108108112

INDEX
S. No Topic Page No.

Week 1
1 Quantum Mechanics: Concept of Wave Particle, Schrodinders Equation 1
2 Quantum Mechanics:Â Particle in a Box 23
Quantum Mechanics:Â Particle in a Box -Â Continued, Harmonic
3 Oscillator 38
Week 2
4 Solids : Formation of Bands, Kronig-Penny Model 58
5 Solids : Kronig-Penny Model - Continued 71
6 Solids : Electrons and Holes 90
7 Solids : Electrons and Holes - Continued 102
8 Solids : Crystals 117
Week 3
9 Density of States 133
10 Density of States - Continued, Fermi Function 154
11 Fermi Function - Carrier Concentration 179
12 Doping 192
13 Doping - Continued 207
Week 4
14 Recombination and Generation 229
15 Recombination and Generation - Continued 243
16 Recombination and Generation - Continued, Charge Transport 264
17 Charge Transport - Continued 278
18 Continuity Equation 291
Week 5
19 Junctions 301
20 Metal Semiconductor Junctions 323
21 Schottky Contact : Electrostatics 337
22 Schottky Contact : Current-Voltage (IV) Characteristics 359
23 Schottky Contact : IV Characteristics - Continued 379
Week 6
24 Schottky Contact : Small Signal Impedance 413
25 PN Junctions : Electrostatics 419
26 PN Junctions : IV Characteristics 437
27 PN Junctions : Small Signal Impedance 449
28 PN Junctions : Non-Idealities 467
Week 7
29 Bipolar Junction Transistors (BJT) 473
30 BJT : IV Characteristics 494
31 BJT : Non-Idealities and Equivalent Circuit Modeling 510
Week 8
32 Metal Oxide Semiconductor Capacitor (MOSCAP) 525
33 MOSCAP - Continued 542
34 MOSCAP : CV Characteristics 559
35 MOSCAP : CV Characteristics - Continued 568
Week 9
36 MOSFET: Introduction 583
37 MOSFET: I-V characteristics 597
38 MOSFET: I-V characteristics - Continued 623
39 MOSFET: I-V characteristics - Cont.. 636
40 Subthreshold swing, Additional concepts 648
Week 10
41 Trapped charge, Body-bias 659
42 Scaling of MOSFETs 679
43 Scaling of MOSFETs - Continued, Leakage currents in MOSFETs 709
44 MOSFET characterization: Parameter extraction 725
45 MOSFET characterization: Trapped charges, contact resistance 740
Week 11
46 MOSFET as a switch 759
47 MOSFET as a switch - Continued 785
48 Amplifiers using MOSFET 797
49 Amplifiers using MOSFET - continued 817
50 Circuits: Frequency Response, Noise 838
Week 12
51 Introduction: Amorphous Semiconductors 860
52 Thin Film Transistors 880
53 Tutorials Session -1 902
Semiconductor Devices and Circuits
Prof. Sanjiv Sambandan
Department of Instrumentation and Applied Physics
Indian Institute of Science, Bangalore
Lecture - 01
Quantum Mechanics: Concept of Wave Particle, Schrodinger’s Equation
(Refer Slide Time: 00:14)
We will start off with the Fundamentals of Quantum Mechanics, ok. Now, most students
are probably at the end of their high school are very familiar with classical mechanics
which is governed by the traditional Newtonian laws of physics, which is very
commonly taught at the high school level. Now, for students are not been exposed
quantum mechanics some of these ideas can turn out to be very strange, and you know
sometimes very puzzling, ok. And there is no better way to introduce quantum
mechanics at least in my opinion then to look at this famous thought experiment by
Richard Feynman, ok. And the experiment starts off you know the discussion on this
starts off like this.
So, many students are familiar with the Young’s double slit experiment, but what we are
going to do we are going to play around with this experiment and try out different ideas
here. So, let us say you have instead of you have a gun, and it is a special kind of a gun
which shoots out marbles, ok. So, essentially what I mean by marbles is any large
discrete particle, ok. So, it could be marbles or you know it could be bullet us it really
1
does not matter large discrete particles that cannot break up that are going to come, that
are going to remain you know structurally which have got structural integrity throughout
this experiment.
