Energy and Band Theory

Electronic Engineering Materials & Nanotechnology Chapter 5
ENERGY BAND THEORY

&
Classification of Solids
Contents:
5-1. Introduction
5-1.1. Energy Levels & Energy Bands
5-1.2. Reference Energy Levels (Vacuum level & Fermi level)
5-2. Crystal Hamiltonian Operator
5-2.1. Adiabatic Approximation
5-2.2. Jellium Model
5-2.3. Hartree Equation (One-electron Model)
5-2.4. Hartree-Fock Equation
5-2.5. Self-consistent Field Method
5-2.6. Post Hartee-Fock Methods
5-3. Bloch Functions
5-4. Kronig-Penney Model
5-5. Effect of Crystal Size & Fine Structure on Energy Bands
5-6. Analogy with Circuit Theory
5-7. Energy Band Structure Calculation Methods
5-7.1 Energy Band Structure (Ab initio Methods)
i. FLAPW method
ii. Density Functional (DFT) method
iii,. Local Density Approximation (LDA)
5-7.2 Energy Band Structure (Approximate Methods)
i. Free electron gas (Fermi Gas)
. ii. Empty lattice model
iv. Quasi-free electron (QFE) model
v. Pseudo-potential method
. vi. Tightly-bound electron (TBE) model
-303-
Prof. Dr. Muhammad EL-SABA
Contents of Chapter 5 (cont.):
5-8. Electron Dynamics & Effective Mass

5-8.1. 1-dimensional Model
5-8.2. 3-dimensional Model (Tensor mass)
5-9. Actual Band Structure and Fermi Surface in 3-D crystals
5-10. Case Studies: Band Structure of Si, Strained Si, GaAs
5-10.1. Band Structure of Si
5-10.2. Band Structure of Strained Si
i. Effect of Biaxial Stress
ii. Effect of Uniaxial Stress
5-10.3. Band Structure of SiC
5-10.4. Band Structure of GaAs
5-11. Hole Concept
5-12. Density of States in Energy Bands
5-13. Classification of Solid Materials
5-14. Classification of Semiconductors
5-15. Direct- and Indirect-gap Semiconductors
5-16. Energy Band Alignment in Heterojunctions
5-17. Graded Bandgap Semiconductors
5-18. Superlattices & Bloch Oscillations
5-19. Low-Dimensional Semiconductors (LDS’s)
5-20. Energy Band Structure of Graphene
5-21. Energy Band Structure of Carbon Nanotubes (CNT’s)
5-22. Energy Band Structure in Silicon NanoWires
5-23. Organic Semiconductors
5-24. Energy-band Structure of Amorphous Semiconductors
5-25. Topological Insulators
5-26. Effect of Magnetic Fields on Energy Bands Structure
5-27. Effect of Magnetic Field on Energy Bands in LDS’s
5-28. Effect of Magnetic Field on Energy Bands in Nanotubes
5-29. Summary
5-30. Problems
5-31. Bibliography
-304-
Energy Band
Theory &
Classification of Solids
5-1. Introduction
As we have seen earlier, the solution of Schrödinger equation for a
particular system results in a set of wave eigenfunctions n and energy
eigenvalues En. The relation between the electron energy and the electron
wave-vector, E(k), which we call dispersion relation of the system, is
obtained by substituting the solution of  into the Schrödinger equation.
In isolated atoms, the electrons are orbiting around their nuclei with certain
energy levels. In this case, the E(k) relation consists of a set of discrete
(quantized) points in the k space. When isolated atoms are brought
together, to form a crystalline solid, various interactions (due to attraction
and repulsion forces) occur between neighboring atoms. At the position of
equilibrium, the forces of attraction and repulsion between atoms are
balanced and the total potential energy of the system becomes minimal.
5-1.1. Energy Levels and Energy Bands

As the electron clouds of individual atoms in a solid are overlapped, the
height of potential wells of different atoms is lowered and electrons of
higher energies are no longer belonging to a specific atom. In contrast, the
electrons that are found in the deep energy levels are somewhat tightly
bound to their nuclei. Figure 5-1 depicts the splitting of energy levels in an
array of diamond atoms. As a consequence of the overlapping process, the
electron energy levels are split into dense groups of levels, known as
energy bands. In fact the exclusion principle dictates that no two electrons
in a given interacting system may have the same quantum state. Therefore,
atoms of identical quantum states modify their energy states such that the
Pauli Exclusion Principle is satisfied. In fact, the original discrete energy
levels of such electrons are split into dense groups of adjacent levels, which
we called energy bands. These dense groups of levels (or energy bands) are
belonging to the crystal rather than to individual atoms. Usually, these
energy bands are separated by successive forbidden regions, that we call
energy gaps. Therefore, the starting point to understand the energy band
-305-
structure of solids is the potential energy diagram of an electron in an

isolated neutral atom and the corresponding energy levels.
Fig. 5-1. Energy levels splitting due to the interaction of adjacent atoms and
overlapping of their probability density functions p(r).
Energy Energy
2p2 Ec
2s2 Ev
2
1s
1s2
R
Nucleus Ro Ro
(a)Isolated diamond atom. (b) Array of diamond atoms.
Fig. 5-2. Energy levels in an isolated diamond atom and in an array of diamond atoms.
-306-
Figure 5-3 shows the development of the electron energy bands in diamond
crystals from the discrete energy levels of an isolated atom (according to
the conventional theory). As shown in figure, when the separation between
atoms is decreased, the energy levels of single atoms (1s, 2s, etc.) are split
into bands of a huge number of adjacent energy levels.
Fig. 5-3. Schematic of energy level splitting and formation of energy bands in a
diamond crystal of N atoms as a function of interatomic distance R. The distance Ro is
the interatomic separation at the state of equilibrium between atomic forces.
Figure 5-4, depicts the energy band diagram at the state of equilibrium,
where the interatomic distance R = Ro. The highest filled energy band, in
the energy band diagram, is usually called: valence band. Also, the energy
band just above the valence band is called the conduction band. The
valence and conduction band are separated by a region called the energy
gap. The height of energy gap is given by:
Eg = Ec - Ev (5-1)
where Ec is the lowest (bottom) level in the conduction band and Ev is the
highest (top) level in the valence band as shown in figure 5-4. Thus, in
crystalline solids, the E-k relation, at a certain point in the physical crystal
lattice, is usually characterized by a sequence of alternating allowed energy
bands and energy gaps. For three-dimensional crystals, the k-space is three-
dimensional. The E-k relation of electrons inside a solid crystal, in the
-307-
various directions of the k-space, is usually called the energy band

structure of that solid.
E
Energy gap
Conduction Band
Ec
Eg
Ev Band of allowed
Valence Band energy levels
V.B.
Fig. 5-4. Schematic of the energy band diagram (versus crystal spatial position).
By studying the energy band structure of electrons inside crystalline solids,

one can understand their electrical properties. For instant, the number of
electrons in the conduction band, the electrons average velocity and the
electron current in a certain crystalline solid can be all calculated if one
knows the E-k relationship (or the energy band structure) of that solid.
Therefore, our fundamental goal is to determine the electrical properties of
semiconductor materials, which are necessary to develop the current-
voltage characteristics of semiconductor devices.
5-1.2. Reference Energy Levels (Vacuum level and Fermi level)

Before proceeding, in the mathematical details of energy bands in solids, it
is wise to settle down some reference energy levels from which we
measure the other energy levels and the whole band structure in solids. We
start by defining the vacuum energy level as the potential energy of an
isolated electron in vacuum at rest. In crystals, electrons are arranged in
energy bands, which are separated by energy gaps. We have seen in
Chapter 3, the arrangement of electron energies follows a probability
distribution function. At equilibrium the distribution function of electron
energies is called the Fermi-Dirac distribution.
According to the Fermi-Dirac distribution, fn(E), all energy levels under a

certain reference energy level (called the Fermi level and denoted by EF)
are occupied at absolute zero temperature. Therefore, fn(E) =1 when E <EF
and fn(E) =0 for E>EF at T=0. That’s electrons fill the energy bands of
-308-
solids, at very low temperatures, up to the Fermi energy level EF, and the
bands are empty for energies that exceed EF. Figure 5-5 depicts the location
of Fermi energy level, EF, with respect to the reference vacuum energy
level, Eo, in different solids. The difference between the vacuum level and
the Fermi level in metals is usually termed the work function (M = Eo - EF).
Also, the difference between the vacuum level and the conduction band
edge in semiconductors and insulators is usually termed the electron
affinity ( = Eo - EF).
Eo Vacuum level Eo

Ec
m s
EF EF
Metal Vacuum Semiconductor Vacuum
or Insulator
Fig. 5-5. Vacuum energy level at metal and semiconductor/insulator–vacuum interface.
Table 5-1: Work function of some metals
Table 5-2: Electron affinity of some semiconductors
-309-
5-2. Crystal Hamiltonian Operator

We know that the one particle Schrödinger equation governs the time
evolution of a complex-valued wavefunction  on the configuration space
( ). The equation represents a quantized version of the total energy of a
classical system evolving under a real-valued potential function V on :

E  =Η  (5-2)
or
For many particles, the equation is the same except that the Hamiltonian

Η as well as and are now on configuration space, ,where N is the
number of particles
The large of particles involved in a typical solid makes direct solution of

the Schrödinger equation completely impossible. Therefore, we are
naturally led to the development of a one-electron approximation. It is
however, necessary to relate the one-electron theory to the actual many-
electron situation, and to estimate the corrections to the results of the one-
electron approximation. In this section we demonstrate how such a one-
electron approximation is developed from the first principles. We start by
defining the Hamiltonian for the entire problem of solid state physics. It
consists of the kinetic energy of all particles in the solid and of their
interaction energies.
The solid is comprised of two groups of electrons: valence electrons, which

contribute to chemical bonding and core electrons which, are tightly bound
in the closed shells of the lattice ions. We usually consider the valence
electrons and the lattice ions as independent constituents of the solid. This
is the first approximation, we usually make in solid state physics. For
simplicity, we shall call the valence electrons, the electrons. The
Hamiltonian then consists of the kinetic energy of these valence electrons
and all ions as well as their interaction energies.

Η = Hel + Hion + He-i + Hext (5-3a)
where we omitted the operator cap sign for the sake of simplicity. For a
non-perturbed system we have Hext = 0. For electrons we have:
-310-
Hel = He (K.E.)+ He-e (electron-electron interaction)

pk2 1 e2
  +  (5-3b)
k 2m 4  o k, k ` rk - rk `
where rk and rk` are the coordinates of the kth and k`th electrons, respectively.
Similarly, for ions, we have:
Hion = Hi (K.E.)+ Hi-i(ion-ion interaction)
+ 1  Vi Ri  R j 
Pi 2

(5-3c)
i 2M i 2 i,j
where Ri and Rj are the coordinates of the ith and jth ions, respectively.
Concerning the electron-ion interaction energy, it may be written as
follows:
H e-i =  Ve-i rk  Ri  (5-3d)

k,i
The ion-ion interaction term Hi-i may be subdivided into two terms, one
accounts to the interaction when the ions are in their equilibrium positions;
the other is a correction to account for the vibrations of the lattice ions
(phonons).
H i -i = H o i -i + H ph , H e-i = H o e-i + H e- ph (5-3e)
Unfortunately it is not always possible to rigorously solve the quantum

mechanical problem. Approximations have to be made. In solid state
theory, two simplifications are commonly made. In solving a given
problem individual terms of the Hamiltonian may be neglected or only
partially considered or handled subsequently as perturbations.
This simplified problem is then further simplified by using the symmetry of

the lattice. The particular approximation needed here depends on the
questions we are seeking to answer and on the nature of the solid material
we want to describe.
-311-
5-2.1. Adiabatic Approximation

We can consider that the solid state problem can be divided into two
components:
 Movement of electrons (electron gas) in a stationary lattice, and

 Movement of ions in a uniform space charge of electrons.
This is the basis of the so-called Born-Oppenheimer (adiabatic)

approximation, which is often used to provide a justification for system
decoupling. The physical background of this approximation is discussed in
chapter 2 of this book.
5-2.2. Jellium Model

According to the adiabatic approximation, Hel can be written as follows:
2 2 1 e2
H el =  k +  + H ve (5-4)
k 2m 8 o k, k ` rk  rk `
where H+ve includes the effect of the ions positive space charge and the
interaction of the electrons with this space charge.
The electron gas is considered here to be embedded in a constant positive

charge background. This model is described as Jellium. Many of the
properties of metals can be described by mean of this approximation. On
the other hand, the movement of ions can be treated in a similar way. So, Hi
can be written as follows:
 V R  R j  + H ve
2 1
H ion=   i2 + i i (5-5)
i 2M i 2 i,j
where H-ve includes the effect of negative space charge of electron

(considered fixed) and the interaction of this charge with ions. It should be
noted that the two terms have H-ve and H+ve exactly cancel each other.
As far as the motion of electrons is concerned, we can therefore adopt a

Schrödinger equation for the electrons:
H el  H e-i   = Ee  (5-6)
-312-
Here He-i is considered with fixed coordinates of ions. The remaining

coupling between electrons and ions vibration is the interaction He-ph. This
coupling can be dealt with in most cases by the perturbation theory. This
theory will be described in a following chapter.
5-2.3. Hartree Equation (One-Electron Approximation)

In the absence of electron-electron interactions, the electron gas problem
(Jellium model) would decouple into one-electron problem, which
describes the motion of an electron in a given potential field. In this case,
the Hamiltonian term which accounts for the electron-electron interaction is
neglected, so that:
 2 2
Η -  k + V rk =  H k (5-7)
k 2m k k
Then, the following Schrödinger equation:
H
k
k  = E
(5-8)
can be reduced, by substituting:
H
k
k  = E (5-9)
and
 r1 ..rN  =  i r1  2 r2 ... N rN  (5-10)
into the following one-electron equations:
Hk  k rk  = E k  k rk  (5-11)
However, the many-body problem (the electron gas which is embedded in a

homogenous, positively charged medium) could be reduced to one-body
problem (one-electron problem) which includes an energy term to account
for the electron interaction. Such a reduction is achieved by the Hartree (or
Hartree-Fock) approximation. In this approximation, the Schrödinger
equation for a single particle writes:
-313-
 2 2 

 2m   VH ( r )  k r   Ek k r  (5-12)
 
where VH is the Hartree potential, which is given by:
 j r 
2
2
e
VH (r )  V (r ) 
4 o

j k r  r
dV (5-13)
Here r and r` are the positions of the ith and kth electrons, respectively. The
above equation is called the Hartree equation. It describes the motion of
the kth electron at the position r in the potential field V(r) of the lattice ions,
and in the Coulomb potential of an average distribution of all other
electron: (j=1, 2, .., N, with j≠k)
It should be noted that the Hartree equation does not account for the Pauli
Exclusion Principle. The spin of electrons is not considered in this
equation. If two electrons are interchanged, the sign of the total wave
N
function will remain unchanged (   k1  K ). This representation is suitable
for bosons but not for Fermions (e.g. electrons) where the total wave
function  must change its sign when two particles (electrons) are
interchanged.
5-2.4. Hartree-Fock Equation

In order to account for the Pauli-exclusion principle we replace the usual
expression of the total wave function 
 = 1 r1  .... N rN  (5-14)
by the following Slater determinant, which satisfies the Pauli-exclusion

principle:
 1 q1  . .  N qi 
1 . .
= (5-15)
N! . .
 i qN  . .  N qN 
-314-
Here, the coordinates qk include the spatial coordinates rk as well as the spin
coordinates of the electrons. Also the factor in front of the determinant is
added for normalization purposes. If two electrons are interchanged, two
columns of the determinate are interchanged, and  changes sign. Also if
two electrons have the same coordinates, two columns will be identical,
and willvanish. Now the interaction term accounts for the Pauli
exclusion term and the Schrödinger equation reads:
 2 2  r     HF r , r  
  V r     k r  = Ek k r 
e
4 o 
 d V  (5-16)
 2m r-r  
where
 r     kH   - e  k r 
2
(5-17)
k k
and
 r  r r  r 
*
j k
*
k j
(5-18)
  HF
 e
j, k
  r  r  k
*
k k
The above form of Schrödinger’s equation is called the Hartree-Fock

equation. Also, the following potential is called Hartree-Fock potential.
 r`    HF r , r 
V `(r )  V r  
e
4 o  r  r`
d V (5-19)
The Hartree-Fock equation may be then written in the following simple

form:
 2 2 

 2m   V`r   k r  = Ek r  (5-20)
 
V`(r) is sometimes called the pseudo-potential.
-315-
5-2.5. Self-Consistent Field Method

The calculation of energy bands is based, in principle, on the Hartree-Fock
equations. The set of Hartree-Fock equations forms a quite complicated
nonlinear integro-differential system. In fact the solution of such equations
must be carried out using iterative numerical methods. The self-consistent
field (SCF) method results in the wave eigenfunctions and energy
eigenvalues satisfying the Hartree (or the Hartree-Fock) equation. The
solution of this equation is undertaken as follows:
1- Choose a set of eigen-functions  (k0 ) as a zero-order approximation

(ground-state functions) for k.
2- Calculate an initial estimate of the Hartree (or the Hartree-Fock)
potential, VH(0) , using equation (5-19) as follows:
e2  jo  r 
V ( 0)
r   V r  +  dV` (5-21)
4  o
H
j k r - r`
3- Substitute the initial estimate of Hartree potential VH(0) into the

Hartree equation (5-13) and solve to obtain a set of new functions  k1 ,
 2 2  1
 2m   VH r   k r  = E k  k r 
(0) 1
(5-22)
 
4- Calculate a new estimate of Hartree potential, VH(1) , using the new set
of functions  k1 , etc.
5- Proceed until you reach the allowed error in  k1 or Ek.

