1
P
We can calculate  = =1 −g ·r in case of a bcc lattice consid-
ered as a simple cubic lattice with a basis:
 = 1 2
r1 = 0
r2 = (2)(x + y + z)
g = (2)(x + y + z)

 = 1 + exp[−g · ((2)(x + y + z))]½= ¾
++ 2 when  +  +  is even
= 1+exp(−(++)) = 1+(−1) =
0 when  +  +  is odd
in reciprocal space this converts the simple cubic lattice with side 2
(the reciprocal of a simple cube with side ) in a fcc lattice with a conventional
cell with side 4, i.e. exactly the reciprocal of a bcc with side , for which
 is 1 (see picture 6.11 Ashcroft-Mermin).
P
Then we calculate  = =1 −g ·r in case of a fcc considered as
a simple cubic lattice with a basis:
 = 1 2
r1 = 0
r2 = (2)(x + y)
r3 = (2)(y + z)
r4 = (2)(x + z)
g = (2)(x + y + z)

c
 = 1 + exp[−g · ((2)(x + y))] + exp[−g · ((2)(x + z))] +

exp[−g · ((2)(y + z))] =
=½1 + exp(−( + )) + exp(−( + )) + exp(−( + )) = ¾
0 if in   : 2 are even and 1 is odd or 2 odd and 1 even
=
4 if in   : 3 are even or 3 are odd
in reciprocal space this converts the simple cubic lattice with side 2
(the reciprocal of a simple cube with side ) in a bcc with a conventional cell
with side 4, i.e. exactly the reciprocal of a fcc with side , for which 
is 1 (see picture 6.11 Ashcroft-Mermin).
P
Finally we calculate  = =1 −g ·r in case of diamond lattice
considered as a fcc with a basis:
 = 1 2
r1 = 0
r2 = (4)(x + y + z)
g = (b1 + b2 + b3 )

b1 = (2)(y + z − x)
b2 = (2)(z + x − y)
b3 = (2)(x + y − z)


⎧= 1 + exp(− 2 ( +  + )) = ⎫
⎨ 2 if  +  +  is two times an even number ⎬
= 1 ± , if  +  +  is odd
⎩ ⎭
0 if  +  +  is two times an odd number
Bragg formulation.
As has already been discussed in the particle scattering, Q = g is equal
to
2 sin  =  (62)
where  is the spacing between to planes in the family of crystal planes
with indices (),  is the angle between k (and k ) and the plane,  is the
wave length of radiation X,  in an integer number.
Q = g is equal Q = k − k = 2 2 
sin  =  2

= g .
Von Laue formulation.

Since |k | = |k | = , then
Q = k − k = g
k = k − g
by squaring both sides of this (and since  = |k − g |) we obtain:
c
g 1
k · = |g | (63)
|g | 2
Powder diffraction:

= 
2 sin (64)
2
where  is the angle of the diffraction (angle between k and k ).
OBSERVATIONS:
The intensity of diffraction is proportional to the Fourier trans-
form of the static autocorrelation function:
(Q) is the scattering amplitude, (Q) is the intensity experimentally
measured with scattering wave vector Q = k −k defined by experimental
conditions.
Z
(Q) ∼  (r)−Q·r r
Z Z
2 ∗ 0 −Q·r0 0
(Q) ∼ |(Q)| ∼ (Q) (Q) ∼  (r ) r  ∗ (r)Q·r r =
Z Z Z
∗ −Q·R
= (r + R) (r) rR =  (R)−Q·R R (65)
 (R) is called static autocorrelation function or Patterson function.
Z
 (R) = (r + R) ∗ (r)r
Z
x-rays:  (R) = (r + R)∗ (r)r
Z X
 X

2
neutrons:  (R) = (r + R − r ) (r − r )r =
 
X

= 2 (R − (r − r )) (66)

The formula for neutrons refers to a monoatomic material, with  equal

for any atom.
c
Liquids (neutron scattering):
Z X

2
(Q) ∼  (R − (r − r ))−Q·R R =

Ã !
X
 X X
 X

= 2 −Q·(r −r ) = 2 1+ −Q·(r −r ) =
   6=
 Z
X
2 2
=  +   (R)−Q·R R =

µ Z ¶
2 −Q·R
=  1 + (R) R (67)
where (R) is the pair-correlation function, i.e. (R)R gives the mean
numbers of atoms in a volume R around distance R from one atom in the
liquid ( (R) = (R): static non local homogeneity hypothesis).
In the hypothesis of statistical isotropy ((R) = ()):
µ Z ¶
2−Q·R
(Q) ∼  1 + () R =
µ Z ¶
2 − cos  2
=  1 + () 2 sin  =
µ Z ¶
2 2 sin 
=  1 + 4 ()  (68)

where () is the radial distribution function, i.e. 42 () is the mean
number of atoms located in a spherical shell with thickness  centered on
an atom in the liquid.
() exhibits characteristic oscillations with maxima in values of  corre-
sponding to a short range order (dynamic statistical order: first neighbours,
second neighbours, etc.).
A monoatomic gas doesn’t exhibit characteristic oscillations in ().
3 Electronic States
Physical and chemical properties in solids strictly depend on electronic mo-
tion, both stationary and not. Considering valence electrons motion as in-
dependent from motion of ions in the lattice (or from motion of nuclei, since
c
core electrons can be considered strictly bounded to them) provides just a

good approximation that will be examined in detail at the beginning of chap-
ter 4. See chapter 4 also for a better comprehension of this complex issue.
However in this chapter we will assume this independency and, moreover,
we will assume nuclei as stationary in their equilibrium positions. If these
equilibrium positions form a Bravais lattice (in case with a basis) the re-
duced many-electron problem can be simplified again and considered as a
one-electron problem in a mean periodic potential (see also paragraph im-
mediately after equation (220)). We start trying to solve the equation for
stationary states for an isolated multielectron system:
b  (r) = ()  (r)
H (69)
where
H b0 = b +  (r) +  (r)
b=H (70)
and
X |pb |2 ~2 X 2
b = =− ∇r (71)

2 2 
is the total kinetic energy operator for valence electrons,
1X 2
 (r) = (72)
2 6= 40 |r − r |
is the Coulomb potential energy associated with repulsive interactions be-

tween electrons,
XX 2
 (r) = − (73)
 l
40 |r − l|
is the Coulomb potential energy associated with attractive interactions electron-
ion.
For simplicity we considered a simple monoatomic crystal: l indicating
the positions of nuclei/ions and  their bare charge.
r = (r1 s1  r2 s2   r s   r s )
represents the set of dynamic coordinates (position and spin) of all valence
electrons.
We neglect relativistic effects (including spin-orbit interactions and direct
magnetic spin-spin interactions). With a procedure similar to the Hartree or
Hartree-Fock self-consistent mean-field approximation used for many-electron
c
atoms (see appendix), we simplify the previous many-body problem to an ef-

fective one-electron problem
b  (r ) =   (r )
H (74)
  
b |pb |2
H = + h (r )i +  (r ) (75)
2
hi representing the average value of electron-electron interactions. Here
we do a further approximation: we assume that the mean periodic potential
h (r )i +  (r ) = (r ) is the same for every electron (a good approx-
imation for macroscopic crystals): for this reason by now on we drop the
subscript .
From a physical point of view the sum of h (r)i and of the bare periodic
potential  (r) is equivalent to the electrostatic screening of the latter.
In other words: the generic valence electron moves in an attractive pe-
riodic potential (r) resulting from the interaction with lattice ions, whose
positive charge  is partially screened from the mean negative charge dis-
tribution of all other valence electrons. This mean field approach is static
in nature and neglects the dynamic correlation effects present within the
screening electrons cloud surrounding ions (and the screening holes cloud
surrounding every generic electron, see below the introduction to the quasi-
particle concept).
In a Fourier representation of periodic potential this Hartree-Fock approx-
imation can be described by a dielectric function depending on wave vector
(see Bloch’s theorem). This dielectric function substitutes 0 in a formula in
which the Fourier transform of  (r) appears.
3.1 Bloch’s Theorem

b n is defined as follows:
The translation operator T
b n (r) = (r + Tn ) = (r+1 a1 + 2 a2 + 3 a3 )
T (76)
In a one dimensional space:

b n () = (+Tn ) = (+)
T (77)
We consider now the problem of determining stationary electronic states in

the lattice within an independent electron approximation in a mean periodic
potential:
(+Tn ) =  (+) =  () (78)
c
We have to solve the eigenvalue equation:

b = 
 (79)
where
2
b = b + ()
 (80)
2
b is invariant under translations such as T
Since  b n:
b T
[ b n] = 0 (81)
we can actually write
b n ()()
T b b + )( + ) = ()
= ( b T b n ()
From a general principle of quantum mechanics, if the commutator between

two operators vanishes, then the two operators share the same set of eigen-
b n commute, it is possible to write:
b and T
functions. In this case, since 
b n () = (n)()
T (82)
b n is an additive operator, which means:
T
T b n2 () = (+1  + 2 ) =
b n1 T (83)
b n1 +n2 ()
= (+(1 + 2 )) = T
Using now the eigenvalue equation we obtain:
(n1 )(n2 )() = (n1 + n2 )() (84)
(n1 )(n2 ) = (n1 + n2 ) (85)
This equation is satisfied by
(n) =  (86)

for any choice of the complex number , provided that it has dimensions of
reciprocal length.
Exploiting periodicity we apply the normalization condition within a sin-
gle primitive cell of volume  :
Z Z ¯ ¯2
2 ¯b ¯
1 = |()|  = ¯Tn ()¯  = (87)
 
Z
2 2
= |(n)| |()| 

c
i.e.
|(n)|2 = 1 = | |2 (88)

then, necessarily
 =  (89)
with  real. Therefore:
b n () = (+Tn ) = (+) =  ()

T (90)
This property is satisfied by any  () (Bloch wavefunction):
 () =  () (91)
in which
 (+Tn ) =  (+) =  () (92)
is a periodic function in . From a physical (intuitive) point of view it is
important to consider that Bloch functions are plane waves modulated by a
periodic factor.
To any eigenfunction of b ( () is a stationary state of an electron in
a periodic potential)
b  () =   ()
 (93)
a wave vector  is associated such that equation (90) holds, and therefore
also equations (91) and (92) are true.
Stationary energy levels  associated to the electron depend on wave
vector  too. In 3D we generalize the above result using the vector notation:
(r + T ) = k·T k (r) (94)

where
k (r) = k (r)k·r (95)
and
k (r + T ) = k (r) (96)
Applying the momentum operator p b = −~∇ to a Bloch wave and to a

plane wave leads to two different result. For a Bloch wave:
bk (r) 6= ~kk (r)

p (97)
This means that a Bloch wave is not an eigenfunction of momentum

operator as instead a simple plane wave is. Thus the term ~k is called
c
quasimomentum or crystal momentum. The mean momentum (the quantum

expectation value of momentum in a Bloch state), instead, is equal to (for a
complete derivation see Appendix E, Ashcroft and Mermin):
 k
 p = =v (98)
~ k
Normally Bloch stationary states are propagating waves unless  k
k
=
0 (In which case we have a standing wave. This can happen at specific
 points: see below other typical bands structures k ). The existence of
stationary electronic propagating waves in a periodic accelerating potential
is a purely quantum effect. Since in real crystals the electron motion usually
encounters some resistance, there must be something that alters the perfect
periodic structure of the crystal, a periodicity defect. Insertion of the Bloch
wave function in the form 91 into 3D version of eq. 93 yields the following
eigenvalue problem:
~2 ¡ 2 ¢
− ∇ k + 2k · ∇k − 2 k +  (r)k = k k (99)
2
id est:
" #
p + ~k)2
(b
+ (r) k = k k (100)
2
b = −~∇ is the momentum operator.
where p
For any fixed k, the eigenvalue problem (100) defines a spectrum of dis-
crete eigenvalues k = k . Every energy level k is labelled with a branch
index  which stands for the set of quantum numbers needed to label the
eigenvalue spectrum at fixed k. Usually these k are degenerate energy lev-
els: all states belonging to the same branch  can be written as different
Bloch functions  
k () = k ()
k·r
, where now  represents the subset of
quantum numbers required to span all  states.  depend on the non-
translational part of the space group of the crystal. By fixing  and letting
k vary we obtain a hypersurface (a line in 1D, a surface in 2D): a specific
electronic branch of the band structure of the crystal. Letting both  and k
vary we obtain the entire band structure. In this context a specific band  is
the whole energy sub-interval spanned by all k values. Sometimes the word
band is used to indicate a branch, generating some confusion. In particular,
adopting our nomenclature, we can say that, in some crystal structures, dif-
ferent energy bands can be (partially) superimposed. As far as the physical
meaning of the single electron energy levels k is concerned, if the peri-
odic potential is the self consistent one obtained by the Hartee-Fock method
c
(see appendix on self consistent mean field) assuming a Slater determinant

crystal wavefunction, describing all itinerant electrons, built by all occupied
spin-orbitals, with orbital parts constituted by the one-electron Bloch states
k () = k ()k·r , we can state two relevant results. First (Koopman’s the-
orem): −k equals the ionization energy of a single electron initially in state
k () being the global initial state the minimum energy one (ground state).

Second: k − k ≈ ∆  approximately equals the transition energy be-
tween the initial state k (), belonging to the global ground state, and the

final state k (), belonging to an excited global state. The two results are
tenable provided all other spin-orbital are unchanged in the involved process
(no relaxation). The fact that both the Coulomb and the exchange inte-
grals present in Hartee-Fock theory scale as 1 where  is the number
of primitive cells make the above statements particularly good for itiner-
ant states such as Bloch waves (see e.g. the tight binding method). For a
proof (including the definitions of Coulomb and exchange integrals) see the
advanced textbook G. Grosso & G. Pastori Parravicini Solid State Physics
(2-nd ed.) Academic Press (2014), pag. 146. If instead the present version
of the Density Functional Approximation is used, the k have generally no
direct physical meaning (ibidem pag. 163).
3.2 Reduction to the first Brillouin zone

In one dimension. Let us first consider  =  ( being any reciprocal lattice
vector). In this case  (), associated to the eigenvalue  , is a periodic
function with the same space periodicity of the lattice.
 ( + ) =  ( + )(+) = (101)


=  () =  ()
We can generalize that to a 3D situation as:
g (r + T ) = g (r) (102)
We consider now, for the sake of simplicity, a generic Bloch wavefunction

 =   ()

in a one dimensional situation; let ,  and , be fixed.

The state  , eigenfunction of  b and related to the eigenvalue/energetic
level  , is also eigenfunction of bn and thus related to the eigenvalue  :
bn 
 = 
 

Since there is a one-to-one correspondence between the eigenvalue 
and the energy level  , and since  is a periodic function of  with
c
periodicity  = 2, also  will be a periodic function of  with the same
periodicity.
The representation of the states as functions of , with −∞    +∞
(repeated-zone scheme), is so highly redundant, since the same branch seg-
ment is shown in all infinitely many equivalent wave vector periods. Fixing
univocally the  corresponding to a specific energy level/state is now neces-
sary.
This is achieved using the reduced-zone scheme where we consider only the
wave vectors belonging to the first Brillouin zone, in such a way to associate
to any state  of branch  a single wave vector .
In one dimension this corresponds to limiting the vector  within the
interval:
 
− ≤ (103)
 
A third representation is the extended-zone scheme in which in any Bril-
louin zone there is just one electronic branch. Starting in the first zone and
increasing, in any  direction, the modulus of the wave vector, we will as-
sign to any zone only one specific branch , in such a way to progessively
increase the  index (the first branch in the first zone, the second branch in
the second zone and so on).
Umklapp of free electron parabola .

As a limiting case, we consider now an empty (periodic!) lattice. This
physically corresponds to a vanishingly small periodic potential. The Bloch
states reduce to plane waves  −12 k·r (they describe the motion of a free
electron) and the energy levels associated to them are given by the follow-
ing formula, which can be considered also as a valid representation in the
extended-zone scheme with an arbitrary space period.
~2 |k|2
k =
2
c
The picture shows the result of this fictitious periodicity of an empty lattice.
In order to obtain the reduced-zone scheme we must back translate (umklapp)
of reciprocal lattice vectors, the different pieces of the parabola from any
higher order Brillouin zone into the first Brillouin zone.
A real periodic potential would introduce a distortion of the parabola and,
generally, the opening of a prohibited energy gap between two consecutive
bands.
Writing the Bloch functions in their characteristic expressions can def-
initely persuade us that the states described with wave vector k and with
wave vector k0 = k − g are exactly the same state.
0

k0
(r) = 
k0 (r)
k ·r
= 
k0 (r)
−g·r k·r
 = 
k (r)
k·r
= 
k (r)
In the formula we defined  

k (r) = k0 (r)
−g·r
, since −g·r and  (r)
¯ ¯0
k
¯  ¯2
in the direct lattice have the same periodicity. Notice that ¯k (r)¯ =
¯ ¯2
¯  ¯
¯k−g (r)¯ .
3.3 Born-von Karman boundary conditions

Since real crystals are finite, they are not perfectly periodic. Yet, by neglect-
ing surface states and introducing appropriate boundary conditions compat-
ible with translational invariance, Bloch’s theorem still holds and provides
a suitable approximation for the real case. In one dimension, considering
N (N1) primitive cells in a crystal of length  =  we can write the
periodic boundary conditions as:
 ( + ) = () =  () (104)

which requires  = 1This last equation implies that the possible (al-
lowed) wave vectors belong to the discrete set of values:
2
 =  (105)

In the
(−   ≤ ) we have exactly N independent  corresponding to
integer  varying between − 2 and 2 :
 
− ≤ (106)
2 2
c
Since ∆ = +1 −  = 2  2, wave vector  nearly form a

continuous set of values. The situation in 3D is very similar to this one. We
consider  = 1 2 3 cells, in the first Brillouin zone the  allowed wave
vectors are given by the following formula:
µ ¶
1 2 3
k(1 2 3 ) = 2 b1 + b2 + b3 (107)
1 2 3
where
 
−   ≤ (108)
2 2
 = 1 2 3
Taking into account spin degeneracy and Pauli principle, it follows that a
single branch k of the band structure (in the first Brillouin zone) can guest
up to 2 electronic states (Bloch waves). As a result, knowing the band
structure of a crystal, including the existence and width of forbidden energy
gaps, it is possible to foresee whether the material is a conductor (metal) or a
semiconductor/insulator. As an example we consider a monovalent element
which crystallizes as a simple crystal, such as  (a body centered cubic
structure). In this case, in the ground state, the set of valence electrons
occupies just the first half of k1 , leaving the second half empty. If acceler-
ated by an external electrical field (see below "Single electron Dynamics")
there will be a net current generated by the highest energy electrons making
transitions to close empty energy states belonging to the same band: such
a crystal is a conductor. Following similar lines of reasoning, to get a more
general rule of thumb, for the sake of simplicity we assume that contigu-
ous allowed bands, spanned by the different branches k , are separated by
forbidden energy intervals called band gaps, i.e. ranges of energy that an
electron in the solid may not have (see e.g. "Nearly-free electron bands").
Let  be the atoms valence and let  be the number of atoms in the basis;
if  is odd, the crystal should be a conductor, otherwise if  is even, the
crystal should be an insulator (or a semiconductor, if the width of the first
prohibited band is less than or equal to   ). However some real situations
are more complicated and there are a lot of important counterexamples of
our rough rule. For example all of the elements in the second column of the
periodic table ( = 2) form metallic crystals, even though they are simple
crystals (one atom per primitive cell). In these cases we should consider the
complex topology of the real band structure in three dimensional  space.
Different branches  with energy k can exhibit a partial superposition of
the relative energy bands, along different directions in  space. This is true
for the above case (e.g for ): as a consequence the first band is not full
c
and the second is also partly occupied giving rise to a metallic behavior. We
will see that the empty states of the first band seem to behave as positive
charge carriers (holes). In one dimension superposition of bands is not al-
lowed because it would imply branch crossing: in a 1D world  would be
an insulator (or a semiconductor) according to our rule of thumb.
3.4 Free electron gas

3.4.1 Plane waves - Fermi energy - Density of states
If we neglect completely the periodic potential, the one electron Hamiltonian
coincides with the kinetic energy operator:
2
b = − ~ ∇2 (109)
2
and the equation for stationary states becomes
~2 2
− ∇ k (r) = k k (r) (110)
2
We already know that eigenvalues (degenerate energy levels) in this case are
given by the following expression
~2 |k|2 ~2 (2 + 2 + 2 )

k = = (111)
2 2
and that the corresponding eigenfunctions are the plane waves
1
k (r) = √ k·r (112)

where  is the crystal volume. Vectors k are the discrete ones given by
the periodical boundary conditions. If we consider the set of allowed wave
vectors (corresponding to a lattice of points in  space) we can notice that
there is a volume (2)3  corresponding to any vector/point/orbital state
(plane wave). The dispersion relation (111) shows that the √ locus of points
of constant energy  is a spherical surface of radius () = 2~. The
number of states with energy between 0 and  is then
4
()3
() = 2 3 3 (113)
(2) 
the factor 2 is introduced to take into account the spin degeneracy. The
density of states thus assumes expression
µ ¶3
  2 2 √
() = = 2  (114)
 2 ~2
c
The free electron gas ground state (at absolute zero) is determined by
ordering the states of the system (112) by increasing energy, and by con-
secutively filling up the unoccupied quantum states with the lowest energy,
according to the Pauli exclusion principle. When all the electrons have been
put in, the Fermi energy  is the energy of the highest occupied state. The
corresponding spherical surface has a radius of  = ( ). The Fermi wave
vector is obtained from (113) by equalizing ( ) to the number of electrons
 contained in the volume 
µ ¶ 13
2
 = 3

¡ ¢ 23
~2 3 2 

 = (115)
2
Both Fermi wave vector and Fermi energy depend on valence electron
density (that is the number of valence electrons per unit volume). The prob-
ability that, at a temperature  , a state with energy  will be occupied is
given by the Fermi-Dirac distribution  (| )
1
(| ) = (116)
exp( −
 
) +1
where  is the chemical potential of an electron. Since  depends weakly
on the temperature (see "Electronic specific heat"), in most of the cases the
approximation ( ) ≈ (0) =  can be used. Wecan now introduce now
the density of occupied states
µ ¶ 32 √
 2 
D(| ) = () (| ) = 2 (117)
2 ~2 exp( −
 
)+1
Obviously for an isolated conductor we have

Z ∞ Z 
D(| ) = () =  (118)
0 0
When the Fermi wavelength  = 2 has a value comparable with the
interatomic distances (in 1 when  = 2) we expect a Bragg reflection
(electron diffraction, see "Scattering theory") to occur off a family of lattice
planes for electrons on the Fermi surface. In this case the free electron model
fails.
c
3.4.2 Chemical potential and electronic specific heat

We start with a property (pointed out by Sommerfeld) of Fermi-Dirac dis-
tribution related to the fact that by differentiating the (116) with respect
to energy we obtain a function similar (but not identical) to the Dirac delta
function (red curve showed in the figure, with a negative sign). The width
at half maximum of this function is ∼   ¿  ≈  . Thus the exact delta
function behaviour is exhibited only at  → 0 
f 3
2.5
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3
E
In this picture chemical potential is 2 a.u.
To make the above assumption stronger, we observe that:

