Semiconductor Physics

Physics of Semiconductors
Dr. M Ramegowda
Associate Professor of Physics
Govt. College (Autonomous)
Mandya - 571401, INDIA
web:https://sites.google.com/site/hmrgowdagcm
email: [email protected]
2 M Ramegowda: Physics of Semiconductors
https://sites.google.com/site/hmrgowdagcm
Contents
1 Introduction 5
1.1 Impurity semiconductors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Carrier concentrations in impurity semiconductors . . . . . . . . . . . . . . . . . 6
1.3 Impurity concentration in the case of extrinsic semiconductors . . . . . . . . . . 10
1.4 Fermi energy in impurity semiconductors . . . . . . . . . . . . . . . . . . . . . . 13
1.5 Drift of electrons in an electric field . . . . . . . . . . . . . . . . . . . . . . . . . 15
1.6 Matthiessen’s rule . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
1.7 Electrical conductivity of impurity semiconductor. . . . . . . . . . . . . . . . . . 18
1.8 Hall effect in intrinsic and impurity semiconductors . . . . . . . . . . . . . . . . 21
1.8.1 Magnetoresistance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
1.9 The mobility of the charge carriers . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.10 Cyclotron Resonance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
3
Chapter 1
Introduction
Semiconductors can be classified into two groups: Elemental Semiconductors and Compound
Semiconductors.
Semiconductors belongs to 14 group of the periodic table, germanium and silicon are referred
to as elemental semiconductors. Germanium has a band gap of 0.72 eV, electron and hole
mobility of 3600 and 1700 centimeter square per volt second respectively. Silicon has band gap of
1.1 eV, electron and hole mobility of 1500 and 500 centimeter square per volt second respectively.
Even though the electron and hole mobility of silicon is lower than that of germanium, the higher
band gap make the silicon devices to operate at 150-200 ◦ C, where as germanium devices will
stop working when temperature reach about 70-80 ◦ C.
Compound Semiconductors are mainly classified into three categories:
1. Binary Compounds: These are the compounds of two elements. It can be from any of
the two groups from the periodic table.
(a) 13-15 semiconductors: Crystallizing with high degree of stoichiometry, most can be
obtained as both n-type and p-type. Many have high carrier mobilities and direct
energy gaps, making them useful for optoelectronics. e.g. GaP, InP, GaAS, etc.
(b) 12-16 semiconductors: These semiconductor materials can crystallise in either the
cubic zinc-blende (sphalerite) (beta) phase or the hexagonal wurzite (alpha) phase,
but, for each material, one or other of those phases is thermodynamically more stable
at 300K. e.g., ZnO, ZnS, ZnSe, ZnTe, CdS, CdSe, CdTe HgS, HgSe, HgTe, etc.
(c) 14-14 semiconductors: In these semiconductors both elements are the 14 group of
periodic table. e.g., SiC, which having energy gap 2.9-3.05 eV and best suited for
high power, high temperature devices.
2. Ternary Compounds: These are the compounds of three elements fromed by 13 and 15
groups of the periodic table. e.g., AlGaAs, InGaAs, etc.
3. Quaternary Compounds: These are the compounds of four elements fromed by 13 and
15 groups of the periodic table. e.g., InAlGaN, InGaAs, etc.
5
A semiconductor material can also be classified into two types on the basis of its energy band
diagram in the energy-momentum space. These two categories of semiconductor materials are
known as the direct and the indirect band-gap semiconductor materials. The figure below shows
a schematic representation of the energy-band diagrams of these two types of semiconductors.
A direct band-gap (DBG) semiconductor
is one in which the maximum energy level of
the valence band aligns with the minimum en-
ergy level of the conduction band with respect
to momentum as shown in Fig. 1.1. On the
other hand, if the two levels are misaligned
with respect to momentum, the semiconduc-
tor is called as indirect band-gap (IBG) semi-
conductor. In a DBG semiconductor, a direct
recombination takes place with the release of
the energy equal to the energy difference be-
Figure 1.1: Direct and indirect band-gap
tween the recombining particles. But in case
of a IBG semiconductor, due to a relative dif-
ference in the momentum, first, the momentum is conserved by release of energy in the form of
phonons. Then recombination occurs accompanied with the release of energy. The probability
of a radiative recombination, hence, in case of IBG semiconductor is much less in comparison
to that in case of DBG semiconductors. Hence, the efficiency factor of a DBG semiconductor is
much more than that of a IBG semiconductor. So, DBG semiconductors are always preferred
over IBG for making optical sources.
Silicon and Germanium are both IBG semiconductors and, hence cannot be used to manufacture
optical sources. The compounds which are formed with 13 group elements are direct band gap.
So most of the indium compounds are direct band gap. Gallium compounds except gallium
phosphide are direct band gap. Aluminum compounds, all of them are indirect band gap
1.1 Impurity semiconductors

The statistical picture of n-type and p-type semiconductors is characterized by the presence of
the Fermi level above (for n-type) or below (for p-type) the position associated with the intrinsic
crystal. In an n-type crystal, for example, there cannot be more electrons than holes unless the
Fermi level is adjusted upward from the intrinsic position, and vice-versa for p-type, as shown
in Fig.1.2. As the temperature and impurity concentration are varied, the position of the Fermi
level changes in a complex way.
1.2 Carrier concentrations in impurity semiconductors

The concentration of electrons and holes in pure semiconductor is given by
Ec −Ef
no = Uc e− kT (1.1)
M Ramegowda: Physics of Semiconductors 7
Figure 1.2: Distribution function, Fermi level, and electron and hole populations for (a) an n-type impurity
semiconductor and (b) a p-type impurity semiconductor.
where
3/2 3/2
2πm∗n kT m∗n

Uc = 2 15
= 4.83 × 10 T cm−3 (1.2)
h2 mo
and
Ef −Ev
po = Uv e− kT (1.3)
where
3/2
2πm∗p kT

Uv = 2 (1.4)
h2
m∗n is the effective mass for electrons in the conduction band, and m∗p is the effective mass
for holes in the valence band. The product no po can easily be shown to be a function only of
the energy gap ∆E, the effective masses, and the temperature, independent of the Fermi level.
Thus, multiplying the expressions 1.1 and 1.3, we may obtain
Ec −Ev ∆E
no po = Uc Uv e− kT = Uc Uv e− kT (1.5)
3
2πkT ∆E
no po = 4 2
(m∗p m∗n )3/2 e− kT (1.6)
h
For a given semiconducting substance, the effective masses and the energy gap ∆E are fixed;
hence the product no po in a given material must be a function only of temperature. This is
essentially a mass-action law governing the relative concentrations of holes and electrons in a
given material. If the semiconductor is in the pure or intrinsic state, then the concentrations of
holes and electrons must be equal, then
po = no = ni (T )
where ni (T ) stands simply for the number of holes or electrons per unit volume in an intrinsic
sample of the semiconductor in question at temperature T . We may then write (using Eqs. 1.5
and 1.6)
3/2
p ∆E
− 2kT 2πkT ∆E
ni (T ) = Uc Uv e =2 2
(m∗p m∗n )3/4 e− 2kT (1.7)
h
According to these results, the concentration of carriers in an intrinsic semiconductor is a strong

function of temperature, increasing rapidly as the temperature rises, and is likewise strongly
dependent upon the energy gap ∆E, decreasing rapidly as ∆E increases.
Consider an impurity semiconductor with the number of electrons per unit volume nd occupying
the donor levels (thus the concentration of unionized donors) Ed is given by
Nd
nd = Ed −Ef (1.8)
1 + 12 e kT
where Nd is the concentration of donor impurity atoms.

