1 Carrier density equations

1.1 Real Density of states

The density of states g(E) dE is established by the full band structure of a material. It is a bumpy function that,
for semiconductors, can be characterized in energy by the valence band maximum (EV ) and the conduction band
minimum (EC ). Between these two energies lies the band gap of width EG = EC − EV . Beyond this general
characterization of EC and EV , the details of the density of states must be extracted from the full band structure
calculations.

1.2 Approximations to the density of states

In the neighborhood of the conduction band minimum and the valence band maximum, the E–⃗k diagram will
have a parabolic-like nature. If we “fit” (approximate) the bands as parabolic, the density of states near the band
edges can be approximated as having a form similar to that we determined for a free electron in a box, using the
electron mass me as a fitting parameter to match the actual density of states; this gives rise to a density of states
effective mass m⋆e that accounts for complexity of all the bands within the full E–⃗k diagram. Within this limit,
the density of states near the band edge in the conduction band is be approximated as
3/2
2m⋆e

1/2
g(E) dE = 4π (E − EC )
h2

and in the valence band as

3/2
2m⋆h

1/2
g(E) dE = 4π (EV − E)
h2

where m⋆e and m⋆h are the fitting parameters and are referred to as the effective mass of the electron and holes
respectively.
It is common to refer to m⋆e /me and m⋆h /me also as the effective mass; the correct interpretation is obvious
given the 9.11 × 10−31 factor difference in values between m⋆e and m⋆e /me .

1.3 Fermi-Dirac statistics

The Fermi-Dirac probability function gives the probability that a state at E will be occupied given the chemical
potential of the electrons µ = EF (also called the Fermi Level). This is given for the bands as
1
PF D (E, EF ) =
1 + e(E−EF )/kT
In the limit that E − EF ≫ kT , this can be approximated with the Boltzmann probability distribution

PF D (E, EF ) ≈ PB (E, EF ) = e−(E−EF )/kT

1.4 Density of carriers in conduction and valence bands

The number of carriers in the conduction and valence band are determined by integrating the density of states
with the Fermi-Dirac probability for occupation. The exact expressions are
Z top
n = g(E) PF D (E, EF ) dE
EC
Z EV
p = g(E) [1 − PF D (E, EF )] dE
bot

As the actual density of states g(E) is not known, this integral is very difficult to do exactly.
If we use the parabolic band approximation, we can replace the actual density of states with the free electron
density of states and the fitting parameter effective mass. If we also assume that the band is wide, the upper

1
limit in the integral can be replaced by ∞ since the Fermi-Dirac function dies exponentially for energies many
kT above the Fermi Level. This gives approximations
Z ∞  ⋆ 3/2
2me 1/2
n = 4π 2
(E − EC ) PF D (E, EF ) dE
EC h
Z EV  ⋆ 3/2
2mh 1/2
p = 4π 2
(EV − E) [1 − PF D (E, EF )] dE
−∞ h

1.5 Effective density of states approximation

If the Fermi Level lies remains within the gap region, EV < EF < EC , these relations can be simplified further
using the effective density of states approximation. The Fermi-Dirac probability distribution is replaced by the
Boltzmann approximation and the integral can be done analytically yielding
n ≈ NC (T ) e−(EC −EF )/kT
p ≈ NV (T ) e−(EF −EV )/kT
where NC (T ) and NV (T ) are called the effective density of states
3/2
2πm⋆e kT

NC (T ) = 2
h2
3/2
2πm⋆h kT

NV (T ) = 2
h2
These values are the number required to ensure that the correct integral is numerically equal to the approximation.
In Si at room temperature (300 K), NC = 2.82×1019 cm−3 and NV = 1.17×1019 cm−3 . At other temperatures,
the T 3/2 scaling behavior of NC and NV can be used to obtain values in Si as a function of temperature
 3/2
T
NC (T ) = NC (300 K)
300 K
 3/2
T
NV (T ) = NV (300 K)
300 K

