4.

3 SEMICONDUCTORS
The Si atom has 14 electrons, which distribute themselves in the various atomic
energy levels as shown in Figure 4.16. The inner shells (n = 1 and n = 2) are full
and therefore “closed.” Since these shells are near the nucleus, when Si atoms come
together to form the solid, they are not much affected and they stay around the par-
ent Si atoms. They can therefore be excluded from further discussion. The 3s and
3p subshells are farther away from the nucleus. When two Si atoms approach, these
electrons strongly interact with each other. Therefore, in studying the formation of
bands in the Si solid, we will only consider the 3s and 3p levels.
The first task is to examine why Si actually bonds with four neighbors, since
the 3s orbital is full and there are only two electrons in the 3p orbitals. The full 3s
orbital should not overlap a neighbor and become involved in bonding. Since only
two 3p orbitals are half full, bonds should be formed with two neighboring Si atoms.
 4.3 SEMICONDUCTORS 329

ψ(3s)
ψ(3py)
Electron energy

3p
ψ hyb
3s y

2p
x
ψ(3px) ψ(3pz)
2s

1s
(a) Isolated Si (b) Si just before bonding

Figure 4.16 The Figure 4.17 (a) Si is in Group IV in the Periodic Table. An isolated Si atom has two
electronic structure electrons in the 3s and two electrons in the 3p orbitals. (b) When Si is about to bond,
of Si. the one 3s orbital and the three 3p orbitals become perturbed and mixed to form four
hybridized orbitals, ψhyb, called sp3 orbitals, which are directed toward the corners of
a tetrahedron. The ψhyb orbital has a large major lobe and a small back lobe. Each ψhyb
orbital takes one of the four valence electrons.

In reality, the 3s and 3p energy levels are quite close, and when five Si atoms approach
each other, the interaction results in the four orbitals ψ(3s), ψ(3px), ψ(3py), and ψ(3pz)
mixing together to form four new hybrid orbitals, which are directed in tetrahedral
directions; that is, each one is aimed as far away from the others as possible as
illustrated in Figure 4.17. We call this process sp3 hybridization, since one s orbital
and three p orbitals are mixed. (The superscript 3 on p has nothing to do with the
number of electrons; it refers to the number of p orbitals used in the hybridization.)
The four sp3 hybrid orbitals, ψhyb, each have one electron, so they are half occu-
pied. This means that four Si atoms can have their orbitals ψhyb overlap to form bonds
with one Si atom, which is what actually happens; thus, one Si atom bonds with
four other Si atoms in tetrahedral directions.
In the same way, one Si atom bonds with four H atoms to form the important
gas SiH4, known as silane, which is widely used in the semiconductor technology
to fabricate Si devices. In SiH4, four hybridized orbitals of the Si atom overlap with
the 1s orbitals of four H atoms. In exactly the same way, one carbon atom bonds
with four hydrogen atoms to form methane, CH4.
There are two ways in which the hybrid orbital ψhyb can overlap with that of the
neighboring Si atom to form two molecular orbitals. They can add in phase (both
positive or both negative) or out of phase (one positive and the other negative) to
produce a bonding or an antibonding molecular orbital ψB and ψA, respectively, with
energies EB and EA as shown in Figure 4.18a to c. Each Si–Si bond thus corresponds
to two paired electrons in a bonding molecular orbital ψB. In the solid, there are
N(∼5 × 1022 cm−3) Si atoms, and there are nearly as many such ψB bonds. The interac-
tions between the ψB orbitals (i.e., the Si–Si bonds) lead to the splitting of the EB energy
level to N levels, thereby forming an energy band labeled the valence band (VB) by
330 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

(a) (b) (c) (d)

ψA
ψhyb Conduction band

ψA

3p
Energy gap, Eg
3s
ψB

Si atom

Valence band
ψhyb ψB

Figure 4.18 (a) Formation of energy bands in the Si crystal first involves hybridization
of 3s and 3p orbitals to four identical ψhyb orbitals, which are at 109.5° to each other
as shown in (b). (c) ψhyb orbitals on two neighboring Si atoms can overlap to form
ψB or ψA. The first is bonding orbital (full) and the second is an antibonding orbital
(empty). In the crystal, ψB overlap to give the valence band (full) and ψA overlap to
give the conduction band (empty) (d). Si crystal

virtue of the valence electrons it contains. Since the energy level EB is full, so is the
valence band. Figure 4.18c and d illustrate the formation of the VB from EB.
In the solid, the interactions between the N number of ψA orbitals result in the
splitting of the energy level EA to N levels and the formation of an energy band
that is completely empty and separated from the full valence band by a definite
energy gap Eg. In this energy region, there are no states; therefore, the electron can-
not have energy with a value within Eg. The energy band formed from NψA orbitals
is a conduction band (CB), as also indicated in Figure 4.18c and d.
The electronic states in the VB (and also in the CB) extend throughout the whole
solid, because they result from NψB orbitals interfering and overlapping each other.
As before NψB, orbitals can overlap in N different ways to produce N distinct wave-
functions ψvb that extend throughout the solid. We cannot relate a particular electron
to a particular bond or site because the wavefunctions ψvb corresponding to the VB
energies are not concentrated at a single location. The electrical properties of solids
are based on the fact that in solids, such as semiconductors and insulators, there are
certain bands of allowed energies for the electrons, and these bands are separated by
energy gaps, that is, bandgaps. The valence and conduction bands for the ideal Si
crystal shown in Figure 4.18d are separated by an energy gap, or a bandgap, Eg,
in which there are no allowed electron energy levels.
 4.3 SEMICONDUCTORS 331

Figure 4.19 A simplified energy band

CB diagram of a semiconductor. CB is the
e– conduction band and VB is the valence
Ec band. At 0 K, the VB is full of electrons
and the CB is empty. If a photon of energy
Electron energy

Photoexcitation Thermal hf > Eg is incident on the semiconductor, it

Eg can be absorbed by an electron in the VB,
hf > Eg excitation
h+ which becomes photoexcited into the CB.
Ev Some electrons in the VB can be
excited into the CB by thermal excitation,
VB that is, occasional rupturing of Si-Si bonds
by energetic lattice vibrations. Thermal
generation creates electron and hole pairs.

We can generalize the energy band diagram of a semiconductor as shown in

Figure 4.19. At absolute zero of temperature the VB will be full of electrons and
the CB will be empty. The conductivity of this ideal semiconductor would be zero
as there are no free carriers to drift. It is possible to excite an electron from the VB
to the CB if a photon of energy hf equal or greater than the bandgap is incident on
this semiconductor. The photon can be absorbed by an electron in the VB, which
becomes photoexcited into the CB5. An electron in the CB is essentially in an empty
band. We can consider this electron in the CB as a free carrier with a certain effec-
tive mass m*.e If there is an electric field Ex along x then this photoexcited electron
will be acted on by a force, F = −eEx, and it will try to move in the −x direction.
For it to do so, there must be empty higher energy levels, so that as the electron
accelerates and gains energy, it moves up in the band. When an electron collides
with a lattice vibration, it loses the energy acquired from the field and drops down
within the CB. Again, it should be emphasized that states in an energy band are
extended; that is, the electron is not localized to any one atom.
Note also that the photogeneration of an electron from the VB to the CB leaves
behind a VB state with a missing electron. This unoccupied electron state has an
apparent positive charge, because this crystal region was neutral prior to the removal
of the electron. The VB state with the missing electron is called a hole and is denoted
h+. The hole can “move” in the direction of the field by exchanging places with a
neighboring valence electron hence it contributes to conduction, as will be discussed
in Chapter 5.
At temperatures above absolute zero, the atoms in a solid vibrate due to their
thermal energy. Some of the atoms can acquire a sufficiently high energy from
thermal fluctuations to strain and rupture their bonds. Physically, there is a possibil-
ity that the atomic vibration will impart sufficient energy to the electron for it to
surmount the bonding energy and leave the bond. The electron must then enter a
higher energy state. In the case of Si, this means entering a state in the CB, as shown
in Figure 4.19. The excitation of electrons from the VB to the CB by lattice vibra-
tions is called thermal generation, and results in the generation of electrons in the

5
In pure intuitive terms, the incident photon has sufficient energy to be able to rupture a Si–Si bond and
release a free electron. An electron is free only in the CB, so this process implies the photoexcitation of an
electron from the VB to the CB.
332 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

CB and holes in the VB as shown in Figure 4.19. The electrons in the CB and holes
in the VB can contribute to conduction, and semiconductors above absolute zero of
temperature have a finite conductivity.

What wavelengths of light can be absorbed by a Si photodetector given Eg = 1.1 eV? Can such a
photodetector be used in fiber-optic communications at light wavelengths of 1.31 μm and 1.55 μm?

SOLUTION

The energy bandgap Eg of Si is 1.1 eV. A photon must have at least this much energy to
excite an electron from the VB to the CB, where the electron can drift. Excitation corresponds
to the breaking of a Si–Si bond. A photon of less energy does not get absorbed, because its
energy will put the electron in the bandgap where there are no states. Thus, hc∕λ > Eg gives
hc (6.6 × 10−34 J s) (3 × 108 m s−1 )
λ < =
Eg (1.1 eV) (1.6 × 10−19 J/eV)
= 1.13 × 10−6 m or 1.1 μm
Since optical communications networks use wavelengths of 1.3 and 1.55 μm, these light waves
will not be absorbed by Si and thus cannot be detected by a Si photodetector.
334 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

4.4 ELECTRON EFFECTIVE MASS

When an electric field Ex is applied to a metal, an electron near the Fermi level can
gain energy from the field and move to higher energy levels, as shown in Figure 4.13.
The external force Fext = eEx is in the x direction, and it drives the electron along
x. The acceleration of the electron is still given by a = Fext∕me, where me is the mass
of the electron in vacuum.
The law Fext = mea cannot strictly be valid for the electron inside a solid, because
the electron interacts with the host ions and experiences internal forces Fint as it
moves around, as depicted in Figure 4.22. The electron therefore has a PE that var-
ies with distance. Recall that we interpret mass as inertial resistance against accel-
eration per unit applied force. When an external force Fext is applied to an electron
in the vacuum level, as in Figure 4.22a, the electron will accelerate by an amount
Fext
avac = [4.4]
me
as determined by its mass me in vacuum.

Ex Ex

Fint
e– Fext Fext
a= a=
me me*
Fext
Vacuum Crystal

x x
(a) An external force Fext applied to (b) An external force Fext applied to
an electron in a vacuum results in an electron in a crystal results in an
an acceleration avac = Fext /me. acceleration acryst = Fext /me*.

Figure 4.22
 4.4 ELECTRON EFFECTIVE MASS 335

When the same force Fext is applied to the electron inside a crystal, the accel-
eration of the electron will be different, because it will also experience internal
forces, as shown in Figure 4.22b. Its acceleration in the crystal will be
Fext + Fint
acryst = [4.5]
me

where Fint is the sum of all the internal forces acting on the electron, which is quite
different than Equation 4.4. To the outside agent applying the force Fext, the electron
will appear to be exhibiting a different inertial mass, since its acceleration will be
different. It would be most useful for the external agent if the effect of the internal
forces in Fint could be accounted for in a simple way, and if the acceleration could
be calculated from the external force Fext alone, through something like Equation 4.4.
This is indeed possible.
In a crystalline solid, the atoms are arranged periodically, and the variation
of Fint, and hence the PE, or V(x), of the electron with distance along x, is also
periodic. In principle, then, the effect on the electron motion can be predicted
and accounted for. When we solve the Schrödinger equation with the periodic PE,
or V(x), we essentially obtain the effect of these internal forces on the electron
motion. It has been found that when the electron is in a band that is not full, we
can still use Equation 4.4, but instead of the mass in vacuum me, we must use
the effective mass m*e of the electron in that particular crystal. The effective mass
is a quantum mechanical quantity that behaves in the same way as the inertial
mass in classical mechanics. The acceleration of the electron in the crystal is
then simply
Fext
acryst = [4.6]
m*e
The effects of all internal forces are incorporated into m*e. It should be
emphasized that m*e is obtained theoretically from the solution of the Schrödinger
equation for the electron in a particular crystal, a task that is by no means
trivial. However, the effective mass can be readily measured. For some of the
familiar metals, m*e is very close to me. For example, in silver, m*e = me for all
practical purposes, whereas in lithium m*e = 2.2me, as shown in Table 4.2. On
the other hand, m*e for many metals and semiconductors is appreciably different
than the electron mass in vacuum and can even be negative. (m*e depends on the
properties of the band that contains the electron as discussed in Section 5.13.)

Table 4.2 Effective mass m*e of electrons in some metals

Metal Ag Au Bi Cu Fe K Li Mg Na Zn
m*e
1.0 1.1 0.008 1.3 12 1.2 2.2 1.3 1.2 0.85
me

Note: Table compiled from multiple sources; values are typical.