Z ∞
(| )
−  = (0| ) − (∞| ) = 1 (119)
0 
The property is the following
Z ∞ Z  µ ¶
2 
()(| ) ≈ () + (  )2 (120)
0 0 6  =
For a proof see, e.g., Ziman, Principles of the theory of solids 2-nd ed.,
Cambridge University Press. There, to derive the previous formula, the
following identity (obtained by integration by parts) is used:
Z ∞ Z ∞ µ ¶
(| )
() (| ) = () −  (121)
0 0 
R 0 0
where () = 0 ( ) . Then (120) can be written in another form
(useful in applications) as:
Z ∞ µ ¶ µ ¶
 (| ) 2  2 
() −  ≈ () + 2
(  )2 (122)
0  6  =
c
When the temperature dependent term can be neglected it is possible to

write −(| ) ≈ ( − ). We can now use eq. (120) to determine
the dependence of the chemical potential on temperature. We write (118) as
Z  Z  µ ¶
 2 
= () ≈ () + (  )2 (123)
0 0 6  
Since Z Z Z
  
() = () + () (124)
0 0 
It is possible to obtain
Z  µ ¶
2 
() ≈ (  )2 ≈ ( ) ( − ) (125)
 6  
Resolving the previous formula with respect to the chemical potential we get
µ ¶
2 
 ≈  − ln () (  )2 (126)
6  
The apparently rough approximation  ≈  is usually acceptable even be-

yond room temperature. The electronic contribution to specific heat could
also be found by differentiating the internal energy with respect to temper-
ature, being Z ∞
= ()(| ) (127)
0
Here we use a more direct approach. Because of the exclusion principle, at
temperature  , only  electrons
Z  +
 
2
 = () ≈ ( )  (128)
 
 − 2
are thermally excited. Each of them contributes to the part of internal energy
depending on the temperature  =   with an individual contribution
 Then
  £ ¤
 ≈ = ( ) (  )2 = 22 ( ) (129)
 
The exact calculation would give us the constant  2 3 instead of the prefactor
2 of the previous formula. In conductors this contribution to specific heat is
important at low temperature and gives us a way to measure ( ).
c
3.4.3 Quasi-particles and Fermi liquid theory (a qualitative intro-

duction)
The dependence of density of occupied states on temperature shows us that,
due to the exclusion principle, just a tiny amount of electrons with energy
close to Fermi energy (in an interval  ± 2 ) can be thermally excited.
In the scheme of single particle states in a mean field, we can say that the
excited state is defined both by the wave functions of the (few) electrons with
an energy bigger than the Fermi one (more precisely, bigger than the chemical
potential) and by the wave functions of the (many) remaining electrons with
an energy smaller than the Fermi one. In the ground state all states up to 
are occupied and all states above it are empty.  is of the order of some 
(for example in  the value is  = 31  ), while at room temperature we
have   ≈ 25×10−3  . The above "electron representation" of the excited
state is very redundant. In order to get closer to the (quasi)electrons and
holes representation we will use later, we introduce here, in a qualitative way,
some important concepts due to Lev Landau. We linearize the dispersion
relation about the energy point  ≈ 
µ ¶

 ≈+ ( −  ) =  ( −  )
 
where the Fermi velocity is  = }−1 () and where  = } is the

momentum of the quasi particles. If we consider  as the new zero of the
energy scale, the dispersion relation becomes  =  ( −  ). This is valid
only for excited electrons with    . In the limit of validity of the lin-
earization we have unoccupied states if    . Afterwards we will see that
the excited state can be described by few particles with negative charge (elec-
trons) for which the previous dispersion relation is valid with  =   0
and  =    and by equally few particles with positive charge (holes)
for which is valid the dispersion relation obtained by the previous formula
multiplied by −1:  = − =  ( −  ) with    . For the holes
this transformation is equivalent to a time reversal transformation, as origi-
nally shown by Feynman. Electrons attract holes and holes attract electrons.
Electrons with a cloud of holes and holes with a cloud of electrons are called
quasi-particles and acquire an effective mass. Furthermore, since the dy-
namic cloud is not perfectly stationary, the quasi-particles have a finite life
time  () beyond which they are annihilated by collisions. Even neglecting
this electron-electron effects (the cloud), just introducing the periodic poten-
tial, the velocity and the mass of electrons and holes can be changed.The
above considerations will help us to explain why the mass of an electron in
a crystal lattice is different from that of the electron in vacuum. Moreover
c
two quasi-electrons repel each other less than two electrons separated by the
same distance in vacuum. This can be represented by introducing a dielectic
constant different from the vacuum one. All the same, mutatis mutandis,
is valid for interactions between two holes and for electron-hole interactions.
The number of quasi-particles is not constant (as is the number of electrons)
but depends on temperature.The complex dynamic system described in this
paragraph is called Fermi liquid. The theoretical framework needed to take
into account many body effects in condensed matter is a particular aspect of
Quantum Field Theory. We will not enter this complex field. The interested
reader is referred to R.D. Mattuck, A Guide to Feynman Diagrams in the
Many-Body Problem, 2-nd ed. Dover 1992.
3.5 Band structure: nearly free electron approxima-

tion
3.5.1 Free electron diffraction
If we allow alkaline atomic elements to crystallize, due to their weak atomic
first ionization potential, we can consider a correspondingly weak periodic
potential () and treat it as a static perturbation (in comparison to kinetic
energy). Then standard 2-nd order perturbation theory (see Appendix on
"Approximate Methods") gives us the following expression for the perturbed
energy levels ():
¯ ¯2
¯ 0 ¯
2
~ |k| 2 X ¯  k| (r)|k  ¯
(k) ≈ +  k|(r)|k  + ³ ´ (130)
2 
~2
|k| 2
− |k 0 2
|
k 6=k 2
Here we have adopted the Dirac notation. The unperturbed states |  are
the plane waves in the vacuum lattice (see above). We have to examine the
matrix elements
Z
0 1 1 0
 k| (r)|k ≡ √ −k·r  (r) √ k ·r r = (131)
  
Z
1 0
= −(k−k )·r (r)r
 
The periodic potential energy () can be represented as a Fourier series

X
(r) = g g·r (132)
g
c
where the  are the reciprocal lattice vectors and where the Fourier coeffi-
cients g can be written as:
Z
1
g = (r)−g·r r (133)
 
Thus:
Z X
0 1 0
 k| (r)|k = g g·r −(k−k )·r r = (134)
  g
Z
1 X 0
= g (g−k+k )·r r =g  k0 k−g
 g 
In order to obtain (134) we used the periodic boundary conditions which

give us a discretized set of wave vectors (105). A similar treatment can be
found above in "Scattering theory" where a continuous set of wave vectors
0
is used. So, unless  =  − , the elements
R of the matrix are zero and the
1
diagonal element  |()|  =   () represents the mean value
of potential energy. If we set this value to zero we can write the expression
for perturbed energy levels as:
~2 |k|2 X |g |2
(k) ≈ + ¡ 2 ¢ (135)
2 g6=0
~2
2
|k| − |k − g|2
This representation makes sense for any  with the exception for the ones
that satisfy the degeneracy condition (525)
 (0) (k) = (0) (k − g) (136)
i.e., using the (111):

|k| = |k − g| (137)
This condition is equivalent to
|k|2 = |k|2 −2k · g + |g|2 (138)
and consequently to
g |g|
k· = (139)
|g| 2
For a better understanding of this equation in terms of Bragg reflection read
the subsection Elastic Scattering - Bragg law. This condition in one dimen-
sion is satisfied by  = ± = ± (). In this case the perturbation theory
c
for degenerate energy levels (see "Approximate methods") must be applied

first, using only two basis wave functions as
1  1 
± = √    ∓ √ −   (140)
2 2
¡ ¢
~2  2
The unperturbed degenerate level  (0) (±  ) = 2 
so splits into two
non-degenerate energy levels
± ~2 ³  ´2 ¯¯ ¯
¯
 = ± ¯ 2 ¯ (141)
2  
between which a forbidden energy gap is generated

¯ ¯
¯ ¯
 = 2 ¯ 2 ¯ (142)

¯ ¯2
If we examine the probability densities ¯± ¯ we can find that
µ ¶
¯ + ¯2 2 
¯ ¯ = sin2 ( ) (143)
 
µ ¶
¯ − ¯2 2 
¯ ¯ = cos2 ( ) (144)
 
¯ − ¯2
¯ ¯ has its maximum values where  =  (we can find the minima of
the
¯ + ¯periodic potential in correspondence with the ions: lattice points), while
¯ ¯2 has its maximum values at interstitial positions between ions, i.e. at
¡ ¢
 + 12  (maxima of the periodic potential). This explains the difference of
energy between the two different perturbed states. Since the degeneration has
been removed it is now possible to use the general formula for non-degenerate
energy levels for all wave vectors in the first Brillouin zone, including the ones
at the zone boundaries, to further improve the accuracy of the method.
3.5.2 Fermi surface and density of states

The Fermi surface is the set of surfaces characterized by equations  =
  (k) for each partially occupied zone . The bigger is the periodic potential,
the stronger is the deviation of the shape of this surface from that of the free
electron sphere. The density of states as a function of energy for a given
branch  is derived by integrating the density of states 2  (2)3 in space
k over the infinitesimal volume bounded by the two surfaces with equation
 =   (k) and  +  =   (k) and dividing the result by :
c
Z
   () 2
 () = = k = (145)
 (2)3 
Z Z
2  
= 3  ⊥ = 3
(2)  4 |∇k   |
where  is the infinitesimal element of the first surface while ⊥ represents
the local distance between the two surfaces. We used the following formula:
 = |∇k   | ⊥ . Thus the total density of states is:
X Z
  X 
() =  () = 3 (146)

4  |∇k   |
If we plot the density of states as a function of energy, where the mean velocity
of a Bloch state  k  = (∇k   ~) is zero (this is a critical point in the first
Brillouin zone) we find discontinuities in () called Van Hove singu-
larities (the virtual discontinuities in () are smoothed out by integration).
Since (−k) =(k), as stated by the Kramers theorem (symmetry − ),
Γ [(k = 0)] is always a critical point. In lattices such as f.c.c., because of
symmetry reasons, also  [k =(2)(100)] and  [k =(2)(12 12 12)]
are critical points. Segment Γ is called Λ while segment Γ is called ∆.
For a given branch , close to a critical point it is possible to write  = (k)
as a quadratic form; using the directions of the local principal axes we get:
=0 +  2 +  2 +  2 (147)

The generic critical point is indicated with  where  represents the number
of negative coefficients in the above expression. As an example we consider
the density of states near 0 in an isotropic situation:  =  =  =  
0. Using polar coordinates |k|   we obtain:
Z p
 |k|2 sin  
() = 3    = 2 32 −0 (148)
4  |k|= − 
0 2 |k| 4 
Using the effective mass ∗ approximation (see below) in this case we would
get  = ~2 2∗ . Similar considerations may be applied to the study of
phonon dispersion relations (see below) in order to obtain their density of
states in terms of frequency.
3.6 Band structure of tightly bound electrons

While in the interstitial space between two lattice ions, where the periodic
potential is nearly constant, the Bloch wave functions are very similar to
c
plane waves, close to the positions where lattice ions are located Bloch waves
resemble localized atomic orbitals. In other words we assume that in the
vicinity of each lattice point in position n the periodic potential can be
approximated by the atomic potential of a single ion located in the same
position n (r − n). In order to describe electronic bands in crystals with
strongly bounded electrons we try to develop a Bloch state as an appropriate
sum, over all positions in the crystal, of atomic orbitals  (r − n) centered
on ions themselves ( is the set of quantum numbers defining one of the
possible stationary state for an electron in the mean field of central forces of
a single atom). Thus  satisfies the stationary Schroedinger equation (we
put  = 0):
∙ ¸
~2 2
− ∇ +0 (r)  (r)=  (r) (149)
2
Then a trial Bloch function of the whole crystal can be written as:
1 X k·n
k (r) = √   (r − n) (150)
 n
Computing k (r + l), being l too a lattice translation like n, and multiplying
the right side of the equation by k·l −k·l we get :
1 X k·(n−l)
k (r + l) = k·l √   (r − (n − l)) =
 n
1 X k·h
= k·l √   (r − h) = k·l k (r)
 h
Since both l and n − l = h are lattice translations, we have just verified that
(150) satisfies Bloch’s theorem. Neglecting superposition integrals, function
(150) satisfies the normalization condition, i.e.:
Z Z
1 X X k·(n0 −n) 0
∗ 
k (r) k (r)r = 1+  ∗ (r − n) (r − n )r ≈1
 n 0
n 6=n
(151)
The expectation value of energy in state k will be then:
Z ∙ ¸
∗ ~2 2

 (k) = k (r) − ∇ +(r) k (r)r (152)
2
where (r) is the exact periodic potential. Introducing ∆(r) = (r)−0 (r),
a function with maximum at r = 0 (we set lattice point n as origin of coor-
dinates), and again neglecting some superposition integrals, we get the final
result:
c
X
  (k) = Eh k·h (153)
h
with:
E0 ≈  (154)
Z
Eh6=0 ≈ ∗ (r + h)∆ (r) (r)r
Equation (153) is the exact Fourier representation of a periodic function of

k (see below Wannier theorem); here we use for it the approximate expression
(154) of Fourier coefficients valid within the tight binding approximation.
Because of the nature of both ∆ (r) and  (r), we expect Eh to be very
small, with the exception of first neighbors. In one dimension ( = ),
considering that E± is negative due to the attractive nature of the periodic
potential with respect to the vacuum level, it is possible to write:
  () ≈ − 2 |E± | cos() (155)
So we have a band with a minimum min =  − 2 |E± | at  = 0 and a

maximum max =  + 2 |E± | at the zone boundary edges where  = ±:
therefore the bandwidth is 4 |E± |. In case of a strong bond the hopping
probability from one site to the next (proportional to Eh ) is small and the
band is quite flat (localized atomic level).
E(k)
ak
"tight binding" band

The black curve represents   (), the red curve the atomic level, the green
curve the parabolic approximation. We have set min = 0. For small values
of  we can approximate the band as:
 ~2  2
 ()≈ ∗ (156)
2
c
Comparing with (155) we obtain for the effective mass:
~2
∗ = (157)
22 |E± |
So, electrons in bands of strongly localized states have a big effective mass.
Remembering the  discrete allowed wavevectors of the first Brillouin zone,
in a crystal (a polystable system with  ions - i.e.  potential wells) the
original atomic level  splits in a band of  levels   () as a result of
the above hopping events (transitions) between different wells. Jumps are
promoted by ∆ , acting as a perturbation operator in (154) which resemble
a matrix element of a transition in perturbation theory (see Appendices).
3.6.1 LCAO Method

A "tight binding" energy band built with a given atomic state  can exhibit
a partial energy superposition with another one built with a different state
 (corresponding to a different atomic energy level  ). Along a given 
direction this generally happens when |k| is greater than a threshold value
in the first zone. In this case there is a loss of significance in the concept
of distinct band  and band  (take the example of band 4 and band 3 in
transition elements as shown, e.g., in Ziman, Theory of Solids (2-nd ed.) pag.
113). This inconvenient can be eliminated using a linear combination of all
(or almost all) different atomic orbitals  (and not only one) corrisponding
approximately to the same atomic energy  ( =  within the same
atomic schell, for example the valence shell 22 of Carbon, containing four
valence electrons and eight possible atomic orbitals, some filled and some
empty) and so obtaining a generalization of the above trial tight binding
wave functions, that is the LCAO Bloch functions:
1 X X k·n

k (r) = √     (r − n) (158)
 n 
The variational coefficients   can be determined minimizing
Z ∙ ¸
∗ ~2 2
 
(k) = k (r) − ∇ +(r) 
k (r)r (159)
2
In interstitial regions atomic states  (r − n) and Bloch states are very dif-
ferent one another; this happens because of energy barriers between pairs
of ions which appear in the periodic potential and which alter the potential
wells produced by isolated ions. Besides bound atomic states are an incom-
plete set for the description of a Bloch state with energy greater than periodic
c
potential maxima (non tunneling states), where the energy spectrum is basi-
cally continuous (see unbound states with positive energy in isolated atoms).
In the figure below the function ∆ is shown together with two different
energy levels greater and smaller than periodic potential maxima. Thus it
was necessary to look for better approximate methods based on basis func-
tions closer to exact Bloch functions, but the description of such methods is
beyond the scope of these lecture notes.
∆() and periodic potential
3.7 Electron dynamics in external fields

Being a consequence of translational invariance, equation (153) has a general
validity and gives us the possibility to prove a theorem which is very crucial
in the study of electron dynamics, when the crystal is plunged in an exter-
nal field. The statement of Wannier theorem is contained in the following
identity
  (−∇)k (r) =   (k)k (r) (160)

where  b =   (−∇) is the operator obtained from function   (k) in which
the following substitution is made:
k → − ∇ (161)
We prove the theorem dropping subscript  and considering only one band in
a one dimensional situation where the lattice positions are given by  = .
We apply operator  b =   (−∇), built starting from (k) expressed as
the Fourier series (153), to Bloch wave function k () and perform just a
few steps:
c
 X 
(− )k () = E   k () = (162)
 
X  1 2
= E (1 +  + ()2 2 + )k () =

 2 
X X
= E k (+) = E  k () = ()k ()
 
c.v.d. In the last step we used Bloch’s theorem.
3.7.1 Equivalent Hamiltonian. Effective mass theorem

We consider now the motion of an electron in an external field (for example
an electric field); let us assume that the electron initially occupies a Bloch
state in a partially occupied band. Let  (r) be the extra potential energy
(to be added to the periodic potential (r)) gained by the electron plunged
in the external field. The wave function of the electron obeys Schroedinger
equation:
 b0  +  (r)
~ = (163)

where
2
b0 = − ~ ∇2 + (r)
 (164)
2
and the isolated crystal problem
b0 k (r) = (k)k (r)
 (165)
has already been solved. We look for a solution of (163) as a wave packet of
Bloch states with time dependent coefficients:
X
(r ) = k ()k (r) (166)
k
where the sum is over an appropriately limited (with respect to zone bound-
ary) interval of values around a specific mean k. Since the set of allowed k
 is almost continuous in a macroscopic crystal, it is even possible to write:
Z
(r ) = k ()k (r)k (167)
This dynamic description of an accelerated wave packet is substantially a

particle description (within the entire crystal volume). In fact in this way the
c
uncertainty in the instantaneous position of the electron is greater than the

lattice parameter but less than the minimum linear size of the crystal volume
(yet, see below Limitations in the effective mass description). Introducing
(167) in (163) we get:
Z
 b 0 k (r)k +  (r)
~ = k () (168)

Using the eigenvalue equation of b0 and the Wannier theorem, it is possible
to write:
Z

~ = k ()(−∇)k (r)k +  (r) = (169)

Z
= (−∇) k ()k (r)k +  (r) =
= (−∇) +  (r)
i.e.:

~ = [(−∇) +  (r)]  (170)

We can now define the equivalent Hamiltonian operator as:
b  = (−∇) +  (r)
 (171)
It is equivalent to the former Hamiltonian operator:

2
b = − ~ ∇2 +  (r) +  (r)
 (172)
2
Formally (−∇) can be viewed as the kinetic energy operator associated
with a particle immersed in an external field  (r). Operator (−∇) con-
tains both the inertial properties of the particle and the effect of the periodic
potential. The result in (170), together with eqs. (174) and (176), is called
effective mass theorem for a Bloch-waves packet. If we assume that k corre-
spond to the minimum of a partially occupied band (it could be the bottom
 (0) of the conduction band in a semiconductor), in 1 it is possible to
write:
~2 2
()≈ (0) + (173)
2∗
where
~2
∗ = ³ 2 ´ (174)
 ()
2
0
c
So the corresponding equivalent Hamiltonian is:

2
b  = − ~ ∇2 +  (0) +  (r)
 (175)
2∗
If we count energy from  (0), we get:
2
b = − ~ ∇2 +  (r)
 (176)
2∗
This Hamiltonian leads the motion of an electron which in the absence of an
external force
F = −∇ (r) (177)
would be free but with a mass different from the one in vacuum: it is just
the effective mass ∗ , a concept we have already introduced in qualitative
terms. In this way the periodic potential has been included in the equivalent
~2 2
kinetic energy operator − 2 ∗∇ .
If we, e.g., carefully observe the trend of LCAO bands around the top
of the valence band and around the bottom of the conduction band in the
first Brillouin zone and if, in particular, we apply the previous procedure to
the top of the valence band, we will notice that in this case electrons have a
negative effective mass, whose absolute value is generally different from that
of electrons at the bottom of conduction band.
0
-2.5 -1.25 0 1.25 2.5
k
The implied negative kinetic energy is difficult to conceive, unless we

remove this oddness reversing the energy axis, counting energy downwards
from the top of the valence band. This seems not to make any sense for
valence band electrons, but will be useful, and physically sound, for holes.
c
3.7.2 Impurity levels

If a tetravalent semiconductor such as  is doped with substitutional atoms
with five valence electrons such as , the exceeding electron can be bound
near the donating atom in a localized state belonging to an energy level within
the forbidden energy gap just below the conduction band. Due to thermal
agitation this electron may gain enough energy to occupy a new (propagating)
Bloch state in the conduction band and potentially become a charge carrier
under the effect of an external field; in this case the initial localized state
has been thermally ionized. In a situation like this one the semiconductor
is said to be doped  and the arsenic atom is called a donor. The energy
levels of this impurity can be described as follows. The exceeding electron
coming from atom , in a  crystal, is plunged in an effective electric
field generated by the positive  ion acquiring an additional Coulomb-like
potential energy:
2
 (r) = − (178)
4
where  is the distance between the electron and the ion and  is the electric
permittivity of . The wave function of the exceeding electron obeys the
effective Schroedinger equation:
µ ¶
~2 2 2
− ∗∇ −  =  (179)
2 4
The energy eigenvalues are the hydrogen-like levels:
∗ 4
 =  = − (180)
82 2 2
These levels lie just below the conduction band. In fact in  ∗ = 10
and  = 10 0 and the ionization energy will be 1 ≈ 10−3 rydberg (1 rydberg
4
= 8
2 2 ≈ 136 eV) which is an amount of energy lower than   at room
0
temperature (25 × 10−3 eV). This means that  doping is an effective way
to increase the electron density in the conduction band of . Since the Bohr
2
radius in  ∗ =  ∗
∗ 2 is a factor  ( 0 ) = 100 greater than the radius
of hydrogen atom, the effective mass theorem can be used in this situation
(the radius is greater than the lattice constant).
3.7.3 Semiclassical dynamics in a crystal

If  (r) depends weekly on position, it is possible that the mean de Broglie
wavelength  = 2 of electrons, conceived as Bloch waves packets, is short
enough not to produce strong electron-wave diffraction. In this case we cope
c
with a semiclassical situation and it isn’t necessary to solve equation (170),

but we can instead start with an equivalent classical Hamiltonian. Here we
follow the most straightforward path to obtain the result.
Let us begin with the original classical Hamiltonian:
|p|2
 = +  (r) +  (r) (181)
2
using the Jordan quantization rules (first quantization)
r→ r (182)
p→ − ~∇ (183)
we obtain the Hamiltonian operator