1.6 Assumptions Underlying the Effective Density of States Approximation

There are three assumptions required to replace the integral over the true density of states with an approximate
integral over the free electron-like density of states. These include:
Parabolic Approximation: The density of states for the actual band diagram can be reasonably be reasonably
modeled as g(E) ∝ (E −EC )1/2 , which is equivalent to modeling the actual bands making up the conduction
band minimum with a parabolic E-vs-k relationship (e.g. parabolic). This must hold within a few kT of the
edge.
Wide Band Approximation: The width in energy of the band is sufficiently large (compared to kT ) that we
can replace the limits in the integral from the top of the conduction band (or bottom of the valence band)
with infinity without significant error
Large Crystal Approximation: The total number of states in a band is on the order of the number of atoms
in the crystal (4N for the conduction band of Si, for example). The crystal must be sufficiently large that
the density of states appears to be a continuum of states rather than discrete.
Under these approximations, the correct integral for the carrier density in the conduction and valence bands can
be written as
Z ∞  ⋆ 3/2
2me 1/2
n = 4π 2
(E − EC ) PF D (E, EF ) dE
EC h
Z EV  ⋆ 3/2
2mh 1/2
p = 4π 2
(EV − E) [1 − PF D (E, EF )] dE
−∞ h

2
 The adddition of Boltzmann’s approximation completes the Effective Density of States approximation and
gives the carrier densities as the numerically simplified expressions

n ≈ NC (T ) e−(EC −EF )/kT

p ≈ NV (T ) e−(EF −EV )/kT

1.7 Physical Interpretation of NC (T ) and NV (T )

In one sense, NC and NV are just prefactors in the expressions n = NC e−(EC −EF )/kT such that the value for
n is equal to that we would obtain doing the integral properly. The actual density of states g(E)dE is temper-
ature independent, but the integral has a strong temperature dependence through the Fermi-Dirac probability
distribution function. This naturally leads to NC being a strong function of temperature.
Another way to think about NC (T ) is that it is the number of states that I would have to put right at the
conduction band edge and nowhere else such that the number of occupied states is the same as occurs in the true
integral. The complex “stadium-like” structure of the band diagram is collapsed to an equivalent set of seats on
a flat floor. Being at a single level, the probability that these seats are filled is much simpler and the expression
bypasses the integral.

1.8 Effective mass

There are at two distinct definitions of the effective mass of electrons and holes in a semiconductor. One, described
above, is to define m⋆ as the value that gives the correct shape for the density of states near the conduction band
minimum and valence band maximum. For Si, this requires electron and hole effectives masses of 1.08 me and
0.56 me respectively. This is an average effective mass that hides more subtle details of the band structure. The
table below lists the effective density of states values for m⋆ for Si, Ge and GaAs at 300 K.
Germanium Silicon Gallium Arsenic
Ge Si GaAs
Bandgap (eV) 0.66 1.12 1.424
Direct gap (eV) 0.8 3.2 1.424
NC (cm−3 ) 1.05 × 1019 2.82 × 1019 4.37 × 1017
NV (cm−3 ) 3.92 × 1018 1.08 × 1019 8.68 × 1018
ni (cm−3 ) 1.83 × 1013 8.81 × 109 2.03 × 106
m⋆e /me 0.56 1.08 0.067
m⋆h /me 0.29 0.56 0.47
A second, and equally valid and perhaps more fundamental, definition of effective mass is based on the
curvature of the E − ⃗k diagram in the vicinty of the band minima and maxima. This is based on pretending that
the wave packets behave like a particle if we accept that this fake patical has a mass given by the curvature of
the band such that from E = h̄2⃗k 2 /2m⋆ ; the result for m⋆ is

h̄2
m⋆ =
d2 E
dk 2
This is an inertial mass that would appropriate for motion under a foce such as an electric or magnetic field.
The problem is that there may be several degenerate bands that contribute to the total density of states, each
with a different curvature. In Si, Ge and GaAs, at the valence band maximum there are two bands with essentially
the same maximum at the Γ point; these are referred to as the light and heavy holes (high curvature implies
light, low curvature heavy). In silicon, these have effective masses given by the curvature of m⋆hh /me = 0.49
and m⋆lh /me = 0.16 where mhh and mlh refer to the heavy and light holes respectively. Interestingly there is an
additional band just slightly below the valence band maximum which is referred to as the split-off band with an
effective mass of m⋆V,so /me = 0.29.
Since the states in the two bands simply add to get the total density of states, and the density of states scales
as (m⋆ )3/2 , we can obtain the density of states effective mass m⋆h,dos as