In a generally similar way one may show that if Na is the concentration of acceptor impurity
atoms and Ea the energy of the acceptor levels, then the concentration of holes associated with
acceptor atoms (thus the density of unionized acceptor sites) pa , will be
Na
pa = Ef −Ea (1.9)
1 + 12 e kT
The density of states associated with the donor and acceptor levels may be represented in terms
of Dirac δ functions by
gd (E) = 2Nd δ(E − Ed ) (1.10)

ga (E) = 2Na δ(E − Ea ) (1.11)
These equations express the fact that there are no donor or acceptor levels at energies other
than Ed and Ea , and that the total number of donor and acceptor quantum states per unit
volume is 2Nd and 2Na , respectively. If the Boltzmann approximation is valid for both the
conduction band and the donor levels (that is, if the Fermi level is several kT units below Ed ),
the exponential factor in the denominator of the expression in ?? is much greater than unity,
and hence the latter may be neglected; Equation ?? may then be written as
Ed −Ef
nd = 2Nd e− kT (1.12)
Since the concentration of electrons in the conduction band is given by

Ec −Ef
no = Uc e− kT (refer Eq.1.1)
we may write the ratio of nd , the number of electrons associated with unionized donors, to
no + nd , the total number of free and loosely bound electrons as
Ed −Ef
nd 2Nd e− kT 1
= E −E Ec −E = Ec −Ed (1.13)
nd + no 2Nd e − dkT f
+ Uc e − kT f 1+ Uc −
e kT
2Nd
3/2
2πm∗n kT

where Uc = 2
h2
Ec −Ed is the donor ionization energy, which is ordinarily smaller than kT ; the exponential factor
above is therefore of the order of unity. If Nd << Uc /2 the ratio of the number of electrons on
unionized donors to the total number will be very small. Since at 300 K Uc is of the order of
1019 cm−3 , this condition is satisfied at that temperature for all donor impurity concentrations
which are small compared to this value. In such instances, it is clear that the donors are almost
completely ionized, and it is usually convenient and accurate to proceed on the assumption that
their ionization is complete, setting Nd equal to zero. At very low temperatures the exponential
factor in equation 1.13 may be appreciably smaller than unity, and the criterion for complete
donor ionization must be written in the more general form
1 Ec −Ed
Nd << Uc e− kT (1.14)
2
In a similar fashion, provided the Boltzmann approximation for holes is valid both in the valence
band and at the acceptor levels, one may show that the acceptor levels are practically completely
ionized whenever
1 Ea −Ev
Na << Uv e− kT (1.15)
2
2πm∗p kT 3/2

where Uv = 2
h2
1
Unless the temperature is very low, this amounts to the requirement Na << Uc . In most
2
circumstances, the Fermi level will be within the forbidden gap, many kT units away from
both acceptor and donor levels, and both acceptor and donor impurities which may be present
will be almost completely ionized. In a semiconductor at equilibrium there must be either a
thermal hole or a positively charged donor ion for every free electron, and a thermal electron or
a negatively charged acceptor ion for every free hole. The entire crystal must thus be electrically
neutral. This electrical neutrality condition can be expressed by equating the algebraic sum of
all negative and positive charges to zero, whereby [recalling that the concentration of ionized
donors is (Nd − nd ) and of ionized acceptors (Na − pa )],
po − no + Nd − nd − (Na − pa ) = 0 (1.16)
For all but the lowest temperatures and the highest values of impurity concentration, the Boltz-
mann approximation will be valid for both conduction and valence bands and for both donor
and acceptor levels, whereby the concentrations of unionized donors and acceptors, pa and nd ,
may be neglected in 1.16, giving
po − no + Nd − Na = 0 (1.17)
On substituting for po = n2i /no to the equation 1.17,
n2o − (Nd − Na )no − n2i = 0 (1.18)
On solving for no by using equation 1.18, we get

1
q
2
no = (Nd − Na ) + (Nd − Na )2 + 4ni (1.19)
2
Similarly, on substituting for no = n2i /po to the equation 1.17, we get

1
q
2 2
po = (Na − Nd ) + (Na − Nd ) + 4ni (1.20)
2
In case of intrinsic semiconductor Nd − Na = 0, no = po = ni . In n-type semiconductor
Nd − Na > 0, no > po and in p-type semiconductor Nd − Na < 0, po > no .
For strongly extrinsic materials, Nd − Na >> ni (for n-type) the equations 1.19 and 1.20
gives
∼ n2i
no = Nd − Na and po ∼
= (1.21)
Nd − Na
and Na − Nd >> ni (for p-type)
n2i
po ∼
= Na − Nd and no ∼
= (1.22)
Na − Nd
Equations 1.21 and 1.22 gives the carrier concentrations of the strongly extrinsic materials.
1.3 Impurity concentration in the case of extrinsic semi-

conductors
For all but the lowest temperatures and the highest values of impurity concentration, the Boltz-
mann approximation will be valid for both conduction and valence bands and for both donor
and acceptor levels, whereby the concentrations of unionized donors and acceptors, pa and nd ,
then the electrical nutrality condition in exrinsic semiconductor can be written as
po − no + Nd − Na = 0 (1.23)
E −Ev Ec −E
− fkT − kT f
Uv e − Uc e + Nd − Na = 0 (using Eqs.1.1 and 1.3) (1.24)
3/2 3/2
2πm∗p kT 2πm∗n kT

where Uv = 2 and Uc = 2
h2 h2
If we let
Ef Ec Ev
α = e kT , βc = e− kT and βv = e kT (1.25)
the equation 1.24 becomes

βv
Uv − Uc βc α + Nd − Na = 0
α
(Nd − Na ) Uv βv
α2 − α− = 0 (1.26)
Uc βc U c βc
On solving the equation 1.26,
" 2 # 21
Ef Nd − Na Nd − Na Uv βv
α=e kT = ± + (1.27)
2Uc βc 2Uc βc Uc βc
Since when Nd = Na = 0, eEf /kT must be a positive quantity. Therefore we must choose the
positive sign of the radical in equation 1.27. On taking logarithm, equation 1.27 can be written
as
 " 2 # 12 
Ef  Nd − Na Nd − Na Uv βv 
= ln + + (1.28)
kT  2Uc βc 2Uc βc Uc βc 
Right hand side of the equation 1.28 is of the type

√ a
ln(a + a2 + x2 ) = ln x + sinh−1 (1.29)
x
Then the equation 1.28, can be written as
s
Ef U v βv −1 Nd − Na
= ln + sinh √ (1.30)
kT U c βc 2 Uc Uv βc βv
On substituting for Uv , Uc , βc and βv , the above equation becomes
∗ 34
mp

1 −1 Nd − Na
Ef = (Ec + Ev ) + kT ln + kT sinh √ ∆E (1.31)
2 m∗n 2 Uc Uv e− 2kT
By using equations 1.7, we can get

−1 Nd − Na
Ef = Ef i + kT sinh (1.32)
2ni
where
34
m∗p

1
Ef i = (Ec + Ev ) + kT ln (1.33)
2 m∗n
is the intrinsic fermi level.
The equation 1.32 gives the Fermi level for an impurity semiconductor in the range over which
the Boltzmann approximation is satisfied, and the donors and acceptors can be regarded as
completely ionized.
Since sinh−1 x is positive for x > 0 and negative for x < 0, it is easily seen that Ef > Ef i for
n-type semiconductors (Nd − Na > 0), and Ef < Ef i , for p-type semiconductors (N d − Na < 0)
as illustrated in Fig.1.3. If the number of donors and acceptors are equal, the inverse hyperbolic
sine function vanishes and the material behaves exactly like an intrinsic semiconductor so far as
electron and hole populations are concerned. Under these conditions the n- and p-type impurities
Conduction band Conduction band