2
b = − ~ ∇2 +  (r) +  (r)
 (184)
2
Applying the Wannier theorem to the band structure
 = (k) (185)
and using the substitution

k → − ∇ (186)
we get the equivalent Hamiltonian operator:
b  = (−∇) +  (r)
 (187)
In a semiclassical situation it is possible to build the equivalent classical

Hamiltonian by reversing the quantization rules (quantization dismounting)
r→ r (188)
p
−∇ → (189)
~
and so obtaining ³p´
 =  +  (r) (190)
~
Now, using Hamilton’s equations
r 
= (191)
 p
p 
= −
 r
c
we get
¡ ¢
r  p~
= (192)
 p
p  (r)
= −
 r
i.e.:
r 1 (k)
= (193)
 ~ k
k  (r)
~ = − (194)
 r
Differentiating the first equation with respect to time and substituting in it
k

from the second equation we obtain the Newton’s law:
µ ¶
2 r 1  2 (k)  (r)
= 2 − (195)
2 ~ kk r
This defines the tensor:
µ ¶
1 1  2 (k)
∗
= (196)
 (k)  ~2  
In 1 the effective mass is thus defined as:
~2
∗ () =  2 ()
(197)
2
and it is a function of 
v,m 2
0
-2.5 -1.25 0 1.25 2.5
k
-1
-2
Speed and effective mass vs. 

c
Analyzing the figure we can understand some interesting applications. Imag-

ine that an electron in a LCAO conduction band at a time  = 0 has  = 0.
The electron is then accelerated by a constant electric field E = −u
(  0). The electron feels the force F = −   ()
u = u . Integrat-
ing the (194) we get:
1
 =  (198)
}
As we can see the wave vector grows linearly with . In the reduced zone
scheme, when  = , a Bragg reflection happens; instantaneously the
0
wave vector becomes  =  −  =  − 2 = − and then it starts
growing linearly again till the new reflection at the edge of the Brillouin zone
occurs. In the figure are shown in black the trend of the mean velocity of
the electron and his effective mass as functions of  (red curves refer to the
valence band). At first ( = 0 and  = 0) the mass is constant and positive
and the velocity grows linearly. At  = (2) the mass becomes infinite,
initially positive then negative, and in a while it stabilizes on a negative value;
the velocity, instead, has already gained its maximum so it decreases until
its value becomes zero at the edge of the Brillouin zone. These anomalies
in the Bloch waves packet show that this description of the electron motion
isn’t completely classical. After the Bragg reflection everything starts all
over again: thus a constant force would produce a periodic motion! In the
real situation (see Transport properties) due to scattering off phonons and
lattice imperfections the electron wave vector has always limited to small
values close to  = 0 being cutt off by scattering processes and the periodic
motion can be hardly observed experimentally.
3.7.4 Limitations in the effective mass description

Under too either intense or fast-varying fields, interband transitions become
possible (tunneling under the energy barrier of hight  ) and the concept
of effective mass cannot be applied any more, at least in clear rigorous terms
(see Optical properties). It is possible to define a upper circular frequency
limit as:

≤ (199)
~
and a upper limit on the applied electric field as:

|E| ≤ (200)

where  is the lattice parameter. Actually these conditions are neither restric-
tive enough nor unique. In a device the particle description (wave packet)
c
needs the charge carrier localization ∆ to be appropriate with respect to

the overall linear dimension: ∆ should be should be small enough so that
the charge carriers could be considered as particles with respect to the size of
the device itself, but, at the same time, big enough to have a quick reaction
∆ to external stimuli. The uncertainty ∆ of wave vector of a Bloch waves
packet corresponds to an energy uncertainty of thermal origin:
~2
∆ = (∆)2 =   (201)
2∗
Using the Heisenberg’s uncertainty principle
~∆∆ ≥  (202)
∆∆ ≥  (203)
we get

∆ ≥ (204)
 

∆ ≥ √ ∗ (205)
2  
At 300  with  = 1  (199) gives us a bandwidth of 2.4 x 1014 Hz while

(204) gives 6.3 x 1012 Hz. At 300  (205) gives
r

∆ ≥ 76  (206)
∗
In a  device with an effective mass of 0.067  the smallest dimension
should be about 29.4 .
3.7.5 Electric current - Electrons and holes

We want now to deepen former considerations considering a semiconductor
with a completely occupied valence band and a completely empty conduction
band.
c
If at  = 0 we apply force F, the response of each electron in the valence band

can be treated as in the previous chapter. The trend of dynamic variables is
shown in the first column of the following picture.
c
At any instant in time the sum of the wave vectors is

X
 = 0 (207)

and the total electric current is proportional to  (which we are going to call
current for the sake of simplicity):
X X 1 k
= (−) h i = (−) =0 (208)
 
~ k
We have so obtained the crucial result that a completely occupied band

cannot conduct current. Let us assume now that, initially, a single electron
occupies the conduction band, thus leaving an empty state in the valence
band.
c
At instant 1 we have, in the conduction band,

X
 = 5  0 (209)

and the current is

X X µ ¶
1 k 1 k
 = (−) h i = (−) = (−) h5 i = (−) 6= 0
 
~ k ~ k 5
(210)
While, in the valence band,
X
 = −5 = 3  0 (211)

with a current equal to

X X µ ¶
1 k 1 k
 = (−) h i = (−) = (−) h3 i = (−) = 
 
~ k ~ k 3
(212)
c
Therefore the total current is  =  +  . Considering a subsequent instant

 = 2 it is possible to notice that the sum of wave vectors associated to all
occupied states in the valence band decreases (it increases as the electric field
increases) and that the total energy increases. The energy of the empty state
decreases while the wave vector and the energy associated to the electron in
the conduction band increase.
The set of electrons in the valence band behaves as a single positive parti-
cle (with a wave vector opposite to the one of the empty state) whose energy
increases, if we reverse the energy axes. This particle has a positive effective
mass, at least in correspondence with small wave vectors, while electrons
would have a negative effective mass. The picture below represents the two
possible representations: the electron representation is shown in the left col-
umn while in the right one the electron-hole representation is depicted.
Basically this is the concept of hole in a dynamic band representation. Since

it is a positive particle, the hole could also be considered as spatially local-
ized describing it as an appropriate Bloch waves packet in the electron-hole
representation. The electron-hole representation is a quite general concept
that can be applied to all systems of Fermions in the framework of second
quantization (see e.g. R.D. Mattuck, A Guide to Feynman Diagrams in the
Many-Body Problem, 2-nd ed. Dover 1992).
In  doped with trivalent impurities (acceptors), such as , in order
c
to complete tetrahedral bonding the substitutional atom captures a valence

electron from , thus becoming a negative ion. Thereby in valence band
remains an empty state that could be described by a hole. This hole is
attracted by the impurity. As long as it is in a localized state, the hole could
be described as stated in section Impurity levels and its energy levels are
situated in the gap just above the valence band. If then the hole is thermally
ionized and submitted to an external electric field, it can carry charge and
contribute to the total current, as already shown for electrons. In this case
we have a p-type semiconductor.
3.7.6 Excitons
In an intrinsic semiconductor each electron-hole pairs could also be generated
by the absorption of a photon with energy greater than  (see Optical
properties). The couple is in a bound metastable state whose energy levels
can be calculated using the same method presented in section Impurity levels
introducing the reduced mass of the system, a sort of hydrogen atom where
the nucleus is the hole (Wannier exciton). In optical absorption experiments
peaks related to transitions between exciton levels can be measured in the
vicinity of the absorption edge.
4 Adiabatic theorem and vibrational motions

Modelling a crystal as a set of independent valence electrons moving in a
mean perfectly periodic potential generated by the interaction with a rigid
lattice of ions does not explain many important physical facts. Among these:
the temperature dependence of specific heat, thermal expansion, melting
(thermodynamic properties), the temperature dependence of electrical re-
sistivity of conductors and of thermal conductivity of insulators, low tem-
perature superconductivity (transport properties), optic absorption of ionic
crystals in the infrared, Brillouin and Raman scattering of light (optic proper-
ties). In this section we discuss an indispensable approximation for a feasible
theory of all the above phenomena: the Born-Oppenheimer approximation
or adiabatic theorem (principle). Before starting the treatment, we warn the
reader that this approximation does not explain other important physical
facts, like superconductivity, which require an "ad hoc" approach. In the
previous sections we assumed that each independent valence electron in a
solid moved in a mean periodic field caused by the lattice of ions and the
other electrons (the symmetry of this field is the same as that of the space
group of the considered crystal). Now we want to consider both the thermal
c
motion of ions about the official atomic points (at a given temperature) and
the motion of valence electrons in the crystal volume; the instantaneous po-
sition of a generic ion with charge  is the sum of a vector of the Bravais
lattice l and a displacement ul () (for the sake of simplicity, for the moment
we consider a simple crystal, a crystal with a basis will be considered later).
The dynamic variables of this quantum many-body problem are the displace-
ments ul and the instantaneous positions r of all valence electrons. Thus
the crystal Hamiltonian is:
b = b + b +  (r) +  (r u) +  (u)

H (213)
where r is the set of all electronic coordinates and u is the set of all ionic
coordinates.
~2 X 2
b = − ∇r (214)
2 
is the total kinetic energy operator of electrons;
~2 X 2
b = − ∇ul (215)
2 l
is the total kinetic energy operator of ions;
1X 2
 (r) = (216)
2 6= 40 |r − r |
is the repulsive Coulomb potential energy of electrons;

XX 2
 (r u) = − (217)
 l
40 |r − l − ul |
is the attractive Coulomb potential energy of ions and electrons;
1X  2 2
 (u) = (218)
2 lh6=l 40 |l + ul − h − uh |
is the repulsive Coulomb potential energy of ions. First step. Since the
electron mass is much smaller than the mass of an ion (if we consider a
single proton, we have:  = 1836  ) as a first approximation we can
assume that the electrons wavefunction  (r|u) has only a parametric de-
pendence on displacements u. In particular, considering stationary states
()
 (r|u) exp(− }), the  (r|u) can be thought as the eigenfunctions
of the reduced Hamiltonian (in which the term b has been neglected) :
c
b0 = b +  (r) +  (r u)

H (219)
b0  (r|u) = () (u) (r|u)
H (220)
()
Electron energy levels  (u) have a parametric dependence on the instanta-
neous positions of ions l + u. The wavefunctions  (r|u) are many-electron
functions (states). However it is possible to solve problem (220) using a
mean field approximation (for example Hartree’s approximation) and obtain
a set of single electron states (independent particles or quasi-particles, see
above). Setting u = 0 in the single-electron mean-field Schrödinger equation,
the total mean potential energy h (r)i +  (r 0) exhibit the translational
invariance of the Bravais lattive and it is possible to use Bloch’s theorem to
calculate monoelectronic states (band structure). As a second step, we look
now for eigenfunctions of the full Hamiltonian H b
b=H
H b0 + b +  (u) (221)
b
HΨ(r u) = Ψ(r u) (222)
where we use the superposition principles of states as:
X
Ψ(r u) = () (u) (r|u) (223)

Multiplying the eigenvalue equation (222) by ∗ (r|u) and integrating over
the electronic coordinates r ( (r|u) form a set of orthonormal functions),
we get: h i
b + () (u) +  (u) () () ()
 (u) =  (u) (224)
where we have neglected several non-adiabatic terms which are listed below.
In this approximation for each electronic stationary state  (r|u) we find
a spectrum of vibrational states () (u) corresponding to vibrational energy
() ()
levels  . In equation (224) the approximate total electronic energy  (u)
has the role of that part of potential energy which depends on ions motion.
In order to obtain (224) we neglected the following terms:
~2 X ³ ´ µZ ¶
() ∗
− 2∇u  (u) ·  (r|u)∇u  (r|u)r + (225)
2 
µZ ¶
~2 X () ∗ 2
−  (u)  (r|u)∇u  (r|u)r
2 
c
The matrix indexed by  and 

Z
∗ (r|u)∇u  (r|u)r (226)
exhibits diagonal elements equal to zero, since

Z Z
 (r|u)∇u  (r|u)r = ∇u ∗ (r|u) (r|u)r
∗
(227)
is the gradient of the (constant) total number of electrons with respect to

ion displacements. Diagonal elements of the other matrix:
Z
−~2
∗ (r|u)∇2u  (r|u)r (228)
2
are multiples of the total kinetic energy of electrons by a factor  and,
thus, negligible. In fact in the worst case, i.e. in case of very strong electron-
ion interaction, it is possible to write  (r|u) =  (r − u) and diagonal
elements become:
Z Z
−~2 2  −~2 2
∗ (r − u) ∇u  (r − u)r = ∗ (r − u)
∇  (r − u)r
2  2 r 
(229)
Generally, instead, all non-diagonal elements are different from zero thus
possibly leading to transitions between electronic states while ions are moving
(electron-phonon interaction).
5 Lattice dynamics
5.1 Lattice specific heat: Einstein model
In solid insulators and at high temperatures the Dulong and Petit’s law is
valid. This law states that the molar specific heat  is constant and equal
to 3 where  =   ' 831  −1  −1 is the ideal gas constant, 
is the Avogadro constant and  the Boltzmann constant. In contrast with
what classical statistical mechanics states (equipartition theorem), experi-
mentally  tends to decrease towards zero at low temperatures. Einstein
gave a first explanation of this behaviour developing a quantum oversimpli-
fied model of vibrational motion of nuclei in crystals. In this model Einstein
assumes that every nucleus performs harmonic oscillations about its equi-
librium position; all atoms oscillate with the same frequency  . In light
c
of modern quantum lattice dynamics, this assumption provides satisfactory

results only at high temperatures, when the motions of nuclei can be con-
sidered as independent. However Einstein model explains in a qualitative
way the dependence of  on temperature  when  → 0. Each nucleus is
considered a quantum harmonic oscillator with oscillation frequency equal to
 ; the spectrum of energy levels is then:
µ ¶
1
 =  + ~  (230)
2
Neglecting the zero-point energy, with a simple derivation we can obtain the
mean energy associated to each oscillator:
~ 
  = ~  =   ~ (231)
 −1 
it is the energy associated to a single quantum of vibrational energy ~

multiplied by the oscillator occupation number given by the Bose-Einstein
distribution.
1
  = ~  (232)
 −1 
Now it is possible to conclude that the internal energy of the crystal (we are
considering one mole) is:
 = 3    (233)
Thus  is equal to:
µ ¶ ~ 
 3 (~  )2   
 = = µ ~ ¶2 (234)
    2 
  
−1
showing a temperature trend qualitatively similar to the experimentally ob-

served one:
lim  ( ) = 0 (235)
 →0
lim  ( ) = 3 (236)
 →∞
Around the Einstein temperature  = ~   the curve exhibits a very

sudden exponential change in the slope. Examining more carefully the exper-
imental results we can notice that, when    ,  tends to 0 more slowly
than as it is stated by the exponential decay in the Einstein model. The low
c
temperature empirical trend is  ∝  3 . In addition, at high temperatures

the experimental  is not exactly constant but it slowly rises as the temper-
ature increases. This last effect would require the inclusion of anharmonic
corrections (see below) and, being almost negligible, will not be considered
further. In order to improve the agreement between theory and experiment
at low temperatures, it is necessary to leave behind the Einstein model and
to develop a more realistic model, able to better describe the true lattice
dynamics. In particular at low temperatures motions of two near nuclei are
generally correlated and this lowers the intensity of internal forces within the
couple.
5.2 Lattice dynamics: Born - Von Karman model

On the basis of the adiabatic principle a full quantum treatment of lattice
dynamics could be built. An alternative approach, phenomenological lat-
tice dynamics, can be started with an initially classical description. The
reason for the "quasi-classical" vibrational behaviour of nuclei in many cir-
cumstances is related to their very short thermal De Broglie wavelength. Let
us consider a particle of mass  accomplishing an oscillatory motion. The
quantum mechanical version of virial theorem states that: 2 h i =  hi
where  is the kinetic energy and  a potential energy of the type  ∝  .
Thus for a harmonic oscillator h i = h i like in classical mechanics and
h2 i
h + i = 2 h i. Then h i can be estimated as 2 = 32   and an average
De Broglie wavelenth of thermal origin as  = √ 2 = √3 

. Inserting
h i
the mass of the proton at room temperature one finds  ≈ 145 ̊ which is
shorter than the typical lattice spacing. This is even more true for heavier
nuclei. According to adiabatic approximation we assume the existence of a
vibrational potential energy  which is a function of displacements of nuclei
with respect to their official positions in the crystal when considered with no
thermal motions (from now on we will use the Born and Huang notation)
µ ¶

r = Rn + r = (237)
n
= 1 a1 + 2 a2 + 3 a3 + r
µ ¶

r is the position, with respect to the origin 0, of nucleus  with
n
mass  situated in the primitive cell (which contains  atoms) identified
by lattice point Rn . r is the position of nucleus  with respect to vec-
tor Rn (internal origin of primitive cell n). In the crystal, ideally with no
c
thermal motions, each nucleus is located exactly in its official position and
the potential energy  attains its minimum value. We introduce now the
displacements of atomic nuclei:
µ ¶ µ ¶ µ ¶
 0  
u =r −r (238)
n n n
0
r indicates the instantaneous position of that nucleus (  n ), which fluctuates
due to vibrational motions. If thermal motions slightly perturb the periodic
structure of the crystal (in thermodynamic equilibrium this means that crys-
tal temperature is much less than melting temperature) a series expansion
of potential energy is feasible and can be arrested to the second order (the
minimum value is put equal to zero):
1 13 µ ¶ µ ¶ µ ¶
1 XXX
0 0
   
 = Φ 0 0   0 0 (239)
2 0 0 0  n n n n
nn  
where we introduced the interatomic force constants tensor

µ 0 ¶ ∙ µ 0 ¶¸
   
Φ 0 = Φ 0 0 (240)
n n n n
defined as: µ 0 ¶
  2
Φ 0 0 = µ ¶  µ 0 ¶ (241)
n n  
  0 0
n n
0
The double sum nn is to be extended to all  crystal cells for both n and
0 ()
n . Physically  is the energy  (u) +  (u) =  (u) which appears in
eq. (224), including here the general situation of crystal with a basis, and
()
it is related to the electronic state  (u) (for example the ground state)
being a perturbation of it. This approximation is called harmonic approxi-
mation. Higher order terms in the series expansion of potential energy, the
anharmonic terms, should be introduced in order to explain melting, thermal
expansion and thermal conductivity. Harmonic approximation explains es-
sential aspects of lattice vibrations-photon scattering (Brillouin and Raman
scattering), the dependence of specific heat on temperature, infrared optical
properties of ionic crystals, insulators and semiconductors and electrical re-
sistivity in conductors (electron-lattice vibrations scattering). Remembering
that a quantum treatment will be finally necessary, we start developing first
a classical dynamics description. The vibrational kinetic energy  of the
c
crystal can be then written as:

1 ¯ µ ¶¯2
1 XX ¯  ¯¯
 =  ¯¯u̇ (242)
2 n  n ¯
Introducing the Lagrangian  =  −  , the Lagrange equations

  
µ ¶− µ ¶ =0 (243)
  
 u̇ u
n n
give birth to a system of classical (Newtonian)

µ ¶ dynamic equations, a system

of 3 equations in the 3 variables 
n
µ ¶ 1
XX µ 0 ¶ µ 0 ¶
   
 ü =− Φ 0 :u 0 (244)
n 0 0
n n n
n 
where : is a rows by columns product of the square matrix corresponding to

0
tensor Φ = [Φ 0 ] multiplied by vector u =[ 0 ] (  = 1 2 3).
µ ¶ 1
XX µ 0 ¶ µ 0 ¶
   
F =− Φ 0 :u 0 (245)
n 0 0
n n n
n 
it is the total elastic force, due to displacements of all the other nuclei, applied
to nucleus  situated in cell n. Consequently it is clear the operational
physical meaning of
µ 0 ¶ µ 0 ¶
   
̃ 0 0 ≡ −Φ 0 0
n n n n
which is the  component of the force applied on nucleus  in cell  caused by

0 0 0
a unitary displacement along direction  of nucleus  in cell  .The number of
independent components in tensor Φ = [Φ 0 ] depends on the point symmetry
of the crystal. If tensor Π =[Π 0 ] represents a symmetry operation of the
point group of the crystal (for example a rotation of order  around an axis
which passes through a fixed point) then (due to the transformation law of
0
tensors after a change in coordinates given by Π : r = Π : r) it must be
0
Φ = ΠΦΠ−1 = Φ (246)
c
where Π−1 is the inverse of matrix Π (Π is a unitary matrix: det Π = 1,

and Π = Π−1 ). Applying a relation like that to any operation in the group
we obtain some constraints determining the independent components of Φ.
Now, due to the translational symmetry of the crystal it is possible to write
0
n = n + h, where h ≡ Th is a lattice translation (in the same way in this
paragraph we will write Rn as n). So we can write:
µ 0 ¶ µ 0 ¶ µ 0 ¶
    
Φ 0 =Φ =Φ (247)
n n n n+h h
0
Notice that Φ depends only on the difference n − n = h. We can eventually
write the dynamic equations as
µ ¶ 1
XX µ 0 ¶ µ 0 ¶
  
 ü =− Φ :u (248)
n 0
h n+h
h 
Where the sum over h involves all the  lattice translations which join
cell n with the other ones (including h = 0). Applying periodic boundary
conditions, as in the case of electrons, and thus imposing that
µ ¶ µ ¶
 
u =u (249)
n + ( − 1)a n
(  = 1 2 3;  = 1 2 3 total number of cells) the dynamic equations (248)

maintain a translational invariance, although they refer to a finite number of
nuclei. The fundamental solutions (lattice waves) should then satisfy Bloch
theorem in its discrete formulation, id est:
µ ¶ µ ¶
 
u =u q·n (250)
n 0
To every lattice wave at least one wave vector q is associated such that the
previous equation is satisfied. As in the case of electrons (k vectors), the set
of q vectors is that composed by the  vectors contained in the first Brillouin
zone. Bloch theorem helps us to reduce the set of 3 dynamic equations to
a set of 3 equations which describes the vibrations of the  nuclei contained
in cell 0. Dropping subscript 0 and considering explicitly the dependence on
q, the new reduced set of equations becomes
Ã !
1
X X µ 0 ¶ ³ 0 ´
 ü ( q) = − Φ q·h : u   q (251)
0
h
 h
c
We introduce now the dynamic matrix

0 ¶ µ

µ 0 ¶ XΦ
 h
D = √ q·h (252)
q  
  0
h
as the discrete Fourier Transform of interatomic force constant tensor. It is

then convenient to write system (251) as:
1 µ 0 ¶ ³ 0 ´
√ Xp 
 ü ( q) = −  0 D : u  q (253)
0
q

We now look for elementary harmonic solutions as:

1
u ( q) = p (q)e ( q) −(q) (254)

where we introduced the polarization unit vector e ( q) and the scalar
complex amplitude (q). Substituting this type of solution into the dynamic
equations we obtain the homogeneous system of linear algebraic equations
(eigenvalue problem):
13 1 ½ µ 0 ¶ ¾ ³ ´
X X  0
2
 0 −  (q) 0  0  0   q = 0 (255)
0 0
q
 
from which it is clear that, for any q, the squares of the natural frequencies are
the eigenvalues of the dynamic matrix while the polarization unit vectors are
its eigenvectors.The eigenvalues are computed solving the equation (solution
condition of the above system):
½ µ 0 ¶ ¾
 2
det  0 −  (q)  0  0 = 0
q
For any value of vector q (there are  independent q vectors in the first
Brillouin zone, as just remembered) the equation has 3 countable solutions
2 (q) = 2 (q) characterized by a branch index . The dynamic matrix is
self-adjoint: thus we have real eigenvalues  2 (q) and a full set of orthonormal
eigenvectors:
1 13
X X ³ 0
´
∗ ( q )   q  =  0 (256)
 
X ³ 0 ´
∗ ( q )  0   q  =  0   0 (257)

c
Solution functions   =   (q) (only positive roots have physical significance

as vibrational frequencies) are called dispersion relations. Considering vi-
brational motions of  nuclei in a primitive cell we cope with 3 degrees of
freedom. Among them 3 refer to the center of mass of the basis (acoustic
branches), while the other 3 −3 refer to internal motion of the basis (optical
branches).
The term acoustic comes from the fact that, when q → 0, acoustic mo-
tions (if coherently, and not thermally, excited) coincide with the macroscopic
elastic waves characterized by  (q) =  (q|q|)q, where  (q|q|), which
depends only on the direction of the propagation q|q|, is the speed of sound
and  corresponds to a quasi-longitudinal wave (1 = ) or to two different
quasi-transverse waves (2 = 1 , 3 = 2 ). These three speeds of sound
are generally different one another. The term quasi indicates that the po-
larization of these waves is mainly longitudinal or mainly transverse. Along
particular symmetry directions and in specific crystals (the cubic ones, as an
example) these waves are completely longitudinal or completely transverse.
The term optical derives from the fact that the corresponding modes can
have a fluctuating electric dipole determining the optical infrared absorption
in the crystal or can have associated a fluctuation of electrical polarizability
determining, in the crystal, the vibrational Raman scattering of photons (in
the infrared, visible and ultraviolet spectrum).
If the crystal is a Bravais lattice with a basis limited to a single atom per
primitive cell, only acoustic modes can exist. In silicon (face centered cubic
lattice with 2 atoms per primitive cell: diamond structure) in addition to the
3 acoustic branches there are 3 × 2 − 3 = 3 optical branches: one of them is
longitudinal, the other two are transverse. Using again Bloch theorem it is
now possible to express the most general solution of the dynamic equations
as the following linear combination of all lattice waves:
µ ¶ X X13
 1
u =p (q)e ( q ) q·n − (q) (258)
n  q∈BZ 
Introducing the 3 normal coordinates (q) = (q)− (q) ((q)

are the amplitudes of normal coordinates) it is possible to write for the real
displacements:
µ ¶ Ã 13
!
 1 1 X X
u = p (q )e ( q ) q·n + 
n 2  q∈BZ 
Substituting these solutions in the vibrational Hamiltonian

c
1
XX µ ¶
 
 = µ ¶ u̇ − 
 n
n   u̇
n
and, finally, using the properties of polarization unit vectors and the fact
that ions displacements (although they have been written so far as complex
quantities) components are real quantities, we get the most important issue
in the harmonic approximation:
13
1 X X© ª
 = | (q )|2 +  2 (q) |(q )|2 (259)
2 q∈BZ 
where

 (q ) = (260)
 ̇(q )
are the kinetic momenta conjugated to normal coordinates (q ). We have
thus proved that the vibrational behaviour of the crystal is equivalent to
that of a set of 3 independent harmonic oscillators with frequencies  2 ();
there is a one to one correspondence between oscillators and lattice waves
(normal modes) (q ). Collective motions (q ) are harmonic motions,
while the motion of a single nucleus is not. For this reason the (q ) are
called collective coordinates and their energy quanta } () behave as Bose
particles, or quasi-particles (see below); as a matter of fact they are not
associated to individual particles (which would be individually localized in
the limit of uncertainty principle). The introduction of normal coordinates
decouples the system of harmonic dynamic equation describing the motion
of nuclei and produces 3 independent elementary equations:
̈(q ) +  2 (q)(q ) = 0 (261)
which can be obtained from  using the Hamilton equations:


= ̇(q ) (262)
 (q)

− = ̇ (q ) (263)
(q)
c
5.3 Linear chain with two atoms per cell: acoustic

modes and optical modes
Let us consider a 1 crystal with lattice parameter  and a basis with two
ions of mass 1 and 2 (1  2 ) located in the official (frozen) positions
µ ¶
1
 = 

µ ¶ µ ¶
2 1
 = + 
 2
and take only into account the interactions between them and with their first
neighbours. It is not necessary for the second ion to be located in the middle
of the cell. However we made this choice in order to use a unique interatomic
force constant . Considering the interaction between ions inside the cell
 = 0 and with ions in the adjacent cells  = ±1, we find  nonzero force
constants which determine the dynamics of ions in the cell  = 0 (we drop
subscripts 11 of Φ)
µ ¶ µ ¶ µ ¶
11 12 21
Φ = 2; Φ = −; Φ = −;
0 0 0
µ ¶ µ ¶ µ ¶
22 21 12
Φ = 2; Φ = −; Φ = −
0 1 −1
by means of which we can build the dynamic matrix :

µ 0 ¶ Ã 2  −
!
 1
− √
1 2
(1 +  )
 =
 − √1 2 (1 + + ) 2
2
The eigenvalues of this matrix are the solutions of the following equation
(solubility conditions of system(255)):
µ ¶µ ¶ ³  ´
2 2 2 2 42
− − − cos2 =0
1 2 1 2 2
i.e.:
r ³´ ³  ´
 
 212 () = ∓ 1−4 sin2
   2
where
1 2
= and  = 1 + 2

c
are, respectively, the reduced mass and the total mass (mass of the center of
mass) of the basis. When  → 0, we asymptotically have
r

1 ≈ ||
2
r µ ¶
2 2 2
2 ≈ 1− (264)
 8
while near the edges of the first Brillouin zone (where  = ±) we get
r
2
1 =

r 1
2
2 = (265)
2
The first branch of eigenvalues has an acoustic nature (see next section)
while the second one an optical (see Optical properties below) nature. More-
over between  1 () and  2 ()
pthere is a gap of prohibited frequencies.
Frequencies greater than  2 (0) = 2 are not possible.
1.5
1.25
y
1
0.75
0.5
0.25
0
-2.5 -1.25 0 1.25 2.5
x
p
Dispersion relations in the linear biatomic chain (y= 
)
In figure the dispersion relations in the first Brillouin zone (   =  ≤ )

with 1 = 22 are shown; the minimum forbidden frequency interval in
correspondence with the edges of the zone is clearly visible.
c
5.4 Simple 1D crystal: acoustic phonons and elastic

waves
This case is even simpler than the previous one: the linear atomic chain is a
1D simple crystal with only one lattice ion (nucleus) of mass  per primitive
cell of size  (the 1D Bravais lattice positions are  = ). We still consider
only first neighbors interactions of strength . The problem is so simple that
we can solve it directly using Newton’s law for the motion of a generic nucleus
 interacting with nuclei  − 1 and  + 1. In this way we get the system of
 ( = 0 1 2   − 1) coupled equations:
̈ = −2 + −1 + +1 (266)
Since we cope with a periodic lattice (1D crystal), Bloch theorem (in its
discrete form) can be used:
 = 0  (267)
in this way we get a single equation which describes the motion of the nucleus
located in the first cell ( = 0):
µ ¶ ³´ ³´ µ ¶
2 −  2
̈0 = − 0 + 0  + 0  = − (1 − cos())0
   
(268)
We now look for a harmonic solution such as
 −
0 () = √  (269)

and we get the solubility condition:
³´ ³ ´
2 2 
 =4 sin (270)
 2
i.e. r ¯ ³ ´¯
 ¯  ¯
=2 ¯sin ¯ (271)
 2
c
y 0.75
0.5
0.25
0
-2.5 -1.25 0 1.25 2.5
p x
Dispersion relation y= 2
In figure the dispersion relation in the first Brillouin zone (−   =  ≤ )

is shown. In the limit  → 0
r

() ≈  || (272)

while r
 
(± ) = 2 (273)
 
is the crystal cutoff frequency for acoustic modes. When  → 0,  =
(2)   and the lattice wave feels the crystal as a continuous elastic
medium. This directly follows from the dynamic equation rewritten as
³´ µ ¶
 −2 + −1 + +1
̈ = (274)
3  2
which, for long wavelengths, can be well approximated by the wave equation
2 2
 =  (275)
2 2
where we have introduced  = 3 (mass density),  =  (Young’s mod-
ulus) and ( ) (continuous field of elastic displacements). The solutions of
this last equation are elastic propagating waves
( ) = (−) (276)
with speed of sound s


 =  = (277)

c
and linear dispersion relation
() =  || (278)
In the discrete (atomic) case, comparing the limits of the dispersion relations
when  → 0, we notice that r

 =  (279)

In other words in a crystal it is possible to identify the acoustic modes with
long wavelength with the elastic waves in the same crystal considered as a
continuous medium (see below Debye model).
5.5 Phonons
Referring to the situation of the previous case, we now introduce the most
general solutions of dynamic equations as a linear combination of all lattice
waves:
1 X
 () = √ () −() (280)
 ∈BZ
The N discrete allowed wavevectors  are those imposed by the periodic
boundary conditions
−1 = 0 (281)
for we can write:
− +1 
2X 2

√  −  2(  )

 () = (282)


We define:
  =  −  (283)
as the normal coordinates and
r ¯ µ ¶¯
 ¯¯  ¯¯
 = 2 sin (284)
¯  ¯
as the corresponding proper frequencies. We can now write the vibrational

Lagrangian as:
0 −1 0−1
1 X 2 1 X
 = ̇ − (+1 −  )2 (285)
2  2 
c
and the Hamiltonian as

0 −1 0−1
1 X 2 1 X
 = + (+1 −  )2 (286)
2  2 2 
where

 = = ̇ (287)
 ̇
are the conjugated momenta. Substituting (282) in the Lagrangian and in
the Hamiltonian, after some lenghty computation we omit here (imposing
 () to be real), we can finally get:
1
−
2X ½¯¯ ¯¯2
+1 
2
2
¾
2
 = ¯̇  ¯ −   |  | (288)
2 
− +1 
1 2 X © 2
2
ª
 = | | +  2 |  |2 (289)
2 
where

 = (290)
 ̇ 
are the canonical conjugated momenta. As in the general 3D case, the pe-
riodic monoatomic linear chain can be considered as a set of N independent
harmonic oscillators, each oscillator being represented by a normal coordi-
nate. Introducing now the Hamiltonian operator
1
−
2X ½¯¯ ¯¯2
+1 
2 ¯ ¯2 ¾
2¯ ¯
̂ = ¯̂ ¯ +   ¯̂  ¯ (291)
2 
where

̂ = −~ (292)
 
it is immediately possible to switch to a quantum description of lattice vi-
brations: the crystal is, from a vibrational point of view, equivalent to a
system of quantum independent harmonic oscillators. In the 3D general case
each oscillator with normal coordinate (q ) owns a frequency (q ) and
a polarization unit vector e(q ) and can assume the discrete energy levels
µ ¶
1
 (q ) = (q ) + ~(q ) (293)
2
c
where (q ) = 0 1 2 3 The total vibrational energy is thus equal to:

13 µ ¶
X X 1
 = (q ) + ~(q ) (294)
q∈BZ 
2
(q ) represents the quantum number -(phonon population number) asso-

ciated to the mode (normal coordinate) (q ). The set of 3 numbers
(q ) completely determines the vibrational quantum state of the crystal
at zero temperature. In thermal equilibrium at temperature  , the vibra-
tional contribution to the internal energy of the crystal is:
13 µ ¶
X X 1
 = h(q )i + ~(q ) (295)
q∈BZ 
2
where
1
h(q )i = ~ (q)
(296)
  
−1
is the Bose-Einstein distribution function. Thus, within the harmonic ap-
proximation, the vibrational properties of the crystal is described by an ideal
Bose-Einstein gas of phonons. Sometimes the term phonon is used to indi-
cate the collective coordinates (and the normal modes) (q ) and not only
their quantum vibrational energy ~(q ). Also in interactions with elec-
trons and photons (emission, absorption and scattering events), the quantized
lattice vibrations behave in the same way as particles associated to collective
coordinates characterized by a crystal momentum ~q (in the reduced zone
scheme or in the repeated zone scheme) and by an energy ~(q ) (which
is a periodic function of q in the repeated zone scheme); lattice waves can
be distinguished in acoustic and optical phonons. When, in an interaction
process, only phonons with wave vectors situated in the first Brillouin zone
are involved, the process is called N process (normal), otherwise, if some
wavevectors exceed the the first Brillouin zone, we have a U process (umk-
lapp).
5.6 Specific heat: Debye model

As in the case of electronic states, in a macroscopic crystal the  phonons
wave vectors (where  coincides with the total number of primitive cells)
quasi-continuously occupy the first Brillouin zone in reciprocal space. This
suggests to approximate this quasi-continuous filling with a truly continuous
one and to introduce the concept of density of vibrational normal modes (if
c
instead we considered a nanometric crystal, this continuous approximation

could not be used). In general in a crystal of volume  containing  primitive
cells, the elementary volume of reciprocal space associated to a single wave
vector is equal to:
83 8 3 8 3
= = (297)
  |a1 · a2 × a3 | 
as turns out from periodic boundary conditions. Furthermore if we consider a

monoatomic simple crystal, to any wave vector there correspond three normal
acoustic modes (one longitudinal and two transverse). For any polarization
(phonon branch) the above quantity may be defined density of normal modes:
it is constant in reciprocal space but, in general, it is frequency dependent in
frequency space. To get this frequency dependent density, we start with the
quantity  () which measures how many normal modes with a given po-
larization  have frequency within 0 and . Fixing  the dispersion relation
 =   (q) defines a surface in reciprocal space and this surface encloses a
volume  () such that:
 ()
 () = ¡ 83 ¢ (298)

Now the density of modes  as a function of  can be computed as the
derivative of  ():
 ()
 () = (299)

In 1 the constant frequency surface degenerates into two discrete points
and the volume  () into a linear segment () bounded by these points.
For example, in the monoatomic 1 chain treated above, for a given  the
equation r ¯ ³ ´¯
 ¯  ¯
=2 ¯sin ¯ (300)
 2
has the two roots
⎛ ⎞
µ ¶
2 1 
± () = ± arcsin ⎝ q ⎠ (301)
 2 1

which defines a segment of length:

⎛ ⎞
µ ¶
4 1 
() = 2+ () = arcsin ⎝ q ⎠ (302)
 2 1

c
thus we have: ⎛ ⎞
() 2 1 
() = ¡ 2 ¢ = arcsin ⎝ q ⎠ (303)

 2 1

to which there corresponds the 1D density:

⎡ ⎛ ⎞⎤
 ⎣ 2 1  1 
() = arcsin ⎝ q ⎠⎦ = q q (304)
  2 1  1 1− 1 2

  4 
In one dimension the density of acousticp modes diverges in correspondence

of the maximum possible frequency 2  at zone boundary, where the
phonon group velocity  is zero, while it is nearly constant for small
wave vectors, where the group velocity is constant and equal to the speed
of sound. In view of the difficulties of generalization of this procedure when
applied to a complex 3 crystal whose lattice dynamics is governed by real-
istic atomic force constants, Peter Debye introduced an approximate model
for the derivation of () considering the crystal as an isotropic elastic con-
tinuous body. In this way only non dispersive acoustic modes (elastic waves)
do exist. The discrete nature of the real finite crystal is lumped into a maxi-
mum allowed frequency, called the Debye frequency   . In a continuous and
isotropic elastic solid body (see Appendix Elasticity and Elastic Waves) a
dynamic deformation process can generally be described as the superposition
of a simple dilatation (compression) field and by a simple shear strain field.
This corresponds to a total displacement field u
u = u + u (305)
where
∇ × u = 0 (306)
∇ · u = 0 (307)
Displacement fields u and u obey these two decoupled wave equations:
 2 u
= 2 ∇2 u (308)
2
 2 u
= 2 ∇2 u (309)
2
where
s
 + 43 
 = (310)

c
is the longitudinal speed of sound and

r

 = (311)

is the transverse speed of sound. In previous equations  is the mass density,

 is the bulk modulus and  is the shear modulus. These two elastic constants
can be expressed by a function of the couple of technical constants  (Young’s
modulus) and  (Poisson’s ratio) using the following formulae:

 = (312)
3(1 − 2)

 = (313)
2(1 + )
For thermodynamic stability reasons it is always    . In case of a simple

uniaxial deformation in the direction of axis :   =  and  =  =
− .   is the axial stress while  ,  ,  are the three diagonal
components of the strain tensor. Moreover it is possible to write:
   
 +  +  = + + = (314)
   
We call  = − 13 (  +   +   ) the hydrostatic pressure and we notice that

 = − (315)

If we consider, for example, the longitudinal acoustic branch, the dispersion
relation is: q
 =  |q| =  2 + 2 + 2 (316)
and, fixing , the (316) represents in space q the equation of a spherical
surface with radius  . Then it is:
µ ¶3
4 
 () =  (317)
3 
³ ´3
4 
3
   3
 () = ¡ 83 ¢ = (318)

6 2 3
µ ¶
  3  2
 () = = (319)
 6 2 3 22 3
c
i.e. we have a parabolic density of modes. Repeating all this for the two
transverse branches (degenerate in the continuous and isotropic elastic mean)
we get the total density of modes:
 2
() = 2 3 (320)
2 
where
1 1 2
3
= 3+ 3 (321)
  
The Debye frequency is then obtained from condition:
Z 
() = 3 (322)
0
as: ¶ µ

 3
= 18  2 3
(323)

With the previous formula it is possible to write:
92
() = (324)
 3
Now we can generalize the specific heat expression in the Einstein model (see
above). We consider the Einstein formula still valid for the single mode with
frequency  but now we sum up over all the modes using () as a weighting
function:
µ ¶ Z  2
~
 (~)   
 = = µ ~ ¶2 () (325)
  0 2
    
−1
This expression for the specific heat fits rather well most experimental data.
Every solid is characterized by its Debye temperature:
~ 
Θ = (326)

Setting:
~
= (327)
 
we can rewrite the expression of the specific heat as (assuming  =  ):
µ ¶3 Z Θ 
  4 
 = 3  3  (328)
Θ 0 ( − 1)2
c
At low temperatures (Θ  → ∞) the integral tends to a constant and the

specific heat tends to zero as  3 . At high temperatures Θ tends to zero,
thus the integral can be approximated as:
Z Θ  Z Θ  µ ¶3
 4  2 Θ
3  ≈ 3   = (329)
0 ( − 1)2 0 
and the specific heat obeys the classical Dulong and Petit law:
 ≈ 3  = 3 (330)
Thus a crystal behaves in a classical way when   Θ and in a quantum

way if   Θ . For example diamond (one of the two allotropes of carbon
in the solid state) with Θ = 1860  and silicon with Θ = 625  both
exhibit a quantum thermodynamic behaviour at room temperature (300 ).
List of Debye temperatures
The figure below shows the universal trend of Debye specific heat (328)
as a function of the reduced temperature  = Θ :
c
Cv/3R 1
0.75
0.5
0.25
0
0 0.5 1 1.5 2
x
Debye specific heat
The success of the Debye model finds a simple physical justification in the
behaviour of Bose-Einstein factor as a function of frequency at low tempera-
~   (q)
−
tures. As a matter of fact when  → 0  we have that   (q) ≈    :
normal modes with a lower frequency has a higher occupational factor. These
modes are just the acoustic modes about the center of the Brillouin zone. For
the contribution of these very modes the Debye model provides a very good
approximation. Simple crystals have only acoustic modes, while more com-
plicated complex crystals need a better theoretical description including the
optical modes.
5.7 Polaritons
The denomination optical modes originates from the fact that these modes
determine the linear optical properties of polar (ionic) crystals in the infrared.
These crystals are non monoelemental crystals in which the chemical bond
is, at least partially, of ionic character. The extreme case is constituted by
alkali metals halides (for example + − ) in which the interatomic forces
have a purely electrostatic nature and the electron states are localized on
individual ions. However also the cases of III-V compounds, such as 
and II-VI compounds, such as  (zincblende), should be considered. In
these crystals, where the bonds are mixed covalent/ionic, the electrons eigen-
states are Bloch waves and the usual band structure picture holds. All these
structures share the absence of a center of symmetry and, as a consequence,
in every primitive cell they exhibit an electric dipole moment (we have a total
electric dipole of ions forming the crystal basis). A characteristic crystallo-
grafic structure is zincblende: its Bravais lattice is f.c.c. (as in the case of
 −  or Si); the basis is formed by two atoms located in (0,0,0) and
in (1/4,1/4,1/4), all along the principal diagonal of the conventional cubic
c
cell, in units of  (cubic lattice parameter). We now imagine that in (0,0,0)

the positive ion is situated, while in (1/4,1/4,1/4) the negative one. Consid-
ering the propagation of electromagnetic waves along direction (111), we can
approximate the lattice dynamic as the one of a linear chain with two atoms
per cell (see above), as long as the branch  2 () of the dispersion relations
is concerned. We study the forced response of the chain under the action of
the (local) electric field of an electromagnetic wave
E = E0 (−) (331)
which propagates in the same direction (111). The corresponding 3 prob-
lem is very difficult to be analyzed since we should consider both the lon-
gitudinal and transverse polarizations, for both the field of the wave in the
crystal (within matter also longitudinal electromagnetic waves may exist) and
for the field of atomic displacements. Using a simplified model, regardless of
the polarization we can write directly the equations of the forced motion of
ions as ( is the absolute value of an ion charge)
µ ¶ µ ¶ µ ¶ µ ¶
1 1 2 2
1 ̈ = −2 +  +  +
   −1
     
((
1 +
1 )−)
+0    (332)
µ ¶ µ ¶ µ ¶ µ ¶
2 2 1 1
2 ̈ = −2 +  +  +
  +1 
     
((
2 +
2 )−)
−0    (333)
Assuming that the wavelength  = 2 is much greater than , we apply
Bloch theorem to previous equations (as already done for the chain with
only one atom per cell) in the hypothesis that  =  (this could be justified
by quantum theory). In this way it is possible to considerably simplify the
equations of motion:
2 0 −
̈1 = − (1 − 2 ) + 
1 1
2 0 −
̈2 = + (1 − 2 ) −  (334)
2 2
Subtracting the second equation from the first, and introducing the reduced
mass , we get:
2 2 0 −
2
(1 − 2 ) + (1 − 2 ) =  (335)
  
c
Introducing now the normal coordinate  = 1 − 2 and the characteristic

frequency r
2
 = =  2 ( = 0) (336)

we obtain:
2  0 −
2
+  2  =  (337)
 