3/2 3/2 3/2
m⋆h,dos = (m⋆hh ) + (m⋆lh )

3
or equivalently
h i2/3
3/2 3/2
m⋆h,dos = (m⋆hh ) + (m⋆lh )

which indeed gives m⋆h,dos /me = 0.55 for Si (close enough to the 0.56 value listed above).
In the case of electrons, things are even more complicated. While the valence band maximum occurs at the
Γ point, the conduction band minimum along the X direction. The constant energy contours of the conduction
minimum are ellipsoids with a different curvature along the [100] direction and transverse to the [100] direction.
This gives different effective masses of in the longitudinal and transverse directions of m⋆e,l /me = 0.98 and
m⋆e,t /me = 0.19. Complicating life further is the fact that there are six equivalent ellipsoids in each of the six
[100] type directions (a degeneracy of 6 in the band). Treating the ellipsoid as an equivalent spherical band, the
effective effective mass is a geometric average of the three directions. In addition, the six degenerate ellipsoids
must sum together to get the total density of states (as for the holes above) adding a 62/3 factor. Combined, the
density of states averaged effective mass for electrons is
1/3
m⋆e,dos = g 2/3 m⋆e,l × m⋆e,t × m⋆e,t


where g is the number of equivalent band minima. Plugging in number for Si, we have
1/3
m⋆e,dos /me = (6)2/3 [0.98 × 0.19 × 0.19] = 1.08

as expected. The various curvature effective masses are given below for Ge, Si and GaAs.
Ge Si GeAs
Bandgap (eV) Eg 0.66 1.12 1.424
Longitudinal electron effective mass m⋆e,l /me 1.64 0.98 1.98
Transverse electron effective mass m⋆e,t /me 0.082 0.19 0.37
Direction of electron minimum ⃗k [111] [100] [111]
⋆
Heavy hole effective mass mhh /me 0.28 0.49 0.45
Light hole effective mass m⋆lh /me 0.044 0.16 0.082
Split-off band energy offset (eV) EV,so -0.028 -0.044 -0.34
Split-off band effective mass m⋆h,so /me 0.084 0.29 0.154

There are numerous other physical processes where it is critical to use an appropriate averaging of the true
“band curvature” effective masses. For example, the mass that would be appropriate in a Drüde model of
conductivity is a weighted inverse mass since acceleration (and hence drift velocity) scales as 1/m. For holes with
the light and heavy bands, we need to account also for the fraction of the holes that would be in each band, which
scales as g(E) dE and hence (m⋆ )3/2 . This gives
−1
(m⋆hh )3/2 (m⋆lh )3/2
 
1
m⋆h,cond = +
(mhh ) + (m⋆lh )3/2
⋆ 3/2 m⋆hh m⋆lh
 ⋆ 3/2
(mhh ) + (m⋆lh )3/2

=
(m⋆hh )1/2 + (m⋆lh )1/2

For Si, we have

(0.49)3/2 + (0.16)3/2
 
m⋆h,cond /me = = 0.37
(0.49)1/2 + (0.16)1/2

which is sufficiently close to the accepted value of 0.36 me .

For electrons, we have a different averaging. All of the electrons are equivalent but have equal probability of
the three directions (one longitudinal and two transverse). Consequently we have to average the inverse masses
of the three directions giving
3
m⋆e,cond =
1/m⋆e,l + 1/m⋆e,t + 1/m⋆e,t

