You are on page 1of 94

Fundamentals

of
Semiconductors

Unit 3: Equilibrium Carrier
Concentrations

Mark Lundstrom
Purdue University
West Lafayette, IN

1
Preface:
This is the full set of lecture slides with notes for Unit 1 of a course on the fundamentals of
semiconductor physics. The course is designed to be broadly accessible to students in any
branch of science or engineering who would like to understand basic semiconductor
physics. Those who use semiconductor devices will gain an understanding of the physics
that underlies the operation of devices. Semiconductor technology developers may find it a
useful starting point for diving deeper into condensed matter physics, statistical mechanics,
thermodynamics, and materials science. The course presents an electrical engineering
perspective on semiconductors, but those in other fields may find it a useful introduction to
the approach that has guided the development of semiconductor technology for more than
50 years. The treatment is physical and intuitive, and not heavily mathematical.

Unit 1: Materials properties and doping
Unit 2: Rudiments of quantum mechanics
Unit 3: Equilibrium carrier concentrations
Unit 4: Carrier transport, generation, and recombination
Unit 5: The semiconductor equations

Unit 3: Equilibrium carrier concentrations

L3.1 The Fermi function
L3.2 Fermi-Dirac integrals
L3.3 Carrier concentration vs. Fermi level
L3.4 Carrier concentration vs. doping density
L3.5 Carrier concentration vs. temperature

Unit 3 Learning Objectives:



Upon completion of Unit 3, you will be able to:

• Explain what the Fermi function is, how it is used, and the effect of the location of the
Fermi level and temperature on the Fermi function.
• Understand the relation between the Fermi level and the equilibrium carrier densities for
degenerate and nondegenerate semiconductors.
• Relate the intrinsic carrier density to the effective densities-of-states and to the bandgap
and understand what the intrinsic Fermi level (intrinsic level) is.
• Relate the equilibrium carrier densities to the location of the Fermi level relative to the
conduction and valence band edges and to the intrinsic level.
• Compute the equilibrium carrier densities given the donor and acceptor concentrations
and the temperature.




2
After studying these materials, you will be introduced to the following terms:

1) Fermi level 12) Extrinsic semiconductor
2) Fermi function 13)Intrinsic semiconductor
3) Nondegenerate 14) Donor degeneracy factor
4) Fermi-Dirac integral 15) Acceptor degeneracy factor
5) Effective DOS 16) Metal-insulator transition
6) np product
7) Intrinsic concentration
8) Intrinsic Fermi level
9) Majority carrier
10) Minority carrier
11) Freeze-out

3
Those who wish to dive deeper into the concepts introduced in this course, may wish to
consult the following two texts.

Advanced Semiconductor Fundamentals (2nd edition) by R.F. Pierret, Pearson
Education, Inc., 2003 [ISBN-0-13-061792-X (paperback)]

Semiconductor Physics: Principles, Theory, and Nanoscale by Sandip Tiwari, Oxford
Electroscience Series, Vol. 3, 2020 [ISBN 978-0-19-875986-7]


Specific recommendations for Unit 3 are listed below.

L3.1 The Fermi function

Pierret, Ch. 4 (4.2)
Tiwari, Appendix E


L3.2 Fermi-Dirac integrals

Pierret, Ch. 4 (4.1.1)


L3.3 Carrier concentration vs. Fermi level

Pierret, Ch.4 (4.5.2)


L3.4 Carrier concentration vs. doping density

Pierret, Ch.4 (4.5.2)


L3.5 Carrier concentration vs. temperature

Pierret, Ch.4 (4.5.2)



4
The first two units of this course have provided the background needed
to dive in to some real semiconductor problems. The topic of Unit 3 is
understanding how we compute the concentrations of electrons and
holes in semiconductors under equilibrium conditions. Before we can do
this, we need to understand the Fermi function, which is the subject of
this lecture.

1
To begin, let's go back to an isolated silicon atom with its discrete
energy levels. We have 14 electrons and fill up the various s, p, d, and f
orbitals associated with the principal quantum numbers n = 1, 2, and 3
to accommodate them. As shown above, we can draw a line, an energy,
such that below the line the states are filled and above the line, the
states are empty. Of course, there is no need to draw this line for the Si
atom, because we can simply put the 14 electrons in, and we know
exactly which energy levels are filled and which ones are empty, but it
will turn out that this line, known as the Fermi energy, is very useful
when the states are distributed in energy, as in a semiconductor.

2
In semiconductors, there are bands of energies, not energy levels.
Shown above is the density of states versus energy for the top of the
valence band and bottom of the conduction band. We saw in Unit 2
how to compute the density of states. We're only interested in the
density of states near the bottom of the conduction band because that's
where the electrons are; it goes as the square root of energy. We also
only need the density of states near the top of the valence
band, because that's where the holes are. That goes as the square root
of energy too. The conduction and valence band densities of states
actually have more complicated shapes and go to zero at the top of the
conduction band and at the bottom of the valence band, but that doesn't
need to concern us because in equilibrium and near equilibrium, the
carriers are near the band edges.

Now, we can draw a line that we will call the Fermi level (or Fermi
energy). It's sometimes also called the electrochemical potential, or
sometimes just the chemical potential. Below the line, the states will
be mostly filled, and above the line, the states are mostly empty. For the
rest of the course, an important question will be: “Where is the Fermi
level in this semiconductor? ”

3
To continue our discussion…

It turns out that there is a very simple function that describes the
probability, a number between 0 and 1, that a state at an energy, E, is
occupied by an electron. That function is the Fermi function, f0. The
subscript zero is to remind us that this function applies to equilibrium
conditions.

The probability in equilibrium that a state at an energy, E, is occupied is


1 / (1 + exp [ (E – EF)/kBT]), where EF is the Femi level. As we will
discuss later in the course, we can also compute this probability out of
equilibrium. In Unit 3, we assume that the semiconductor is in
equilibrium, so f0 is the probability that energy states are occupied.

4
We won’t derive the Fermi function in this course, but you can see that
it makes sense. Electrons like to occupy the lowest energy states, so it
makes sense that the states with low energy should have a high
probability of being occupied. States with high energies are less
occupied.