E
EF
Ed
EFi EFi
Ea
EF
Valence band Valence band
n-type p-type
Figure 1.3: Relative positions of band edges, intrinsic Fermi level, and actual Fermi level in (a) n-type and (b)
p-type semiconductors.
are said to be fully compensated. If the net impurity density |Nd − Na | is much larger than ni ,
the number of thermally excited carriers will be small compared with the total number; in this
case the argument of the inverse hyperbolic sine function in equation 1.32 is very large. Since
sinh−1 x = ±ln|2x| for large values of x, it is clear that under these circumstances equation 1.32
becomes

|Nd − Na |
Ef = Ef i ± kT ln (1.34)
ni

|Nd − Na |
Ef = Ef i + kT ln for n-type (Nd > Na ) (1.35)
ni

|Nd − Na |
Ef = Ef i − kT ln for n-type (Nd < Np ) (1.36)
ni
In n-type semiconductor Na = 0 and in p-type semiconductor Nd = 0, then Eq. 1.35 and Eq.
1.36 can be written as

Ef − Ef i
Nd = ni exp (1.37)
kT

Ef i − Ef
Na = ni exp (1.38)
kT
For lightly doped materials, at normal temparature Nd = nd Na = na

Ef − Ef i
nd = ni exp (1.39)
kT

Ef i − Ef
na = ni exp (1.40)
kT
On substituting Eq. 1.35 (with Na = 0) in Eq. 1.1,

Uc Nd Ec − Ef i
no = exp − (1.41)
ni kT
On substituting Eq. 1.36 (with Nd = 0) in Eq. 1.3,

Uv Na Ef i − Ev
po = exp − (1.42)
ni kT
Eq.1.41 and Eq.1.42 gives the electron and hole concentrations at thermal equilibrium for n-
type and p-type semiconductors respectively, interms of intrinsic charge carriers and doping
concentration.
1.4 Fermi energy in impurity semiconductors

At very low temperatures there may not be sufficient thermal energy available to maintain
complete ionization of donor and acceptor impurity levels. At zero absolute temperature, for
example, in an n-type semiconductor, the donor levels must be completely occupied and the
conduction band completely devoid of electrons. Both these conditions cannot be realized unless
the Fermi level lies between the donor levels and the conduction band, as depicted in Fig.1.4
In this case the Boltzmann approximation can no longer be used to describe the occupation
probability associated with the donor levels, since the Fermi level is now above rather than
below the donor energy levels. The Boltzmann approximation can, however, still be used for
the electrons in the conduction band, because the Fermi level will remain many kT units below
the bottom of the conduction band; it should be recalled in this connection that kT is, only
about 0.001 eV at 10K. In case of a purely n-type semiconductor where Na = 0, and we shall
E E
Ec Ec
EF
Ed
Ea
EF
Ev Ev
f(E)=0 f(E)=1 f(E)=0 f(E)=1

f(E)
n-type p-type
Figure 1.4: Distribution function in an n-type semiconductor at a very low temperature.
assume that T is so low that the concentration of thermally created holes in the valence band is
negligible. Since Na = 0, of course, pa must also vanish. Then the electrical nutrality condition
in exrinsic semiconductor can be written as
no − Nd + nd = 0 (1.43)
or, using 1.1 and 1.8,

 
Ec −E
− kT f 1
Uc e − Nd 1 − Ed −Ef
 = 0
1
1+ 2
e kT
Ec −Ef Nd
Uc e− kT − Ed −Ef = 0
1 + 2e− kT
Ec −Ef Ec +Ed −2Ef
Uc e− kT + 2e− kT − Nd = 0 (1.44)
Letting
Ef Ec Ed
α = e kT , βc = e− kT and βd = e− kT (1.45)
the equation 1.44 becomes

α Nd
α2 + − = 0 (1.46)
2βd 2Uc βd βc
On solving the equation 1.46,
12
Ef 1 1 Nd
α = e kT = − ± + (1.47)
4βd 16βd2 2Uc βc βd
Since when Nd = 0, eEf /kT must be a positive quantity. Therefore we must choose the positive
sign of the radical in equation 1.47. On taking logarithm, equation 1.47 can be written as
" 21 #
Ef 1 1 Nd
= ln − + + (1.48)
kT 4βd 16βd2 2Uc βc βd
Right hand side of the equation 1.48 is of the type

√ x
ln(x + x2 + a2 ) = ln a + sinh−1
a
Then the equation 1.48, can be written as
s s !
Ef Nd Uc βc
= ln + sinh−1 (1.49)
kT 2Uc βd βc 8Nd βd
On substituting for Uc , βc and βd , the above equation becomes

r !
1 kT Nd U c − ∆Eid
Ef = (Ec + Ed ) + ln + kT sinh−1 e 2kT (1.50)
2 2 2Uc 8Nd
where ∆Eid = Ec − Ed is the donor ionization energy. At absolute zero, equation 1.50 gives
1
(Ec + Ed ), which means that the Fermi level is midway between the donor levels and the
2
conduction band. As the temperature increases, the Fermi level first increases slightly (remaining
nevertheless below Ec ) and then decreases, moving down through the donor levels toward the
center of the gap. For temperatures such that the Fermi level is several kT units below the
donor energy Ed at which point the donors are almost completely ionized. Equation 1.50 above
reduces, approximately, to 1.35.
It can be shown in a similar way that for a p-type semiconductor under these same circum-
stances,
r !
1 kT Na Uv − ∆Eia
Ef = (Ea + Ev ) + ln + kT sinh−1 e 2kT (1.51)
2 2 2Uv 8Na
where ∆Eia = Ea − Ev is the acceptor ionization energy. At absolute zero, equation 1.51 gives
1
(Ev + Ea ), which means that the Fermi level lies midway between the acceptor levels and
2
the valence band, and with increasing temperature moves first downward a bit, then upward,
approaching the value given by equation 1.36 as the acceptor levels become fully ionized.
In both n-type and p-type semiconductors the Fermi level approaches the intrinsic value Ef i
as given by Eq. 1.33 for sufficiently high temperatures, since at some high temperature the
number ni of thermally excited carriers will be far in excess of the number |Nd − Na | contributed
by donor and acceptor impurities. Beyond this point the behavior of the material will in every
way approximate that of an intrinsic semiconductor. In some instances, however, this transition
temperature may exceed the melting point of the semiconductor. The Fermi level, as determined
by Eq. 1.50 and Eq. 1.51 in the low temperature range and by Eq. 1.34 at higher temperatures
is shown as a function of temperature for both n-type and p-type materials in Fig.1.5.
Figure 1.5: Variation of Fermi energy with temperature for (a) n-type and (b) p-type semiconductors, of various
impurity densities. The curves marked ”1” correspond to relatively low impurity densities, those marked ”2” to
intermediate densities, and those marked ”3” to relatively high impurity densities.
1.5 Drift of electrons in an electric field

The interaction between a free electron and phonons or crystal defects can be viewed as a series
of collisions obeying the principles of conservation of energy and momentum. As a consequence
electrons are never at rest and are submitted to a perpetual random motion that can be compared
to the Brownian motion of fine particles in a liquid. The trajectory of electrons is thus a series of
random velocity vectors. In the absence of an applied external force all these small movements
average out and the net displacement of the electron is zero (Fig. 1.6) When an electric field is
applied, on the other hand, a net drift of the electron in the opposite direction of the electric field
is observed It is worthwhile noting that the random thermal velocity of electrons is much larger
than the velocity produced by imposing an electric field. To obtain the current flow resulting
from this process one must calculate the average drift velocity of the electrons caused by the
electric field. The analogy with Brownian motion in a liquid allows us to write two hypotheses
Figure 1.6: Electron motion: A: in the absence of electric filed, B: in the presence of electric field. The stars
represent collisions.
concerning the motion of an electron:

1. Each electron in the conduction band moves freely in the crystal between each collision.
The average time between two collisions is called relaxation time, τn .
2. The direction of the electron motion after a collision is random. Collision events, are
therefore, isotropic.
If no is the number of electrons in the conduction band can undergo a collision at any instant
t = to and n is the number of electrons which have not yet undergone a collision at any instant
t > to . The population of these electrons, n decreases between t and t + dt by an amount dn
according to the equation:
n
dn = − dt (1.52)
τn
The total number of electrons that have not undergone a collision in the time interval between
t and to is
Z n Z t
dn 1
= − dt
no n to τn
−(t−to )
n = no e τn (1.53)
The equation for the movement of a quasi-free electron with an effective mass m∗n in the presence
of electric filed E is
dv
m∗n = −eE
dt
(1.54)
The drift velocity gained by electron from time to to t is

Z v Z t
eE
v= dv = − ∗
dt
0 to mn
eE
v = − ∗ (t − to ) (1.55)
mn
The average drift velocity of the electron population, i.e. the drift velocity, vdn resulting from
the application of the electric field is
Z 0
1 eE
vdn = − ∗ (t − to )(−dn) (1.56)
no no mn
On differentiationg Eq. 1.53 and substituting for dn in Eq. 1.56, yields:
eE ∞ 1
Z
vdn = − ∗ (t − to )e−(t−to )/τn dt (1.57)
mn to τn
Let
(t − to ) dt
x= =⇒ dx =
τn τn
the Eq. 1.57 becomes
eEτn ∞ x
Z
= − ∗ xe dx
mn 0
eEτn
vdn = − ∗ = µn E (1.58)
mn
where
eτn
µn = − (1.59)
m∗n
is called the mobility of electrons in the conduction band. From the Eq. 1.59, it can be seen that
the mobility is proportional to the relaxation time of the electrons and inversely proportional
to their effective mass. Since mobility is proportional to the relaxation time it decreases with
temperature because thermal lattice vibrations increase with increasing temperature. Similarly,
impurities and defects cause electron scattering (collisions), and therefore, mobility decreases
with increasing impurity or defect concentration.
A similar derivation can made for holes in the valence band and yields:
eEτn
vdp = − = µp E (1.60)
m∗p
where
eτp
µp = − (1.61)
m∗p
is the hole mobility.
1.6 Matthiessen’s rule

It is normally a very good approximation to combine the scattering of charge carriers due to
both impurities and lattice phonons influences the mobility.
1 1 1
= + (1.62)
µ µimpurities µlattice
where µ is the actual mobility, µimpurities is the mobility that the material would have if there
was impurity scattering but no other source of scattering and µlattice is the mobility that the
material would have if there was lattice phonon scattering but no other source of scattering.
Matthiessen’s rule can also be stated in terms of the relaxation time:
1 1 1
= + (1.63)
τ τimpurities τlattice
A more thorough analysis of the scattering of electrons by phonons yields:
3
µlattice = T − 2 (1.64)
and the dependence of mobility on impurity concentration, N

3
T2
µimpurities = (1.65)
N
1.7 Electrical conductivity of impurity semiconductor.

The current density carried by holes and by electrons is equal to the product of their respective
charge densities and velocities. Thus, the current density, or the charge density per unit time
transported by holes and electrons is
jn = −nevn = ρn vn (1.66)
jp = pevp = ρp vp (1.67)
where ρn and ρp refer to the charge densities associated with the electron and hole densities
no and po , and vn and vp , are the average vector velocities of electrons and holes, respectively.
Since the average electron velocity is µn E and the average hole velocity is µp E, where µn and
µp are the electron and hole mobilities and E is the field strength, we may write equation 1.66
and 1.67 as
jn = no eµn E (1.68)
jp = po eµp E (1.69)
whereby the total electrical current density j may be expressed as
j = e(no µn + po µp )E = σo E (1.70)
The electrical conductivity σ is then given by
σo = e(no µn + po µp ) (1.71)
σo = eµp (bno + po ) (1.72)
where b is defined as the ratio of electron to hole mobility, that is,

µn
b = (1.73)
µp
Using Eq. 1.42 and Eq. ?? in Eq. 1.72

eµp
q
2 2
σo = (b − 1)(Nd − Na ) + (b + 1) (Nd − Na ) + 4ni (1.74)
2
Consider for the moment an n-type semiconductor, for which (Nd − Na ) is positive. At low
temperatures, ni << (Nd − Na ), whereupon, according to Eq. 1.74,
σo ∼
= eµp b(Nd − Na ) = eµn (Nd − Na ) (1.75)
which is independent of temperature except insofar as µn may depend on T. The semiconductor

is then a strongly extrinsic material, most of the carriers being contributed by impurity atoms.
As the temperature is increased, ni will increase rapidly, according to Eq. 1.7, and at length ni
will equal, then exceed, (Nd − Na ). For temperatures such that ni >> (Nd − Na ), Eq. 1.74 gives
σo ∼
= eµp ni (b + 1) (1.76)
which is the expression of conductivity for an intrinsic semiconductor, and then rises rapidly
with increasing temperature due to the rapid increase of ni . The sample is now an essentially
intrinsic semiconductor. The transition between these two extremes of behavior occurs at the
temperature where ni = (Nd − Na ). As the impurity density (Nd − Na ) increases this transition
temperature increases, and for different semiconducting substances of equal impurity density the
transition temperature increases as the band gap ∆E increases, because the temperature required
to generate a given density of intrinsic carriers increases with ∆E. The same qualitative behavior
may be shown to follow from Eq. 1.74 for p-type materials, wherein (Nd − Na ) is negative. A
semilogarithmic plot of conductivity versus 1/T as predicted by Eq. 1.74 for several samples of
varying impurity content is shown in Fig. 1.7.
At extremely low temperatures (less than about 30K) there is not sufficient thermal energy
available to maintain the ionization of donor and acceptor levels. In this temperature range
the conductivity again falls rapidly with decreasing temperature as free electrons and holes are
”frozen out” of the conduction and valence bands into donor and acceptor levels. In crystals of
quite high impurity density the wave functions of electrons on neighboring donor atoms or of
holes on neighboring acceptor atoms may overlap appreciably. The result of this is to broaden
the donor or acceptor level into a narrow energy band, just as atomic energy levels broaden
into a band, in the tight binding approximation, as the electronic wave functions of neighboring
atoms begin to overlap. Since there are twice as many states in this band as there are electrons
(or holes), a small conductivity may be observed in the partially filled impurity band, which is
termed as residual impurity band conductivity.
(a) (b)
Figure 1.7: (a) Electrical conductivity, (b) carrier concentration for a series of n-type germanium samples covering
a wide range of donor impurity content.
1.8 Hall effect in intrinsic and impurity semiconductors

If electric field, E along the x direction and magnetic field Bo along the z direction are applied
to a semiconductor material, potential difference VH develops along the y direction (Fig.1.8).
The force F on the charge carrier (say a hole) move with velocity v is
Figure 1.8: Hall Effect
dv
m∗p = eE + e(v × Bo ) (1.77)
dt
Since Bo is along z direction, the velocity of charge carrier having two components along x and
y directions, then Eq.1.77 can be written as
dvx eEx e
= + (v × Bo )x
dt m∗p m∗p
eEx eBo
= + ∗ vy
m∗p mp
eEx
= + ωo vy (1.78)
m∗p
dvy eEy
similarly = − ωo vx (1.79)
dt m∗p
where
eBo
ωo = (1.80)
m∗p
On multiplying i to the Eq.1.79 and adding to Eq.1.78,
d e
(vx + ivy ) = (Ex + iEy ) − iωo (vx + ivy )
dt m∗p
dv eE
+ iωo v = (1.81)
dt m∗p
where the complex quantities v and E are defined by
v = vx + ivy and E = Ex + iEy (1.82)
On multiplying by eiωo t to the Eq.1.81
dv iωo t eE iωo t
e + iωo veiωo t = e
dt m∗p
d eE iωo t
veiωo t =

e
dt m∗p
On integrating,
eE iωo t
veiωo t = e +C
iωo m∗p
where C is constant of integration. At time t = 0, v = vo , then

eE
C = vo −
iωo m∗p
Therefore,
eE
v = vo e−iωo t + 1 − e−iωo t

∗
(1.83)
iωo mp
The average velocity over the exponential distribution of relaxation time (of holes), τp is
R ∞
ve−t/τp dt
0
hvi = R ∞
e−t/τp dt
0
Z ∞ Z ∞
1 eE
−(1/τp +iωo )t
−t/τp −(1/τp +iωo )t