4
which, again for Si yields
3
m⋆e,cond /me = = 0.26
1/0.98 + 1/0.19 + 1/0.19
which is the accepted value.
The effective mass for conduction in Ge, Si and GaAs are given below:
Ge Si GeAs
Bandgap (eV) Eg 0.66 1.12 1.424
electron conduction effective mass m⋆e,cond /me 0.12 0.26 0.067
hole conduction effective mass m⋆h,cond /me 0.21 0.36 0.34
Other processes, such as free carrier absorption at long wavelengths (intraband absorption of low energy
photons) have yet another mass scaling and must be treated accordingly.
Just to complicate things further, the effective mass is really a tensor property since we should consider the
cross-curvature terms as well as the pure curvature. The tensor for the inertial effective mass can be defined as
 2 
 −1  1 ∂ E
Minertial ij = 2
h̄ ∂ki ∂j
where ki and kj are the wavevector components in different directions.
And finally, it must be noted that Si has yet another complication. The heavy hole band in Si doesn’t approach
anything like a spherical band (more of a squashed cube) and hence these approximately are more challenging
that presented here.

1.9 Summary table for effective mass

Ge Si GeAs
Bandgap Eg (eV) 0.66 1.1242 1.424
Direct gap) Eg,direct (eV) 0.8 3.2 1.424
Electron effective density of states NC (cm−3 ) 1.05 × 1019 2.86 × 1019 4.37 × 1017
Hole efffective density of states NV (cm−3 ) 3.92 × 1018 3.10 × 1019 8.68 × 1018
Intrinsic carrier density ni (cm−3 ) 1.83 × 1013 1.07 × 1010 2.03 × 106
Longitudinal electron effective mass m⋆e,l /me 1.64 0.9163 1.98
Transverse electron effective mass m⋆e,t /me 0.082 0.1905 0.37
Direction of electron minimum ⃗k [111] 0.85 × [100] [111]
Heavy hole effective mass m⋆hh /me 0.28 0.49 0.45
Light hole effective mass m⋆lh /me 0.044 0.16 0.082
Split-off band energy offset (eV) EV,so -0.028 -0.0441 -0.34
Split-off band effective mass m⋆h,so /me 0.084 0.29 0.154
DOS electron effective mass m⋆e,dos /me 0.56 1.08 0.067
DOS hole effective mass m⋆h,dos /me 0.29 0.56 0.47
conduction electron effective mass m⋆e,cond /me 0.12 0.26 0.067
conduction hole effective mass m⋆h,cond /me 0.21 0.36 0.34

1.10 Mass action law

The density of electrons in the conduction band increases exponentially as the Fermi Level approaches the con-
duction band edge. Simultaneously, the density of holes in the valance band falls exponentially since the Fermi
Level moves away from the valence band edge. Within the Effective Density of States Approximation, the product
of the electron and hole density is a constant, commonly called n2i .
h i h i
n × p = n2i = NC (T ) e−(EC −EF )/kT × NV (T ) e−(EF −EV )/kT
= NC (T )NV (T ) e−(EC −EV )/kT
= NC (T )NV (T ) e−Eg /kT
= NC NV e−Eg /kT

5
where Eg is the bandgap of the material.
The term n2i refers to the intrinsic carrier concentration ni . In a pure semiconductor in equilbrium with no
impurity doping (or other influences), the number of holes and electrons must be equal (n = p) and hence both
will equal ni , the intrinsic (pure) carrier concentration.
This mass action law is extremely powerful. As soon as the electron concentration is a few orders of magni-
tude greater than ni , the hole concentration becomes negligible (minority component). The holes may still be
important, but their numbers will be small compared to the number of electrons. The converse is similarly true
when the hole concentration becomes large.
Since the effective density of states NC (T ) and NV (T ) scale as T 3/2 , n2i will scale as T 3 .

n × p = n2i = NC (T )NV (T ) e−Eg /kT

 3
T
= NC (300 K)NV (300 K) e−Eg /kT
300 K

Note that the mass action law does cease to be valid if the Fermi Level rises above the conduction band
minimum or below the valence band maximum. In these regimes, the product np will fall below the value
predicted by the effective density of states approximation.