6- The so-obtained wave functions are called the self-consistent
wave functions.
If the Hartree-Fock potential V`(r) is taken as the total potential energy of

an electron in the crystal (due to other electrons and ions), then we can
write:
 2 2 
    V ' ( r ) . (r )  E. (r )
 2m  (5-23)
-316-
This equation is very interesting because it yields valid results without

having to determine the self-consistent functions, if we start with an
acceptable pseudo potential V(r). It can be shown that V`(r) in the crystal
lattice is a periodic function of lattice coordinates. This feature is exploited
in the so-called “Bloch functions” that we shall discuss in a next section.
Other derived and simplified methods, like the density functional theory
(DFT) or the tightly-bound electron (TBE) method, have been developed
in such the way that they either neglect or approximate some of the
integrals, which are involved in the HF and SCF methods. The approximate
methods will be discussed in section 5-7 of this chapter.
5-2.6. Post-Hartee-Fock Methods

We have seen in the above discussion that the introduction of the self-
consistent potential (Hartree or Hartree-Fock potential), enables the
problem of a system of interacting particles to be reduced to a one-particle
problem. The Hartree–Fock based models have been successfully applied
to atoms and molecules. Therefore, the Hartee-Fock (HF) is considered as
one of the ab initio computational methods in solid-state physics. The main
two problems of the technique are the neglect of electrons and ions
correlation effects and the high computational demands. In fact the
Hartree–Fock methods can only be applied to systems with small numbers
of atoms.
The so-called post-Hartree–Fock methods are a category of computational

methods, which are developed to improve the Hartree–Fock and self-
consistent field methods. They add electron correlation terms to account for
the repulsion between electrons rather than being averaged in the Hartree–
Fock method. For the great majority of systems, in particular for excited
states and processes, the correlation term is the most important. The post-
Hartree–Fock methods give more accurate results than the traditional HF
method, with the price of numerical computational costs.
The accuracy and reliability of the above methods is hard to assess. In large
scale atomic calculations where one deals with hundreds or perhaps
thousands of atoms, and where translational symmetry may be broken, new
approaches are required. In Chapter 9, we present some formulations of
quantum mechanics that significantly reduces the number of variables, but
still enables us to calculate electronic properties of electronic systems and
devices with high degrees of accuracy.
-317-
5-3. Bloch Functions

When we try to find out the wave functions for electrons in a crystalline
solid, the task may seem impossible because of the huge number of atoms
and electrons. However, the problem is greatly simplified if we consider
that the electrons of the solid belong to the crystal rather than to individual
atoms. If the total potential energy function V(r) of the lattice is periodic, as
shown in figure 5-6, then the solution is greatly simplified by using the so-
called Bloch functions. Bloch showed that if the lattice potential V(r) is
periodic such that:
V r  = V r + R (5-24)
where R= n1 a1 + n2 a2 + n3 a3 and a1, a2 & a3 are the basic lattice vectors,

then the solution of the one-electron Schrödinger equation can be written in
the following form :
 k r  = e  j k.r uk r  (5-25)
where uk(r) is a periodic function and has the same periodicity as the
crystal lattice, i.e.,
uk r  = uk r  R (5-26)
In fact the functions uk are three-fold periodic and should satisfy:
V r  =  u r  exp  j G
h,k,l
hkl hkl . r (5-27)
where Ghkl is the reciprocal lattice vector, which is given by the relation:
Ghkl = hb1 + kb2 + lb3 and b1, b2 & b3 are the reciprocal lattice unit vectors.
As we have seen so far in Chapter 1, the product Ghkl.Rijk (or simply G.R) is
equal to 2N, where N is an integer. The functions uk(r) are sometimes
called the atomic part of the Bloch functions because they contain the
essential features of the unit cell and reflect the atomic properties. In
contrast, the plane-wave part of (r), i.e. exp(jk.r), reflects the large scale
variations of (r), and becomes significant for free electrons.
-318-
Fig. 5-6. Schematic of the crystal periodic potential.
Note 5-1. Bloch Functions versus Wannier Functions
The Bloch wave functions ukn(r) are the eigenfunctions of the Schrödinger
equation with periodic potential. They are characterized by the wavevector
k and the band index n. The Wannier function Wn(r-R) are equivalent set of
orthogonal functions, which are defined as follows:
Wn r-R = u nk r .dk
1

 j k.R
e
V BZ (5-28a)
The integration is taken over the first Brillouin zone (BZ) of volume V.
There exist , for each band n, a set of Wannier functions, all identical
except each translational lattice vector R. Alternatively, the set of
orthogonal Wannier functions may be defined as follows:
WR r  =  uk r 
1  j k.R
e
N k (5-28b)
where uk(r) has the same periodicity as the crystal lattice and N is the
number of primitive cells in the crystal. The Wannier functions are not
eigenfunctions of the Hamiltonian. However they may serve as a
convenient basis for the expansion of electron states in certain regimes.
-319-
5-4. Kronig-Penney Model

Even when using the Bloch functions approach, the mathematics would be
almost impossible if one were to use realistic periodic potentials. In the
Kronig-Penny model (1931), the crystal potential V(r) is modeled by a
periodic chain of square wells. Fig. 5-7 depicts the Kronig-Penny model in
one dimension. Here we have:
V(x) = Vo for -b < x < 0

= 0 for 0<x<a
Then the time-independent

Schrödinger equation is written as
follows:
Fig. 5-7. Kronig-Penney Model.
1- Inside the potential well (0<x<a) where V=0
d 2
2
+  2 = 0 (5-29)
dx
with
 2 = 2m E/ 2 (5-30)
2- Inside the box (-b < x < a) where V=Vo
d 2
2
+  2 = 0 (5-31)
dx
with
 2=
2m
E -Vo  (5-32)
2
-320-
By applying the Bloch theorem  k  x  = e uk  x , where uk(x) is a

 j k.x
periodic function with period (a+b), we get:
u`` + 2 j k u` +  2  k 2 u  0   (5-33)
which has the solution :
u1 = A exp  j  k x+B exp  j  k x (5-34)
and
u`` + 2j k u` -  2  k 2 u  0   (5-35)
which has the following solution
u2 = C exp  j   jk  x + D exp   j   jk  x (5-36)
where k is the wave number. Applying the conditions of continuity of 

and d/dx at the two boundaries, at x=0 and at x= a, results:
u1(0) = u2(0), u1`(0) = u2`(0), u1(a) = u2(-b), u1`(a) =u2`(-b) (5-37)
Equations (5-34), (5-36) will have a solution, for the constants A, B, C and
D, only if the determinant of their coefficients is null. Therefore,
+1 1 1 1
   k     k     jk     jk  =0
 j   k  a    j   k  a   j    jk b  j    jk b
+e e e e
 j   k  a    j   k  a   j    jk b  j    jk b
+   k  e    k  e    jk  e    jk  e
(5-38a)
This happens if the following condition is satisfied:
-321-
 2  2
sh b  sin  a + cosh b  cos a  = cos k a+b (5-38b)
2 
In the case where the periodic potential well is approximated by a series of

delta functions (i.e., when Vo   , b  0) such that:
V x  = a Vo   x - m a  (5-39)
m
Then, equation (5-40) reduces to


P sin  a 
+ cos  a  = cos k a  (5-40a)
a

where P = m Vo b a /  2 . The above equation is not a solution of the
Schrödinger equation but rather it is the condition for finding a solution for
the wave function in a periodic potential well. In order to satisfy this
condition, the left-hand side (LHS) of this identity must be bound within
the limits of cos(ka) which are  1. Figure 5-8 shows the plot of the LHS
of equation (5-40a) versus (a) and the regions where the solution exists
(dark areas). Figure 5-9 shows the E-k diagram, which we can obtain by the
aid of this relation.
The E-k relation is obtained by defining the regions where a have a

solution (and hence E is allowed), and then calculating the corresponding
values of k using the following relation:
1
k   cos 1 

 P sin 2mE a / 
+ cos
    2n
2mE a /   
a  2 mE a /   a (5-40.b)
The so-called reduced E-k diagram can be obtained by folding the original
E-k diagram inside the region (-/a, /a) for positive and negative k-values
(Figure 5-10). This region is called the first Brillouin zone.
-322-
Fig. 5-8. Regions where there exist a solution of the Kronig-Penney model.
Fig. 5-9. Electronic E-k diagram, according to the Kronig-Penney model.
-323-
E E
E=ħ2k2/2m
k k
-2/a -/a 0 /a 2/a -/a 0 /a
Fig. 5-10. The E-k dispersion relation and the reduced E-k diagram.
-324-
5-5. Effect of Crystal Size & Fine Structure on Energy Bands

In a more realistic model for a one-dimensional crystal with restricted
length L, we can prove that only certain (quantized) values of k (and E) are
possible.
E
V(x)
Vo 2/L
x
0 L
k
0
(a) Potential well of finite size L and finite height Vo and its E(k) diagram.
E
V(x)
a
Vo
x
0
k
-2/a -/a 0 /a 2/a
(b) Periodic potential well of infinite size and finite height Vo and its E(k) diagram
E
2/L
V(x)
a
Vo
x
0 L
k
-2/a -/a 0 /a 2/a
(c) (c) Periodic potential well of finite size L and finite height Vo and its E(k) diagram
Fig. 5-11. Effect of crystal size and fine structure on the E-k relation..
-325-
As shown in figure 5-11, we can observe the effect of crystal size and fine
structure (crystal periodicity) on small- and large-scale quantization of
energy bands. The crystal periodic structure results in discrete and
successive allowed and forbidden bands of energy, whereas the crystal
finite size results is quantized energy levels at each allowed energy band.
Table 5-3. Energy gap of famous semiconductor materials and compounds.
Material Symbol Band gap (eV) at 302K

Silicon Si 1.11
Selenium Se 1.74
Germanium Ge 0.67
Silicon carbide SiC 2.86
Aluminium phosphide AlP 2.45
Aluminium arsenide AlAs 2.16
Aluminium antimonide AlSb 1.6
Aluminium nitride AlN 6.3
Diamond C 5.5
Gallium(III) phosphide GaP 2.26
Gallium(III) arsenide GaAs 1.43
Gallium(III) nitride GaN 3.4
Gallium(II) sulfide GaS 2.5
Indium antimonide InSb 0.17
Indium(III) nitride InN 0.7
Indium(III) phosphide InP 1.35
Indium(III) arsenide InAs 0.36
Zinc oxide ZnO 3.37
Zinc sulfide ZnS 3.6
Zinc selenide ZnSe 2.7
Zinc telluride ZnTe 2.25
Cadmium sulfide CdS 2.42
Cadmium selenide CdSe 1.73
Cadmium telluride CdTe 1.49
Lead(II) sulfide PbS 0.37
Lead(II) selenide PbSe 0.27
Lead(II) telluride PbTe 0.29
Copper(II) oxide CuO 1.2
Copper(I) oxide Cu2O 2.1
-326-
5-6. Analogy with Circuit Theory

The existence of allowed and forbidden energy bands is a characteristic of
waves propagating through periodic structures. This does not only happen
for electron waves in a crystal lattice, but also in many other physical
systems as well. A typical example is the propagation of electromagnetic
waves through a long coaxial line, which is periodically loaded (e.g., each
d interval) as shown in figure 5-12. In such a system the frequency of
propagating modes is quantized and there exist allowed and forbidden
frequency bands, as shown in figure 5-10. In the electromagnetic theory,
the forbidden frequency modes are sometimes evanescent modes.
 d  d  d 
Fig. 5-12. A coaxial transmission line and its equivalent periodic loaded L-C circuit.

2c/d Forbidden band

c/d Allowed band
k
0 /d 2/d
Fig. 5-13. The dispersion characteristics (-k ) of a periodically-loaded

transmission line.
-327-
5-7. Energy Band Structure Calculation Methods

The energy band structure of a solid can be illustrated by the relation
between the carrier wave-vector, k (or momentum p=k), and its energy, E,
i.e. the E(k) relation. As we have seen so far, the (E-k) relation of a
crystalline solid is usually obtained by solving the Schrödinger equation
using the so-called: one-electron approximation. The calculations needed
for solving the Schrödinger equation and the determination of the (E-k)
relation in a three-dimensional solid may be then quite complicated,
although approximations are always used. For instance, the solution of the
Hartree-Fock equation (5-22) is usually carried out using the self-
consistent field method, where, the choice of a good zero-approximation
for k (the self-consistent functions) is not an easy task.
Generally speaking, the electronic band structure calculation methods can

be grouped into two general categories, namely:
ab initio Methods,
Empirical (Approximate) Methods
The ab initio methods such as Hartree-Fock (HF) and variant methods,

such as the Density Functional Theory (DFT), calculate the electronic
structure from first principles, without the need for empirical fitting
parameters. In general, these methods utilize a variational approach to
calculate the ground state energy of a many-body system, where the system
is defined at the atomic level. Typically, the original calculations are
performed on systems containing a few atoms (up to 1000 atoms), because
these methods are computationally expensive, in terms of CPU time and
memory storage. The computational load of the HF method scales as N4
(N being the number of basis wavefunctions). In practice, it can scale closer
to N3 as the program can identify and neglect zero and small integrals.
The empirical methods, such as the quasi-free electron (QFE), which is

known as rhe Orthogonalized Plane Wave (OPW) method or the tight-
binding (TBE) method, which is sometimes known as the Linear
Combination of Atomic Orbitals (LCAO) method.
In addition to the above basic methods, there exist other variant methods
which are improved versions of the OPW or TB methods. For instance, the
pseudopotential method involves some empirical constants or fitting
parameters. These parameters are chosen to fit the experimental data of
band-to-band transitions, derived from optical absorption experiments.
-328-
We dully note that, in all these methods the electronic structure is

calculated by solving a one-electron Schrodinger equation. However,
empirical methods are computationally less expensive than ab initio
calculations and provide a relatively easy means of generating the
electronic band structure.
5-7.1. Energy Band Structure Approximate Methods

There exist two basic approximate (or empirical) methods, which are
widely used for the calculation of energy bands in solids, namely:
The quasi-free electron (QFE) model, and
The tightly bound electron (TBE) model.
Both of the above approximate methods use the perturbation theory. But
they differ in their zero approximation. The quasi-free electron (QFE)
method (which represents the physicists approach) takes the free electron
model as a zero-order approximation and the periodic field of the lattice as
a perturbation. This results in the so-called pseudopotential methods. The
tightly bound electron (TBE) method (which represents the chemists
approach) takes the electron in an isolated atom as a zero approximation
and the periodic potential of the lattice as perturbation. This results in the
so called LCAO (linear combination of atomic orbitals) methods. In the
following sub-sections, we examine the fundamental results of these
approximate methods. We begin by the free electron model as an extreme
case, where electrons are not subject to any interaction or crystal forces.
i. Gas of Free Electrons (Fermi Gas Model)

One can explain many physical properties of metals by the model of free
electrons. In this model, we neglect the interaction between the valence
electrons (which we consider as free) and their positive ions (or cations).
Therefore, the potential energy of electrons is neglected and the total
energy is simply equal to the kinetic energy. In fact, it was shown
experimentally that free electrons in metals can move in straight lines
(without interaction or collision with other electrons) for long distances. At
room temperature this distance may be about 1 cm, 1000 atomic distances.
The wave function (r) of a free electron (without acting field force) is
described by the following Schrödinger equation:
2   2 2  2 
 + +  r   E  r   0
2m   x 2  y 2  z 2 
(5-42)
-329-
If the electron is enclosed in a box (or a sample of metal) of edge length L,

the wave function (r) and the energy eigenfunction E will be exactly the
same as those given by (4-48) and (4-47). These equations describe the
wave functions and energy eigenvalues for an infinite cubic potential well,
with the edge length a is replaced by L. However, we know that the
solution of (r) must satisfy the cyclic boundary conditions at the box
limits such that:
(x, y, z) = (x+L, y, z)
= (x, y+L, z) (5-43)
= (x, y, z+L)
Therefore, for free electrons the wave function satisfying Schrödinger

equation and the above boundary conditions should take the following form
of propagating waves:
(r) = A exp (j k.r) (5-44)
where A is a normalization constant. Note that, when the electron is bound

(e.g., in a potential well), the waves described by equation (5-44) are
subject to multiple reflections at the potential-well walls, such that (x) ~
[exp(jkx) - exp(-jkx)] ~ sin(kx), and the resulting waves are called standing
waves. The wave vector k in the above solution is quantized such that:
k 2 = k x2 + k y2 + k z2 (5-45)
and
2 n 2 m 2l
k x= , k y= , k y= (5-46)
L L L
where n, m and l are integers. Substituting the solution (5-44) into the
Schrödinger wave equation (5-42) results is the (E-k) relation of the free
electron:
 2 k 2  2  2 
2
E  (5-47)
  .( n  m  l )
2 2 2
2m 2m  L 
-330-
The (E-k) relation is then parabolic1. Figure 5-14 depicts the E-k relation of
a free electron in a one-dimensional lattice of length L. As the crystal
length L is much greater than the inter-atomic distances, the separation
between adjacent k-values (2/L) is negligible. Then the E-k relation of
such free electrons may be considered as a continuous curve. In the 3-
dimension k-space, equation (5-47) is usually represented by constant
energy spheres (for different values of E) as shown in figure 5-14. Also the
allowed states in the k-space are represented as points inside such spheres
of constant energy. The elemental volume in the k-space is termed by d3k
and is given by:
 2 
3
d k = dk x dk y dkz = 
3
 (5-48)
 L 
According to Pauli’s exclusion principle, each elemental volume (cell) in

the k-space can accommodate up to 2 electrons (of different spins). The
probability of occupation of these cells is given by the Fermi-Dirac
distribution. Thus, in the fundamental state of an electron gas of N free
electrons, the occupied states can be represented by points inside a sphere
of radius kF. The Fermi energy EF is related to kF by the quadratic relation:
2 2
EF = kF (5-49)
2m
E(k) kz
ky
k kx
E = constant
-2/L 0 2/L
Fig. 5-14. The E-k relation of a free electron, and the 3-dimensional representation in
the form of spherical constant energy surfaces.
1
This relation can be directly obtained from the known relation E = p2 /2m for a free electron.
-331-
By definition, all energy levels under the Fermi-level are occupied and all
those above it are empty in the system fundamental state (at 0K). Then, the
total number of electrons inside the Fermi sphere is equal to its volume
divided by the volumes occupied by each electron,
N
4 / 3 k F3 = V k F2
1 2 / L 3 3 2 (5-50)
2
Here we considered V=L3. Therefore, the Fermi-energy EF can be related to

the density of electrons n = N/V and mass m by the following relation:
2  N  
2
EF     EF 
2m  2 L 
2
2m
3  2
n 
2/3 (5-51)
It should be noted that, on the free-electron model, there is no band

structure and the existence of Brillouin zone is irrelevant since the electron
cannot detect the crystal structure.
-332-
ii. Empty Lattice Model

The empty lattice model is just as the free electron model, but with
superimposed symmetry of the appropriate lattice symmetry. In this case,
you will have certain band structure, without energy gaps. For instance, the
following figure depicts the band structure of the free electron model, with
diamond lattice symmetry, in the first Brillouin zone (BZ). Obviously, the
band structure of a free electron is the same whether it is in a zinc-blende or
a diamond crystal. Notice that if a finite crystal potential is added, energy
gaps will open up at the BZ edges.
Fig. 5-15. Empty lattice band structure for a free electron in a diamond lattice.
iii. Quasi-Free Electron Model (Weakly-bound electron model) (A)

The second classical approximation is that of weakly bound or nearly free

electron model. In this approximation, the Hamiltonian operator Η is
considered as the sum of two terms:

Η = Ho + H1 (5-52)
-333-
The first term Ho = p2/2m, which corresponds to the Hamiltonian of a free

electron, is considered as the zero approximation. The Hamiltonian
perturbation term H1 is equal to the crystal potential V(r). The crystal
potential V(r) is assumed to have a constant average value Vo plus a weak
periodic component v(r) such that
H1 = V(r) = V o + v(r) (5-53)
The Schrödinger wave equation may be then written in the following form:
2 2
  +  - v r   = 0 (5-54)
2m
where = E - V o and v(r) is assumed small compared to . The detailed

mathematical analysis of the nearly free electron model is just an exercise
on Fourier analysis. The potential energy can be expanded as a Fourier
series and truncated to first terms. In one-dimensional lattice, the potential
energy can be expanded as follows:
i= 
V ( x )   v i exp ( jki x ) (5-55)
i=-
where vi are the Fourier series coefficients and ki =2i /a. vi are given by:
a
1
vi  a  V ( x) exp ( j ki x)dx (5-56)
o
Evidently, vo is equal to the average of the potential, i.e. vo =V0. The

corresponding solution of  is then given by:
+
 ( x)  u m exp  j( k + G m ) x  (5-57)
m = -
where Gm is the translation vector in one-dimensional reciprocal lattice and

given by : Gm = 2m/a with m integer. The stress is made on the plane
wave part of the Bloch functions. The atomic part, which is strongly
dependent on the periodic variation of V(r), is only considered as a small
perturbation. In fact, the atomic functions are considered as constants in
-334-
this method (the coefficients um). If the Fourier series are truncated to three
terms such that:
  2 x  
V ( x )  v1 exp ( jk 1 x ) + v o + v -1 exp (  jk 1 x )  Vo  1 - cos  
  a   (5-58)
Thereupon, the solution of  is given by:
 