To get comfortable with the Fermi function, we can plot it as shown


above on the right. The Fermi function goes between 0 and 1, as a
probability should. At energies much lower than EF, then we have exp
[large negative number], which is close to 0, so you can see that f0
approaches 1. At energies much higher than the Fermi energy, then we
have exp [large number], so you can see that f0 approaches 0. For a
state located at the Fermi energy, E = EF, and we can see that the
probability that a state at the Fermi energy is occupied is 1/2.

To summarize, states below the Fermi energy have a small probability


of being empty, and states above the Fermi energy have a small
probability of being filled.

5
Let’s flip the Fermi function on its side and plot f0 vs. E as shown
above. At the Fermi energy, the probability that a state is occupied if a
state exists is one-half. As shown above, the Fermi function transitions
between 1 and 0 in a range of energies centered at EF.

6
As shown above, we can plot the Fermi function vs. energy at different
temperatures. If we increase the temperature, the width in energy of the
transition from 1 to 0 gets larger. As we decrease the temperature, the
width of the transition region gets narrower. If we go all the way to T = 0
K, the Fermi function becomes a step function. All the states below the
Fermi level are filled, and all the states above the Fermi level are empty.

We can see that temperature has important effects on the Fermi


function.

7
A question that we will often ask is: “Where is the Fermi
level?” Typically, it will be somewhere inside the bandgap. For a very
heavily doped N-typed semiconductor it might get a little ways into the
conduction band, and for a very heavily doped P-type semiconductor, it
might get a little ways into the valence band, but we’ll generally find it in
the shaded region above. States way up in the conduction band that are
above the Fermi energy will be empty, and the states deep in the
valence band way, below the Fermi energy, will be filled. Near the
bottom of the conduction band there's a small probability that some
states might be filled, and near the top of the valence band, there's a
small probability that some of the states might be empty. This is what
gives rise to the holes and to the electrons in a semiconductor.

8
Let's apply the Fermi function to the conduction band, and assume that
the Fermi energy is located in the forbidden gap, as shown above. The
probability that a state at EF is occupied is one-half, but there are no
states in the forbidden gap so there are no electrons there. You can see
that the probability that a state in the conduction band is occupied is
small because all of the states in the conduction band are well above
EF.

There is an important approximation to the Fermi function when the


probability that states are occupied is small (i.e. when the energies of
the states are well above EF). A nondegenerate N-type
semiconductor is one in which the conduction band states have a small
probability of being filled. As shown above, for a nondegenerate
semiconductor, f0 simplifies to an exponential, f0 ~ exp[EF –
E)/kBT]. The order of the arguments is easy to remember. The higher
the Fermi energy moves, the higher the probability that states in the
conduction band are occupied, so it’s (EF – E) in the argument of the
exponential. This is an approximation that we're going to make use of
frequently as we work out semiconducting problems.

9
There's a similar approximation that we can make for nondegenerate P-
type semiconductors. For a nondegenerate P-type semiconductor, there
is a small probability that the states in the valence band are empty. So,
if we look at the valence band, that’s well below the Fermi energy, the
energy is much less than the Fermi energy so there's a small probability
of being empty.

What we’re interested in now is not the probability that the state is
filled, but, rather, the probability that the state is empty. As shown
above, as long as the Fermi energy is a bit above the top of the valence
band, then (1-f0) ~ exp [(E – EF)/kBT]. The order of the arguments is
easy to remember. The lower the Fermi energy moves, the higher the
probability that states in the valence band are empty, so it’s (E – EF) in
the argument of the exponential.

10
In calculations, we will frequently assume nondegenerate
semiconductors. A nondegenerate semiconductor is one in which the
states in the conduction band are always well above the Fermi
energy, and the states in the valence band are well below the Fermi
energy. So as long as the Fermi energy lies in the forbidden gap and
doesn't get too close to the conduction band or too close to the valence
band, this nondegenerate assumption will be a good assumption to
make. When we make that nondegenerate assumption, then the
probability that a state at EC is occupied is just exp [(EF – EC)/kBT. The
probability that a state at the top of the valence band is empty is just
exp [(EV – EF)/kBT].

11
We will usually show the Fermi level when we draw an equilibrium
energy band diagram. Shown above is an energy band diagram for an
intrinsic semiconductoer. For an intrinsic semiconductor, there are equal
numbers of electrons and holes, so we expect the Fermi level to be
near the middle of the gap so that there's an equal probability of the
states in the conduction band being filled and the states in the valence
band being empty. It turns out it's not exactly in the middle because, as
we'll discuss later, the density of states in the valence band is slightly
different from the density of states in the conduction band.

12
If we see an energy band diagram with a Fermi level, our question will
be: “What are the electron and hole densities, n0 and p0?”

The nondegenerate approximation says that the probability of states in


the conduction band are filled is given by an exponential, as shown
above. Most of the electrons in the conduction band are near the
bottom of the conduction band, near EC, so we expect the electron
concentration to be proportional to exp [(EF – EC)/ kBT].

In the valence band, we're interested in the probability that the state is
empty. The nondegenerate approximation says that the probability that
states in the valence band are empty is gives by an exponential, as
shown above. Most of the holes in the valence band are near the top of
the valence band, so we expect the hole concentration to be
proportional to exp [(EV – EF)/ kBT].

And we work out the math in the next lecture, we'll see that that is
indeed the case.

13
This lecture was an introduction to the Fermi function. When we
compute carrier densities in the next lecture, we'll see how to make use
of the Fermi function.

The Fermi function is simply an expression that gives the probability


that a state at a certain energy, E, is occupied in equilibrium (if a state
exists at that energy). There are two parameters in the Fermi
function. One is the Fermi level, EF, and the second is the temperature,
T.

In the next lecture we will make use of this Fermi function as we begin
to compute carrier concentrations in semiconductors.

14
In the previous lecture, we discussed the Fermi function. We need to
understand the Fermi function because it is used to compute carrier
densities in semiconductors. When we do the calculation, we'll
encounter an integral known as the Fermi-Dirac integral. That integral
and how to simplify the result so that we obtain simple expressions that
relate the electron and hole concentrations to the location of the Fermi
level is to subject of this lecture.