= vo e dt + e − e dt
τp
0 iτp ωo m∗p 0

vo eE 1
= + 1−
1 + iωo τp iωo m∗p 1 + iωo τp

1 eEτp
hvi = vo + (1.84)
1 + iωo τp m∗p
If we assume that τp is a function of the magnitude v, but that it is independent of direction,

then the average associated with vo is always zero. Then, on substituting for E = Ex + iEy ,

eτp Ex + iEy
hvi =
m∗p 1 + iωo τp

eτp (Ex + iEy )(1 − iωo τp )
=
m∗p 1 + ωo2 τp2

eτp Ex ωo τp eτp Ey ωo τp
= + Ey + i ∗ − Ex
m∗p 1 + ωo2 τp2 1 + ωo2 τp2 mp 1 + ωo2 τp2 1 + ωo2 τp2
ωo2 τp3 ωo τp2 ωo τp2

e τp
hvi = τp − Ex + Ey + i Ey − Ex (1.85)
m∗p 1 + ωo2 τp2 1 + ωo2 τp2 1 + ωo2 τp2 1 + ωo2 τp2
In general, the average velocity v must be computed as
v = vx + ivy (1.86)
On comparision of Eq.1.85 and Eq.1.86,

" #
τp2 τ 3

e p
vx = τp Ex + ωo Ey − ωo2 Ex (1.87)
m∗p 1 + ωo2 τp2 1 + ωo2 τp2
" #
τp2

e τp
vy = Ey − ωo Ex (1.88)
m∗p 1 + ωo2 τp2 1 + ωo2 τp2
The x- and y-components of the current density may, in the customary way, be expressed as
Ix = po evx Iy = po evy (1.89)
In the usual experimental arrangement, as shown in Fig.1.8, the y-component of current Iy = 0,

whereby, the Eq.1.87 in Eq.1.88 gives
τp2 /(1 + ωo2 τp2 )

Ey = ωo Ex (1.90)
τp /(1 + ωo2 τp2 )
Eq.1.90 and Eq.1.87 in Eq.1.89 yeilds,

" 2 !#
po e2 τp2 /(1 + ωo2 τp2 )
Ix = τp + ωo2 − τp3 /(1 + ωo2 τp2 ) Ex (1.91)
m∗p τp /(1 + ωo2 τp2 )
If τp is independent of v, then f (τp ) = f (τp ) and the Eq.1.91 and Eq.1.90 reduce to
po e2 τp3 τp3

2
Ix = τp + ωo − Ex
m∗p 1 + ωo2 τp2 1 + ωo2 τp2
po e 2 τ p
Ix = Ex = σo Ex (1.92)
m∗p
and Ey = ωo τp Ex (1.93)
Using Eq.1.80 and Eq.1.92 in Eq.1.93,
Bo
Ey = Ix (1.94)
po e
It is clear that the simple expression for the Hall coefficient obtained previously is correct even
when there is a Boltzmann distribution of velocities, provided that τ is independent of velocity.
It is also evident from Eq.1.92 that in this case the electrical conductivity is independent of the
magnetic field; there is no magnetoresistance.
Consider a semiconductor crystal which is not strongly extrinsic, but which contains both free
electrons and holes in significant concentrations. The hole and electron velocity components are
then expressed as equations having the form of Eq.1.87 and Eq.1.88. Since ωo τ < 1 for both
holes and electrons, 1 + ωo2 τ 2 ∼
= 1, then the current component expressions can be written as
Ix = −no ev̄nx + po ev̄px (1.95)

and Iy = −no ev̄ny + po ev̄py (1.96)
Using Eqs.1.87 in 1.95,

no e 2 h 2 2 3
i p e2 h
o 2 2 3
i
Ix = τ̄n Ex − ωon τn Ey − ωon τn Ex + ∗ τ̄p Ex + ωop τp Ey − ωop τp Ex
m∗n mp
" # " #
τ 2 τ 3
τn2 2 nτ 3
p 2 p
Ix = no eµn Ex − ωon Ey − ωon Ex + no eµp Ex + ωop Ey − ωop Ex (1.97)
τ̄n τ̄n τ̄p τ̄p
where
eBo eBo eτ¯n eτ¯p
ωon = , ωop = , µn = and µp = (1.98)
m∗n m∗p m∗n m∗p
Using Eqs.1.88 in 1.96,
no e2 h 2
i p e2 h
o 2
i
Iy = τ̄n Ey + ωon τn Ex + ∗ τ̄p Ey − ωop τp Ex
m∗n mp
!
τn2 τp2
= Ex no eµn ωon − po eµp ωop + (no eµn + po eµp )Ey (1.99)
τ̄n τ̄p
In the usual experimental arrangement, as shown in Fig.1.8, the y-component of current Iy = 0,

whereby, the Eq.1.99 gives
τ2 2
po µp ωop τ̄pp − no µn ωon ττ̄nn
Ey = Ex (1.100)
no µn + po µp
Eq.1.99 gives the Hall field. On substituting for Ex using Ix = σo Ex ,
τ2 2
po µp ωop τ̄pp − no µn ωon ττ̄nn Ix
Ex =
no µn + po µp σo
2 p τ2 2 τn 2
1 po µp Bo τ̄p2 − no µn Bo τ̄n2 Bo µ
= Ix ∵ σo = e(no µn + po µp ) and ωo =
e (no µn + po µp )2 τ̄
pτ2 2 τn 2
1 po τ̄p2 − no b τ̄n2 µn
Ex = Bo Ix = RBo Ix ∵b= (1.101)
e (no b + po )2 µp
where
p τ2 2 τn 2
1 po τ̄p2 − no b τ̄n2
R = (1.102)
e (no b + po )2
gives the Hall coefficient in a nonextrinsic semiconductor. If the dominant scattering process for
τp2 τ2 3π
both electrons and holes is acoustical mode phonon scattering then 2 = n2 = and Eq.1.102
τ̄p τ̄n 8
then yields
3π po − no b2
R = (1.103)
8e (no b + po )2
For p-type semiconductor po >> no , then Eq.1.103 yeilds
3π 1
Rp−type = (1.104)
8 epo
For n-type semiconductor no >> po , then Eq.1.103 yeilds

3π 1
Rn−type = − (1.105)
8 eno
For an intrinsic semiconductor po = no = ni and Eq.1.103 reduces to
3π 1 1 − b
Rintrinsic = (1.106)
8 eni 1 + b
The Hall effect, aside from any fundamental interest it may excite, is important because it affords
a practical way of telling whether a sample is n-type or p-type and determining the concentration
of carriers. When combined with measurements of conductivity, the Hall effect can also be used
to determine the mobility of the charge carriers. In samples which are highly extrinsic and in
which the predominant scattering mechanism is acoustical mode phonon scattering, it is clear
from Eq.1.104 and 1.105 that a measurement of R can be used directly to obtain no or po and
the conductivity type follows from the sign of R.
1.8.1 Magnetoresistance
If electric field, E along the x direction and magnetic field Bo along the z direction are applied
to a semiconductor material, potential difference VH develops along the y direction (Fig.1.8).
The current Ix through the material and developed Hall field Ey due to the charge carriers (say
a holes) are given by (see Eq.1.91 and Eq.1.90 )
" 2 !#
po e2 τp2 /(1 + ωo2 τp2 )
Ix = τp + ωo2 − τp3 /(1 + ωo2 τp2 ) Ex (1.107)
m∗p τp /(1 + ωo2 τp2 )
τp2 /(1 + ωo2 τp2 )