 ( x ) ~ exp j(k + ) x  exp j(k - ) x (5-59)
a a
E = (ħ2/2m)(k - 2/a)2 E(k) E = (ħ2/2m)(k + 2/a)2
Ec
Eg
Ev
k
-2/a 0 2/a
Fig. 5-16. The E-k relation in the quasi-free electron model.
Substituting the above solution into Schrödinger equation results in two

algebraic equations whose solution implies the following E-k relation, as
shown in figure 5-16:
 2  2  Eg  2k 2  
1  4  2  
2 2 2
E (k )      (5-60)
2m  a  2 2m  2mEg  a  
 
where Eg = 2 |V0 |. As shown in figure, there exist two allowed solutions for
the E-k relation. Each solution corresponds to an allowed energy band. At k
= 0, we have:
 2  2 
2
Eg
E( 0 )     (5-61)
2m  a  2
-335-
Therefore, there exists an energy gap between the two allowed energy
bands described by (5-60). The height of this gap is given by Eg = 2 |V0 |.
The mathematical structure of the quasi-free electron model shows that it
cannot be valid in the neighborhood of certain planes in the k-space.
iv. Pseudopotential Method

The pseudo-potential method is based upon the observation that the core
electrons are relatively unaffected by the environment of an atom.
Therefore, we assume that their contribution to the total binding energy
does not change when isolated atoms are brought together to form a solid
crystal. The actual energy differences of interest are the changes in the
outer valence electron energies. In fact the strong nuclear Coulomb
potential and highly localized core electron wave-functions are difficult to
represent computationally. For instance, in silicon (1s22s22p63s23p2) we can
neglect the contribution of the core electrons (1s22s22p6) and consider only
the influence of the outer valence electrons (3s23p2). Since the core states
are localized in the vicinity of the nucleus, the valence states must oscillate
rapidly in this core region in order to maintain this orthogonality with the
core electrons.
Fig. 5. 17. Pseudo-potential Vps and the pseudo-potential wavefunction ps.
-336-
The rapid oscillation of core electrons results in a large kinetic energy for
the valence electrons in the core region, which roughly cancels the large
potential energy due to the strong Coulomb potential. Thus the valence
electrons are much more weakly bound than the core electrons. It is
therefore convenient to replace the strong Coulomb potential V(r)=-Ze/r of
core electrons by an effective pseudo-potential Vps, which is smoothly
varying. This effective potential replaces the valence electron wave-
functions , which oscillate rapidly, in the core region. Figure 5-17 depicts
the pseudo-potential function Vps and the pseudo-potential wavefunction
ps. The subsequent figure 5-18 depicts the pseudopotential Vps and its
Fourier transform in the reciprocal lattice wavevector space (q-space).
Fig. 5-18. Pseudopotential Vps and its Fourier transform in the reciprocal k-space.
The E-K diagram and the properties of the semiconductor can be calculated
using the following one-electron Schrödinger wave equation:
(5-62a)
with k(r) denoting the pseudo wavefunction and V(r) the pseudopotential.
This function is a good approximation to the true wave function outside the
core region and therefore can be used to calculate the physical properties of
the semiconductors which are dependent on the valence and conduction
electrons only. Since pseudopotentials are weak perturbations on the free-
electron band structure, a good starting point for diagonalzing (5-62) is to
expand k as a sum of plane waves (similar to those of nearly free
-337-
electrons). The following expansion can be used to diagonalize the last

equation:
(5-62b)
Here, G is the reciprocal lattice vector and the Dirac notation |k+G> is
used to denote the plane wave with wavevector K+G.
The matrix elements are determined by the Fourier components VG of the

pseudopotential
(5-63a)
These are frequently called the form factors of the pseudopotential. When
there exist more than one atom in the primitive unit cell, a structure factor
is introduced. This structure factor SG depends on the relative position rn of
the respective atom in the primitive unit cell, and is defined as follows:
(5-63b)
Here N is the number of atoms in the primitive unit cell. The pseudo-
potential V(r) can be expressed in terms of the structure factor and the form
factors as follows:
(5-63c)
For diamond structures, there are two atoms in the primitive unit cell, at the
positions r1 , r2. Taking the midpoint between the two atoms in the unit cell
as origin, the positions of the atoms are then given by r1 =-r2= (ao/8)(1,1,1).
Thus, the structure factor is given by
(5-63d)
-338-
where  =r1=(ao/8)(1,1,1). In unstrained diamond structures the reciprocal

lattice vectors in order of increasing magnitude are (in units of 2/ao):
(5-64)
Form factors with reciprocal lattice vectors larger than G2>11(2/ao)2 may
be neglected, since typically VG decreases as 1/G2 for large G. Assuming
that the atomic pseudopotentials are spherically symmetric, the form
factors only depend on the absolute value of the reciprocal lattice vector.
The form factors belonging to the reciprocal lattice vectors G3 have an
absolute value of √3(2/ao) and are conventionally labeled V3. Only three
pseudopotentials form factors V3, V8 and V11 are required to calculate the
band structure. The following table lists the parameters employed in the
band structure calculation of Si and Ge
Table 5-4. Parameters employed in the band structure calculation of Si and Ge
Pseudopotentials Si Ge Units
-0.2241 -0.221 Rydberg
-0.052 0.019 Rydberg
-0.0724 0.056 Rydberg
0.03 0.275 Rydberg
1.06 1.22 Å
0.00023 0.000965 Rydberg
7.5589 10.0911 Å
the employed parameters in the empirical pseudopotential calculations

consist of three local form factors V3, V8 and V11, two parameters (Ao, Bo) to
model the nonlocal correction, and two parameters ( ) entering the spin-
-339-
orbit interaction term. The following figure depicts the energy band
structure of Si, as obtained by the pseudopotential method.
Fig. 5-19. Energy band structure of Si, according the pseudopotential method.
There are two approaches for calculating pseudopotential form factors.

They can be determined by fitting a small number of experimental data,
such as the position of peaks in optical reflectivity spectra. This approach is
known as the Empirical Pseudopotential Method (EPM). The
disadvantage of the EPM is that it requires experimental inputs. With the
availability of high-speed computers, however, it is possible to determine
the pseudopotential form factors from first principles without any
experimental input. These first-principles pseudopotential methods are
known as self-consistent or ab initio pseudopotential methods.
-340-
v. Tightly-Bound Electron Model (TBE)

The second basic approximation is called the tightly bound electron (TBE)
model. It is sometimes called linear combination of atomic orbitals
(LCAO). This is the complementary approximation to the quasi-free
electron model. For an isolated atom of potential V(r), the Schrödinger
equation yields a solution which differs depending on whether we consider
the energy shell s or p, etc. The mathematical analysis used in this
approximation has its origin from the perturbation theory. For a given
energy level (s or p, etc.), the final solution of the energy E is composed of
a zero order approximation term E 01 (denoting a certain energy level of an
isolated atom), plus corrections terms, which depend on the lattice constant
a, as shown in figure 5-20.
E  E10   a   2 a  Cos ka + Higher order terms (5-65)
Figure 5-19 depicts the formation of an energy band from an energy

level, E 01 , in the TBE model. A more complete description of the TBE
method can be found in the article of Pantelides and Harrison [12].
E(k) E( k, a)
k = /a
Eo1 + + 2
Eo1
4
Eo1 + - 2
k=0
k a
-/ao 0 /ao ao
Fig. 5. 20. The energy band structure in the tightly bound electron model.
The concept of bonding and antibonding orbitals introduced for molecules

can be easily extended to crystals if one assumes that the orbitals of each
atom in the crystal overlap with those of its nearest neighbors only. This is
a reasonable approximation for most solids. The results of orbital overlap
in a solid is that the bonding and antibonding orbitals are broadened into
bands. Those occupied by electrons form valence bands while the empty
-341-
ones form conduction bands. Figure 5-21 shows schematically how the s
and p orbitals evolve into bands in a tetrahedral semiconductor.
Fig. 5. 21. Evolution of the atomic s and p orbitals into valence and conduction bands
in a semiconductor (left). Case of Si (right)
Fig. 5. 22. The valence band structure of Si calculated by the tight-binding method
(broken curves) and by the empirical pseudopotential method (solid lines)
-342-
5-7.2. Energy Band Structure in Solids (ab initio Methods)

The ab initio quantum methods are computational methods based on
quantum physics, to find out the energy band structure of a matter. Ab initio
electronic structure methods have the advantage that they can be made to
converge to the exact solution, when all approximations are sufficiently
small. In this case, when all possible configurations are included, such
methods tend to the exact solution of the electronic Schrödinger equation
(in the Born-Oppenheimer approximation). The convergence, however, is
usually not monotonic. The Hartee-Fock, (HF) technique, which is the
simplest ab inito method, has been introduced in the beginning of this
Chapter2, in conjunction with the self-consistence methods. The
computational load of the HF method scales as N4 (N being the number of
basis wavefunctions). However in practice it can scale closer to N3 as the
program can identify and neglect zero and small integrals. Density
functional theory (DFT) methods use functionals, which include Hartree–
Fock exchange scale in a similar manner to Hartree–Fock method, but with
a larger proportionality term and are thus more expensive than an
equivalent Hartree–Fock calculation. DFT methods that do not include
Hartree–Fock exchange can scale better than Hartree–Fock. The dark side
of ab initio methods is their computational cost. They often take enormous
amounts of CPU time and memory.
i. The k.p Method and Luttinger-Kohn-Hamiltonian

A widely used approach for modeling the near-band-edge electronic states
in semiconductors is the k.p method. By this method we can obtain an
analytical expression for the band structure near a given point k0 in the k-
space, for which the band structure is already known. It also allows the
extrapolation of the band structure over the entire Brillouin zone from the
energy gaps and optical matrix elements at the zone center. In this approach
the electronic state (r) is approximated in terms of d bands, as follows:
k(r) = exp(jk. r) v v(x; k).u(r) (5-66)
The index k=(kx, ky) indicates in-plane vector, the functions u(r) are lattice
periodic, zone-center Bloch functions varying on the atomic scale and the
functions (x; k) are the corresponding envelope functions describing the
variation of the wave function on the (larger) nanoscale.
2
The Hartee-Fock and self consistent methods have been already described in section §5-2.5.
-343-
There is a hierarchy of k.p-models including 4-band, 6-band and 8-band

Hamiltonians. Depending on the Hamiltonian model, effects such as
quantum confinement, band-mixing, spin-orbit interaction and mechanical
strain can be treated consistently. The basic stage in this model hierarchy is
the 4-band Luttinger-Kohn-Hamiltonian, which describes the band-
mixing between the heavy holes and the light holes.
Fig. 5-21. Illustration of the k.p Hamiltonian matrix.
ii. Full-potential Linearized Augmented Plane Waves (FLAPW ) Method

The FLAPW (full-potential linearized-augmented plane-wave) method is
one of the most accurate first-principles (ab initio) methods for determining
electronic and magnetic properties of crystals and surfaces. In the past the
FLAPW method has been limited to systems of about a hundred atoms due
to the lack of an efficient parallel implementation to exploit the power and
memory of parallel computers. Nowadays, there exist very efficient
implementations of this method, which make use of parallel processing of
the plane-wave components for each state by several processors.
iii. Density Functional Theory (DFT)
The density-functional theory (DFT) is a specific technique to solve the
many-body Schrödinger equation, in terms of single-particle equations and
an effective potential. Today, density functional theory is widely employed
in the description of bulk materials and molecules in physics, chemistry, and
biology. The beginning of the DFT was in the papers of Thomas and Fermi
in the 1927 and its extension due to Dirac in 1928, but only during the 1960s
the theory became complete and accurate in the work of Kohn, Sham and
Hohenberg. In much the same way as the Hartree equation, the DFT
provides a way to decouple the electron-electron interactions in the material,
and solve the one electron equations by a self-consistent method to find out
the electron wave-functions and the corresponding energies. However, the
solution is performed starting from the ground-state electron density, like the
self-consistent field method. Therefore, the many-electron Schrödinger
equation is replaced by the so-called Kohn-Sham equation:
-344-
 2 2 ' 
 2m   Veff  i (r )  Ei i (r ), i  1,2,... N (5-67)
 
where N is the number of electrons in the system and the effective

potentials, Veff, is defined as follows:
Veff = V(r) + VH + Vxc (5-68)
The first term in equation V(r) is the external potential and includes the
potential originating from the nuclei as well as an external applied electric
field if present. The term VH corresponds to the classical Coulomb potential
(Hartree potential) for an electron density n(r). This term is obtained by
solving the Poisson equation for the scalar potential
2VH(r) = e.( /= - (e2/) n(r) (5-69)
Finally, the exchange-correlation potential Vxc takes into account the many-
body effects, and is defined as the functional derivative of the exchange
energy, EXC with respect to the electron density Vxc =∂Exc /∂n. The last term
contains the remaining contributions to the potential that we do not know
exactly. There exist N eigenfunctions i for each electron, from which the
electron density can be calculated using the Thomas-Fermi model
(5-70)
Here, he sum is taken over all the occupied states and fi = f(Ei - EF) is the
occupation probability (the Fermi-Dirac distribution in equilibrium).
Also, the total energy is given by:
ET =  Ei . f(Ei - EF) (5-71)
where the sum is taken over the total number of electrons (N).
Note 5-2. Hohenberg-Kohen (HK) Theorem

The Hohenberg-Kohen (HK) theorem states that: for any system
constituted by interacting particles, for a fixed interaction, there exists a
bijective relation among the external potential, the ground–state many–
body wave–function and the ground–state density of the system,
-345-
Vext ⇐⇒ 0 ⇐⇒ ρ0 ; ρ0 = <0|ρ|0>.
Thereforte, the ground–state energy Eo and density o can be determined by
minimizing a functional of the charge–density n. This functional has a
minimum when ρ = ρ0. As a consequence of the HK theorem, the
knowledge of the ground–state wave–function 0 gives access to all
physical observables. Moreover the external potential fixes the Hamiltonian
and knowing the Hamiltonian we can in principle access the excited states
of the system. It is surprising that given the ground–state density we can
access all this physical information
Fig. 5-22. Illustration flowchart of the Kohen-Sham method.
-346-
The solutions of the Kohn-Sham equations in principle give the ground

state properties of the system. However, the term Exc is not known exactly
and some kind of approximation is needed. The two types of
approximations which are generally used are the local density
approximation (LDA) and the generalized gradient approximation (GGA).
In both LDA and GGA, the exchange-correlation potential Vxc is a local
functional of the charge density. They are accurate when the charge density
n(r) is a smoothly varying function. However, for systems where strong
correlation effects are important, i. e., where the charge density n(r)
changes sharply, these approximations most likely fail.
DFT is in practice used only for ground–state properties, as there is no

practical scheme able to describe excited stated starting from the ground–
state density. Excited state properties can be obtained from the extension to
the time domain of DFT: this is Time–Dependent DFT (TDDFT).
δρ(1) = χKS (1, 1′)δvKS(1′)
where χKS(1, 2) is the response function of the KS non interacting system,
while δvKS(1) is the variation of the TDKS potential needed to follow the
evolution of the density when an external potential δVext(1) is applied. The
analogous version to the Hohenberg-Kohn theorem which includes time-
dependent potentials is known as the Runge-Gross theorem.
iv. Local Density Approximation (LDA)

The local density approximation (LDA) was proposed by Kohn and Sham.
They showed that it could be applied to the limiting case of a slowly
varying electron density. Therefore, one considers exchange energy EXC to
be that of a uniform electron gas of the partial density n(r). In this case, the
Kohn-Sham equation (5-63a) may be written as follows:
 2 2 n( r ' ) 
 2m   V (r )   | r  r '|dr'  VLDA   i (r )  E (r ), i  1,2,... N
'
(5-72)
 
where VLDA = ∂E/∂n is the local approximation to the Kohn-Sham

exchange-correlation potential and V(r) is the external potential energy.
The Kohn-Sham-LDA can be further extended to the spin dependent case
by replacing the scalar external potential V(r) by a spin dependent potential
and replacing the electron charge density n(r) by the density matrix. The
DFT in the local density approximation (DFT-LDA) has been very
-347-
successful in the analysis of interacting-electron ground states. Many

efforts have also been devoted to DFT-based schemes for excited states.
5-8. Electron Dynamics and Effective Mass

The electron (or hole) effective mass is the mass it seems to carry in a
crystal. It can be shown that, under most conditions, electrons and holes in
a crystal respond to electric and magnetic fields as if they were free
particles in a vacuum, but with a different mass. In fact, electrons with
energy close to the conduction band minimum behave almost like free
electrons. The presence of the periodic potential of crystal atoms, changes
the properties of electrons. Therefore, the mass of an electron inside a
crystal differs from the free electron mass, m0 = 9.1x10-31 kg.
Note 5-4. E(k) Relation Expansion

The E(k) relation can be developed, in 1-dimensional k-space, around an
extremum (minimum or maximum) point Eo=E(ko) using the Taylor series
as follows:
E(k)  E k o   k  k o  
dE
dk ko

1
2
 k  k o
2 d E 
2
 d k 
2
ko
 ... (i)
As E(ko) corresponds to an extremum, then the first order derivative

(dE/dk) must vanishes at k = ko. Also we can neglect higher order terms
near the edge where the difference (k - ko) is very small. So:
E ( k )  E k o    k  k o 
1 
2 d E
2
 (ii)
2  d k 
2
ko
Substituting m*   2 / d 2 E / dk 2  and E(ko) = Eo results in:

 k  k o
2 2

E( k )  E o  (iii)
2m*
5-8.1. One-dimensional Model (Scalar Effective Mass)

Let us begin our study by the simple case where the k-space is one
dimensional and the carrier motion is collinear. The equation of motion of
a wave packet (associated to a moving electron) can be determined from
the work exerted by a force F that produces a change in the electron energy
E during a time t. The electron group velocity vg is given by:
1 dE 1 E
vg   (5-73)
 dk  k
-348-
where the force F is related to vg and the increase is energy E during t

by the following relation:
E = F. vg t (5-74)
But E = (dE/dk) k. So, we can write:
F =  dk / dt (5-75)
which is equivalent to F = dp/dt, with p being the electron (or lattice)

momentum. This relation, which is equivalent to the classical equation of
motion, is known as the Ehrenfest theorem3. On the other hand, the
acceleration of electrons  is given by:
dvg d  1 dE  1  dk   d 2 E 
       (5-76)
dt dt   dk    dt   d k 2 
Substituting F from (5-65) into results in:
 = F(1/) dE/dk (5-77)
The above relation may written in the following classical form, which is
analog to Newton’s second law:
F = m*  (5-78)
with
2
m* = 2  2

d E dk  (5-79)
The parameter m* has the dimension of mass and usually called the
effective mass of the electron. Figure 5-23 depicts the variation of m* in a
1-dimensional k-space as derived from the E-k relation. The effective mass
m* is different from the gravitational or rest mass mo of the electron
because of the inclusion of the lattice forces through the E-k relation of the
electron. As shown in figure, the effective mass near a band edge is
approximately constant. This is because the E-k diagram is almost
3
According to the Ehrenfest theorem, we have F= dp/dt and v = dr/dt in both classical and quantum
mechanics.
-349-
parabolic near band edges. For instance, the effective nass of electrons near
the bottom of conduction band is almost 0.36 mo in Si and 0.55 mo in Ge
and 0.45 mo in GaAs. The (E-k) relation near an extremum point E(ko)= Eo
can be developed in the following form:
 k  k o
2 2
E (k )  E  o * (5-80)
2m
The  sign corresponds to a minimum or a maximum edge of E(k) relation.
E(k)
k
-/a -/2a 0 /2a /a
vg
-/a -/2a 0 /2a /a

m*
-/a -/2a 0 /2a /a
Fig. 5-23. Variation of E, vg and m* in one dimensional k-space
For a conduction band whose minimum at E(kc) = Ec we have:
E( k )  E c 
 k  kc
2
  2
(5-81)
2m*
-350-
Also for a valence band whose maximum in E (kv) = Ev, we have:
E( k )  E v 
2 k  k v 
2
(5-82)
2m*
5-8.2. Three-dimensional Model and Tensor Mass (A)

In the three dimensional k-space, we have:
1
vg   E (5-83)
 k
where  k is the nabla differential operator in the k-space:
E E E
 E akx  a ky + akz
k x k y kz
k
(5-84)
with akx, aky and akz being the unit vectors in the Cartesian k-space. Also the
acceleration of the electron is given by:
dvg
 