336 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

4.5 DENSITY OF STATES IN AN ENERGY BAND

Although we know there are many energy levels (perhaps ∼1023) in a given band,
we have not yet considered how many states (or electron wavefunctions) there are
per unit energy per unit volume in that band. Consider the following intuitive argu-
ment. The crystal will have N atoms and there will be N electron wavefunctions ψ1,
ψ2, . . . , ψN that represent the electron within the whole crystal. These wavefunctions
are constructed from N different combinations of atomic wavefunctions, ψA, ψB,
ψC, . . . as schematically illustrated in Figure 4.23a,6 starting with
ψ1 = ψA + ψB + ψC + ψD + · · ·
all the way to alternating signs
ψN = ψA − ψB + ψC − ψD + · · ·
and there are N(∼10 ) combinations. The lowest-energy wavefunction will be ψ1
23

constructed by adding all atomic wavefunctions (all in phase), and the highest-energy
wavefunction will be ψN from alternating the signs of the atomic wavefunctions,
which will have the highest number of nodes. Between these two extremes, especially
around N∕2, there will be many combinations that will have comparable energies
and fall near the middle of the band. (By analogy, if we arrange N = 10 coins by
heads and tails, there will be many combinations of coins in which there are 5 heads
and 5 tails, and only one combination in which there are 10 heads or 10 tails.) We
therefore expect the number of energy levels, each corresponding to an electron
wavefunction in the crystal, in the central regions of the band to be very large as
depicted in Figure 4.23b and c.
Figure 4.23c illustrates schematically how the energy and volume density of
electronic states change across an energy band. We define the density of states g(E)
such that g(E) dE is the number of states (i.e., wavefunctions) in the energy interval
E to (E + dE) per unit volume of the sample. Thus, the number of states per unit
volume up to some energy E′ is
E′
Sv (E′) = ∫0
g(E) dE [4.7]

which is called the total number of states per unit volume with energies less than
E′. This is denoted Sv(E′).
To determine the density of states function g(E), we must first determine the
number of states with energies less than E′ in a given band. This is tantamount to
calculating Sv(E′) in Equation 4.7. Instead, we will improvise and use the energy
levels for an electron in a 3D potential well. Recall that the energy of an electron in
a cubic PE well of size L is given by
h2
E= (n21 + n22 + n23 ) [4.8]
8meL2

6
This intuitive argument, as schematically depicted in Figure 4.23a, is obviously highly simplified because the
solid is three-dimensional (3D) and we should combine the atomic wavefunctions not on a linear chain but on a
3D lattice. In the 3D case there are large numbers of wavefunctions with energies that fall in the central regions
of the band.
 4.5 DENSITY OF STATES IN AN ENERGY BAND 337

E
ψN

ψ1
g( E)
❘❚❯❱❲❳ ❨❩❚❬
(a) (b) (c)
Figure 4.23 (a) In the solid there are N atoms and N extended electron wavefunctions from ψ1
all the way to ψN. There are many wavefunctions, states, that have energies that fall in the central
regions of the energy band. Note that although only eight atoms are shown, these are eight
sequential atoms among N atoms, and N is very large. Overall, the wavefunctions for N atoms
must be symmetric or antisymmetric. (b) The distribution of states in the energy band; darker
regions have a higher number of states. (c) Schematic representation of the density of states
g(E) versus energy E.

where n1, n2, and n3 are integers 1, 2, 3, . . . . The spatial dimension L of the well
now refers to the size of the entire solid, as the electron is confined to be some-
where inside that solid. Thus, L is very large compared to atomic dimensions,
which means that the separation between the energy levels is very small. We will
use Equation 4.8 to describe the energies of free electrons inside the solid (as in
a metal).
Each combination of n1, n2, and n3 is one electron orbital state. For example,
ψn1,n2,n3 = ψ1,1,2 is one possible orbital state. Suppose that in Equation 4.8 E is given
as E′. We need to determine how many combinations of n1, n2, n3 (i.e., how many ψ)
have energies less than E′, as given by Equation 4.8. Assume that (n12 + n22 + n32) = n′2.
The object is to enumerate all possible choices of integers for n1, n2, and n3 that
satisfy n21 + n22 + n23 ≤ n′2.
The two-dimensional 2D case is easy to solve. Consider n21 + n22 ≤ n′2 and the
2D n-space where the axes are n1 and n2, as shown in Figure 4.24. The 2D space
is divided by lines drawn at n1 = 1, 2, 3, . . . and n2 = 1, 2, 3, . . . into infinitely
many boxes (squares), each of which has a unit area and represents a possible state
ψn1,n2 . For example, the state n1 = 1, n2 = 3 is shaded, as is that for n1 = 2, n2 = 2.
Clearly, the area contained by n1, n2 and the circle defined by n′2 = n21 + n22
( just like r2 = x2 + y2) is the number of states that satisfy n21 + n22 ≤ n′2. This area
is 14 (πn′ 2 ) .
338 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

n2 n3

n12 ❛ n22 ❭ n'2 In here n12 + n22 + n32 ≤ n'2

5 Vol. = 1 ( 4 πn'3 )
8 3
4
n'
3
n1 ❭ ❪
2
n2 ❭ ❫ n2
1
➊ n1 n1
0 1 2 3 4 5 6

➊ n2 n1 ❭ ❴❵ n2 ❭ ❴ n1
Figure 4.24 Each state, or electron wavefunction in Figure 4.25 In three dimensions, the volume
the crystal, can be represented by a box at n1, n2. defined by a sphere of radius n′ and the positive
axes n1, n2, and n3, contains all the possible
combinations of positive n1, n2, and n3 values that
satisfy n21 + n22 + n23 ≤ n′2.

In the 3D case, n12 + n22 + n32 ≤ n′2 is required, as indicated in Figure 4.25. This is
the volume contained by the positive n1, n2, and n3 axes and the surface of a sphere of
radius n′. Each state has a unit volume, and within the sphere, n12 + n22 + n32 ≤ n′2 is
satisfied. Therefore, the number of orbital states Sorb(n′) within this volume is given by

Sorb (n′) = ( πn′ 3) = πn′ 3

1 4 1
8 3 6
Each orbital state can take two electrons with opposite spins, which means that
the number of states, including spin, is given by
1
S(n′) = 2Sorb (n′) = πn′ 3
3
We need this expression in terms of energy. Substituting n′2 = 8me L2E′∕h2 from
Equation 4.8 in S(n′), we get
πL3 (8meE′) 3∕2
S(E′) =
3h3
Since L3 is the physical volume of the solid, the number of states per unit volume
Sv(E′) with energies E ≤ E′ is
π(8meE′) 3∕2
Sv (E′) = [4.9]
3h3
Furthermore, from Equation 4.7, dSv∕dE = g(E). By differentiating Equation 4.9
with respect to energy, we get

g(E) = (8π21∕2 ) ( 2 ) E1∕2

Density of me 3∕2
states [4.10]
h
 4.5 DENSITY OF STATES IN AN ENERGY BAND 339

Equation 4.10 shows that the density of states g(E) increases with energy as E1∕2
from the bottom of the band. As we approach the top of the band, according to our
understanding in Figure 4.23c, g(E) should decrease with energy as (Etop − E)1∕2,
where Etop is the top of the band, so that as E → Etop, g(E) → 0. The electron mass
me in Equation 4.10 should be the effective mass m*e as in Equation 4.6. Further,
Equation 4.10 strictly applies only to free electrons in a crystal. However, we will
frequently use it to approximate the true g(E) versus E behavior near the band edges
for both metals and semiconductors.
Having found the distribution of the electron energy states, Equation 4.10, we
now wish to determine the number of states that actually contain electrons; that is,
the probability of finding an electron at an energy level E. This is given by the
Fermi–Dirac statistics.
As an example, one convenient way of calculating the population of a city is to
find the density of houses in that city (i.e., the number of houses per unit area),
multiply that by the probability of finding a human in a house, and finally, integrate
the result over the area of the city. The problem is working out the chances of actu-
ally finding someone at home, using a mathematical formula. For those who like
analogies, if g(A) is the density of houses and f (A) is the probability that a house is
occupied, then the population of the city is

n= ∫ City
f (A)g(A) dA

where the integration is done over the entire area of the city. This equation can be
used to find the number of electrons per unit volume within a band. If E is the
electron energy and f (E) is the probability that a state with energy E is occupied,
then

n= ∫Band
f (E)g(E) dE

where the integration is done over all the energies of the band.
 4.5 DENSITY OF STATES IN AN ENERGY BAND 341

DENSITY OF STATES IN A BAND Given that the width of an energy band is typically EXAMPLE 4.8
∼10 eV, calculate the following, in per cm3 and per eV units:
a. The density of states at the center of the band.
b. The number of states per unit volume within a small energy range kT about the center.
c. The density of states at kT above the bottom of the band.
d. The number of states per unit volume within a small energy range of kT to 2kT from the
bottom of the band.

SOLUTION

The density of states, or the number of states per unit energy range per unit volume g(E), is
given by

(h 2)
3∕2
me
g(E) = (8π21∕2 ) E1∕2

which gives the number of states per cubic meter per Joule of energy. Substituting E = 5 eV,
we have

gcenter = (8π21∕2 ) [
(6.626 × 10−34 ) 2 ]
9.1 × 10−31 3∕2
(5 × 1.6 × 10−19 ) 1∕2 = 9.50 × 1046 m−3 J−1

Converting to cm−3 and eV−1, we get

gcenter = (9.50 × 1046 m−3 J−1)(10−6 m3 cm−3)(1.6 × 10−19 J eV−1)

= 1.52 × 1022 cm−3 eV−1

If δE is a small energy range (such as kT), then, by definition, g(E) δE is the number
of states per unit volume in δE. To find the number of states per unit volume within kT at
the center of the band, we multiply gcenter by kT or (1.52 × 1022 cm−3 eV−1)(0.026 eV) to get
3.9 × 1020 cm−3. This is not a small number!
At kT above the bottom of the band, at 300 K (kT = 0.026 eV), we have

g0.026 = (8π21∕2 ) [ −34 2 ]

9.1 × 10−31 3∕2
(0.026 × 1.6 × 10−19 ) 1∕2
(6.626 × 10 )
= 6.84 × 1045 m−3 J−1

Converting to cm−3 and eV−1 we get

g0.026 = (6.84 × 1045 m−3 J−1)(10−6 m3 cm−3)(1.6 × 10−19 J eV−1)

= 1.10 × 1021 cm−3 eV−1

Within kT, the volume density of states is

(1.10 × 1021 cm−3 eV−1)(0.026 eV) = 2.8 × 1019 cm−3

This is very close to the bottom of the band and is still very large.
 4.6 STATISTICS: COLLECTIONS OF PARTICLES 343

4.6 STATISTICS: COLLECTIONS OF PARTICLES

4.6.1 BOLTZMANN CLASSICAL STATISTICS
Given a collection of particles in random motion and colliding with each other,7 we
need to determine the concentration of particles in the energy range E to (E + dE).
Consider the process shown in Figure 4.27, in which two electrons with energies E1
and E2 interact and then move off in different directions, with energies E3 and E4.
Let the probability of an electron having an energy E be P(E), where P(E) is the
fraction of electrons with an energy E. Assume there are no restrictions to the elec-
tron energies, that is, we can ignore the Pauli exclusion principle. The probability
of this event is then P(E1)P(E2). The probability of the reverse process, in which
electrons with energies E3 and E4 interact, is P(E3)P(E4). Since we have thermal
equilibrium, that is, the system is in equilibrium, the forward process must be just
as likely as the reverse process, so
P(E1)P(E2) = P(E3)P(E4) [4.11]

Furthermore, the energy in this collision must be conserved, so we also need

E1 + E2 = E3 + E4 [4.12]

We can show that P(E) = A exp(−βE), where A and β are constants, is a solution
by simply substituting this expression into Equations 4.11 and 4.12. Further, we can
also show that β must be 1∕kT, where k is the Boltzmann constant and T is the
temperature, by comparing the average energy calculated from using P(E) with that
observed in experiments.8

P(E) = A exp(−
kT )
Boltzmann
E
[4.13] probability
function
Equation 4.13 is the Boltzmann probability function and is shown in Figure 4.28.
The probability of finding a particle at an energy E therefore decreases exponentially
with energy. We assume, of course, that any number of particles may have a given
energy E. In other words, there is no restriction such as permitting only one particle
per state at an energy E, as in the Pauli exclusion principle.
Suppose that we have N1 particles at energy level E1 and N2 particles at a higher
energy E2. Then, by Equation 4.13, we have

= exp(−
kT )
N2 E2 − E1 Boltzmann
[4.14]
N1 statistics

If E2 − E1 ≫ kT, then N2 can be orders of magnitude smaller than N1. As the

temperature increases, N2∕Nl also increases. Therefore, increasing the temperature
populates the higher energy levels.
Classical particles obey the Boltzmann statistics. Whenever there are many more
states (by orders of magnitude) than the number of particles, the likelihood of
7
From Chapter 1, we can associate this with the kinetic theory of gases. The energies of the gas molecules,
which are moving around randomly, are distributed according to the Maxwell–Boltzmann statistics.
8
See Question 4.10.
344 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

∝ exp(–E/kT )

E2

ψ4
ψ2 E4
E2 Interaction
E1

ψ1
ψ3 0
E1 N(E)
E3 N2 N1

Figure 4.27 Two electrons with initial Figure 4.28 The Boltzmann
wavefunctions ψ1 and ψ2 at E1 and E2 interact energy distribution describes the
and end up at different energies E3 and E4. statistics of particles, such as
Their corresponding wavefunctions are ψ3 electrons, when there are many
and ψ4. more available states than the
number of particles.

two particles having the same set of quantum numbers is negligible and we do not have
to worry about the Pauli exclusion principle. In these cases, we can use the Boltzmann
statistics. An important example is the statistics of electrons in the conduction band
of a semiconductor where, in general, there are many more states than electrons.

