Вы находитесь на странице: 1из 73

Vasileska/Computational Electronics 64835_C002 Page Proof page 23 19.1.

2010 11:34am Compositor Name: PG1421

Introductory Concepts

This chapter serves as an overview of some basic concepts that will be used extensively later
in the book. Naturally, when discussing semiconductors, it is necessary to first describe the
crystal structure (Section 2.1), which incorporates three important topics: the classification
of crystals by symmetry, Miller indices, and the concept of the reciprocal space.
After introducing crystal structures, we then introduce semiconductors (Section 2.2)
and discuss their properties. An understanding of semiconductors is incomplete without
the discussion of their electronic structure (Section 2.3). More details about various
semiempirical methods for electronic structure calculation are given in Appendix A.
A summary of the important semiconductors used today is provided in Section 2.4.
The discussion about the electronic structure is naturally followed by an introduction of
the concept of the effective mass (Section 2.5), the density of states calculation for the case
of analytic band structures (Section 2.6), and an introduction to the concept of the electron
mobility (Section 2.7).
Semiconductor statistics are discussed in Section 2.8 as it serves as a prelude in the des-
cription of the operation of the most commonly used semiconductor devices described in
Section 2.9, such as diodes, bipolar junction transistors (BJTs), metal oxide semiconductor field
effect transistors (MOSFETs), silicon-on-insulator (SOI) devices, metal semiconductor
field effect transistors (MESFETs), and high electron mobility transistors (HEMTs).

2.1 Crystal Structure

In mineralogy and crystallography [1–3], a crystal structure is a unique arrangement of
atoms in a crystal. A crystal structure is composed of a basis, a set of atoms arranged in a
particular way, and a lattice, as illustrated schematically in Figure 2.1. The basis is located
upon the points of a lattice spanned by lattice vectors, which is an array of points repeated
periodically in three dimensions (3D). The set of points forming a volume that can
completely fill the space of the lattice when translated by integral multiples of the lattice
vectors is called the unit cell, as shown in the bottom half of Figure 2.1. The length of the
edge of a unit cell and the angle between the adjacent sides are called the lattice parameters.
The symmetry properties of the crystal are embodied in its space group. A crystal’s
structure and symmetry play a role in determining many of its properties, such as cleavage
planes, electronic band structure, and optical properties.
Although there are an infinite number of ways to specify a unit cell, for each crystal
structure there is a conventional unit cell, which is chosen to display the full symmetry of the
crystal. However, the conventional unit cell is not always the smallest possible choice.
A primitive unit cell of a particular crystal structure is the smallest possible volume one can
construct with the arrangement of atoms in the crystal such that, when stacked, it com-
pletely fills the space. This primitive unit cell does not always display all the symmetries

Vasileska/Computational Electronics 64835_C002 Page Proof page 24 19.1.2010 11:34am Compositor Name: PG1421

24 Computational Electronics

+ =

2D lattice Basis 2D crystal

The “basis” may be a single atom,
or a small group of atoms.

a (x̂,ŷ,ẑ)


Face-centered cube + Basis = Unit cell

Definition of a lattice, basis and crystalline structure. Top panel: Two-dimensional. Bottom panel:

inherent in the crystal. A Wigner–Seitz cell (named after E.P. Wigner and Frederick Seitz) is
a particular kind of primitive cell that has the same symmetry as the lattice. In a unit cell,
each atom has an identical environment when stacked in a 3D space. In a primitive
cell, each atom may not have the same environment. There are only seven possible crystal
systems that atoms can pack together to produce an infinite 3D space lattice in such a way
that each lattice point has an identical environment to that around every other lattice point.

2.1.1 Classification of Crystals by Symmetry

The defining property of a crystal is its inherent symmetry, by which we mean that under
certain operations the crystal remains unchanged. For example, rotating the crystal 180
degrees about a certain axis may result in an atomic configuration that is identical to the
original configuration. The crystal is then said to have a twofold rotational symmetry about
this axis. In addition to rotational symmetries like this, a crystal may have symmetries in
the form of mirror planes and translational symmetries, and also the so-called compound
symmetries that are a combination of translation and rotation=mirror symmetries. A full
classification of a crystal is achieved when all of these inherent symmetries of the crystal
are identified [4].
The crystal systems are a grouping of crystal structures according to the axial system
used to describe their lattice. Each crystal system consists of a set of three axes in a
particular geometrical arrangement. There are seven unique crystal systems. The simplest
and most symmetric, the cubic (or isometric) system, has the symmetry of a cube, that is, it
exhibits four threefold rotational axes oriented at 109.58 (the tetrahedral angle) with respect
Vasileska/Computational Electronics 64835_C002 Page Proof page 25 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 25

a a a

a a a
a a a
Simple cubic Body-centered cubic Face-centered cubic

Bravais lattices that form the cubic crystal system.

to each other. These threefold axes lie along the body diagonals of the cube. The other six
systems, in order of decreasing symmetry, are hexagonal, tetragonal, rhombohedral (also
known as trigonal), orthorhombic, monoclinic, and triclinic. Some crystallographers con-
sider the hexagonal crystal system not to be its own crystal system, but instead a part of the
trigonal crystal system. The crystal system and the Bravais lattice of a crystal describe the
(purely) translational symmetry of the crystal. The three Bravais lattices that form the cubic
crystal system are shown in Figure 2.2.
The simple cubic system consists of one lattice point on each corner of the cube. Each atom
at the lattice points is then shared equally between eight adjacent cubes; the unit cell
therefore contains in total one atom (1=8)  8 ¼ 1. The body-centered cubic (BCC) system
has one lattice point in the center of the unit cell in addition to the eight corner points. It has AQ1
a contribution of two lattice points per unit cell ((1=8)  8 þ 1) ¼ 2. Finally, the face-centered
cubic (FCC; see Figures 2.2 and 2.3) has lattice points on the faces of the cube of which each AQ2
unit cube gets exactly one half contribution, in addition to the corner lattice points, giving a
total of four atoms per unit cell ((1=8 for each corner)  8 corners þ (1=2 for each face) 
6 faces) ¼ 4. Attempting to create a C-centered cubic crystal system would result in a
simple tetragonal Bravais lattice. There are eight lattice points on a simple cubic for each
corner of the shape. There are nine lattice points for a BCC system because of the extra
point in the center of the unit. There are 14 lattice points on a FCC lattice.
When the crystal systems are combined with the various possible lattice sites, we arrive
at the Bravais lattices. They describe the geometric arrangement of the lattice points,

Interpenetrating face-centered cubic lattice that is a basic
building block of C, Si, Ge, and GaAs.
Vasileska/Computational Electronics 64835_C002 Page Proof page 26 19.1.2010 11:34am Compositor Name: PG1421

26 Computational Electronics

and thereby the translational symmetry of the crystal. In 3D, there are 14 unique Bravais
lattices that are distinct from one another in the translational symmetry they contain.
All crystalline materials recognized until now (not including quasi-crystals) fit in one of
these arrangements. The Bravais lattices are sometimes referred to as space lattices.
The crystal structure consists of the same group of atoms, the basis, positioned around
each and every lattice point. This group of atoms therefore repeats indefinitely in 3D
according to the arrangement of one of the 14 Bravais lattices. The characteristic rotation
and mirror symmetries of the group of atoms, or unit cell, is described by its crystallo-
graphic point group.
The crystallographic point group or crystal class is the mathematical group comprising
the symmetry operations that leave at least one point unmoved and that leave the appear-
ance of the crystal structure unchanged. These symmetry operations can include reflection,
which reflects the structure across a reflection plane; rotation, which rotates the structure a
specified portion of a circle about a rotation axis; inversion, which changes the sign of the
coordinate of each point with respect to a center of symmetry or inversion point; and improper
rotation, which consists of a rotation about an axis followed by an inversion. Rotation axes
(proper and improper), reflection planes, and centers of symmetry are collectively called
symmetry elements. There are 32 possible crystal classes. Each one can be classified into one
of the seven crystal systems.
The space group of the crystal structure is composed of the translational symmetry
operations in addition to the operations of the point group. These include pure translations,
which move a point along a vector; screw axis, which rotate a point around an axis while
translating parallel to the axis; and glide planes, which reflect a point through a plane
while translating it parallel to the plane. There are 230 distinct space groups.
Twenty of the 32 crystal classes are so-called piezoelectric and crystals belonging to one
of these classes (point groups) display piezoelectricity. All 20 piezoelectric classes lack a
center of symmetry. Any material develops a dielectric polarization when an electric
field is applied, but a substance that has such a natural charge separation even in the
absence of a field is called a polar material. Whether or not a material is polar is determined
solely by its crystal structure. Only 10 of the 32 point groups are polar. All polar crystals
are pyroelectric, so the 10 polar crystal classes are sometimes referred to as the pyroelectric

2.1.2 Miller Index

Miller indices [3,5] are a notation system in crystallography for planes and directions in
crystal lattices. In particular, a family of lattice planes is determined by three integers, l, m,
and n, the Miller indices. They are written (lmn) and denote planes orthogonal to a direction
(l, m, n) in the basis of the reciprocal lattice vectors (Figure 2.4). By convention, negative
integers are written with a bar, as in 3 for 3. The integers are usually written in lowest
terms, i.e., their greatest common divisor should be 1. There are also several related
notations [lmn] (see Figure 2.5), with square instead of round brackets, which denote a
direction in the basis of the direct lattice vectors instead of the reciprocal lattice. The
notation {lmn} denotes all planes that are equivalent to (lmn) by the symmetry of the
crystal. Similarly, the notation hlmni denotes all directions that are equivalent to [lmn] by
There are two equivalent ways to define the Miller indices: via a point in the reciprocal
lattice or as the inverse intercepts along the lattice vectors. Both definitions are given
below. In either case, one needs to choose the three lattice vectors a1, a2, and a3 as described
Vasileska/Computational Electronics 64835_C002 Page Proof page 27 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 27

(001) (100) (010)

(101) (110) (011)

(111) (111) (111)
Planes with different Miller indices in cubic crystals.

[001] [111] [011]

z [110]
0 x [110]

Examples of directions.

above. Given these, the three primitive reciprocal lattice vectors are also determined
(denoted as b1, b2, and b3). Then, given the three Miller indices l,m,n (lmn) denotes planes
orthogonal to lb1 þ mb2 þ nb3. That is, (lmn) simply indicates a normal to the planes in the
basis of the primitive reciprocal lattice vectors (Figure 2.6). Because the coordinates
are integers, this normal is itself always a reciprocal lattice vector. The requirement of
lowest terms means that it is the shortest reciprocal lattice vector in the given direction.

Plan (221)

z =1
Plan (111)

y = 1/2
z =1 O
x =1/2

y =1 FIGURE 2.6
Examples of determining indices for a plane
O x =1 using intercepts with axes; left (111), right (221).
Vasileska/Computational Electronics 64835_C002 Page Proof page 28 19.1.2010 11:34am Compositor Name: PG1421

28 Computational Electronics

Equivalently, (lmn) denotes a plane that intercepts the three Plan (1122)
points a1=l, a2=m, and a3=n or some multiple thereof. That is,
the Miller indices are proportional to the inverses of the inter- [001]
cepts of the plane, in the basis of the lattice vectors. If one of
the indices is zero, it means that the planes do not intersect that
axis (the intercept is ‘‘at infinity’’). The related notation [lmn] R =1/2
denotes the direction la1 þ ma2 þ na3. That is, it uses the direct
lattice basis instead of the reciprocal lattice. Note that [lmn] is not [010] S =–1/2
generally normal to the (lmn) planes, except in a cubic lattice as Q =1
described below. P =1
For the special case of simple cubic crystals, the lattice [110]
vectors are orthogonal and of equal length; similarly for the
reciprocal lattice. So, in this common case, the Miller indices FIGURE 2.7
Miller-Bravais index.
(lmn) and [lmn] both simply denote normal=directions in
Cartesian coordinates. For FCC and BCC lattices, the primitive
lattice vectors are not orthogonal. However, in these cases, the Miller indices are conven-
tionally defined relative to the lattice vectors of the cubic Bravais lattice, and hence are
again simply the Cartesian directions.
With hexagonal (see Figure 2.7) and rhombohedral crystal systems, it is possible to
use the Bravais–Miller index, which has four numbers (h, k, i, l) where i ¼ h – k and h, k,
and l are identical to the Miller index. The (001) plane has a threefold symmetry, it remains
unchanged by a rotation of 1=3 (2p=3 rad, 1208). The [100], [010], and the [110] directions
are really similar. If S is the intercept of the plane with the [110] axis, then i ¼ 1=S and i is
redundant and not necessary.
The crystallographic directions are fictitious lines linking nodes (atoms, ions, or mol-
ecules) of a crystal. The crystallographic planes are fictitious planes linking nodes. Some
directions and planes have a higher density of nodes (see Figure 2.8); these dense planes
have an influence on the behavior of the crystal:

. Optical properties: in condensed matter, the light ‘‘jumps’’ from one atom to the
other with Rayleigh scattering; the velocity of light thus varies according to
the directions, whether the atoms are close or far; this gives rise to birefringence
. Adsorption and reactivity: the adsorption and the chemical reactions occur on
atoms or molecules, these phenomena are thus sensitive to the density of nodes

Face (100)
e (21


Dense crystallographic planes.
Vasileska/Computational Electronics 64835_C002 Page Proof page 29 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 29

. Surface tension: the condensation of a material means that the atoms, ions, or
molecules are more stable if they are surrounded by other similar species; the
surface tension of an interface thus varies according to the density on the surface
. Pores and crystallites tend to have straight grain boundaries following dense
. Cleavage
. Dislocations (plastic deformation)
. The dislocation core tends to spread on dense planes (the elastic perturbation is
‘‘diluted’’); this reduces the friction (Peierls–Nabarro force), the sliding occurs
more frequently on dense planes
. The perturbation carried by the dislocation (Burgers vector) is along a dense
direction: the shift of one node in a dense direction is a lesser distortion
. The dislocation line tends to follow a dense direction, the dislocation line is often
a straight line, a dislocation loop is often a polygon

For all these reasons, it is important to determine the planes and thus to have a notation

2.1.3 Reciprocal Space

In crystallography, the reciprocal lattice of a Bravais lattice is the set of all vectors K such
that eiKR ¼ 1 for all lattice vectors R. The reciprocal lattice is itself a Bravais lattice and the
reciprocal of the reciprocal lattice is the original lattice. For a 3D lattice, defined by its
primitive vectors (a1, a2, a3), its reciprocal lattice can be determined by generating its three
reciprocal primitive vectors, through the formula

a2  a3
b1 ¼ 2p ,
a1  (a2  a3 )
a3  a1
b2 ¼ 2p , (2:1)
a2  (a3  a1 )
a1  a2
b3 ¼ 2p :
a3  (a1  a2 )

Using column vector representation of (reciprocal) primitive vectors, the formula above
can be rewritten using matrix inversion

[b1 b2 b3 ]T ¼ 2p[a1 a2 a3 ]1 : (2:2)

This method allows generalization to arbitrary dimensions. Curiously, the cross-product

formula dominates introductory materials on crystallography.
Each point (hkl) in the reciprocal lattice corresponds to a set of lattice planes (hkl) in
the real space lattice. The direction of the reciprocal lattice vector corresponds to the normal
to the real space planes, and the magnitude of the reciprocal lattice vector is equal to the
reciprocal of the interplanar spacing of the real space planes. The reciprocal lattice plays a
fundamental role in most analytic studies of periodic structures, particularly in the theory
of diffraction. For Bragg reflections in neutron and x-ray diffraction [6], the momentum
Vasileska/Computational Electronics 64835_C002 Page Proof page 30 19.1.2010 11:34am Compositor Name: PG1421

30 Computational Electronics

difference between the incoming and diffracted x-rays of a crystal is a reciprocal lattice
vector. The diffraction pattern of a crystal can be used to determine the reciprocal vectors
of the lattice. Using this process, one can infer the atomic arrangement of a crystal.
The Brillouin zone [7] is a primitive unit cell of the reciprocal lattice. In mathematics
and solid-state physics, the first Brillouin zone is a uniquely defined primitive cell of
the reciprocal lattice in the frequency domain. It is found by the same method as for the
Wigner–Seitz cell in the Bravais lattice. The importance of the Brillouin zone stems from

Brillouin zone Brillouin zone

(a) (b)

Construction of the first Brillouin zone on (a) square and (b) hexagonal structure.

Specifications of Critical Points
Symbol Description
G Center of the Brillouin zone
Simple cube
M Center of an edge
R Corner point
X Center of a face
Face-centered cubic
K Middle of an edge joining two hexagonal faces
L Center of a hexagonal face
U Middle of an edge joining a hexagonal and a square face
W Corner point
X Center of a square face
Body-centered cubic
H Corner point joining four edges
N Center of a face
P Corner point joining three edges
A Center of a hexagonal face
H Corner point
K Middle of an edge joining two rectangular faces
L Middle of an edge joining a hexagonal and a
rectangular face
M Center of a rectangular face
Vasileska/Computational Electronics 64835_C002 Page Proof page 31 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 31

the Bloch wave description of waves in a periodic medium, in which it is found that the
solutions can be completely characterized by their behavior in a single Brillouin zone.
Taking the surfaces at the same distance from one element of the lattice and its neighbors,
the volume included is the first Brillouin zone shown in Figure 2.9. Another definition is
as the set of points in k-space that can be reached from the origin without crossing any
Bragg plane. There are also second, third, etc. Brillouin zones, corresponding to a sequence
of disjoint regions (all with the same volume) at increasing distances from the origin,
but these are used more rarely. As a result, the first Brillouin zone is often called simply
the Brillouin zone. (In general, the nth Brillouin zone consists of the set of points that can be
reached from the origin by crossing n  1 Bragg planes.)
Several points of high symmetry are of special interest—these are called critical points.
These critical points are summarized in Table 2.1.
Reciprocal lattices for the cubic crystal system are as follows. We find that the simple
cubic Bravais lattice, with a cubic primitive cell of side a, has for its reciprocal lattice a
simple cubic structure with a cubic primitive cell of side 2p=a (1=a in the crystallographer’s
definition). The cubic lattice is therefore said to be dual, having its reciprocal lattice being
identical (up to a numerical factor). The reciprocal lattice to an FCC lattice is the BCC
lattice. The reciprocal lattice to a BCC lattice is the FCC lattice. It can be easily proven that
only the Bravais lattices, which have 908 between (a1, a2, a3) (cubic, tetragonal, orthorhom-
bic) have (b1, b2, b3) parallel to their real-space vectors. The first, second, and third Brillouin
zone of simple cubic, FCC, and BCC lattice are shown in Figure 2.10. In Figure 2.11,
we show the first Brillouin zone for a BCC lattice along with the high symmetry points.