Ey = ωo Ex (1.108)
τp /(1 + ωo2 τp2 )
Since ωo τ << 1 for both holes and electrons, 1 + ωo2 τ 2 ∼

= 1. This condition will be satisfied
best for small values of the magnetic induction and small values of τp , which are generally
associated with higher temperatures. It is quite generally satisfied for most substances under
commonly encountered experimental conditions down to the liquid nitrogen temperature range.
then Eq.1.107 can be written as
" 2 !#
2 3
po e2 τp τ p τ p
Ix = 1 + ωo2 − Ex
m∗p τp 2 τp
" 2 !#
τ 3τ − τ 2
p p p
= σpo 1 − ωo2 Ex (1.109)
τp 2
po e 2 τ p
where σpo = is conductivity of the material due to holes. It is apparent now that the
m∗p
electrical conductivity is dependent on the magnetic field; this effect is called magnetoresistance.
It can be shown that the second term in Eq.1.109 is always positive, so that the conductivity is
invariably reduced by the magnetic field. Eq.1.109 can be rewritten as
" 2 !#−1
τ 3τ − τ 2
p p p
Ex = ρo 1 − ωo2 2 Ix
τp
" 2 !#
τ 3τ − τ 2
p p p
= ρo 1 + ωo2 Ix
τp 2
Ex = (ρo + ∆ρ)Ix (1.110)
1
where ρo = is zero-field resistivity and
σpo
2!
τ 3τ − τ 2
p p p
∆ρ = ρo ωo2
τp 2
2!
e2 Bo2 τp3 τp − τp2
= ρo ∗ 2
m τp 2
2!
∆ρ e2 τp3 τp − τp2
= (1.111)
ρo Bo2 m∗ 2 τp 2
∆ρ
The ratio is called magnetoresistance coefficient.
ρo Bo2
Physically, the magnetoresistance effect is due to the fact that the magnetic field deflects the
particles upward or downward along the y-axis in a direction normal to the direction of the
current vector. The resulting trajectories are curved rather than straight, and the average drift
distance along the current flow direction between successive collisions is thereby reduced. The
magnetoresistive term in Eq.1.111 being small compared to the zero-field term for ωo τp << 1.
Then Eq.1.110 becomes
m∗p
Ex = ρo Ix = Ix (1.112)
po e 2 τ p
The Hall field Ey , using the same approximation, ωo τp << 1 in Eq.1.108, becomes
τp2
Ey = ωo Ex (1.113)
τp
eBo
Using ωo = and Eq.1.112 in Eq.1.113,
m∗P
1 τp2
Ey = Ix = RBo Ix (1.114)
po e τ p 2
where the Hall coefficient R is given by
1 τp2 νp
R = 2 = (1.115)
po e τp po e
τp2
where νp = is called Hall factor.
τp 2
τ̄ is the mean free time between carrier collisions, which depends on the carrier energy. For
example, for semiconductors with spherical constant-energy surfaces, τ ∝ E −1/2 for phonon
scattering and τ ∝ E 3/2 for ionized impurity scattering. In general,
τ = C1 E −s (1.116)
where C1 and s are constants. From Boltzmann distribution for nondegenerate semiconductors,
the average value of the nth power of τ̄ is
R ∞ n 3/2 −E/kT
τ E e dE
τ¯n = 0R ∞ 3/2 −E/kT (1.117)
0
E e dE
On solving the integral, we obtain
C1n (kT )−ns Γ 5

2
− ns
τ¯n = (1.118)
Γ 52

where Γ(n) is the gamma function defined as

Z ∞
Γ(n) = xn−1 e−x dx (1.119)
0
λ
For phonon scattering C1 = √ ; where λ is the wavelength of elastic waves and s = 12 , then
2
λ (kT )−1/2 Γ(2) λ 1 4λ

τ̄ = √ 5
=√ √ = √ (1.120)
2 Γ 2 2kT (3/4) π 3 2kT π
2 Γ 3

2
√ 2
λ λ (1/2) π λ
τ¯2 = 2
5
= √ = (1.121)
2kT Γ 2 2kT (3/4) π 3kT
λ3 Γ(1) λ3 1 λ3 4
τ¯3 = 3/2 5
= 3/2
√ = 3/2
√ (1.122)
(2kT ) Γ 2 (2kT ) (3/4) π (2kT ) 3 π
Using Eq.1.120 and Eq.1.121 in Eq.1.115,
3π 1
R = (p-type) (1.123)
8 po e
3π 1
R = − (n-type) (1.124)
8 no e
eBo
Using ωo = , Eq.1.120, 1.121 and Eq.1.122 in Eq.1.109,
m∗P
e2 Bo2 λ2 4 − π

Ix = σpo 1 − ∗2 Ex (1.125)
mP KT 8
From this expression, it can be seen that the magnetoresistance is very large at low temperatures.
1.9 The mobility of the charge carriers

The Hall coefficient R is given by
p τ2
2 τn 2
1 po τ̄p2 − no b τ̄n2
R = (Ref. Eq.1.102)
e (no b + po )2
2 p τ2
2 τn 2
1 po µp τ̄p2 − no µn τ̄n2

µn
= b= (1.126)
e (no µn + po µp )2 µp
For p-type semiconductor po >> no , then Eq.1.126 yeilds
τ2 τ2
µp τ̄p2 µp τ̄p2 τ̄p2 τ̄p2
p p
R = = =⇒ µp = Rσo = µHp (1.127)
epo µp σo τp2 τp2
For n-type semiconductor no >> po , then Eq.1.126 yeilds
2
µn ττ̄n2
2
µn ττ̄n2 p τ̄n2 τ̄n2
R = − n
=− =⇒ µn = −Rσo = µHn (1.128)
epo µn σo τn2 τn2
where σo = po eµp /no eµn is the conductivity of p-type / n-type semiconductor. µHp = Rσo /
µHn = −Rσo is often defined as the Hall Mobility in p-type / n-type semiconductor. The
quantities µp and µn are the true drift mobilities or conductivity mobilities; the Hall mobility
on the other hand is simply a quantity with the dimensions of mobility which is easily obtained
τ̄ 2
from quantities which can be quite simply measured experimentally. The factor must be
τ2
known to convert Hall mobility into the true drift mobility. Since this factor is usually not far
from unity, however, the Hall mobility gives a rough indication of the true mobility in most
cases which arise in connection with experiment.
Figure 1.9: Electron mobility for a series of n-type germanium samples covering a wide range of donor impurity
concentration.
1.10 Cyclotron Resonance