4.6.2 FERMI–DIRAC STATISTICS

Now consider the interaction for which no two electrons can be in the same quantum
state, which is essentially obedience to the Pauli exclusion principle, as shown in
Figure 4.27. We assume that we can have only one electron in a particular quantum
state ψ (including spin) associated with the energy value E. We therefore need those
states that have energies E3 and E4 to be not occupied. Let f(E) be the probability
that an electron is in such a state, with energy E in this new interaction environment.
The forward event in Figure 4.27 requires that we have an electrons at E1 and E2 and,
at the same time, E3 and E4 must be unoccupied (empty). Thus, the probability of
the forward event is given by.
f (E1) f (E2)[1 − f (E3)][1 − f (E4)]
The square brackets represent the probability that the states with energies E3 and E4
are empty. In thermal equilibrium, the reverse process, the electrons with E3 and E4
interacting to transfer to E1 and E2, has just as equal a likelihood as the forward
process. Thus, f (E) must satisfy the equation
f(E1) f(E2)[1 − f (E3)][1 − f (E4)] = f(E3) f (E4)[1 − f (E1)][1 − f (E2)] [4.15]

In addition, for energy conservation, we must have

E1 + E2 = E3 + E4 [4.16]
 4.6 STATISTICS: COLLECTIONS OF PARTICLES 345

Paul Adrien Maurice Dirac (1902–1984)

received the 1933 Nobel prize for
physics with Erwin Schrödinger.
His first degree was in electrical
engineering from Bristol University.
He obtained his PhD in 1926 from
Cambridge University under Ralph
Fowler.
© Pictorial Press Ltd./Alamy Stock
Photo.

By an “intelligent guess,” the solution to Equations 4.15 and 4.16 is

1
f (E) =
1 + A exp(
kT )
[4.17]
E

where A is a constant. You can check that this is a solution by substituting Equation
4.17 into 4.15 and using Equation 4.16. The reason for the term kT in Equation 4.17
is not obvious from Equations 4.15 and 4.16. It appears in Equation 4.17 so that at
sufficiently high energies Equation 4.17 becomes the same as the Boltzmann distri-
bution in Equation 4.13 in agreement with experiments.9 In a more rigorous approach
we would use a constant 1∕β instead of kT in Equation 4.17, and then show that β
must be 1∕kT by comparing the predictions based on Equation 4.17 with experiments.
Letting A = exp(−EF∕kT), we can write Equation 4.17 as
1 Fermi–Dirac
f (E) =
1 + exp(
kT )
[4.18]
E − EF statistics

where EF is a constant called the Fermi energy. The probability of finding an

electron in a state with energy E is given by Equation 4.18, which is called the
Fermi–Dirac function.
The behavior of the Fermi–Dirac function is shown in Figure 4.29. Note
the effect of temperature. As T increases, f (E) extends to higher energies. At

9
If N1 and N2 are the number of electrons at energies E1 and E2, then the Boltzman distribution predicts
Equation (4.14) for N1∕N2. At sufficiently high energies Equation 4.17 gives the same prediction for N1∕N2. The
reason is that at very high energies there are very few electrons compared with the available number of states
at these energies so that it very unlikely that two electrons will try to occupy the same state; that is the Pauli
exclusion principle is not needed and the electron statistics is simply the Boltzmann distribution.
346 CHAPTER 4 ∙ MODERN THEORY OF SOLIDS

T2 > T1

T=0
EF

T1

Figure 4.29 The Fermi–Dirac function f(E) describes the statistics

f(E)
of electrons in a solid. The electrons interact with each other and 0 1 1
the environment, obeying the Pauli exclusion principle. 2

energies of a few kT (0.026 eV) above EF, f (E) behaves almost like the Boltzmann
function

f (E) = exp[ −
kT ]
(E − EF )
(E − EF ) ≫ kT [4.19]

Above absolute zero, at E = EF, f (EF) = 12 . We define the Fermi energy as that
energy for which the probability of occupancy f (EF) equals 12 . The approximation to
f (E) in Equation 4.19 at high energies is often referred to as the Boltzmann tail to
the Fermi–Dirac function. Notice that the spread of f (E) around EF increases with
temperature. This spread around EF is typically several kTs.
412 CHAPTER 5 ∙ SEMICONDUCTORS

5.1 INTRINSIC SEMICONDUCTORS

5.1.1 SILICON CRYSTAL AND ENERGY BAND DIAGRAM
The electronic configuration of an isolated Si atom is [Ne]3s2p2. However, in the
vicinity of other atoms, the 3s and 3p energy levels are so close that the interactions
result in the four orbitals ψ (3s), ψ (3px), ψ (3py), and ψ (3pz) mixing together to form
four new hybrid orbitals (called ψ hyb) that are symmetrically directed as far away
from each other as possible (toward the corners of a tetrahedron). In two dimensions,
we can simply view the orbitals pictorially as in Figure 5.1a. The four hybrid orbit-
als, ψhyb, each have one electron so that they are half-occupied. Therefore, a ψhyb
orbital of one Si atom can overlap a ψhyb orbital of a neighboring Si atom to form
a covalent bond with two spin-paired electrons. In this manner one Si atom bonds
with four other Si atoms by overlapping the half-occupied ψhyb orbitals, as illustrated
in Figure 5.1b. Each Si–Si bond corresponds to a bonding orbital, ψB, obtained by
overlapping two neighboring ψhyb orbitals. Each bonding orbital (ψB) has two spin-
paired electrons and is therefore full. Neighboring Si atoms can also form covalent
bonds with other Si atoms, thus forming a three-dimensional network of Si atoms.
The resulting structure is the Si crystal in which each Si atom bonds with four Si

ψhyb orbitals
Valence
electron

Si ion core (+4e)

(a)

Electron energy

Ec+χ
Conduction band (CB)
Empty of electrons at 0 K.

Ec
Bandgap = Eg
ψB
Ev

Valence band (VB)

Full of electrons at 0 K.

0
(b) (c)

Figure 5.1 (a) A simplified two-dimensional illustration of a Si atom with four hybrid orbitals ψhyb.
Each orbital has one electron. (b) A simplified two-dimensional view of a region of the Si crystal
showing covalent bonds. (c) The energy band diagram at absolute zero of temperature.
 5 .1 INTRINSIC SEMICONDUCTORS 413

Figure 5.2 A two-dimensional pictorial view of the Si

crystal showing covalent bonds as two lines where each
line is a valence electron.

atoms in a tetrahedral arrangement. The crystal structure is that of a diamond, which

was described in Chapter 1. We can imagine the Si crystal in two dimensions as
depicted in Figure 5.1b. The electrons in the covalent bonds are the valence electrons.
The energy band diagram of the silicon crystal is shown in Figure 5.1c.1 The
vertical axis is the electron energy in the crystal. The valence band (VB) contains
those electronic states that correspond to the overlap of bonding orbitals (ψB). Since
all the bonding orbitals (ψB) are full with valence electrons in the crystal, the VB is
also full with these valence electrons at a temperature of absolute zero. The conduc-
tion band (CB) contains electronic states that are at higher energies, those corre-
sponding to the overlap of antibonding orbitals. The CB is separated from the VB
by an energy gap Eg, called the bandgap. The energy level Ev marks the top of the
VB and Ec marks the bottom of the CB. The energy distance from Ec to the vacuum
level, the width of the CB, is called the electron affinity χ. The general energy band
diagram in Figure 5.1c applies to all crystalline semiconductors with appropriate
changes in the energies.
The electrons shown in the VB in Figure 5.1c are those in the covalent bonds
between the Si atoms in Figure 5.1b. An electron in the VB, however, is not local-
ized to an atomic site but extends throughout the whole solid. Although the electrons
appear localized in Figure 5.1b, at the bonding orbitals between the Si atoms this is
not, in fact, true. In the crystal, the electrons can tunnel from one bond to another
and exchange places. If we were to work out the wavefunction of a valence electron
in the Si crystal, we would find that it extends throughout the whole solid. This
means that the electrons in the covalent bonds are indistinguishable. We cannot label
an electron from the start and say that the electron is in the covalent bond between
these two atoms.
We can crudely represent the silicon crystal in two dimensions as shown in
Figure 5.2. Each covalent bond between Si atoms is represented by two lines cor-
responding to two spin-paired electrons. Each line represents a valence electron.

5.1.2 ELECTRONS AND HOLES

The only empty electronic states in the silicon crystal are in the CB (Figure 5.1c).
An electron placed in the CB is free to move around the crystal and also respond to

1
The formation of energy bands in the silicon crystal was described in detail in Chapter 4.
414 CHAPTER 5 ∙ SEMICONDUCTORS

Electron energy

Ec + χ
CB
Ec
hf > Eg Free e– hf
Eg hole
e–
Hole
Ev

VB

0
(a) (b)
Figure 5.3 (a) A photon with an energy greater than Eg can excite an electron from
the VB to the CB. (b) When a photon breaks a Si–Si bond, a free electron and a hole in
the Si–Si bond are created.

an applied electric field because there are plenty of neighboring empty energy levels.
An electron in the CB can easily gain energy from the field and move to higher
energy levels because these states are empty. Generally we can treat an electron in
the CB as if it were free within the crystal with certain modifications to its mass,
as explained later in Section 5.1.3.
Since the only empty states are in the CB, the excitation of an electron from the
VB requires a minimum energy of Eg. Figure 5.3a shows what happens when a
photon of energy hf > Eg is incident on an electron in the VB. This electron absorbs
the incident photon and gains sufficient energy to surmount the energy gap Eg and
reach the CB. Consequently, a free electron and a “hole,” corresponding to a missing
electron in the VB, are created. In some semiconductors such as Si and Ge, the
photon absorption process also involves lattice vibrations (vibrations of the Si atoms),
which we have not shown in Figure 5.3b.
Although in this specific example a photon of energy hf > Eg creates an electron–
hole pair, this is not necessary. In fact, in the absence of radiation, there is an electron–
hole generation process going on in the sample as a result of thermal generation.
Due to thermal energy, the atoms in the crystal are constantly vibrating, which cor-
responds to the bonds between the Si atoms being periodically deformed. In a certain
region, the atoms, at some instant, may be moving in such a way that a bond becomes
overstretched, as pictorially depicted in Figure 5.4. This will result in the over-
stretched bond rupturing and hence releasing an electron into the CB (the electron
effectively becomes “free”). The empty electronic state of the missing electron in
the bond is what we call a hole in the valence band. The free electron, which is in
the CB, can wander around the crystal and contribute to the electrical conduction
when an electric field is applied. The region remaining around the hole in the VB
is positively charged because a charge of −e has been removed from an otherwise
 5 .1 INTRINSIC SEMICONDUCTORS 415

e–

h+

Figure 5.4 Thermal vibrations of atoms can break

bonds and thereby create electron–hole pairs.

h+ CB h+
e–
(a) h+ Eg h+ (d)
VB

h+e– e–
h+
(b) h+ h+ (e)

e–
(c) h+ (f)
h+

Figure 5.5 A pictorial illustration of a hole in the valence band wandering around the crystal due to the tunneling
of electrons from neighboring bonds.

neutral region of the crystal. This hole, denoted as h+, can also wander around the
crystal as if it were free. This is because an electron in a neighboring bond can “jump,”
that is, tunnel, into the hole to fill the vacant electronic state at this site and thereby
create a hole at its original position. This is effectively equivalent to the hole being
displaced in the opposite direction, as illustrated in Figure 5.5a. This single step can
reoccur, causing the hole to be further displaced. As a result, the hole moves around
the crystal as if it were a free positively charged entity, as pictured in Figure 5.5a
416 CHAPTER 5 ∙ SEMICONDUCTORS

to d. Its motion is quite independent from that of the original electron. When an
electric field is applied, the hole will drift in the direction of the field and hence
contribute to electrical conduction. It is now apparent that there are essentially two
types of charge carriers in semiconductors: electrons and holes. A hole is effectively
an empty electronic state in the VB that behaves as if it were a positively charged
“particle” free to respond to an applied electric field.
When a wandering electron in the CB meets a hole in the VB, the electron has
found an empty state of lower energy and therefore occupies the hole. The electron
falls from the CB to the VB to fill the hole, as depicted in Figure 5.5e and f. This
is called recombination and results in the annihilation of an electron in the CB and
a hole in the VB. The excess energy of the electron falling from CB to VB in certain
semiconductors such as GaAs and InP is emitted as a photon. In Si and Ge the excess
energy is lost as lattice vibrations (heat).
It must be emphasized that the illustrations in Figure 5.5 are pedagogical pictorial
visualizations of hole motion based on classical notions and cannot be taken too seri-
ously, as discussed in more advanced texts (see also Section 5.13). We should remem-
ber that the electron has a wavefunction in the crystal that is extended and not localized,
as the pictures in Figure 5.5 imply. Further, the hole is a concept that corresponds to
an empty valence band wavefunction that normally has an electron. Again, we cannot
localize the hole to a particular site, as the pictures in Figure 5.5 imply.