First, second, and third Brillouin zones for simple cubic, FCC, and BCC lattices.
Vasileska/Computational Electronics 64835_C002 Page Proof page 32 19.1.2010 11:34am Compositor Name: PG1421

32 Computational Electronics

X [001]



Γ [010]
X X [2nd BZ]
FIGURE 2.11 [100]
First Brillouin zone of FCC lattice that corresponds to the [110]
first Brillouin zone for all diamond and zinc-blende materials
(C, Si, Ge, GaAs, InAs, CdTe, etc.). There are eight hexagonal
faces (normal to [111]) and six square faces (normal to [100]).
The sides of each hexagon and square are equal.

2.2 Semiconductors
A semiconductor is a solid that has electrical conductivity in between that of a conductor and
that of an insulator and can be controlled over a wide range, either permanently or
dynamically. Semiconductors are tremendously important in technology. Semiconductor
devices [8,9], electronic components made of semiconductor materials, are essential in
modern electrical devices. Examples range from computers to cellular phones to digital
audio players. Silicon is used to create most semiconductors commercially, but dozens of
other materials are used as well.
Semiconductors are very similar to insulators. The two categories of solids differ pri-
marily in that insulators have larger band gaps—energies that electrons must acquire to be
free to move from atom to atom. In semiconductors at room temperature, just as in
insulators, very few electrons gain enough thermal energy to surmount the band gap
from the valence band to the conduction band, which is necessary for electrons to be
available for electric current conduction. For this reason, pure semiconductors and insula-
tors in the absence of applied electric fields have a roughly similar resistance. The smaller
band gap of semiconductors, however, allows for other means besides temperature to
control their electrical properties.
The property of semiconductors that makes them most useful for constructing electronic
devices is that their conductivity may easily be modified by introducing impurities into
their crystal lattice. The process of adding controlled impurities to a semiconductor is
known as doping. The amount of impurity, or dopant, added to an intrinsic (pure) semi-
conductor varies its level of conductivity. Doped semiconductors are often referred to as
extrinsic (see Figure 2.12).
The materials chosen as suitable dopants depend on the atomic properties of both the
dopant and the material to be doped. In general, dopants that produce the desired
controlled changes are classified as either electron acceptors or donors. A donor atom
that activates (that is, becomes incorporated into the crystal lattice) donates weakly-bound
valence electrons to the material, creating excess negative charge carriers. These weakly-
bound electrons can move about in the crystal lattice relatively freely and can facilitate
Vasileska/Computational Electronics 64835_C002 Page Proof page 33 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 33

(0, 12 , 12 )
Each silicon atom bonds

( )
1 1 to four nearest neighbors.
, 0,
2 2

Each bond is composed of two

electrons, one from each atom at
( 1 1
, ,0
2 2 ) the ends of the bond.


• The principle of doping involves REPLACING a small fraction of the silicon atoms
with atoms with a DIFFERENT number of valence electrons, giving an EXCESS
number of holes or electrons:

(0, 12 , 12 ) (0, 12 , 12 )
(12 , 0, 12 ) (12 , 0, 12 )
( 1 1
, ,0
2 2 ) (12 , 12 , 0)
(0,0,0) (0,0,0)

Arsenic has one more valence electron Boron has one less valence electron
than silicon and so when we replace a than silicon and so when we replace a silicon
silicon atom with an arsenic one we atom with an boron one we obtain an
obtain an extra electron that can vacant bond that can conduct electricity.
conduct electricity, as it can be freed This leaves a hole in the valence band,
from the atom easily. and since the bond moves relatively easily,
so does the hole.

Covalent bond (top panel), arsenic impurity (bottom left) and boron impurity (bottom right).

conduction in the presence of an electric field. (The donor atoms introduce some states
under, but very close to, the conduction band edge. Electrons at these states can be easily
excited to a conduction band, becoming free electrons, at room temperature.) Conversely,
an activated acceptor produces a hole. Semiconductors doped with donor impurities are
called n-type, while those doped with acceptor impurities are known as p-type. The n- and
p-type designations indicate which charge carrier acts as the material’s majority carrier.
The opposite carrier is called the minority carrier, which exists due to thermal excitation at
a much lower concentration compared with the majority carrier.
For example, silicon has four valence electrons. In silicon, the most common dopants are
column or group III elements and column V elements. Group III elements all contain three
valence electrons, causing them to function as acceptors when used to dope silicon. Group
V elements have five valence electrons, which allows them to act as a donor. Therefore, a
silicon crystal doped with boron creates a p-type semiconductor whereas one doped with
phosphorus results in an n-type material.
Vasileska/Computational Electronics 64835_C002 Page Proof page 34 19.1.2010 11:34am Compositor Name: PG1421

34 Computational Electronics

The concentration of dopant introduced in an intrinsic semiconductor determines

its concentration and indirectly affects many of its electrical properties. The most important
factor is that doping directly affects the material’s carrier concentration. In an intrinsic
semiconductor under thermal equilibrium, the concentration of electrons and holes
is equivalent. That is, n ¼ p ¼ ni where n is the concentration of conducting electrons, p is
the electron hole concentration, and ni is the material’s intrinsic carrier concentration.
Intrinsic carrier concentration varies between materials and is dependent on temperature.
ni in silicon, for example, is roughly 1  1010 cm3 at 300 K (room temperature). In general,
an increase in doping concentration affords an increase in conductivity due to the higher
concentration of carriers available for conduction. Degenerately (very highly) doped semi-
conductors have conductivity levels comparable with metals and are often used in modern
integrated circuits as a replacement for metal. Often the superscript plus and minus
are used to denote relative doping concentrations in semiconductors. For example, nþ
denotes an n-type semiconductor with a high, often degenerate, doping concentration.
Similarly, p would indicate a very lightly doped p-type material. It is useful to note that
even degenerate levels of doping imply low concentrations of impurities with respect
to the base semiconductor. In crystalline intrinsic silicon, there are approximately
5  1022 atoms=cm3. The doping concentration for silicon semiconductors may range
anywhere from 1013 cm3 to 1020 cm3. Doping concentrations above about 1018 cm3
are considered degenerate at room temperature (see Figure 2.13). Degenerately doped
silicon contains a proportion of impurity to silicon on the order of parts per thousand.
This proportion may be reduced to parts per billion in very lightly doped silicon. Typical
concentration values fall somewhere in this range and are tailored to produce the desired
properties in the device that the semiconductor is intended for. Boron, arsenic, phosphorus,
and antimony, among other substances, are commonly used dopants in the semiconductor
Doping a semiconductor crystal introduces allowed energy states within the band gap
but very close to the energy band that corresponds with the dopant type. In other words,
donor impurities create states near the conduction band while acceptors create states
near the valence band. The gap between these energy states and the nearest energy band
is usually referred to as dopant-site bonding energy or EB and is relatively small. For
example, EB for boron in silicon bulk is 0.045 eV, compared with silicon’s band gap of
about 1.12 eV. Because EB is so small, it takes little energy to ionize the dopant atoms
and create free carriers in the conduction or valence bands. Usually the thermal energy
available at room temperature is sufficient to ionize most of the dopant.
Dopants also have the important effect of shifting the material’s Fermi level towards
the energy band that corresponds with the dopant with the greatest concentration.
Since the Fermi level must remain constant in a system in thermodynamic equilibrium,

Hydrogenic impurity states overlap form impurity band

Density of states


Schematic description of band gap 0
narrowing due to heavy doping
with donor impurities. For heavy doping the impurity band overlaps with the VB or CB
Vasileska/Computational Electronics 64835_C002 Page Proof page 35 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 35

stacking layers of materials with different properties leads to many useful electrical
properties. For example, the p-n junction properties are due to the energy band bending
that happens as a result of lining up the Fermi levels in contacting regions of p-type and
n-type material. This effect is shown in a band diagram. The band diagram typically
indicates the variation in the valence band and conduction band edges versus some spatial
dimension, often denoted as x. The Fermi energy is also usually indicated in the diagram.
Sometimes the intrinsic Fermi energy, Ei, which is the Fermi level in the absence of doping,
is shown. These diagrams are useful for explaining the operation of many kinds of
semiconductor devices. Junctions between regions of n- and p-type semiconductors create
electric fields, which cause electrons and holes to be available to move away from them,
and this effect is critical to semiconductor device operation. Also, a density difference in the
amount of impurities produces a small electric field in the region, which is used to
accelerate nonequilibrium electrons or holes.
In addition to permanent modification through doping, the resistance of semiconductors
is normally modified dynamically by applying electric fields. The ability to control
resistance=conductivity in regions of semiconductor material externally through the appli-
cation of electric fields is the feature that makes semiconductors useful. It has led to the
development of a broad range of semiconductor devices, like transistors and diodes.
Semiconductor devices that have dynamically controllable conductivity, such as tran-
sistors, are the building blocks of integrated circuit devices like the microprocessor.
These ‘‘active’’ semiconductor devices (transistors) are combined with passive components
implemented from semiconductor material, such as capacitors and resistors, to produce
complete electronic circuits (see Figure 2.14).
In most semiconductors, when electrons lose enough energy to fall from the conduction
band to the valence band (the energy levels above and below the band gap), they often
emit light, a quantum of energy in the visible electromagnetic spectrum. This photoemis-
sion process underlies the light-emitting diode (LED) and the semiconductor laser, both of
which are very important commercially. Conversely, the semiconductor absorption of
light in photodetectors excites electrons to move from the valence band to the higher
energy conduction band, thus facilitating the detection of light and varying with its
intensity. This is useful for fiber optic communications and for providing the basis for
energy from solar cells.

PCs (left) and a magnified view of a Pentium IV processor.
Vasileska/Computational Electronics 64835_C002 Page Proof page 36 19.1.2010 11:34am Compositor Name: PG1421

36 Computational Electronics

Semiconductors may be elemental materials, such as silicon and germanium; compound

semiconductors, such as gallium arsenide (GaAs) and indium phosphide; or alloys, such
as silicon germanium or aluminum gallium arsenide.

2.3 Band Structure

There are two popular ways to describe the electronic structure of a crystal. The first starts
from single atoms. An atom has discrete energy levels. When two atoms come close to each
other, the energy levels split into an upper and a lower level (bonding and antibonding
states in covalent bond theory), whereby they delocalize across the two atoms. With more
atoms, the number of levels increases and the groups of levels form bands. Semiconductors
contain many bands. If there is a large distance between the highest occupied state and
the lowest unoccupied state, then a gap will likely remain between occupied and unoccu-
pied bands even after band formation.
A second approach starts with free electrons waves. Upon the introduction of the
electrostatic potential associated with the atomic cores, the waves undergo Bragg reflection
and cannot penetrate the bulk, which opens the band gap in the free electron energy
momentum (dispersion) relations.
In Figure 2.15, we show the results of the simulations in which the piece-wise-constant
potential barrier tool model on the NCN Nanohub (www.nanoHUB.org) has been used.
In this example, the ‘‘crystal’’ is a set of quantum wells with a finite height that holds a certain
number of discrete energy levels. One can clearly see that when N such identical wells are
brought together, they interact among each other through the overlap of the bound states and
in well, and the degeneracy of the levels in the well is lifted, thus forming energy bands. On the
bottom panel of Figure 2.15, we also show the formation of the cosine-like bands and the
associated energy gaps for the case of 40 wells. One can also observe that as these are quasi-
bound states [10], the transmission coefficient equals unity at resonance. This fact is shown in
the plot of the total transmission coefficient that is shown in the middle panel of Figure 2.15.
The actual calculated band structure for the Si diamond lattice is shown in Figure 2.16,
which is more complicated due to the 3D nature of the problem. Shown are the energy
versus crystal momentum along certain high symmetry directions in the first Brillouin zone
of the Si reciprocal lattice.
Like in other solids, the electrons in semiconductors can have energies only within
certain bands (i.e., ranges of levels of energy) between the energy of the ground state,
corresponding to electrons tightly bound to the atomic nuclei of the material, and the free
electron energy, which is the energy required for an electron to escape entirely from the
material. The energy bands each correspond to a large number of discrete quantum states
of the electrons, and most of the states with low energy (closer to the nucleus) are full, up to
the top of a particular band called the valence band. Semiconductors and insulators are
distinguished from metals because the valence band in semiconductor materials is very
nearly full under the usual operating conditions, with relatively few electrons in the
conduction band associated with thermal excitation across the gap for intrinsic materials.
The ease with which electrons in a semiconductor can be excited from the valence band
to the conduction band depends on the band gap between the bands, and it is the size of
this energy band gap that serves as an arbitrary dividing line (roughly 4 eV) between
semiconductors and insulators, as shown in Figure 2.17.
Vasileska/Computational Electronics 64835_C002 Page Proof page 37 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 37

Formation of bands on the example of multiple quantum wells interacting with each other. Top (left) panel:
Distribution of energy states for 40 quantum interacting wells. Top (right) panel: Transmission coefficient through
the structure. Notice perfect transmission at resonances. Bottom panel: Formation of energy bands with cosine like
dispersion relation. These results are obtained with the piece-wise constant potential barrier tool (PCPBT available
on the NanoHUB, www.nanohub.org).

Γ΄25 X1
0 EV
E (eV)



L Л Γ Δ X U, K Σ Γ More complex representation of band structure in
k silicon showing wavevector dependence.
Vasileska/Computational Electronics 64835_C002 Page Proof page 38 19.1.2010 11:34am Compositor Name: PG1421

38 Computational Electronics



Electron energy
Fermi level Bandgap

FIGURE 2.17 Valence

Band structure of a semiconductor band
showing the nearly full valence band
and the nearly empty conduction band. Metal Semiconductor Insulator

A more complete view of the band structure takes into account the periodic nature of a
crystal lattice. The Schrödinger equation is solved for the periodic potential associated with
the crystal lattice [11,12], which is in the form of Bloch waves

cnk (r) ¼ eikr unk (r), (2:3)

k is called the wavevector and is related to the direction of motion of the electron in the
n is the band index
unk(r) has the periodicity of the real space lattice

The wavevector k takes on values within the first Brillouin zone (BZ) corresponding to the
crystal lattice, and particular high symmetry directions=points in the BZ are assigned
conventional names like G, D, L, S, etc. The available energies for the electron also depend
upon k, as shown for silicon in Figure 2.16. In this diagram, the topmost energy of the
valence band is labeled Ev and the bottom energy in the conduction band is labeled Ec.
The top of the valence band is not directly below the bottom of the conduction band (one is
at G, the other at X), so silicon is called an indirect gap material. In other semiconductors
(for example, GaAs), both are at G and these materials are called direct gap materials.
Although electronic band structures are usually associated with crystalline materials,
quasi-crystalline and amorphous solids may also exhibit band structures. However, the
periodic nature and symmetrical properties of crystalline materials makes it much easier to
examine the band structures of these materials theoretically. In addition, the well-defined
symmetry axes of crystalline materials makes it possible to determine the dispersion
relationship between the momentum (a 3D vector quantity) and energy of a material. As
a result, virtually all of the existing theoretical work on the electronic band structure of
solids has focused on crystalline materials. Examples of the band structures of the most
commonly used semiconductors are shown in Figure 2.18, calculated using the band
structure lab that resides on NanoHub (www.nanohub.org).
The energy versus crystal moment, k, is referred to as the dispersion relation between the
energy and momentum of electrons in reciprocal space and is related to the actual momen-
tum of electrons in the crystal. For crystalline structures, the dispersion relation of the
electrons is periodic, and the Brillouin zone is the smallest repeating space within this
periodic structure (see Figure 2.11). For an infinitely large crystal, if the dispersion relation
Vasileska/Computational Electronics 64835_C002 Page Proof page 39 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 39

Band structure for Si (top left), Ge (top right), GaAs (bottom left), and InAs (bottom right). These figures were
generated with the tool called Bandstructure Lab that can be found at www.nanohub.org.

for an electron is defined throughout the Brillouin zone, then it is defined throughout the
entire reciprocal space. This Ansatz is a special case of electron waves in a periodic crystal
lattice using Bloch waves as treated generally in the dynamical theory of diffraction. Every
crystal is a periodic structure that can be characterized by a Bravais lattice, and for each
Bravais lattice we can determine the reciprocal lattice, which encapsulates the periodicity in
a set of three reciprocal lattice vectors (b1, b2, b3). Now, any periodic potential, V(r), which
shares the same periodicity as the direct lattice, can be expanded out as a Fourier series
whose only nonvanishing components are those associated with the reciprocal lattice
vectors. So the expansion can be written as

V(r) ¼ VK eiKr , (2:4)

where K ¼ m1b1 þ m2b2 þ m3b3 for any set of integers (m1, m2, m3). From this theory, an
attempt can be made to predict the band structure of a particular material.
The nearly free electron approximation in solid-state physics is similar in some
respects to the hydrogen-like atom of quantum mechanics in that interactions between
electrons are completely ignored. This allows the use of Bloch’s theorem, which states
Vasileska/Computational Electronics 64835_C002 Page Proof page 40 19.1.2010 11:34am Compositor Name: PG1421

40 Computational Electronics

that electrons in a periodic potential have wavefunctions and energies that are periodic in
wavevector up to a constant phase shift between neighboring reciprocal lattice vectors.
The consequences of periodicity are described mathematically by the Bloch wave-
function cnk(r) ¼ eikrunk(r), where unk(r) is periodic over the crystal lattice, that is,
unk(r) ¼ unk(r  R). Here index n refers to the nth energy band, wavevector k is related to
the direction of motion of the electron, r is the position in the crystal, and R is the location
of an atomic site. (For more details on the semiempirical energy band structure calculation,
see Appendix A).
The opposite extreme to the nearly free electron approximation assumes the electrons in
the crystal behave much like an assembly of constituent atoms. This tight-binding model
assumes the solution to the time-independent single electron Schrödinger equation is
well approximated by a linear combination of atomic orbitals cn (r). Namely
C(r) ¼ bn,R cn (r  R), (2:5)

where the coefficients bn,R are selected to give the best approximate solution of this form.
Index n refers to an atomic energy level and R refers to an atomic site. A more accurate
approach using this idea employs Wannier functions, defined by
an (r  R) ¼ dkeik(Rr) unk , (2:6)

in which unk is the periodic part of the Bloch wave and the integral is over the Brillouin
zone. Here, index n refers to the nth energy band in the crystal. The Wannier functions
are localized near atomic sites, like atomic orbitals, but being defined in terms of
Bloch functions they are accurately related to solutions based upon the crystal potential.
Wannier functions on different atomic sites R are orthogonal. The Wannier functions can
be used to form the Schrödinger solution for the nth energy band as
cnk (r) ¼ eik(Rr) an (r  R): (2:7)

A detailed description of the semiempirical methods for the electronic structure calculation
is given in Appendix A.