And you take this gun and you start firing marbles one at a time, and there is a certain
spread to this firing. So, it is not going in a straight line there is a certain spread. And
then you place a barrier right in front of this gun, and this barriers got 2 gaps they got 2
fine gaps through which the marble can get through and we will call these gaps a slit 1
and slit 2, ok. And what we are interested in seeing is we want to know what the pattern
of impact is on a screen that is placed behind this barrier. So, we have a recording screen
here, and this these little marbles will come and try to stick on to this recording screen,
and what we want to find out is how many marbles were collected at some point x from
some reference x equal to 0.
So, if you define this point on the screen as x equal to 0 and you vary this distance as x, I
want to know what is the probability, that a marble came and struck the screen at this
point and a simple way to estimate this probability is simply by counting the number of
marbles that came in impacted the screen at this point divided by the total number of
marbles where fired from this gun, it is a very simple way of looking imagining this. So,
I need to find out I want to get a distribution of the impact profile on the screen, ok. So,
that is that is the simple experiment.
2
Now, let us try to play around with this. So, let us say we closed one of these gaps, ok.
So, this gap is closed. So, slit 2 is closed and slit 1 is open and you fire your marble gun
and you will find that all the marbles you know they cannot get through the barrier
beyond this point in these regions, and the only possible way to you know impact screen
is by travelling through this gap here. So, you will find that the marbles randomly travel
through this gap. So, it might go there you know bounce off this edge some of the
marbles might bounce off that edge and some of them might just get through and so on.
And what you would expect is you would expect most of the marbles that ever got
through this gap to strike the screen at some point here, and there would be some sort of
a gradual decrease in this probability that a marble really impacted the screen at the other
places. And this is what any this is what say common sense might tell us, ok. And it is a
fairly good argument because there is only one entrance to this entire path to the screen.
On the other hand if you were to close slit 1 and keep slit 2 open you would expect the
same profile, but the only thing is it shifted in x, ok. So, you do not expect the profile
there because there is no path to strike the screen over there into the marbles are now,
going to enter is going are going to enter this barrier through slit 2 and they are going to
impact the screen in this manner you will find a lot of marbles impacting here and very
few in the other regions. So, you have some kind of a profile like this.
Now, what is key is to understand that the marbles are going through one marble is going
through only one slit, a marble cannot decide to go through both slits together, ok. That
is what that is what we feel that is what our intuition is, ok, the reality might be different,
ok. So, if you were to go talk to a layman and say you know this marble travel through
both slits together it would appear a bit puzzling at least initially, ok. And you with that
intuition you do expect this distribution and this is what you might actually see if you
were to conduct this experiment.
3
Now, if you keep both slits open you will find that the marbles are able to go through
both slits simultaneously you will see a profile of marbles you know impacting the
screen in various paths and you expect the profile to look more or less like this. So, I
probably cover it with all my writings, but you would expect the effective profile to be a
summation of these two profiles that you see and you would expect something of this
kind, ok. And of these slits are very far apart you might you might see a slight dip up
here in this.
So, essentially the probability of finding a marble the probability that a marble impacted
the screen at some point x, from this reference is let us say p of x and that is simply the
summation of the probability that the marble came through slit 1 and impact to the screen
and that the marble came through slit 2 an impact in the screen. So, it is just a summation
of these two probabilities. And what is important is that we have this intuition that a
marble can only go through one of the two slits, ok. So, it cannot go through both slits.
So, now, that is what the experiment with marbles tell us, ok. So, now, let us change the
experiment of it.
4
Instead of having a gun which shoots out marbles we will have a wave generator, ok. So,
you can think of it as a way a gun that shoots out waves. And these are the wave fronts,
so you can conduct this entire experiment in water or oil if you like and these are the
wave fronts, of this wave. And this waveform front propagates through and we keep the
same barrier that we use in the previous experiment maybe change the sizes a bit, but
essentially the idea is the same and once again we need to keep a recording screen and
we want to record the intensity of the wave observed on the screen, ok. And we want to
find out what is how does the intensity vary with the x, ok.