The forced solution is:

= 0 − (338)
2 − 2
Thus the associated electric dipole moment is:
2 
 =  = 0 − (339)
2 −  2
and the polarizability of the single cell (of the basis)  =  (0 ) may be
written as:
1 2 
= (340)
0  2 − 2
Neglecting local-field effects, if there are  cells per unit volume, the dielectric
susceptibility of the crystal can be written as:
1 2 
 =  = (341)
0  2 −  2
Consequently the dielectric constant becomes:

µ ¶
2
 = 0 (1 + ) = 0 1 + 2 (342)
 − 2
where we introduced the plasma frequency of lattice ions:

s
2
 = (343)
0 
q
Note that when the frequency is equal to   = 2 + 2 we have () = 0.
Moreover it is (Lyddane-Sachs-Teller relation):
 2 (0)
2
= (344)
 (∞)
c
and (Born equation): µ ¶

 2
(0) = (∞) 1 + 2 (345)

epsilon rel.
3.75
2.5
1.25
0
0 0.5 1 1.5 2 2.5
omega/omegaT
-1.25
Relative dielectric constant of GaAs
As shown in the picture (black curve), in the frequency interval ∆ between

 and   the dielectric constant is negative. Since the speed of propagation
of the electromagnetic wave is:
r
 0
= ≈ (346)
 
where  is the refractive index, we have that in the interval ∆ the elec-
tromagnetic wave cannot propagate in the crystal. An e.m. wave like that
would be totally reflected from the surface of the crystal (Restrahlen effect).
Thus, independently from periodicity effects, in a ionic crystal there is an
energy gap prohibited to photons ∆ = ~∆. If in the model we include
dissipative interactions, the trend of the dielectric function (the real part)
is the one without divergences shown in the picture (red curve: absorption
and dispersion in infrared). The second figure shows both the real and the
imaginary part of the refractive index, including dissipation (cfr eq. 347).
c
n 2.5
1.5
0.5
0
0 0.5 1 1.5 2
omega/omrgaT
Complex index of refraction (GaAs)

In order to describe the absorption (attenuation of the electromagnetic wave)
we introduce a complex wave vector. Thus the dispersion relation of electro-
magnetic waves in the crystal can by described as:
r
 0 
= =  = r³ ´ (347)
   2
1 + 2 −2

Solving with respect to  we get the complex wave vector:

sµ ¶
 2
= 1+ 2 (348)
  − 2
ck 5
3.75
2.5
1.25
0
0 0.5 1 1.5 2
omega/omegaT
Polariton dispersion relations

In the last figure (relative to  not including dissipation) the real part of
the product of the wave vector and the speed of light in vacuum  as a function
of the frequency is shown in black. The imaginary part, corresponding to
the attenuation of the electromagnetic wave, is shown in red. The group
c
velocity  can never exceed . The asymptotic phononic and photonic
trends are evident. The mixed photon-optical phonons excitations introduced
in this section are quantized (their energies are ~()) and they are called
polaritons. In a more realistic 3 model, in which transverse optical phonons
 are distinguished from longitudinal optical phonons  (and transverse e.m.
wave are distinguished from longitudinal ones) we have   =   (0) and   =
 (0). Moreover the e.m. transverse waves couple exclusively with transverse
optical phonons and they can propagate outside the interval ∆. When  =
 mixed longitudinal optical phonon/longitudinal e.m. wave, stationary
modes exist. In the previous picture the relative dispersion relation would
be represented as a vertical line. Generally   and   lie in the infrared (in
   = 51 × 1013 s−1 and   = 55 × 1013 s−1 ). In a ionic crystal a
moving electron carries with itself a cloud of optical phonons which changes
its mass (the mass increases): this quasi-particle is called polaron. Also
in non-polar crystals, such as silicon, optical modes interact with photons,
but this interaction causes nonlinear optical properties (Raman scattering)
and the effect is evident also in the visible and in the ultraviolet. From a
quantum point of view the description of scattering processes needs a second
order correction in dynamic perturbation theory (see Appendix: Quantum
approximate methods).
5.8 Inelastic scattering

Let us assume that the interaction potential in a crystal can be described as
a sum of atomic potentials  (for the sake of simplicity consider a simple
crystal in which the atoms occupy the nodes of the Bravais lattice):
X
(r) =  (r − rn )
n
As we already learned in the theory of scattering, the scattering amplitude

is given by the following expression:
Z Z X
−Q·r
(Q) ∼ (r) r ∼ −Q·r  (r − rn )r =
n
Z X
= −Q·(r−rn ) −Q·rn  (r − rn )r =
n
X Z
−Q·rn
=  −Q·(r−rn )  (r − rn )r ∼ (349)
n
X
∼  (Q) −Q·rn
n
c
P
Where n −Q·rn is the structure factor of the whole crystal, and the atomic
shape factor is: Z
(Q) = −Q·R  (R)r

When atoms occupy equilibrium positions (there are no lattice vibrations)
rn = n =1 a1 + 2 a2 + 3 a3
it is X
−Q·n = (Q − g)
n
where g is a vector of the reciprocal lattice, and Bragg law for elastic diffrac-
tion is recovered. In fact for any vector m in the lattice:
X X X
−Q·n = −Q·(n+m) = −Q·m −Q·n (350)
n n n
P
i.e. is either n −Q·n = 0 or, if Q = g (a vector of the reciprocal lattice),
P it−Q·n
n = .
Otherwise, if atoms vibrate about their equilibrium positions, we can
write their displacements as a superposition of all normal modes summing
over index q. For the sake of simplicity we drop branch index ; moreover
there is not any  index since we are considering just one atom per cell:
X
rn = n+ (un (q) + u∗n (q)) (351)
q
where we added the complex conjugate in order to have real displacements.

The sum is confined in the first Brillouin zone and the sum over  is implied.
Using Bloch theorem we again write
un (q) = u0 (q)q·n
and
u0 (q) =  (q)e(q)−(q)
where e(q) is the polarization vector of the normal mode q, and  (q) √ is the
amplitude of the corresponding normal coordinate, equal to (q)  in
the notation used in the description of lattice vibrations. Thus considering
equations (351) and (349) we have:
X X  ∗
(Q) ∼ −Q·rn = −Q·(n+ q (un (q)+un (q))) =
n n
X Y
−Q·n −Q·(un (q)+u∗n (q))
=  
n q
c
Considering only small lattice displacements

∗
−Q·(un (q)+un (q)) ≈ 1 − Q · (un (q) + u∗n (q)) − |Q · u0 (q)|2 +  (352)
and just the linear terms, i.e.:

Y ∗ X
−Q·(un (q)+un (q)) ≈ 1 − Q · (un (q) + u∗n (q)) (353)
q q
we get:
Ã !
X X
(Q) ∼  −Q·n
1− Q · (un (q) + u∗n (q)) =
n q
X X X
= −Q·n − −Q·n Q · (u0 (q)q·n +u∗0 (q)−q·n ) =
n n q
X X
= (Q − g) − (Q · e(q)) (q)−(q) −Q·(n−q) +
q n
X X
(q) −Q·(n+q)
− (Q · e(q)) (q) 
q n
where normal coordinates have real amplitudes. We consider now the Bravais
lattice inversion symmetry, in this case the lattice sum over q is equal to the
one over −q, and we have:
X
(Q) ∼ (Q − g) −  (Q · e(q)) (q)−(q) (Q − q − g)+
q
X
− (Q · e(q)) (q)(q) (Q − q − g) (354)
q
We now apply the Fermi’s golden rule (where  is the transition proba-
bility per unit time), id est (Q) ∼  = 2 ~
|h1|0 |2i|2 (2 − 1 ± ~),
in the form valid for interaction with a time harmonic potential 0 ± (in
this case we have a term ±~ in a Dirac delta function which represents the
conservation of energy). We consider as initial state |1i ∼ k ·r , and as final
state h2| ∼ k ·r (initially free particles with a defined wavevector interacting
with a crystal). InR this case the matrix element h1|0 |2i is equal to the scat-
tering amplitude 0 (r)−Q·r r , where Q = k − k . Thus the first term
in equation (354) gives us the elastic contribution in the expression of the
intensity and is proportional to (Q − g)(2 − 1 ) | (Q)|2 . In this way we
find the Bragg law, with the condition that scattered particles and incident
particles share the same energy. The second and the third term give us an
inelastic contribution in the expression of intensity, corresponding to a first
c
P
order inelastic scattering, and proportional to | (Q)|2 q |Q · e(q)|2  2 (q)
(Q − q − g)(2 − 1 ± ~(q)). This means that there is an active energy
exchange mechanism between the particle (a neutron, for example) or be-
tween a photon of the electromagnetic field (light or X rays, for example)
and the quanta of vibrational energy. A vibrational quantum with energy
~(q) can be created or destructed. Factor (Q − q − g) imposes the rela-
tion:
Q = k − k = q + g (355)
If the equality is satisfied when g = 0, id est Q is in the first Brillouin zone
of the reciprocal lattice, we have a Normal or N process, otherwise, if Q is
outside the first Brillouin zone, id est equation (355) is satisfied when g 6= 0,
we have a Umklapp or U process. In the first situation equation (355) can be
seen as a principle of conservation of quasi-momentum (exchange between the
crystalline momentum ~q of a phonon and the momentum of the particle):
~Q = ~k − ~k = ~q
In the second situation the additional term ~g corresponds to a term of vari-

ation of the momentum associated to the motion of the center of mass of
the entire crystal. The Umklapp process can be considered as the creation
(or destruction) of a phonon associated to a Bragg reflection. In a quantum
description of lattice vibrations, the mean value (it is a thermal mean) of
h 2 (q)i depends on the population of the vibrational mode with wave vec-
tor q, described by the Bose-Einstein distribution. Actually, if we considered
the quadratic terms in the expansion (352), both the elastic term and the
inelastic one in the expression of intensity should be multiplied by a factor
of: Y  2
(1 − |Q · u0 (q)|2 ) = − q |Q·u0 (q)| = −2
q
where −2 is the Debye-Waller factor, where

1X 1X
 = |Q · u0 (q)|2 = |Q · e(q)|2  2 (q) =
2 q 2 q
1X 2 (q)
= |Q · e(q)|2
2 q 
√
In the last expression we wrote the normal coordinate (q)  with
the notation used in the description of lattice vibrations ((q) shouldn’t be
confused with the absolute value of Q = k − k , which is a scattering wave
vector). When lattice vibrations are considered, the intensity of Bragg peaks
c
(and the one of inelastic peaks) is thus decreased due to a disorder which
introduces a "noise" in the crystal translational symmetry. In a D quantum E
2 (q)
description of lattice vibration, the mean value (thermal mean) of 
¡ ¢ ~ 
is equal to (q) + 12  , where (q) is the occupation number for bosons
in the quantum harmonic oscillator corresponding to the mode q:
1
(q) =
exp( ~(q)

) −1
The Debye-Waller factor depends on the temperature, and in particular 
is proportional to  at high temperatures (above the Debye temperature),
while it assumes a constant value, still different from zero (due to the vibra-
tional zero-point energy), at low temperatures. If we decide to consider also
the nonlinear terms, which contain products of more than one term such as
Q · (un (q) + u∗n (q)), in the expression of intensity we would have also terms
corresponding to second order scattering (with a smaller intensity with re-
spect to the first order), id est corresponding to an exchange of energy and
momentum with two or more phonons (multiple phonons processes).
Space-time autocorrelation function:
In a "static" crystal we have that:
Z
(Q) ∼ (Q) ∼  (R)−Q·R R
which is the static structure factor, is equal to the spatial Fourier transform
of the static autocorrelation function:
Z
 (R)= (r + R) ∗ (r)r
In a crystal or in a mean with vibrational motions we have that:

Z
(Q ) ∼ (Q ω) ∼  (R)−Q·R  R
which is the dynamic structure factor of the crystal, it is equal to the space-
time Fourier transform of the dynamic autocorrelation function (where the
time integration is done on time intervals larger than characteristic times in
vibrational motions, the integral corresponds to a thermal average operation):
Z Z ¿Z À
∗ ∗
 (R) ∼ (r + R + ) (r )r ∼ (r + R) (r 0)r

c
6 Optical properties
Generally the linear optical properties in solids (dispersion, absorption, re-
flection, refraction) depend on macroscopic interaction processes between
photons, electrons and phonons. In the figure below the qualitative trend
of the absorption coefficient (358) in a doped polar semiconductor is shown
where it is possible to see all possible microscopic mechanisms.
Absorption coefficient
In this section we will analyze explicitly just the principal band and the
continuous background (in the previous section we introduced the polari-
tonic absorption), however the principles already treated would give us the
possibility of understanding from a quantum point of view all the peaks in
the figure. To start we build a general connection between the macroscopic
optical parameters (refractive index and extinction coefficient) and the tran-
sition probabilities between different quantum states in a solid excited by an
electromagnetic radiation. The electric field of a monochromatic plane wave,
with polarization unit vector e, which propagates in a solid along the  axis
is described, in the typical complex form, by expression:

E = 0 e(−) (356)
n o
(
−)
Thus the effective electric field will be e Re 0  . Macroscopically
the dispersion and absorption (extinction) processes can be described intro-
ducing a complex wave vector e  =  +  and we can write:
E = 0 e− (−) (357)

c
Id est, we have a wave attenuation. The intensity  of the wave (averaged

³ n o´2
over a period), proportional to Re 0 (−) = 12 |E |2 , exhibits an
exponential decay in space:
 = 0 −2 = 0 −Σ (358)
Introducing the dispersion relation

e
=  (359)
e

with complex refractive index:
0 00
e =  + 
 (360)
we have ³ ´
Σ=2 ” (361)

0 ”
We
¡ introduce now the complex relative dielectric constant e =  +√ =
0¢
1 +  + ” ( is the complex dielectric susceptibility). Since 
e ≈ e , it
is:
0 0
³ 0 ´2 ³ 00 ´2
 = 1 +  =  −  (362)
”
³ 0
´ ³ 00
´
 = ” = 2  
Using the second equation in (362) we can thus write:

³  ´ ”
Σ=
 0
But, since the imaginary part in the refractive index is always less than the
real part, using the first equation in(362) we get:
³ ´ ” ³ ´
Σ= p ' ”
 1 + 0 
the approximation being valid especially about the absorption frequencies,

0
where  ' 0. Now we want to underline the connection between the extinc-
tion coefficient and the microscopic transition probabilities (per unit time)
induced by the coupling between the solid and the electromagnetic field of
the wave. We consider now the electromagnetic energy density associated to
the wave (averaged on a period):
c
1 ³ 0 ´2
 = 0  |E |2 (363)
2
and the Poynting vector (averaged overe a period too):
1 0
 = 0  |E |2 (364)
2
The conservation equation for the electromagnetic energy density in the sta-
tionary state is:
 
=−  (365)
 
where  is the power density absorbed by the solid; the same power is lost
by the wave. So it is:
  
= 0  (366)
  
Since the intensity  of the wave is proportional to  ,  decreases in
space as stated by the law (358) thus it is:
   ³  ´ ” 1 ³ 0 ´2
= − 0 Σ = − 0 2  0  |E |2 (367)
    2
Using, again, the second equation in (362), we finally get:
 1
= 0 ” () |E |2 (368)
 2
Now, introducing the probability per unit volume and time   of a single
microscopic event which leads to the extinction of a photon to happen, we
get:
1 X 
0 ” () |E |2 = ~ (369)
2 

The quantity   can be estimated using the Fermi’s golden rule (see
Appendix: Quantum approximate methods). Solving with respect to ” ()
we get: µ ¶
” 2~ X 
 () = (370)
0 |E |2  
6.1 Kramers e Kronig relations

”
Once the ” () =  () has been calculated with (369), it is possible to
0 0
obtain  () =  () − 1 using an integral relation (and vice versa). It is
c
a general equation which comes both from the linear relation between the
polarization vector P() and the electric field () and from the principle
of causality (linear response theory). In presence of delay effects (due to
dissipation and dispersion), the most general linear relation between P()
and () can be written as:
Z ∞
P() = 0 ( )( −  ) (371)
0
If (− ) = (− ) it is P() = 0 (); which gives () its own meaning.
In particular () is equal to zero when   0 (principle of causality: effects
cannot precede causes): so we integrate between 0 and ∞. Equation (371) im-
plies also the time invariance (the hereditary integral is a convolution: id est
(  ) = ( −  )). If the polarizing field is harmonic: () =  exp(−)
we have µZ ∞ ¶
P() = 0 
( )   − (372)
0
So P() = P exp(−) and we can write the important relation:
e()
P = 0  (373)
being Z ∞
e() =
 ( )  (374)
0
from which it results immediately that:
 e∗ ()
e(−) =  (375)
More generally  e() is a tensor  e () which depends on the symmetry of
the crystal. In cubic crystals it reduces to a scalar quantity. Considering the
principle of causality, in equation (374) we could integrate between −∞ and
+∞ thus concluding that  e() is the Fourier transform of (). We can make
now an analytic continuation of  e() in the upper semiplane, introducing
0 0 ”
the complex frequency  =  +  ” . Since lim” −→∞   −  = 0,  e() is
analytic in the upper semiplane. As a consequence (using Cauchy theorems):
I
e()

 = 0 (376)
−

for any circuit  in the upper semiplane which does not contain . Let us
assume  ≥ 0 on the real axis. We chose as a circuit  = 1 ∪ 2 ∪ 3 ∪ 4
0
where 1 (− ≤  ≤ −; ” = 0), 2 ( = +    =  − 0), 3 (+ ≤
c
0
 ≤ ; ” = 0) and 4 ( =     = 0 − ). Solving the four line integrals
in the limit  −→ 0 and  −→ ∞, we get from (376):
Z
+∞ 0
e( ) 0

 − e
() = 0 (377)
0 − 
−∞
hR R +∞ i
−
where we should consider the integral as   : Cauchy principal value: lim→0 −∞
+ +
.
We can rewrite the previous equation as:
Z
+∞ 0
1 e( ) 0

e() =
  (378)
 0 − 
−∞
Introducing again the real and the imaginary part of the susceptibility, we
obtain the Kramers e Kronig relations:
Z
+∞ 0
0 1 ” ( ) 0
 () =  (379)
 0 − 
−∞
Z
+∞ 0 0
” 1  ( ) 0
 () = −  (380)
 0 − 
−∞
0
 () is the Hilbert transform of ” () and vice versa. Since from (375) we
0
get that  () is an even function, while ” () is an odd function, multiplying
0
the numerator and denominator by  + , we can rewrite equations (379)
as:
Z
+∞ 0 0
0 2  ” ( ) 0
 () =  (381)
 0 2 − 2
0
Z
+∞ 0 0
2  ( ) 0
” () = − 02 2
 (382)
  −
0
From the first equation of (381) we obtain the static relative dielectric con-
stant as:
Z ” 0
+∞
0 0 2  ( ) 0
 (0) = 1 +  (0) = 1 +  (383)
 0
0
c
The (383) is called sum rule. As a direct application of (381), we assume

that, in (369),

∝ ( − 0 ) (384)

So from (370) it is ” () = ( − 0 ), where  is a constant. Using the
(383) we notice that  = 0 [ (0) − 1] 2. Now, using the first equation in
(381), we finally get:
£0 ¤
0 0  20  (0) − 1
 () = 1 +  () = 1 + (385)
 20 −  2
Tthis equation can be compared with the one obtained in the polariton the-
ory.
6.2 Probability per unit time of optical transitions

When a crystal is perturbed by an electromagnetic wave whose electric field
in vacuum is
1
E = e0 (k·r−) +  (386)
2
in (370) the transition probability per unit time (and volume) may be evalu-
ated by means of the Fermi’s golden rule (see Appendix: Quantum approxi-
mate Methods). If we consider only the first order absorption processes in the
perturbation theory (the initial state is the ground state with single electron
()
energies  ) the rule is:
 2 ¯¯ c ¯¯2 ³ () ()

´
= ¯h | |i¯   −  − ~ (387)
 ~
where the matrix elements of the time independent part of the perturbation
operator can be written as:
c|i = 0
h | h|k·r e·b
p|i (388)
2
where pb = −~∇ is the momentum operator and  is the electron mass in
vacuum. In order to explain the (388) we should remember that, using the
vacuum gauge, the electromagnetic field can be derived using just the vector
potential A (with divergence equal to zero: ∇ · A = 0) as:
A
E = − (389)

B = ∇×A
c
In this situation the Hamilton function of a particle with charge  and iner-
tial momentum p =v, moving in an external electromagnetic field, can be
written as:
|p − A|2
 = (390)
2
If the charge is also under the effect of an additional electrostatic field (due
to other fixed charges) with potential  (r), and thus it has a potential energy
of  (r) =  (r), the total Hamilton function becomes:
|p − A|2
= +  (r) (391)
2
Developing the square of the absolute value, and substituting to p the mo-
mentum operator, we obtain the quantum Hamiltonian operator as:
b =
 b0 + 
b (392)
where
b p|2
|b
0 = +  (r) (393)
2
b  |A|2 
 = − A·b p+ ≈ − A·b p (394)
 2 
In the last approximation we neglected the quadratic term, according to first
order perturbation theory. This is consistent with the linear response theory
we are developing, see (371). In this way we cannot take into account nonlin-
ear optical phenomena. This treatment of e.m. radiation-matter interaction
in which particles obey the Schroedinger nonrelativistic quantum mechanics
and e.m. fields obey Maxwell equations is called semiclassical. This semi-
classical treatment does not explain spontaneous emission and high energy
relativistic processes. If the wave electric field has an expression such as (386)
then, using (389), the expression of the vector potential becomes:
0 (k·r−)
A= e +  (395)
2
So, if the charged particle is an electron with  = −, it is:
b  = 0 k·r e·b
 p− +  = c− +  (396)
2
We neglect the c.c., accounting for stimulated emission in which the initial
state is an excited state whose energy is higher than that of the final state,
and get the (387) and the (388) in the form:
 2 2 |0 |2 ¯¯ ¯2 ³ ´
= k·r
h| e·b ¯ () ()
p|i  (1 −  )  −  − ~ (397)
 ~ 4 2 2
c
Using the (370) we get:

2 X ¯¯ ¯2 ³ ´
” () = h |k·r
e·b
p |i¯  (1 −  )  () −  () − ~ (398)
 
0  2 2 
where we introduced the probability  that the initial state is occupied and
the probability 1 −  that the final state is unoccupied. Using the first of
0 0
the Kramers and Kronig equations we get  () = 1 +  (). Id est the unity
plus a sum of other terms such as the one in the right side of equation (385):
one term for each possible transition between the initial and the final state
of each electron in the solid (assuming unitary volume).
6.3 Interband optical transitions