2.4 Preparation of Semiconductor Materials

Semiconductors with predictable, reliable electronic properties are necessary for mass
production. The level of chemical purity needed is extremely high because the presence
of impurities even in very small proportions can have large effects on the properties of
the material (Table 2.2). A high degree of crystalline perfection is also required, since AQ3
faults in crystal structure (such as dislocations, twins, and stacking faults) interfere
with the semiconducting properties of the material. Crystalline faults are a major cause
Vasileska/Computational Electronics 64835_C002 Page Proof page 41 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 41

Material Properties of Germanium, Silicon, and Gallium Arsenide
Name Symbol Germanium Silicon Arsenide
Band minimum at k ¼ 0
Minimum energy Eg,direct (eV) 0.8 3.2 1.424
Effective mass m*e =m0 0.041 ?0.2? 0.067 AQ4
Band minimum not at k ¼ 0
Minimum energy Eg,indirect (eV) 0.66 1.12 1.734
Longitudinal effective mass m*e,l=m0 1.64 0.98 1.98
Transverse effective mass * =m0
me,t 0.082 0.19 0.37
Longitudinal direction (111) (100) (111)
Heavy hole valence band maximum at E ¼ k ¼ 0
Effective mass * =m0
mhh 0.28 0.49 0.45
Light hole valence band maximum at k ¼ 0
Effective mass mlh* =m0 0.044 0.16 0.082
Split-off hole valence band maximum at k ¼ 0
Split-off band valence band energy Ev,so (eV) 0.028 0.044 0.34
Effective mass mh:so* =m0 0.084 0.29 0.154
m0 ¼ 9.11  1031 kg is the free electron rest mass.

of defective semiconductor devices. The larger the crystal, the more difficult it is to
achieve the necessary perfection. Current mass production processes use crystal ingots
between 4 and 12 in. (300 mm) in diameter, which are grown as cylinders and sliced
into wafers.
Because of the required level of chemical purity and the perfection of the crystal
structure, which are needed to make semiconductor devices, special methods have been
developed to produce the initial semiconductor material. A technique for achieving high
purity includes growing the crystal using the Czochralski process. An additional step that
can be used to further increase purity is known as zone refining. In zone refining, part of
a solid crystal is melted. The impurities tend to concentrate in the melted region, while
the desired material recrystallizes leaving the solid material more pure and with fewer
crystalline faults. In manufacturing semiconductor devices involving heterojunctions
between different semiconductor materials, the lattice constant, which is the length of
the repeating element of the crystal structure, is important for determining the compati-
bility of materials.
Semiconductor materials are insulators at a temperature of absolute zero that conduct
electricity in a limited way at room temperature. The defining property of a semiconductor
material is that it can be doped with impurities that alter its electronic properties in a
controllable way. Because of their application in devices like transistors (and therefore
computers) and lasers, the search for new semiconductor materials and the improvement
of existing materials is an important field of study in materials science. The most com-
monly used semiconductor materials are crystalline inorganic solids. These materials
can be classified according to the periodic table groups from which their constituent
atoms come. The group III nitrides have high tolerance to ionizing radiation, making
Vasileska/Computational Electronics 64835_C002 Page Proof page 42 19.1.2010 11:34am Compositor Name: PG1421

42 Computational Electronics

them suitable for radiation-hardened electronics. A comprehensive list of semiconductor

materials used in different device type fabrications is given in following table.
List of semiconductor materials

. Group IV elemental semiconductors

. Diamond (C)
. Silicon (Si)
. Germanium (Ge)
. Group IV compound semiconductors
. Silicon carbide (SiC)
. Silicon germanide (SiGe)
. Group III–V semiconductors
. Aluminum antimonide (AlSb)
. Aluminum arsenide (AlAs)
. Aluminum nitride (AlN)
. Aluminum phosphide (AlP)
. Boron nitride (BN)
. Boron phosphide (BP)
. Boron arsenide (BAs)
. Gallium antimonide (GaSb)
. Gallium arsenide (GaAs)
. Gallium nitride (GaN)
. Gallium phosphide (GaP)
. Indium antimonide (InSb)
. Indium arsenide (InAs)
. Indium nitride (InN)
. Indium phosphide (InP)
. Group III–V ternary semiconductor alloys
. Aluminum gallium arsenide (AlGaAs, AlxGa1xAs)
. Indium gallium arsenide (InGaAs, InxGa1xAs)
. Indium gallium phosphide (InGaP)
. Aluminum indium arsenide (AlInAs)
. Aluminum indium antimonide (AlInSb)
. Gallium arsenide nitride (GaAsN)
. Gallium arsenide phosphide (GaAsP)
. Aluminum gallium nitride (AlGaN)
. Aluminum gallium phosphide (AlGaP)
. Indium gallium nitride (InGaN)
. Indium arsenide antimonide (InAsSb)
. Indium gallium antimonide (InGaSb)
Vasileska/Computational Electronics 64835_C002 Page Proof page 43 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 43

. Group III–V quaternary semiconductor alloys

. Aluminum gallium indium phosphide (AlGaInP, also InAlGaP, InGaAlP, AlInGaP)
. Aluminum gallium arsenide phosphide (AlGaAsP)
. Indium gallium arsenide phosphide (InGaAsP)
. Aluminum indium arsenide phosphide (AlInAsP)
. Aluminum gallium arsenide nitride (AlGaAsN)
. Indium gallium arsenide nitride (InGaAsN)
. Indium aluminum arsenide nitride (InAlAsN)
. Group III–V quinary semiconductor alloys
. Gallium indium nitride arsenide antimonide (GaInNAsSb)
. Group II–VI semiconductors
. Cadmium selenide (CdSe)
. Cadmium sulfide (CdS)
. Cadmium telluride (CdTe)
. Zinc oxide (ZnO)
. Zinc selenide (ZnSe)
. Zinc sulfide (ZnS)
. Zinc telluride (ZnTe)
. Group II–VI ternary alloy semiconductors
. Cadmium zinc telluride (CdZnTe, CZT)
. Mercury cadmium telluride (HgCdTe)
. Mercury zinc telluride (HgZnTe)
. Mercury zinc selenide (HgZnSe)
. Group I–VII semiconductors
. Cuprous chloride (CuCl)
. Group IV–VI semiconductors
. Lead selenide (PbSe)
. Lead sulfide (PbS)
. Lead telluride (PbTe)
. Tin sulfide (SnS)
. Tin telluride (SnTe)
. Group IV–VI ternary semiconductors
. lead tin telluride (PbSnTe)
. Thallium tin telluride (Tl2SnTe5)
. Thallium germanium telluride (Tl2GeTe5)
. Group V–VI semiconductors
. Bismuth telluride (Bi2Te3)
. Group II–V semiconductors
. Cadmium phosphide (Cd3P2)
Vasileska/Computational Electronics 64835_C002 Page Proof page 44 19.1.2010 11:34am Compositor Name: PG1421

44 Computational Electronics

. Cadmium arsenide (Cd3As2)

. Cadmium antimonide (Cd3Sb2)
. Zinc phosphide (Zn3P2)
. Zinc arsenide (Zn3As2)
. Zinc antimonide (Zn3Sb2)
. Layered semiconductors
. Lead(II) iodide (PbI2)
. Molybdenum disulfide (MoS2)
. Gallium Selenide (GaSe)
. Tin sulfide (SnS)
. Bismuth Sulfide (Bi2S3)
. Others
. Copper indium gallium selenide (CIGS)
. Platinum silicide (PtSi)
. Bismuth(III) iodide (BiI3)
. Mercury(II) iodide (HgI2)
. Thallium(I) bromide (TlBr)
. Miscellaneous oxides
. Titanium dioxide: anatase (TiO2)
. Copper(I) oxide (Cu2O)
. Copper(II) oxide (CuO)
. Uranium dioxide (UO2)
. Uranium trioxide (UO3)
. Organic semiconductors
. Magnetic semiconductors

By far, silicon (Si) is the most widely used material in semiconductor devices. Its combin-
ation of low raw material cost, relatively simple processing, and a useful temperature
range make it currently the best compromise among the various competing materials.
Silicon used in semiconductor device manufacturing is currently fabricated into rods
that are large enough in diameter to allow the production of 300 mm (12 in.) wafers.
Germanium (Ge) was a widely used early semiconductor material but its thermal sensi-
tivity makes it less useful than silicon. Today, germanium is often alloyed with silicon
for use in very-high-speed SiGe devices; IBM is a major producer of such devices. GaAs is
also widely used in high-speed devices, but so far, it has been difficult to form large-
diameter boules of this material, limiting the wafer diameter to sizes significantly smaller
than silicon wafers, thus making mass production of GaAs devices significantly more
expensive than silicon. Other less common materials are also in use or under investigation.
Silicon carbide (SiC) has found some application as the raw material for blue LEDs and
is being investigated for use in semiconductor devices that could withstand very
high operating temperatures and environments with the presence of significant levels of
ionizing radiation. IMPATT diodes have also been fabricated from SiC. Various indium
compounds (indium arsenide, indium antimonide, and indium phosphide) are also being
Vasileska/Computational Electronics 64835_C002 Page Proof page 45 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 45

used in LEDs and solid-state laser diodes. Selenium sulfide is being studied in the
manufacture of photovoltaic solar cells.
Wide band gap semiconductors are semiconductor materials with electronic band gaps
larger than two to three electron volts (eV) (the exact threshold ‘‘wideness’’ often depends
on the application though). They have many applications in optoelectronic and power
devices, especially where high-temperature operation is important. Gallium nitride (GaN)
is used to make blue LEDs and lasers, for example. In the future, high brightness, long-life
white LEDs may replace incandescent bulbs in many situations. The next generation of
DVD players (the Blu-ray and HD DVD formats) uses GaN based lasers. Aluminum
nitride can be used to fabricate ultraviolet LEDs with wavelengths down to 200–250 nm.
Due to its cost, it is so far used mostly in military applications. Devices for high power and
high temperature applications have been developed. Both gallium nitride and silicon
carbide are robust materials well suited for such applications. Cubic boron nitride is
used as well. Most of these are for specialist applications in space programs and military
systems. However, they have not begun to displace silicon from its leading place in the
general power semiconductor market.

2.5 Effective Mass

In solid-state physics, a particle’s effective mass is a concept related to the semiclassical model
of transport in a crystal where it represents the mass with which electrons respond to
external forces like an electric field. It can be shown that, under most conditions, electrons
and holes in a crystal respond to electric and magnetic fields almost as if they were
free particles in a vacuum, but with a different mass. This mass is usually stated in units
of the ordinary mass of an electron me (9.11  1031 kg).
When an electron is moving inside the material, the force between other atoms will affect
its movement and it will not be described directly by Newton’s law. So we introduce
the concept of effective mass to describe the movement of electrons in Newton’s law.
The effective mass can be less than or greater than the free electron mass or negative in the
case of holes. The effective mass is defined by an analogy with Newton’s second law
F ¼ ma. Using quantum mechanics, it can be shown that for an electron in an external
electric field E

1 d2 E
a¼ qE, (2:8)
h 2 dk2

a is the acceleration
h is the reduced Planck’s constant
k is the wave number (often loosely called momentum since k ¼ p=h  for free electrons)
e(k) is the energy as a function of k or the dispersion relation as it is often called

From the external electric field alone, the electron would experience a force of F ¼ qE,
where q is the charge. Hence, under the model in which only the external electric field acts,
effective mass m* becomes
Vasileska/Computational Electronics 64835_C002 Page Proof page 46 19.1.2010 11:34am Compositor Name: PG1421

46 Computational Electronics

d2 E(k)
m* ¼ h 2 : (2:9)

For a free particle, the dispersion relation is a quadratic, and so the effective mass would
be constant (and equal to the real mass). In a crystal, the situation is far more complex.
The dispersion relation is not even approximately quadratic, over the whole BZ. However,
wherever a minimum occurs in the dispersion relation, the minimum can be approximated
by a quadratic curve in the small region around that minimum. Hence, for electrons that
have energy close to a minimum, effective mass is a useful concept. In energy regions far
away from a minimum, effective mass can be negative or even approach infinity. Effective
mass, being generally dependent on direction (with respect to the crystal axes), is a tensor.
However, for most calculations the various directions can be averaged out.
Many semiconductors can be described as having a one band minimum at k ¼ 0 as well
as several equivalent anisotropic band minima at k 6¼ 0. In addition, there are three band
maxima of interest that are close to the valence band edge. As an example, we consider the
band structure of silicon as shown in the figure below.
The E-k diagram for Si within the first Brillouin zone and along the (100) direction
is shown in Figure 2.19. The energy is chosen to be to zero at the edge of the valence
band. The lowest band minimum at k ¼ 0 and still above the valence band edge occurs at
Ec,direct ¼ 3.2 eV. This is not the lowest minimum above the valence band edge since there
are also six equivalent minima at k ¼ (x, 0, 0), (x, 0, 0), (0, x, 0), (0, x, 0), (0, 0, x), and
(0, 0, x) with x ¼ 5 nm1 (see Figure 2.20). The minimum energy of all these minima
equals 1.12 eV ¼ Ec,indirect. The effective mass of these anisotropic minima is characterized
by a longitudinal mass along the corresponding equivalent (100) direction and two trans-
verse masses in the plane perpendicular to the longitudinal direction. In silicon, the
longitudinal electron mass is me:l ¼ 0.98m0 and the transverse electron masses are
* ¼ 0.19m0, where m0 ¼ 9.11  1031 kg is the free electron rest mass.
Two of the three hole band maxima (a hole is essentially a missing electron in the valence
band) occur at 0 eV (see Figure 2.19). These bands are referred to as the light and heavy
hole bands with a light hole mass of mlh* ¼ 0.16m0 and a heavy hole mass of mhh * ¼ 0.46m0.
In addition, there is a split-off hole band with its maximum at Ev,so ¼ 0.044 eV and a split-
off hole mass of m*v,so ¼ 0.29m0. The constant energy surfaces of the heavy-hole, light-hole,
and split-off bands are shown in Figure 2.21.
The values of the energy band minima and maxima as well as the effective masses for
germanium, silicon, and GaAs are listed in Table 2.3. AQ5
The effective mass for density of states calculations equals the mass that provides the
density of states using the expression for one isotropic maximum or minimum or

Energy (eV)

FIGURE 2.19 –2
E-k diagram within the first Brillouin zone along the –6 –4 –2 0 2 4 6
(100) direction. Wave number (1/nm)
Vasileska/Computational Electronics 64835_C002 Page Proof page 47 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 47

The first brillouin zone

001 111

010 100 111


001 111
111 111

X valleys of Si L valleys of Ge

Constant energy surfaces of the conduction band of Si and Ge. Note that in the case of Ge we have four conduction
band minima (since the band minima occurs on the edge of the BZ1), in the case of Si we have six conduction band
equivalent valleys. In the case of GaAs (not shown in this figure) we have only one constant energy surface at the
center of the Brillouin zone.

6E+09 3E+09 2E+09

KY (m–1)

KY (m–1)

KY (m–1)

0 0 0 0.4

–6E+09 –3E+09 –2E+09

0 6E+09 0 3E+09 0 2E+09
KX (m–1) KX (m–1) KX (m–1)

Three-dimensional equi-energy surfaces of heavy hole, light hole, and split off band in Si for kz ¼ 0.

Effective Mass and Energy Band Gap of Ge, Si, and GaAs
Name Symbol Germanium Silicon Arsenide
Smallest energy band gap at 300 K Eg (eV) 0.66 1.12 1.424
Effective mass for density of states calculations
Electrons m*e:dos=m0 0.56 1.08 0.067
Holes * =m0
mh:dos 0.29 0.57=0.81a 0.47 AQ6
Effective mass for conductivity calculations
Electrons * =m0
me:cond 0.12 0.26 0.067
Holes * =m0
mh:cond 0.21 0.36=0.386a 0.34
m0 ¼ 9.11  1031 kg is the free electron rest mass.
Vasileska/Computational Electronics 64835_C002 Page Proof page 48 19.1.2010 11:34am Compositor Name: PG1421

48 Computational Electronics

8p 2 *3=2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
gc (E) ¼ me E  EC for E  Ec (2:10a)

for the density of states in the conduction band and

8p 2 *3=2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
gV (E) ¼ mh EV  E for E  EV (2:10b)

for the density of states in the valence band.