1
Just a reminder that the Fermi function gives us the probability that a
state at energy, E, is occupied in equilibrium, if that state exists. It's
given by the expression shown above. The two key parameters are the
Fermi level EF and the temperature T.

2
Our goal is to understand how the electron density in the conduction
band, n0, and the hole density in the valence band. p0, depend o the
location of the Fermi level. We have shown that in the nondegenerate
approximation the probability that a state at the bottom of the
conduction band is occupied is exp [(EF – EC)/kBT], so we expect n0 to
be proportional to that.

We have also shown that the probability that a state near the top of the
valence band is empty is exp [(EV – EF)/kBT], so the hole concentration
should be proportional to that. These conclusions apply to
nondegenerate semiconductors, but, what is the constant of
proportionality, and what if the semiconductor is not nondegenerate?
Those questions will be answered in this lecture.

3
There are two ways to do this calculation. Our goal is to relate the
carrier density to the Fermi energy and to the other properties of the
semiconductor, and we can do the calculation in k-space or we can do it
in energy space. We’ll first discuss how we do it in k-space, but we
won't actually do the calculation. Then we'll discuss how to do the
calculation in energy space using the density of states, DOS(E).

4
Consider again a volume, OMEGA, of semiconductor, a cube carved
out of an infinite semiconductor. We want to calculate the number of
electrons in this volume. We saw earlier that when we apply periodic
boundary conditions, we get a finite number of k-states that are very
closely spaced. We just add up the probability that each one of those
states is occupied in equilibrium, to find the total number of electrons in
this volume. That is the first sum shown in the slide above. The volume
itself is arbitrary, so we’re really more interested in the carrier density –
the number per unit volume, so we would divide the result by volume.
That is the second equation shown in the slide above.
Since we are dealing with a significantly size volume, the k-states are
very finely spaced, so we can convert the sum to an integral as long as
we have the correct number of states in each element of k-space, dk.
The correct number will be given by the density of states in k-space, Nk,
that we have seen earlier, This is 2 for spin times volume,
OMEGA, over (2 pi)3.

Putting this all together, converting the sum to an integral, and then
performing the integral, we get the answer that we're after.

5
We won’t do that k-space calculation here, but you might give it a try
and see if you get the answers we're going to get by doing the
calculation in energy space. It's a calculation that proceeds straight
forwardly using the density of states in k-space as shown above and
then converting the sum over the finely spaced states in the volume
OMEGA into an integral. When doing this calculation, be sure to
include the valley degeneracy – e.g. remember that Si has six
equivalent valleys in the conduction band along the six directions, +x, -
x, +y, -y, +z, and -z.

6
We will do the calculation in energy space. To do this calculation, we're
going to need the density of states in energy. As shown above, we know
the DOS for parabolic energy bands 1D, in 2D, and in 3D. We’ll do the
calculation in 3D using the DOS that goes to zero at the bottom of the
conduction band and then increases as the square root of energy for
energies above the bottom of the conduction band. To test your
understanding, you can work our the corresponding results in 1D and
2D.

7
Let's first take a look at an energy in the conduction band and
ask: “How many electrons are there in the shaded energy range shown
above?” The number of states in that region, is DOS(E)dE. To find the
number of electrons in that region, I take the number of states in that
region, DOS(E)dE and multiply by the probability that those states are
occupied, f0(E).

For now, let’s make the nondegenerate approximation to the Fermi


function, to get the exponential approximation shown above. Remember
that DOS(E) goes as the square root of energy and then we conclude
that the number of electrons in dE, n(E)dE, is proportional to the
square root of energy times an exponential factor.

8
If we plot n(E)dE, we get the result shown above, which tells us where
the electrons are located in the conduction band. As the energy
increases, the square root starts to increase, and n(E)dE increases, but
as we continue to increase the energy, the exponential factor
dominates and n(E)dE begins to decrease. There is a peak in n(E)dE.
To find the energy of maximum n(E)dE, we differentiate it with respect
to energy and set it equal to zero. If you do that calculation, you'll be
able to show that this peak occurs at an energy that is 1/2 kBT
above the bottom of the conduction band. That's a relatively small
energy, If kBT at room temperature is 0.026 eV, the peak occurs at
0.013 eV above the bottom of the conduction band. The band gap is
1.11 eV, so this verifies my previous assertion that most of the electrons
in the conduction band are near the bottom of the conduction
band. They are actually are very close to the bottom of the conduction
band. Similarly the holes are located within about 1/2 kBT from the top
of the valence band.

We conclude that electrons and holes are located very near the band
edges.

9
Let’s now do the calculation and compute the electron density by
integrating the number of electrons in an energy range, n(E)dE, over
the entire conduction band. In this case, we will not make the
nondegenerate approximation, so we’ll use the Fermi function and not
its exponential approximation.
As seen in the first equation in the slide above, we will integrate from
the bottom of the conduction band to infinity. You might worry about our
density of states, the sqrt(E – EC) form only describes the DOS near the
bottom of the conduction band, so it's not going to be a
good approximation as we go to very high energies. But the Fermi
function helps out. There will be virtually zero probability that states well
above the bottom of the conduction band are occupied so it doesn't
matter that we make an error in the density of states.
We multiply the number of states, times the probability that the states
are occupied and integrate from the bottom of the conduction band to
infinity. We can see that the result is going to depend on Fermi level and
temperature, because those two parameters are in the Fermi
function. The result is also going to depend on the effective mass of the
semiconductor and the number of equivalent valleys because those
quantities are in the density of states.

10
It is convenient to make a couple of definitions, First, we’ll define a
normalized Fermi level with respect to the bottom of the conduction
band, (EF – EC ) in units of kBT; call this etaF. We'll also define a
normalized energy, (E – EC ) in units of kBT; call this eta. With these
definitions, we can rewrite the integral as shown in the third line on the
left in the slide above.

Now it's simply a matter of evaluating the integral and then we have our
answer. There's only one problem, that integral cannot be done
analytically. It has to be done numerically with Simpson's rule or
whatever your favorite technique for numerically evaluating integrals is.