6.3.1 Direct transitions
As an application of (398) we consider an intrinsic semiconductor at low
temperature: initially the valence band is completely occupied and the con-
duction band completely unoccupied, then an e.m. radiation represented by
the plane wave (386) hits the crystal. Transitions between occupied Bloch
states in the valence band
|i = k (r)k ·r (399)
with energy   (k ) and unoccupied Bloch states in conduction band
| i = k (r)k ·r (400)
with energy   (k ) occur, generating electron-hole pairs. Considering the

matrix elements in the (388) we notice that they are proportional to:
Z
¡ ¢
−~h| ∇|i = −~ ∗
k·r
k (r)
−k ·r k·r
 ∇ k (r)k ·r r (401)

where the integral is over the whole crystal volume. Introducing the relative
electronic coordinates r0 = r − n (where n is a generic node of the Bravais
lattice) and integrating over the volume of a single primitive cell, we can
write:
X Z 0 ³ ´ 0
0 0 0
k·r −(k −k−k )·n k ·r0
h | ∇|i =  ∗
k (r )−k ·r k·r
 ∇ 
k (r ) r
n 0
P (402)
−(k −k−k )·n
Using (350), we notice that the lattice sum n  is zero, unless:
k − k = k + g (403)
c
where g is a vector of the reciprocal lattice. This conservation principle

of the total wave vector in the crystal is a consequence of the translational
symmetry. Considering just the first BZ (g = 0) and observing that the
photon wave vector k is negligible with respect to the dimensions of the
Brillouin zone (electric dipole approximation), the previous relation becomes
k = k . So in the band representation (reduced zone scheme) the allowed
transitions are only the vertical ones. Since each initial state is orthogonal
to each final state we finally get:
Z
k·r 0 0 0
−~h | ∇|i ≈ ∗ 
k (r ) (−~∇) k (r )r (404)
0

Using the following relation:

 ³ (0) (0)
´
h|b
p|i =  −  h|r|i (405)
~
which comes from the time derivative of operator r:
r 1 h b i b
p
= r0 = (406)
 ~ 
we get: Z
k·r 0 0 0 0
−~h| ∇|i ≈   (k ) ∗ 
k (r )r k (r )r (407)
0
where:
[  (k ) −   (k )]
  (k ) = (408)
~
Here we can notice how the transition probability depends on the crystal
point symmetry. In case of a crystal with a center of symmetry, a transition
can only occur when the parity of the initial state is different from that of
the final state. Now we can finally get a more compact form of the imaginary
part of the dielectric susceptibility as a sum over all k0  in the first Brillouin
zone:
2 X
”
 () = 2 2
| (k)|2  [  (k) −   (k) − ~] (409)
0   k
where: Z ³ ´
0 0 0 0
 (k) =  (k ) ∗
k (r ) e · r k (r )r (410)
0
Generally  (k) can be considered as almost constant, so:
” 2 | |2 X
 () ≈  [ (k) − ~] (411)
0  2 2 k
c
where  (k) =   (k) −   (k) is the vertical difference in energy between
the conduction and the valence band. The latter equation can be rewritten
as:
” 2 | |2 ~2
 () ≈  (~) (412)
0 2 (~)2
where: X
 (~) =  [ (k) − ~] (413)
k
is by definition the joined density of states. Considering that the density of
states in the BZ is 2 (2)3 and considering a unitary volume, the joined
density of states can also be written as:
Z
2
 (~) ≈  [ (k) − ~] k (414)
(2)3 
and, using (146): Z
1 k
 (~) ≈ 3 (415)
4 |∇k  (k)|
where k is the constant energy surface of equation  (k) = ~. Equa-
tion (414) is equal to (415) due to the following property of the Dirac delta
function: Z P
( )
 [()]  () = 0 (416)
| ( )|
where  are the roots of equation () = 0. The assumption we made that
both the band  and  are doubly degenerate because of the spin, is always
valid for centrosymmetric crystals but it is not applicable to, e.g., zincblende
structures. We have a critical point whenever ∇k  (k) vanishes; the joined
density of states exhibist then a Van Hove singularity (147). Assuming par-
abolic (and isotropic) bands around Γ we can write:
 ~2 |k|2
 (k) =  (k) −  (k) ≈ +  (417)
2∗
where:
∗ (0)∗ (0)
∗ = (418)
∗ (0)+∗ (0)
is the reduced mass of the electrons and holes effective masses and  =
  (0) −   (0). Now, applying the methods learned in subsection Density
of states, we get, in the vicinity of the fundamental absorption edge, the
approximate form:
 (∗ )32 p
52
2
(~) = ~− (419)
3
c
.
Band structure in Germanium ” of Germanium
6.3.2 Indirect transitions

In the previous subsection the absorption properties of a semiconductor with
a direct gap such as ,  and all the II-VI compound are explained.
But, when the maximum of the valence band and the minimum of the con-
duction band do not share the same k the related direct transition is not
allowed not being vertical (see the figure below). In this case to find out
the absorption threshold one must consider more complex transition paths
involving phonons.
c
Indirect interband transitions

This is the case, for example, of , , . In silicon, e.g., the transition
 (Γ) →  () is prohibited by the electric dipole selection rules. In presence
of phonons an indirect transition can happen in two subsequent steps. These
indirect processes are also important in semiconductors with a direct gap
when vertical transitions are forbidden by the point symmetry (eq. 407),
even though only in Γ, as it happens in 2 . Referring to the adiabatic
theorem (eq. 220) the interaction  (r u) can be written as the following
lattice sum:
X X
 (r u) ≈  (r − n) − un ·∇ (r − n) (420)
n n
where we introduced the screened pair potentials  (r − n − un ). Develop-

ing un in normal coordinates (258) and considering, for the sake of simplicity,
only one active normal mode, the electron-phonon perturbation operator can
be written as:
Ã !
(q) X
b  = − √ e (q) ·∇ (r − n)q·n − (q) +  (421)
2  n
The term − (q) is responsible for the absorption (destruction) of a phonon,
while the c.c. is responsible for the stimulated emission (creation) of a
phonon. A deeper analysis of the matrix elements of operator (421) between
initial and final states such as |() (u) (r|0)  (see adiabatic theorem in
order to understand the meaning of symbols) shows that the allowed tran-
sitions conserve the total wave vector (see also equation (355) and inelastic
scattering theory):
k − k = ±q + g (422)
c
Normal  events correspond to g = 0 and Umklapp  events to g 6= 0. In the

presence of the perturbation (421) and of the photon-electron perturbation
b (396) the indirect transition can occur in two different ways. In the first
one there is at first a direct vertical transition  (Γ) →  (Γ) and then,
due to a creation or a destruction of a phonon with wave vector q(), the
quasi horizontal transition  (Γ) →  () occurs. In the second one at
first, due to a creation or destruction of a phonon with wave vector q(), the
transition  (Γ) →  () takes place and then there is the vertical transition
 () →  () caused by the absorption of a photon. The probability per
unit time of these events can be calculated using the Fermi’s golden rule up
to second order:
¯ Ã !¯2
 X ¯¯ 2 X h | b  |ih|
b |i h | b  |i ¯¯
b  |ih|
= ¯ + ¯ (423)
 ¯~  −  − ~  −  ± ~ (q) ¯
k k 
× [ (k ) −  (k ) − ~ ± ~ (q)]
Since the previous formula contains (q), the amplitude of the phonon
normal coordinate, the probability of indirect transitions depends on tem-
perature. (q) is a fluctuating random variable whose quadratic mean
value can be calculated by means of the Bose-Einstein distribution. We can
actually use the fact that the mean total energy of the oscillator is twice its
mean potential energy:
® ~ (q)
 2 (q) 2 (q)  ∝ ~   (q)
(424)
  
−1
In (423) the factor multiplying the delta function weakly depends on the wave
vector. Moreover, since the second denominator is always greater than the
first one, the second type of transition can be neglected. In this way, repeat-
ing considerations similar to those in the previous subsection, the imaginary
part of  for an indirect transition reduces to:
Z Z
”
2 ®2
 ∝  (q )   [ (k ) −  (k ) − ~ ± ~(q)] k k (425)
 
Assuming parabolic bands around stationary points we get, near the indirect
band gap  =  () −  (Γ),
” ∝ (~ ∓ ~(q) −  )2 (426)
when ~ ≥  ± (q) and 0 otherwise. Neglecting the dispersion of phonon

bands (as in the Einstein model), we set as mean frequency (q) =  (q),
c
the frequency corresponding to wave vector q = q . Because the (425)

strongly depends on temperature it is possible to experimentally ascertain
the existence of an indirect gap measuring the temperature dependence of
absorption. The electron-hole pairs with different wave vectors created during
indirect processes generally have a longer lifetime than that of the pairs with
the same wave vector generated in direct processes.
6.4 Intraband transitions and plasmons

In metals and in extrinsic semiconductors, besides the interband transitions
at high frequencies already treated in the previous subsection, there are quasi-
free charge carriers which populate partially occupied bands and cause the
continuous background shown in the introductive figure on optical absorp-
tion. A quantum treatment of intraband indirect transitions (vertical intra-
band transitions do not exist) could proceed exactly as in the previous sub-
section just considering occupied and unoccupied states in the same band.
The states associated to quasi-free electrons are perturbed by both electron-
phonon and photon-electron interactions. The model adopted to describe
this situation is similar to that used in the theory of stationary transport
phenomena (see below) in which just the electron-phonon interaction is con-
sidered (for the dynamic part) and the states are Bloch wave packets obeying
semiclassical equations. Here we limit ourselves to some simple considera-
tions in the free electrons approximation. We proceed in analogy with the
coupling between photons and oscillating ions plasma described in subsection
polaritons. We can apply the same theory to the jellium: a plasma consti-
tuted by free electrons and by a uniform background of positive ions. The
main difference lies in the fact that here the resonance frequency correspond-
ing to binding energy is zero: thus wepwrite  = 0. Besides the plasma
frequency is that of free electrons  =  2 . The dielectric constant in
the plasma then becomes:
µ ¶
 2
 = 0 (1 + ) = 0 1 − 2 (427)

and the refractive index: r

 2
= 1− (428)
2
Consequently the dispersion relation of transverse e.m. waves becomes:
sµ ¶
  2
= 1− 2 (429)
 
c
ck 2.5
1.5
0.5
0
0 0.5 1 1.5 2
omega
Jellium dispersion relation
A thorough analysis of Maxwell equations in jellium would show that

transverse e.m. waves can propagate only if their frequency is above the
plasma frequency (real , black line in the figure). Below the plasma fre-
quency waves are damped (imaginary , red in the figure). Exactly at the
plasma frequency  (  ) = 0 stationary longitudinal waves, associated to a
longitudinal oscillating electric field, are established in jellium (mixed collec-
tive non propagating modes). These stationary waves are quantized and their
energy quanta ~ are called plasmons. In metals the plasma frequency oc-
curs in the ultraviolet. Plasmons are the evidence of collective motions whose
understanding requires going beyond the approximation of independent elec-
trons used in normal band theory. The oversimplified model we have used
here has classical bases (Drude-Lorentz theory). A quantum description of
the free electron gas would lead to a wave vector depending dielectric function
(spatial dispersion) as it is introduced in the theory of potential screening
(217).
7 Transport phenomena
7.1 Phenomenology
In a crystal containing mobile charges (conductor or semiconductor) the pres-
ence of an electric field E of external origin causes an irreversible charge
transport phenomenon (electric current) removing the body from thermody-
namic equilibrium. On the other hand, in all crystals a temperature gradient
∇ causes an irreversible heat flux called conduction. From a macroscopic
c
point of view (adopted by the thermodynamics of irreversible processes) when

the cause of the flux is small, we are within the linear response regime and
two famous phenomenological laws, Ohm’s local law and Fourier’s law, well
describe both transport processes:
j = E (430)
q = −k  ∇ (431)
where j is the charge current density vector. Its flux through a surface 
represents the amount of charge per unit time, current , flowing through the
surface: Z
 = j · n (432)

The principle of electric charge conservation can be written as the continuity
equation for the charge density  (charge per unit volume).

+ ∇ · j =0 (433)

In the Fourier’s law q is the heat current density vector. Its flux through a
surface  represents the amount of heat flowing through the surface per unit
time: Z
†
= q · n (434)
 
q is strictly connected to the entropy current density vector q  whose
divergence appears in the balance equation for entropy density  (entropy per
unit volume)
 ³q ´

+ ∇· = [] (435)
 
 [] is the entropy production (inherently positive). The latter equation
shows, in a local form, the second principle of thermodynamics. For electrical
conductivity we use the symbol  while we use the symbol k  for thermal
conductivity. More precisely in crystals both conductivities are second order
tensors and the two phenomenological laws should be generalized as:
13
X
 =    (436)

13
X 
 = −  (437)


c
These new laws show that generally the current density vectors have not
the same direction of the electric field and of the temperature gradient. As a
matter of fact even more general kinetic coefficients exist: a temperature gra-
dient can generate an electric current and an electric field can create a heat
flux. The kinetic microscopic theory (see below) explains these correlated
effects called thermoelectric effects (Seebeck and Peltier effects). Further-
more, the Onsager theorem, based on microscopic reversibility of dynamic
laws, states that the two conductivities are symmetric tensors:  =   ,
 =  . Since every symmetric tensor, with an appropriate rotation of the
coordinate system, can be reduced to the diagonal form:
⎛ ⎞
1 0 0
σ = ⎝ 0 2 0 ⎠ (438)
0 0 3
it turns out that the number of independent components of both σ and k
can be, at most, three. Due to crystal symmetry this number can be further
decreased. Actually, let R be a symmetry operation and R the unitary matrix
which transforms the atom coordinates (position vectors) according to the
symmetry operation, then it should be:
0
σ = RσR −1 = σ (439)
Repeating this consideration for each symmetry operation, it is shown that

triclinic, monoclinic and orthorhombic crystals have three independent com-
ponents: 1 6=  2 6=  3 ; tetragonal, trigonal and hexagonal crystals have
two independent components:  1 =  2 6=  3 ; and cubic crystals only one:
 1 =  2 =  3 (isotropy). The situation is analogous for thermal conductivity.
As far as the temperature dependence is concerned, the electrical resistivity
( = 1) of metals shows a monotonically increasing trend. At low temper-
atures it increases as 0 +  5 . 0 is called residual resistivity depending on
static lattice defects (vacancies/interstitials, dislocations, packing defects).
At high temperatures the resistivity increases linearly:  ∝  . Generally,
as stated by Matthiessen law, the total resistivity is the sum of both an
athermal (0 ) and a thermal contribution.
c
Electric conductivity vs T
The case of semiconductors is more complex since their resistivity depends

also on the possible doping and its treatment needs special considerations. In
metals the thermal conductivity at low temperatures linearly increases with
 , reaches a maximum value and then rapidly decreases until it reaches a
constant value at high temperatures. In insulators, instead, at low tempera-
tures the thermal conductivity increases as  3 , reaches a maximum value and
then decreases as  −1 at high temperatures. In metals, at very low temper-
atures, the ratio between electrical and thermal conductivity is substantially
proportional to temperature and does not depend on the particular metal
(Wiedemann-Franz law).
Metals Insulators
c
7.2 Kinetic theory

Since transport coefficients refer to irreversible processes in non-equilibrium
conditions, for a theoretical treatment we cannot exploit the traditional
bridge connecting the microscopic (quantum) dynamic description to equilib-
rium macroscopic properties (functions of state): the equilibrium statistical
mechanics. As long as electrons are concerned, for equilibrium situations
major information is given by the Fermi-Dirac distribution (116). In the
present non-equilibrium case we should use, instead, the Boltzmann kinetic
theory adapted to electrons in the semiclassical dynamic scheme. At first
we start with a completely classical description of a gas with  non inter-
acting electrons. In the phase space (r p) of one electron we can introduce
the single particle distribution function: (r p)rp = representing the
number of electrons whose instantaneous state is within a volume Γ = rp.
 =  = (r p)rp represents the probability of occupation of the
instantaneous state (r p). If, moreover, the electrons are under the action of
an external field of forces F(r) = −∇U (r), the motion is led by the single
particle Hamilton function  = |p|2 2 + U (r). For a given energy , the
dynamic trajectory of an electron in phase space obeys Hamilton equations
(191) from which one can demonstrate the Liouville theorem: (r p) (and,
consequently, (r p)) is constant along all dynamic trajectories:
 (r p)
=0 (440)

The proof is shown in the appendix below. If now we want to introduce in
the simplest way the effect of interactions (collisions) with other particles
(electrons and/or phonons) we can correct ad hoc the Liouville theorem as
follows:
µ ¶
 (r p)  (r p)  (r p) − 0 (r p)
= =− (441)
    (p)
The previous equation is based on the heuristic hypothesis that, in the ab-
sence of external forces and temperature gradients, thermal equilibrium is
restored by scattering between particles in a mean relaxation time  (p). In
the classical case 0 (r p) is the Maxwell-Boltzmann distribution
0 (r p) =  −1 exp(−0 (r p)  ) (442)
where Z
= exp(−0 (r p)  )rp (443)
is the partition function.

c
Computing the total derivative in equation 441 and using the Hamilton
equations we get the Boltzmann equation:
  ³ p ´  (r p) − 0 (r p)

+ F(r) · + · =− (444)
 p  r  (p)
Using the effective mass theorem (equivalent Hamiltonian) we now modify

the above equation to make it applicable to a quantum electrons system.
Electrons will be described by semiclassical Bloch wave packets moving in a
partially occupied band  = (k) neglecting interband transitions. This is
the case of a conductor under the effect of a weak electric field E: F(r) =
−E. Employing the results of subsection Semiclassical dynamics, we can
now write the Boltzmann equation in the stationary case  = 0 as:
F(r)  (r k) (r k) (r k) − 0 (k)

· + v(k) · =− (445)
~ k r  (k)
where:
1 (k)
v(k) = (446)
~ k
In the fermion quantum situation
1
0 (k) ≈ h i (447)
(k)−
exp   (r)
+1
is the Fermi-Dirac distribution. The density of occupied states in k space

depends also on r and its expression per unit volume is:
2
D(r k| (r)) =  (r k) (448)
(2)3
where we have considered the spin degeneracy and assumed a local temper-
ature field  (r) with a weak gradient. We can thus apply a linear response
approximation. To do that we rewrite equation (445) as:
F(r)  (r k)  (r k)

 (r k) = 0 (k) −  (k) · −  (k)v(k) · (449)
~ k r
We solve now this equation approximately up to the first step of an iterative
procedure:
F(r) 0 (k) 0 (k)

(r k) ≈ 0 (k) −  (k) · −  (k)v(k) · (450)
~ k r
c
Computing the partial derivatives

0 (k) 0
= ~ v(k) (451)
k  µ ¶
0 (k) 0  0 (k) −  
= = −
r  r   (r) r
we finally get:
∙ ¸
0 ( −  ) 
(r k) ≈ 0 (k) −  (k) v(k) · F(r) − =
  (r) r
= 0 (k) +  (r k) (452)
7.2.1 Appendix: Liouville theorem

The motion of a set of N particles in phase space is equivalent to the motion
of a fluid with density  = (r p); thus the conservation principle of mass
corresponds to the conservation of the number of particles. This principle
can be written in the form of a continuity equation:

+ ∇· (v) = 0

In the phase space of a particle (independent particles gas in an external
field):
µ ¶ X µ ¶
(r p) X 
3 3
  
+  +  =0
 =1
  =1
 
Using the Hamilton equations we get:
(r p)
+ { } = 0

where:
X 3
  X  
3
{ } = −
=1
  =1
 
is the Poisson bracket. The total derivative of , as already discussed in the
derivation of (444), is:
 (r p)
= + { }
 
and, consequently, it is immediately obtained that  = 0 (c.v.d.). Note
that the last equation is very similar to that for the time derivative of a
quantum operator b
b  b  h b bi
= +   . (453)
  ~
c
7.3 Electrical conductivity

Once we have solved the Boltzmann equation (445) in the linear response
regime (eq. 452), the current density vector in the local Ohm law (430) can
be computed (as a function of the electric field) as the integral over the whole
k space :
Z Z
−2
j = − D(r k| (r))v(k)k = (r k)v(k)k (454)
(2)3
To a first approximation, neglecting the thermoelectric effects due to the
thermal gradient, we get
Z
22
j=  (k)( −  ) (v(k) · E) v(k)k
(2)3
Where, based on Kramers theorem ((−k) =(k) → v(−k) = −v(k)) and
0 (−k) = 0 (k), we have taken into account that
Z
2
0 (k)v(k)k = 0 (455)
(2)3
and made the further approximation 0  ≈ −( −  ). Thus (see
subsection density of states):
Z
22 
j= 3
 (k)( −  ) (v(k) · E) v(k) 
(2) |∇k (k)|
and finally Z
2
j =  (k) (v(k) · E) v(k) (k) (456)

where the integral is over the Fermi surface (k) =  ; we now define the
density of states on the Fermi surface as:
2 1
 (k) = 3
(457)
(2) |∇k (k)|
The electric conductivity tensor can be eventually written in the form:

Z
2
σ=  (k) (k)v(k)v(k) (458)

Notice that only electrons on the Fermi energy can contribute to the charge
transport process. In metals the dependence of conductivity on temperature
is lumped in  (k) =  (k), on the Fermi surface, which is mainly due to
c
electron-phonon scattering. The probability of a scattering event per unit

time is (  (k))−1 and can be computed by means of the Fermi’s golden rule
up to first order using the matrix elements of an appropriate electron-phonon
interaction operator. At high temperatures this implies a dependence of
(  (k))−1 on h2 (q)i which, in turn, is ∝  and explains the linear trend
of resistivity with respect to temperature. At low temperatures, instead,
elastic scattering with lattice imperfections prevails, and this explains the
athermal residual resistivity. At intermediate temperatures a simple theory
combining the different processes in a sufficiently accurate way does not exist.
Actually it is very complicated to account for  (Umklapp) processes. In
simple metals the Fermi surface is spherical (free electrons model), to a first
approximation, and all the quantities that have to be integrated are constant.
So in this case we get, after a few simple steps:
2 2 (2∗ )32 p

 =    (459)
3 2 2 ~3
Using the expressions of Fermi energy (115) and Fermi speed ~ ∗ we
finally get an equation often used in applications:
µ ¶
2 
= ∗  (460)
  
In the case of transition metals (where there is a superposition between bands
 and , and the Fermi level is within the  band) it is possible to use
the previous formula in an additive way (460) for the two different types of
electrons, with very different values for effective masses and relaxation times.
In the case of noble metals (where there is a superposition between bands 
and , and the Fermi level is above band  within a  band) wave vectors on
the Fermi surface are very near the edge of the first Brillouin zone producing
a strong deviation of the Fermi surface from the spherical shape with some
band regions having a negative curvature and, thus, implying conduction due
to holes. Also in this situation an additive (two bands) model with two types
of charges, with opposite signs, is used.
7.4 Electronic thermal conductivity

Here we start with the thermodynamic relation for the differential of specific
entropy:
1 
 =  −  (461)
 
where  =  is the internal energy per unit volume and  =  is
the number of electrons per unit volume.  is the chemical potential, which
c
can be approximated by the Fermi energy  . By analogy we can make the

assumption that the entropy flux is connected to energy and electrons fluxes
through the following equation:
1 
j = j − j (462)
 
Then irreversible thermodynamics gives the following formula for entropy
production (eq. 435):
µ ¶ ³´
1
 [] = ∇ · j + ∇ · j (463)
 