For instance, for a single band minimum described by a longitudinal mass and two
transverse masses, the effective mass for density of states calculations is the geometric
mean of the three masses. Including the fact that there are several equivalent minima at the
same energy, one obtains the effective mass for density of states calculations from

* ¼ MC (ml ml mt )1=3 ,
me,dos (2:11)

where Mc is the number of equivalent band minima. For silicon, one obtains

* ¼ (ml mt mt )1=3 ¼ (6)2=3 (0:89  0:19  0:19)1=3 m0 ¼ 1:08m0 :

me,dos (2:12)

The effective mass for conductivity calculation is the mass that is used in conduction
related problems accounting for the detailed structure of the semiconductor. These calcu-
lations include mobility and diffusion constant calculations. Another example is the
calculation of the shallow impurity levels using a hydrogen-like model. As the conductiv-
ity of a material is inversely proportional to the effective masses, one finds that the
conductivity due to multiple band maxima or minima is proportional to the sum of the
inverse of the individual masses, multiplied by the density of carriers in each band, as each
maximum or minimum adds to the overall conductivity. For anisotropic minima contain-
ing one longitudinal and two transverse effective masses, one has to sum over the effective
masses in the different minima along the equivalent directions. The resulting effective mass
for bands, which have ellipsoidal constant energy surfaces, is given by

* ¼ , (2:13)
(1=ml ) þ (1=ml ) þ (1=mt )

provided the material has an isotropic conductivity as is the case for cubic materials. For
instance, electrons in the X minima of silicon have an effective conductivity mass given by

* ¼ 3  (1=ml ) þ (1=mt ) þ (1=mt )1
¼ 3  (1=0:89) þ (1=0:19) þ (1=0:19)1 m0
¼ 0:26m0 : (2:14)

When the electric field applied to the semiconductor is small, the kinetic energy of carriers
is small, i.e., carriers occupy states near the band edges. In this region, the curvature of the
bands is parabolic and the energy-wavevector relationship is of the form

h 2 k2 h 2  2 
EK ¼ ¼ kx þ ky2 þ kz2 , (2:15)
2m* 2m*
Vasileska/Computational Electronics 64835_C002 Page Proof page 49 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 49

Ek is the kinetic energy
m* is the effective mass at the band edge
kx , ky , and kz are the wavevector components along the x-, y-, and z-axis, respectively

When the applied electric fields are large, as is always the case in normal device operation,
carriers can have very high kinetic energy, i.e., occupy states that are high in energy
relative to the conduction band minima. In this region, the E-k relationship may differ
from a parabolic band structure and the nonparabolic energy band structure described by

h 2 k2
Ek (1 þ aEk ) ¼ ¼ g(k) (2:16)

gives a more appropriate description of the actual band structure. In Equation 2.16, the
nonparabolicity factor a is defined as
1 mc
a¼ 1 , (2:17)
Eg m0

Eg is the energy difference between the conduction band and the valence band at the
G point
m0 is the electron rest mass
mc is the conductivity mass

The value of a is around 0.47 for silicon.

Having defined the E-k relationship, we can calculate the carrier group velocity using

v ¼ rk E k , (2:18)

where rk is a delta operator with respect to k. Substituting Equation 2.3a into Equation 2.5
leads to the following expression

h k
v¼ (2:19)

for the carrier velocity for parabolic band structure. For nonparabolic bands, the velocity
wavevector relation is of the form

h k 1
v¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (2:20)
m* 1 þ 4ag(k)

Experimentally, effective masses were measured using cyclotron resonance, a method in

which microwave absorption of a semiconductor immersed in a magnetic field goes through
a sharp peak when the microwave frequency equals the cyclotron frequency vc ¼ eB=m*.
In recent years, effective masses have more commonly been determined through the meas-
urement of band structures using techniques such as angle-resolved photoemission (ARPES)
Vasileska/Computational Electronics 64835_C002 Page Proof page 50 19.1.2010 11:34am Compositor Name: PG1421

50 Computational Electronics

or, most directly, the de Haas-van Alphen effect. Effective masses can also be estimated
using the coefficient g of the linear term in the low-temperature electronic specific heat at a
constant volume of cv. The specific heat depends on the effective mass through the density of
states at the Fermi level and as such is a measure of degeneracy as well as band curvature.
Since carrier mobility depends on the ratio of carrier collision lifetime t to effective mass,
masses can in principle be determined from transport measurements, but this method is not
practical since carrier collision probabilities are typically not known a priori. Groups III–V
compounds based on GaAs and InSb have far smaller effective masses than tetrahedral
group IV materials like Si and Ge. In the simplest Drude picture of electronic transport, the
maximum obtainable charge carrier velocity is inversely proportional to the effective mass
v ¼ mE, where the mobility is given by m ¼ et=m* with e being the electronic charge.
The ultimate speed of integrated circuits depends on the carrier velocity, so the low effective
mass is the fundamental reason that GaAs and its derivatives are used instead of Si in high-
bandwidth applications like cellular telephony.
In the rest of this section, we describe the key steps in determining the effective masses
for a general conduction band ellipsoid as illustrated in Figure 2.22. In the ellipse coord-
inate system (ECS), the constant energy ellipsoid can be expressed as

(k̂3) Device
Y Source Top gate Drain
Bottom gate
(k̂΄3) (001)

(100) (111)

Crystal ( )

k 2
k 1

Schematic description of the three orthogonal coordinate
systems: device coordinate system (DCS), crystal coord-
inate system (CCS), and ellipse coordinate system (ECS).
(From Lundstrom and co-workers.) AQ7
Vasileska/Computational Electronics 64835_C002 Page Proof page 51 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 51

h 2 kk2
h 2 k?1 þ k?2
E¼ þ : (2:21)
2ml 2mt

In Equation 2.21, the k-space origin is translated to the conduction-band minima, which
serves as the reference for the electronic energy. In compact vector notation, Equation 2.21
can be written as

h 2 T 1

E¼ k M kE , (2:22)
2 E E

where kE ¼ kk k?1 k?2 consists of the components of an arbitrary wavevector in the ECS
and the inverse M1 1 1 1
E is a 3  3 diagonal matrix with ml , mt , mt along the diagonal. For a
given channel material and for a given conduction band ellipsoid, the directions of the unit
basis vectors kk , k?1 , k?2 relative to the crystal coordinate system (CCS) are known, thus
allowing one to write the 3  3 rotation matrix RE C , which transforms the components of
0 0 0
an arbitrary vector kC ¼ (k1  k2  k3 )T defined in the CCS, to its components in the ECS, i.e.,

kE ¼ R E C kC : (2:23)

A similar rotation matrix RC D transforms a wavevector kD ¼ (k1  k2  k3 )T in the device

coordinate system (DCS) to kC in the CCS as

kC ¼ RC D kD : (2:24)

By combining Equations 2.23 and 2.24, we obtain

kE ¼ RE D kD , (2:25)

where the rotation matrix is defined as RE D ¼ RE C RC D. Thus,

h 2 T 1

E¼ k M kD , (2:26)
2 D D

where the inverse effective mass in the DCS is

D ¼ RTE D M1
E RE D: (2:27)

The above procedure is valid for zinc-blende material systems. A similar procedure could
be applied for wurtzite material systems.
The listing of the Fortran code that performs the following coordinate transformation is
given below:

C Program that calculates the effective masses for different

C Crystalographic orientations
implicit real*8(a-h, o-z)
real rml, rmt
real*8 R1(3,3),R2(3,3),R3(3,3)
Vasileska/Computational Electronics 64835_C002 Page Proof page 52 19.1.2010 11:34am Compositor Name: PG1421

52 Computational Electronics

real*8 RD(3,3)
real*8 A1(3,3),A2(3,3),A3(3,3)
real*8 Inv_m1(3,3)
C Define matrices R1, R2 and R3
R1(1,1) ¼ 1
R1(2,2) ¼ 1
R1(3,3) ¼ 1
R2(1,2) ¼ 1
R2(2,3) ¼ 1
R2(3,1) ¼ 1
R3(1,3) ¼ 1
R3(2,1) ¼ 1
R3(3,2) ¼ 1
rml ¼ 0.916
rmt ¼ 0.19
C Read input parameters
print*, ‘define transport direction’
print*, ‘define width direction’
print*, ‘define wafer orientation’
C Construct matrix RD
rd(1,1) ¼ xt=sqrt(xt*xtþyt*ytþzt*zt)
rd(2,1) ¼ yt=sqrt(xt*xtþyt*ytþzt*zt)
rd(3,1) ¼ zt=sqrt(xt*xtþyt*ytþzt*zt)
rd(1,2) ¼ xw=sqrt(xw*xwþyw*ywþzw*zw)
rd(2,2) ¼ yw=sqrt(xw*xwþyw*ywþzw*zw)
rd(3,2) ¼ zw=sqrt(xw*xwþyw*ywþzw*zw)
rd(1,3) ¼ xo=sqrt(xo*xoþyo*yoþzo*zo)
rd(2,3) ¼ yo=sqrt(xo*xoþyo*yoþzo*zo)
rd(3,3) ¼ zo=sqrt(xo*xoþyo*yoþzo*zo)
do i ¼ 1,3
print*,(rd(i,j),j ¼ 1,3)
C equivalent valley pair 1 masses: mx, my, mz
do i ¼ 1,3
do j ¼ 1,3
sum ¼ 0
do k ¼ 1,3
sum ¼ sum þ R1(i,k)*rd(k,j)
A1(i,j) ¼ sum
Vasileska/Computational Electronics 64835_C002 Page Proof page 53 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 53

do i ¼ 1,3
do j ¼ 1,3
term1 ¼ a1(1,i)*a1(1,j)
term2 ¼ a1(2,i)*a1(2,j) þ a1(3,i)*a1(3,j)
inv_m1(i,j) ¼ term1=rml þ term2=rmt

conf_mass1 ¼ 1=inv_m1(3,3)
rm1p_inv ¼ inv_m1(1,1)-inv_m1(3,1)*inv_m1(3,1)=inv_m1(3,3)
transport_mass1 ¼ 1=rm1p_inv
rm2p_inv ¼ inv_m1(2,2)-inv_m1(2,3)*inv_m1(2,3)=inv_m1(3,3)
rm21p_inv ¼ inv_m1(1,2)-inv_m1(2,3)*inv_m1(3,1)=inv_m1(3,3)
rm2dp_inv ¼ rm2p_inv - rm21p_inv*rm21p_inv=rm1p_inv
width_mass1 ¼ 1=rm2dp_inv

print*,‘ ’
print*, ‘valley pair 1’
print*,transport_mass1, width_mass1, conf_mass1

C equivalent valley pair 2 masses: mx, my, mz

do i ¼ 1,3
do j ¼ 1,3
sum ¼ 0
do k ¼ 1,3
sum ¼ sum þ R2(i,k)*rd(k,j)
A2(i,j) ¼ sum

do i ¼ 1,3
do j ¼ 1,3
term1 ¼ a2(1,i)*a2(1,j)
term2 ¼ a2(2,i)*a2(2,j) þ a2(3,i)*a2(3,j)
inv_m1(i,j) ¼ term1=rml þ term2=rmt

conf_mass2 ¼ 1=inv_m1(3,3)
rm1p_inv ¼ inv_m1(1,1)-inv_m1(3,1)*inv_m1(3,1)=inv_m1(3,3)
transport_mass2 ¼ 1=rm1p_inv
rm2p_inv ¼ inv_m1(2,2)-inv_m1(2,3)*inv_m1(2,3)=inv_m1(3,3)
rm21p_inv ¼ inv_m1(1,2)-inv_m1(2,3)*inv_m1(3,1)=inv_m1(3,3)
rm2dp_inv ¼ rm2p_inv - rm21p_inv*rm21p_inv=rm1p_inv
width_mass2 ¼ 1=rm2dp_inv

print*,‘ ’
print*, ‘valley pair 2’
print*,transport_mass2, width_mass2, conf_mass2
Vasileska/Computational Electronics 64835_C002 Page Proof page 54 19.1.2010 11:34am Compositor Name: PG1421

54 Computational Electronics

C equivalent valley pair 3 masses: mx, my, mz

do i ¼ 1,3
do j ¼ 1,3
sum ¼ 0
do k ¼ 1,3
sum ¼ sum þ R3(i,k)*rd(k,j)
A3(i,j) ¼ sum
do i ¼ 1,3
do j ¼ 1,3
term1 ¼ a3(1,i)*a3(1,j)
term2 ¼ a3(2,i)*a3(2,j) þ a3(3,i)*a3(3,j)
inv_m1(i,j) ¼ term1=rml þ term2=rmt
conf_mass3 ¼ 1=inv_m1(3,3)
rm1p_inv ¼ inv_m1(1,1)-inv_m1(3,1)*inv_m1(3,1)=inv_m1(3,3)
transport_mass3 ¼ 1=rm1p_inv
rm2p_inv ¼ inv_m1(2,2)-inv_m1(2,3)*inv_m1(2,3)=inv_m1(3,3)
rm21p_inv ¼ inv_m1(1,2)-inv_m1(2,3)*inv_m1(3,1)=inv_m1(3,3)
rm2dp_inv ¼ rm2p_inv - rm21p_inv*rm21p_inv=rm1p_inv
width_mass3 ¼ 1=rm2dp_inv
print*,’ ‘
print*, ‘valley pair 3’
print*,transport_mass3, width_mass3, conf_mass3

2.6 Density of States

To calculate various optical properties, such as the rate of absorption or emission and how
electrons and holes distribute themselves within a solid, we need to know the number of
available states per unit volume per unit energy. We first calculate the available states in
k-space and then use the energy-momentum relation in parabolic bands to give the density
of states in terms of energy.
By considering the electrons in a solid as a free electron gas, that is, the electrons are free
to wander around the crystal without being influenced by the potential of the atomic
nuclei, we can obtain a relationship for the number of available states in a solid. A free
electron has a velocity v and a momentum p ¼ mv. Its energy consists entirely of kinetic
energy (V ¼ 0), therefore,

1 jpj2
E ¼ mv2 ¼ : (2:28)
2 2m*
Vasileska/Computational Electronics 64835_C002 Page Proof page 55 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 55

Louis De-Broglie first hypothesized the idea of particle-wave duality, which associates an
electron wave with a wave number k. For free electrons, k ¼ p=h  . In this way, the electron
can be represented by a vector in velocity v, momentum p, or k-space. If we choose to
represent the electron state as a vector, it points in a direction given by the components
magnitude of the basis vectors in k-space. It should be apparent that vectors of the same
magnitude have the same energy forming spherical shells. This can be understood better if
we consider the equation for the energy of the electron in terms of k.

h 2  2  h 2 jkj2
E¼ kx þ ky2 þ kz2 ¼ : (2:29)
2m* 2m*

Classically, all values of energy would be allowed and there would be no restriction on the
number of electrons with the same value of k. However, at atomic scales, the effects of
quantum mechanics dominate and two further important principles come into play. These
are the Heisenberg uncertainty principle and the Pauli exclusion principle. Together, these
two rules mean that the wavefunction for the electron must satisfy the Schrödinger
equation, subject to boundary conditions. The solution of the Schrödinger equation leads
to wavefunctions of the form

c(r) ¼ C exp(ik  r): (2:30)

Application of the periodic boundary condition to Equation 2.30 leads to the followed set
of allowed k-values

2pnx 2pny 2pnz

kx ¼ , ky ¼ , kz ¼ , nx , ny , nz ! integers: (2:31)

The result given in Equation 2.31 suggests that the volume in k-space associated with one
state (if we do not consider the spin of the electron) is V3D ¼ (2p=L)3 . Let us now consider
the spherical shell shown in Figure 2.23. The number of states that can be accommodated in
the shell is

4pk2 dk V
M(k) ¼ 3
¼ 2 k2 dk ¼ g0 (E)dE, (2:32)
(2p=L) 2p

where g0 (E) represents the number of states per unit energy function. For a parabolic band,
the density of states g(E) per unit energy is then given by (taking spin degeneracy into
1 2m* 3=2 1=2
g3D (E) ¼ E dE: (2:33)
2p2 h 2

In two-dimensional (2D) structures, such as a quantum well, the procedure is much the
same but this time one of the k-space components is fixed. Instead of finding the number of
k-states enclosed within a sphere, the problem is to calculate the number of k-states lying in
an annulus of radius k to k þ d k. The k-space (see Figure 2.24) would be completely filled if
each state occupied an area of V2D ¼ (2p=L)2 . Using an identical procedure as the one used
in the case of the 3D density of states function calculation, for 2D systems one gets
Vasileska/Computational Electronics 64835_C002 Page Proof page 56 19.1.2010 11:34am Compositor Name: PG1421

56 Computational Electronics

(k̂3) Device
Y Source Top gate Drain
Bottom gate
(k̂΄3) (001)

(100) (111)

Crystal ( )

k 2
k 1


Visualization of k-space showing values of k as points.
The number of allowed states is the number of these
points contained in the shell of radius k and thickness dk.



k-Space in 2D. The density of states at an energy E is the number
of k-states per unit volume contained with the annulus of radius
k and thickness dk.
Vasileska/Computational Electronics 64835_C002 Page Proof page 57 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 57

g2D (E) ¼ : (2:34)
The one-dimensional (1D) density of states function is calculated using
1 m* 1=2 1
g1D (E) ¼ : (2:35)
p h 2 E1=2

Table 2.4 summarizes the ratio of allowed energies to ground state energy and degeneracy
of the energy level for 2D, 1D, and 0D structures, while the graph shows the density of
states functions plotted against energy. The DOS functions for 3D, 2D, 1D, and 0D cases are
illustrated in Figure 2.25.