5
So, once again if you go back through our thought experiment which is to close the slit
and keep that slit open the wave front strikes this barrier and it would sort of diffract
through this little opening and it would strike the screen giving you a distribution and
intensity distribution that probably looks something like this, ok. And this distribution is
very strongly of course, it is very strongly dependent on the size of the slit, but that is not
the main point of all this argument, ok. So, you have an intensity distribution that looks
like this.
Now, if you were to close that slit and keep this one open you will find another intensity
distribution that looks like this.
Now, if you were to keep both slits open the wave front enters through both slits at the
same time and what you will see is an interference of these two waves. So, you see all
these points where both these wave fronts are interfering with each other, ok. So, what
you will see is an interference pattern, wherein the effective intensity at some point x
depends on the intensity of the wave coming the intensity 1 which is basically the
intensity or the waves coming through slit one and intensity 2 in this manner, and this is
the interference term where this is the phase difference between these waves.
So, what is key here is unlike bullet us we imagine the waves to be going through both
slits simultaneously and that is why we have interference, ok. Now, the bullet us did not
do this, that is what our intuition told us but on the other hand the wave front you see it
6
approaches both slits simultaneously. And it goes through both of the slits
simultaneously and therefore, you have interference. And this is you know sort of very
well expected and we have experiments such as the Young’s double slit experiment
which sort of sort of provides an empirical or an experimental corroboration with all this
theory, ok.
Now, let us come down to a third experiment. And what we do here is we go back to a
marble gun, ok, but instead of firing marbles we have to fire electrons, ok. So, we have
an electron gun. So, essentially you can imagine these to be very tiny particles. So, what
I mean by an electron gun is it has to be very tiny particles, ok. And it is alright at this
point think of these particles as discrete, ok. And we perform the same experiment and
we fire one electron at a time.
So, let us say you do not had to fire all of them together you do not have to fire you do
not have to fire them at any specific rate let us assume that you are firing one electron at
a time. So, you have one electron, you wait for some time, you fire a second electron you
wait for some time and go on so forth, and you perform the same experiment, you still
have your barrier with 2 slits and you have a recording screen. And the question is what
do you think will happen here, ok.
Now, if you follow the intuition of the marble if you think of electrons as discrete
particles they are very tiny you would expect that the electrons that one particular
7
electron chooses to go through any one slit, right. So, maybe this electron went through
this slit the second electron went through this slit and so on, and what you would expect
to see it is a distribution that looks not very different from our experiment on marbles
right you would expect to see a distribution that looks like that, but the reality is quite
surprising.
If you were to do this experiment what is expected is you would see a distribution that
looks more like what the way more like the manner in which the waves behaved as
compared to the bullet us, ok. And that is largely slow because it is not intuitive, and
there is this a puzzle because we were to believe that the electron went through only one
slit at a time, right. The electron this particular electron decided to either go through this
slit or this one it did not go through both slits that was our intuition but that is not what
the experiment says.
The experiment is telling you that despite the electrons being fired one at a time you
know with even some sufficient space and time between the firings. Each electron has
gone through both slits simultaneously much like a wave and not like a particle and that
is a big puzzle, ok.
8
So, we say that the, this is not quite possible we say that it is impossible for the particle
like electron to be behaving like a wave and producing an intensity pattern that is
expected when we have waves, ok. And we say that we do not believe the fact that the
electron went through both slits simultaneously and therefore, we are now, going to place
a detector at this location in order to identify the position of the electron. Or in other
words they are going to place this detector that is going to exchange information with
electron it is going to ping the electron at the photon let us say and in obtained
information with regards to the position of the electron.
And the point placing the detector is to identify the slit through which the electron is
going to enter, and then see has to what kind of recording these obtained on the screen.
So, let us say we keep the experimental setup the same the recording screen is the same,
the position of the slits are the same, the electron gun is the same etcetera and the only
difference is the appearance of this detector. So, what kind of pattern do we think that we
will see the electrons create on the recording screen? So, when we run the experiment we
have a very surprising result. We see that when the experiment is run with the detector
and place the electrons produce an intensity pattern that was seen when bullet us were
fired or when large particles were fired.