Remembering that the heat current density vector is defined as q =  j ,
we obtain:
q = j − j = v − v =v − v = (−)j (464)
We approximate  by  , and then, using again the method of microscopic

kinetic theory (see previous section), we can compute q as:
Z
2
q = (r k)(− )v(k)k (465)
(2)3
using now the following expression:
0 ( −  ) 
(r k) =  (k)v(k) · (466)
  r
which is valid in the absence of external fields. Remembering Fourier law,
we can finally write thermal conductivity as follows:
Z µ ¶
2 ((k) −  )2 0
 =  (k)v(k)v(k) − k (467)
(2)3  
In this case we cannot simply approximate −0  with ( −  ), since
now we would obtain a null result. Instead we must use the better Sommer-
feld approximation (eq. 122). Using, as usual, the result
E
k =  (468)
|∇k (k)|
and again defining the density of states on the Fermi surface as
2 1
 = , (469)
(2)3 |∇k (k)|
c
after an integration over energy , we finally get

Z
1 2 2
 =     (k) (k)v(k)v(k) (470)
3 
But we have already demonstrated that the above integral coincides with the
electrical conductivity divided by the square of the electron charge (eq. 456).
Thus:
 2 2
 =  (471)
3 2
The above equation is called Wiedemann-Franz law. If we make use of this
result, knowing the trend of electrical conductivity, we can get the trend
of thermal conductivity with respect to temperature. Figures are shown
in subsection Phenomenology. We have assumed that all relaxation times
(scattering processes) involved in heat transport in metals coincide with those
involved in charge transport. This is not exact and can induce a deviation
from Wiedemann-Franz law, mainly at intermediate temperatures.
7.5 Lattice thermal conductivity (insulators)

In insulating crystals heat conduction is due to the non-equilibrium kinetics of
lattice waves packets in presence of a temperature gradient. In this situation
the carrier particles are phonons packets moving with the group velocity:
  (q)
v (q) = (472)
q
and carrying the energy quantum ~  (q). The distribution function in ther-
modynamic equilibrium is the Bose-Einstein factor (mean occupation num-
ber    of a normal mode with wave vector q):
1
0 (q) = ~   (q)
(473)
 −1
 
The phonons chemical potential is zero, since the number of phonons is not
conserved. The heat density flux vector can be written as:
Z
1 X
q =  (r q)~  (q)v (q)q (474)
(2)3 
Thus in equation (450) we can write, in the present case:
~   (q)
0 (q) 0 (q)  1 ~  (q)   

= = µ ¶2 (475)
r  r   2 ~   (q) r
  
−1
c
which can be written as:

0 (q)  (q) 
= (476)
r ~ (q) r
introducing the specific heat of each phonon   (q) (eq. 234):

~   (q)
1 ~2  2 (q)  
 (q) = µ ¶2 (477)
  2 ~   (q)
  
−1
The heat density flux can thus we written as (see (450) with q instead of k):
Z
1 X 
q = − 3   (q)v (q)v (q) (q) q (478)
(2)  r
And then, using Fourier law, the lattice thermal conductivity tensor can be
expressed as:
Z
1 X
k =   (q)v (q)v (q) (q)q. (479)
(2)3 
Sometimes the phonon mean free path is introduced as Λ (q) = v (q)  (q).
Using a simplest treatment, strictly inspired by the kinetic theory of gases,
we could immediately write
1
k = Λ (480)
3
where  is the total specific heat per unit volume and  and Λ are oppor-
tune average values of v (q) and Λ (q). The above equation can be derived
by the following elementary considerations. Phonons propagating a distance
∆ =   = Λ corresponding to a temperature drop ∆ , carry a quantity
of heat per particle  = ∆ , being  the contribution of a single particle
(phonon) to specific heat. Then the heat flux is p = ∆  where  is
the number of particles per unit volume and  = h2 i. If the temperature
2
gradient is   we can write
p  = − ( )   = − Λ ( )
and  = 3, being  = h2 i. In a Debye model, neglecting the con-
tribution of optical phonons (which have nearly zero group velocities), the
velocity of acoustic phonons is constant and is equal to the mean speed of
sound. For phonon-phonon interactions (anharmonic effects) we can con-
sider that Λ( ) ∝ (   )−1 . At high temperatures    ∝  and the
specific heat is constant. At low temperatures the specific heat follows  3 ,
c
while the mean free path is limited by the crystal size Λmax = . From a de-
tailed microscopic point of wiew, it can be shown that only phonon-phonon
Umklapp  scattering guarantees local thermodynamic equilibrium and is
responsible for a finite thermal conductivity.Without the contribution given
by  scattering events, the thermal conductivity would diverge, as proved
by Peierls. Anyhow, without the size limitation to mean free path, at low
temperatures k would diverge as (   )−1 ≈ exp(~ min   ), where
min is the minimum frequency of phonons which may encounter a  scat-
tering process. Two phonons can participate to a  scattering process when
the sum of their wavevectors goes beyond the first Brillouin zone. An im-
portant role is also played by elastic scattering off lattice defects. Whenever
the wavelength of a phonon is greater than the size of a defect a Rayleigh
scattering occurs, for which Λ ∝ |q|−4 .
8 Appendix: Quantum approximate meth-

ods
8.1 Time-dependent perturbations
A given system, initially isolated, since time instant 0 is exposed to a dy-
namic perturbation (i.e. to the time-dependent interaction with a second
coupled system) until time instant 0 +  . The response of the first system
(induced quantum transitions) can be obtained solving Schroedinger equa-
tion:
 ³ b ´
~ =  + b ()(0   )  (481)

Function (0   ) is 1 when 0   ≤ 0 +  and 0 otherwise. b () is the
time dependent perturbation operator.  b  is the unperturbed Hamiltonian of
the initially isolated system the spectrum of which is assumed to be known.
()
Let us first assume that the unperturbed energy levels  are discrete and
non-degenerate.
b   = () 
 (482)
If at time 0 the system is in the stationary state  (it is not necessarily
the ground state), then:
(0 ) =  (483)
Following Dirac, we look for a solution of the Schroedinger equation (481)
such as:
X ()

() =  () − ~  (484)

c
That is, we use a generalized form of superposition principle with time-

dependent coefficients  (). At time 0 +  the perturbation ceases. In view
of the chosen initial condition, the square of the absolute value of  (0 +  )
represents the transition probability → ( ) from initial state | i = |i
to final state | i = |i during a time  (by now on we will consider 0 = 0):
→ ( ) = | ( )|2 (485)
Substituting the wave function (484) in the Schroedinger

R equation and using
the orthonormality of the set of all  (h|i = ∗  r =   ), we get a
system of ordinary differential equations for the coefficients  ()
 () X b
~ = h| ()|i   () (486)
 6=
where we have defined:

() ()
 − 
  = (487)
~
and with the initial conditions:
 (0) =   (488)
The matrix elements of the perturbation operator are defined as:

Z
 () = h| ()|i = ∗ b () r
b (489)
and we assumed that all the diagonal elements are zero, as it is true in many
important cases (see below). At first we integrate each equation of the system
with respect to time obtaining (for  6= ):
Z
1 X b 0 0 0 0
 () = h| ( )|i   ( ) (490)
~ 0 6=
Then we proceed with an iterative solution method assuming that all the
zero-th order solutions coincides with the initial conditions:
(0)
 () =  (0) =   (491)
Then the first order solutions are obtained substituting the zero-th order
solutions in the right side of equations(490) :
Z
(1) 1 X b 0 0
(0) 0 0
 () = h| ( )|i   ( ) (492)
~ 0 6=
c
thus obtaining
Z X
1 0
h|b ( )|i     =
0 0
(1)
 () = (493)
~ 0 6=
Z 
1 0
h|b ( )|i  
0 0
~ 0
The iteration proceeds substituting

Z 
1 0
h|b ( )|i  
0 0
(0)
 () + (1)
 () =   + (494)
~ 0
again in the right side of equations (490):

Z 0 h i 0
(2) 1 X b 0    (0) 0 (1) 0
 () = h| ( )|i  ( ) +  ( )  (495)
~ 0 6=
and then obtaining:

Z
(2) 1  b 0 0 0
 () = h| ( )|i   + (496)
~ 0
µ ¶2 X Z  Z 0
1 0 00 00
h|b ( )|i  h|b ( )|i   
0   00 0
~ 6= 0 0
and so on, after  substitutions:

Z 0 h i 0
(+1) 1 X b 0   (0) 0 (1) 0 () 0
 () = h| ( )|i  ( ) +  ( ) +  +  ( ) 
~ 0 6=
(497)
We consider now the first order approximation in the situation in which
b () = 
c0 . From (493 485) we get:
¯ ¯2
¯ c ¯ ¡ ¢
¯ h|  |i¯ 2   
2 sin
0
(1) 2
→ ( ) =  ¡ ¢ (498)
~2    2
2
c
y 1
0.75
0.5
0.25
0
-20 -10 0 10 20
x
2
Function y= sin(2)
(2)
2 ;  =   
If   À 1 we obtain that the following function

¡
2   
¢
2 sin 2
 ¡ ¢ (499)
  2
2
can be approximated with an isosceles triangle with base 4 and height
 2 : the area under the principal maximum is thus approximately equal to
2 . In this limit
³ the function´ can be approximated by a Dirac delta function
() ()
 ( ) = ~  −  . Summarizing, in the limit    À 1 ( → ∞)
we can write the transition probability as:
(1) 2 ¯¯ c ¯2 ¡
¯ ¢
→ ( ) ≈ ¯h| 0 |i¯  () − 
()
(500)
~
From an experimental point of view, it turns out that if the initial state |i is
an excited one it is not a truly stationary state and it tends to spontaneously
decay also in the absence of a any perturbation (spontaneous decay) following
an exponential law (in the simplest cases):
| ()|2 = −  (501)
where the mean lifetime   of state |i depends on the experimental energy
() ()
uncertainty ∆ , around the mean value  , with which the energy of
the state is known through the empirical inequality:
()
∆  ≥ ~ (502)
From a physical point of view, then, the 500 can be used when −1
 ¿  ¿
 .
c
8.1.1 Harmonic perturbation

For the time dependence:
b () =  c− −
c+  +  (503)
³ ´†
c
where  = − c +
, and in the limit  → ∞ ( −1 ¿  ¿   ), with a
procedure similar to that just used, we get the transition probability per unit
time e→ in the form:
(1)
2 ¯¯ c± ¯2 ¡
¯ ¢
e→
(1)
≈ ¯h|  |i¯  () − 
()
± ~ (504)
~
In this case the delta function tells us that transitions can happen only if the
total energy of the two coupled systems is conserved:
() − 
()
± ~ = 0 (505)
in agreement with the Bohr postulate of old quantum theory when ~ is the
energy of a light quantum perturbing an electron system. Using equations
(496), instead, we would obtain the second order result (useful to describe
e.g. non linear optics):
2 ¯¯ c± ¯2 ¡
¯ ¢
e→
(2)
( ) ≈ ¯h| |i¯  () −  ()
± ~ + (506)
~ ¯ ¯2
2 ¯¯ X h|
c± |ih| c± |i ¯¯ ¡
() ()
¢
¯ ¯   −  ± ~ ± ~
~ ¯6= () − ()
± ~ ¯
Note that the number of energy quanta ~ of the second system involved in
the process is equal to the order of the perturbation. Previous equations
describe the absorption or the stimulated emission of one or two energy
quanta ~, respectively. At the second order the transition can be seen as
the sequence of two transitions through intermediate states |i. Intermediate
transitions do not conserve energy. If the final state is a quasi-stationary state
the singularity of the Dirac delta function can be removed. Introducing the
density of final states () and integrating the (504) we get:
Z
e (1) 2 ¯¯ c± ¯2 ¡
¯ ¢
→ ≈ ¯h| |i¯  () −  ()
± ~ (() )() =(507)
~
2 ¯¯ c± ¯2
¯ ()
¯h| |i¯ ( ∓ ~)
~
usually written as:
e (1) 2 ¯¯ c± ¯¯2
→ ≈ ¯h| |i¯ ( ) (508)
~
c
the above equation is called Fermi’s golden rule, where  =  ∓~. Study-
ing the matrix elements h|c± |i we get the so called selection rules, strictly
connected to the symmetry of the system.
8.1.2 Photon-electron interactions: electric dipole selection rules

The theory outlined in the previous paragraph can be applied to the interac-
tion between an electromagnetic wave and the electrons of an atomic system
(atoms, molecules, solids, in particular crystals...); it becomes particularly
simple in case of a plane, monochromatic, linearly polarized wave of wave-
length  À , where  is the characteristic linear dimension of the elementary
volume occupied by electrons (in the case of crystals it is the volume of the
primitive cell). For a more general treatment see Optical properties. If the
wave is polarized along axes  and propagates along axis  For the sake of
simplicity we refer to the single electron of a hydrogen atom,  is the atomic
radius and we set the origin at the nucleus. The wave electric field can be
written as follows:
E = u cos( − ) (509)
where  = (2) is the wave vector. Thus (||) ≈  = 2 ¿ 1,
and we are lead to the simpler form:
1 ¡ ¢
E ≈ u cos() = u  + − (510)
2
As far as this field does not appear anymore as a propagating wave, we can
neglect magnetic effects and consider the atom as an electric dipole μ = −r
( is the absolute value of the electron charge) interacting quasi-statically
with the instantaneous electric field. Thus we assume as perturbation oper-
ator
1 1
b = −bμ·E = b   + b
 − (511)
2 2
b is then written in the form (503) whith  c+ =  c− = 1 b
2
 = 12 . In this
situation the matrix elements of the perturbation operator are h | c± |i =
1
2
h ||i. In order to see when they vanish we must consider the following
integral: Z
h ||i = ∗  r (512)
In case of an hydrogen atom the orbital part of the 22 wavefunctions be-
longing to the same energy level
136
() ≈ − eV; ( = 1 2 ) (513)
2
c
can be written in spherical coordinates (  ) as:

 ()
 (  ) =  ( ) (514)

where  () is the radial wavefunction,  ( ) is a spherical harmonic, 
is the principal quantum number,  is the orbital quantum number and 
the magnetic one. The parity of these states with respect to a coordinate
inversion is defined and depends exclusively on  ( ); it is (−1) (when
r → −r,  = |r| remains the unchanged). In order to have a nonzero integral,
since  is an odd function, it should be  − = ∆ = ±  , i.e. the
parity of the final state should be different from that of the initial state. A
further consideration about the angular momentum of the electron-photon
system imposes the stronger limitation: ∆ = ±1 (optical selection rule).
In an atom with atomic number  in the mean field approximation, the
single electron energy levels depend on () and the degeneracy reduces to
2(2 + 1). However in the (514)  while the  () change as functions of ,
the  ( ) remain the same: the selection rules don’t change. In a crystal
instead the initial and final states are Bloch waves normally belonging to
different bands and one must consider the total fluctuating electric dipole of
the crystal basis. This leads to specific selection rules which can also involve
phonons when the adiabatic approximation is not applicable.
8.2 Static perturbation theory

8.2.1 Adiabatic switching on
Consider a perturbation operator (without the (0   ) modulation) such
as:
b () = lim 
c (515)
→0
where   0 and  ¿ 1. Setting 0 = −∞ and 0 +  = 0, equations (493)

can be written as:
Z
(1) 1 0 0 0
c |i  0 =(516)
 (0) = h|
~ −∞
Z 0 c|i c|i
 c|i
0 0 0  h| h|
h|     = =  () ()
~ −∞ ~   +   −  + 
Notice that lim→0 has not been taken yet. Taking the limit now, it is the
c, we obtain:
adiabatic switching on of the static perturbation 
c|i
h|
(1)
 (0) =  () ()
(517)
 − 
c
and substituting this result in (484), for a generic perturbed eigenfunction

(1)
 we get:
X h|
c|i
(1)
 =  +  () ()
 (518)
6=  − 
Using these perturbed eigenfunctions we can now compute the perturbed en-
ergy levels as: ³ ´
 ≈ h(1)
 | b  + c |(1)
 i (519)
Up to second order in  and considering the necessary condition for series
convergence ¯ ¯
¯ () ¯
c|i¯¯
¯ − () ¯ À ¯¯h| (520)
after long calculations we get:
¯ ¯2
¯ c ¯
X ¯h| |i¯
()
 =  c|i + 2
+ h| (521)
() ()
6=  − 
(1)
In many applications keeping in the final formulae for the coefficients  (0)
the initial expression:
c|i
h|
 () ()
(522)
 −  + 
it is possible to account for spontaneous deacay and for other dissipative
phenomena provided the value of  is adjusted in order to have a better
match with experimental data. Finally, with an infinitesimal  it is possible
to prove that (PP = Cauchy principal part, see also Kramers and Kronig
relations):
1 1 ¡ () ¢
() ()
≈ PP () ()
−   − () . (523)
 −  +   − 
This last equation and the following one, a new representation of Dirac delta
function coming from the Green functions formalism, are very commonly
used. µ ¶
1 1 1 
() ≈ − Im = (524)
  +    + 2
2
c
8.2.2 Degenerate energy levels

When we use formulae such as (521) and n(518) owe are assuming that degen-
()
erate levels in the unperturbed spectrum  do not exist. As a matter
()
of fact if to a given level  there correspond more than one eigenfunction
| i in the perturbation series there would be divergent terms such as:
c| i
h |
() ()
(525)
 − 
We can overcome this problem as follows: we assume that to the degenerate
energy level  () there correspond the two unperturbed eigenfunctions 1 =
|1i e 2 = |2i. To a first approximation we write the perturbed wavefunction
as:
 = 1 + 2 (526)

substituting now  in the equation for perturbed stationary states
³ ´ ¡ ¢
b c
 +  (1 + 2 ) =  () + ∆ (1 + 2 ) (527)
and using the orthonormality condition for functions 1 and 2 we get the
homogeneous linear algebraic system:
Ã !µ ¶ µ ¶
c|1i − ∆
h1| c|2i
h1|  0
c|1i c|2i − ∆ = (528)
h2| h2|  0
c|1i = h1|
Considering that h2| c|2i∗ , the solubility condition is:
³ ´³ ´ ¯ ¯2
c|1i − ∆ h2|
h1| c|2i − ∆ − 2 ¯¯h1| c|2i¯¯ = 0 (529)
The two roots are:

" #
1 ³ ´ 1 r³ ´2 ¯ ¯2
∆ ± =  c|1i + h2|
h1| c|2i ± c|1i − h2|
h1| c|2i¯¯
c|2i + 4 ¯¯h1|
2 2
(530)
When the diagonal matrix elements are zero, we simply have:
¯ ¯
± ¯ c ¯
∆ = ± ¯h1| |2i¯ (531)
The unperturbed degenerate level  () so splits into the two perturbed non-
degenerate levels: ¯ ¯
± () ¯ c ¯
 =  ±  ¯h1| |2i¯ (532)
c
¯ ¯
¯ c ¯
Between them there is an energy gap  = 2 ¯h1| |2i¯. Using the first of
the (528)
c|2i± = 0
−∆ ± ± + h1| (533)
and the normalization condition
||2 + ||2 = 1 (534)

c|2i  0, we get:
when h1|
1 1
± = √ ; ± = ∓ √ (535)
2 2
To the two levels  + and  − (532) correspond respectively the two eigen-
functions + and − :
1 1
± = √ 1 ∓ √ 2 (536)
2 2
After having used this procedure for all unperturbed degenerate levels, we
can use the (521) in order to have more accurate level evaluation. The
above procedure succeeds, that is we get two different solutions, when the
perturbation breaks the symmetry of the unperturbed system.
8.3 Dirac delta function

The Dirac delta function () is defined by the following integral properties:
Z
+∞
 ()( − ) =  () (537)

−∞
Z
+∞
() = 1 (538)
−∞
where () is any normal function. Strictly speaking no function with these
properties can exist and () should be better called a distribution, a new
mathematical object clarified by the French mathematician Schwartz in the
1950’s. In physical applications the above properties can be exhibited up to
a good approximation by series of functions depending on a parameter. For
example the Gaussian functions
2
− 22
∆(|) = √ (539)
 2
c
satisfy exactly the second condition for any  and very well the first in the
limit  → 0. In fact
Z
+∞ Z
+∞
lim  ()∆( − |) ≈ () lim ∆( − |) =  () (540)

→0 →0
−∞ −∞
In quantum mechanics (scattering theory and dynamic perturbation theory)

one meets integrals of the type
Z
+∞
  (541)
−∞
where both  and  are real. We can write

Z
+∞ Z
+ µ ¶
   − − sin()
  = lim   = lim = lim 2 = 2()
→∞ →∞  →∞ 
−∞ −
(542)
In fact for very big  values the function 2 sin()
has a very sharp maximum

(equal to 2) for  = 0 and becomes null about the origin at  = ±.
Then for larger || values it has symmetric vanishingly small damped oscil-
lations. Thus in the limit of big  values we can approximate the plot of the
function by only the (extremely narrow) isosceles triangle with huge height
2 and vanishing basis 2This narrow triangular peak has area 2 and
the function (in the limit  → ∞) behaves similarly to 2∆(|) in the limit
 → 0. In 3D we have
Z
(r)(r − a)r = (a) (543)
with (r − a) = ( −  )( −  )( −  ). Even though this last nota-
tion is not correct (as the product of distributions is not defined), it has an
operational meaning thinking of each unidimensional delta function as the
limiting value of a series of functions such as ∆(|). Thus we can write
Z
1
3 k·r r =(k) (544)
(2)
Another useful representation is
1 
() = lim (545)
→0  2 + 2
c
Often the property

X ( −  )
 [()] = ¯¡ ¢ ¯ (546)
¯  ¯
 ¯  = ¯

is used, where the  are the roots of equation () = 0.
8.4 Green functions

Referring to the time independent Schroedinger equation:
³ ´
b −  =0 (547)
0
the solution (r r ) of the following equation is called Green function
³ ´
b 0
 −  (r r |) = (r − r )
0
(548)
0
We obtain the (r r |) as a function of eigenvalues  and eigenfunctions
b (spectral representation). Since the  (r) are a set of complete
 (r) of 
and orthonormal functions we can write:
0
X 0
(r r |) =  (r |) (r) (549)

and thus: ³ ´X X
b − 0
 (r |) (r) =
0
∗ (r ) (r) (550)
 
8
. By multiplying both sides of equation by ∗ (r) and integrating over the
whole space we get:
0 0
 (r |) ( − ) = ∗ (r ) (551)
and then
0
X ∗ (r0 ) (r)
 
(r r |) = (552)

 − 
The local density of states (r|E) is defined as:
X
(r|E) = | (r)|2 ( −  ) (553)

8
P 0 0 0
Condition  ∗ (r ) (r) =(r − r ) can be written expanding the (r − r ) as a
series of eigenfunctions
R and determining the coefficients using the orthonormality and the
0 0
definition of :  (r)(r − r )r = (r ).
c
Using the following identity (see Kramers and Kronig relations):

µ ¶
1 1
= () +   (554)
 −  
where  is infinitesimal, and introducing the complex energy  →  − , we
then get:
1
(r|E) = lim Im (r r| − ) (555)
→0 
and the density of states ():
Z
1
(E) = lim Im (r r| − )r (556)
→0 
R 0
(r r|)r is called trace of the Green function (r r |) considered as
a
¡ matrix whose elements are labelled with the pair of continuous indexes
0¢
r r .
8.5 Selfconsistent mean field