5.1.3 CONDUCTION IN SEMICONDUCTORS

When an electric field is applied across a semiconductor as shown in Figure 5.6, the
energy bands bend. The total electron energy E is KE + PE, but now there is an

Figure 5.6 When an electric field is applied, Ex

electrons in the CB and holes in the VB can Electrostatic PE(x)
drift and contribute to the conductivity. V(x)
(a) A simplified illustration of drift in Ex.
(b) Applied field bends the energy bands
since the electrostatic PE of the electron is
−eV (x) and V (x) decreases in the direction
of Ex, whereas PE increases. Ex
CB
Electron energy

CB

VB
Hole energy

VB

x=0 x=L
(a) (b)
 5 .1 INTRINSIC SEMICONDUCTORS 417

additional electrostatic PE contribution that is not constant in an applied electric

field. A uniform electric field Ex implies a linearly decreasing potential V(x), by
virtue of (dV∕dx) = −Ex, that is, V = −Ax + B. This means that the PE, −eV(x), of
the electron is now eAx − eB, which increases linearly across the sample. All the
energy levels and hence the energy bands must therefore tilt up in the x direction,
as shown in Figure 5.6, in the presence of an applied field.
Under the action of Ex, the electron in the CB moves to the left and immediately
starts gaining energy from the field. When the electron collides with a thermal vibra-
tion of a Si atom, it loses some of this energy and thus “falls” down in energy in
the CB. After the collision, the electron starts to accelerate again, until the next
collision, and so on. We recognize this process as the drift of the electron in an
applied field, as illustrated in Figure 5.6. The drift velocity vde of the electron is μeEx
where μe is the drift mobility of the electron. In a similar fashion, the holes in the
VB also drift in an applied field, but here the drift is along the field. Notice that
when a hole gains energy, it moves “down” in the VB because the potential energy
of the hole is of opposite sign to that of the electron.
Since both electrons and holes contribute to electrical conduction, we may write
the current density J, from its definition, as
J = envde + epvdh [5.1]

where n is the electron concentration in the CB, p is the hole concentration in the
VB, and vde and vdh are the drift velocities of electrons and holes in response to an
applied electric field Ex. Thus,
Electron and
vde = μeEx and vdh = μhEx [5.2] hole drift
velocities
where μe and μh are the electron and hole drift mobilities. In Chapter 2, we derived
the drift mobility μe of the electrons in a conductor as
Drift mobility
eτe
μe = [5.3] and scattering
me time
where τe is the mean free time between scattering events and me is the electronic
mass. The ideas on electron motion in metals can also be applied to the electron
motion in the CB of a semiconductor to rederive Equation 5.3. We must, however,
use an effective mass m*e for the electron in the crystal rather than the mass me in
free space. A “free” electron in a crystal is not entirely free because as it moves it
interacts with the potential energy (PE) of the ions in the solid and therefore expe-
riences various internal forces. The effective mass m*e accounts for these internal
forces in such a way that we can relate the acceleration a of the electron in the CB
to an external force Fext (e.g., −eEx) by Fext = m*e a just as we do for the electron in
vacuum by Fext = mea. In applying the Fext = m*e a type of description to the motion
of the electron, we are assuming, of course, that the effective mass of the electron
can be calculated or measured experimentally. It is important to remark that the true
behavior is governed by the solution of the Schrödinger equation in a periodic lattice
(crystal) from which it can be shown that we can indeed describe the inertial resistance
of the electron to acceleration in terms of an effective mass m*e . The effective mass
depends on the interaction of the electron with its environment within the crystal.
418 CHAPTER 5 ∙ SEMICONDUCTORS

We can now speculate on whether the hole can also have a mass. As long as we
view mass as resistance to acceleration, that is, inertia, there is no reason why the
hole should not have a mass. Accelerating the hole means accelerating electrons
tunneling from bond to bond in the opposite direction. Therefore, it is apparent that
the hole will have a nonzero finite inertial mass because otherwise the smallest
external force will impart an infinite acceleration to it. If we represent the effective
mass of the hole in the VB by m*h , then the hole drift mobility will be
eτh
μh = [5.4]
m*h
where τh is the mean free time between scattering events for holes.
Taking Equation 5.1 for the current density further, we can write the conductivity
Conductivity
of a semiconductor as
of a semi- σ = enμe + epμh [5.5]
conductor
where n and p are the electron and hole concentrations in the CB and VB, respec-
tively. This is a general equation valid for all semiconductors.

5.1.4 ELECTRON AND HOLE CONCENTRATIONS

The general equation for the conductivity of a semiconductor, Equation 5.5, depends
on n the electron concentration, and p, the hole concentration. How do we determine
these quantities? We follow the procedure schematically shown in Figure 5.7a to d
in which the density of states is multiplied by the probability of a state being occu-
pied and integrated over the entire CB for n and over the entire VB for p.
We define gcb(E ) as the density of states in the CB, that is, the number of states
per unit energy per unit volume. The probability of finding an electron in a state
with energy E is given by the Fermi–Dirac function f (E ), which is discussed in
Chapter 4. Then gcb(E )f(E ) is the actual number of electrons per unit energy per
unit volume nE(E ) in the CB. Thus,
nE dE = gcb(E) f (E)dE
is the number of electrons in the energy range E to E + dE. Integrating this from
the bottom (Ec) to the top of the CB gives the electron concentration n, number of
electrons per unit volume, in the CB. In other words,
Top of CB Top of CB
n= ∫Ec
nE (E)dE = ∫Ec
gcb (E) f (E)dE

We will assume that (Ec − FF) ≫ kT (i.e., EF is at least a few kT below Ec)
Boltzmann so that
tail of
f (E) ≈ exp[−(E − EF)∕kT]
Fermi–Dirac
distribution We are thus replacing Fermi–Dirac statistics by Boltzmann statistics and thereby
inherently assuming that the number of electrons in the CB is far less than the num-
ber of states in this band.
 5 .1 INTRINSIC SEMICONDUCTORS 419

E E E
[1 – f(E)]
gcb(E) ∝ (E–Ec)1/2
CB
f(E) for Area = ∫nE(E)dE = n
electrons
nE(E)
Ec Ec

EF EF

Ev Ev pE(E)
Area = p
For holes
VB

0
g(E) f(E) or [1 – f(E)] nE(E) or pE(E)
(a) (b) (c) (d)
Figure 5.7 (a) Energy band diagram. (b) Density of states (number of states per unit energy per unit
volume). (c) Fermi–Dirac probability function (probability of occupancy of a state). (d) The product of
g(E) and f (E) is the energy density of electrons in the CB (number of electrons per unit energy per unit
volume). The area under nE(E) versus E is the electron concentration.

Further, we will take the upper limit to be E = ∞ since f (E ) decays rapidly with
energy so that gcb(E)f(E) → 0 near the top of the band. Furthermore, since gcb(E)f(E)
is significant only close to Ec, we can use
Density of
(π8 √2)m*e 3∕2 1∕2 states in
gcb (E) = (E − Ec )
h3 conduction
band
for an electron in a three-dimensional PE well without having to consider the exact
form of gcb(E) across the whole band. Thus

(E − Ec ) 1∕2 exp[ −
kT ]
∞
(π8 √2)m*e 3∕2 (E − EF )
n≈
h3 ∫Ec
dE

which leads to
Electron
n = Nc exp[ − ]
(Ec − EF ) concentration
[5.6]
kT in CB
where
Effective
Nc = 2( )
3∕2
2πm*e kT density of
[5.7]
h2 states at CB
edge
The result of the integration in Equation 5.6 seems to be simple, but it is an
approximation as it assumes that (Ec − EF) ≫ kT. Nc is a constant, that is, independent
420 CHAPTER 5 ∙ SEMICONDUCTORS

of the Fermi energy, and is called the effective density of states at the CB edge.
Notice that Nc depends on the effective mass2 and has a small temperature depen-
dence as apparent from Equation 5.7. Equation 5.6 can be interpreted as follows. If
we take all the states in the conduction band and replace them with an effective
concentration Nc (number of states per unit volume) at Ec and then multiply this
simply by the Boltzmann probability function, f (Ec) = exp[−(Ec − EF)∕kT ], we
obtain the concentration of electrons at Ec, that is, in the conduction band. Nc is thus
an effective density of states at the CB band edge.
We can carry out a similar analysis for the concentration of holes in the VB.
Multiplying the density of states gvb(E) in the VB with the probability of occupancy
by a hole [1 − f (E)], that is, the probability that an electron is absent, gives pE,
the hole concentration per unit energy. Integrating this over the VB gives the hole
concentration
Ev Ev
p= ∫0
pE dE = ∫0
gvb (E)[(1 − f (E)]dE

With the assumption that EF is a few kT above Ev, the integration simplifies to

p = Nv exp[ − ]
Hole
(EF − Ev )
concentration [5.8]
in VB kT
where Nv is the effective density of states at the VB edge and is given by
Effective

Nv = 2( )
density of 2πm*h kT 3∕2
[5.9]
states at VB h2
edge
We can now see the virtues of studying the density of states g(E) as a function
of energy E and the Fermi–Dirac function f(E). Both were central factors in deriv-
ing the expressions for n and p. There are no specific assumptions in our derivations,
except for EF being a few kT away from the band edges, which means that Equations
5.6 and 5.8 are generally valid.
The general equations that determine the free electron and hole concentrations
are thus given by Equations 5.6 and 5.8. It is interesting to consider the product np,

np = Nc exp[ − ]Nv exp[ − ] = Nc Nv exp[ − ]

(Ec − EF ) (EF − Ev ) (Ec − Ev )
kT kT kT
or

np = Nc Nv exp(−
kT )
Eg
[5.10]

where Eg = Ec − Ev is the bandgap energy. First, we note that this is a general

expression in which the right-hand side, Nc Nv exp(−Eg∕kT ), behaves as if it were
a constant for a given material at a given temperature; it depends on the bandgap
Eg but not on the position of the Fermi level. In the special case of an intrinsic

2
The effective mass in Equation 5.7 is called the density of states effective mass, and is not the same as that
used in describing the electron drift mobility.
 5 .1 INTRINSIC SEMICONDUCTORS 421

semiconductor, n = p, which we can denote as ni, the intrinsic concentration, so

that Nc Nv exp(−Eg∕kT ) must be n2i. From Equation 5.10 we therefore have

np = n2i = Nc Nv exp(−
kT )
Eg Mass action
[5.11a]
law
This is a general equation that is valid as long as we have thermal equilibrium.
External excitation, such as photogeneration, is excluded. It states that the product
np is constant at a given temperature and depends only on the semiconductor mate-
rial. Equation 5.11a is called the mass action law. If we somehow increase the
electron concentration, then we inevitably reduce the hole concentration. The con-
stant ni has a special significance because it represents the free electron and hole
concentrations in the intrinsic material. From Equation 5.11a,

ni = (Nc Nv ) 1∕2 exp(−

2kT )
Eg Intrinsic
[5.11b]
concentration
An intrinsic semiconductor is a pure semiconductor crystal in which the elec-
tron and hole concentrations are equal. By pure we mean virtually no impurities in
the crystal. We should also exclude crystal defects that may capture carriers of one
sign and thus result in unequal electron and hole concentrations. Clearly in a pure
semiconductor, electrons and holes are generated in pairs by thermal excitation
across the bandgap. It must be emphasized that Equation 5.11b is generally valid
and therefore applies to both intrinsic and nonintrinsic (n ≠ p) semiconductors.
When an electron and hole meet in the crystal, they “recombine.” The electron
falls in energy and occupies the empty electronic state that the hole represents.
Consequently, the broken bond is “repaired,” but we lose two free charge carriers.
Recombination of an electron and hole results in their annihilation. In a semicon-
ductor we therefore have thermal generation of electron–hole pairs by thermal exci-
tation from the VB to the CB, and we also have recombination of electron–hole pairs
that removes them from their conduction and valence bands, respectively. The rate
of recombination R will be proportional to the number of electrons and also to the
number of holes. Thus
R ∝ np
The rate of generation G will depend on how many electrons are available for
excitation at Ev, that is, Nv; how many empty states are available at Ec, that is, Nc; and
the probability that the electron will make the transition, that is, exp(−Eg∕kT), so that

G ∝ Nc Nv exp(−
kT )
Eg

Since in thermal equilibrium we have no continuous increase in n or p, we must

have the rate of generation equal to the rate of recombination, that is, G = R. This
is equivalent to Equation 5.11a.
In sketching the diagrams in Figure 5.7a to d to illustrate the derivation of the
expressions for n and p (in Equations 5.6 and 5.8), we assumed that the Fermi level EF
is somewhere around the middle of the energy bandgap. This was not an assumption
422 CHAPTER 5 ∙ SEMICONDUCTORS

in the mathematical derivations but only in the sketches. From Equations 5.6 and
5.8, we also note that the position of Fermi level is important in determining the
electron and hole concentrations. It serves as a “mathematical crank” to determine
n and p.
We first consider an intrinsic semiconductor, n = p = ni. Setting p = ni in Equa-
tion 5.8, we can solve for the Fermi energy in the intrinsic semiconductor, EFi, that is,