1.11 Donors and Acceptors

In silicon, the addition of an impurity atom from column VB on the periodic chart will bring an extra electron
to the system. Phosphorus, for example, has 15 electrons compared to silicon’s 14 electron. The extra electron
is “most happy” when localized to the phosphorus atom. However, due to the large dielectric constant of the Si
crystal, the binding energy of this extra electron is small and it can be readily “ionized” to become delocalized
in the conduction band.
Acceptors are the analog attached to the valence band. In silicon, a column IIIB impurity such boron will be
deficit one electron in the valence band. The empty electron state will normally be localized around the boron
atom, but thermal excitations can cause a neighboring valence electron to fill that state with the “empty” bond
becoming delocalized as a hole in the valence band.
We can roughly estimate the binding energy of an electron on a donor (or an extra electron on an acceptor)
by treating the impurity as a hydrogen-like atom in the polarizable medium of the crystal lattice. The radius of
the first electron orbital is approximately

r ≈ ϵr a0

where ϵr is the relative dielectric permittivity of the matrix and a0 is the Bohr radius

4πϵ0 h̄2
a0 =
m e e2
= 52.917721 pm = 0.53 Å = 0.053 nm

Similarly, the binding energy is approximately

RH
Eb ≈
ϵ2r

where RH is the Rydberg constant (energy)

2
e2

me
RH =
4πh̄ϵ0 2
= 13.605692 eV

For Si, with ϵr = 11.9, the estimated binding energy is then on the order of 0.1 eV. In practice, for shallow donors
and acceptors, the actual binding energy to ionize a donor or acceptor is on the order of 0.05 eV.
The slight preference for an electron to be localized on the donor atom, and the hole to be localized on the
acceptor atom, results in the formation of donor states approximately 0.05 eV below the conduction band edge

6
and acceptor states approximately 0.05 eV above the valence band edge. The donors are ionized when the donor
states are empty, and the acceptors are ionized when the acceptor states are occupied.
Donor and acceptor states are considered localized states and hence follow a slightly modified Fermi-Dirac
occupation probability distribution function. For these states, the probability that the energy level is filled is
given by
1
P localized (E, EF , g) =
states 1 (E−EF )/kT
1+ e
g
where g is the degeneracy of the states. In Si, g is typically 2 for donors like As and P, and 4 for acceptors like B.
One must be careful to recognize when the donors and acceptors are ionized. Donors are ionized when empty,
but acceptors are ionized when filled. This is detailed in the table below.
Filled Empty
Donors not ionized ionized
Acceptors ionized not ionized
As donors are ionized when the donor state is empty, the concentration of ionized donors is given by
h i
+
ND = ND 1 − P localized (ED , EF , g)
states

ND
=
1 + g e(EF −ED )/kT
+
where ND is the concentration of donor impurities and ND is the concentration of ionized donors.
For acceptors, the impurity is ionized when the states is filled so the concentration of ionized acceptors is
given by
h i
NA− = NA P localized (EA , EF , g)
states

NA
=
1 (EA −EF )/kT
1+ e
g

where NA and NA− are the concentration of acceptor impurities and the concentration of ionized acceptors re-
spectively.
Under a wide range of conditions, the Fermi level will fall between the acceptor and donor levels (EA < EF <
ED ) and one can invoke the full ionization approximation where ND +
≈ ND and NA− ≈ NA .

1.12 Charge Balance

Every excess electron in the conduction band (n) must come from either an ionized donor (electron transfer to
conduction band), or it must leave behind an unoccupied state in the valence band (a hole p). Similarly, holes
come either as a result of an electron promoted into the conduction band or from ionization (electron capture) of
an acceptor atom. This charge balance can be very simply expressed as
+
ND − NA− = n−p

where n and p are the concentration of electrons and holes, ND +

is the concentration of ionized donors and NA− is
the concentration of ionized acceptors.
By the mass action law, normally only either n or p are significant, since the product np must be a constant.
In Si at room temperature n2i is 1020 cm−6 , so by the time the electron concentration reaches 1013 cm−3 , the
concentration of holes is a million times smaller and can often be neglected. The charge balance can be rewritten
as
n2i
+
ND − NA− = n−
n
n2i
NA− − ND
+
= p−
p

7
where the first is useful when the donor concentration exceeds the acceptor concentration, and the latter for
the inverse condition. The second term on the right hand side is often negligible leading to much simplified
calculations.