We'll give our integral a name; it is called a Fermi-Dirac integral of


order 1/2. The terms that we have out front depend on the effective
mass and the temperature. Just as we expected that they would. We
give those terms a name, NC, the effective density of states.

We have derived the expression that we're after. Next, let's talk a little
bit more about this Fermi-Dirac integral.

11
The general definition of a Fermi-Dirac integral of order j is shown in the
shaded box above. The integral that we encountered when we worked
out the 3D carrier density for parabolic bands had a value of j = 1/2, but
it's simply an example of a more general function.
The Fermi-Dirac integral has a very nice property – if the normalized
Fermi energy is much less than zero (which means that the Fermi
energy is well below EC) then the FD integral simplifies and becomes an
exponential. This is a nondegenerate semiconductor. In a
nondegenerate semiconductor the Fermi-Dirac integral reduces to an
exponential. There's also a nice property related to the derivative.
In this course, we will for the most part assume nondegenerate
semiconductors and avoid FD integrals. If you should need to make use
of FD integrals, the notes listed on the slide above contain all you need
to know about using Fermi-Dirac integrals in semiconductor problems.
Finally, when you are reading papers or books, be careful. The Fermi-
Dirac integral is sometimes defined without the factor out front that
contains a GAMMA function. Without that factor, the FD integral is
written as a roman F; with the factor, is it written as a script
F. Sometimes people aren’t careful about which F they use, so be
certain that you understand which version of the Fermi-Dirac integral is
being used.

12
Let’s take a closer look at the Fermi-Dirac integral of order 1/2, that
occurs when we work out the carrier density in 3D. That function,
scriptF1/2 vs. etaF is plotted above. When etaF < 0, the Fermi level is
below the conduction band, and the semiconductor is
nondegenerate. When etaF > 0, the Fermi level is above the bottom of
the conduction band and the semiconductor is degenerate.

As shown in the plot above, for nondegenerate semiconductors, the FD


integral approaches the exponential, which simplifies matters
greatly. For a given argument, etaF, the FD integral is always equal to or
less than the corresponding exponential.

For a nondegenerate semiconductor, there is a very simple relation


between the carrier density and the Fermi level. It's just the exponential
relation shown above at the lower right. This is the expression that we
will use in the remainder of this course to relate the carrier
concentration to the Fermi level.

13
Just a reminder that a nondegenerate semiconductor is one for which
the Fermi level stays within the shaded region shown above. It doesn't
get too close to the bottom of the conduction band or too close to the
top of the valence band. We’ll sometimes use the exponential
approximation to the FD integral outside of the safe region, but we’ll
understand that the results will only be approximate.

14
We have succeeded in relating the electron density to the Fermi energy.
In general, that relation involves Fermi-Dirac integrals, as shown in the
shaded box above, but these are a little complicated, because we need
to do them numerically. For a nondegenerate semiconductor the
relation becomes much simpler, the Fermi-Dirac integral reduces to an
exponential and the electron density becomes the effective density of
states, NC, times exp [(EF – EC)/kBT].

15
We could go through the same kind of calculation for holes, but in this
case, we are interested in the probability that the state is empty. When
we do the calculation, we get answers that look much like the electron
density results. We get an effective density of states for the valence
band times a Fermi-Dirac integral that depends on the normalized Fermi
energy. In this case, the normalized Fermi energy is how far the Fermi
energy is below the top of the valence band in units of kBT. The
effective density of states for the valence band, involves the effective
mass for holes. For a nondegenerate P-type semiconductor, the hole
density becomes the effective density of states, NV, times exp [(EV –
EF)/kBT].

16
To summarize, we have discussed how one uses the Fermi function,
and the density of states to to relate the concentration of electrons in
the conduction band and holes in the valence band to the location of the
Fermi energy. In general (for parabolic bands), these relations involve
a Fermi-Dirac integral of order 1/2, but if the semiconductor is
nondegenerate the Fermi-Dirac integrals simplify to exponentials and
we are left with the two simple expressions shown above. You should
remember these, because, we're going to use them over and over and
over again.

It’s easy to remember the order EF and EC in the expression for the
electron density because the higher the Fermi level, the more probable
it is that states in the conduction bands are occupied. It’s also easy to
remember the order EV and EF in the expression for the hole density,
because the lower the Fermi level, the more probable it is that states in
the valence band are empty.

In the next lecture, we will discuss in a little more detail, how the carrier
concentrations vary as a function of the Fermi level.

17
In the previous lecture, we learned how to relate the charge carrier
densities, the electron and hole concentrations, to the position of the
Fermi level. In this lecture, we continue that discussion.

1
To review, for the electron density, we will focus on the nondegenerate
expression shown above. There is a corresponding expression for the
hole density, which is also shown above.
The effective density of states for the semiconductor is related to the
density of states effective mass of the semiconductor. Our focus in this
lecture is on using these nondegenerate expressions.

2
To illustrate how to use these expressions and to calibrate ourselves on
the general size of the numbers, we'll be doing many of our
calculations and examples for silicon. The density of states effective
mass and the corresponding effective density of states for electrons in
the conduction band of silicon are shown above. These numbers will be
used for the examples in this lecture.

3
For holes in silicon, there is a different effective mass, so we get a
different effective density of states. The numbers shown above are the
ones that we will use as we do some calculations and examples in this
lecture.

4
Let's look now at example shown above on the right. The Fermi level is
0.2 eV below the bottom of the conduction band, so EF - EC is -0.2 eV.
The thermal energy is kBT = 0.026 eV at room temperature. We multiply
the exponential by the effective density of states, put in numbers, and
we find the electron concentration. This would be considered to be a
moderate electron concentration.

5
How do we get the hole concentration? Well, the Fermi level is located
well above the top of the valence band, so we don't expect there to
be too many holes. We can use our expression for p0 above
recognizing that EV - EF is a large negative number. Plugging numbers
in, we get a very small hole concentration.

6
It is easy to go from the Fermi level to the carrier concentration, but
sometimes, we may be given the carrier concentrations and asked for
the Fermi level. That is straightforward too. We just solve the equations
in the shaded box for the Fermi level.

For an N-type semiconductor, we might be given the electron


density. We can then determine where the Fermi level is that resulted in
that electron density simply by using first expression above on the right.