The electrons and holes in semiconductors may be made to exhibit a resonance in the pres-
ence of a large constant magnetic field (along the z-axis) and a small transverse oscillating
radiofrequency or microwave electromagnetic field whose electric vector lies in the xy-plane.
The resulting hole or electron orbits are circular at resonance, and energy is absorbed from
the rf-field every half cycle, as in a cyclotron. The situation is, in fact, precisely analogous to
what happens to nuclear charged particles in a cyclotron, and for this reason the phenomenon
is termed cyclotron resonance.
Let us initially investigate the case of spherical energy surfaces and a single scalar effective mass.
The Lorentz force on a hole in a p-type sample is given by
dv
F = m∗p = eE + e(v × B) (1.129)
dt
If the effective mass m∗p is independent of direction, and if the electric vector of the oscillating
field, whose frequency is ω, is assumed to vibrate along the x-direction, then
Ex = Eo eiωt , Ey = Ez = 0 (1.130)
while Bx = By = 0, Bz = Bo (1.131)
and the components of the Lorentz force (Eq.1.129) give for the equations of motion
d2 x
Fx = m∗p = eEx + e(v × B)x = eEo eiωt + evy Bo (1.132)
dt2
d2 y
Fy = m∗p 2 = eEy + e(v × B)y = −evx Bo (1.133)
dt
d2 z
Fz = m∗p 2 = 0 (1.134)
dt
It has been assumed that the magnetic vector of the rf-field is negligible compared with the
constant field Bo . The z-component equation. The x- and y-component of Eq.1.132 and 1.133
may be written
d2 x eEo iωt dy
2
= ∗
e + ωo (1.135)
dt mp dt
d2 y dx
= −ωo (1.136)
dt2 dt
eBo
where ωo = ∗ . The solutions of the Eq.1.135 and 1.136 may be of the form
mp
x = xo eiωt and y = yo eiωt (1.137)
and the Eq.1.136 and 1.135 reduces to

ωo
−ω 2 yo = −iωo ωxo =⇒ yo = i xo
ω
eEo eEo /m∗p
−ω 2 xo = + iωo ωyo =⇒ xo = 2 (1.138)
m∗p ωo − ω 2
iωo eEo /m∗p
yo = (1.139)
ω(ωo2 − ω 2 )
When ω = ωo resonance occurs and the amplitudes become very large. The resonance frequency
ωo is often referred to as the cyclotron frequency. Clearly, if ωo is measured experimentally and
if the magnetic induction Bo is accurately known, the effective mass m∗ , can be determined.
From Eq.1.138 and Eq.1.138, it can be seen that as ωo → ω
yo = ixo = xo eiπ/2
whereby x(t) = xo eiωt
y(t) = yo eiωt = xo eiωt+π/2
At resonance, then, x(t) and y(t) are orthogonal harmonic vibrations of equal amplitude which
differ in phase by π/2. It is easily seen that the resultant particle trajectory, representing the
hole orbit, is circular.
In practice, the resonance is usually observed by measuring the Q-factor of a microwave resonant
cavity as the magnetic field Bo (and hence the frequency ωo ) is varied. When ωo equals the
microwave excitation frequency, a sharp decrease in the Q of the cavity is obtained, because at
resonance a great deal of electromagnetic field energy is absorbed when carriers are excited to
large resonance amplitudes and then are scattered by impurities or lattice vibrations.
When the cyclotron resonance effect in silicon and germanium was studied experimentally, it
was found for both substances that there were more than two resonances (one for electrons, one
for holes). Moreover, it was observed that some of the cyclotron resonance frequencies underwent
significant changes as the orientation of the crystal with respect to the static magnetic field was
altered, while others were practically independent of sample orientation. These effects were
finally explained using a model in which the constant energy surfaces for electrons in k-space
are ellipsoidal rather than spherical, and in which the presence of three separate valence bands,
two of which are degenerate at k = 0, are considered. The constant energy surfaces for holes in
∂ 2E
these valence bands are roughly spherical, although since differs for the different valence
∂k 2
bands there is more than one possible effective mass for holes.
It has been found, however, that the minimum value of energy within the reduced zone
need not necessarily be at k = 0, but may lie elsewhere within the zone. This is indeed the
observed state of affairs for electrons in the conduction band of germanium and silicon. Since
the diamond lattice is really two interpenetrating face-centered cubic lattices separated along
p
the cube diagonal by a distance a 3/4 (where a is the cube edge). A plot of E versus k
along the (100) direction in this Brillouin zone for electrons in the conduction band of silicon is
shown in Fig.1.10. The smallest value of E is reached at a value of k which is about 0.8(2π/a).
For energies greater than Ec , one may construct constant energy surfaces about the minimum
energy point. It is clear that the energy versus crystal momentum curve shown in Fig.1.10(a)
is roughly parabolic about the minimum point, corresponding to free electron behavior with an
appropriate effective mass which we shall call m∗k . If one were to plot the E versus k curve along
some path passing through the minimum point, but perpendicular to the kx axis one would also
obtain a curve which would be parabolic about the minimum point, but there is now no aspect
of crystal symmetry which would require that the curvature in this direction be the same as it
is along the kx direction. (Fig.1.11.) As a result the effective mass related to changes in the
momentum component along this direction has a value which is different from m∗k , and which
we shall denote as m∗⊥ . It turns out, as we shall soon see, that requirements of crystal symmetry
demand that the effective mass have the same value m∗⊥ for any direction in k-space normal to
the kx -direction.
Figure 1.10: The E versus k relation for (a) the conduction band of silicon, plotted along the kx -direction, (b)
the conduction band of germanium, plotted along a (111) direction in k-space.
Figure 1.11: The E versus k relation for the conduction band of silicon plotted along the kx -direction (solid
curve) and along a line normal to the kx -axis which passes through the point kmin (dashed curve). The different
curvatures at the minimum point reflect the anisotropy of the effective mass.
The energy in excess of Ec represents the kinetic energy of a free electron in the conduction
band, whereby
(px − pxo )2 p2y + p2z

E − Ec = + (1.140)
2m∗k 2m∗⊥
where pxo = ~kxo represents the value of px at the bottom of the band, as shown in Fig. 1.11.
According to this the surface of constant energy for an electron of energy s slightly greater
than ac is seen to be an ellipsoid of revolution whose center is at kx = kxo , ky = kz = 0. Since,

however, the six 100 directions of a cubic lattice are crystallographically equivalent, the E versus
k curve along any of these six directions must be the same. There must then be six equivalent
energy minima along the six 100 directions, and associated with each of these minima there
must be a family of ellipsoidal constant energy surfaces similar to those described by Eq.1.140
and is illustrated in Fig.1.12(a).
The situation for electrons in the conduction band of germanium is somewhat similar, except
that there are a set of eight equivalent minimum energy points which lie along the 111 directions
in k-space at the intersection of those directions with the surface of the Brillouin zone. These
minima thus lie at the centers of the hexagonal faces of the Brillouin zone. The constant energy
surfaces are again ellipsoidal in form, but since the energy minima are at the zone boundary, and
since there is a rather large forbidden energy region which must be surmounted before electrons
can be excited outside the Brillouin zone in these directions, there are eight half-ellipsoids which
extend into the interior of the Brillouin zone along the 111 directions, as shown in Fig.1.12(b).
For many purposes these may be considered to be equivalent to four full ellipsoids. In both
germanium and silicon the longitudinal effect mass m∗k is much larger than the transverse mass
m∗⊥ , the ellipsoids thus being quite long and thin.
Figure 1.12: Ellipsoidal constant energy surfaces in k-space for (a) silicon and (b) germanium. In silicon the
major axes of the ellipsoids are along 100 directions, while in germanium, since the energy minimum is on the
zone boundary, the constant energy surfaces form eight half-ellipsoids whose major axes lie along 111 directions.
Consider a single ellipsoid, which we shall assume to be symmetric about the kz -axis and to be
centered on the point ko , as shown in Fig.1.13. The magnetic induction B, makes an angle θ
with the major axis of the ellipsoid, and for convenience we shall assume that the oscillating
electric field vector has only a y-component. The equation of motion (for a particle of positive
charge e) may be written in terms of the tensor which represents the reciprocal of the effective
mass as
Figure 1.13: Vector geometry used for the calculation of cyclotron resonance effect in materials with energy
surfaces are ellipsoidal
dv 1 1
= ∗
· F = ∗ [eE + e(v × Bo )] (1.141)
dt m m
The relation between e and k for the case of the ellipsoidal energy surface shown in Fig.1.13
according to Eq.1.140 with p = ~k,
" #
2 2 2
kx + ky (kz − k zo )
E − Ec = ~2 + (1.142)
2m∗⊥ 2m∗k
An effective mass tensor associated with each ellipsoid given by
1 ∂ 2E