Degeneracy of Energy States for Quantum Well (2D), Quantum Wire (1D), and Quantum Box (0D)
2D 1D 0D
State E=E0 n(E) E=E0 Degenerate States n(E) E=E0 Degenerate States n(E)
1 1 1 2 (1,1) 1 3 (1,1,1) 1
2 4 1 5 (2,1),(1,2) 2 6 (2,1,1),(1,2,1),(1,1,2) 3
3 9 1 8 (2,2) 1 9 (2,2,1),(1,2,2),(2,1,2) 3
4 16 1 10 (3,1),(1,3) 2 11 (3,1,1),(1,3,1),(1,1,3) 3
5 25 1 13 (3,2),(2,3) 2 12 (2,2,2) 1
6 36 1 17 (4,1),(1,4) 2 14 (3,2,1),(3,1,2),(2,3,1) (2,1,3),(1,3,2),(1,2,3) 6
7 49 1 18 (3,3) 1 17 (3,2,2),(2,3,2),(2,2,3) 3
8 64 1 20 (4,2),(2,4) 2 18 (4,1,1),(1,4,1),(1,1,4) 3
9 81 1 25 (4,3),(3,4) 2 21 (4,2,1),(4,1,2),(1,4,2) (1,2,4),(2,4,1),(2,1,4) 6
10 100 1 26 (5,1),(1,5) 2 27 (3,3,3) 1

Density of states

Bulk (3D)
Quantum well (2D)
Quantum wire (1D)
1E+20 Quantum dot (0D)
0 10 20 30 40 50 60 70 80 90 100 110 120
Energy (meV)

Density of states for bulk (3D blue), quantum well (2D red), quantum wire (1D green), and quantum Dot AQ8
(0D black).
Vasileska/Computational Electronics 64835_C002 Page Proof page 58 19.1.2010 11:34am Compositor Name: PG1421

58 Computational Electronics

2.7 Electron Mobility

Electron mobility (or simply, the mobility) is a quantity relating the drift velocity of
electrons to the applied electric field across a material, according to the formula vd ¼ mE,
where vd is the drift velocity, E is the applied electric field, and m is the mobility. In
semiconductors, mobility can also apply to holes as well as electrons. In a solid, electrons
(and in the case of semiconductors, holes) will move around randomly in the absence of an
applied electric field. Therefore, if one averages the movement over time, there will be no
overall motion of charge carriers in any particular direction. However, upon applying an
electric field, electrons will be accelerated in an opposite direction to the electric field. The
summation of the time between the acceleration of electrons due to the electric field and the
deceleration of electrons due to collisions and lattice scattering events (caused by phonons,
crystal defects, impurities, etc.) over the mean free path between scattering events results in
the electrons having an average drift velocity (Chapter 3). This net electron motion must be
orders of magnitude less than the normally occurring random motion, otherwise the
mobility equation is not valid (i.e., typical drift speeds in copper being of the order of
104 m=s compared with the speed of random electron motion of 105 m=s). In a semicon-
ductor, the two charge carriers, electrons, and holes will typically have different drift
velocities for the same electric field. Since mobility is usually a strong function of material
impurities and temperature, and is determined empirically, mobility values are typically
presented in a table or chart form. The mobility of electrons and holes in silicon at room
temperature is shown in Figure 2.26.
The electron and hole mobilities have a similar doping dependence: for low doping
concentrations, the mobility is almost constant and is primarily limited by phonon scatter-
ing. At higher doping concentrations, the mobility decreases due to ionized impurity
scattering with the ionized doping atoms. The actual mobility also depends on the type of
dopant. The above figure is for phosphorous and boron doped silicon and is calculated using

(1414  68:5)
mn (N) ¼ 68:5 þ cm2 =V s, (2:36a)
1 þ (N=9:2  1016 )0:711


(470:5  44:9)
mp (N) ¼ 44:9 þ cm2 =V s: (2:36b)
1 þ (N=2:23  1017 )0:719

Mobility (cm2/V s)

FIGURE 2.26 1.0E+14 1.0E+16 1.0E+18 1.0E+20
Electron and hole mobility versus doping density for
Doping concentration (cm–3)
Vasileska/Computational Electronics 64835_C002 Page Proof page 59 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 59

These are empirical relations obtained by fitting experimental values. Empirical relations
for the temperature as well as doping dependence of the carrier mobilities in silicon are
available as well and are listed below:

T 7:4  108 T2:33
mn (N, T) ¼ 88 þ cm2 =V s,
300 1 þ (N=(1:26  1017 (T=300)2:4 ))0:88(T=300)0:146
T 1:35  108 T 2:33
mp (N, T) ¼ 54:3 þ cm2 =V s:
300 1 þ (N=(2:35  1017 (T=300)2:4 ))0:88(T=300)0:146

The surface and interface mobility of carriers is affected by the nature of the adjacent layer
or surface. Even if the carrier does not transfer into the adjacent region, its wavefunction
does extend over 1–10 nm so that there is a nonzero probability for the particle to be in the
adjacent region. The net mobility is then a combination of the mobility in both layers.
Carriers in the inversion layer of a MOSFET have an up to three times lower mobility, since
the mobility in the amorphous silicon dioxide is much lower than that in the silicon. The
presence of charged surface states further reduces the mobility just like ionized impurities
would. For n-channel or p-channel MOSFETs, the electron or hole mobility at the silicon
dioxide=silicon interface has a very strong effect on the speed of the device. More details on
low field transport and methods for mobility calculation in different types of semiconduct-
ors are given in Chapter 3 of this book.

2.8 Semiconductor Statistics

The conduction band electron and the valence band hole concentrations in a semicon-
ductor are given by averaging over the density of states discussed in Section 2.6 and the
Fermi Dirac distribution governing the occupancy of electrons and holes at equilibrium


n¼ gc (E)f (E)dE, (2:38a)



p¼ gv (E)(1  f (E))dE: (2:38b)


For nondegenerate semiconductors (where the Fermi energy is more than 3kBT away from
both the valence and conduction band), the equations for n and p simplify to

n ¼ NC e(EF EC )=KT , p ¼ NV e(EV EF )=KT , (2:39)

Vasileska/Computational Electronics 64835_C002 Page Proof page 60 19.1.2010 11:34am Compositor Name: PG1421

60 Computational Electronics

where NC and NV are the effective density of states in the conduction and valence bands,
respectively, and are

 3=2  3=2
2pmn* kT 2pmp* kT
NC ¼ 2 , Nv ¼ 2 : (2:40)
h2 h2

When the semiconductor is undoped, n ¼ p ¼ ni and EF ¼ Ei, which leads to the following
equations for n and p:

n ¼ ni e(EF Ei )=KT , (2:41a)

p ¼ ni e(Ei EF )=KT : (2:41b)

If the semiconductor is not doped, then the concentration of electrons in the conduction
band n is equal to the concentration of holes in the valence band p, hence n ¼ p ¼ ni where ni
is defined as the intrinsic carrier concentration. Using the equations above, it is easy to
show that the product np is given by

np ¼ n2i ¼ NC Nv e(Ev EC )=KT ¼ NC Nv eEgap =KT , (2:42a)

where Egap ¼ EC  EV. Thus, we have an equation for the intrinsic carrier concentration
ni ¼ NC NV eEgap =KT : (2:42b)

Besides the np product, we can equate n and p to obtain the intrinsic Fermi energy level

NC e(EF EC )=KT ¼ NV e(EV EF )=KT ,

EC þ EV kT NC EC þ EV 3kT m*p
Ei ¼ þ ln , Ei ¼ þ ln : (2:43)
2 2 NV 2 4 mn*

The Fermi energy level is a function of the carrier concentration, thus the carrier concen-
tration of the semiconductor may be analyzed as a function of temperature. The ratio of
ionized donor atoms to total donors and the ratio of ionized acceptor atoms to total
acceptor atoms is given by

NDþ 1
¼ gD ¼ 2, (2:44a)
ND 1 þ gD e(EF ED )=KT
NAþ 1
¼ gA ¼ 4, (2:44b)
NA 1 þ gA e(EA EF )=KT

ED is the donor energy level
EA is the acceptor energy level
gD and gA are the degeneracy factors
Vasileska/Computational Electronics 64835_C002 Page Proof page 61 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 61

2.9 Semiconductor Devices

Semiconductor devices are electronic components that exploit the electronic properties of
semiconductor materials, principally silicon, germanium, and gallium arsenide. Semicon-
ductor devices have replaced thermionic devices (vacuum tubes) in almost all applications.
Semiconductor devices are manufactured both as single discrete devices and as integrated
circuits, which consist of a number—from a few to billions—of devices manufactured and
interconnected on a single semiconductor substrate. A list of common semiconductor
devices is given below. They can be separated into the following:

Two-terminal devices:

. Avalanche diode (avalanche breakdown diode)

. Diode (rectifier diode)
. Gunn diode
. IMPATT diode
. Laser diode
. Light-emitting diode (LED)
. Photocell
. PIN diode
. Schottky diode
. Solar cell
. Tunnel diode
. Zener diode

Three-terminal devices:

. Bipolar transistor (BJT)

. Darlington transistor
. Field effect transistor (FET)
. Insulated gate bipolar transistor (IGBT)
. Silicon controlled rectifier (SCR)
. Thyristor
. Triac
. Unijunction transistor

Four-terminal devices:

. Hall effect sensor (magnetic field sensor)

Vasileska/Computational Electronics 64835_C002 Page Proof page 62 19.1.2010 11:34am Compositor Name: PG1421

62 Computational Electronics

Multi-terminal devices:

. Charge-coupled device (CCD)

. Microprocessor
. Random access memory (RAM)
. Read-only memory (ROM)

All transistor types can be used as the building blocks of logic gates, which are fundamental
in the design of digital circuits. In digital circuits like microprocessors, transistors act as on-
off switches; in the MOSFET, for instance, the voltage applied to the gate determines
whether the switch is on or off. Transistors used for analog circuits do not act as on-off
switches; rather, they respond to a continuous range of inputs with a continuous range of
outputs. Common analog circuits include amplifiers and oscillators. Circuits that interface
or translate between digital circuits and analog circuits are known as mixed-signal circuits.
Power semiconductor devices are discrete devices or integrated circuits intended for
high current or high voltage applications. Power integrated circuits combine integrated
circuit technology with power semiconductor technology, these are sometimes referred
to as ‘‘smart’’ power devices. Several companies specialize in manufacturing power semi-

2.9.1 Diode
The pn junction diode is a two-terminal device made from a pn-junction. At the junction of a p-
type and an n-type semiconductor, a space charge region forms due to the interdiffusion of
positive and negative charges, called the depletion region, which blocks current conduction
from the n-type region to the p-type region, but allows current to conduct from the p-type
region to the n-type region. Thus, when the device is forward biased, with the p-side at higher
electric potential, the diode conducts current easily; when reversed biased, the current is on the
other hand. Exposing a semiconductor to light can generate electron–hole pairs, which
increases the number of free carriers and its conductivity. Diodes optimized to take advan-
tage of this phenomenon are known as photodiodes. Compound semiconductor diodes can
also be used to generate light, as in light-emitting diodes and laser diodes.
The conduction and valence band, total charge density, the potential profile, and the
electric field profiles in a pn-diode under equilibrium conditions are shown in Figure 2.27.
In Figure 2.27, Vbi is the built-in voltage, W is the width of the depletion region, and Emax
is the maximum electric field value. Within the depletion charge approximation, these
quantities are given by

qVbi ¼ (Ei  EF )p þ (EF  Ei )n ,

EF  Ei
nn0 ¼ ni exp ,
kB T
Ei  EF (2:45a)
pp0 ¼ ni exp ,
kB T
kB T pp0 nn0 NA ND
Vbi ¼ ln V T ln ,
q n2i n2i
Vasileska/Computational Electronics 64835_C002 Page Proof page 63 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 63

p-side n-side



−xp xn
x FIGURE 2.27
Emax Equilibrium conduction band, charge density, potential, and electric
field profile variation in a diode.

9 sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
xn þ xp ¼ W =
2ks e0 (NA þ ND )Vbi
Vp (0) ¼ Vn (0) ! W ¼ , (2:45b)
; qNA ND
NA xp ¼ ND xn

Emax ¼  ¼  , (2:45c)
dx x¼0 ks e0 (NA þ ND )

NA and ND are the doping at the p- and n-side of the junction
pp0 and nn0 are the equilibrium majority hole and electron concentrations in the quasi-
neutral regions

Under applied bias, the quasi-Fermi levels for the electrons and holes on the n-side and the
p-side, respectively, split and we have the situation depicted in Figure 2.28.



−xp xn FIGURE 2.28

Energy-band diagram of a diode under forward applied bias.
Vasileska/Computational Electronics 64835_C002 Page Proof page 64 19.1.2010 11:34am Compositor Name: PG1421

64 Computational Electronics

Then, inside and at the edges of the depletion region

np ¼ n2i exp ,
np (xp ) ¼ np0 exp , (2:46)
pn (xn ) ¼ pn0 exp ,

npo is the equilibrium minority electron concentration in the p-region
pno is the equilibrium minority hole concentration in the n-region

The excess minority carrier concentration on both the p-side and the n-side of the junction
(in the quasi-neutral regions) is next found as a solution of the minority carrier diffusion
equation, which gives

Dpn (x) ¼ pn0 (eV=VT  1)e(xxn )=Lp ,

Dnp (x) ¼ np0 (eV=VT  1)e(xþxp )=Ln :

The excess variation of the minority carrier concentration in the quasi-neutral regions is
shown in Figure 2.29. The corresponding diffusion components of the currents of the
minority carriers are related to the gradient of this variation

qDp pn0 V=VT

Jpdiff (x) ¼ (e  1)e(xxn )=Lp ,
qDn np0 V=VT
Jndiff (x) ¼ (e  1)e(xþxp )=Ln :

The various components of the current in the diode are schematically shown in Figure 2.30.
Under the assumption that there is no recombination in the space-charge region, the total
diode current equals the sum of the minority carrier components of the diffusion currents.
This assumption leads us to the following final expressions for the current:

np (x) pn (x)

Forward bias
region W


−xp xn
Reverse bias

Spatial variation of the minority carrier density in the quasi-neutral regions of a diode.
Vasileska/Computational Electronics 64835_C002 Page Proof page 65 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 65

Jtot = Jpdiff (xn) +Jndiff (−xp)

Majority Jpdiff + Jpdrift Majority Jndiff + Jndrift


Minority Jpdiff
Minority Jndiff

−xp xn

No SCR generation/recombination

Spatial variation of drift and diffusion components of the current due to both minority and majority carriers.

Dp pn0 Dn np0 V=VT
Itot ¼ Ipdiff (xn ) þ Indiff (xp ) ¼ qA þ (e  1), (2:49)
Lp Ln

where the reverse saturation current is given by

Dp pn0 Dn np0 Dp Dn
Is ¼ qA þ ¼ qAn2i þ : (2:50)
Lp Ln Lp ND Ln NA

The current–voltage (I–V) characteristics of Ge, Si, and GaAs diodes are schematically
shown in Figure 2.31. A physical explanation for the origins of the forward and reverse
component of the current is shown in Figure 2.32.

2.9.2 BJT Transistor

BJTs are formed from two pn-junctions in either a n-p-n or p-n-p configuration. The middle,
or base, region between the junctions is typically very narrow. The other regions, and their
associated terminals, are known as the emitter and the collector. A small current injected
through the junction between the base and the emitter changes the properties of the base-
collector junction so that it can conduct current even though it is reverse biased. This
creates a much larger current between the collector and emitter, controlled by the base-
emitter current. A schematic description and the configurations of a pnp and npn transistor
are shown in Figure 2.33.

Ge Si GaAs

Schematic of I–V characteristics of a Ge, Si, and GaAs diode.
Vasileska/Computational Electronics 64835_C002 Page Proof page 66 19.1.2010 11:34am Compositor Name: PG1421

66 Computational Electronics

Forward bias: Reverse bias:


EC Ln qV
q(Vbi −V ) q(Vbi + |V |)

qV EFp


Reverse saturation current is due to

minority carriers being collected over a
distance on the order of the diffusion length.

Description of the origin of the various current components under forward (left panel) and reverse (right panel)
bias conditions.

PNP - transistor NPN - transistor



E p+ n C E n+ p n C
+ +
Symbols (top panel) and configuration + +
(bottom panel) of pnp (left) and npn
(right) transistors.

Forward active region Saturation region
(emitter-base FB, collector-base RB) (both junctions forward biased)
Cutoff region Inverted active region
Four modes (regimes) of operation of
(both junctions reverse biased) (emitter-base RB, collector-base FB)
a pnp BJT.

Since this is a three-terminal device, there are four regions of operation of a BJT transis-
tor. The example shown in Figure 2.34 is for a pnp BJT.
Since it has three leads, there are three possible amplifier types that are schematically
shown in Figure 2.35.
The qualitative description of the transistor operation along with some notation that will
be followed in this section are shown in Figure 2.36.
Vasileska/Computational Electronics 64835_C002 Page Proof page 67 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 67

p p+
E C nn VEC n VEC
p+ n p
p p
(a) Common-base (b) Common-emitter (c) Common-collector

Amplifier types.

p+ n p • Emitter doping is much larger than

IEp { base doping
• Base doping larger than collector
• Current components:
IE = IEp + IEn
IB2 ICn IC = ICp + ICn
IEn IB = IE − IC = IB1 + IB2 − IB3
• IB1 = current from electrons being
back injected into the forward-biased
emitter-base junction
• IB2 = current due to electrons that
replace the recombined electrons in
the base
• IB3 = collector current due to
ICp thermally-generated electrons in the
collector that go in the base

Definition of the emitter, base and collector current components and their physical origin.