1.13 Effective Density of States without Boltzmann Approximation

The approximation of the Fermi-Dirac probability distribution by the Boltzmann approximation limits the validity
to conditions where the Fermi Level remains within the bandgap of a material. This can be relatively easily relaxed
by treating the integral more precisely with the Fermi-Dirac distribution, leading to what is referred to as the
Fermi Integral.
Consider the expression for the electron density without the Boltzmann approximation, but replacing the true
g(E) dE with the parabolic, spherical, wide band approximation.
3/2 Z ∞
2m⋆e (E − EC )1/2

n = 4π dE
h2 EC 1 + e(E−EF )/kT

We use the substituion x = (E − EC )/kT and dE = kT dx. For the exponential term in the denominator, it is
split into two terms

e(E−EF )/kT = e(EC −EF )/kT e(E−EC )/kT

This reduces the integral to

3/2 Z ∞
2πm⋆e kT x1/2

4
n = √ dx
π h2 0 1+ ex e(EC −EF )/kT
3/2 ∞
2πm⋆e kT x1/2
 Z
4
= √ e−(EC −EF )/kT dx
π h2 0 A + ex

where A is a constant equal to

A = e−(EC −EF )/kT

In the limit where EC − EF ≫ kT , A becomes negligible and the integral is readily evaluated
∞ ∞ −x
x1/2 x1/2
Z Z
dx ≈ dx
0 A + ex 0 e
  √
3 π
= Γ =
2 2

This recovers the expected Effective Density of States Approximation with the Boltzmann approximation.
3/2
2m⋆e

n = 2 e−(EC −EF )/kT
h2

However, the integral expression can be evaluated readily even when EC − EF is negative (A just becomes
very significant). It is perhaps useful to rewrite the correct integral as
3/2 Z ∞ 1/2
2πm⋆e kT
  
2 x
n = 2 e−(EC −EF )/kT √ dx
h2 π 0 A + ex

The first term is just the Boltzmann Effective Density of States estimate for the carrier concentration, and the
second integral term is a multiplicative correction factor. Unfortunately the integral must be done numerically
or as a series approximation. However, it should be obvious that the value from the full expression will be less
than the Boltzmann estimate since A is positive definite.

8
1.14 Fermi-Integral
It is relatively easy to show that the number of electrons in the conduction band in the parabolic, spherical, wide
band and large crystal limit is given exactly by the integral:
 
EF − EC
n = NC (T )F1/2
kT

where Fj (x) is the Fermi-Dirac integral given by

∞
tj
Z
1
Fj (x) = dt
Γ(j + 1) 0 1 + e−x ej

and NC (T ) is the conventional effective density of states. The generalization to Fj from F1/2 allows use in other
than 3 dimensions (for the free electron like density of states).
For large negative x, it is also easy to show that the Fermi-Integral approaches the Boltzmann limit

lim Fj (x) = ex
x→−∞

Rewriting the integral slightly, we can obtain an expansion using the Taylor series for 1/(1 + x) = 1 + x −
x2 + x3 − . . ..
Z ∞
1 tj
Fj (x) = dt
Γ(j + 1) 0 1 + e−x et
Z ∞
ex tj
= dt
Γ(j + 1) 0 et + ex
Z ∞
ex tj
≈ dt
Γ(j + 1) 0 et [1 + e−t ex ]
Z ∞ j "  x 2  x 3 #
ex t ex e e
= 1− t + − + . . . ] dt
Γ(j + 1) 0 et e et et
Z ∞ j Z ∞ j Z ∞ j Z ∞ j
ex

t x t 2x t 3x t
= dt − e dt + e dt − e dt + . . .
Γ(j + 1) 0 et 0 e2t 0 e3t 0 e4t
∞ j ∞ j ∞ j ∞ j
ex ex e2x e3x
Z Z Z Z 
t t t t
= dt − j+1 dt + j+1 dt − j+1 dt + . . .
Γ(j + 1) 0 et 2 0 et 3 0 et 4 0 et
ex e2x e3x e4x
 