For a P-type semiconductor, we might be given the hole density. We


can then determine where the Fermi level is that resulted in that hole
density simply by using second expression above on the right

So we can go either way. Given n0 or p0, we can compute EF. Given EF,
we can compute n0 and p0.

7
The product of the electron and hole concentrations in equilibrium is a
very important quantity that we will discuss next.

8
We have expressions for the equilibrium electron and hole
densities, and we know what the effective densities of states are, so
let's multiply the two together. As shown above, the Fermi level drops
out of the product. In equilibrium, the np product is independent of
the location of the Fermi level. (This is the case for the
nondegenerate semiconductor that we are assuming.)
As shown above in the third equation on the left, the equilibrium np
product depends on the electron and hole effective masses (which are
in the effective densities of states) and on the band gap of the
semiconductor. For an intrinsic semiconductor, we know that n0 = p0 =
ni, so in equilibrium , n0p0 = ni2. Since n0p0 doesn't depend on the Fermi
level, n0p0 = ni2 for a doped semiconductor as well – n0p0 always equals
ni2 for a semiconductor in equilibrium.
We discussed the intrinsic carrier concentration earlier in the course,
and we pointed out that it depends exponentially on band gap and
temperature. We've now derived an expression that allows us to
compute ni from the material parameters of the semiconductor. As
shown by the last equation on the left, ni does, indeed, depend
exponentially on the band gap and on the temperature. This is a very
important result.

9
It is important to remember that the equilibrium np product is
independent of where the Fermi level is located. That is not the case
when we have a degenerate semiconductor; then we have to use
Fermi-Dirac integrals.

We've learned that the equilibrium np product (ni2) depends


exponentially on band gap. The wider the band gap, the harder it is to
break covalent bonds and produce intrinsic carriers. It also depends
exponentially on temperature. The higher the temperature, the easier it
is for the thermal energy to break covalent bonds. And for silicon, if we
plug numbers in and are careful about how we do the calculation, we
find that ni is about 1010 per cubic centimeter, the number that we've
been using in earlier lectures.

10
Let’s back up now and redo a problem that we did earlier on slide 6, but
do it a different way. Remember, for this problem, we had an N-type
semiconductor with a Fermi level near the conduction band. We first
computed the electron density, and then we computed the hole density
using the expression shown above, which relates the hole
concentration to location of the Fermi level with respect to the top of the
valence band. Doing the math, we found p0 = 1.14 x 104 holes per cubic
centimeter. We can do this problem a different way.

11
There is actually an easier way to do this problem. We begin as we did
earlier, by computing the electron concentration since it's an N-type
semiconductor, but then we use the fact that n0p0 = ni2, and we know
ni2. We find that p0 = ni2/n0, which gives us 0.68 x 103 holes per cubic
centimeter
But wait a minute. When we did it first way, which should be
equivalent, we got a different answer, 1.14 x 104. If round both of them
off they're about 104 holes per cubic centimeter. That's 12 orders of
magnitude less than the electron concentration, so that’s not really too
bad, but why are they a little different?
To see the reason, plug in the numbers that we've been using for NC,
NV, and EG and compute ni. You will see that it doesn't give us exactly 1
x 1010. The intrinsic carrier concentration is very sensitive to these
parameters. People who need an accurate intrinsic carrier
concentration spend a lot of time worrying about temperature-
dependent band gaps, temperature-dependent effective masses, etc.,
and they make small corrections for details, like an excitonic correction
to the band gap. So the band gap and effective densities of states we
used don’t give us exactly the ni we assumed. That’s why we get two
different answers.

12
To continue the discussion…

Both approaches give us about the same answer. I bring this up just to
sensitize you to the fact that ni is difficult to compute exactly. If you're
encountering a problem, ask yourself: “Do I have more confidence in
the value of the intrinsic carrier density or in the effective densities of
states and band gap?” then choose the approach that you have more
confidence in.

13
The np product is an important quantity to understand. Reading an
energy band diagram and determining whether the semiconductor is N-
type or P-type is also important. If we look at an energy band diagram
and see the Fermi level up near the conduction band, we know that
there will be more electrons than holes, so this would be an N-type
semiconductor, but n0p0 is still ni2.

14
If we look an energy band diagram and see a Fermi level near the top
of the valence band, we know that there will be more holes than
electrons, so this is a P-type semiconductor.

15
If we look at an intrinsic semiconductor, we know that the Fermi level
will be about in the middle of the gap, and there will be equal numbers
of electrons and holes. But is the Fermi level exactly at the middle of the
gap?

16
We can calculate the location of the intrinsic Fermi level (or just
intrinsic level).

We know how the electron density is related to the Fermi level, and we
know how the hole density is related to the Fermi level. We also know
that in an intrinsic semiconductor, the electron density is equal to the
hole density, and both are equal to ni, a known quantity. By convention,
we will call the Fermi level under intrinsic conditions Ei instead of EF; ,Ei
is the intrinsic Fermi energy.

If we do the calculation, we get the result shown above on the lower left.
We find that Ei is offset from the middle of the band gap by a small
correction that depends on how much different the effective density of
states in the valence and conduction bands are. Remember that those
two quantities depend on the effective masses, so the Fermi level is
either a little above or a little below the middle of the band
gap depending on the relative values of the effective masses.

Can you explain why different effective masses require that the intrinsic
Fermi level not be exactly at the center of the forbidden gap?

17
Let's work out some numbers for silicon. We know what the effective
densities of states are, so we can compute the correction term and
see how far away from the middle of the band gap we are. Putting
numbers in, we find that for Si, the intrinsic level is just 0.007 eV below
the middle of the band gap. The intrinsic Fermi is, indeed, located very
close to the middle of the band gap. There are some semiconductors
like gallium arsenide where there's more of an asymmetry between the
conduction band and the valence band, but we take the logarithm of
(NV/NC), so the intrinsic Fermi level never gets very far from the middle
of the band gap.

Remember: The intrinsic level is very near the middle of the band gap.