Regarding circuit definitions, the base transport factor is defined as

aT ¼ ICp =IEp : (2:51)

If there were no recombinations in the base, the base transport factor is equal to unity. On
the other hand, the emitter injection efficiency is calculated according to

g¼ ¼ , (2:52)
ICp þ IEp IE

and it approaches unity when the emitter doping is much higher than the base doping. The
current amplification under DC operating conditions for a common base configuration is
then defined as

IC ICp þ ICn ICp

adc ¼ ¼ ¼ aT g: (2:53)
IE IEp þ IEn IEp þ IEn
Vasileska/Computational Electronics 64835_C002 Page Proof page 68 19.1.2010 11:34am Compositor Name: PG1421

68 Computational Electronics

The corresponding current amplification for the current emitter configuration is

IC IC adc
bdc ¼ ¼ ¼ : (2:54)
IB IE  IC 1  adc

Another set of parameters that are important for the description of the operation of BJT are
the reverse saturation currents in the common emitter and common base configurations,
which are given by the following expressions

IBC0 ¼ ICn ! IC ¼ ICp þ ICn ¼ adc IE þ IBC0 , (2:55a)

adc IBC0
IC ¼ adc (IC þ IB ) þ IBC0 ! IC ¼ IB þ ,
1  adc 1  adc
! IC ¼ bdc IB þ IEC0 ,
! IEC0 ¼ (1 þ bdc )IBC0 : (2:55b)

From the above expressions, it is evident that a small base current IB forces the E-B junction
to be forward biased, and injects a large number of holes that traverse the base to the
Following the same procedure as for pn-diodes, we solve for the minority carrier
concentration in the different regions, which gives the spatial variations of the minority
carrier diffusion currents. This procedure is illustrated in Figure 2.37.

pB(0) Saturation
Forward pB(x) nC(x΄)
nE(x˝) nC0
active p
x˝ x΄
0˝ 0 Cut-off W 0΄

IE = InE(0˝) + IpB(0) IC = InC(0΄) + IpB(W )

InE(0˝) IpB(x)
InE(x˝) InC(0΄) InC(x΄)
IB2 = IpB(0) – IpB(W )
x˝ x΄
0˝ 0 W 0΄
Base recombination current

Variation of the minority carrier concentration in the quasi-neutral regions of a pnp-BJT (top panel) and the
corresponding diffusion currents (bottom panel).
Vasileska/Computational Electronics 64835_C002 Page Proof page 69 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 69

Using the diagrams shown in Figure 2.37, one obtains the general expressions for the
emitter, base, and collector currents under arbitrary bias conditions of the following form:
IE ¼ Aqn2i þ coth (eVEB =VT  1)  Aqn2i (eVCB =VT  1),

IC ¼ Aqni (e  1)  Aqni 2
þ coth(W=LB ) (eVCB =VT  1),
LB NB sinh (W=LB ) LC NC LB NB
IB ¼ Aqni
þ coth  (eVEB =VT  1)
LE NE LB NB LB sinh(W=LB )
þ Aqni2
þ coth  (eVCB =VT  1): (2:56)
LC NC LB NB LB sinh(W=LB )

The graphical representation of the various current components in the so-called forward
active mode is shown in Figure 2.38.
For the forward-active mode, when the collector base junction is reverse biased and the
emitter-base junction is forward biased, Equation 2.56 simplifies to
IE Aqn2i þ coth eVEB =VT ¼ IEn þ IEp ,
DB 1
IC Aqn2i eVEB =VT ¼ ICp ,
LB NB sinh(W=LB )
DE DB cosh(W=LB )  1
IB Aqni2
þ eVEB =VT
LE NE LB NB sinh(W=LB )
DC DB cosh(W=LB )  1
 Aqni 2
þ :
LC NC LB NB sinh(W=LB )

Under these bias conditions, the emitter injection efficiency reduces to


g¼ ¼ ! : (2:58)
IEp þ IEn 1 þ (LE NE DB =LB NB DE ) coth(W=LB ) short 1 þ (LE NE DB =WNB DE )

p+ n p

IEp { } ICp

IEn { ICn

Recombination in the

Emitter, base, and collector current components in the forward active mode of operation.
Vasileska/Computational Electronics 64835_C002 Page Proof page 70 19.1.2010 11:34am Compositor Name: PG1421

70 Computational Electronics

The base transport factor is calculated using

ICp 1 W2
aT ¼ ¼ ! 1  2 , (2:59)
IEp cosh (W=LB ) short 2LB

and the common emitter current gain is

(LE NE DB =LB NB DE ) coth(W=LB ) L E NE DB

bdc ¼ ! : (2:60)
1 þ 2(LE NE DB =LB NB DE ) coth(W=LB ) sinh2 (W=2LB ) short

The graphical representation of the various current components for the case of saturation
and cut-off modes of operation are shown in Figure 2.39.
The form of the input and the output characteristics for the common base and common
emitter situation is shown in Figure 2.40.

2.9.3 MOSFET
Another type of transistor, the field effect transistor, operates on the principle that semi-
conductor conductivity can be increased or decreased by an external electric field. The field
may be applied by a reverse-biased p-n junction, forming a junction field effect transistor
(JFET), by a reversed biased metal-semiconductor Schottky contact to form a MESFET, or
by an electrode isolated from the bulk material by an oxide layer, forming a MOSFET.
The MOSFET is the most common field effect transistor in use today. The gate electrode is
biased to produce an electric field that controls the conductivity of a ‘‘channel’’ between

p+ n p

{ } ICp
{ } ICp΄

{ ICn

Recombination in the
p+ n p


IB1 IB3 Recombination in the base

IB is ignored in this diagram

Emitter, base and collector current components in saturation (top panel) and cut-off (bottom panel) mode of
Vasileska/Computational Electronics 64835_C002 Page Proof page 71 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 71

Common-base configuration:
Forward active
VCB < –3VT Saturation

IE ≠ 0
VCB = 0
IE = 0

Common-emitter configuration:
IC VCB = 0 Forward active
VEC = 0
IB ≠ 0


IB = 0

Input (left) and output (right) characteristics of a pnp transistor in the common-base (top panel) and common
emitter (bottom panel) configurations.

two terminals, called the source and drain. Depending on the type of carrier in the channel,
the device may be an n-channel (for electrons) or a p-channel (for holes) MOSFET. Although
the MOSFET is named in part for its historically ‘‘metal’’ gate, in modern MOSFETS,
heavily doped polysilicon is typically used instead. The configuration, symbol, and trans-
fer characteristics for n-channel and p-channel enhancement (normally off) or depletion
mode (normally on) devices are shown in Figure 2.41.
To better understand the operation of a MOSFET device, it is important to understand
the role of the gate and the drain electrode separately on the energy band diagram of a
MOSFET. We begin with a MOSFET in equilibrium to first understand the influence of the
gate electrode for the case of an n-channel MOSFET. A positive bias applied to the gate
electrode results in the following:

1. It reduces the potential energy barrier seen by the electrons flowing from the
source contact to the drain region.
2. It inverts the surface, i.e., changes the carrier type at the surface from holes to
electrons, and increases the conductivity of the channel.

This is schematically shown in Figure 2.42. Note that we have a flux of carriers: one flux of
carriers goes from the source to the drain and is balanced by the flux of carriers that go from
the drain to the source, so the net current in equilibrium, as it should be expected, is zero.
The effect of a positive bias applied to the drain electrode on n-channel MOSFET
operation for the two gate bias conditions from Figure 2.42 is shown in Figure 2.43. The
drain bias pulls the energy bands down on the drain side, resulting in a lowering of the
barrier seen by electrons in the source, which is controlled by the gate potential.
As can be seen from the schematic band diagrams shown in Figure 2.43, the balance of
the carrier fluxes from source-drain and drain-source is disturbed with the application of a
Vasileska/Computational Electronics 64835_C002 Page Proof page 72 19.1.2010 11:34am Compositor Name: PG1421

72 Computational Electronics

n+ n+
p-Type SC
(a) n-channel, enhancement mode device

n+ n+
p-Type SC
(b) n-channel, depletion mode device

Symbol VG
p+ p+
n-Type SC

(c) p-channel, enhancement mode device

Symbol VG
p+ p+
n-Type SC

(d) p-channel, depletion mode device

Cross-sectional diagrams, circuit symbols, and transfer characteristics of the four basic MOSFET configurations.

VG = 0 Source Drain

VG > VT Source Drain

Equilibrium band diagram from source to drain
with no gate bias (upper), and with a positive
bias applied for an n-channel MOSFET.

drain bias, allowing a net electron flux from source to drain. For VG ¼ 0, only a small
portion of carriers can overcome the source barrier which, in turn, gives rise to a small
drain current, basically corresponding to the subthreshold mode of operation of the
MOSFET (top panel in Figure 2.43). When VG > VT, the barrier in the source region is
significantly reduced, allowing many more carriers to flow from the source to the drain
and the MOSFET is said to be operating in an on-state (linear or saturation, as we will see
later in the text).
Another way of qualitatively describing the MOSFET operation is shown schematically
in Figure 2.44. For VG > VT, the channel forms and for small applied drain bias, the channel
Vasileska/Computational Electronics 64835_C002 Page Proof page 73 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 73

VG = 0, VD > 0



Large potential barrier allows only few EC

electrons to go from the source to the drain Drain
(subthreshold conduction)

VG > VT, VD > 0



Smaller potential barrier allows a large EC

number of electrons to go from the source Drain
to the drain

The role of the drain electrode.

(a) VG >VT, VD > 0 (small) S D

Variation of electron density n+ n+

along the channel is small:
ID VD p-Type SC

(b) VG >VT, VD > 0 (larger) S D

n+ n+
Increase in the drain current
reduces due to the reduced
conductivity of the channel p-Type SC
at the drain end.

(c) VG >VT, VD =VG –VT S D
n n+
Pinch-off point. Electron
density at the drain-end of
p-Type SC
the channel is identically zero.

(d) VG >VT, VD >VG –VT G


Post pinch-off characteristic. n+ ΔL n


The excess drain voltage is FIGURE 2.44

dropped across the highly p-Type SC Channel and depletion region formation for
resistivepinch-off region different conditions ranging from subthres-
denoted by ΔL.
hold to pinch-off.
Vasileska/Computational Electronics 64835_C002 Page Proof page 74 19.1.2010 11:34am Compositor Name: PG1421

74 Computational Electronics

is relatively uniform (top panel in Figure 2.44). If we increase the drain bias (second panel
of Figure 2.44), two things occur: (1) the drain current increases due to the increase of the
lateral electric field that accelerates the carriers from the source to the drain region and (2)
the channel is nonuniform, i.e., the larger drain bias leads to smaller gate to channel voltage
at the drain side of the channel, which in turn reduces the concentration of inversion layer
electrons there. The current remains constant along the channel due to current continuity,
hence the velocity of the carriers at the drain end of the channel is much larger when
compared with the velocity of the carriers near the source end of the channel where the
concentration is higher, since the product env is constant (the current is mostly drift current
along the channel in the on-state, the contribution of the diffusion component of the current
is insignificant in this regime). The I–V characteristics, shown in Figure 2.45, correspond to
the various bias conditions given in Figure 2.44.
Using simple capacitor model, the threshold voltage, VT, of a MOSFET may be
derived as

1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
VT ¼ 2wF þ 2qNA ks e0 (2wF ) þ VFB , (2:61)

wF ¼ 1qjEF  Ei j is the bulk potential
NA is the channel doping
Cox is the oxide capacitance

The flat-band voltage is given by

1 Qit Qf Qot Qm
VFB ¼ fMS þ þ þ got þ gm , (2:62)
q Cox Cox Cox Cox

fMS is the metal-semiconductor workfunction
Qit is the interface-trap density
Qf is the fixed oxide charge density
Qox is the oxide charge
Qm is the mobile charge density

In general, the oxide charges are minimized in the device fabrication process.
Analytical MOSFET models are usually based on the so-called gradual channel approxi-
mation (GCA). Contrary to the situation in the ideal two-terminal MOS device, where the


(c) (d)
Linear (b)

FIGURE 2.45 Saturation

(a) region
Current voltage (I–V) characteristics correspond-
ing to the various bias conditions shown in VD
Figure 2.44.
Vasileska/Computational Electronics 64835_C002 Page Proof page 75 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 75

charge density profile is determined from a 1D Poisson’s equation, the potential distribu-
tion in a MOSFET is generally 2D. The reason is that the geometric effects and the
application of a drain-source bias create a lateral electric field component in the channel,
perpendicular to the vertical field associated with the gate capacitor. The GCA states that,
under certain conditions, the electrostatic problem of the gate region may be expressed in
terms of two coupled 1D equations—a Poisson’s equation for determining the vertical
charge density profile under the gate and a charge transport equation for the channel. This
allows us to determine self-consistently both the channel potential and the charge profile at
any position along the gate. A direct inspection of the 2D Poisson’s equation for the
channel region shows that the GCA is valid if we can assume that the electric field gradient
in the lateral direction of the channel is much less than that in the vertical direction
perpendicular to the channel, an assumption that breaks down for short channel, nano-
scale devices. Typically, we find that the GCA is valid for long-channel MOSFETs, where
the ratio between the gate length and the vertical distance of the space charge region from
the gate electrode, the so-called aspect ratio, is large. However, if the MOSFET is biased in
saturation, the GCA is invalid near drain as a result of the large lateral field gradient that
develops in this region.
Based on a simple charge control model, in which the gate charge is completely balanced
by the inversion charge, the following expressions for the drain current in the linear
(VD < VG  VT) and saturation region (VD > VG  VT) may be derived:
Wmeff Cox 1
ID ¼ (VG  VT )VD  VD2 for VD  VG  VT ,
L 2
Wmeff Cox
ID ¼ (VG  VT )2 for VD  VG  VT ,

the so-called square law theory due to the quadratic dependence of the saturated drain
current with the gate voltage. When the gate charge is balanced with the inversion charge AQ9
plus the depletion charge (bulk charge theory), the following results can be derived using a
space-charge control model (SCCM):
(  "   #)
Wmeff Cox 1 2 4 VD 3=2 3VD
ID ¼ (VG  VT )VD  VD  VW wF 1 þ  1þ ,
L 2 3 2wF 4wF
2ks e0 qNA (2wF ) WT qNA
VW ¼ ¼ ,
Cox Cox
8sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9
< V V    
G T VW 2 VW =
VDsat ¼ VG  VT  VW þ 1þ  1þ : (2:64)
: 2wF 4wF 4wF ;

The results of the square law and bulk charge theory are shown schematically in Figure
2.46. It is evident that the deviation between the square law and bulk charge theory is
larger for larger channel doping. The role of the series source and drain resistance is
illustrated in Figure 2.47.
When the channel length is scaled to nanometer dimensions, first velocity saturation and
then velocity overshoot effects start to dominate the device behavior. Velocity saturation
arises because of the fact that at large in-plane fields, the effective carrier mobility is not
Vasileska/Computational Electronics 64835_C002 Page Proof page 76 19.1.2010 11:34am Compositor Name: PG1421

76 Computational Electronics



increasing Bulk charge
Square-law versus bulk-charge theory.

VDS =VD + (RS + Rd )ID

n+ RS RD n+

p-Type SC

Transfer characteristic Output characteristic

ID Rs = Rd = 0
Rs = Rd = 0
Rs ≠ 0
Rd ≠ 0


Effect of Source and drain series resistances on the transfer and output characteristics. (After Pierret.)

constant but decreases with an increase in the electric field. When the device is velocity
saturated, then

ID ¼ qAeff nvd , Aeff ¼ Zyeff

% - (2:65)
Device width Effective thickness of the inversion layer,

i.e., the velocity limited drain current equals to

ID ¼ qZyeff nvd ¼ qyeff nZvd ¼ Zvd Cox (VG  VT ): (2:66)


Comparing the above expression with the mobility-limited one for a long-channel device,
we find that

meff (VG  VT )
vsat $ ) For VG  VT ¼ 5 V we get L ¼ 1:25 mm (vsat ¼ 107 cm=s):
Vasileska/Computational Electronics 64835_C002 Page Proof page 77 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 77

In Figure 2.47, we give an example of the output characteristics of a velocity saturated

device. The main difference is that when velocity saturation becomes important, the gate
voltage dependence of the saturated drain current goes from a square law dependence to a
linear dependence.

2.9.4 SOI Devices PD=FD SOI Devices
Both partially depleted silicon-on-insulator (PD-SOI) devices and fully depleted silicon-on-
insulator (FD-SOI) devices (shown schematically in Figures 2.48 and 2.49) show promise
for high-performance CMOS, microprocessors, and system-on-a-chip designs. In ultra-thin
body (UTB) SOI structures, the control of short-channel effects (SCE) and threshold voltage
(VT) adjustment can be realized with little or no channel doping. In FD-SOI, the thin silicon
body (<40 nm) is depleted of mobile carriers under all bias conditions. Because a depletion
charge does not exist in the channel, carriers in the inversion encounter a smaller average
electric field than in standard bulk silicon devices with heavy channel doping.
FD-SOI devices also provide significantly better subthreshold slopes than bulk silicon
transistors. Work by the INTEL corporation has shown that in comparison with bulk
PMOS, FD-SOI devices with raised source-drain contacts achieved better switching per-
formance, comparable channel mobility, and consumed 30% less power. A schematic
description of the inversion charge in conventional MOSFETs, dual-gate (DG) devices,
and FD-SOI devices is shown in Figure 2.48, while the structures of both n-channel and
p-channel SOI devices are shown in Figure 2.49. As shown in Figure 2.48, due to the
broadening of the inversion charge distribution away from the surface in DG and UTB
structures, and the reduced surface field at the oxide-semiconductor interface, gate leakage
is correspondingly reduced.

VG = 5 V

Channel length = 0.2 μm

Current (0.2 mA/div)

Current (0.2 mA/div)

Current (0.2 mA/div)

VG = 5 V
VG = 5 V
2V 2V
1V 1V 1V
Drain voltage (1 V/div) Drain voltage (1 V/div) Drain voltage (1 V/div)
Experimental Calculated Calculated
device IV-characteristics with IV-characteristics
velocity saturation without velocity
saturation effect

Example of a velocity saturated device with channel length 0.2 mm compared with the constant mobility case.
Vasileska/Computational Electronics 64835_C002 Page Proof page 78 19.1.2010 11:34am Compositor Name: PG1421

78 Computational Electronics

Structures and energy

band diagrams

Bulk MOSFET Inversion layer

charge distribution

Source Drain

Substrate Gate Substrate


Double-gate MOSFET

Source Drain


Gate Body Gate
SiO2 SiO2

Ultrathin body MOSFET


Source Drain

FIGURE 2.49 Buried Tbody

Device structure, band diagram, and inversion charge in oxide
Gate Body Buried
conventional MOSFETs, dual-gate device structures and SiO2 oxide
fully-depleted SOI device structures.

Gate Gate
tox1 tox1
SiO2 SiO2
Source-N+ P Drain-N+ tSi Source-P+ N Drain-P+ tSi

Buried SiO2 tox2 Buried SiO2 tox2

Back gate (substrate) Back gate (substrate)

n-Channel p-Channel

n-Channel and p-channel SOI device structures.

Figure 2.50 illustrates the difference between PD-SOI and FD-SOI devices. In the PD case,
the depletion regions associated with the front gate and the back gate do not touch because
the SOI layer is too thick; whereas in the FD case, they overlap and the SOI layer is
completely depleted. The simulated output characteristics of the PD and FD SOI device
structure are shown in the left and right panels of Figure 2.51, respectively. In the case of
Vasileska/Computational Electronics 64835_C002 Page Proof page 79 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 79

Front Back Front Back

gate gate gate gate


Depletion regions Depletion region

Depletion regions in PD and FD devices.