1 1
= ∇ k (∇k E) = (1.143)
m∗ αβ ~2 ~2 ∂kα ∂kβ
and from this it is apparent that the reciprocal effective mass tensor is symmetric, that is

1 1
= (1.144)
m∗ αβ m∗ βα
∂ 2E ∂ 2E 1
According to Eq.1.144, = 0 for α 6= β and = ∗ for α = β, so that all the
∂kα ∂kβ ∂kα ∂kβ m
off-diagonal elements of the tensor are zero. The diagonal elements are being

1 1 1 1 1
∗
= ∗
= ∗ and ∗
= ∗ (1.145)
m xx m yy m⊥ m zz mk
The effective mass tensor then has the form

1∗ 0 0
m⊥
1 1

= 0 m⊥∗ 0 (1.146)
m∗
1
0 0

m∗ k
and

1∗ 0 0 Fx Fx
m⊥ m∗⊥

1
0 1

Fy
· F = m∗⊥
0 · Fy =

m∗⊥
(1.147)
m∗
1
Fz

0 0 Fz

m∗
k
m∗k
where Fx , Fy , Fz are the Lorentz force components. Using Eq.1.141, the equation of motion can
be written as
dvx
Fx = m∗⊥ = eEx + e(vy Bz − vz By )
dt
dvy
Fy = m∗⊥ = eEy + e(vz Bx − vx Bz ) (1.148)
dt
dvz
Fz = m∗k = eEz + e(vx By − vy Bx )
dt
According to Fig.1.13, Bx = Bo sin θ, Bz = Bo cos θ, By = 0, Ey = Eo eiωt , Ex = Ez = 0, the
Eq.1.148 yeilds
dvx
= ω⊥ vy cos θ
dt
dvy eEo
= ω⊥ vz sin θ − ω⊥ vx cos θ + ∗ eiωt (1.149)
dt m⊥
dvz
= −ωk vy sin θ
dt
where
eBo eBo
ω⊥ = and ωk = (1.150)
m∗⊥ m∗k
We may assume oscillatory solutions of the form
x = xo eiωt , y = yo eiωt and z = zo eiωt
then Eqs.1.149 can be written as
ω 2 xo = −iω⊥ ωyo cos θ (1.151)

eEo
ω 2 yo = −i(ω⊥ ωzo sin θ − ω⊥ ωxo cos θ) − (1.152)
m∗⊥
ω 2 zo = iωk ωyo sin θ (1.153)
Eq.1.151 and 1.153 in Eq.1.152 gives

eEo
ω 2 yo = ω⊥ ωk yo sin2 θ + ω⊥
2
yo cos2 θ −
m∗⊥
eEo 1
yo = − ∗
(1.154)
m⊥ (ω − ω⊥ ωk sin2 θ − ω⊥
2 2
cos2 θ)
Eq.1.154 in Eq.1.151 and 1.153 gives
eEo iω⊥ cos θ
xo = ∗
(1.155)
m⊥ ω(ω 2 − ω⊥ ωk sin2 θ − ω⊥
2
cos2 θ)
eEo iωk sin θ
zo = − ∗ (1.156)
m⊥ ω(ω 2 − ω⊥ ωk sin2 θ − ω⊥
2
cos2 θ)
The amplitudes xo , yo and zo become very large when
q
ω = ω⊥ ωk sin2 θ − ω⊥
2
cos2 θ) (1.157)
and this is accordingly the cyclotron resonance frequency for the particular ellipsoid which we
are considering. The resonance frequency depends only upon the magnetic field, the effective
masses and the angle θ between the field Bo and the major axis of the ellipsoid. For ellipsoidal
energy surfaces, it is clear that the resonance frequency is a strong function of the orientation of
the sample relative to the magnetic field. At the resonant frequency the particle orbits can be
shown from Eq.1.154, 1.155 and 1.156 to be elliptical in form and to lie in a plane perpendicular
to the field Bo . The above results have been derived for positive particles of charge e, but for
electrons we need only replace e by −e (in which case we must also replace ωk by −ωk and ω⊥ by
−ω⊥ ). The resonant frequency is obviously unaltered by these substitutions; the only difference
is that electrons traverse their orbits in the opposite sense from that found for positive particles.

Semiconductor Physics

Uploaded by

Copyright:

Available Formats

Semiconductor Physics

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Semiconductor Physics

Uploaded by

Copyright:

Available Formats

Physics of Semiconductors

1.1 Impurity semiconductors

1.2 Carrier concentrations in impurity semiconductors

According to these results, the concentration of carriers in an intrinsic semiconductor is a strong

where Nd is the concentration of donor impurity atoms.

gd (E) = 2Nd δ(E − Ed ) (1.10)

Since the concentration of electrons in the conduction band is given by

On substituting for po = n2i /no to the equation 1.17,

n2o − (Nd − Na )no − n2i = 0 (1.18)

On solving for no by using equation 1.18, we get

1.3 Impurity concentration in the case of extrinsic semi-

the equation 1.24 becomes

Right hand side of the equation 1.28 is of the type

Conduction band Conduction band

Valence band Valence band

1.4 Fermi energy in impurity semiconductors

f(E)=0 f(E)=1 f(E)=0 f(E)=1

or, using 1.1 and 1.8,

the equation 1.44 becomes

Right hand side of the equation 1.48 is of the type

On substituting for Uc , βc and βd , the above equation becomes

1.5 Drift of electrons in an electric field

concerning the motion of an electron:

The drift velocity gained by electron from time to to t is

On differentiationg Eq. 1.53 and substituting for dn in Eq. 1.56, yields:

the Eq. 1.57 becomes

is the hole mobility.

1.6 Matthiessen’s rule

and the dependence of mobility on impurity concentration, N

1.7 Electrical conductivity of impurity semiconductor.

whereby the total electrical current density j may be expressed as

The electrical conductivity σ is then given by

σo = eµp (bno + po ) (1.72)

where b is defined as the ratio of electron to hole mobility, that is,

which is independent of temperature except insofar as µn may depend on T. The semiconductor

1.8 Hall effect in intrinsic and impurity semiconductors

Figure 1.8: Hall Effect

On multiplying i to the Eq.1.79 and adding to Eq.1.78,

where the complex quantities v and E are defined by

v = vx + ivy and E = Ex + iEy (1.82)

On multiplying by eiωo t to the Eq.1.81

where C is constant of integration. At time t = 0, v = vo , then

If we assume that τp is a function of the magnitude v, but that it is independent of direction,

In general, the average velocity v must be computed as

On comparision of Eq.1.85 and Eq.1.86,

Ix = po evx Iy = po evy (1.89)

In the usual experimental arrangement, as shown in Fig.1.8, the y-component of current Iy = 0,

τp2 /(1 + ωo2 τp2 )

Eq.1.90 and Eq.1.87 in Eq.1.89 yeilds,

Using Eq.1.80 and Eq.1.92 in Eq.1.93,

Ix = −no ev̄nx + po ev̄px (1.95)

Using Eqs.1.87 in 1.95,

In the usual experimental arrangement, as shown in Fig.1.8, the y-component of current Iy = 0,

For n-type semiconductor no >> po , then Eq.1.103 yeilds

τp2 /(1 + ωo2 τp2 )

Since ωo τ << 1 for both holes and electrons, 1 + ωo2 τ 2 ∼

On solving the integral, we obtain

C1n (kT )−ns Γ 5

where Γ(n) is the gamma function defined as