Nv exp[ − ] = (Nc Nv ) 1∕2 exp(−

2kT )
(EFi − Ev ) Eg
kT
which leads to

EFi = Ev + Eg − kT ln( )
Fermi energy
1 1 Nc
in intrinsic [5.12]
semiconductor 2 2 Nv
Furthermore, substituting the proper expressions for Nc and Nv we get

Eg − kT ln( )
Fermi energy
1 3 m*e
in intrinsic EFi = Ev + [5.13]
semiconductor 2 4 m*h
It is apparent from these equations that if Nc = Nv or m*e = m*h , then
1
EFi = Ev + Eg
2
that is, EFi is right in the middle of the energy gap. Normally, however, the effective
masses will not be equal and the Fermi level will be slightly shifted down from
midgap by an amount 34 kT ln(m*e∕m*h ), which is quite small compared with 12 Eg . For
Si and Ge, the hole effective mass (for density of states) is slightly smaller than the
electron effective mass, so EFi is slightly below the midgap.
The condition np = n2i means that if we can somehow increase the electron
concentration in the CB over the intrinsic value—for example, by adding impurities
into the Si crystal that donate additional electrons to the CB—we will then have
n > p. The semiconductor is then called n-type. The Fermi level must be closer to
Ec than Ev, so that
Ec − EF < EF − Ev
and Equations 5.6 and 5.8 yield n > p. The np product always yields n2i in thermal
equilibrium in the absence of external excitation, for example, illumination.
It is also possible to have an excess of holes in the VB over electrons in the CB,
for example, by adding impurities that remove electrons from the VB and thereby
generate holes. In that case EF is closer to Ev than to Ec. A semiconductor in which
p > n is called a p-type semiconductor. The general band diagrams with the appro-
priate Fermi levels for intrinsic, n-type, and p-type semiconductors (e.g., i-Si, n-Si,
and p-Si, respectively) are illustrated in Figure 5.8a to c.
It is apparent that if we know where EF is, then we have effectively determined
n and p by virtue of Equations 5.6 and 5.8. We can view EF as a material property
that is related to the concentration of charge carriers that contribute to electrical
conduction. Its significance, however, goes beyond n and p. It also determines the
 5 .1 INTRINSIC SEMICONDUCTORS 423

χ CB
Φ Φn
Ec Ec Ec Φp
EFn
EFi
EFp
Ev Ev Ev

VB

(a) (b) (c)

Figure 5.8 Energy band diagrams for (a) intrinsic, (b) n-type, and (c) p-type semiconductors.
In all cases, np = ni2.

energy needed to remove an electron from the semiconductor. The energy difference
between the vacuum level (where the electron is free) and EF is the work function
Φ of the semiconductor, the energy required to remove an electron even though there
are no electrons at EF in a semiconductor.
The Fermi level can also be interpreted in terms of the potential energy per
electron for electrical work similar to the interpretation of electrostatic PE. Just as
eΔV is the electrical work involved in taking a charge e across a potential difference
ΔV, any difference in EF in going from one end of a material (or system) to another
is available to do an amount ΔEF of external work. A corollary to this is that if
electrical work is done on the material, for example, by passing a current through
it, then the Fermi level is not uniform in the material. ΔEF then represents the work
done per electron. For a material in thermal equilibrium and not subject to any
external excitation such as illumination or connections to a voltage supply, the Fermi
level in the material must therefore be uniform, ΔEF = 0.
What is the average energy of an electron in the conduction band of a semicon-
ductor? Also, what is the mean speed of an electron in the conduction band? We
note that the concentration of electrons with energies E to E + dE is nE(E) dE or
gcb(E)f (E) dE. Thus the average energy of electrons in the CB, by definition of the
mean, is

ECB =
1
n ∫
CB
Egcb (E) f (E) dE

where the integration must be over the CB. Substituting the proper expressions for
gcb(E) and f(E) in the integrand and carrying out the integration from Ec to the top
of the band, we find the very simple result that
Average
3
ECB = Ec + kT [5.14] electron
2 energy in CB
Thus, an electron in the conduction band has an average energy of 32 kT above Ec.
Since we know that an electron at Ec is “free” in the crystal, 32 kT must be its average
kinetic energy.
424 CHAPTER 5 ∙ SEMICONDUCTORS

Table 5.1 Selected typical properties of Ge, Si, InP, and GaAs at 300 K

Eg χ Nc Nv ni μe μh
(eV) (eV) (cm−3) (cm−3) (cm−3) (cm2 V−1 s−1) (cm2 V−1 s−1) m*e ∕me m*h∕me εr
19 18 13
Ge 0.66 4.13 1.04 × 10 6.0 × 10 2.3 × 10 3900 1900 0.12a 0.23a 16
0.56b 0.40b
Si 1.10 4.01 2.8 × 1019 1.2 × 1019 1.0 × 1010 1400 450 0.26a 0.38a 11.9
1.08b 0.60b
InP 1.34 4.50 5.2 × 1017 1.1 × 1019 1.3 × 107 4600 190 0.079a,b 0.46a 12.6
0.58b
GaAs 1.42 4.07 4.4 × 1017 7.7 × 1018 2.1 × 106 8800 400 0.067a,b 0.40a 13.0
0.50b

NOTE: Ge and Si are indirect whereas InP and GaAs are direct bandgap semiconductors. Effective mass related to conductivity
(labeled a) is different than that for density of states (labeled b). In numerous textbooks, ni is taken as 1.45 × 1010 cm−3 and is therefore
the most widely used value of ni for Si, though the correct value is actually 1.0 × 1010 cm−3. (Green, M.A., Journal of Applied Physics, 67,
2944, 1990.) (Data combined from various sources.)

This is just like the average kinetic energy of gas atoms (such as He atoms) in
a tank assuming that the atoms (or the “particles”) do not interact, that is, they are
independent. We know from the kinetic theory that the statistics of a collection of
independent gas atoms obeys the classical Maxwell–Boltzmann description with an
average energy given by 32 kT . We should also recall that the description of electron
statistics in a metal involves the Fermi–Dirac function, which is based on the Pauli
exclusion principle. In a metal the average energy of the conduction electron is 35 EF
and, for all practical purposes, temperature independent. We see that the collective
electron behavior is completely different in the two solids. We can explain the dif-
ference by noting that the conduction band in a semiconductor is only scarcely
populated by electrons, which means that there are many more electronic states than
electrons and thus the likelihood of two electrons trying to occupy the same elec-
tronic state is practically nil. We can then neglect the Pauli exclusion principle and
use the Boltzmann statistics. This is not the case for metals where the number of
conduction electrons and the number of states are comparable in magnitude.
Table 5.1 compares some of the properties of the important semiconductors, Ge,
Si, InP, and GaAs.

EXAMPLE 5.1 INTRINSIC CONCENTRATION AND CONDUCTIVITY OF Si Given that the density of
states related effective masses of electrons and holes in Si are approximately 1.08me and
0.60me, respectively, and the electron and hole drift mobilities at room temperature are 1400
and 450 cm2 V−1 s−1, respectively, calculate the intrinsic concentration and intrinsic resistivity
of Si.
SOLUTION

We simply calculate the effective density of states Nc and Nv by

Nc = 2( ) Nv = 2( )
2πm*e kT 3∕2
2πm*h kT 3∕2

2
and 2
h h
 5 .1 INTRINSIC SEMICONDUCTORS 425

Thus

Nc = 2[ ]
2π(1.08 × 9.1 × 10−31 kg) (1.38 × 10−23 J K−1 ) (300 K) 3∕2

−34 2
(6.63 × 10 J s)
= 2.81 × 1025 m−3 or 2.81 × 1019 cm−3
and

Nv = 2[ ]
2π(0.60 × 9.1 × 10−31 kg) (1.38 × 10−23 J K−1 ) (300 K) 3∕2

(6.63 × 10−34 J s) 2
25 −3
= 1.16 × 10 m or 1.16 × 1019 cm−3
The intrinsic concentration is

ni = (Nc Nv ) 1∕2 exp(−

2kT )
Eg

so that

ni = [ (2.81 × 1019 cm−3 ) (1.16 × 1019 cm−3 ) ] 1∕2 exp[ −

2(300 K) (8.62 × 10−5 eV K−1 ) ]
(1.10 eV)

10 −3
= 1.0 × 10 cm
The conductivity is
σ = enμe + epμh = eni(μe + μh)
that is,
σ = (1.6 × 10−19 C)(1.0 × 1010 cm−3)(1400 + 450 cm2 V−1 s−1)
= 3.0 × 10−6 Ω−1 cm−1
The resistivity is
1
ρ= = 3.3 × 105 Ω cm
σ
Although we calculated ni = 1.0 × 1010 cm−3, the most widely used ni value in the literature
has been 1.45 × 1010 cm−3. The difference arises from a number of factors but, most impor-
tantly, from what exact value of the effective hole mass should be used in calculating Nv.
Henceforth we will simply use3 ni = 1.0 × 1010 cm−3, which seems to be the “true” value.
426 CHAPTER 5 ∙ SEMICONDUCTORS

5.2 EXTRINSIC SEMICONDUCTORS

By introducing small amounts of impurities into an otherwise pure Si crystal, it is
possible to obtain a semiconductor in which the concentration of carriers of one
polarity is much in excess of the other type. Such semiconductors are referred to as
extrinsic semiconductors vis-à-vis the intrinsic case of a pure and perfect crystal.
For example, by adding pentavalent impurities, such as arsenic, which have a valency
of more than four, we can obtain a semiconductor in which the electron concentra-
tion is much larger than the hole concentration. In this case we will have an n-type
semiconductor. If we add trivalent impurities, such as boron, which have a valency
of less than four, then we find that we have an excess of holes over electrons. We
now have a p-type semiconductor. How do impurities change the concentrations of
holes and electrons in a semiconductor?
 5.2 EXTRINSIC SEMICONDUCTORS 427

5.2.1 n-TYPE DOPING

Consider what happens when small amounts of a pentavalent (valency of 5) element
from Group V in the Periodic Table, such as As, P, Sb, are introduced into a pure
Si crystal. We only add small amounts (e.g., one impurity atom for every million
host atoms) because we wish to surround each impurity atom by millions of Si atoms,
thereby forcing the impurity atoms to bond with Si atoms in the same diamond
crystal structure. Arsenic has five valence electrons, whereas Si has four. Thus when
an As atom bonds with four Si atoms, it has one electron left unbonded. It cannot
find a bond to go into, so it is left orbiting around the As atom, as illustrated in
Figure 5.9. The As+ ionic center with an electron e− orbiting it is just like a hydro-
gen atom in a silicon environment. We can easily calculate how much energy is
required to free this electron away from the As site, thereby ionizing the As impurity.
Had this been a hydrogen atom in free space, the energy required to remove the
electron from its ground state (at n = 1) to far away from the positive center would
have been given by −En with n = 1. The binding energy of the electron in the H
atom is thus
mee4
Eb = −E1 = = 13.6 eV
8ε2oh2
If we wish to apply this to the electron around an As+ core in the Si crystal
environment, we must use εr εo instead of εo, where εr is the relative permittivity of
silicon, and also the effective mass of the electron m*e in the silicon crystal. Thus,
the binding energy of the electron to the As+ site in the Si crystal is
Electron
= 2 2 2 = (13.6 eV) ( )( 2 )
m*e e4 m*e 1
E Si
b [5.15] binding
8εr εoh me εr energy at a
With εr = 11.9 and m*e ≈ 13 me for silicon, we find EbSi = 0.032 eV, which is donor
comparable with the average thermal energy of atomic vibrations at room temperature,
∼3kT (∼0.07 eV). Thus, the fifth valence electron can be readily freed by thermal
vibrations of the Si lattice. The electron will then be “free” in the semiconductor,

Figure 5.9 Arsenic-doped Si crystal.

The four valence electrons of As allow it to bond just
like Si, but the fifth electron is left orbiting the As site.
The energy required to release the free fifth electron
into the CB is very small.