1
F .k  k E (5-85)
dt 2
Therefore, the effective mass is a 2nd order tensor (matrix with 9 elements):
  2E  2E  2E   1 1 1 
 
 k x
2
k x k y k xk z   m*xx m*xy m*xz 

1 1   2E  2E  2E   1 1 1 
  
m*  2  k y k x k y2 k y k z   m*yx m*yy m*yz 
  2E  2E  2E   1 1 1  (5-86)
 
 k z k x k z k y k z2   mzx m*zz 
*
m*zy
This may be written in the tensorial form:
1
m * -1  2
kk E (5-87)

-351-
Simply, the coefficients of the inverse effective mass tensor can be written
as follows:
* 1 1  2E
m  2 (5-88)
 ki k j
ij
with ( i, j = x, y, z ).
When the solid material is isotropic then the effective mass tensor reduces
to a scalar quantity (zero-order tensor), such that: m*xx  m*yy  m*zz and other
coefficients are null.
1 1   2E  1   2E 
m*    is positive m* 1    is regative
 2  k 2   2  k 2 
( Electrons in Conduction Band ) ( Electrons in Valence Band )
Fig. 5-24. Schematic representation of the conduction & valence bands in 2-

dimensional k-space
The effective mass method is valid for non-perturbed states with energies
close to the extreme values (minima and maxima of energy bands). The
effective mass theory may be also valid for small perturbations, i.e. for
small and slow external applied fields. It should be noted, that external
fields bend the energy bands of an ideal crystal so that the displacement of
all bands is the same.
-352-
5-9. Actual Band Structure & Fermi Surface in 3-D Crystals

The crystalline semiconductor properties are mainly determined from the
states near to both sides of the energy gap, i.e. near the top edge of valence
bands and near the bottom edge of conduction bands. In 3-dimensional k-
space, the E-k relation near a band edge (at k=ko) may be expanded into a
Taylor series as follows:
E k   

k  ko m  d m E k 
 dk m  (5-89)
m0 m!   k  ko
or
2 d E 
E(k)  E k o   k  k o 
dE  2

1
k  k   2  ....
 dk  k ko 2 (5-90)
o
 dk  k k
o
2 2
where dE/dk=0 at band edges (a minimum or a maximum). Also d E/dk
is a tensor of 2nd order (dyadic) whose components are ( E/kikj). By
2
appropriate choice of the coordinates x, y and z, this tensor may be

diagoralized such that ( E/ ki kj)=0 for i  j. Thergore, the E-k
2
relation in the k-space may be written as follows, if the Taylor series is

truncated after the 2nd order terms:
E k   E ko   * k x  k xo   * k y  k yo   * k z  k zo 
2 2 2 2 2 2
(5-91a)
2mxx 2m yy 2mzz
where m*xx , m*yy and mzz

*
are defined as follows:
2 2 2
m*xx  , m*
 and m*
 (5-91b)
 2 E  k x2 yy
 2 E  k y2 zz
 2 E  k z2
Equation (5-91), which represents the E-k relation near a band edge, is a 2nd
order surface in the k-space (i.e. a sphere, or ellipsoid, according to the
magnitude and sign of its coefficients). For instance, if mxx=myy =mzz = m*,
then equation (5-91) reduces to the following simple parabolic E-k relation:
2
E k   E k o   *
k  k o 2 (5 - 92)
2m
-353-
This equation describes a sphere in the k-space, whose center ko =

(kxo,kyo,kzo). The constant energy surface is the surface on which the energy
E is constant in the k-space. In particular, the constant energy surface on
which E is equal to the Fermi energy EF is called the Fermi surface. For
instance, the Fermi surface of a gas of free electrons (for which the mass is
constant in all directions and equal to the electron rest mass) is simply a
sphere in the k-space.
The electrical properties of a material are determined by the shape of the

Fermi surface. We'll see later on that the current flowing across a solid is
related to the change in the occupancy of states near the Fermi surface of
the material. For systems of interacting particles such as electrons inside
crystal lattice, the Fermi surface is not generally spherical. Unlike free
electrons, the quasi-free electrons have non-spherical constant energy
surfaces. Figure 5-25 depicts the Fermi surface of copper according to the
nearly free electron model.
Fig. 5-25. The First Brillouin zone of the FCC lattice and the Fermi energy surface of
electrons in Copper.
The E-k relation in certain valleys of some semiconductors cannot be

accurately described by a simple parabolic relation. In this case, a non-
parabolicity factor is usually introduced in the E-k relation of the
semiconductor:
2
     ( E )  E k  1   E( k )
2
k k o (5–93a)
2m *
-354-
where  is the non-parabolicity factor . For the main conduction band of

Si, the non-parabolicity factor is defined as follows:
1  m* 
 .1   (5–93b)
Eg  mo 
The nonparabolicity factor a is 0.5 eV-1 in Si , Ge and 0.7 eV-1 in GaAs.
Note 5-5. Tensors
A 2nd order tensor is a square matrix of rank 3. The components of a 2 nd

order tensor are called dyadic terms. Similarly the components of a 3 rd
order tensor are called triadics terms and so, on (tetradic, pentadic, etc).
The tensorial product T of two vectors A=[A1 A2 A3] and B=[B1 B2 B3] is
termed as T=AB or Tij=AiBj with (i,j =1,2,3).
The main features of the band theory that are tightly related to the
electronic conduction phenomena are:
1- Non-sphericity of the Fermi surfaces. This is because electrons are not
completely free inside the crystal.
2- Variation of electron group velocity vg over the Fermi surface.
2- Variation of effective mass m* over the Fermi surface.
3- In non-cubic crystals, the non-sphericity of Fermi-surface can lead to
anisotropic conductivity. This is because the effective mass is no
longer a constant scalar quantity. Rather, it becomes a 2 nd order tensor
whose components (dyadics) are k-dependent.
-355-
5-10. Case Studies: Band Structure of Si, Ge and GaAs(A)

The energy band structure is simply the plot of the E-k relation inside the
first Brillouin zone, for each allowed band. The E-k relation can be also
plotted in different directions of the k-space using 2-dimensional plots. The
band structure of silicon, germanium and gallium arsenide is shown in
figure 5-26, in different directions of the k-space.
Fig. 5-26. Energy band diagram of Ge, Si, and GaAs (from left to right, respectively).
5-10.1. Case Study 1: Band Structure of Silicon (A)

As mentioned so far, in Chapter 1, Si crystallizes in the diamond structure.
The first Brillouin zone of silicon is a truncated octahedral. The detailed
band structure of silicon is shown in figure 5-27. As shown in figure, there
exist 4 conduction bands and 4 valence bands in the silicon band structure.
These 8 bands of silicon are arising from the 3s and 3p electrons (2+2
electrons). In fact when an electron is shared between two atoms in a
covalent bond it may be in one of two states. The first state called the
bonding state and has a lower energy than the original energy level of
isolated atom. The second state is called the anti-bonding state and has a
-356-
higher energy level. The 4 valence bands in the band structure of silicon are
built from the bonding states and the 4 conduction bands from the anti-
bonding states.
Fig. 5-27. The E-k relation of Si in the directions [111] and [100], Ec is located at
about 0.85 (2 /a) (1,0,0) or its five equivalent positions in the k-space. Also Ev is
located at (0,0,0) and labeled 2.
The minimum of the principle (lowest) conduction band lies near the point
X towards the [100] direction. As there exist 6 equivalent directions [100],
this conduction band has 6 equivalent minima (valleys), as shown in figure
5-29.
-357-
Fig. 5-28. Hybridization of electron orbitals in silicon crystals.
Near these minima the E-k is quadratic and can be described by the
following relation for the minimum (or valley) situated at (km, 0, 0)
E  Ec 
2
2ml *
2 2
2mt

k x  km   * k y2  k z2  (5-94a)
where ml* and mt* are longitudinal and transverse effective masses. They
are given by:
2 2 2
m  2
*
, m  2
*
 (5-94b)
l
 E  k x2  E  k y2  2 E  k z2
t
Both ml* and mt* can be measured by cyclotron resonance4. Their values
are as follows:
ml*  0.90 mo , mt*  0.192 mo (5-95a)
where the rest mass of electrons in free space mo = 9.11x10-31 kg. The other
minima or valleys have the same shape but different orientations, as shown
in figures 5-29. It is clear in this figure that the minima (edges) of the 6
equivalent valleys lie at E = Ec and kx = 0.85 ko, ky =0.85 ko and kz =
4
The cyclotron resonance is a magnetic field induced physical phenomenon. It will be
discussed later in chapter 7.
-358-
0.85 ko where ko =2 /a. The valley of the second conduction band of Si
(at ) is situated 1eV above the X-valley. The longitudinal and transverse
mass of the L-valley of silicon are as follow at the band edge:
ml*  1.59 mo , mt*  0.121 mo (5-95b)
Fig. 5-29. Constant energy surfaces, of the principal conduction band in Silicon,
The valence bands of silicon are approximately quadratic. The constant

energy surfaces of the two upper bands (for heavy and light holes) are
fluted spheres and may be described by the following relations near k= 0,
E( k )  Ev 
2
2mo
Ak 2
 
 B 2 k 4  C 2 k y2 k z2  k z2 k x2  k x2 k y2 
1
2
 (5-96)
where A, B and C are constants5 and the  signs correspond to light holes
and heavy holes bands, respectively. For the valence band of light holes
(with plus sign and designated by the letter l), the effective mass is usually
denoted mlh*. At the band edge, the light holes mass mlh* is given by:
mlh*  mo A  
B 2  1 6 C 2  0.153 mo (5-97a)
Also, for the valence band of heavy holes (with minus sign and designated
by the letter h), the effective mass is usually denoted mhh*. At the band
edge, the heavy hole mass mhh* is given by:
5
According to Dexter and Coll, we may take A= 4.0  0.1, B =1.1  0.4, C = 4.1  0.4.
-359-
m*hh  mo A  
B 2  1 6 C 2  0.537 mo (5-97b)
(a) light holes (b) heavy holes
Fig. 5-30. Constant energy surfaces in the valence band of Silicon.
The effective mass of electrons and holes in the most common

semiconductors are shown in Table 5-4. Equation (5-96), which describes
the E(k) relation of the two upper valence bands in Si (the l and h valence
bands), is sometimes written using polar k-coordinates in the form:
2k 2
E( k , , )  E v  *
g(  ,  ) (5-98)
2mv
where mv* is the isotropic hole effective mass, as given by (5-97), while
g(, ) contains the l and h valence bands anisotropy information. The third
valence band, which is called the split-off band (s-band), has its top
0.044 eV below the l and h bands. This band is only populated at higher
hole energies, and its E(k) relation may be described by the following
simple parabolic relation:
2k 2
E( k )  E v  s  A (5-99)
2 mo
where s = 0.044 eV is the shift between the top of the s-band and the top
of the l and h valence bands. The effective mass of split-off holes is given
-360-
by mh,so* = 0.234 mo. The effective mass of electrons and holes in energy
bands of Si and other semiconductors is indicated in Appendix D.
Figure 5-31 shows the constant energy surface at E=25 meV below the
valence band edge for the light hole (a) and for the heavy hole (b).
In the unstrained case, both heavy and light hole are degenerate at k=0. The
color bars on the left side of each plot correspond to the 2D slices in these
pictures. The following figure shows the constant energy contours in the
(kx,ky) plane of a 2D slice through the hole light (a) and the heavy hole (b)
E(kx,ky,kz) dispersion at kz=0, i.e. E(kx,ky,0). As shown, the light hole is
much more isotropic than the heavy hole.
(a) (b)
Fig. 5-31. Two-dimensional energy contours in Silicon, of (a) the valence band of light
holes and (b) the valence band of heavy holes.
5-10.2. Case Study 2: Band Structure of Strained Silicon

The SiGe has a lattice constant that is larger than Si, thus Si is strained
when it is deposited over SiGe, as shown in figure 5-32. Stress affects the
E(k) diagram of Si and affects the material properties. Applied strain may
be either: tensile or compression. Also, stress may be biaxial or uniaxial.
i. Effect of Biaxial Stress

For the case of biaxial stress in the (001) plane of Si, the 6-fold degenerate
6-valleys are split into a 2-fold degenerate 2 valley pair (located along
the [001] direction) and a 4-fold degenerate 4 valleys pair. Figure 5-32
shows the effect of tensile stress on the valence bands of Si.
-361-
Fig. 5-32. Effect of biaxial tensile stress on the conduction bands of Si
Fig. 5-33. Effect of biaxial tensile stress on the valence bands of Si
-362-
Figure 5-33 shows the dispersion along the [110] and [100] directions for
biaxially strained Si on Si0.87Ge0.13. Note that, the directions [100], [010], [-
100] and [0-10] are equivalent. Therefore, if the biaxial stress is directed
along one of these directions, the result is the same in the biaxially strained
Si. However, the heavy and light holes are now mixed and there are no
purely heavy hole (HH) nor light hole (LH) states. At the zone center
(k=0), the highest valence band has light hole character. Thus the
degeneracy of the HH and LH bands at the  point (k=0) is lifted.Due to
the positive hydrostatic strain we obtain a reduced band gap with respect to
the unstrained Si. Furthermore, the degeneracy of the heavy and light hole
at k=0 is lifted by 69 meV and the light hole has been shifted above the
heavy hole. Also, the anisotropy of the holes along the different directions
[100] and [110] is very pronounced. The biaxial strain can be expressed as
applied stress (ij) in units of GPa.
Fig. 5-34. Effect of biaxil tensile stress on the valence band of Si
The figure 5-35 shows the constant energy surface at E=25 meV and 94
meV below the valence band edge (a) and the constant energy surface
below the valence band edge (b). The degeneracy of heavy and light hole at
k=0 is lifted by 69 meV. The figure 5-36 shows the constant energy
contours in the (kx,ky) plane of a slice through the ground state hole (a) and
the first excited hole state (b) E(kx,ky,kz) dispersion at kz=0, i.e. E(kx,ky,0).
-363-
At figure 5-36(a) one can see that around k=0, the highest hole state has
light hole character and for larger k values the heavy hole character
dominates. At figure 5-36(b), the second highest hole state is shown which
is separated by 0.069eV from the light hole (valence band edge). Around
k=0 it has heavy hole character and at larger k values it has light hole
character as can be understood by comparing these plots to the unstrained
dispersion curves.
Fig. 5-35. Effect of tensile stress on the constant enegy surfaces of Si valence bands.
(a) valence band of light holes and (b) valence band of heavy holes.
Fig. 5-36. Effect of tensile stress on the E-k contours of Si valence bands.
(a) valence band of light holes and (b) valence band of heavy holes.
-364-
ii. Effect of Uniaxial Stress

We know that the diamond lattice of unstrained (relaxed) silicon is
composed of two interpenetrating face-centered cubic (FCC) lattices. The
number of symmetry operations in this structure is 12. Under the
application of a uniaxial stress along the [111] direction, the FCC lattice of
unstrained Si is modified to a triagonal system (rhombohedron with edges
of arbitrary angles). The number of symmetry operations is then reduced by
a factor 4 compared to the unstrained case. Because this direction coincides
with the diagonal of the cube, all the 6-valleys are symmetrically oriented
with respect to this direction. The resulting strain tensor thus has equal
magnitudes of the diagonal components. Shear strain components are also
present with xy = yz =xz. Consequently, the degeneracy of the valleys is
not lifted and electrons are populating all the valleys equally. In the valence
bands, the energy dispersion in the [100] and [001] directions are the same.
Fig. 5-37. Effect of uniaxial tensile & compressive stress on the energy bands of Si.
-365-
The hydrostatic shift and uniaxial splitting of the conduction and valence
bands for (100) substrates are shown in the following figure for both
compressive and tensile strain. As, shown in figure, uniaxial strain splits
the degeneracy of bands e.g. the valleys in the conduction band of silicon
and the degeneracy of light holes (LH) and heavy holes (HH) valence
bands. The deformation potentials of silicon (which can be measured
experimentally) determine the exact amount of splitting of energy bands.
5-10.3. Case Study 3: Band Structure of SiC

Silicon carbide (carborundum) is a very promising material and a
considerable effort is currently undertaken to create high power high
frequency devices using this material. Silicon carbide crystallizes in
different (about 250) modifications or polylypes. However, the SiC crystals
have two famous structures, namely -SiC and -SiC. Alpha silicon
carbide (α-SiC) is the most commonly encountered polytype; it is formed at
temperatures greater than 1700 °C and has a hexagonal crystal structure
(similar to Wurtzite). All other non-cubic silicon carbide structures (e.g.,
hexagonal structures, like 6H-SiC) are referred to as -SiC. The beta silicon
carbide (β-SiC) is sometimes called 3C-SiC This is the only SiC cubic
structure.. It has a zincblende crystal structure. The -SiC has a repetitive
sequence of covalent bonded tetrahedra of SiC4 and Si4C. Figure 5-38
depicts the main polytypes of crystalline SiC. The following figures depict
the main features and the detailed band structure of 3C-SiC.
3C-SiC ( SiC) 6H-SiC ( SiC)
4H-SiC
Fig. 5-38. The main polytypes of SiC.
-366-
Fig. 5-39. The main features in the E-K diagram of 3C-SiC.
Fig. 5-40. The detailed band structure of 3C-SiC.
-367-
5-10.4. Case Study 4: Band Structure of GaAs

It is well known that the semiconductor III-V compounds crystallize in the
zincblende (spharelite) structure. The lattice is composed of two
interpenetrating face-centered cubic (FCC) lattices6. The reciprocal lattice
is a body-centered cubic (BCC) lattice. The first Brillouin zone of such
compounds is a truncated octahedral. The band structure of GaAs is shown
in figure 5-41. Like all III-V compounds, the top of the valence band of
GaAs is located in the center of Brillion zone (  point). Also, the
conduction band edge of III-V materials is found either at the  point or
near the L or the X point. The valence bands of GaAs are almost quadratic.
The constant energy surfaces of the two upper bands of light and heavy
holes ( h) are fluted spheres. The third valence split-off band (s-band),
has its top 0.034eV below the l and h bands.
Fig. 5-41. The main features of the E-k relation of GaAs. The top of valence band Ev is
located at the center of the Brillouin zone (0, 0, 0) and labeled . The bottom of
conduction band Ec is located at , directly above the valence band.
6
Note that we consider that zincblende structures as two inter-penetrating face-centered cubes (FCC),
separated by ¼ the cube length.
-368-
(5-100a)
(5-100b)
Fig. 5-42. Detailed energy band structure of GaAs..
-369-
The following figure shows the 2-dimensional constant energy contours of

the heavy holes valence band and Conduction band constant energy surface.
For more details about the different bands of the GaAs, and other compound
semiconductors, refer to Appendix D.
(a) (b)
Fig. 5-43. (a) Contours of constant energy surfaces of the heavy holes valence band.
(b)Conduction band constant energy surface
5-11. Hole Concept

The phenomenon of conduction is of major interest in the study of solid-
state physics in general and for the study of semiconductors in particular.
As we have seen so far, each energy band in the band structure of a given
solid may be filled (occupied) or empty or partially filled. Each simple
band cannot contain more than 2M electrons where M is the number of
atoms in the crystal. If the number of electrons in a simple band is N, then
the average current density which is carried by these electrons is given by:
e N
J  vgi
V i 1
(5-101a)
where V is the solid volume. If the band is completely filled, than no

conduction takes place and we have:
-370-
e N
J  vgi  0
V i1
(5-101b)
If the Valence band has P empty states, then the corresponding current
density is given by:
e N  P e P
J   v gi   v gi (5-101c)
V i 1 V i 1
An empty state in a filled band is called a "hole". According to equation