4 ×10–4 4 × 10–4
VG =1 V-no impact tSi = 0.3 μm VG = 1 V-no impact tSi = 0.1 μm
3.5 ×10–4 VG = 2 V-no impact 3.5 × 10–4 VG = 2 V-no impact
VG = 3 V-no impact VG = 3 V-no impact
VG = 1 V-with impact
VG = 1 V-with impact
Drain current ID (A/μm)

Drain current ID (A/μm)

3 ×10–4 3 × 10–4 VG = 2 V-with impact
VG = 2 V-with impact VG = 3 V-with impact
VG = 3 V-with impact
2.5 ×10–4 2.5 × 10–4

2 ×10–4 2 × 10–4

1.5 ×10–4 1.5 × 10–4

1 ×10–4 1 × 10–4

5 ×10–5 5 × 10–5

0 × 100 0 ×100
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
Drain voltage VD (V) Drain voltage VD (V)

Output characteristics of PD (left panel) and FD (right panel) SOI devices.

the PD device, there is a noticeable ‘‘kink-effect’’ in the I–V characteristics due to the
presence of impact ionization, which generates excess holes that in turn modify the body
potential. The kink-effect does not appear in the FD device because the impact generated
holes at high drain bias are swept very fast by the source contact and do not lead to
modification of the body potential.
The variation of the subthreshold slope for different Si film thickness is shown in Figure
2.52. Note that the FD device has much better turn-off characteristics when compared with
the PD device. For this particular device geometry, the transition from PD to FD behavior
occurs around 100 nm Si film thickness.

2.9.5 MESFET
A cross-section of a MESFET is shown in Figure 2.53, which consists of a metal semicon-
ductor Schottky barrier contact that controls the conductivity of the region between the
source and the drain. To understand the operation of a MESFET [13], we consider the
section under the gate of Figure 2.53. The source is grounded, the gate is zero or reverse
biased, and the drain is zero or forward biased; that is, VG  0 and VD  0.
Vasileska/Computational Electronics 64835_C002 Page Proof page 80 19.1.2010 11:34am Compositor Name: PG1421

80 Computational Electronics


105 Transition region between

FD and PD SOI-device

Sumthreshold slope (mV/decade)






50 100 150 200 250 300 350
Thickness of the Si film (nm)

Subthreshold slope versus Si film thickness.

The drain voltage along the channel is shown on the right side of Figure 2.53. The voltage
drop across an elemental section dx of the channel can be obtained from Figure 2.53 and is
given by

ID dx
dV ¼ , (2:68)
qmn ND Wg [a  W(x)]

and the depletion-layer width at a distance y from the source is given by

2es [V(x) þ VG þ Vbi ]
W(x) ¼ : (2:69)

where Vbi is the built-in potential. Now, the drain current ID is constant, independent of x.
Then one can rewrite

ID dx ¼ qmn ND Wg [a  W(x)]dV: (2:70)

The differentiation of the drain voltage dV is obtained from Equation 2.69. Substituting dV
into Equation 2.70 and integrating from x ¼ 0 to x ¼ Lg gives
"     #
VD 2 VD þ VG þ Vbi 3=2 2 VG þ Vbi 3=2
I ¼ IP  þ , (2:71)
VP 3 VP 3 VP


Wg mn q2 ND2 a3 qND a2
IP ¼ and VP ¼ : (2:72)
2es Lg 2es
Vasileska/Computational Electronics 64835_C002 Page Proof page 81 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 81

The voltage VP is called the pinch-off voltage, that is, the total voltage (VD þ VG þ Vbi) at
which W2 ¼ a.
For high-frequency application of MESFETs, an important figure of merit is the cutoff
frequency fT, which is the frequency at which the MESFET can no longer amplify the input
signal. It is calculated using

gm IP =VP qm ND a2
fT ¼ < n 2 : (2:73)
2pCG 2pWg Lg (es =W) 2pes Lg

Gm is the device transconductance gm ¼ qID =qVGS jVDS ¼const: and is a measure of how
effectively the gate controls the channel charge. From Equation 2.73, one can see that to
improve high-frequency performance, a MESFET with high carrier mobility and short
channel length should be used. The derivation of Equation 2.71 is based on the assumption
that the carrier mobility in the channel is a constant, independent of the electric field.
However, for very high-frequency operations, the longitudinal field, i.e., the electric field
direct from the source to the drain, is sufficiently high that the carriers travel at their
saturation velocity. In this case,

IDsat ¼ Aqnvs ¼ Wg (a  W)qND vs : (2:74)

The cutoff frequency under the saturation velocity condition is then calculated using

gm Wg vs es =W vs
fT ¼ ¼ ¼ : (2:75)
2pCG 2pWg Lg es =W 2pLg

Therefore, to increase fT, we must reduce the gate length Lg and employ a semiconductor
with a high velocity.

2.9.6 HEMTs
GaAs-based HEMTs and pseudomorphic HEMTs (pHEMTs) [14] are rapidly replacing
conventional MESFET technology in military and commercial applications requiring low
noise figures and high gain, particularly at millimeter-wave frequencies. The application of
pHEMTs for high-efficiency power amplification is gaining popularity. Other commonly
used names for HEMTs include modulation doped FET (MODFET), two-dimensional
electron gas FET (TEGFET), and selectively doped heterojunction transistor (SDHT).
Since HEMTs and pHEMTs are field-effect transistors, the basic principles of their oper-
ation are very similar to those of the MESFETs described in Section 2.9.5. The main
difference between HEMTs and MESFETs is the epitaxial layer structure. In the HEMT
structure, compositionally different layers are grown in order to optimize and to extend
the performance of the FET. For group III–V semiconductors using a GaAs substrate, the
common materials used are AlxGa1xAs and GaAs. For most device applications, the
AlxAs mole fraction is between 0.2 < x < 0.3. The pHEMT also incorporates InxGa1xAs,
where InxAs is constrained to x < 0.3 for GaAs-based devices. These different layers form
heterojunctions since each layer has a different band gap. Structures grown with the same
lattice constant but different band gaps are simply referred to as lattice-matched HEMTs.
Those structures grown with slightly different lattice constants are called pseudomorphic
HEMTs or pHEMTs. Figure 2.53 shows the band gap energy as a function of the lattice
constant for the group III–V semiconductors [15].
Vasileska/Computational Electronics 64835_C002 Page Proof page 82 19.1.2010 11:34am Compositor Name: PG1421

82 Computational Electronics

W(x) W2
Source Drain
Lg ID x
VD 0 Lg
(Source) (Drain)

Cross section of the channel region of a MESFET (left), and drain voltage variation along the channel (right).

Epitaxial Layer Composition for Basic GaAs-Based HEMT and pHEMT Devices Compared
to Those of MESFET
þ þ þ
Ohmic contact n GaAs n GaAs n GaAs
Schottky contact n GaAs n AlGaAs n AlGaAs
Donor nþ AlGaAs or Si pulse doping nþ AlGaAs or Si pulse doping
Spacer Undoped AlGaAs Undoped AlGaAs
Channel nþ GaAs Undoped GaAs Undoped InGaAs
Buffer p GaAs p GaAs p GaAs

The epitaxial structure of a basic HEMT is illustrated in Figure 2.54. Similar to the
MESFET, the HEMT structure is grown on a semi-insulating GaAs substrate using molecu-
lar beam epitaxy (MBE), or less commonly, metal–organic chemical vapor deposition
(MOCVD). Table 2.5 contains the common MESFET, HEMT, and pHEMT epitaxial struc-
tures. The buffer layer, also typically GaAs, is epitaxially grown on the substrate in order to
isolate defects from the substrate and to create a smooth surface upon which to grow the
active layers of the transistor. Many pHEMT structures contain a superlattice structure to
further inhibit substrate conduction. A superlattice structure is a periodic arrangement of
undoped epitaxial layers used to realize a thicker epitaxial layer of a given property. For
example, alternating layers of AlxGa1xAs and GaAs form a typical pHEMT superlattice.
The AlxGa1xAs has a larger band gap than GaAs, making it superior to GaAs as a buffer.
However, due to strain problems, the AlxGa1xAs layer thickness is limited. To resolve this
problem, the AlxGa1xAs is grown to just below its critical thickness limit and a thin layer
of GaAs is grown on top. The GaAs relieves the strain and allows another layer of
AlxGa1xAs to be grown. This process is typically repeated 10 to 15 times, creating a
layer that is ‘‘essentially’’ a thick buffer of AlxGa1xAs.
In the conventional HEMT structure, the channel is grown next. In the ideal system, all of
the electron conduction would take place in this channel. The most important point about
the channel layer in the HEMT and pHEMT devices is the two-dimensional electron gas
(2DEG) that results from the band gap difference between AlxGa1xAs and GaAs (or
AlxGa1xAs and InxGa1xAs, in the case of the pHEMT). Figure 2.55 illustrates the band
Vasileska/Computational Electronics 64835_C002 Page Proof page 83 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 83

InSb Diamond structure crystal

0.64 Binary III-V semiconductor
Ternary III-V semiconductor

0.62 GaSb
Lattice constant a [nm]

InAs GaSb
0.60 InP
InP In0.52Al0.48As

0.58 In0.53Ga0.47As
GaAs0.5Sb0.5 GaAs
0.56 Ge GaAs
0 0.5 1 1.5 2 2.5
Band gap Eg (eV)

Band gap energy as a function of lattice constant.

diagram of a generic HEMT showing the 2DEG formed by the different band gaps. The
2DEG is formed since the higher band gap of AlxGa1xAs allows free electrons to diffuse
from the AlxGa1xAs to the lower band gap GaAs (or InxGa1xAs) near the interface.
A potential barrier then confines the electrons to a thin sheet of charge known as the
2DEG. In contrast to the MESFET, which has a doped channel and consequently a high
density of ionized donors, the donors in the HEMT structure are spatially separated from
the 2DEG (called modulation doping), which results in significantly less Coulomb scatter-
ing and very high channel mobilities. The remainder of the HEMT structure contains an
AlxGa1xAs spacer layer, a donor layer nþ AlxGa1xAs, an n AlxGa1xAs Schottky contact
layer, and a highly doped nþ GaAs layer. The spacer layer serves to separate the 2DEG
from any ionized donors generated by the pulse doping or nþ active layer. The drawback
of the spacer layer, however, is that the sheet carrier concentration (total amount of charge)
in the channel is reduced as the spacer layer thickness is increased. The donor layer or
Schottky layer is an nþ AlxGa1xAs layer and serves as the source of electrons. To avoid the
possibility of electron conduction in the AlxGa1xAs (which has low electron mobility), the
thickness of the Schottky must be chosen so that the depletion region of the gate overlaps
the depletion at the AlxGa1xAs=2DEG interface for depletion mode devices. The nþ GaAs
is present to realize low-resistance ohmic contacts. Most of the structures used today are
variants of this basic HEMT structure, having been optimized for performance and appli-
cations. For instance, many pHEMTs used for power applications will incorporate two
silicon modulation doping pulses, the second one below the channel, to increase the total
charge available.
The fabrication and basic operation of HEMT and pHEMT devices are very similar to
those for the MESFET. There exist some differences, mainly related to the presence of the
AlxGa1xAs in the epitaxial structure. As mentioned previously, AlxGa1xAs has a larger
band gap energy than GaAs and the band gap increases with the AlAs mole fraction.
HEMTs require ohmic contacts directly to the 2DEG, which is made more difficult with
increased AlAs mole fraction. An advantage of the AlGaAs is the higher Schottky barrier
height resulting from the deposition of the gate metal on the AlGaAs. Unfortunately, the
Vasileska/Computational Electronics 64835_C002 Page Proof page 84 19.1.2010 11:34am Compositor Name: PG1421

84 Computational Electronics

high doping in the donor layer decreases the breakdown voltage. However, power
HEMT and pHEMT structures with higher breakdown voltages (>10 V) have been
engineered using either double recess technology or by reducing the doping in the
Schottky layer.
Under operation, HEMTs and MESFETs are biased similarly. When a negative gate bias
is applied to the HEMT device, the Schottky layer becomes depleted. As the gate is biased
further, the 2DEG becomes depleted. This results in the modulation of the channel (2DEG)
by a negatively applied gate bias where gain and amplification occur until the channel is
pinched off (i.e., fully depleted). The transconductance is given by

ensat Wg
gm ¼ , (2:76)

e is the permittivity of InxGa1xAs
nsat is the saturated velocity of InxGa1xAs
Wg is the unit gate width of the device
d is the distance from the gate to the 2DEG

Under high-electric-field conditions, the HEMT shows a higher saturated velocity over
the MESFET. Since the conduction of electrons from the source to the drain takes place in a
channel that is well confined, gm will remain very high at low drain currents. This is
somewhat contrasted with the MESFET because at low drain currents, the distance d
will increase because the edge of the depletion region enters the tail of the doping profile.
This results in a compression of the gm. The higher mobility of the HEMT results in
lower parasitic drain and source resistances. As a result, ft ¼ gm=(2pCgs) and fmax are
increased from the MESFET case for a given gate length leading to a lower noise figure
and higher gain.
The reliability of HEMTs and pHEMTs [16] is affected by the epitaxial structure, device
fabrication, and device geometry. One of the drawbacks of using AlxGa1xAs in the
material structure is the occurrence of traps, called DX centers, for AlxAs mole fractions
around x ¼ 0.26. These traps are deep donor levels that can lead to reduced drain current,
an increase in low frequency noise and photoconductivity, and are particularly problem-
atic at low temperatures. Furthermore, the creation of DX centers increases with higher
doping of the AlxGa1xAs. DX centers are avoided by keeping the AlxAs content below
x ¼ 0.24 for n-type doped AlxGa1xAs. A second possible reliability problem that can occur
is the deconfinement of the 2DEG under high-temperature conditions. Thermally acceler-
ated testing has shown that the Al can migrate laterally into the gate, resulting in a change
in the conduction band discontinuity.
To take advantage of the high gm, HEMTs and pHEMTs rely on small geometries for
optimum performance. Paralleling the MESFET, HEMTs, and pHEMTs suffer the same
electromigration and metal interdiffusion reliability problems associated with the ohmic
and gate metallizations under device operation. In addition, hot electron traps resulting
from the generation of avalanche electrons are a problem for HEMTs and pHEMTs. The
hot electrons cause degradation in the current and in the gain and power under microwave
drive as they become trapped in the passivation or the AlxGa1xAs passivation interface.
For power devices, catastrophic failures or ‘‘burnout’’ can also be an issue due to the high
channel temperatures resulting from the large currents required for high power.
Vasileska/Computational Electronics 64835_C002 Page Proof page 85 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 85

Source Drain
Source passivation Gate passivation Drain

Source cap Drain cap

AIGaAs barrier δ-doping

AIGaAs spacer
InGaAs channel
Back spacer

Backside (δ–)doping


Semi-insulating substrate

Schematic cross section of a High Electron Mobility Transistor (HEMT).

ΔEC Fermi




Energy band diagram of a generic AlGaAs-GaAs HEMT showing the two-dimensional electron gas (2DEG)
quantum well channel.
Vasileska/Computational Electronics 64835_C002 Page Proof page 86 19.1.2010 11:34am Compositor Name: PG1421

86 Computational Electronics

25 100
Gain and power of HEMTs
of HBTs
20 of other FETs
Gain (dB/per stage)

Output power (W)




0 0.01
0 10 20 30 40 50 60 70 80 90 100
Frequency (GHz)

Gain per amplifier stage and output power of amplifiers with HEMTs, HBTs, and other FETs. (Wu, Y.-F. et al.,
Proceedings of the 24th International Symposium on Compound Semiconductors (ISCS 97), WA3, San Diego, CA, 1997.
With permission.)

2.1 What is the number of the nearest-neighbor atoms for simple cubic, FCC, and BCC
lattice. What is the number of the second nearest neighbors for simple cubic, FCC, and
BCC lattice.
2.2 What are the Miller indices for the planes shown in Figure P.2.2a and b?