As+
e–
428 CHAPTER 5 ∙ SEMICONDUCTORS

Electron energy
CB

Ec
~0.03 eV E
d
As+ As+ As+ As+

Figure 5.10 Energy band diagram for an

n-type Si doped with 1 ppm As. Ev x Distance into
There are donor energy levels just below Ec crystal
around As+ sites. As atom sites every 106 Si atoms

or, in other words, it will be in the CB. The energy required to excite the electron
to the CB is therefore 0.032 eV. The addition of As atoms introduces localized
electronic states at the As sites because the fifth electron has a localized wavefunction,
of the hydrogenic type, around As+. The energy Ed of these states is 0.032 eV below
Ec because this is how much energy is required to take the electron away into the
CB. Thermal excitation by the lattice vibrations at room temperature is sufficient to
ionize the As atom, that is, excite the electron from Ed into the CB. This process
creates free electrons but immobile As+ ions, as shown in the energy band diagram
of an n-type semiconductor in Figure 5.10. Because the As atom donates an electron
into the CB, it is called a donor atom. Ed is the electron energy around the donor
atom. Ed is close to Ec, so the spare fifth electron from the dopant can be readily
donated to the CB. If Nd is the donor atom concentration in the crystal, then provided
that Nd ≫ ni, at room temperature the electron concentration in the CB will be nearly
equal to Nd, that is n ≈ Nd. The hole concentration will be p = n2i∕Nd, which is less
than the intrinsic concentration because a few of the large number of electrons in
the CB recombine with holes in the VB so as to maintain np = n2i. The conductivity
will then be

σ = eNd μe + e( )μh ≈ eNd μe

n-type n2i
[5.16]
conductivity Nd
At low temperatures, however, not all the donors will be ionized and we need
to know the probability, denoted as fd (Ed), of finding an electron in a state with
energy Ed at a donor. This probability function is similar to the Fermi–Dirac function
f (Ed) except that it has a factor of 12 multiplying the exponential term,
Occupation 1
fd (Ed ) =
1 + exp[ ]
[5.17]
probability at 1 (Ed − EF )
a donor 2 kT
The factor 12 is due to the fact that the electron state at the donor can take an
electron with spin either up or down but not both4 (once the donor has been occupied,

4
The proof can be found in advanced solid-state physics texts.
 5.2 EXTRINSIC SEMICONDUCTORS 429

a second electron cannot enter this site). Thus, the concentration of ionized donors
at a temperature T is given by
Nd+ = Nd × (probability of not finding an electron at Ed )
= Nd [1 − fd (Ed )]
Nd Ionized donor
=
1 + 2 exp[ ]
[5.18]
(EF − Ed ) concentration
kT

5.2.2 p-TYPE DOPING

We saw that introducing a pentavalent atom into a Si crystal results in n-type doping
because the fifth electron cannot go into a bond and escapes from the donor into the
CB by thermal excitation. By similar arguments, we should anticipate that doping a
Si crystal with a trivalent atom (valency of 3) such as B, Al, Ga, or In will result
in a p-type Si crystal. We consider doping Si with small amounts of B as shown in
Figure 5.11a. Because B has only three valence electrons, when it shares them with
four neighboring Si atoms, one of the bonds has a missing electron, which of course
is a hole. A nearby electron can tunnel into this hole and displace the hole further
away from the boron atom. As the hole moves away, it gets attracted by the negative
charge left behind on the boron atom and therefore takes an orbit around the B− ion,
as shown in Figure 5.11b. The binding energy of this hole to the B− ion can be
calculated using the hydrogenic atom analogy as in the n-type Si case. This binding
energy turns out to be very small, ∼0.05 eV, so at room temperature the thermal
vibrations of the lattice can free the hole away from the B− site. A free hole, we

h+
B– B–

Free

(a) (b)
Figure 5.11 Boron-doped Si crystal.
B has only three valence electrons. When it substitutes for a Si atom, one of its
bonds has an electron missing and therefore a hole, as shown in (a). The hole orbits
around the B− site by the tunneling of electrons from neighboring bonds, as shown
in (b). Eventually, thermally vibrating Si atoms provide enough energy to free the
hole from the B− site into the VB, as shown.
430 CHAPTER 5 ∙ SEMICONDUCTORS

Electron energy
B atom sites every 106 Si atoms

Ec x Distance
into crystal

Figure 5.12 Energy band diagram

B– B– B– B–
for a p-type Si doped with 1 ppm B. Ea
There are acceptor energy levels Ea ~0.05 eV
h+
just above Ev around B− sites. These Ev
acceptor levels accept electrons
VB
from the VB and therefore create
holes in the VB.

Table 5.2 Examples of donor and acceptor ionization energies (eV) in Si

Donors Acceptors

P As Sb B Al Ga

0.045 0.054 0.039 0.045 0.057 0.072

recall, exists in the VB. The escape of the hole from the B− site involves the B atom
accepting an electron from a neighboring Si–Si bond (from the VB), which effectively
results in the hole being displaced away and its eventual escape to freedom in the VB.
The B atom introduced into the Si crystal therefore acts as an electron acceptor and,
because of this, it is called an acceptor impurity. The electron accepted by the B atom
comes from a nearby bond. On the energy band diagram, an electron leaves the VB
and gets accepted by a B atom, which becomes negatively charged. This process leaves
a hole in the VB that is free to wander away, as illustrated in Figure 5.12.
It is apparent that doping a silicon crystal with a trivalent impurity results in a
p-type material. We have many more holes than electrons for electrical conduction
since the negatively charged B atoms are immobile and hence cannot contribute to
the conductivity. If the concentration of acceptor impurities Na in the crystal is much
greater than the intrinsic concentration ni, then at room temperature all the acceptors
would have been ionized and thus p ≈ Na. The electron concentration is then deter-
mined by the mass action law, n = n2i∕Na, which is much smaller than p, and con-
sequently the conductivity is simply given by σ = eNaμh.
Typical ionization energies for donor and acceptor atoms in the silicon crystal
are summarized in Table 5.2.

5.2.3 COMPENSATION DOPING

What happens when a semiconductor contains both donors and acceptors? Compen-
sation doping is a term used to describe the doping of a semiconductor with both
 5.2 EXTRINSIC SEMICONDUCTORS 431

donors and acceptors to control the properties. For example, a p-type semiconductor
doped with Na acceptors can be converted to an n-type semiconductor by simply
adding donors until the concentration Nd exceeds Na. The effect of donors compen-
sates for the effect of acceptors and vice versa. The electron concentration is then
given by Nd − Na provided the latter is larger than ni. When both acceptors and
donors are present, what essentially happens is that electrons from donors recombine
with the holes from the acceptors so that the mass action law np = n2i is obeyed.
Remember that we cannot simultaneously increase the electron and hole concentra-
tions because that leads to an increase in the recombination rate that returns the
electron and hole concentrations to satisfy np = n2i. When an acceptor atom accepts
a valence band electron, a hole is created in the VB. This hole then recombines with
an electron from the CB. Suppose that we have more donors than acceptors. If we
take the initial electron concentration as n = Nd, then the recombination between
the electrons from the donors and Na holes generated by Na acceptors results in the
electron concentration reduced by Na to n = Nd − Na. By a similar argument, if we
have more acceptors than donors, the hole concentration becomes p = Na − Nd, with
electrons from Nd donors recombining with holes from Na acceptors. Thus there are
two compensation effects:
n2i
1. More donors: Nd − Na ≫ ni n = (Nd − Na) and p=
(Nd − N a ) Compensation
n2i doping
2. More acceptors: Na − Nd ≫ ni p = (Na − Nd) and n=
(Na − Nd )
These arguments assume that the temperature is sufficiently high for donors and
acceptors to have been ionized. This will be the case at room temperature. At low
temperatures, we have to consider donor and acceptor statistics and the charge neu-
trality of the whole crystal, as in Example 5.9.

RESISTIVITY OF INTRINSIC AND DOPED Si Find the resistance of a 1 cm3 pure silicon EXAMPLE 5.4
crystal. What is the resistance when the crystal is doped with arsenic if the doping is 1 in 109,
that is, 1 part per billion (ppb) (note that this doping corresponds to one foreigner living in
China)? Note that the atomic concentration in silicon is 5 × 1022 cm−3, ni = 1.0 × 1010 cm−3,
μe = 1400 cm2 V−1 s−1, and μh = 450 cm2 V−1 s−1.
SOLUTION

For the intrinsic case, we apply

σ = enμe + epμh = en(μe + μh)
−19
so σ = (1.6 × 10 C)(1.0 × 1010 cm−3)(1400 + 450 cm2 V−1 s−1)
= 2.96 × 10−6 Ω−1 cm−1
Since L = 1 cm and A = 1 cm2, the resistance is
L 1
R= = = 3.47 × 105 Ω or 347 kΩ
σA σ
432 CHAPTER 5 ∙ SEMICONDUCTORS

When the crystal is doped with 1 in 109, then

NSi 5 × 1022
Nd = 9
= = 5 × 1013 cm−3
10 109
At room temperature all the donors are ionized, so
n = Nd = 5 × 1013 cm−3
The hole concentration is
n2i (1.0 × 1010 ) 2
p= = = 2.0 × 106 cm−3 ≪ ni
Nd (5 × 1013 )
Therefore,
σ = enμe = (1.6 × 10−19 C)(5 × 1013 cm−3)(1400 cm2 V−1 s−1)
= 1.12 × 10−2 Ω−1 cm−1
L 1
Further, R= = = 89.3 Ω
σA σ
Notice the drastic fall in the resistance when the crystal is doped with only 1 in 109 atoms.
Doping the silicon crystal with boron instead of arsenic, but still in amounts of 1 in 109,
means that Na = 5 × 1013 cm−3, which results in a conductivity of
σ = epμh = (1.6 × 10−19 C)(5 × 1013 cm−3)(450 cm2 V−1 s−1)
= 3.6 × 10−3 Ω−1 cm−1
L 1
Therefore, R= = = 278 Ω
σA σ
The reason for a higher resistance with p-type doping compared with the same amount of
n-type doping is that μh < μe.

EXAMPLE 5.5 COMPENSATION DOPING An n-type Si semiconductor containing 1016 phosphorus (donor)
atoms cm−3 has been doped with 1017 boron (acceptor) atoms cm−3. Calculate the electron
and hole concentrations in this semiconductor.
SOLUTION

This semiconductor has been compensation doped with excess acceptors over donors, so
Na − Nd = 1017 − 1016 = 9 × 1016 cm−3
This is much larger than the intrinsic concentration ni = 1.0 × 1010 cm−3 at room tempera-
ture, so
p = Na − Nd = 9 × 1016 cm−3
The electron concentration
n2i (1.0 × 1010 cm−3 ) 2
n= = = 1.1 × 103 cm−3
p (9 × 1016 cm−3 )
Clearly, the electron concentration and hence its contribution to electrical conduction is
completely negligible compared with the hole concentration. Thus, by excessive boron doping,
the n-type semiconductor has been converted to a p-type semiconductor.
 5.2 EXTRINSIC SEMICONDUCTORS 433

THE FERMI LEVEL IN n- AND p-TYPE Si An n-type Si wafer has been doped uniformly EXAMPLE 5.6
with 1016 antimony (Sb) atoms cm−3. Calculate the position of the Fermi energy with respect
to the Fermi energy EFi in intrinsic Si. The above n-type Si sample is further doped with
2 × 1017 boron atoms cm−3. Calculate the position of the Fermi energy with respect to the
Fermi energy EFi in intrinsic Si. (Assume that T = 300 K, and kT = 0.0259 eV.)

SOLUTION

Sb gives n-type doping with Nd = 1016 cm−3, and since Nd ≫ ni (=1.0 × 1010 cm−3), we have
n = Nd = 1016 cm−3
For intrinsic Si,

ni = Nc exp[ − ]
(Ec − EFi )
kT
whereas for doped Si,

n = Nc exp[ − ] = Nd
(Ec − EFn )
kT
where EFi and EFn are the Fermi energies in the intrinsic and n-type Si. Dividing the two
expressions,

= exp[ ]
Nd (EFn − EFi )
ni kT
so that

EFn − EFi = kT ln( ) = (0.0259 eV) ln(

1.0 × 1010 )
Nd 1016
= 0.36 eV
ni
When the wafer is further doped with boron, the acceptor concentration is
Na = 2 × 1017 cm−3 > Nd = 1016 cm−3
The semiconductor is compensation doped and compensation converts the semiconductor to
p-type Si. Thus
p = Na − Nd = (2 × 1017 − 1016) = 1.9 × 1017 cm−3
For intrinsic Si,

ni = Nv exp[ − ]
(EFi − Ev )
kT
whereas for doped Si,

p = Nv exp[ − ] = Na − Nd
(EFp − Ev )
kT
where EFi and EFp are the Fermi energies in the intrinsic and p-type Si, respectively. Dividing
the two expressions,

= exp[ − ]
p (EFp − EFi )
ni kT
434 CHAPTER 5 ∙ SEMICONDUCTORS

so that

EFp − EFi = −kT ln( ) = −(0.0259 eV) ln(

1.0 × 1010 )
p 1.9 × 1017
ni
= −0.43 eV
 5.1 Carrier Drift 157

5.1 | CARRIER DRIFT

An electric field applied to a semiconductor will produce a force on electrons and
holes so that they will experience a net acceleration and net movement, provided
there are available energy states in the conduction and valence bands. This net move-
ment of charge due to an electric field is called drift. The net drift of charge gives
rise to a drift current.