= ex 1 − j+1 + j+1 − j+1 + j+1 + . . .
2 3 4 5
∞
X (−ex ) n−1
= ex
n=1
nj+1

The series converges rapidly as long as ex is sufficiently small. If used for x < −0.3, 100 terms will give 15 decimal
digits of precision (obviously fewer needed for more negative values). For x < −35, the Boltzmann approximation
is accurate to 15 digits of precision.
For large positive x, the denominator in the integrand is unity for t < x, and we can obtain the asymptotic
expression
Z ∞
1 tj
Fj (x) = dt
Γ(j + 1) 0 1 + e−x et
Z x
1
≈ tj dt
Γ(j + 1) 0
1
= xj+1
(j + 1)Γ(j + 1)
xj+1
=
Γ(j + 2)

9
1.15 Where is the Fermi Level
In equilibrium and in the absence of any external influences, the distribution of electrons in the varies possible
electron states is determined by the position of the Fermi Level. However, one can also take the view that the
Fermi Level will adopt that position that satisfies the known distribution of carriers. Simply put, the actual
occupation of electron states must be consistent with the Fermi Level.
Given a density of donor and acceptor impurities, the Fermi Level EF will adjust itself such that the following
equations are simultaneously satisfied.
Z
n = g(E)PF D (E)dE ≈ NC (T )e−(EC −EF )/kT
Z
p = g(E)[1 − PF D (E)]dE ≈ NV (T )e−(EF −EV )/kT

+ ND
ND =
1 + ge(EF −ED )/kT
N
NA− =   A
1 (EA −EF )/kT
1+ e
g
+
ND − NA− = n−p
While it may not be immediately obvious, there will be a solution for physical situations.
There are several approximations that make determining the Fermi Level simpler. In many cases, we know
either n or p. For example, given ND = 1015 cm−3 , it is reasonable to “guess” that the full ionization approxi-
+
mation is valid so ND = ND . And for Si at room temperature where n2i = 1020 cm−6 , the hole concentration will
+
be negligible compared to the electron concentration. Charge balance hence give n = ND = ND . Using the first
expression, we can then determine the Fermi Level position relative to the conduction band edge
n ND
EC − EF ≈ −kT ln = −kT ln
NC NC
" −3 #
1015 cm
= −(0.02586 eV) ln
2.82 × 1019 cm−3
= 0.265 eV
The Fermi level will be 0.265 eV below the conduction band edge. Once one has this position, it is critical to verify
the assumptions. In this case, we assumed the full ionization approximation which requires that ED − EF ≫ kT ,
which would be satisfied for donors where EC − ED ≈ 0.05 eV.
Each situation is different, but similar concepts must be used. As the concentration of donors (or acceptors)
+
increases, the full ionization approximation fails and one must deal with ND < ND . Similarly, at extremely high
concentrations the Fermi Level will rise above the conduction band edge and the Boltzmann Effective Density of
States approximations becomes invalid.
Note: One big mistake is to forget that temperature plays a big role in many of these equations. NC and
NV scale as T 3/2 , and n2i is exponential in temperature.
Programs such as Fermi.exe deal with solving the full set of equations with all of the constraints.

1.16 Minority Carrier Lifetime

If the concentration of electrons and holes is out of equilibrium, they will return to equilibrium either by thermal
excitation of electrons from the valence band to the conduction band (creating a hole and an electron), or by
recombination of an electron in the conduction band with a hole in the valence band.
The dynamics of the recombination process will be dominated by minority carrier concentration since there
will be lots of majority carriers present to recombine with the few minority carriers. The mean time for a minority
carrier to recombine with its counterpart is called the minority carrier lifetime τ , normally designated τe or τh
for electrons and holes respectively. The rate of recombination per unit time is then n/τe or p/τh for electrons
and holes.
The minority carrier lifetime is primarily determined by the concentration of the majority carriers since each
is a potential suitor for the minority carrier. It is also governed by the quality of the material (purity and defects),

10
temperature and the presence of specific carrier traps and recombination centers. Values range from picoseconds
in materials designed as high speed detectors to seconds in extremely pure and undoped semiconductors.

11