The band structure of electronic levels in a crystal has been considered within
the independent electron approximation in a periodic potential equal for any
electron. We thus assumed that the quantum motion of each electron was
described by a single-electron wavefunction  (r  s ) depending only on the
coordinates and the spin of the th electron. Here  represents the set of
quantum numbers defining the stationary Bloch state of the th electron: the
wavevector k and the branch index  (which depends on the non transla-
tional symmetry). Generally the periodic potential is not known a priori (as
instead we have always assumed) and should be determined together with
functions  (r  s ). Historically the problem was solved for the first time
by Hartree with the heuristic solution we are going to explain, then it was
considered by Fock in a variational scheme which included the exchange sym-
metry of the many body wavefunction (r1  s1  r2  s2   r  s )9 in a system
with  interacting electrons, leading to the Hartree-Fock mean field. We first
outline the second more rigorous method. In order to simplify the treatment
we consider a many electrons atom in which the nuclear potential (attractive
for electrons) exhibits a spherical symmetry. So the Hamiltonian operator
describing the motion of  electrons ( = 1 2  ) in the atom is:
2 X X 2
b =−~ 1X 2
 ∇2r − + (557)
2  
40 |r | 2 6=
40 |r − r |
9
With the notation used in section 4 it is the   (r|0), where  corresponds to the
ground state with minimum energy.
c
and can be rewritten as a sum of single particle Hamiltonians plus an inter-

action Hamiltonian: X X
b= b + 1
  (558)

2 6=
2 2
b = − ~ ∇2r − 
 (559)
2  40 |r |
2
 =  (560)
40 |r − r |
Instead, to study the crystal case, we should consider only valence electrons
in a periodic potential, as stated in section 4. In the Hamiltonian (557) we
neglected the spin-orbit interactions, the spin-spin interactions and all other
relativistic effects.
The expectation value of the electrons total energy should be calculated
as:
Z
b = ∗ (r1  s1  r2  s2   r  s )(r
 || b 1  s1  r2  s2   r  s )Π r
(561)
The Hartree equations (see below) may be derived assuming that:
(r1  s1  r2  s2   r  s ) = Π  (r  s ) (562)
from the following variational condition:
Ã !
X
b   (r  s )  −
  Π  (r  s )||Π    (r  s )| (r  s )  = 0


(563)
where the Lagrange multipliers  are needed to define the necessary condition
for minimum of the functional  || b  conditioned by the  normaliza-
XR
tion conditions   (r  s )| (r  s ) = ∗ (r  s ) (r  s )r = 1.
s
We look for the specific functional form of each  (r  s ) which minimizes
the energy of the electron system. It is, however, necessary to notice that
the form (562) of the many body wave function does not have the necessary
exchange symmetry. Fock solved this problem using a Slater determinant
wavefunction
¯ ¯
¯  (r1  s1 )  (r2  s2 )   (r  s ) ¯
¯ 1 1 1 ¯
1 ¯¯ 2 (r1  s1 ) 2 (r2  s2 )  2 (r  s ) ¯¯
(r1  s1  r2  s2   r  s ) = √ ¯ ¯
! ¯     ¯
¯  (r1  s1 )  (r2  s2 )   (r  s ) ¯
  
(564)
c
in the same variational scheme considering, always by means of Lagrange

multipliers, the orthonormality conditions   (r s )| (r s ) =   . The
Slater wavefunction guarantees the exchange antisymmetry (if two columns
are exchanged the determinant changes its sign) and the Pauli exclusion prin-
ciple (if two electronic states are identical, there are two identical rows and
then the determinant is zero). Now instead we will deduce Hartree equations
following an heuristic method based on a electrostatic statistical approxima-
tion for the mean field. If the many body wave function is the product (562),
thus implying the statistical independence of the single electronic motions,
we can assume that the total charge density at position r, due to all the
electrons except , is:
X ¯¯ ¯2
¯
 (r) = − ¯ (r)¯ (565)
6=
for the sake of simplicity we have not considered the spin. This mean density
distribution creates a mean electrostatic potential  (r) obeying the Poisson
equation:
 (r)
∇2  (r) = − (566)
◦
The solution of this equation can be written as the Coulomb integral:
¯ ¯2
Z ¡ 0¢
X Z ¯¯ (r0 )¯¯
1  r 0   0
 (r) = 0 r = − 0 r (567)
4◦ |r − r | 4◦ 6= |r − r |
If we now consider electron , its wave function satisfies the Schroedinger

equation: ∙ 2 2 ¸
~∇ 2
− − −  (r)  (r) =   (r) (568)
2 4◦ |r|
We added together the Coulomb attractive potential energy of nucleus and
the mean repulsive Coulomb potential energy  (r) = −  (r) which takes
into account the screening of the nuclear charge due to all other electrons
with respect to th electron. Actually this situation can be considered as
2
that of a single attractive force field − 4(r)|r|2 u due to the interaction with
the pointlike charge  in a medium where the dielectric function depends

on position as:
◦
(r) = 4◦ (569)
1 +  |r|  (r)
Since  (r) 0, (r)  ◦ . Using the explicit expression of  (r) we get the
c
system of Hartree equations:

⎡ ¯ ¯2 ⎤
Z ¯ 0 ¯
⎢ ~2 ∇
2
2 2 X ¯ (r )¯ 0⎥
⎣− − + 0 r ⎦  (r) =   (r) (570)
2 4◦ |r| 4◦ 6=
|r − r |
This nonlinear system of integro-differential equations can be solved in an it-

(0)
erative way, starting from the hydrogen-like expressions  (r) for the  (r)
in the square bracket corresponding to  (r) =0 At any iteration () we get
(+1)
a better approximation  (r) for the  (r) and  (+1) (r) for the mean
field  (r) through equation (567). Once the convergence is obtained, we
have the selfconsistent mean field and the associated single-electron wave
functions which minimize the system total energy.
8.6 Semiclassical dynamics in a conservative force field

In a conservative force field weakly depending on position through the po-
tential energy (r), an electron with large momentum behaves almost like
a particle obeying Newtonian classical mechanics. In one-dimension, the
Hamiltonian operator is:
2
b = b + ()
 (571)
2
We use the equation of the time derivative for the expectation value of a
physical quantity  =  ( )
* + Z ( )
 h i  b 1 h b bi  b
 1 h i
= +  = ∗ + b 
b  (572)
  ~  ~
and we apply this equation to both position and momentum:

 hi 1 Dh b iE hi
= b  =
 (573)
 ~ 
¿ À
 hi 1 Dh b iE 
= b  = − (574)
 ~ 
Differentiating the first equation with respect to time and substituting the
expression of the derivative of the momentum into the second equation, we
get: ¿ À
2 hi 
 =− (575)
2 
If the dynamics was be completely
 ® ¡ classical
¢ at any instant of time
¡ we¢ would
have  = hi and −  = −  =hi =  ().  () = −  is the
c
classical force acting on the particle. Yet in quantum mechanics position and
momentum obey Heisenberg uncertainty principle:
2® 2® ~2
( − hi) ( − hi) ≥ (576)
4
In the hypotesis that q( ) is a wave packet with a small (but not zero)
®
uncertainty of position ( − hi)2 about the mean value hi we can derive
®
an approximate  
expanding 
in a power series of ( − hi) about hi
up to second order :
¿ À Z Z (µ ¶ µ 3 ¶ )
   1  
= ∗ ( ) ( ) ≈ + ( − hi)2 |( )|2 
   =hi 2 3 =hi
³ 2 ´ (577)
  2
We neglected 2 ( − hi) |( )| because its integral is identically
=hi
zero. So the previous equation can be written as:
¿ À µ ¶ µ ¶
  1 3 2®
≈ + ( − hi) (578)
  =hi 2 3 =hi
If the following inequality is satisfied

¯µ ¶ ¯ ¯µ ¶ ¯
¯
1¯   3 ¯ ® ¯ ¯
¯ 2 ¯  ¯
¯ ¯ ( − hi) ¿ ¯ ¯ (579)
2 ¯ 3
 =hi ¯ ¯  =hi ¯
®
− 

is very similar to the classical force  (). We write again the latter
condition as: ¯¡ ¢ ¯
¯  ¯
2 ¯  =hi ¯
2®
( − hi) ¿ ¯¯¡ 3 ¢ ¯
¯ (580)
 
¯ 3 =hi ¯
Given the uncertainty of the position, respecting the above inequality is easier
if () depends weakly on . Now, using the uncertainty principle, we can
write:
® ~2
( − hi)2 ≥ ® (581)
4 ( − hi)2
But the mean kinetic energy is:
®
h2 i hi2 ( − hi)2
= + (582)
2 2 2
c
Figure 5:
c
For kinetic energy the condition to heve semiclassical motion is:

®
hi2 À ( − hi)2 (583)
® ®
The figure shows in which regions of the ( − hi)2  ( − hi)2 plane
the conditions for a semiclassical motion are verified. The region below the
hyperbola is forbidden since there Heisenberg uncertainty principle would be
violated; this zone is labelled as unphysical (UP UnPhysical). In the region
above the hyperbola (QM Quantum Mechanics) there is a curvilinear triangle
inside which the semiclassical conditions are verified. This area is wider if
the potential energy depends weakly on position and in case of large mean
linear momentum (in this case the de Broglie wavelength associated with the
electron is very small, having introduced a mean de Broglie wavelength  =
 hi associated to the wave packet). In general this graphical representation
could change with time. The following case of a simple periodic potential is
particularly simple and significative:
2
 () = 0 sin( ) (584)

Here, considering both semiclassical conditions and uncertainty principle, we
obtain at the end the unique condition  ¿  which guarantees that the
wave packet representing the electron does not exhibit significant diffraction.
In a crystal, let  be the lattice parameter, this condition is never verified,
since it would imply a wave vector  À 2 greater than the edge of the
first Brillouin zone. Yet the relevance of semiclassical dynamics applyed to
crystals is well known. This apparent contradiction is solved considering
the effective mass theorem: when we consider the equivalent semiclassical
dynamics in a crystal,  () does not include the periodic potential (which
is lumped in the electron effective mass) but, instead, just represents the
external force field  (generally not periodic in space) which weakly depends
on position. Unfortunately the considerations shown in this paragraph are
not completely general since they require that () is a continuous and
differentiable function up, at least, to third derivative and that none among
the derivatives is identically zero. In conclusion we cannot apply this method
to an harmonic oscillator where () = 12 2 2 .
9 Appendix: Elasticity and Elastic Waves

9.1 Stresses, strains and elastic constants
As for all other physical properties, the mechanical response of materials de-
pends on the scale at which it is measured. For 3D compact materials (e.g. di-
c
amond), when the length scale is much bigger than the interatomic distance,
the mechanical behaviour is conveniently modelled by standard continuum
mechanics. Yet there exist special materials with a complex mesostructure
for which this approach may not be appropriate. Aerogels and some types
of cluster assembled carbon films belong to this cathegory. In the follow-
ing we shall concentrate on elasticity, neglecting more complex phenomena
like plasticity and fracture. Engineers usually deal with polycrystalline or
multi-phase solids which can be characterized by macroscopic elastic para-
meters such as the bulk modulus  and the shear modulus . Such materials
are elastically isotropic at the macro-scale and two parameters are sufficient
to describe their behaviour. Physicists are traditionally more interested to
anisotropic single crystals, like perfect diamond (cubic) and perfect graphite
(hexagonal). The scalar variables volume and pressure are not adequate to
define the deformed state of solids: tensor quantities are needed. To describe
the kinematics of the deformation process, let a vector field r spans all points
within the material volume  in an undeformed equilibrium state. If now a
material deformation is produced by some external and/or internal agents,
the old (undeformed) material positions are mapped into the new ones as
r0 = r + u(r) by the displacement vector field u(r), describing both deforma-
tions (volume and shape variations) and rigid rotations. In the small strain
regime, the symmetric part of the gradient of the displacement field u(r) is
the strain tensor 
µ ¶
1  
 =  (∇u) = + (585)
2  
This tensor describes how infinitesimal cubic volume elements  = 
change their volume and/or their shape. The diagonal components represent
0
tensile strains (e.g.,  = ( − )), while the off-diagonal components
represent shear strains. More precisely, assuming in the following summation
from 1 to 3 over repeated vector and tensor suffixes, the relative volume
0
variation ( −  ) is equal to  , the trace of  .  is either a
pure dilation (  0) or a pure hydrostatic compression (  0). Instead,
 =  − 13   is a pure shear strain (a strain deviator, where   is the
unit tensor): in fact  = 0, proving  is a pure shear. In the undeformed
state the angle, e.g.,  between x and y is 90 . After a shear deformation
the angle is reduced of  =   = 2 = 2 . Any deformation is the
sum of a pure shear and a dilation (compression) as it is evident from the
identity  =  + 13    .
Provided the produced strains are small and reversible, a generalized
Hooke’s law describes the linear elastic material response:
c
 =   (586)

In the above equation   is the stress tensor.   is the −component of
the force per unit area acting on a face of a cubic element whose outgoing
normal is the unit vector of the −th cartesian axis. The diagonal compo-
nents represent tractions, while the off-diagonal components represent shear
stresses. The elastic material properties are embodied in the fourth-rank ten-
sor   the elastic constant tensor with 34 = 81 elements. Because of the
symmetry of   (see below) and  , the number of independent components
of  diminishes to 36. This number is further lowered to 21 by energetic
considerations. Using then Voigt’s contraction scheme
  or   11 22 33 23 or 32 13 or 31 12 or 21
 or  1 2 3 4 5 6
eq. 586 can be given a simpler matrix form
 =   (587)
where the elastic constant 6 × 6 matrix  is symmetric.

The material symmetry produces further reduction. In the case of cu-
bic crystals, for instance, the independent constants are 3 and only 2 in a
isotropic material, as anticipated. In the simplest case of isotropic elasticity
eq. 587 becomes:
⎛ ⎞ ⎛ ⎞⎛ ⎞
1  + 43  − 23  − 23 0 0 0 1
⎜ 2 ⎟ ⎜  − 23  + 43  − 23 0 0 0 ⎟ ⎜ 2 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟
⎜ 3 ⎟ ⎜  − 23  − 23  + 43 0 0 0 ⎟ ⎜ 3 ⎟
⎜ ⎟=⎜ ⎟⎜ ⎟
⎜ 4 ⎟ ⎜ 0 0 0 2 0 0 ⎟ ⎜ 4 ⎟
⎜ ⎟ ⎜ ⎟⎜ ⎟
⎝ 5 ⎠ ⎝ 0 0 0 0 2 0 ⎠ ⎝ 5 ⎠
6 0 0 0 0 0 2 6
(588)
Uniaxial stress. In engineering two different elastic constants, namely the
Young modulus  and the Poisson’s ratio  are usually introduced. If a
uniaxial stress  =  1 =  is applied (with all other   = 0) only strains
 = 1 =  ,  = 2 and  = 3 = 2 are non-zero. Then  is defined as
 and  as −2  = −3 . Since there are only two independent elastic
c
constants,  and  can be obtained from  and  by the following formulae:

9
 = (589)
3 + 
3 − 2
 = (590)
2(3 + )
The meaning of bulk modulus  appears clearly in hydrostatic pressure

conditions:  1 =  2 =  3 = − and  4 =  5 =  6 = 0. In this case
1 = 2 = 3 = −3 and 4 = 5 = 6 = 0. Thus the relative volume
variation  = 1 + 2 + 3 = −. From the last equation it follows
that the isothermal compressibility  , whose positiveness is a condition for
material thermodynamic stability, is just the inverse of 
µ ¶
−1 1 
 =  = − 0 (591)
  
Since  too is always positive (see eq. 592),  can vary between −1 (when
 = 0) and 12 (when  = 0). In practice there are no materials known for
which   0.
Shear: In the case of simple shear, e.g. 4 =  and all other   = 0, then
4 =  2 and all other  = 0 If one introduces the angular shear strain
measure  4 = 24 = , the last equation reads  =  , which illustrates the
physical meaning of .
Cubic lattices have three independent elastic parameters 11 , 12 and 44 :
in this case (as for lower symmetry crystals)  and  are direction dependent
and  can exceed the limits of isotropic materials. As the cubic case reduces
to the isotropic case when 11 − 12 = 244 , a possible measure of the degree
of anisotropy is (11 − 12 − 244 )  (11 − 44 ). Hexagonal lattices have five
independent elastic parameters 11 , 12  13 , 33 and 44 .
solid 11 12 13 33 44 

silicon (cubic) 165.7 63.9 79.6 2.33
diamond (cubic) 1076.0 125.0 575.8 3.51
graphite (hexagonal) 1060.0 180.0 15.0 36.5 4.5 2.26
The elastic moduli are in   and the density  in 3
c
Graphite is globally highly anisotropic but completely isotropic for any

rotation around the c axis perpendicular to a graphene plane. Its extremely
low 44 value explains its use as a solid lubricant material. Instead the in-
plane moduli are of the same order of those of diamond, the hardest known
material.
The elastic energy density E of a strained isotropic material is the quadratic
form:
1
E =  ( )2 +   (592)
2
In this formula spherical symmetry (isotropy) has been fully taken into ac-
count. In fact  and   are invariants of the strain tensor with respect
to rotations. The limitation to second order in the strains leads to linear
elasticity (  = E ), as stated by eq. 588), with both  and  positive
for E must be minimum in the undeformed state (stability).
In mechanical equilibrium the internal stresses in every volume element
must balance. This is realized when the stress tensor field obeys the following
equations
 
+ = = 0 and   =   (593)

where = is an external body force. If external surface forces  act on the
points of the outer surface  bounding the volume  , also equilibrium bound-
ary conditions must be taken into account. They read
(   )r∈ =  (594)
where n is the outgoing normal of the surface element  centered at r ∈ .
The boundary conditions will remain the same even in the dynamic case.
9.2 The acoustic waves and their phonons

Though the traditional ways to measure the mechanical properties of ma-
terials and, in particular, the elastic constants are based on quasi-static de-
formation processes (e.g. the tensile test to measure the Young modulus),
many important methods are acoustic in nature or make use of acoustic
phenomena (ultrasound propagation, acoustic microscopy, acoustic emission,
Brillouin scattering, laser induced surface acoustic waves). In the simplest
case one measures the time required for a longitudinal ultrasonic pulse to
travel back and forth inside a cylindrical sample along its axis. Knowing
the length of the cylinder theppulse velocity  is obtained. If the cylinder
is a long very thin rod,  = ,  being the rod density. To face less
naive methods and more complex geometries, some basic results of classical
elastodynamics must be employed.
c
The elastodynamic equation substituting eq. 593, written in terms of the

dynamic displacement field u = u(r ), reads:
 2u
2
= 2 ∇2 u + (2 − 2 )∇(∇ · u) (595)

q¡ ¢ p
where  =  + 43   is the longitudinal sound velocity and  = 
is the transverse sound velocity. Remembering that the most general defor-
mation process is the superposition of a simple dilation and of a simple shear,
one is tempted to try a solution of the type u = u + u with ∇ × u = 0
and ∇ · u = 0. The second condition is identical to that holding for electro-
magnetic waves in vacuum. This trick perfectly works (instead anisotropic
materials require a more complex treatment) leading to two decoupled wave
equations for u and u :
 2 u 2 2  2 u
2
=  ∇ u and 2
= 2 ∇2 u (596)
 
The fundamental bulk solution of eq. 595 is then the superposition of three
independent monochromatic plane waves, one longitudinal () and two
(mutually perpendicular) transverse ( 1   2 ), of the type:
© ª
u = < q eq [q·r− (q))] (597)
where q is the wavevector,  is a branch index ( =  1  2 ), q is the

complex amplitude of the normal coordinate   (q) = q − and eq
a polarization unit vector (eq ||q; eq ⊥q). Moreover: ω  (q) =  |q| and
ω1 2 (q) =  |q|. The above classical description can be translated into the
language of quantum mechanics of the systems of independent harmonic os-
cillators: the quanta of the fields u are the long-wavelength acoustic phonons
whose possible energies are:
µ ¶
1
 (qα) = q + ~ω (q) (598)
2

Solid State Physics Lecture Notes

Uploaded by

Copyright:

Available Formats

Solid State Physics Lecture Notes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Solid State Physics Lecture Notes

Uploaded by

Copyright:

Available Formats

Lecture Notes of

Solid State Physics

a.y. 2017/18 - February 2018

4 Adiabatic theorem and vibrational motions 64

7 Transport phenomena 112

8 Appendix: Quantum approximate methods 124

9 Appendix: Elasticity and Elastic Waves 142

1.1 Order and symmetry

1.1.1 Simple Crystals: lattices and translational order

A lattice translation Tnn0 is the vector diﬀerence between two vectors n

whose Fourier coeﬃcients are computed as

then one writes

( ) = |a1 · a2 × a3 | (13)

In this way it is always true that a · b =   and the (8) is authomatically

1. Every reciprocal lattice vector g is perpendicular to a family of (di-

2. If the components of g have no common factors, then in general the

Besides from its translational symmetry each Bravais lattice is charac-

1.1.2 Complex crystals: lattice and basis

In the most general case a crystal can be conceptually built decorating

1.2 Crystal binding

the weak interaction between instantaneous atomic dipoles, fluctuating with

1.3 Tensorial observables

In this formula, as in all tensorial equations, the sumPfrom 1 to 3 over all

tensor is diagonal:  =   . In this case the nine components of χ collapse

If the considered rotation is a symmetry operation of the crystal (that is,

information is extractable from diﬀraction of the wave associated with the

2.1 Elementary theory of elastic scattering

the angle  is not too small, in the simplest

which is a function of the scattering angle as |Q| = 2 sin(2)In

2.2 Quantum particle scattering amplitude: Born ap-

 (r) =k ·r (23)

The plane wave satisfies the homogeneous equation (empty lattice):

is associated to the scattered particle beam. The integral equation can be

Within Born approximation potential energy is considered as a perturbation

2.3 Elastic scattering - Bragg law

g being the reciprocal lattice vectors, with Fourier coeﬃcients

Then the scattering amplitude can be written as:

to all considered planes. Let us consider the necessary condition to have

|g | − 2 |k | sin  = 0 (46)

2 sin  =  (47)

considering that (see Appendix)

The form sin()

2.4 X-Ray Diﬀraction

reconducted to the vector Helmholtz equation7

the Helmholtz equation is obtained.

once that the diﬀraction condition is set Q = g .

where  depends on the nature of the pth atom.

Monoatomic lattice with a basis:

g = (2)(x + y + z)

g = (2)(x + y + z)

 = 1 + exp[−g · ((2)(x + y))] + exp[−g · ((2)(x + z))] +

g = (b1 + b2 + b3 )

Von Laue formulation.

 (R) is called static autocorrelation function or Patterson function.

The formula for neutrons refers to a monoatomic material, with  equal

Liquids (neutron scattering):

core electrons can be considered strictly bounded to them) provides just a

is the Coulomb potential energy associated with repulsive interactions be-

r = (r1 s1  r2 s2   r s   r s )

atoms (see appendix), we simplify the previous many-body problem to an ef-