5.1.1 Drift Current Density

If we have a positive volume charge density  moving at an average drift velocity vd,
the drift current density is given by
Jdrf  vd (5.1a)
In terms of units, we have

Jdrf  _
Coul  _
cm3 
 cm
s   cm2  s  cm2
Coul
__ _A (5.1b)

If the volume charge density is due to positively charged holes, then

Jpdrf  (ep)vdp (5.2)
where Jpdrf is the drift current density due to holes and vdp is the average drift velocity
of the holes.
The equation of motion of a positively charged hole in the presence of an electric
field is
F  m*cp a  eE (5.3)
where e is the magnitude of the electronic charge, a is the acceleration, E is the
electric field, and m*cp is the conductivity effective mass of the hole.1 If the electric
field is constant, then we expect the velocity to increase linearly with time. However,
charged particles in a semiconductor are involved in collisions with ionized impu-
rity atoms and with thermally vibrating lattice atoms. These collisions, or scattering
events, alter the velocity characteristics of the particle.
As the hole accelerates in a crystal due to the electric field, the velocity increases.
When the charged particle collides with an atom in the crystal, for example, the particle
loses most, or all, of its energy. The particle will again begin to accelerate and gain
energy until it is again involved in a scattering process. This continues over and over
again. Throughout this process, the particle will gain an average drift velocity which,
for low electric fields, is directly proportional to the electric field. We may then write
vdp  p E (5.4)
where p is the proportionality factor and is called the hole mobility. The mobility is
an important parameter of the semiconductor since it describes how well a particle

1
The conductivity effective mass is used when carriers are in motion. See Appendix F for further
discussion of effective mass concepts.

nea29583_ch05_156-191.indd 157 12/11/10 10:15 AM

 158 CHAPTER 5 Carrier Transport Phenomena

Table 5.1 | Typical mobility values at T  300 K and low doping

concentrations
n (cm2/V-s) ␮p (cm2/V-s)
Silicon 1350 480
Gallium arsenide 8500 400
Germanium 3900 1900

will move due to an electric field. The unit of mobility is usually expressed in terms
of cm2V-s.
By combining Equations (5.2) and (5.4), we may write the drift current density
due to holes as
Jpdrf  (ep)vdp  eppE (5.5)
The drift current due to holes is in the same direction as the applied electric field.
The same discussion of drift applies to electrons. We may write
Jndrf  vdn  (en)vdn (5.6)
where Jndrf is the drift current density due to electrons and vdn is the average drift
velocity of electrons. The net charge density of electrons is negative.
The average drift velocity of an electron is also proportional to the electric field
for small fields. However, since the electron is negatively charged, the net motion of
the electron is opposite to the electric field direction. We can then write
vdn  n E (5.7)
where n is the electron mobility and is a positive quantity. Equation (5.6) may now
be written as
Jndrf  (en)(n E)  en nE (5.8)
The conventional drift current due to electrons is also in the same direction as the
applied electric field even though the electron movement is in the opposite direction.
Electron and hole mobilities are functions of temperature and doping concen-
trations, as we will see in the next section. Table 5.1 shows some typical mobility
values at T  300 K for low doping concentrations.
Since both electrons and holes contribute to the drift current, the total drift current
density is the sum of the individual electron and hole drift current densities, so we may
write

Jdrf  e(n n  p p)E (5.9)

EXAMPLE 5.1 Objective: Calculate the drift current density in a semiconductor for a given electric field.
Consider a gallium arsenide sample at T  300 K with doping concentrations of Na  0
and Nd  1016 cm3. Assume complete ionization and assume electron and hole mobilities given
in Table 5.1. Calculate the drift current density if the applied electric field is E  10 V/cm.

nea29583_ch05_156-191.indd 158 12/11/10 10:15 AM

 5.1 Carrier Drift 159

■ Solution
Since Nd  Na, the semiconductor is n type and the majority carrier electron concentration,
from Chapter 4 is given by
______________

  n
Nd  Na Nd  Na 2
n  __  __ 2
i  1016 cm3
2 2

The minority carrier hole concentration is

n2i (1.8  106)2

n 
p_ __  3.24  104 cm3
1016

For this extrinsic n-type semiconductor, the drift current density is

Jdrf  e(n n  p p)E  en Nd E

Then

Jdrf  (1.6  1019)(8500)(1016)(10)  136 A /cm2

■ Comment
Significant drift current densities can be obtained in a semiconductor applying relatively small
electric fields. We may note from this example that the drift current will usually be due primar-
ily to the majority carrier in an extrinsic semiconductor.

■ EXERCISE PROBLEM
Ex 5.1 A drift current density of Jdrf  75 A/cm2 is required in a device using p-type silicon
when an electric field of E  120 V/cm is applied. Determine the required impurity
doping concentration to achieve this specification. Assume that electron and hole
mobilities given in Table 5.1 apply. (Ans. Na  8.14  1015 cm3)

5.1.2 Mobility Effects

In the previous section, we defined mobility, which relates the average drift velocity
of a carrier to the electric field. Electron and hole mobilities are important semicon-
ductor parameters in the characterization of carrier drift, as seen in Equation (5.9).
Equation (5.3) related the acceleration of a hole to a force such as an electric
field. We may write this equation as
dv  eE
F  m*cp _ (5.10)
dt
where v is the velocity of the particle due to the electric field and does not include
the random thermal velocity. If we assume that the conductivity effective mass and
electric field are constants, then we may integrate Equation (5.10) and obtain
v_ eEt (5.11)
mcp
*

where we have assumed the initial drift velocity to be zero.

Figure 5.1a shows a schematic model of the random thermal velocity and mo-
tion of a hole in a semiconductor with zero electric field. There is a mean time

nea29583_ch05_156-191.indd 159 12/11/10 10:15 AM

 160 CHAPTER 5 Carrier Transport Phenomena

4
1
1 4

3 2
2 3

E field
(a) (b)

Figure 5.1 | Typical random behavior of a hole in a semiconductor (a) without an

electric field and (b) with an electric field.

between collisions which may be denoted by cp. If a small electric field (E-field)
is applied as indicated in Figure 5.1b, there will be a net drift of the hole in the
direction of the E-field, and the net drift velocity will be a small perturbation on
the random thermal velocity, so the time between collisions will not be altered ap-
preciably. If we use the mean time between collisions cp in place of the time t in
Equation (5.11), then the mean peak velocity just prior to a collision or scattering
event is

 
ecp
vdpeak  _ E (5.12a)
mcp
*

The average drift velocity is one half the peak value so that we can write

 
ecp
1 _
vd  _ E (5.12b)
2 mcp
*

However, the collision process is not as simple as this model, but is statistical in
nature. In a more accurate model including the effect of a statistical distribution, the
factor _12 in Equation (5.12b) does not appear. The hole mobility is then given by
vdp ecp
p  _  _ (5.13)
E m*cp
The same analysis applies to electrons; thus, we can write the electron mobility as
ecn
n  _ (5.14)
m*cn
where cn is the mean time between collisions for an electron.
There are two collision or scattering mechanisms that dominate in a semicon-
ductor and affect the carrier mobility: phonon or lattice scattering, and ionized impu-
rity scattering.
The atoms in a semiconductor crystal have a certain amount of thermal energy
at temperatures above absolute zero that causes the atoms to randomly vibrate about
their lattice position within the crystal. The lattice vibrations cause a disruption in
the perfect periodic potential function. A perfect periodic potential in a solid allows

nea29583_ch05_156-191.indd 160 12/11/10 10:15 AM

 5.1 Carrier Drift 161

electrons to move unimpeded, or with no scattering, through the crystal. But the ther-
mal vibrations cause a disruption of the potential function, resulting in an interaction
between the electrons or holes and the vibrating lattice atoms. This lattice scattering
is also referred to as phonon scattering.
Since lattice scattering is related to the thermal motion of atoms, the rate at
which the scattering occurs is a function of temperature. If we denote L as the
mobility that would be observed if only lattice scattering existed, then the scattering
theory states that to first order
L  T 32 (5.15)
Mobility that is due to lattice scattering increases as the temperature decreases. In-
tuitively, we expect the lattice vibrations to decrease as the temperature decreases,
which implies that the probability of a scattering event also decreases, thus increas-
ing mobility.
Figure 5.2 shows the temperature dependence of electron and hole mobilities
in silicon. In lightly doped semiconductors, lattice scattering dominates and the car-
rier mobility decreases with temperature as we have discussed. The temperature de-
pendence of mobility is proportional to T n. The inserts in the figure show that the
parameter n is not equal to _32 as the first-order scattering theory predicted. However,
mobility does increase as the temperature decreases.
The second interaction mechanism affecting carrier mobility is called ionized im-
purity scattering. We have seen that impurity atoms are added to the semiconductor
to control or alter its characteristics. These impurities are ionized at room temperature
so that a coulomb interaction exists between the electrons or holes and the ionized im-
purities. This coulomb interaction produces scattering or collisions and also alters the
velocity characteristics of the charge carrier. If we denote I as the mobility that would
be observed if only ionized impurity scattering existed, then to first order we have
T 32
I  _ (5.16)
NI
 
where NI  Nd  Na is the total ionized impurity concentration in the semiconduc-
tor. If temperature increases, the random thermal velocity of a carrier increases, re-
ducing the time the carrier spends in the vicinity of the ionized impurity center. The
less time spent in the vicinity of a coulomb force, the smaller the scattering effect
and the larger the expected value of I. If the number of ionized impurity centers
increases, then the probability of a carrier encountering an ionized impurity center
increases, implying a smaller value of I.
Figure 5.3 is a plot of electron and hole mobilities in germanium, silicon, and
gallium arsenide at T  300 K as a function of impurity concentration. More ac-
curately, these curves are of mobility versus ionized impurity concentration NI. As
the impurity concentration increases, the number of impurity scattering centers in-
creases, thus reducing mobility.
If L is the mean time between collisions due to lattice scattering, then dtL is the
probability of a lattice scattering event occurring in a differential time dt. Likewise, if I
is the mean time between collisions due to ionized impurity scattering, then dtI is the
probability of an ionized impurity scattering event occurring in the differential time dt.

nea29583_ch05_156-191.indd 161 12/11/10 10:15 AM

 162

nea29583_ch05_156-191.indd 162
5000 1000
4000

2000 NA  1014
␮n T 22 NA  1016
1000
2000
ND  1014 500 N  1014 cm3
D
16 NA  1017
ND  10 100 200 500 1000
T (K)
1000

ND  1017 NA  1018
500 100

␮n (cm2/ V-s)
␮p (cm2/ V-s)
NA  1019
1000
ND  1018
500
T 22
␮p
ND  1019
200
100 NA  1014 cm3

100
100 200 500 1000
T (K)
50 10
50 0 50 100 150 200 50 0 50 100 150 200
T ( C) T ( C)
(a) (b)

Figure 5.2 | (a) Electron and (b) hole mobilities in silicon versus temperature for various doping concentrations. Inserts show
temperature dependence for “almost” intrinsic silicon.
(From Pierret [8].)

12/11/10 10:15 AM
 5.1 Carrier Drift 163

104
␮n T  300 K

Ge
103 ␮p

102
104
Mobility (cm2/V-s)

␮n
103 Si
␮p

102
104
␮n

GaAs
103
␮p

102
1014 1015 1016 1017 1018 1019
Impurity concentration (cm3)

Figure 5.3 | Electron and hole mobilities versus impurity

concentrations for germanium, silicon, and gallium arsenide
at T  300 K.
(From Sze [14].)

If these two scattering processes are independent, then the total probability of a scatter-
ing event occurring in the differential time dt is the sum of the individual events, or
dt  _
_ dt  _
dt (5.17)
 I L
where  is the mean time between any scattering event.
Comparing Equation (5.17) with the definitions of mobility given by Equation (5.13)
or (5.14), we can write

1
1 _ 1
 
_ _ (5.18)
I L

where I is the mobility due to the ionized impurity scattering process and L is the
mobility due to the lattice scattering process. The parameter  is the net mobility.
With two or more independent scattering mechanisms, the inverse mobilities add,
which means that the net mobility decreases.

nea29583_ch05_156-191.indd 163 12/11/10 10:15 AM

 164 CHAPTER 5 Carrier Transport Phenomena

EXAMPLE 5.2 Objective: Determine the electron mobility in silicon at various doping concentrations and
various temperatures.
Using Figure 5.2, find the electron mobility in silicon for:
(a) T  25°C for (i) Nd  1016 cm3 and (ii) Nd  1017 cm3.
(b) Nd  1016 cm3 for (i) T  0°C and (ii) T  100°C.

■ Solution:
From Figure 5.2, we find the following:
(a) T  25°C; (i) Nd  1016 cm3 ⇒ n  1200 cm2/V-s.
(ii) Nd  1017 cm3 ⇒ n  800 cm2/V-s.
(b) Nd  10 cm3; (i) T  0°C ⇒ n  1400 cm2/V-s.
16

(ii) T  100°C ⇒ n  780 cm2/V-s.

■ Comment
The results of this example show that the mobility values are strong functions of the doping
concentration and temperature. These variations must be taken into account in the design of
semiconductor devices.