(5-91), the hole can be considered as a positive charge carrier (with positive
charge +e). Therefore, when a hole is influenced by an external field, it
will be moving in the direction of field (against electrons). Also, the hole
has the following characteristics which relate its velocity vp, effective mass
mp*, wave vector kp and energy Ep to the corresponding missing electron
parameters vn, mn*, kn and En This is because the total wave vector of a
filled band is zero. Therefore, for holes we have:
k p
  kn (5-102a)
N N
 ki  
i 1 i 1
p i
 0, (5-102b)
E p
  En , (5-102c)
*
  mn ,
*
(5-102d)
m p
v p
  vn (5-102e)
Note 5-6. Hole Current

Relation (5-91) can be written in the following form:
N1 N

J  e 
V N
v
i 1
gi   e n vn

where n=N/V is the density of electrons and vn is their average velocity in
the band. When a band is filled and all levels are occupied then the average
velocity of electrons is zero (vn=0). Consequently, the average current due
to these electrons is null.
-371-
Notice that m*n at the top of a band (a valence band) is negative, so that m*p
is positive. Finally, it should be emphasized that, it is more convenient to
talk about holes rather than electrons in valence bands, like speaking about
a bubble in a spirits. The reason the concept of holes simplifies the analysis
is that the density of states function of a whole band can be rather complex.
However, it can be dramatically simplified if only states close to the band
edge need to be considered.
Because of the anisotropy of the effective mass and the presence of

multiple equivalent band minima, we define two types of effective mass,
the effective mass for density of states calculations and the effective mass
for conductivity calculations. Electrons in GaAs have an isotropic effective
mass so that the conductivity effective mass equals the density of states
effective mass. The effective mass values for electrons and holes are listed,
with the value of the smallest energy gap, in Table 5-5.
Table 5-5 Energy gap and effective mass of carriers in Ge, Si and GaAs, at 300K.
Symbol Ge Si GaAs
Minimum Energy Gap Eg 0.66 1.12 1.424
Electron density of states effective mass m*nd /mo 0.55 1.08 0.067
Hole density of states effective mass m*pd /mo 0.37 0.811 0.45
Electron conductivity effective mass m*nc /mo 0.12 0.26 0.067
Hole conductivity effective mass m*pc /mo 0.21 0.386 0.37
-372-
5-12. Density of States in Energy Bands

As we have shown earlier, the k-space is quantized to discrete points (or
states) because of the finite crystal dimensions. Each point in the k-space
corresponds to a vector in the k-space. The points of these vectors can only
be situated at the corners of a simple reciprocal unit cell of volume vk
 2  3
vk  (5-103a)
V
where V is the direct crystal lattice volume. Each of these elemental cells
in the k-space corresponds to an energy level. As each energy level can
accommodate two electrons of two different states (with different spins),
then the density of states in the k-space is given by:
1 V
G ( k)   (5-103b)
1 2 vk 4 3
It follows from the above relation that the total number of electrons in a
certain band is given by:
 G k f k d k
3
N (5-104a)
Band
However, as the distribution function in quantum physics is usually

expressed as a function of energy f(E), it is preferable to have another
expression for the density of states per unit energy, G(E), so that:
N  G ( E) f ( E)dE (5-104b)
Band
It follows from the above two relations that:
G E  dE   G k  d k
3
(5-105a)
Cons. E . Surface
This relation is usually written in the following integral form:
G E  
2V ds

2  3
 
Const. E. Surface | k E k |
(5-105b)
-373-
where dS is the elemental surface area of a constant energy surface in the k-

space, as shown in figure 5-44. Also, the integral is taken over a surface of
constant energy.
Constant kz dS
energy surface,
E = Constant dk
ky
kx
Fig. 5-44. Elemental surface area on a constant energy surface.
In order to calculate the density of states in a certain energy band, one has
to know the E-k relation and the shape of the constant energy surface of
this band. G(E) is simply the fractional number of states between the two
energy surfaces E and E+ dE. For instance, for a spherical energy band we
have S = 4 k2 such that:
32
 2m *
G(E) spheric 
V
  E 12
(5-106a)
2  2 
2
There exists another more convenient form of the density of states, which
does not need the knowledge of the crystal lattice volume. The density of
states per unit energy per unit volume is expressed as g(E) = G(E)/V.
g E  
1 ds
2 3   
Const. E. Surface |  k E k |
(5-106b)
-374-
For the matter of illustration, the electrons in a metal have a parabolic

energy relation E(k), as given by (5-47), and hence their density of states
can be readily calculated as follows:
4 (2mo )3 / 2
g E  
2   3
. E (5-107)
3
It should be noted that a factor of 2, corresponding to 2 electrons per

energy level, is sometimes explicitly put in the integral before g(E). In this
case g(E) is merely the density of energy levels in the band. The actual
form of g(E) is complicated and may contain several critical points (where
k E =0). Evidently, the band edges (minima or maxima) represent critical
points.
The typical shape of g(E) in non-overlapped bands is illustrated in the

figure 5-45. As shown in figure, the density of states in the conduction
band starts to decrease at threshold energy Eth. In silicon, we have Eth =2 eV
and 1.8 eV in GaAs. Near band edges, the constant energy surfaces are
generally quadratic and g(E) will have either of the following forms:
g E    E  Ec 1 2 near a minimum
(5-108)
g E   E v  E 
12
near a maximum
g(E)
Maximu Minimu
m m
Valence Conduction
Band Band
E
Ev Ec Eth
Fig. 5-45. Typical form of the density of states in non-overlapped energy bands.
-375-
Therefore, in a parabolic band, the density of states in three dimensional k-

space, is proportional to (E-Ec)1/2. Also, one can prove that the density of
states is independent of energy in two dimensional structurs. In addition,
the density of states is proportional to (E-Ec)-1/2 in one dimensional
structures, as shown in the figure 5-46.
E(k) E(k) E(k)
3D
Ec Eci Eci
kx, ky, kz kx, ky kx
g(E)  E1/2 g2D(E) = constant g1D(E)  E-1/2
2D
3D 1D
E E E
Ec Eci Eci
3-dimensional 2-dimensional 1-dimensional

semiconductor semiconductor semiconductor
Fig. 5-46. The E-k relation and density of states in 1-D, 2-D and 3-D semiconductors.
The following table summarizes the relations describing the dispersion

relations and density of states (DOS) of parabolic semiconductors of
different dimensions. More details about the so-called “low-dimensional
semiconductors” will be given later at the end of this chapter.
-376-
Table 5-6. Dispersion relations and density of states (DOS) of parabolic semi-
conductors of different dimensions. Note that u(E-Ec) is a unit step function. Here Ec
and Eci denote the conduction band (or subband) edge.
Dimension Dispersion E(K) Dos g(E)

3D (Bulk) E=Ec -[ħ2/2m*](kx2+ky2 +kz2) [(2m*/ħ2)3/2/22] .(E-Ec)1/2
2D (Well) E=Eci - [ħ2/2m*].(kx2 + ky2) [m*/ħ2]. u(E-Ec)
1D (wire) E= Eci -[ħ2/2m*].(kx2 ) [√2.m*3/2/h].(E-Ec)-1/2
0D (Dot) - 2δ(E-Ec)
Because of the geometrical effect of a two-dimensional surface, atoms at a

surface of a semiconductor have fewer neighbors than similar atoms in the
bulk. Therefore, the electronic energy levels at surfaces are different from
what they are in the bulk. For example, a silicon atom in bulk silicon is
surrounded by four tetrahedral bonds. On the surface, a silicon atom will
have fewer bonds, and the surface valence electrons that do not participate
in bonding are described as dangling bonds. These surface valence
electrons give rise to new electronic states called surface states.
Fig. 5-47. Density of states at the surface of a semiconductor (surface states)
-377-
5-13. Classification of Solids

The observed conductivity of solid materials covers a very wide range
(1:1020), as shown in figure 5-48. The band theory explains this wide range
of variation in terms of the density of mobile electrons, which lie in
partially filled bands.
Metals Semiconductors Insulators

Resistivit
y
10-8 10-4 104 108 1012 1016  cm
Fig. 5-48. Classification of solid materials according to their resistivity.
It follows from the foregoing discussion, that energy bands may be filled or
empty or partially filled according to the atomic structure of the constituent
atoms of the solid material. Accordingly, the solid-state materials may be
classified into two categories:
1- Conductors (metals and semi metals),

2- Non conductors (insulators and semiconductors).
Metals are substances with partially filled (or overlapping bands). Free
states are available because the number of available states is greater than
the number of electrons and bands may overlap, as shown in figure 5-42.
Metals, whose conduction band is partially filled with slight overlapping,
are sometimes called semimetals. Therefore, the Fermi energy in metals
separates occupied states from non-occupied states. Thereore, the shape of
the Fermi surface in metals define the metal properties. Non-conductors
contain empty (conduction) bands and completely filled (valence) bands at
0K. Non-conductors may be further classified into insulators and
semiconductors.
The energy gap in semiconductors is usually in the order of 1 eV, as shown

in Table 5-7. Hence, the thermal excitations can raise a few electrons from
valence band to conduction band in semiconductors. Therefore, the
properties of a semiconductor, is determined from the shape of the energy
band diagram near bottom of conduction band and top of valence band. On
the other hand, insulators are characterized by a huge energy gap in the
order of 5eV.
-378-
E E
C.B.
Ev C.B. Ec
Eg
Ec V.B. Ev
V.B.
Semimetals Metals
(with overlapped bands) (Non overlapped bands)
(a) Conductors.
E E
C.B. C.B.
Ec Ec
Eg ≈ 1 eV Eg > 5eV
Ev Ev
V.B. V.B.
Semiconductors Insulators
(b) Non-Conductors.
Fig. 5-49. Classification of solids to metals, semiconductors and insulators.
Table 5-7. Energy gap and resistivity of some materials.
Eg (eV)  (.cm)
Material Category
0K 300K 300K
SiO2 Insulator 9 9 10 -1016
14
Si3N4 Insulator 5 5 1014

GaN Compound Semiconductor 3.36 3.3 1011
Poly acetylene Polymer Semiconductor 2.5 2.5
Ga As Compound III-V Semiconductor 1.52 1.43 108
Si Element IV Semiconductor 1.17 1.12 2x105
Ge Element IV Semiconductor 0.74 0.66 44
In As Compound III-V Semiconductor 0.42 0.36 0.3
Al Metal - - 2.64x10-6
Cu Metal - - 1.67x10-6
-379-
5-14. Classification of Semiconductors

Semiconductor materials can be classified according to their constituent
atoms to:
Elemental semiconductors (e.g., Si, Ge, and diamond),
Compound semiconductors (e.g., GaAs, InSb, AlGaAs, AlGaAsSb).
Simple or elemental semiconductors are all from the IV group in the

periodic table (e.g., Si and Ge) and crystallize in a diamond structure.
Compound semiconductors are composed of atoms of two or more

elements, most frequently from the III and V groups (III-V compounds,
like GaAs) or the II and VI groups (II-VI compounds, like CdS, ZnS and
ZnTe). Also some compound semiconductors are composed of two or
more elements from the IV group (IV-IV compounds, like SiC). When
compound semiconductors are formed from two elements, such as GaAs,
they are called binary compounds.
Many of the III-V compound semiconductors crystallize in the zincblende

structure, while others crystallize in the Wurtzite or rock salt structures.
Note that Zincblende is a cubic structure while Wurtzite is non-cubic.
Some II-VI compound semiconductors, like cadmium sulfide CdS,

crystallize in the HCP (hexagonal closed packing) Wurtzite structure,
while others, like zinc telluride ZnTe, crystallize in the zincblende
structure.
Some IV-IV compound semiconductors, like silicon carbide SiC crystallize

in close-packed structures having a special one-dimensional polymorphism,
called polytypism. Polytype crystals are identical in the two dimensions of
the close-packed planes and different in the perpendicular stacking
direction. The SiC crystals have two famous structures, namely -SiC and
-SiC. The -SiC is a poly-type of SiC, which has a repetitive sequence of
covalent bonded tetrahedra of SiC4 and Si4C. Also, -SiC has a zinc-blende
structure. The -SiC is sometimes called 3C-SiC. This is the only SiC cubic
structure7. All other non-cubic structures (e.g., rhombohedral or hexagonal
structures, like 6H-SiC) are referred to as -SiC.
7
Note that we consider that zincblende structures as two inter-penetrating face-centered cubes (FCC),
separated by ¼ the cube length.
-380-
Fig.5-50. Main semiconductor elements and compounds in the periodic table and there
typical energy gaps
-381-
When compound semiconductors are formed from more than one group III
elements (randomly distributed on III sites) or more than one group V
elements (randomly distributed on V sites), these compounds are called
crystalline solid solutions.
Crystalline solid solutions that have 3 elements are called ternary

compounds, and have the notation AxB1-xC (e.g., AlxGa1-xAs). Note that x
and 1-x are the mole fractions of the first two elements, which belong to the
same group, in the crystal solution. Also, crystalline solid solutions that
have 4 elements are called quaternary compounds, and have the notation
AxB1-xCyD1-y (e.g., GaxIn1-xAsyP1-y). Here, the subscript x is the mole fraction
of Ga in the group III elements and the subscript y is the mole fraction of
As in the group V elements. For more details about the variation of Eg with
mole fraction x, refer to table 5-8.
Semiconductors can be also classified according to their bandgap width as

follows:
Narrow-gap semiconductors (like PbS, PbTe, and SnTe),
Wide-gap semiconductors (like ZnS, GaN and SiC).
Narrow gap semiconductors are frequently used in microwave and infrared

detectors. Interestingly, a certain number of semiconductors, like HgTe, are
referred to as null-gap semiconductors. Such materials are classified as
semiconductors because they do not have partially filled bands at 0K.
As for wide-gap semiconductors, they are currently used in power devices,

optoelectronics and space applications. Interest in Group-III nitrides (like
GaN) is motivated by the possible use of these materials in optoelectronic
devices which operate in the blue to UV region of the electromagnetic
spectrum. As mentioned above, the Group-III nitrides are normally grown
in the wurzite structure, but can also be produced in the zincblende
structure, which is more amenable to p-type doping and device fabrication.
-382-
5-15. Direct and Indirect Gap Semiconductors

One of the most important observations on the E-k diagram of certain
semiconductors is that the valence band maximum may be found at a
different value of k than the principal conduction band. For instance, the
silicon material has its top valence bands at k =0, while the bottom of
principle conduction band lies at k =0.85 (2/a) in directions [100], as
shown in figure 5-51. Such semiconductors are called indirect-gap
semiconductors.
The electron transitions from the minimum point in the conduction band to
the maximum point of the valence band in indirect-gap semiconductors
require some change in k. Therefore, the conservation of crystal momentum
(or wave vector) requires a phonon emission (or absorption) during the
band-to-band transitions in indirect-gap semiconductors. This is only
possible if there are occupied electron and hole states directly on top of
each other. This is why the Si, which has an indirect gap, is not used as a
light source in optoelectronic devices.
However, there exists other type of semiconductors, like GaAs, whose

energy gap is direct. That is to say, the top of valence band and the bottom
of conduction band are situated at the same k-value. So, electrons, which
make transitions from the conduction band to the valence band, do so
without change in their k-value. This type of transitions is usually
associated by an emission of photons.
E E
C.B. C.B.
0.3 eV
Ec Ec
phonon
photon Eg =1.42 eV
Eg =1.12 eV
Ev Ev
V.B. V.B.
[111] [100] [111] [100]
k k
GaAs Si
Fig. 5-51. Electron transitions in direct (GaAs) and indirect (Si) semiconductors.
-383-
5-16. Energy Band Alignment in Heterostructures (A)

Heterojunction devices are special category of semiconductor devices,
which consist of several layers of different semiconductors, with different
energy gaps. The energy bands alignment at the interface of heterojunctions
may be explained by the band affinity model (Anderson’s rule). This is
based on the line-up of the electron affinity χ of materials in contact.
The electron affinity is the energy required to take an electron from the
bottom of the conduction band to the vacuum level where it can escape
from the crystal. According to Anderson’s rule, the vacuum level of the
two materials of a heterojunction should line up, such that :
Ec = χ = χ1 - χ2 & Ev = Eg2 - Eg1 - Ec (5-109)
For instance, if we have a GaAs - Al0.3Ga0.7As heterojunction, where the

GaAs has χ = 4.07 eV and the Al0.3Ga0.7As has χ = 3.74 eV, then we predict
that Ec= χ = 0.33eV at their interface. Also, the energy gap changes by
Eg =1.79-1.42 = 0.37eV, and hence Ev = 0.04eV, as shown in figure 5-
52. More examples are shown in figure 5-53.
Fig. 5-52. Alignment of energy bands at a heterstructure interface, according to the

band affinity model (Anderson’s rule).
Note 5-7 Band Alignment Theories

Actually, there exist 3 theories, for the heterojunction energy bands alignment. The
Anderson rule (the band affinity model), is widely-accepted in the literature. It dictates
the align-up of the vacuum level of all materials in contact. Other theories presume the
align-up of the conduction-band edge, or the intrinsic level, which is mid-way between
the conduction band and the valence band edges, of the two semiconductors in contact.
-384-
Fig. 5-53. Alignment of energy bands at the interface of GaAs - Al0.3Ga0.-As hetero-
junction, where Egg= 1.422 + 1.247x. For x = 0.3, Ecc = 0.33 eV, Eg = 0.37 eV and
Evv = 0.04 eV at 300K.
Table 5-8. Variation of important parameters of AlxGa.As1-x as function of the mole

fraction x, for x < 0.45 and T = 300 K
Parameter Formula
Energy gap [eV] Eg =
 Valley = 1.422 + 1.247x
L Valley = 1.707 + 0.642x
X Valley = 1899 + 0.125x
Electron effective mass mn*/mo =
 Valley = 0.067 + 0.083x
L Valley (total) = 0.55 + 0.12x
X Valley (total) = 0.85 - 0.07x
Hole effective mass mp*/mo = 0.48 + 0.31 x

Dielectric constant r = 13.1 - 3x
-385-
Fig. 5-54. Bandgap offset of the hetero-junctions of some semiconductors.
-386-
5-17. Graded Bandgap Semiconductor Structures (A)

Some devices need alloy systems with very low strain-induced defects.
This can be achieved in some compound semiconductors, like AlxGa1-xAs,
GaxIn1-x, AsxP1-x and GaxIn1-xAsyP1-y, by grading the mole fraction indexes x
and y. The resultant material is a graded band gap semiconductor. It is
interesting to show that some compound semiconductors can be changed
from direct to indirect gap according to the mole fraction of their
constituent elements. For instance, figure 5-55 depicts the band structure of
of GaAs1-xPx at different values of x. As shown in figure, the energy gap
height Eg of such a compound semiconductor varies piecewise linearly with
the mole fraction x such that:
Eg(x) = Eg + 1.25 x for 0<x<0.45 (5-110a)
Eg(x) = EgX - 0.45 x for 0.45<x<1 (5-110b)
where Eg = 2.42eV and EgX = 2.26eV are the extreme values when x=0
(GaAs) and x=1 (GaP), respectively. When x=0, the compound is simply a
GaAs direct gap semiconductor having Eg=Eg=2.42eV. Also, the material
changes from direct band gap to indirect band gap semiconductor for
x>0.45. When x is further increased to 1, the compound becomes a GaP,
which is an indirect gap semiconductor having Eg =EgX =1.26eV.
Eg (eV) E x  X
3.0 =1.0
x
=0.4 C.B.
2.7 x=
0.0
Ec
2.4
1.42 2.2
Indirec eV eV
2.1
t Ev
1.8 Direct
V.B.
1.5 GaAs GaP [111] [100]
0 0.2 0.4 0.6 0.8 k
1.2 1.0 x
Fig. 5-55. Band structure dependence on the mole fraction x of the GaAs1-xPx .
-387-
Other semiconductor compounds, like AlxGa1-xAs, show the same behavior

and their energy gap height can be adjusted by the stoechmetric change of
the alloy. Such materials enabled us to construct tailored (or engineered)
semiconductor structures, which have some particular band structure. The
so-called graded bandgap semiconductors are constructed by arranging
several adjacent layers of slightly different mole fraction. Such structures
have many applications in imaging and optical devices. As we’ll illustrate
in chapter 14, such a process could be performed by various epitaxial
techniques like the molecular beam epitaxy (MBE).
Fig. 5-56. Energy band diagram of a graded bandgap semiconductor.
It should be noted that there exist a strong correlation between the lattice
spacing constant (lattice parameter) of ternary compound semiconductor
and their energy gap height. Consequently, the problems of crystal
mismatch and strain induced defects in various heterojunction devices 8 can
be overcome by use of graded bandgap structures. Therefore, in graded
energy-gap semiconductors and alloys not only the energy gap, but also all
the other physical parameters such as the effective mass and the dielectric
constant change with the composition mole fraction. Table 5-8 depicts the
formulae describing such parameters in AlxGa1-xAs.
8
As we stated above, Heterojunction devices are those devices employing different semiconductor
materials within the same device.
-388-
Fig. 5-57. Correlation between the lattice constant and the energy gap of some
compound semiconductors.
-389-
5-18. Superlattices & Bloch Oscillations (A)