18
Knowing the intrinsic level, we can relate the carrier densities to the
Fermi level in a different way. The two relations on the upper left relate
the carrier densities to the Fermi level and they're perfectly adequate,
but there's another way to do this that's commonly done and it is useful
also.
So, for example, we know that ni is related to the Fermi level (the
intrinsic level). By setting EF = Ei in the first two equations on the left,
and by setting n0 and p0 equal to ni, we can solve these equations for
NC and NV, and eliminate NC and NV in the first two equations. The result
is the two equations shown above at the right.
Frequently, we know ni for a semiconductor, so expressing the carrier
densities in terms of ni and the Fermi level with respect to Ei is
convenient. As shown by the equations at the right, if the Fermi level is
above the intrinsic level, there are more electrons, and if the Fermi level
is below the intrinsic level, there are more holes.
These two expressions at the upper left and the two on the right are
equivalent. We'll use one or the other depending on which one is more
convenient for the problem at hand. You should get used to being able
to work back and forth and use either one of these approaches.

19
It is easy to read an energy band diagram and determine whether we
have an N-type or a P-type semiconductor. If the Fermi level is above
the middle of the gap (the intrinsic level), it's N-type. If the Fermi level is
below the middle of the gap, it's P-type. If the Fermi level is closer to the
conduction band than to the valence band, it's N-type. If it's closer to the
valence band than to the conduction band, it's P-type. So very easy to
read an energy band diagram quickly and determine whether we're
dealing with an N-type or a P-type semiconductor.

20
To summarize, shown at the top in the slide above are two equivalent
ways to related the equilibrium electron and hole concentrations to the
location of the Fermi level.

As shown in the gray box above, we also have expressions for the
effective densities of states.

As also shown in the gray box above, we developed an expression for


the intrinsic carrier density, and we found that the n0p0 product is always
equal to ni2 – independent of Fermi level for a nondegenerate
semiconductor.

We will continue this discussion of carrier densities in the next lecture,


which is about how the carrier concentrations vary with doping density.

21
Let's continue our discussion about carrier concentrations in
semiconductors under equilibrium conditions. The topic for this lecture
is on how the carrier concentration depends on doping.

1
Just as a reminder that we have developed the expressions shown
above that relate the Fermi level to the electron concentration, and the
Fermi level to the hole concentration, and we are focusing, as is
typically done in semiconductor work, on using the non-degenerate
approximation to the Fermi Dirac integral, realizing that from time to
time, if we're solving real problems, we may have to do the extra
work of evaluating Fermi-Dirac integrals.

2
Let’s begin by discussing what we mean by space charge
neutrality. Think about a bulk, uniformly doped semiconductor. Let’s
look at a region of the semiconductor shown in white above. In this
region, there are some ionized donors with a positive charge, and there
are some mobile electrons with a negative charge. In this region, there
are also some ionized acceptors with a negative charge, and there are
some positively-charged holes.

The electrons and holes, with concentrations, n and p, are mobile and
free to move around. The dopants, with concentrations ND and NA, are
fixed in the lattice and cannot move unless we heat the lattice up to very
high temperatures (e.g. on the order of 1000 degrees centigrade).

What is the net charge in the region of the semiconductor shown in


white? It's simply the magnitude of the charge on an electron or hole, q,
times the concentration of positively charged holes minus the
concentration of negatively charged electrons, plus the concentration of
positive charges due to ionized donors, minus the concentration of
negative charges due to ionized acceptors.

3
You may have heard the phrase: “nature abhors a vacuum.” Nature also
doesn't like a charge imbalance. It wants the net charge to be zero. So
what happens is that the mobile charges will move around, positive
charges will attract negative charges, and in the end for this uniform
semiconductor the net charge should be zero. The charge due to the
ionized dopants has been canceled by the mobile charges. The result is
that we can assume that for our uniformly doped semiconductor, the
net charge is zero. That's will be our starting point for our
calculations. If the semiconductor is not quite uniform, if it's got a
gradient in the doping profile for example, then it might not be exactly
neutral. Sometimes we talk about quasi-neutrality and make this
approximation anyway. If there's a strong nonuniformity like in PN
junctions that we use for building some types of devices, then it won't
be neutral at all and we won't be able to use this approach. But for a
uniformly doped semiconductor this will be our starting point.

4
We will also assume in this lecture that the dopants are fully
ionized, every donor has donated its electron to the conduction
band, and every acceptor is negatively charged because an electron
from the valence band has occupied each acceptor. The donors have
positive charges and the acceptors have negative charges. There are
mobile carriers too. We will assume that ND+. the concentration of
ionized donors, is equal to the concentration of donors. ND. The
concentration of ionized acceptors, NA-, is assumed to be equal to NA,
the concentration of acceptors.

5
To compute the carrier densities from the given doping densities, we will
begin by assuming space charge neutrality, the equation in the shaded
box above. We will also we assume fully ionized dopants in this lecture,
which is the second equation on the left. Finally, we know that n0p0 is
ni2.

So, assuming that we know the number of dopants, the two equations
in the middle of the slide above are two equations in two unknowns –
the electron and hole concentrations. We just need to solve these
equations.

6
As shown above, we start with the statement of charge neutrality, step
1). Next, we assume fully ionized dopants, step 2). Then we recall the
fact that the equilibrium np product is equal to ni2, step 3) . The result
shown in step 4) is that we can either eliminate p0 and get an
expression just in terms of n0, or we can eliminate n0, and get an
expression in just terms of p0.

We end up with quadratic equations that can be solved for n0 or p0.


Generally, we would solve for p0 in a P-type semiconductor and for n0 in
an N-type semiconductor.

7
The equation for n0 is shown above. Solving the quadratic equation, we
get an equation for the equilibrium electron density, n0.

We then compute the hole density from n0p0 = ni2.

8
The calculation for a P-type semiconductor proceeds similarly. The
equation for p0 is shown above. Solving the quadratic equation, we get
an equation for the equilibrium hole density, p0.

We then compute the electron density from n0p0 = ni2

9
Let's do a couple of examples. Let's begin with silicon doped with
moderate concentration of phosphorous. We’ll also assume that the
temperature is 300 K, so it’s safe to assume that the dopants are fully
ionized. What are the electron and hole concentrations?

Remember that at 300 K, ni is about 1010 per cc. We can evaluate the
expression, but you can see that for ND >> ni and NA = 0, it reduces to
n0 = ND.