■ EXERCISE PROBLEM
Ex 5.2 Using Figure 5.2, find the hole mobility in silicon for:
(a) T  25°C for (i) Na  1016 cm3 and (ii) Na  1018 cm3, and
(b) Na  1014 cm3 for (i) T  0°C and (ii) T  100°C.
(b) (i) p  550 cm2/V-s, (ii) p  300 cm2/V-s)]
[(Ans. (a) (i) p  410 cm2/V-s, (ii) p  130 cm2/V-s;

5.1.3 Conductivity
The drift current density, given by Equation (5.9), may be written as
Jdrf  e(n n  p p)E  E (5.19)
where  is the conductivity of the semiconductor material. The conductivity is
given in units of ( -cm)1 and is a function of the electron and hole concentra-
tions and mobilities. We have just seen that the mobilities are functions of impurity
concentrations; conductivity, then is a somewhat complicated function of impurity
concentration.
The reciprocal of conductivity is resistivity, which is denoted by  and is given
in units of ohm-cm. We can write the formula for resistivity as2
1 ___
_ 1
  e(n n  p p) (5.20)

Figure 5.4 is a plot of resistivity as a function of impurity concentration in silicon,

germanium, gallium arsenide, and gallium phosphide at T  300 K. Obviously, the
curves are not linear functions of Nd or Na because of mobility effects.
2
The symbol  is also used for volume charge density. The context in which  is used should make it
clear whether it stands for charge density or resistivity.

nea29583_ch05_156-191.indd 164 12/11/10 10:15 AM

 172 CHAPTER 5 Carrier Transport Phenomena

TEST YOUR UNDERSTANDING

TYU 5.1 Consider a sample of silicon at T  300 K doped at an impurity concentration of
Nd  1015 cm3 and Na  1014 cm3. Assume electron and hole mobilities given
in Table 5.1. Calculate the drift current density if the applied electric field is
E  35 V/cm. (Ans. 6.80 A/cm2)
TYU 5.2 Silicon at T  300 K is doped with impurity concentrations of Nd  5  1016 cm3
and Na  2  1016 cm3. (a) What are the electron and hole mobilities? (b) Deter-
mine the conductivity and resistivity of the material.
-cm] [Ans. (a) n  1000 cm2/V-s, p  350 cm2/V-s; (b)   4.8 ( -cm)1,   0.208

TYU 5.3 For a particular silicon semiconductor device at T  300 K, the required
material is n type with a resistivity of 0.10 -cm. (a) Determine the re-
quired impurity doping concentration and (b) the resulting electron mobility.
[Ans. (a) From Figure 5.4, Nd  9  10 16 cm3; (b) n  695 cm2/V-s]

5.2 | CARRIER DIFFUSION

There is a second mechanism, in addition to drift, that can induce a current in a
semiconductor. We may consider a classic physics example in which a container, as
shown in Figure 5.9, is divided into two parts by a membrane. The left side contains
gas molecules at a particular temperature and the right side is initially empty. The
gas molecules are in continual random thermal motion so that, when the membrane
is broken, the gas molecules flow into the right side of the container. Diffusion is the
process whereby particles flow from a region of high concentration toward a region
of low concentration. If the gas molecules were electrically charged, the net flow of
charge would result in a diffusion current.

5.2.1 Diffusion Current Density

To begin to understand the diffusion process in a semiconductor, we will consider a sim-
plified analysis. Assume that an electron concentration varies in one dimension as shown
in Figure 5.10. The temperature is assumed to be uniform so that the average thermal
velocity of electrons is independent of x. To calculate the current, we will determine the
net flow of electrons per unit time per unit area crossing the plane at x  0. If the distance
l shown in Figure 5.10 is less than the mean-free path of an electron, that is, the average

x0

Figure 5.9 | Container

divided by a membrane with
gas molecules on one side.

nea29583_ch05_156-191.indd 172 12/11/10 10:15 AM

 5.2 Carrier Diffusion 173

n(x)
n(l)

n(0)

n(l)

x  l x0 x  l x

Figure 5.10 | Electron concentration versus distance.

distance an electron travels between collisions (l vth cn), then on the average, electrons
moving to the right at x  l and electrons moving to the left at x  l will cross the
x  0 plane. One half of the electrons at x  l will be traveling to the right at any instant
of time and one half of the electrons at x  l will be traveling to the left at any given
time. The net rate of electron flow, Fn, in the x direction at x  0 is given by
1 n(l)v  _
Fn  _ 1 n(l )v  _
1 v [n(l )  n(l )] (5.29)
th th
2 2 2 th
If we expand the electron concentration in a Taylor series about x  0 keeping
only the first two terms, then we can write Equation (5.29) as
1v
Fn  _ dn  n(0)  l _
n(0)  l _ dn (5.30)
2 th dx dx
which becomes
dn
Fn  vth l _ (5.31)
dx
Each electron has a charge (e), so the current is
dn
J  eFn  evth l _ (5.32)
dx
The current described by Equation (5.32) is the electron diffusion current and is pro-
portional to the spatial derivative, or density gradient, of the electron concentration.
The diffusion of electrons from a region of high concentration to a region of low
concentration produces a flux of electrons flowing in the negative x direction for this
example. Since electrons have a negative charge, the conventional current direction
is in the positive x direction. Figure 5.11a shows these one-dimensional flux and

nea29583_ch05_156-191.indd 173 12/11/10 10:15 AM

 174 CHAPTER 5 Carrier Transport Phenomena

Electron concentration, n
Electron flux

Electron diffusion
current density

x
(a)

Hole concentration, p
Hole flux

Hole diffusion
current density

x
(b)

Figure 5.11 | (a) Diffusion of electrons due to a density

gradient. (b) Diffusion of holes due to a density gradient.

current directions. We may write the electron diffusion current density for this one-
dimensional case, in the form

dn
Jnxdif  eDn _ (5.33)
dx

where Dn is called the electron diffusion coefficient, has units of cm2/s, and is a posi-
tive quantity. If the electron density gradient becomes negative, the electron diffu-
sion current density will be in the negative x direction.
Figure 5.11b shows an example of a hole concentration as a function of distance
in a semiconductor. The diffusion of holes, from a region of high concentration to
a region of low concentration, produces a flux of holes in the negative x direction.
Since holes are positively charged particles, the conventional diffusion current den-
sity is also in the negative x direction. The hole diffusion current density is propor-
tional to the hole density gradient and to the electronic charge, so we may write

dp
Jpxdif  eDp _ (5.34)
dx

for the one-dimensional case. The parameter Dp is called the hole diffusion coef-
ficient, has units of cm2/s, and is a positive quantity. If the hole density gradient be-
comes negative, the hole diffusion current density will be in the positive x direction.

nea29583_ch05_156-191.indd 174 12/11/10 10:15 AM

 5.2 Carrier Diffusion 175

Objective: Calculate the diffusion current density given a density gradient. EXAMPLE 5.5
Assume that, in an n-type gallium arsenide semiconductor at T  300 K, the electron con-
centration varies linearly from 1  1018 to 7  1017 cm3 over a distance of 0.10 cm. Calculate
the diffusion current density if the electron diffusion coefficient is Dn  225 cm2/s.

■ Solution
The diffusion current density is given by
dn  eD _
Jndif  eDn _ n
n
dx x
 
1  1018  7  1017  108 A /cm2
 (1.6  1019)(225) ____
0. 10

■ Comment
A significant diffusion current density can be generated in a semiconductor material with only
a modest density gradient.

■ EXERCISE PROBLEM
Ex 5.5 The hole density in silicon is given by p(x)  1016 e(xLp) (x  0) where Lp  2  104 cm.
Assume the hole diffusion coefficient is Dp  8 cm2/s. Determine the hole dif-
fusion current density at (a) x  0, (b) x  2  104 cm, and (c) x  103 cm.
[Ans. (a) Jp  64 A/cm 2; (b) Jp  23.54 A/cm 2; (c) Jp  0.431 A/cm2]

5.2.2 Total Current Density

We now have four possible independent current mechanisms in a semiconductor.
These components are electron drift and diffusion currents and hole drift and diffu-
sion currents. The total current density is the sum of these four components, or, for
the one-dimensional case,
dn  eD _dp
J  enn Ex  epp Ex  eDn _ p (5.35)
dx dx
This equation may be generalized to three dimensions as
J  enn E  epp E  eDnn  eDpp (5.36)
The electron mobility gives an indication of how well an electron moves in a
semiconductor as a result of the force of an electric field. The electron diffusion
coefficient gives an indication of how well an electron moves in a semiconductor as
a result of a density gradient. The electron mobility and diffusion coefficient are not
independent parameters. Similarly, the hole mobility and diffusion coefficient are
not independent parameters. The relationship between mobility and the diffusion
coefficient is developed in the next section.
The expression for the total current in a semiconductor contains four terms. For-
tunately in most situations, we will only need to consider one term at any one time at
a particular point in a semiconductor.

nea29583_ch05_156-191.indd 175 12/11/10 10:15 AM

 180 CHAPTER 5 Carrier Transport Phenomena

*5.4 | THE HALL EFFECT

The Hall effect is a consequence of the forces that are exerted on moving charges by
electric and magnetic fields. The Hall effect is used to distinguish whether a semicon-
ductor is n type or p type3 and to measure the majority carrier concentration and majority
carrier mobility. The Hall effect device, as discussed in this section, is used to experi-
mentally measure semiconductor parameters. However, it is also used extensively in
engineering applications as a magnetic probe and in other circuit applications.
The force on a particle having a charge q and moving in a magnetic field is given by
F  qv  B (5.48)
where the cross product is taken between velocity and magnetic field so that the force
vector is perpendicular to both the velocity and magnetic field.
Figure 5.13 illustrates the Hall effect. A semiconductor with a current Ix is placed
in a magnetic field perpendicular to the current. In this case, the magnetic field is in
the z direction. Electrons and holes flowing in the semiconductor will experience a
force as indicated in the figure. The force on both electrons and holes is in the (y)
direction. In a p-type semiconductor ( p0  n0), there will be a buildup of positive
charge on the y  0 surface of the semiconductor and, in an n-type semiconductor
(n0  p0), there will be a buildup of negative charge on the y  0 surface. This net
charge induces an electric field in the y direction as shown in the figure. In steady

Bz


VH
 d
EH
e
h

EH W
z y

x
L

Ix
 
Vx

Figure 5.13 | Geometry for measuring the Hall effect.

*Indicates sections that will aid in the total summation of understanding of semiconductor devices, but
may be skipped the first time through the text without loss of continuity.
3
We will assume an extrinsic semiconductor material in which the majority carrier concentration is
much larger than the minority carrier concentration.

nea29583_ch05_156-191.indd 180 12/11/10 10:15 AM

 5.4 The Hall Effect 181

state, the magnetic field force will be exactly balanced by the induced electric field
force. This balance may be written as
F  q[E  v  B]  0 (5.49a)
which becomes
qEy  qvx Bz (5.49b)
The induced electric field in the y direction is called the Hall field. The Hall field
produces a voltage across the semiconductor which is called the Hall voltage. We
can write
VH  EH W (5.50)
where EH is assumed positive in the y direction and VH is positive with the polarity
shown.
In a p-type semiconductor, in which holes are the majority carrier, the Hall volt-
age will be positive as defined in Figure 5.13. In an n-type semiconductor, in which
electrons are the majority carrier, the Hall voltage will have the opposite polarity.
The polarity of the Hall voltage is used to determine whether an extrinsic semicon-
ductor is n type or p type.
Substituting Equation (5.50) into Equation (5.49) gives
VH  vx WBz (5.51)
For a p-type semiconductor, the drift velocity of holes can be written as
Jx Ix
ep  (ep)(Wd )
vdx  _ __ (5.52)

where e is the magnitude of the electronic charge. Combining Equations (5.52) and
(5.50), we have
Ix Bz
VH  _ (5.53)
epd
or, solving for the hole concentration, we obtain
Ix Bz
p_ (5.54)
edVH
The majority carrier hole concentration is determined from the current, magnetic
field, and Hall voltage.
For an n-type semiconductor, the Hall voltage is given by
Ix Bz
VH  _ (5.55)
ned
so that the electron concentration is
Ix Bz
n  _ (5.56)
edVH
Note that the Hall voltage is negative for the n-type semiconductor; therefore, the
electron concentration determined from Equation (5.56) is actually a positive
quantity.

nea29583_ch05_156-191.indd 181 12/11/10 10:15 AM

 182 CHAPTER 5 Carrier Transport Phenomena

Once the majority carrier concentration has been determined, we can calculate
the low-field majority carrier mobility. For a p-type semiconductor, we can write
Jx  epp Ex (5.57)
The current density and electric field can be converted to current and voltage so that
Equation (5.57) becomes
Ix eppVx
_  __ (5.58)
Wd L
The hole mobility is then given by
Ix L
p  __ (5.59)
epVx Wd
Similarly for an n-type semiconductor, the low-field electron mobility is determined
from
Ix L
n  __ (5.60)
enVx Wd

EXAMPLE 5.8 Objective: Determine the majority carrier concentration and mobility, given Hall effect
parameters.
Consider the geometry shown in Figure 5.13. Let L  101 cm, W  102 cm, and
d  103 cm. Also assume that Ix  1.0 mA, Vx  12.5 V, Bz  500 gauss  5  102 tesla,
and VH  6.25 mV.

■ Solution
A negative Hall voltage for this geometry implies that we have an n-type semiconductor.
Using Equation (5.56), we can calculate the electron concentration as
(103)(5  102)
n  ______  5  1021 m3  5  1015 cm3
(1.6  1019)(105)(6.25  103)
The electron mobility is then determined from Equation (5.60) as
(103)(103)
n  _______  0.10 m2/V-s
(1.6  10 )(5  1021)(12.5)(104)(105)
19

or

n  1000 cm2/V-s

■ Comment
It is important to note that the MKS units must be used consistently in the Hall effect equations
to yield correct results.

■ EXERCISE PROBLEM
Ex 5.8 A p-type silicon sample with the geometry shown in Figure 5.13 has parameters
L  0.2 cm, W  102 cm, and d  8  104 cm. The semiconductor parameters
are p  1016 cm3 and p  320 cm2/V-s. For Vx  10 V and Bz  500 gauss 
5  102 tesla, determine Ix and VH .(Ans. Ix  0.2048 mA, VH  0.80 mV)

nea29583_ch05_156-191.indd 182 12/11/10 10:15 AM