Superlattices are special layered semiconductor structures, whose energy
band diagram is deliberately engineered to look like that of periodic square
potential well. These structures are usually constructed by arranging
several thin layers of small gap and large gap semiconductor layers. The
thickness of these stacked layers, which correspond to the width of
resulting square potential wells, is made very small (in the order of de-
Broglie wavelength) to enable tunneling across different layers.
The superlattices, which are constructed from different semiconductors in

this manner, are hetero-junctions. Superlattices may be also implemented
by repeated “nipi” structure of a single semiconductor (homo-junctions).
Such structures have many applications in optical and microwave devices,
particularly the so-called resonant tunneling devices (RTD’s) and the real-
space-transfer (RST) devices.
y Material I (Wider gap)

z
V(x) a
ECI – ECII
EgI EgII
I II I II I EvI – EvII
x x
a
Material II (Narrower Gap)
Fig. 5-58(a). A superlattice with periodic square barrier.
The effect of the superlattice on electron transport is to divide the

conduction band into mini bands and mini gaps. An electron on the top or
in the bottom of a mini band will have a zero velocity. If scattering is
completely neglected, electrons in mini bands will oscillate with frequency
controlled by the time needed by the electron to make a tour of the first
Brillouin zone. This phenomenon is called Bloch oscillation. It has been
observed experimentally in 1983 by T. Sollner et al and exploited for the
detection and mixing of 2.5 THz oscillations. In order to visualize the
Bloch oscillation, consider the electrons inside a thin semiconductor layer
(of thickness a, as shown in figure 5-58), which is subjected to an electric
field ζ such that:
-390-
dk x e
  (5-111a)
dt 
Consider an electron at t=0 with k=0. Assume the electron is drifted during
a time t under the electric field effect. Then we have:
e
k x ( t )    xt (5-111b)

When the electron reaches k=/a it is Bragg-reflected to k=-/a. Then, it
moves from -/a to /a again. The period of motion is given by:
 2
T . (5-112)
e x a
As shown in figure 5-58, the semiconductor layer thickness a and electric

field ζ should be controlled such that the period T becomes smaller that the
mean free time between collisions (T < ) for sustained oscillations.
Fig. 5-58(b). Illustration of the Bloch oscillations
In real-space-transfer devices (RTD’s), electrons are heated up, by the

electric field effect, such that they transfer from the wide-gap layers (e.g.,
AlGaAs) to the narrow-gap layers (e.g., GaAs), where their mobility is
substantially lower, giving rise to a negative differential conductivity and a
possibility of oscillation.
-391-
5-19. Low-Dimensional Semiconductors (LDS’s)

We have learned so far that conduction electrons in the bulk of a
semiconductor can move freely, in all the three dimensions of the physical
space. However, in certain semiconductor devices, free electrons are only
permitted to move in one or two dimensions. Such semiconductors, are
called low-dimensional semiconductors (LDS’s). For instance, the
electrons filling the thin inversion layer in a conventional MOSFET
(Metal-Oxide-semiconductor Field Effect Transistor) are only permitted to
move freely in the 2 dimensions of the inversion layer. The motion of free
electrons in the third dimension (which is normal to the thin layer plane) is
confined and quantized.
The semiconductor structure is considered as a quasi two-dimensional

Q2D, when one of its three spatial dimensions is getting so small (in the
order of de-Broglie wavelength) compared to the other two dimensions.
The gas of electrons filling such structures is sometimes called two-
dimensional electron gas (2DEG) or a quantum well. Thus, the electrons
filling the thin inversion layer in conventional MOSFET devices act as
quantum potential wells and form a 2DEG. Quantum wells are also usual in
hetero-junction semiconductor devices. On the other hand, the structure is
considered as a quasi-one-dimensional Q1D, when two of its three spatial
dimensions are so small compared to the third dimension. Such structures
are sometimes called quantum wires. Many LASER devices are currently
built from 1-D structures, including nanowires and nanotubes. Also, if all
the three dimensions are so small the structure is called a quasi-zero-
dimensional Q0D structure or a quantum dot.
The idea behind confinement is all about keeping electrons trapped in a

small area. The size we are talking about for confinement is in the order of
10 nm. One is probably wondering why confinement is so important. For
one thing, it leads to new electronic properties that are not present in
classical 3-D semiconductor devices. Consider the quantum dot. The
typical quantum dot is anywhere between 3-60 nm in diameter. That’s still
30 to 600 times the size of a typical atom. A quantum dot exhibits 0-D
confinement, meaning that electrons are confined in all three dimensions.
The only things in nature that have true 0-D confinement are atoms. So a
quantum dot can be loosely described as an artificial atom. One can easily
deduce the density of states (DOS) and the density of electrons n in LD’s,
by the aide of equation (5-94). For simplicity, let’s assume a parabolic E-k
relation for electrons in such structures.
-392-
5-19.1. 0D Structures
As we mentioned above, the quantum dots, are three-dimensionally
constrained semiconductor nanostructures, whose dimensions are in the
order of de-Brouglie wavelength (1-100 nm for electrons). Quantum dots
are usually fabricated from compound semiconductors (e.g., InGaAs),
which are used in optical and fluorescent applications (e.g., Laser and solar
cells).
Quantum dots also have other applications in biological tagging and

quantum computing. In fact the quantum dots are 2-D analogies for real
atoms and suitable for atomic physics experiments, that cannot be done on
real atoms. For instance, a magnetic flux quantum (fluxoid) in an atom
requires a very high magnetic field of 1M Tesla, while for a quantum dot it
is about 1Tesla, which is accessible. Therefore, quantum dots can be
thought of as artificial atoms. As shown in figure 5-59, the fabrication of
quantum dots may be carried out by creating different semiconductor layers
on top of each other, using one of the modern epitaxial techniques. These
techniques are detailed in chapter 10.
g(E)
Q0D
Fig. 5-59. The E-k relation and density of states in zero-dimensional semiconductors.
Cr
Cr Cr AlGaAs
AlGaAs AlGaAs GaAs
GaAs GaAs
AlGaAs
Fig. 5-60. Array of quantum dots on top of AlGaAs substrate.

-393-
The one-dimensional structures, such as quantum wires, are sometimes
called quasi-one-dimensional (Q1D) devices or one-dimensional electron
gas (1DEG). The motion of electrons in Q1D structures is free in one
dimension and quantized in the other two dimensions. Therefore, the E-k is
transformed to a series of sub-bands, as shown in figure 5-61. In one-
dimensional structures, the density of states (per unit length per unit
energy) gc1(E) is given by:
mn*
g c1 ( E ) 
1
E  Ec 1/ 2 (5-113)
 2 2
In two-dimensional structures, the density of states (per unit area per unit
energy) gc2(E) is independent of energy and given by:
mn*
gc2 (E)  (5-114)
 2
Note that, in Q2D structures, the motion of electrons is only free in two
dimensions and quantized in the third dimension. So, the E-k is transformed
to a series of sub-bands, as shown in figure 5-59.
Hence, the two-dimensional electrons density is related to the Fermi energy

and is given by the following relation in a given sub-band:
m*n k BT   Ec  E F 
n2 d  .ln 1  exp   (5-115a)
 2   k BT 
Therefore, the total density of electrons (in all sub-bands) at equilibrium is

given by the following sum:
mi* k BT   Eci  EF 
n2 d  .ln 1  exp
 
 (5-115b)
i   2
  k BT 
-394-
where the sum (for i=0 to ∞) is taken on all sub-bands and Eci is the edge of
the ith sub-band of the conduction band. If the conduction band of the
semiconductor has several equivalent valleys, then:
Mc mij* k BT   Eij  EF 
n2 d   .ln 1  exp  

(5-115c)
j 1 i  2
  k BT 
where the first sum (for i=0 to ∞) is taken on all sub-bands and the outer
sum (for j=1 to Mc) is taken on all equivalent valleys of the conduction
band. Also, Eij is the energy of the ith sub-band in the jth valley.
In the case of silicon, we’ve six equivalent valleys (Mc = 6). In silicon
inversion layers, which are usually induced in MOSFET devices 9, we’ve j
denotes the 4 transverse direction valleys (parallel to <100>) and the 2
longitudinal direction valleys (perpendicular to <100>).
mt* k BT   E  EF 
n2 d  4 .ln 1  exp  ci  
i  2
  k BT 
(5-115d)
m k BT *
  Eci  EF 
2 l
. ln 1  exp  
i  2
  k BT 
where mt* =0.19 mo and ml* =0.916 mo. If the 2DEG potential well is
triangular, we can express the energy of the ith sub-band, Eci, in terms of the
quantum number i by the following relation:
1/ 3
 h 2   3 e 
2/3
Eci  Ec (0)    
* 
 i  14 2 / 3 (5-116)
 2m    2 
where ζ is the applied electric field and Ec(0) is the conduction band edge
of the 2DEG structure (e.g., of the silicon inversion layer). Also m┴*
denotes the effective mass in the perpendicular direction. So, when j
denotes the transverse direction valleys, then m┴* = ml*, and when j denotes
the longitudinal direction valleys, then m┴* = mt*.
9
Metal-Oxide-semiconductor (MOSFET) devices are usually fabricated on <100> silicon wafers.
-395-
In the above relations we assume the Fermi-Dirac statistics, which are valid
in thermal equilibrium. We also assume that the electron motion is bound
in the perpendicular directional to the 2-dimensional space. A more general
relation of the electron density should take the probability of finding an
electron along the perpendicular direction, so that:
mi* ( z )k B T   Ei  E F 
    
2
(5-117)
n2 d ( z ) . ln  1 exp 
 2 i
i    k BT 
where i(z) is the electron wavefunction in the i th subband and |i(z)|2 is

the probability of finding an electron, of the i th subband, along the z-
direction.
E(k) E(k) E(k)
3D Q2D Q1D
kx, ky, kz kx, ky kx
g(E)  E1/2 gc2(E) gc1(E)

Q2D
Q1D
3D
E EC EC1 EC2 E E
Ec Ec1 Ec2
3-dimensional Quasi 2-dimensional Quasi 1-dimensional

Fig. 5-61. The E-k relation and density of states in low-dimensional semiconductors.
-396-
Note that the effective mass of the i th sub-band, mi*, should be taken as
function of z, in heterostructures of several 2-dimensional layers of
different semiconductors. Therefore, the electron wave-functions in each
sub-band can be calculated by solving the Schrödinger equation (SE) in the
perpendicular direction.
 2 d  1 d i 
  *   eV ( z )  Eci ( z ) i  Eci i (5-118)
2 dz  mi ( z ) dz 
where Eci is the band discontinuity, in the case of multiple semi-conductor

layers, along the z-direction. Also, the potential energy V(z) can be
calculated by solving the Poisson equation.
5-19.4. Example of 2D Semiconductor (Quantum Well)

The following figure shows a heterojunction of a 2D semiconductor,
sandwiched inside another semiconductor with wider energy gap. Note the
discrete energy levels and the stair case density of states, inside the 2D
semiconductor. The surrounding 3D semiconductor has a continuum of
energy levels (energy bands).
Fig. 5-62. Example of a 2-dimensional semiconductor (quantum well).
-397-
5-20. Energy Bands in Graphene (A)

Graphene is a single layer of carbon packed in a hexagonal (honeycomb)
lattice, with a carbon-carbon distance of 0.142 nm. It is the first truly two-
dimensional crystalline material. The band structure of graphene is shown
in figure 5-63. The electronic structure of graphene is rather different from
usual three-dimensional materials. Its Fermi surface is characterized by six
double cones, as shown in figure. Using the tight-binding model, the
energy of the electrons with wavenumber k is:
E (k )    o 1  4 cos32 ao k y . cos  2
3

ao k x  4 cos2  2
3
ao k x  (5-119)
where γ0 ≈ 2.8 eV is called the nearest-neighbor hopping energy and ao ≈

1.42 nm is the lattice constant.
Fig. 5-63. Energy bands structure of graphite sheets (graphene)
Close to the Fermi level the dispersion relation for electrons and holes is
linear. Since the effective masses are given by the curvature of the energy
bands (m*=2/2E(k)/k2), this corresponds to zero effective mass. The
equation describing the excitations in graphene is formally identical to the
Dirac equation for massless fermions which travel at a constant speed.
Therefore, the connection points of the cones (Fermi points) are sometimes
called the Dirac points10. This gives rise to interesting analogies between
graphene and nuclear particles, which are valid for energies up to
approximately 1eV, where the dispersion relation starts to be nonlinear.
10
One result of this special dispersion relation, is that the quantum Hall effect becomes unusual in
graphene, as we’ll see in Chapter 7 of this book.
-398-
Note 5-8. The Tight Binding Model for Graphene

The graphene plane is a hexagonal lattice with two atoms per unit cell (say
A and B) and a basis defined by two unit vectors (say a1, a2). The condition
ai·bj = 2δi,j allows one to obtain the reciprocal lattice vectors (b1, b2). Every
carbon atom possesses four valence electrons (two 2s and two 2p
electrons). When the atoms are placed onto the graphene hexagonal lattice
the electronic wave functions from different atoms overlap. However, the
overlap between the pz orbitals and the s or px and py electrons is strictly
zero by symmetry. Consequently, the pz electrons, which form the π bonds
in graphene, can be treated independently of other valence electrons.
Within this π-band approximation, the A atom (or B atom) is uniquely
defined by one orbital per atom site pz(r−rA) or pz(r−rB). To derive the
electronic spectrum of the total Hamiltonian, the corresponding
Schrödinger equation has to be solved, and by applying the Bloch theorem,
the wave functions can be written as follows
where
with k is the crystal momentum, N is the number of unit cells in the

graphene sheet, l is the cell position index and J = A,B
The E(k) relation is derived by solving the Schrödinger equation which
reduces to diagonalizing the following 2×2 matrix
Neglecting the overlap <A|B>, restricting interactions to nearest neighbors

only and setting HAA = HBB = 0 as energy reference the dispersion relation
are then:

E (k )    o 1  4 cos32 ao k y . cos 2
3

ao k x  4 cos2  2
3
ao k x 
where γ0 is the nearest-neighbor hopping energy and ao is the lattice
constant. These are the π and the π∗ bands in the TB model. One of the
above two bands (which represent the valence and the conduction bands),
is completely filled and the other completely empty. Moreover they
intersect only in two points, known as Dirac points, in the BZ.
-399-
For this reason graphene is a semi–metal with a one–dimensional Fermi

surface. For the same reason CNTs are predicted to be either metallic or
semi–conducting according to whether or not the Dirac points belong to the
set of allowed k–points. The predictions of the TB model are partially
confirmed by Ab–initio calculations even if some differences appear. For
example the DFT band structure of graphene is not symmetric with respect
to the chemical potential, as the TB.
5-21. Energy Bands in Carbon Nanotubes (A)

Carbon nanotubes are quasi-one-dimensional form of carbon with nano–
metric diameter. They can be formed from graphene sheets which are
rolled up to form tubes. We have seen that carbon nanotubes may be
metallic or semiconducting according to their geometry and chirality. In
fact, the unique energy band structure of graphite allows for it to behave as
both a metal and a semiconductor. The electronic and mechanical
properties of metallic single walled nanotubes have many similarities with
graphene. In order to know the electronic structure and properties of carbon
nanotubes one can apply specific quantization condition, since electrons are
bound in the circumferential direction of the tube. There exists an
approach, known as the zone folding approximation (ZFA), which utilizes
such quantization conditions.
Fig. 5-64. Energy band diagram in one and two-dimensional k-space of metallic and
semiconductor nanotubes.
-400-
The quantization condition is kc.R=2n, where kc is the wave-vector in the

circumferential direction, R is the chirality vector of the nanotube, and n
is an integer. With all boundary conditions in place we can calculate the
energy band structure for the carbon nanotube, from the graphene band
structure. Note that the energy gap is zero at the K-point (Fermi point),
where k = (4/3a), and a =√3ao. This can be verified, by substituting k
into (5-109). If the graphene sheet is rolled in that direction to form a CNT,
the resulting CNT is metallic.
Carbon nanotubes do not all have the same band gap. For each diameter
there is a set of allowed valences and conduction states. As nanotube
diameters increase, more states are allowed. The relation between the
energy gap (Eg, in eV) and diameter (d, in nm) is approximately given by:
Eg ≈ 2 Eo (a / d) = 0.9 / d (5-120)
where a = 1.42 nm is the distance between neighbor carbon atoms in a flat

graphene sheet and Eo =2.7 eV is their binding energy. In this way,
different-size nanotubes can have bandgaps as low as zero (like a metal), as
high as the band gap of silicon, and maybe in between. Therefore, the
geometry of nanotubes limits electrons to select a few slices of the graphite
energy states. Depending on the diameter of the nanotube, one of these
slices can include the narrow path that joins electrons with conduction
states. This point is called the Fermi point and makes two thirds of the
nanotubes metallic. Otherwise, if the slices miss the Fermi point, the
nanotube is semiconductor, as shown in figure 5-64. It worth noting that
twisting of CNT’s transforms the metallic nanotube to a semiconducting
one with a band-gap that varies with the twist angle. The following figure
illustrates the energy gap of a SWNT as a function of its diameter.
Fig. 5-65(a). Energy gap of carbon nanotube, as a function of tube diameter.