To find the hole concentration, it's just ni2/n0.

Note that this is an extrinsic semiconductor, ND >> ni. We really don’t


need to solve the quadratic equation for problems like this.

10
Here is another example with the same concentration of N-type
dopants, but now we have some P-type dopants. Again, we want to
determine the electron and hole concentrations

We could evaluate the quadratic expression, but you can see that for
(ND – NA) >> ni , so it simplifies to n0 = ND -NA.

To find the hole concentration, it's just ni2/n0 again.

This is also an extrinsic semiconductor because ND -NA >> ni. We


don’t need to solve the quadratic equation for problems like this.

For problems like this, the majority carrier concentration is just the net
doping density and the minority carrier concentration is obtained from
n0p0 = ni2.

11
The conclusion is that when the net doping density is much greater than
the intrinsic carrier concentration and the dopants are fully ionized, then
if it's an N-type sample, the electron concentration is equal to the net N-
type doping, and we determine the minority hole concentration from the
n0p0 product.

If it's a P-type sample, the hole density is equal to the net P-type
doping, and we get the minority electron concentration from the n0p0
product.

12
The carrier concentration vs. temperature is shown above. In this
lecture, we have been working in the extrinsic region, where the
temperature is moderate, the dopants fully ionized, and where the
dopants overwhelm the intrinsic carriers.

It is very easy to calculate the carrier concentrations in the extrinsic


region. This is the region that we will stay in for most of the remainder of
the course, but we will sometimes encounter low temperatures and
sometimes encounter high temperatures. How to compute the carrier
concentrations at low temperatures and at high temperatures is the
subject of the next lecture.

13
Our goal in Unit 3 is to understand equilibrium carrier concentrations in
semiconductors. As we wrap up this unit, there's one more topic that
we need to discuss – how the carrier concentration varies with
temperature.

1
Earlier in the course, we discussed qualitatively how the carrier
concentration varies with temperature. At low temperatures, the
dopants freeze out, and there are no carriers. In the previous lecture,
we discussed the extrinsic region. There is also an intrinsic
region, where the intrinsic carriers overwhelm the dopants. This lecture
is a more quantitative discussion of the freeze out and intrinsic regions,
but first, let’s quickly review the extrinsic region.

2
The extrinsic region is a simple, nice region to operate in. In this region,
the dopants are fully ionized (the temperature is sufficient to do that),
and the net doping density overwhelms the intrinsic carriers so that we
don't need to worry about the intrinsic carriers.

3
Under extrinsic conditions, it's very simple and easy to calculate the
electron and hole densities. The dopants overwhelm the intrinsic
carriers, so the electron density is equal to the net N-type doping
density, and the hole concentration is ni2/n0. The story is similar for P-
type semiconductors.

4
Let's shift our attention now to higher temperatures, the intrinsic
region. We've seen, as we developed an expression for the intrinsic
carrier concentration, that the intrinsic carrier concentration
increases exponentially with temperature, so as we increase the
temperature, we'll get to a point where the intrinsic carriers, created by
breaking covalent bonds and exciting electrons across the
bandgap, can overwhelm the dopants. The dopants continue to be fully
ionized, but the net doping density is not large compared to the intrinsic
carrier concentration anymore.

5
It’s not difficult to calculate the electron concentration at high
temperatures. We start with the quadratic equation we developed and
solve it for the electron density. The only difference is, we cannot ignore
ni2. So, the calculation gets a little more complicated.

Once we’ve calculated the electron density, we get the hole density
from ni2/n0 using the appropriate ni for the higher temperature.

6
It’s the same for P-type semiconductors, we solve a quadratic equation
for the holes, including the intrinsic carrier density.

7
Let's do an example and see how this works. We have been looking at
an example with Si doped at ND = 2 x 1015 donors per cc, but let's make
the temperature 600 K now. What are n0 and p0?

At 300 K, ni was 1010 and could be ignored, but at 600 K, it will be


significantly higher. We could compute it from the intrinsic carrier
concentration formula shown above. Computing it actually requires
some care. This formula probably would not give us the correct
answer, because we have to account for things like the temperature
dependence of the bandgap, the temperature dependence of effective
masses, and details like that.

8
The correct ni at 600 K can be found in the reference listed above. For
this example, we'll take the value of ni at 600 K as 4 x 1015 per cc. We
solve our quadratic equation, and plug numbers in for the doping
density, and for intrinsic carrier concentration squared. We see that we
can't neglect the ni2 term. We end up with an electron concentration that
is not equal to the doping density, but is significantly higher, 5.1 x 1015
( the N-type doping is only 2 x 1015). The minority hole concentration is
still equal to ni2/n0, and we can compute that as well.

9
When the temperature gets higher, we have to be more careful, and
include the effects of the intrinsic carriers. When we do that, we will get
a significantly higher electron concentration, and we also get a
significantly higher hole concentrations but the calculations are
reasonably straight-forward using the quadratic equations that we
developed from the assumption of space charge neutrality.

10
As we go to higher temperatures, and the semiconductors become
intrinsic, we lose the effects of the doping. Selective doping in various
regions of the semiconductor is how we use create devices, so at high
temperatures, devices can fail. “Going intrinsic” determines the upper
limit operating temperature of a semiconductor device. If we're looking
for semiconductor devices that need to operate in higher
temperatures, we would look for semiconductors such as silicon carbide
or gallium nitride, semiconductors that have larger bandgaps than
silicon, because they will have smaller intrinsic carrier
concentrations, and these effects will take place at higher
temperatures than they do in silicon. High temperature electronic
devices tend to be made from materials that are different from silicon.

11
Let's shift our attention now to the low temperature, the freeze-out
region. At low temperatures, there is not enough thermal energy to
completely ionize the dopants, they're only partially ionized, or maybe
not ionized at all. The thermal energy is also too low to create intrinsic
carriers. Even though they are only partially ionized, the ionized doping
density can still be much greater than the intrinsic carrier concentration.