4 4

y y
3 5
x x
(a) (b)


2.3 What are the Miller indices for the lines formed by the intersection of the crystal plane
shown in Figure P.2.2b) with planes [001], [010], and [100]?
Vasileska/Computational Electronics 64835_C002 Page Proof page 87 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 87

2.4 Find the angle a between the two nearest bonds in the diamond lattice.
2.5 Assume that the maximum donor concentration can be achieved when donor
atoms effectively ‘‘touch’’ each other. The donor atoms are regarded as spheres
with radii
es m0
aBd ¼ aB
e0 mdn

where aB ¼ 0.53 Å is the Bohr radius. Assuming that the semiconductor is GaAs,
estimate the maximum doping concentration. Use es=e0 ¼ 12.9 and mdn=m0 ¼ 0.067.
Hint: Assume that the maximum doping level donors are as closely packed as
possible, even though this is not necessarily a very realistic assumption.
2.6 The primitive translation vectors of the hexagonal space lattice may be taken as
pffiffiffi a a pffiffiffi a a
a1 ¼ 3 x þ y;
a2 ¼  3 x þ y; a3 ¼ cz
2 2 2 2
pffiffiffi 2
(a) Show that the volume of the primitive cell is 3=2a c.
(b) Show that the primitive translation vectors of a reciprocal lattice are AQ11

2.7 Show that the internal energy per unit volume of a 3D Fermi gas of free particles at
absolute zero is 3NEF(0)=(5V); show that the corresponding result for a 2D Fermi gas
is NEF(0)=(2V).
2.8 Determine the position of the Fermi level in a Si sample at T ¼ 0 K for the following
doping densities:
(a) NA ¼ 1016 cm3 , ND ¼ 0 (p-type sample)
(b) NA ¼ 1016 cm3 , ND ¼ 1015 cm3 (compensated sample)
(c) NA ¼ 1016 cm3 , ND ¼ 1017 cm3 (compensated sample)
(d) NA ¼ 0, ND ¼ 1017 cm3 (n-type sample)
Assume that both the acceptor and the donor type impurities introduce energy levels
within the band gap.
2.9 Given an NA ¼ 1014 cm3 doped Si sample,
(a) Present a qualitative argument for the approximate position of the Fermi level in
the material as T ! 0 K.
(b) Compute and plot the Fermi level as a function of T from T ¼ 300 K to T ¼ 500 K.
(c) Comment on the general position of the Fermi level as a function of T.
(d) How will the above answers be modified if the Si sample was doped with donors
instead of acceptors?
2.10 Determine the equilibrium electron and hole concentrations inside a uniformly
doped sample of Si under the following conditions:
(a) Room temperature, NA
ND , ND ¼ 1015 cm3
(b) Room temperature, NA ¼ 1016 cm3 , ND
(c) Room temperature, NA ¼ 9  1015 cm3 , ND ¼ 1016 cm3
(d) T ¼ 450 K, NA ¼ 0, ND ¼ 1014 cm3
(e) T ¼ 650 K, NA ¼ 0, ND ¼ 1014 cm3
Vasileska/Computational Electronics 64835_C002 Page Proof page 88 19.1.2010 11:34am Compositor Name: PG1421

88 Computational Electronics

2.11 Calculate the DOS function for a 2D and a 1D system. Use the approach for the
calculation of the 3D DOS function, which was described in class. Sketch the 3D, 2D,
and 1D DOS functions.
2.12 Find the relationship between the Fermi level EF and the electron concentration n for a
2D degenerate electron gas.
2.13 An effective density of states in the conduction band of silicon is equal to
2:8  1019 cm3 (at room temperature). The density of states effective mass for silicon
is 1:182me at 300 K and 1:077me at 77 K, where me is the free electron rest mass.
Consider a sample doped at 1016 cm3 by shallow ionized donors. Find an expression
for the number of electrons per unit energy, dn=dE, in the conduction band as a
function of energy in this sample and plot it at T ¼ 77 K and T ¼ 300 K.
2.14 Consider a silicon sample in the thermal equilibrium at room temperature (300 K).
The intrinsic carrier concentration of Si at this temperature is about 1:5  1010 cm3 .
Plot the concentration of electrons and holes versus acceptor doping in the range from
1013 cm3 to 1017 cm3 for the following shallow-donor concentrations:
(a) 1015 cm3
(b) 1016 cm3
2.15 Find the total charge of electrons injected into the p-region of an nþp silicon diode as a
function of the bias voltage V. The doping density of the p-region is NA, intrinsic
carrier concentration is ni, and the diode temperature is T. The length of the p-region is
L and the diffusion length of electrons in the p-region is Ln. Consider the following
three cases:
(a) Arbitrary relation between L and Ln
(b) L Ln
(c) L
Assume that at the contacts n ¼ np0, where np0 is the equilibrium concentration of
2.16 Derive the expression for the space-charge region width, the maximum electric field,
and the depletion capacitance of a linearly graded junction.
2.17 Consider a pn-junction diode. The concentration of holes in the n-section of the device
is described by the continuity equation

d2 pn pn  pn0
Dp  ¼ 0:
dx2 tp

The concentration of shallow ionized donors in the n-section is equal to 1015 cm3.
The intrinsic carrier concentration is 1010 cm3. The forward voltage applied to the
diode is 0.5 V. Assuming that the length of the n-section, L, is much smaller than the
diffusion length Lp, calculate and sketch the hole distribution in the n-section of the
device. Also, assuming that Dp ¼ 12 cm2=s and the lifetime tp ¼ 1 ms, how short does
the n-section have to be to satisfy the condition that L
Lp (use L ¼ Lp =10 as a
Vasileska/Computational Electronics 64835_C002 Page Proof page 89 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 89

2.18 The forward IV-characteristics of a pn-diode are shown in Figure P.2.18.

(a) Explain the origins for the deviation of the measured IV characteristics from the
ideal model predictions.
(b) Calculate the series resistance of the diode. Explain how you arrived at your



Current (A)





0 0.1 0.2 0.3 0.4 0.5
Voltage (V)


2.19 Show diagrammatically and with analytical expressions (simple ones) the current
components in reverse active mode and in the cut-off mode (Figure P.2.19).


(a) Reverse active mode:


(b) Cut-off mode


2.20 The output characteristics of a BJT in a common-emitter configuration is given in

Figure P.2.20:
(a) What is the small signal current gain for VCE ¼ 10 V?
(b) What is the output conductance of this device?
(c) Estimate the Early voltage of the transistor. What is the reason for the Early
Vasileska/Computational Electronics 64835_C002 Page Proof page 90 19.1.2010 11:34am Compositor Name: PG1421

90 Computational Electronics

30.00 m
lb = 120 μA

lb = 90 μA
Collector current (mA)

20.00 m

lb = 60 μA

Change in lc
10.00 m Change in VCE
lb = 30 μA

lb = 0 μA

0.00 5.00 10.00 15.00 20.00
Collector emitter voltage (V)


2.21 Sketch the energy band diagram and the charge distribution in a MOS capacitor with
an n-type semiconductor for the following bias conditions:
(a) Accumulation
(b) Depletion
(c) Inversion
2.22 The field distribution in an ideal MOS capacitor is shown in Figure P.2.22.
(a) Sketch the potential and the charge distribution profiles. Is the MOS capacitor
biased in accumulation, depletion, or inversion regime?
(b) Calculate the voltage drop in the semiconductor and in the oxide.
(c) What is the magnitude of the oxide capacitance Cox?
(d) Calculate the threshold voltage VTH for the MOS capacitor.
The dielectric permittivity of silicon is es ¼ 1:05  1010 F=m, the oxide permittivity
is eox ¼ 3:45  1011 F=m, the intrinsic carrier concentration equals to ni ¼ 1:5
1010 cm3 and the temperature is T ¼ 300 K.
Electric field (kV/cm)




0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 x (μm)

Vasileska/Computational Electronics 64835_C002 Page Proof page 91 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 91

2.23 Plot the high-frequency CV curves for the three structures shown in Figure P.2.23 on
the same plot. Explain how you arrived at your answer. In your analysis, assume that
the gate voltage is swept very slowly.


p+ p+ n+ n+
p-Type SC p-Type SC p-Type SC

(a) (b) (c)


2.24 A charge density of 2.4  107 C=cm2 is uniformly distributed in the oxide in a
MOSFET structure. The oxide thickness is dox ¼ 150 nm.
(a) Calculate the contribution to the threshold voltage from these charges.
(b) Calculate for the case when all charges are located at the semiconductor=oxide
(c) Calculate when the charges are located 20 nm away from the semiconductor-oxide
interface into the oxide layer.
In your calculations, use eox ¼ 3:45  1011 F=m for the dielectric permittivity of the
2.25 A charge density of 2:4  107 C=cm2 is uniformly distributed in the oxide in a
MOSFET structure. The oxide thickness is dox ¼ 150 nm.
(a) Calculate the contribution to the threshold voltage from these oxide charges.
(b) Calculate for the case when all the charges are located at the silicon–silicon
dioxide interface.
(c) Calculate for the linear charge distribution within the oxide (zero charge density at
the Si–SiO2 interface and maximum charge density at the metal–SiO2 interface).
2.26 Consider nearly two identical silicon MOSFETs. The only difference between the two
devices is that the oxide layer in one of them is perfectly clean and the other one is
contaminated with sodium ions that produce a positive charge density. The concen-
tration of sodium ions equals to 2  1016 cm3 and the thickness of the oxide layer is
0.1 mm. The permittivity of SiO2 is 3:45  1011 F=m.
(a) What is the difference (including sign) in the device threshold voltages if these
devices are n-channel devices?
(b) What is the difference (including sign) in the device threshold voltages if these
devices are p-channel devices?
(c) Assume that the threshold voltage of the n-channel clean device is 1 V and that the
threshold voltage of the p-channel clean device is 1 V. Sketch the qualitative
dependencies of the drain-to-source saturation current on the gate voltage for
all four devices (clean and contaminated n-channel and clean and contami-
nated p-channel devices). Label the thresholds and shifted thresholds on the
gate-voltage axis.
Vasileska/Computational Electronics 64835_C002 Page Proof page 92 19.1.2010 11:34am Compositor Name: PG1421

92 Computational Electronics

2.27 Derive the expressions for the extrinsic transconductance and the extrinsic drain
conductance of a transistor with finite source and drain series resistances in terms
of those of an ideal transistor with zero series resistances.
2.28 Calculate the dependence of the drain current ID upon the drain voltage VDS for
VGS ¼ 5 V for a silicon MOSFET with the following values of the source (RS) and drain
resistance (RD): RS ¼ RD ¼ 0 V, and RS ¼ RD ¼ 100 V. The device parameters are as
Gate length: L ¼ 4 mm
Gate width: W ¼ 100 mm
Electron mobility in the channel: mn ¼ 1000 cm2=V s
Dielectric permittivity of gate oxide: eox ¼ 3.45  1011 F=m
Dielectric permittivity of silicon: esc ¼ 1.05  1010 F=m
Flat-band voltage: VFB ¼ 0 V
Substrate bias: Vsub ¼ 0 V
Temperature: T ¼ 300 K
Substrate doping: NA ¼ 1015 cm3
Gate oxide thickness: dox ¼ 20 nm
Intrinsic carrier concentration: ni ¼ 1.5  1010 cm3
In your calculations, use:
(a) Square-law theory
(b) Bulk-charge theory
Repeat the problem for NA ¼ 1016 cm3 and NA ¼ 1017 cm3. Discuss the validity of
the square-law theory versus substrate doping and the influence of the series resist-
ance on the drain current characteristics of these devices.
2.29 Derive an expression for the drain saturation current of an n-channel MOSFET using
the square-law theory and neglecting velocity saturation effects, i.e., assuming the
constant mobility model, but taking into account the source series resistance RS. Use
the following MOSFET parameters:
Gate oxide thickness: dox ¼ 17.5 nm
Device gate width: W ¼ 100 mm
Gate length: L ¼ 4 mm
Threshold voltage: VT ¼ 1 V
Electron mobility in the channel: mn ¼ 800 cm2=V s
Dielectric permittivity of gate oxide: eox ¼ 3.45  1011 F=m
Gate voltage: VGS ¼ 5 V
Substrate bias: Vsub ¼ 0 V
Plot IDsat versus RS for 2V < RS  20 V.
2.30 Calculate and plot the subthreshold current for a long-channel Si MOSFET as a
function of the gate voltage VGS that varies in the range between VT  1 (V) and VT.
For the drain voltage, assume VDS ¼ 0.01, 0.1, and 10 V. Use the following parameters:
Threshold voltage: VT ¼ 1 V
Substrate doping: NA ¼ 1015 cm3
Electron mobility in the channel: mn ¼ 800 cm2=V s
Device gate width: W ¼ 100 mm
Gate length: L ¼ 20 mm
Gate oxide thickness: dox ¼ 50 nm
Energy band gap: EG ¼ 1.12 eV
Vasileska/Computational Electronics 64835_C002 Page Proof page 93 19.1.2010 11:34am Compositor Name: PG1421

Introductory Concepts 93

Effective density of states in the conduction band: NC ¼ 3.22  1019 cm3

Effective density of states in the valence band: NV ¼ 1.83  1019 cm3
Dielectric permittivity of gate oxide: eox ¼ 3.45  1011 F=m
Dielectric permittivity of silicon: esc ¼ 1.05  1010 F=m
Temperature: T ¼ 300 K
2.31 Consider the following idealized ion implantation profile near the semiconductor-
insulator interface in a Si MOSFET (x ¼ 0 corresponds to the semiconductor-insulator
interface). Calculate the threshold voltage shift as a function of Ni for 1014 cm3 
Ni  1017 cm3 for dimp ¼ 0.08 mm. Assume
Semiconductor background doping: NA ¼ 1015 cm3
Gate oxide thickness: dox ¼ 50 nm
Energy band gap: EG ¼ 1.12 eV
Effective density of states in the conduction band: NC ¼ 3.22  1019 cm3
Effective density of states in the valence band: NV ¼ 1.83  1019 cm3
Dielectric permittivity of gate oxide: eox ¼ 3.45  1011 F=m
Dielectric permittivity of silicon: esc ¼ 1.05  1010 F=m
Temperature: T ¼ 300 K
Assume shallow ionized acceptors. Define the threshold voltage as the gate
voltage for which the electron concentration at the surface n(0) is equal to NA þ Ni
(Figure P.2.31).

Doping density

Ni dimp


x=0 Distance x


2.32 The current–voltage characteristics of a long-channel MOSFET are described with

( Zm  
eff Cox
(VG  VT )VD  12 VD2 , VD < VG  VT
ID ¼ L
Zmeff Cox 2
2L (VG  VT ) , VD > VG  VT

meff is the effective electron mobility in the channel
L is the effective channel length
Z is the channel (gate) width
Cox is the gate capacitance
ID is the drain current
VD is the drain voltage (with the source as reference)
VG is the gate voltage
VT is the threshold voltage
Vasileska/Computational Electronics 64835_C002 Page Proof page 94 19.1.2010 11:34am Compositor Name: PG1421

94 Computational Electronics

When the drain and gate are connected, consider the following two cases:
(a) VT > 0
(b) VT < 0
Find and plot ID in terms of VD in each of these two cases.
2.34 A GaAs MESFET is fabricated using an epitaxial layer doped to ND ¼ 1017 cm3 that is
0.2 mm thick. The Schottky barrier metallization has a barrier height of 0.75 eV.
(a) Evaluate the built-in voltage Vbi.
(b) Evaluate the pinch-off voltage Vpo.
(c) Evaluate the threshold voltage VT.
(d) Is the resulting transistor a depletion or enhancement mode device?
(e) What is the depletion layer thickness with zero gate voltage?
2.35 Write a program to evaluate the drain current as a function of VD for a specified gate
voltage. Use the transistor parameters of Problem 2.34. The program should be valid
for all channel lengths; assume that mn ¼ 5000 cm2=V s and vs ¼ 1.3  107 cm=s.
(a) Plot ID(VD) with VG ¼ 0 for a transistor with L ¼ 10 mm and Z ¼ 100 mm.
(b) Repeat (a) for L ¼ 0.2 mm and Z ¼ 100 mm.
(c) Compare VDsat in (b) with the value VDsat ¼ EsL, which is the extreme velocity
saturation limit. Does this limit apply to a transistor with L ¼ 0.2 mm?

1. R. G. Burns, Mineralogical Applications of Crystal Field Theory (Science - Cambridge University
Press, Cambridge, U.K., 1970).
2. C. Kittel, A. Zettl, and P. McEuen, Introduction to Solid State Physics, John Wiley & Sons Inc.
(November 2004).
3. N. W. Ashcroft and N. D. Mermin, Solid State Physics (Harcourt, New York, 1976).
4. C. Thomas, Mechanical Behavior of Materials (Waveland Press, Long Grove, IL, 2000), pp. 85.
5. J. W. Edington, Practical Electron Microscopy in Materials Science (N. V. Philips’ Gloeilampenfab-
rieken, Eindhoven, 1976), ISBN 1-878907-35-2, Appendix 2.
6. L. V. Azároff, R. Kaplow, N. Kato, R. J. Weiss, A. J. C. Wilson, and R. A. Young, X-ray Diffraction
(McGraw-Hill, New York, 1974).
7. D. J. Chadi, Special points for Brillouin-zone integrations, Phys. Rev. B 16, 1746–1747 (1977).
8. S. M. Sze, Modern Semiconductor Device Physics (John Wiley & Sons, Incorporated, New York,
October 1997).
9. M. S. Shur, Introduction to Electronic Devices (John Wiley & Sons, Incorporated, August 1995).
10. D. K. Ferry, Quantum Mechanics: An Introduction for Device Physicists and Electrical Engineers
(Taylor & Francis, Inc., January 2001).
11. E. Mertzbacher, Quantum Mechanics (John Wiley & Sons, Incorporated, December 1997).
12. W. A. Harrison, Electronic Structure and the Properties of Solids: The Physics of the Chemical Bond
(Dover Publications, New York, July 1989).
13. J. C. Bean, Materials and technologies, in High-Speed Semiconductor Devices, S. M. Sze, ed. (John
Wiley & Sons, New York, 1990).
14. J. Mun, ed., GaAs Integrated Circuits, (MacMillan Publishing Company, New York, 1988).
15. F. Ali, and A. Gupta, eds., HEMTs and HBTs: Devices, Fabrication and Circuits (Artech House,
Boston, MA, 1991).
16. Y.-F. Wu, D. Kapolnek, P. Kozodoy, S. Keller, B. P. Keller, S. P. Denbaars, and U. K. Mishra,
AlGaN=GaN MODFETs with low ohmic contact resistance through source-drain nþ
Re-growth, in Proceedings of the 24th International Symposium on Compound Semiconductors
(ISCS 97), WA3, San Diego, CA (1997).
17. A. Christou, ed., Reliability of GaAs MMICs (John Wiley and Sons, West Sussex, U.K., 1992).
Vasileska/Computational Electronics 64835_C002 Page Proof page 95 19.1.2010 11:34am Compositor Name: PG1421

[AQ1] In the sentence beginning with ‘‘The body-centered cubic . . . ’’, please check whether ‘‘BCC’’ is the
correct acronym for this term.
[AQ2] In the sentence beginning with ‘‘Finally, the face-centered cubic . . . ’’, please check whether
‘‘FCC’’ is the correct acronym for this term.
[AQ3] Please check the in-text citation of Table 2.2 for appropriateness.
[AQ4] Please check the appearance of question marks with the data of Table 2.2 and update.
[AQ5] Table 2.2 cited at this occurrence is renumbered as Table 2.3 corresponding to the text discussed.
Please check.
[AQ6] Footnote ‘‘a’’ cross-referred in Table 2.3 is not defined. Please update.
[AQ7] Please provide complete details for the source line of Figures 2.22 and 2.47, and also include the
same in references list.
[AQ8] Figures are to appear in grayscale. Modify the phrase referring to color in the caption of Figure 2.25
[AQ9] Please review the sentence beginning with ‘‘the so-called square law theory . . . ’’ for clarity.
[AQ10] Please provide caption for all problem figures if appropriate.
[AQ11] Please review the sentence beginning with ‘‘Show that the primitive . . . ’’ for clarity.