-401-
The following figure shows the Katuara plots for the same parameter. The
Katuara plot is a theoretical graph was designed in 1999 by Hiromichi
Kataura to rationalize some experimental findings. It relates the nanotube
bandgap energy and its diameter. The oscillating of the of the Kataura plot
branches reflects the strong dependence of the SWNT properties on the
chirality (n, m) index, as well.
Fig. 65(b). Katwara plots for the energy gap of carbon nanotube,.
Figure 5-66 depicts the dispersion relation E(k) of electrons and the density
of states in both metallic and semiconductor nanotubes. The following
figure schematically shows the electronic density of states expected for the
π- electrons in a semiconducting SWNT. The spikes in the density of states
are called van Hove singularities (vHSs). They arise from the quasi one-
dimensional character of electronic motion in systems with very high
aspect ratios. Each van Hove singularity is labeled with the index of the
subband to which it belongs. These singularities are of great interest for
optical transitions, in a variety of optical phenomena. In fact, the unique
optical properties observed in SWNTs are due to the one-dimensional (1D)
confinement of their electronic structure. Note that the density of states
(DOS) per unit length in one dimension is given by:
-402-
(5-121)
Fig. 5-66. Energy bands and density of states of metallic and semiconductor nanotubes.
-403-
Fig. 5-67. Details of the density of states of a semiconductor nanotube.
The above remarkable results can be obtained from a variety of

considerations, starting from the so-called Zone Folding Approach (ZFA),
based on knowledge of the electronic properties of the graphene, to the
direct study of nanotubes using Tight–Binding (TB) approaches. The
comparison with sophisticated Ab–initio calculations and available
experimental results, permits to find the limits of these simple approaches
Note 5-9: The Zone Folding Approach (ZFA)
In the ZFA the wave–functions n(x)=exp[i(kxx+kzz)].un(x) of the graphene

sheet are used to describe electrons in CNTs assuming that the curvature of
the sheet gives negligible effects if the tube radius is large enough. Then
the only difference the electrons feel, with respect to graphene, is the
quantization of the angular momentum lz = kxR (we are assuming here that
the tube direction is the z-axis). Accordingly only the wave–functions n(x)
are considered in the model.
-404-
In the ZFA each CNT can be identified by two numbers, which represent
the circumference vector in the basis of the direct lattice vectors a 1, a2 of
graphene. Boundary conditions are then imposed which results in selecting
specific k–points in the reciprocal space.
As shown in the right top of the following figure, the Brillouin Zone (BZ)
of graphene is represented and the lines correspond to the k points which
respects the boundary conditions of a(5,5) CNT. The right bottom panel,
instead, shows the energy surface of the π−π∗ bands of graphene, cut by the
allowed k–points which can be used to construct the CNT band structure.
The predictions of the ZFA give a good description of the properties of

CNTs but some corrections have to be considered both for SWCNTs, to
include the curvature of the tubes, and for MWCNTS, to include the effect
of the interaction among different tubes. For example the position of the so
called Dirac points (see the definition of the Dirac points) has to be shifted
in the Brillouine Zone due to curvature effects. Finally the ZFA results
depend on the method used to compute the band structure of graphene. The
energy surfaces in the above figure have been calculated within the TB
model, for example.
Within the TB+ZFA scheme, it can be show that for example all (n,n) and
(3n,0) CNTs are metallic, with n any integer, while all the remaining (n, 0)
tubes are semi–conducting. CNTs of different kinds, (n,m) CNTs, are said
to be chiral and can be either metallic or semi–conducting. The general rule
is that a CNT is metallic if n−m is a multiple of 3.
-405-
5-22. Energy Bands in Silicon Nanowires and Nanotubes(A)

Silicon nanowires (SiNWs) are direct bandgap materials with a Brillouin
zone half the length of that of bulk Si along the Δ line. In the conduction
band, valley splitting reduces the averaged mobility mass along the axis of
the wire, but quantum confinement increases the transverse mass of the
conduction band edge.
Fig. 5-68. Silicon nanowire (SiNW) and its unit cell.
Fig. 5-69. Energy bands in an 1.54 nm x 1.54 nm SiNW.
-406-
For the wire thickness range up to 3nm, the effective mass at the
conduction band edge is at least 35% heavier than that of transverse mass
of bulk Si.
Quantum confinement has the largest effect on the carrier effective masses
in the valence band. The effective mass at the valence band edge is at least
6 times heavier than that of the bulk in the [100] direction. The effective
mass of the next highest band is even heavier. Small energy splitting also
occurs at the conduction band minimum. Figure 5-70 shows the shape of
hydrogenated SiNW and its unit cell. Also, Fig. 5-71 depicts the energy
band structure of a 1.54 nm SiNW. For wires greater than 1.54 nm thick,
the 4 bulk valleys which compose the conduction band minimum are split
into 3 energies. The center energy is twofold degenerate roughly evenly
split between the lowest and highest energy so that the conduction band
minimum becomes non-degenerate except for spin.
The silicon nanotubes (SiNTs) are some sort of hollow SiNWs, with tube-
wall-thickness of several nanometers. SiNTs are expected to have more
potential applications than SiNWs. Recently, crystalline silicon nanotubes
(c-SiNTs), have been successfully synthesized. Figure 5-60 shows a TEM
picture of a c-SiNT. The following figure shows the effect of SiNT
thickness on the energy band diagram and energy gap of a uniform SiNT of
diameter D=1.80nm. The band gap increases significantly as the thickness
becomes smaller due to quantum confinement.
Fig. 5-70. TEM image of a synthesized c- SiNW.

-407-
Fig. 5-71. Effect of tube thickness () on the energy band structure and energy gap of a
uniform c-SiNT of diameter D= 1.8 nm.
-408-
5-23. Organic Semiconductors (A)

It has been demonstrated, in the late 1970's, by Alan Heeger and his
colleagues that some molecular structures of certain organic crystals and
polymers (plastics) behave as semiconductors. Organic semiconductors,
like acenes, oligothiophenes, fullerenes and their related compounds, have
very attractive characteristics for electronic and opto-elelctronic devices.
Such organic semiconductors can be deposited on thin plastic films and so
are flexible and inexpensive. In fact, the so-called polymer electronics (or
plastic electronics) is a new emerging topic in electron devices.
The question is: why are some organic materials (polymers) behave like
semiconductors? The answer is as simple as follows. In all systems of
conductive organic materials, the conducting chain of the molecule consists
of unsaturated carbon atoms (e.g. --CH-CH-CH--; each atom is only three-
fold coordinated). Each carbon atom contributes a single pz-electron to a
bond with the neighboring atoms. This situation is unstable to small
symmetry-breaking distortions (called Jahn-Teller distortions). The
bandgap that is opened up by this symmetry lowering is in the order of
2.5eV and this is in the range of semiconductors. Therefore, organic
semiconductors mostly fall in the category of wide-band-gap
semiconductors (Table 5-9). Electrons may be excited from the valence
band to the conduction band, by thermal and optical excitations, or by
doping. In fact, most organic semiconductors have a sufficient doping level
to make them p-type.
There are two major classes of organic semiconductors. These are:
organic charge-transfer complexes and
linear-backbone conductive polymers derived from polyacetylene.
Linear backbone organic semiconductors include polyacetylene itself and

its derivatives polypyrrole, and polyaniline. At least locally, charge-transfer
complexes often exhibit similar conduction mechanisms to inorganic
semiconductors. Such mechanisms arise from the presence of hole and
electron conduction layers separated by a band gap. As with amorphous
semiconductors, tunneling, localized states, mobility gaps, and phonon-
assisted hopping also contribute to conduction, particularly in poly-
acetylenes..
-409-
Semiconducting molecules are distinguished from insulating molecules by

the large spatial extent of their frontier orbitals. In the case of hole
transport, the orbital of relevance is the highest occupied molecular orbital
(HOMO), whereas, for electron transport, it is the lowest unoccupied
molecular orbital (LUMO). In insulating molecules, these orbitals are
small, localized on just a few atoms. As a result, the spatial overlap
between frontier orbitals on neighboring molecules is also small and so the
likelihood of charge transfer between them is vanishing. In semiconducting
molecules however, the frontier orbitals are spatially delocalized, covering
much of the molecule. Delocalization arises from the strong interaction
(conjugation) of partially filled orbitals on neighboring atoms, most often
p-orbitals on C, O2, S or N2. Because of delocalization, the spatial overlap
between orbitals on neighboring molecules (or conjugated segments, in
polymers) is larger than in insulators, and therefore the probability of
charge transfer is greater.
Fig. 5-72, Two monomers of poly phenylenevinylene (PPV), showing (a) the chemical
structure, (b) the HOMO, and (c) the LUMO.
The strength of the coupling between front orbitals defines not only the rate
of charge transfer between molecules, but also its mechanism. When
neighboring molecules are well coupled, the charge is delocalized across
both molecules. Conversely, when the molecules are poorly coupled, the
charge is strongly localized on a single molecule and only able to transfer
to the other molecule via a thermally assisted hopping.
-410-
5-24. Energy Bands in Amorphous Semiconductors

The nature of the energy band structure in amorphous semiconductors is of
both fundamental and applied interest. Several authors have proposed
different theories and band models for amorphous semiconductors in order
to explain their electrical conductivity. However, since the seminal work of
Anderson it has been known that disorder induces localization of electron
states. Therefore, almost all amorphous semiconductor models invoke the
concept of localized states in the band tail.
The so called “mobility edge” is defined as the energy separating localized

and non-localized states in the conduction or valence bands of a non-
crystalline material, or the impurity band of a doped semiconductor.
There have been a very large number of studies on electron states in

amorphous materials, and especially amorphous silicon (a-Si). The
calculations ranged from simplified tight-binding (TB) approaches to the
density functional theory (DFT).
A difficulty with many calculations of electron states in amorphous Si is

the usual challenge of obtaining a realistic starting model. Amorphous Si is
well known to be almost entirely four-coordinated. However, the analysis
is usually limited to three-dimensional systems, since in one or two
dimensions all states become localized. Figure 5-73 depicts the electronic
states in the band-gap region of -Si.
Fig. 5-73. Electronic states in the band-gap region of a-Si. The valence band tail is
almost exponential with width of 190 meV as estimated with fitting (dashed line).
-411-
5-25. Topological Insulators

The topological insulator is a new state of matter with unusual properties
that manifest themselves on the surface of insulators or at the interface of
narrow band gap materials where gapless (metallic) surface/interface states
appear. These surface/interface states are excitated by massless fermions.
Protected by time-reversal symmetry, the spin of such fermions (called
Kramers partners) is locked at a right angle to their momenta due to spin-
orbit coupling, making them protected against scattering. This means the
larger the spin-orbit coupling, the better is the protection.
Topological insulators (TI) can be found in a subclass of narrow-gap semi-
conductors and heterostructures where the band gaps are smaller than the
relativistic corrections to the band structure. In fact, the heterostructures of
some compound semiconductors such as Bi1-x Sbx and Pb1-xSnxTe exhibit a
strong topological phase. In Bi1-xSbx, with increasing x the gap at the L
point decreases and the L+ and L- bands cross at x = 0:04. After further
increase in the mole fraction x, the gap reopens with inverted band.
Fig. 5-74. Band inversion between two end members in Pb1-xSnxTe., Energy spectrum of
the inverted contact. The solid lines are Weyl states and dashed lines are additional
branches appear for contact thickness l > lo.
Some 2D topological insulators, such as HgTe quantum wells, exhibit the

quantum spin Hall Effect (SQHE). In the interface of the quantum well of
such 2D TI’s, the spin changes sign, thereby creating a pair of gapless
helical edge states inside the bulk energy gap. The SQHE is presented in
Chapter 7 of this book.
-412-
5-26. Effect of Magnetic Field on the Energy Band Structure of Solids

In the presence of a magnetic field B, the one-electron Hamiltonian may be
written as follows:
 P2 ( p  eA) 2
Η 
2m
V( r ) 
2m
 V r  (5-122a)
where p = -jħ is the generalized momentum and A is the vector potential,

which is defined by B=xA. The quantity P = (p+eA) is called the
kinematic momentum. Using the effective mass approximation, the effect
of the crystal potential V(r) can be included in the electron effective mass
m* such that:
 ( p  eA) 2
Η 
2m*
(5-122b)
Substituting such a Hamiltonian into the Schrödinger equation and solving,

we can obtain the E-k relation describing the electron states. In a certain
direction of the k-space, say kz, the E-k relation is as follows:
 2 k z2
E  E n  Ec    c (n  21 ) (5-123)
2m*
wherec = eBz/mz* is called the cyclotron frequency, Bz is the component

of B along the z-axis of the lattice space, mz* is the component of effective
mass in the z-direction and Ec is a constant (energy reference). We note
here an additional quantized term, in comparison with the simple parabolic
E-k relation E=Ec + ħ2k2/2m*, with no applied magnetic field. We note also
that the additional term is only significant for strong magnetic fields, where
c is in range of microwave frequencies or higher. Therefore, the simple
parabolic E-k relation of the quasi-free electron is transformed into a family
of quantized parabolas (called sub-bands), in the presence of a strong
magnetic field.
The quantized energy levels described by (5-112) are called Landau levels.
Figure 5-75 depicts the effect of magnetic field on the simple E-k relation
of a quasi-free electron and the constant energy surface. Figure 5-76 shows
how the free electron Fermi surface (dotted sphere) is transformed into
concentric cylinders - called the Landau tubes. Also, figure 5-69 illustrates
the transformation of quasi-continuous energy band spectrum of electrons
-413-
in a crystal into discrete energy levels, in the presence of a strong magnetic

field. When the magnetic field is not strong enough, such that ħc < ħ,
where  is the energy state lifetime, then, the energy broadening 11 in
Landau levels (E=ħ ) will be in the order of the distance between them
(ħc) and no quantization can be observed. Note that, in presence of
magnetic field, the lowest energy level in the conduction band is given by:
Ec(B) = Ec + ½ ħc (5-124)
Thu, the magnetic field tends to increase the energy gap (by 1/2 ħc). We
have seen so far, the density of states in one-dimensional k-space is
proportional to E-1/2 (with no applied magnetic field). In the presence of
magnetic field, the density of energy states will have similar energy
dependence. Therefore, in the direction of magnetic field we have:
g E   B  E  E c   c ( n  21 )
1 2
n
(5-125)
where the sum is taken over all subbands (Landau levels). Thus, the
density of states in magnetic fields tends to infinity at the apexes of Landau
levels (where E = Ec + ħc(n+1/2) ) and decreases with increase in energy,
as shown in figure 5-77.
E-Ec
n=2
ħ2k2/2m*+1/2
ħc n=1 ħc
n=0
k
Fig. 5-75. Illustration of the quantized Landau levels, in presence of magnetic field.
11
The energy broadening in a quantum state is due to the uncertainty in energy by E=ħt, where tis
the state lifetime. The state lifetime is roughly the meantime between two successive collisions.
-414-
It worth notice that the change of the shape of density of states to the
peaked form, shown in figure, is expected to have many consequences on
the physical characteristics of the material. For instant, the energy states
near the apexes will be depopulated (unfilled) because of their very large
number. In addition, the Fermi energy EF must coincide with B such that it
becomes over the filled states. When magnetic field is raised, one of the
Landau levels will cross the Fermi energy. Hence, the electron should drop
into a lower energy level. One can observe an oscillatory free energy with
increasing magnetic field. This oscillation is known as the Shubnikov-de
Haas effect. Figure 5-78 illustrates the oscillation of Fermi level as the
magnetic field is increased. The maxima of these oscillations occur at
regular intervals of 1/B such that:
1 2e 1
   . (5-126)
B  Aext
The above equation relates the periodicity of oscillations of the free energy
to the external cross-sectional area Aext of the Fermi surface. So, it should
be noted that the presence of localized energy levels in the energy gap (due
to impurity atoms or other defects in the crystal) broadens the transition of
Fermi energy between Landau levels.
Fig. 5-76.. Transformation of the constant energy surface of a free electron (dashed
sphere) into quantized cylinders (Landau levels), in presence of a magnetic field.
-415-
E E – Ec - ħ 2k2/2m*
n=3 7/2ħ
n=2 5/2
c
Band
n=1 ħ c
3/2
Ec n=0 ħc
1/2
ħc
Eg Eg +1/2
ħc
Band
Quasi-continuous energy bands without Discrete energy levels with magnetic field.
magnetic field.
Fig. 5-77. Quantization of energy bands in presence of a magnetic field.
B B=
gn(E)
0 0
n=0 n=1 n=2 n=3 n=4

E-Ec
1/2ħc 3/2ħc 5/2ħc 7/2ħc
Fig. 5-78. Effect of magnetic field on the density of states in the direction of field.
E- Ec - ħ2k2/2m*
Locus of Fermi
n=3
Energy level
n=2
n=1
Average
n=0
Fig. 5-79. Schematic of the locus of Fermi energy as magnetic field increases.
-416-
5-27. Effect of Magnetic Fields on the Energy Bands in LDS’s

The effect of magnetic on low-dimensional semiconductor devices such as
superlattices and quantum cascade (QC) structures has been a fruitful
subject during the last few years.
When a magnetic field is applied parallel to the layers of a superlattice, it

shifts the E(k) dispersion curves, as shown in figure 5-80. The shift in k-
states is proportional to the magnetic field density, and given by:
eB
k  . z  (5-127)

When a magnetic field is applied perpendicularly to the layers of a

superlattice, as shown in figure 5-80, we can separate the Hamiltonian in
perpendicular and in-plane components. The in-plane states are further
localized in Landau levels. A perpendicular magnetic field offers the
possibility of simulating a quantum box structure, with a confining
potential that depends upon the intensity of magnetic field. The magnetic
size of the quantum box is then given by:
lc =[h / (eB)] 1/2 (5-128)
(a)
(b)
Fig. 5-80. Effect of (a) parallel and (b) perpendicular magnetic fields on the
dispersion curves of a superlattice.
-417-
The above magnetic effects are very important when they are coupled with
optical phenomena in superlattices, as shown in figure 5-81. It has been
observed that parallel magnetic fields brought a quenching of the intensity
of electro-luminescence signals, together with a broadening and a blue-shift
of the photo-emission in superlattices. On the other hand, a perpendicular
magnetic field will allow transitions only between Landau levels and
optical phonon emission is strongly decreased.
Fig. 5-81. Effect of magnetic field on the spectra of a quantum cascade LASER, as
detected by InSb cyclotron resonance detector.
5-28. Effect of Magnetic Fields on Nanotubes (Aharonov-Bohm effect)

As we have seen so far, the electronic properties of a carbon nanotube
depend on its chirality, which, in turn, depends on the direction in which
the graphene sheet has been rolled up to form the nanotube. Therefore,
about two-thirds of nanotubes are semiconductors, and the remaining third
are metals. Although the electronic fate of a nanotube is determined once it
has been grown, this is not quite the end of the story. For instance,
researchers have tried to alter the band gaps of nanotubes by twisting or
bending nanotubes or by filling them with molecules. Recently, several
researchers have demonstrated that this fate can be achieved with a
magnetic field.
-418-
Indeed, the experiments show that one can convert a semiconducting

nanotube into a metallic one, and vice versa, at least in principle. The
magnetic field comes into play via the so-called Aharonov-Bohm (AB)
effect. In this quantum effect, the wavefunction of an electron acquires a
phase shift as it follows a trajectory that encloses a magnetic flux. This
phase shift alters the way the electron trajectories interfere, and therefore
affects the electronic properties of the material. Therefore, by increasing
the magnetic field, all the electronic properties of the CNTs will oscillate
with period o = h/e.
Fig. 5-82. Illustration of the AB effect and resistance oscillation of a nanotube ring.
As shown in figure 5-82, the resistance of a nanotube is maximum at zero

magnetic field, because electrons are propagating around the ring in
opposite direction interfere with each other constructively at the point of
injection, forming a standing wave. Applying a magnetic field breaks time-
reversal invariance, and the constructive interference is lost. However, the
magneto-optical and magneto-electronic nanotube devices based on the AB
effect will probably always be out of reach because they require a magnet
of about 1000T.
-419-
Fig. 5-83. Resistance of a nanotube ring as a function of the magnetic field.
Note 5-10. The Aharonov-Bohm Effect in Carbon Nanotubes