12
How would we solve for the carrier concentration in the freeze out
region? We begin with space charge neutrality, but since ni2 is so small,
we can neglect the minority carriers. If this is an N-type sample, we will
neglect the holes. The result is that the electron concentration is just
equal to the net N-type ionized doping concentration. All we have to do
is to compute the fraction of dopants that are ionized and we have our
answer.

13
The slide above summarizes the situation. We're at a low temperature,
so only a few of the bonds of the fifth electron to the donor have been
broken. There is one electron in the conduction band for each donor
that is ionized. We want to compute the fraction of the dopants that are
ionized, ND+/ND.

14
To compute ND+/ND, we might think that the fraction of these states that
are occupied is given by a Fermi function. (This will get us close, but
we’ll see on the next slide that it is not quite correct.)

If we assume that the probability that an electron is on one of the donor


sites is given by the equilibrium Fermi function, then the probability that
the donors are ionized is given by 1 - f0(ED).

15
Assuming ND+/ND, = 1 - f0(ED) is close, but it's not exactly right; the
actual answer is a little bit different. The correct answer includes a
donor degeneracy factor, gD, as shown above. This occurs because
the localized donor states are different from the de-localized, extended
states in the conduction band.

Each donor can accept a fifth electron that has spin up or spin
down, but once it accepts an electron of either spin, the donor atom is
neutral and cannot accept another one. This doesn't occur for the
distributed conduction band states, so the occupation statistics for
donor states and conduction band states are different as shown above.

16
To summarize, the way we describe the fraction of donor or acceptor
states that are ionized is by an expression that looks almost like a Fermi
function, but with a donor or acceptor degeneracy factor. The
standard factor used for the conduction band is two, and for the valence
band, it is four. If you'd like to learn a little bit more about these factors
and where they come from, please see the reference listed above.

17
Let's see how the calculation is done for an N-type semiconductor at a
low temperature. We’ll just summarize the approach, for more details,
see the reference listed above.

As shown in the first equation at the right, the fraction of donors that are
ionized (empty donor states) is given by expression that is similar to (1-
f0).

As shown in the second equation on the right, the electron


concentration is equal to the density of ionized donors.

Finally, the third equation relates the electron concentration to the Fermi
level.

If we equate the second equation to the third, and make use of the first
equation, we get an equation that can be solved for the Fermi level.
Once we know the Fermi level, the third equation gives us the electron
concentration.

18
The result turns out to be a little bit complicated, but it is displayed
above. This equation describes the drop in the free carrier
concentration as the temperature decreases and fewer and fewer
dopants are ionized.

19
If we look at our result at very low temperatures where only a small
fraction of donors are ionized, the complicated expression
simplifies. The electron concentration depends exponentially on the
binding energy of the fifth electron and on temperature. In fact, the low
temperature expression looks familiar. It looks like our expression for ni.

You can see that we have a square root of effective density of states for
the conduction band, but instead of an effective density of states for the
valence band, we have the density of donor states divided by the donor
degeneracy factor. The thermal energy is so small that there is no
probability that an electron from the valence band will be excited to the
conduction band, so the valence band is out of the picture. In fact, the
donor levels play the role of the valence band. For the intrinsic
density, we had ni ~ exp [ -(EC – EV)/kBT], but in this case we have n0 ~
exp [ -(EC – ED)/kBT], so the result is intuitively what we might have
expected.

20
When dopants freeze out, we lose the effects of the dopants. Dopant
freeze out can cause semiconductor devices to fail at low
temperatures.

21
We should briefly mention the metal-insulator transition. Recall that a
metal conducts electricity at T = 0 K because the Fermi level is right in
the middle of the band to begin with, so there are plenty of electrons –
we don’t rely on dopants. Semiconductors, however, become insulators
at T= 0 K, because the dopants freeze out and there are no intrinsic
carriers, so there are no carriers to conduct electricity. Heavily doped
semiconductors are different.
Recall that conduction band states arise when the wave functions of
adjacent atoms overlap, the states smear out and became delocalized
bands that allow electrons to travel throughout the crystal. When we
dope heavily, the wave functions of the adjacent dopants begin to
overlap and delocalize. Instead of electrons sitting on specific
dopants, these levels smear out into a donor band, and electrons are
free to move around and conduct.
If we dope a semiconductor heavily enough, the dopants are close
enough together so that the discreet isolated energy levels of the
dopants merge into an impurity band. When heavily doped
semiconductors are cooled down to low temperatures, they continue to
conduct electricity; they behave as a metal. There is no freeze out. This
is known as the metal-insulator transition.

22
Now that we understand how the carrier concentration varies with
temperature, can we also understand how the Fermi level varies with
temperature?

23
Let's consider an N-type semiconductor and ask how the Fermi level
varies as we sweep the temperature from a low to a high
temperature. The expected Fermi level vs. temperature characteristic is
sketched above.

At low temperatures, all of the donors are occupied, so the Fermi level
needs to be above the donor level. At very high temperatures, the
semiconductor is intrinsic, so the Fermi level needs to be near the
middle of the gap. As the temperature increases, the Fermi level
smoothly transitions from a high energy near EC to mid-gap as shown
above. Something similar happens for a P-type semiconductor. At low
temperatures, none of the acceptors are ionized, which means there
aren't electrons on the acceptors, so the Fermi level has to be below the
acceptor level. When the P-type semiconductor is very hot, it becomes
intrinsic and the Fermi level is near the middle of the gap, and EF(T) just
smoothly vary as we increase the temperature.

24
We can summarize Unit 3 in one slide. We have tried to
understand how the carrier concentration varies with temperature. In
the extrinsic region, where the donors are fully ionized and where the
dopants overwhelm the intrinsic carries, life is simple and easy. In an N-
type semiconductor, the electron concentration is just equal to the net
donor density. In a P-type semiconductor, the hole concentration is just
equal to the net acceptor density. When the semiconductor becomes
hot, the intrinsic carriers become important, and the carrier
concentration increases. Finally, as the semiconductor is cooled down,
the dopants cannot be ionized, so the carrier density decreases, unless
the semiconductor is doped very heavily.

This qualitative understanding of how the carrier density varies with


temperature is something that you should have a quantitative
understanding of now.

This wraps up Unit 3. In the next lecture, we’ll summarize the key
points that you should take away from this unit.

25

You might also like