The Aharonov–Bohm (AB) effect is a purely quantum mechanical effect
which does not have a counterpart in classical mechanics. Classically, a
magnetic field confined in a closed region of space alters the kinematics of
charged particles only if they move inside this region. Electron dynamics,
instead, are governed by the Schrödinger equation, and are influenced even
if electrons move on other paths where the magnetic field is confined, and
the Lorentz force is zero. If this closed region is the inner part of a CNT,
electrons traveling around the cylinder are expected to manifest a shift of
their phase. The mathematical interpretation of this effect is connected with
the definition of the vector potential, which, in the case of confined
magnetic fields, cannot be nullified everywhere. This extraordinary effect,
first predicted by Aharonov and Bohm (AB) in 1960, was interpreted as a
proof of the reality of the electromagnetic potentials. The idea that
electrons could be affected by electromagnetic potentials without being in
contact with the fields was skeptically received by the scientific
community. At the same time the AB paper spawned a flourishing of
experiments and extension of the original idea. The first experiment aimed
at proving (or disproving) the AB effect revealed a perfect agreement with
the theoretical predictions. Nevertheless only in 1986, the experiment
which can be considered as a definitive proof of the correct interpretation
of the AB effect was realized. Tonomura et al, were in fact able to
observed AB oscillations, using superconducting niobium cladding.
-420-
Nowadays the AB effect can be used in a wide range of applications, like

the investigation of mesoscopic conductors and superconductors. One of
the most well–known case is given by carbon nanotubes (CNTs) that, if
immersed in a uniform magnetic field aligned with the tube axis, have been
predicted to show peculiar oscillations of the electronic gap and all other
properties of the CNT. These oscillations are characterized by a period
given by the magnetic flux quantum o = πr2B = (h/e) and are commonly
interpreted as caused by the change in the wave functions of the electrons
localized on the tube surface induced by the Aharonov-Bohm effect. As
the flux period o = πr2B, one can find the average radius r of the outer shell
of a CNT, which can be verified by measurement.
The first experiment carried on CNTs, in 1999, described the oscillations in

the electronic conductivity but with period of (h/2e). This deviation from
the predicted AB oscillation period has been explained in terms of the weak
localization effect induced by defects and dislocation. Only in 2004, a clear
proof of the existence of AB oscillations with (h/e) period has been given
by Coskun et colleagues by measuring the conductance oscillations in
quantum dots.
-421-
5-29. SUMMARY
One of the most important results of the application of quantum mechanics
to the description of electrons in solids is that the allowed energy levels of
electrons in individual atoms are grouped into bands. In fact, when atoms
are brought together to form a crystalline solid, the energy levels of
individual atoms are split and broadened (because of the Pauli exclusion
principle) and transformed into energy bands. You may imagine energy
bands as some sort of channels in which electrons move freely. These
energy bands are usually separated by the so-called energy gaps.
The electrons of lower energy levels are tightly bound to their nuclei and
form narrow energy bands. But the electrons of higher energy levels
(valence shells) are quasi-free and form wider energy bands. The energy
gaps are narrower the higher is the corresponding band. In higher levels,
energy bands may overlap. Each simple band cannot contain more than 2N
electrons where N is the number of atoms in the crystal. The highest filled
band is termed valence band, and the highest empty or partially filled band
is termed conduction band. The lowest level in the conduction band is
-422-
termed by Ec, while highest level in the valence band is termed by Ev. The
distance between the top of valence band and the bottom of conduction
band is called the energy gap Eg. Evidently we have:
Eg = Ec - Ev
The lowest level in the conduction band, Ec, designates the energy of a
conduction electron at rest. The energy of an electron at rest is its potential
energy. The kinetic energy of an electron in the conduction band is merely
the difference E-Ec. Solid materials may be classified to conductors and
non-conductors. Conductors (metals and semimetals) are substances with
partially filled bands. Non conductors (semiconductors and insulators)
contain empty and filled bands. The energy gap in semiconductors is
usually in the order of 1eV and greater than 5eV in insulators.
At very low temperatures electrons fill the energy bands of solids, up to an

energy level called the Fermi level EF, and the bands are empty for
energies that exceed EF. In 3-dimensional k-space, the E(k) relation may be
drawn as constant energy surfaces E(k) = constant. Specifically, the
constant energy surface E(k) = EF is called the Fermi-energy surface.
The electrons moving inside crystalline solids may be considered as free

electrons whose effective mass is given by:
-423-
m*  
kk E
In 1-dimensional k-space (k=kx), the effective mass is scalar and given by
2

m* = 2
d E dk 2  
The effective mass is different from the electron rest mass mo, because it
lumps the effect of lattice forces through the E-k relation. The effective
mass m* in a certain band is inversely proportional to the bandwidth.
Thus the effective mass in a crystalline solid is proportional to the energy
gap width of that solid.
The concept of holes is introduced in semiconductors since it is easier to

keep track of the missing electrons in an "almost-full" band, rather than
keeping track of the actual electrons in that band. An empty state in the
valence band is called a hole. The hole has a positive charge +e and a
position effective mass m*p. The following tables depict the effective mass
(mo/m*) of electron and holes in some famous semiconductor materials
Holes; mo/m* Electrons; mo/m*
IV; III-V II-VI IV-VI IV; III-V II-VI IV-VI
C 0,25 AlSb 0,98 CdS 0,80 PbS 0,25 C 0,2 AlSb 0,12 CdS 0,21 PbS 0,25
0,04 0,98
Ge 0,28 GaN 0,60 CdSe 0,45 PbTe 0,2 Si 0,19 GaN 0,19 CdSe 0,13 PbTe 0,17
0,16 1,64
Si 0,49 GaSb 0,40 ZnO Ge 0,082 GaSb 0,042 ZnO 0,27
SiC 1,00 GaAs 0,082 ZnS SiC 0,6 GaAs 0,067 ZnS 0,40
GaP 0,60 GaP 0,82
InSb 0,40 InSb 0,014
InP 0,64 InP 0,077
-424-
Generally speaking, there exist two basic categories of computational

methods, namely to develop the energy band structure of a matter:
 Ab initio methods
 Approximate (or empirical) methods
The ab initio computational methods are based on first-principles quantum

physics, to find out the energy band structure of a matter. Ab initio
electronic structure methods have the advantage that they can be made to
converge to the exact solution, when all approximations are sufficiently
small. In this case, when all possible configurations are included, such
methods tend to the exact solution of the Schrödinger equation. The
Hartee-Fock, (HF) technique is the simplest ab inito method. Density
functional theory (DFT) methods use functionals, which include Hartree–
Fock exchange scale. Although its history goes back to the early thirties of
the 20th century, DFT has been formally established in 1964 by two
theorems due to Hohenberg and Kohn.
• 1920s: Introduction of the Thomas-Fermi model.

• 1964: Hohenberg-Kohn paper proving existence of exact DF.
• 1965: Kohn-Sham scheme introduced.
• 1970s and early 80s: LDA. DFT becomes useful.
• 1985: Incorporation of DFT into molecular dynamics
• 1988: Becke and LYP functionals. DFT useful for some chemistry.
• 1998: Nobel prize to W.Kohn in chemistry for developing DFT
The dark side of ab initio methods is their computational cost. They often
take enormous amounts of CPU time and memory. Therefore, so many
researchers have been utilizing the so-called approximate or empirical
methods to develop the energy band structure of crystalline solids.
Generally speaking, there exist two basic approximate methods, which are
widely used for the calculation of energy bands in solids, namely:
 The quasi-free electron (QFE) model, and
 The tightly bound electron (TBE) model.
Both of the above approximate methods use the perturbation theory. But
they differ in their zero approximation. The quasi-free electron (QFE)
method (which represents the physicists approach) takes the free electron
model as a zero-order approximation and the periodic field of the lattice as
-425-
a perturbation. This results in the so-called pseudopotential methods. The

tightly bound electron (TBE) method (which represents the chemists
approach) takes the electron in an isolated atom as a zero approximation
and the periodic potential of the lattice as perturbation. This results in the
so called linear combination of atomic orbitals (LCAO) methods.
In the so called low-dimensional semiconductors (LDS’s), free electrons
are only permitted to move in one or two dimensions. For instance, the
electrons filling the thin inversion layer in a conventional MOSFET
(Metal-Oxide-semiconductor Field Effect Transistor) are only permitted to
move freely in the 2 dimensions of the inversion layer. The motion of free
electrons in the third dimension (normal to the layer plane) quantized. The
next table summarizes the dispersion relations E(k) and density of states
(DOS) of parabolic semiconductors of different dimensions.
One of the most interesting low dimensional structures is Carbon nanotubes

(CNTs). CNTs are rolled sheets of graphene (graphite layers). They may be
metallic or semiconducting according to their geometry and chirality. The
energy band structure of graphite allows for it to behave as both a metal
and a semiconductor. Graphene is a single atomic layer of graphite which
is a honeycomb lattice of carbon atoms. The energy band structure of a
CNT can be described by the following relation:

E (k )    o 1  4 cos32 ao k y . cos 2
3

ao k x  4 cos2  2
3
ao k x 
where γ0 is the nearest-neighbor hopping energy and ao is the lattice
constant.
-426-
5-30. Problems
5-1) In device fabrication it is sometimes useful to be able to see through

a silicon wafer in order to align a pattern on the front of the wafer with one
on the back. This requires the use of an infrared microscope, since silicon
is opaque to visible light. Given that the energy band gap of silicon is
1.12eV, estimate the minimum wavelength of light that could be used in
such microscope.
5-2) If the energy of a crystal excitation varies with the wave vector k
like:
2  k x2 k y2 k z2 
E( k )  E o   
2 m  m  m 
 x y z 
Show that the density of states per unit energy per unit volume is given by:
g( E ) 
N
2 m m m 1/ 2
( E  E o )1 / 2
2 
2 2 x y z
5-3) A 1 cm3 of a metal has a density of states g(E) =51021 E1/2, where E
is measured from the bottom of conduction band. If the number of
electrons in a small energy interval between E=4.6 eV and E+ dE = 4.601
eV is equal to 2.0  10 15 electron.
i) Determine the fraction of states between E, E+dE, which are
occupied by electrons.
ii) Calculate the position of Fermi level.
iii) Determine the number of electrons at T= 0K.
5-4) For photons, where the relationship between energy and wavelength
is E = pc = hc/, show that the density of states is given by:
g(E) = 8E2/(h c)3
5-5) State two results of using the Kronig–Penney model with the
Schrodinger’s wave equation. Plot the function f(a) = 9sin(a)/a + cos(a)
for 0 ≤ a ≤ 6π. Also, given the function f(a) = cos(ka), indicate the allowed
values of a which will satisfy this equation.
5-6) Show that the average kinetic energy of a hole in the valence band of
a semiconductor is 3/2 kBT.
-427-
5-7) Prove that the ground-state solution to the case of a particle-in-a-

cylinder (like an electron inside a nanotube or a nanowire) is given by the
wave function:
where J0(01·r/R) is the zero-order Bessel function, 01 the first zero of the
zero-order Bessel function, L the length of the cylinder, and
and NL = (2/L)1/2 are normalization factors for the radial and axial portions
of the wave function.
Prove also that the confinement energy increase in the band gap of such
tubes, is then given by
where h is the Planck’s constant and m* is called the reduced effective

excitonic mass, m* = me.mh/(me +mh).
5-8) How does the band gap indicate whether or not your substance is an
insulator, semiconductor or conductor?
5-9) What is a direct bandgap semiconductor? What is an indirect bandgap

semiconductor?
5-10) What is the meaning of the density of states function?

5-11) What are the benefits of graded-bandgap semiconductors?
5-12) What is the purpose of understanding band theory?
-428-
5-31. BIBLIOGRAPHY
[1] M. Born and R. Oppenheimer, Zur Quantentheorie der
Molekeln, Ann of Physics (Leipzig), vol. 84, No 20, p.457, 1927.
[2] E. Fermi, Un metodo statistico per la determinazione di alcune
proprietà dell'atome, Rend. Accad. Naz. Lincei 6, p.602, 1927.
[3] A.Wilson, “The Theory of Electronic Semiconductors,”
Proceedings of the Royal Society of London, Series A, Vol.133,
p.458, 1931.
[4] W. Shockley, The Empty Lattice Test of the Cellular Method in
Solids, Phys. Rev. vol. 52, pp. 866 – 872, 1937.
[5] Gregory H. Wannier. The structure of electronic excitation levels
in insulating crystals. Phys. Rev. vol. 52, p. 191, August 1937.
[6] F. Seitz, Modern Theory of Solids, McGraw-Hill, New York, 1940.
[7] Conyers Herring. A new method for calculating wave functions in
crystals. Phys. Rev. 57, 1169 June 1940.
[8] W. Shockley, Electrons and Holes in Semiconductors, Van
Nostrand, New York, 1950.
[9] J. R. Reitz, Method of One-Electron Theory of Solids, pp. 2-96.
Solid State Physics, Vol. 1, Edited by Frederick Seitz and David
Turnbull, Academic Press. Inc., New York, 1955.
[10] J. C. Slater, Electronic Structure of Solids, Handbuch der Physik,
vol. XIX, edited by Flagge, Springer-Verlag, Berlin, 1956.
[11] G. H. Wannier, Elements of Solid State Theory, Cambridge
University Press, New York, 1959.
[12] J. Callaway, Electron Energy Bands in Solids, Academic Press, New
York, 1958, 1964, 1990.
[13] J.M. Luttinger, Analytic properties of single-particle propagators for
many fermion systems. Phys. Rev, 121(4):942, 1961.
[14] E. I. Blount, Formalisms of Band Theory, vol. 13 of Solid State
Physics, Academic Press, New York, 1962.
[15] Jacques des Cloizeaux, Energy bands and projection operators in a
crystal: Analytic and asymptotic properties. Phys. Rev. 135 (3A), 1964.
[16] P. Hohenberg and W. Kohn. Inhomogeneous electron gas, Phys.
Rev. vol. 136 (3B), p.864, Nov.1964.
[17] W. Kohn and L. J. Sham, “Self-consistent equations including
exchange and correlation effects”. Phys. Rev. 140 (4A), p.1133, 1965.
[18] Walter A. Harrison, Pseudopotentials in the theory of metals, W.A.
Benjamin, New York, 1966.
[19] Volker Heine. The Pseudopotential Concept, vol. 24 of Solid State
Physics, p. 1, Academic Press, New York, 1970.
-429-
[20] J. M. Ziman. Principles of the Theory of Solids, chap. 6, pp. 200-

203, Cambridge University Press, 2nd Edition, 1972.
[21] T. L. Gilbert, “Hohenberg-Kohn theorem for nonlocal external
potentials”, Physical Review B, vol. 12, No. 6, p.2111, 1975.
[22] P. S. Kireev, Semiconductor Physics, Mir Publisher, Moscow, 1975.
[23] C. Kittle, Introduction to Solid-State Physics, 2nd edition, John
Wiley & Sons, New York, 1976.
[24] N.W. Ashcroft, N.D.Mermin, Solid State Physics, Saunders, 1976.
[25] A. Animalu, Intermediate Quantum Theory of Crystalline Solids,
Prentice-Hall, Englewood. Cliffs, New Jersey, 1977.
[26] O. Madelung, Introduction to Solid-State Theory, Springer-Verlag,
Berlin, 1978.
[27] D. C. Langreth and M. J. Mehl. Beyond the local-density
approximation in calculations of ground-state electronic properties.
Phys. Rev. B , Vol.28, No.4, p.1809, August 1983.
[28] H. Englisch and R. Englisch. Exact density functionals for ground-
state energies. Phys. Stat. Sol. (b) 124, 373 1984.
[29] A. P. Sutton, M. W. Finnis, D. G. Pettifor and Y. Ohta, The tight-
binding bond model, J. Phys. C 21, Vol. 35, 1988.
[30] S. Wang, Fundamentals of Semiconductor Theory and Device
Physics. Englewood Cliffs, NJ: Prentice Hall, 1988.
[31] G. Bastard, Wave Mechanics Applied to Semiconductor Hetero-
structures, Les Editions de Physique, Les Ulis, France, 1990.
[32] T. A. Arias, M. C. Payne and J. D. Joannopoulos, Ab initio
molecular-dynamics techniques extended to large-length-scale systems.
Phys. Rev. B , vol.45, no. 4, p.1538, Jan. 1992.
[33] S. Goedecker and M. Teter, Tight-binding electronic-structure
calculations and tight-binding molecular dynamics with localized
orbitals, Phys. Rev. B,,vol. 51 , No15, p.9455, 1995.
[34] O. Madelung, Editor, “Semiconductor Basic Data,” Springer-
Verlag, Berlin, 2nd edition, 1996.
[35] Handbook of Organic Conductive Molecules and Polymers, Wiley,
New York, 1997.
[36] John H. Davies, The physics of low-dimensional semiconductors,
Cambridge Press, 1998.
[37] P.Y. Yu, M. Cardona, Fundamentals of Semiconductors, Graduate
Texts in Physics, 4th Edition, Springer-Verlag Berlin Heidelberg 2010.
[38] J. E. Moore, The birth of topological insulators. Nature, vol. 464, p.
194, 2010.
[39] M. K. Y. Chan and G. Ceder, Efficient Band Gap Prediction for
Solids Phys. Rev. Lett. 105, 196403 –November 2010.
-430-

Energy and Band Theory

Uploaded by

Copyright:

Available Formats

Energy and Band Theory

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

Energy and Band Theory

Uploaded by

Copyright:

Available Formats

What are some of the main topics covered in the chapter?

What are some of the main topics covered in the chapter?

How are energy bands formed in solids?

How are energy bands formed in solids?

Electronic Engineering Materials & Nanotechnology Chapter 5

ENERGY BAND THEORY

Contents of Chapter 5 (cont.):

5-8. Electron Dynamics & Effective Mass

5-1.1. Energy Levels and Energy Bands

structure of solids is the potential energy diagram of an electron in an

(a)Isolated diamond atom. (b) Array of diamond atoms.

various directions of the k-space, is usually called the energy band

By studying the energy band structure of electrons inside crystalline solids,

5-1.2. Reference Energy Levels (Vacuum level and Fermi level)

According to the Fermi-Dirac distribution, fn(E), all energy levels under a

Fig. 5-5. Vacuum energy level at metal and semiconductor/insulator–vacuum interface.

Table 5-1: Work function of some metals

Table 5-2: Electron affinity of some semiconductors

5-2. Crystal Hamiltonian Operator

The large of particles involved in a typical solid makes direct solution of

The solid is comprised of two groups of electrons: valence electrons, which

Hel = He (K.E.)+ He-e (electron-electron interaction)

Hion = Hi (K.E.)+ Hi-i(ion-ion interaction)

H e-i =  Ve-i rk  Ri  (5-3d)

H i -i = H o i -i + H ph , H e-i = H o e-i + H e- ph (5-3e)

Unfortunately it is not always possible to rigorously solve the quantum

This simplified problem is then further simplified by using the symmetry of

5-2.1. Adiabatic Approximation

 Movement of electrons (electron gas) in a stationary lattice, and

This is the basis of the so-called Born-Oppenheimer (adiabatic)

5-2.2. Jellium Model

The electron gas is considered here to be embedded in a constant positive

where H-ve includes the effect of negative space charge of electron

As far as the motion of electrons is concerned, we can therefore adopt a

Here He-i is considered with fixed coordinates of ions. The remaining

5-2.3. Hartree Equation (One-Electron Approximation)

Then, the following Schrödinger equation:

 r1 ..rN  =  i r1  2 r2 ... N rN  (5-10)

into the following one-electron equations:

Hk  k rk  = E k  k rk  (5-11)

However, the many-body problem (the electron gas which is embedded in a

5-2.4. Hartree-Fock Equation

 = 1 r1  .... N rN  (5-14)

by the following Slater determinant, which satisfies the Pauli-exclusion

The above form of Schrödinger’s equation is called the Hartree-Fock

The Hartree-Fock equation may be then written in the following simple

V`(r) is sometimes called the pseudo-potential.

5-2.5. Self-Consistent Field Method

1- Choose a set of eigen-functions  (k0 ) as a zero-order approximation

3- Substitute the initial estimate of Hartree potential VH(0) into the

5- Proceed until you reach the allowed error in  k1 or Ek.

If the Hartree-Fock potential V`(r) is taken as the total potential energy of

This equation is very interesting because it yields valid results without

5-2.6. Post-Hartee-Fock Methods

The so-called post-Hartree–Fock methods are a category of computational

5-3. Bloch Functions

where R= n1 a1 + n2 a2 + n3 a3 and a1, a2 & a3 are the basic lattice vectors,

In fact the functions uk are three-fold periodic and should satisfy:

Fig. 5-6. Schematic of the crystal periodic potential.

Note 5-1. Bloch Functions versus Wannier Functions