Now, this is even more surprising than the previous experiment. So, what the electrons
are saying is that when there is no detector around, we are going to behave very
9
differently; we are going to behave like waves and create an intensity pattern that is
expected of waves. But the moment we identified that a detector is watching us, they are
going to behave very differently and we are going to behave like bullet us and create a
very different intensity pattern. And this is the puzzle of quantum mechanics, and this is
what quantum mechanics tells you which is that the, that at the very basic level you treat
every particle as a wave and a particle or a wave particle ok. And from this point on they
are going to go ahead and develop this concept and try to understand this idea of a wave
particle.
As far as this part of the experiment is concerned, which is the interaction of the detector
with the electron, creating the making the electron behave differently it is a very
wonderful ground for philosophers to act on and there are a lot of interesting debates, ok.
while there is while it is quite at least qualitatively understood that as to what happens,
there are a lot of interesting debates between physicists of the twentieth century in trying
to argue this factor out. So, it makes so very for some very good reading for the student
who is interested.
So, let us now try to understand this key concept of quantum mechanics, which is called
as the wave particle duality, ok. And the idea of this concept is that both a wave like and
particle like treatment is needed to explain electron explain the experiments involving
electrons photons or any other tiny particles. So, for example, if you look at the photon
10
we have, its wave like nature appear in experiments such as the Young’s double slit
experiment. And we have its particle like nature appear in experiments such as such as
the photoelectric effect, ok.
And if you look at the electron, we have its wavelike nature appear in the experiments
that we just discussed which was which was like the Young’s double slit experiment.
And we have its particle like nature appear and experiments which involves a scattering,
ok. So, we need both of wave like treatment as well as a particle like treatment to explain
the behaviour of these entities, and coming to the electron you know, which are basically
the particles of interest with regards to this course of course, it is not that the photon is
not of interest, but we would like to elect understand the electron very well.
So, coming to the coming to the electron just like how you know its particle like aspect
can be defined with the with terms like the momentum, which is say the mass into the
velocity and terms like the kinetic energy, which is half into the mass into the velocity
square. Just like how you can define say the particle attributes of the electron we need to
also identify the wavelike attributes of the electron. So, what is the wave length of an
electron? And this relation between the wavelike nature and particle like nature is
defined by the De Broglie’s equation. So, it is a very very essential idea or essential
concept, which connects the wavelength of the wave particle to the momentum of the
wave particle.
So, we say that 𝜆 which is the wavelength is defined as h/p, where h is a constant and it
is called as a Planck’s constant and it is got a value of this. It is the same constant that is
used to define the energy of a photon. So, the energy of a photon is h times nu, where nu
is the frequency of photon, we are talking about the same Planck’s constant and p is the
momentum of the electron. So, if you know the momentum of the electron.
So, let us say the electron is fired from a gun with a certain amount of energy kinetic
energy. So, we know the kinetic energy of the electron which is nothing, but the
momentum squared by 2m, where m is the mass of the electron and therefore, we know
the momentum of the electron and if we know the momentum of the electron we know
the wavelength of the electron via the De Broglie’s equation. So, let us get some
numbers and some estimates and you know try to see as to how we can use this equation.
11
So, the electron has got all these properties it is got a charge of minus 1.6X10-19
coulombs, it is got a mass of 9.1X10-31 kilograms. So, that is a rest mass of the electron.
Now, let us say the electron is fired from an electron gun you know for the purpose of
this example, we say that it is got a velocity of 105 meters per second, and therefore, it is
got a momentum of mass into velocity which is about 9.1 to 10 power minus 26 k g
meter per second. And therefore, it has got a kinetic energy and this kinetic energy can
be represented in Joules, but by dividing it by q, which is the magnitude of the charge of
the electron.
We can also represent energy in terms of a unit called as electron volts. So, if you have
energy in Joules, so let us say you have 1 Joule in order to get this energy in electron volt
units electron volts is a unit of energy in order to get this into electron volts units you
take E and divide it by the magnitude of an electron charge, which is 1 joule divided by
1.6X10-19, and you have the energy in electron volts. So, joule is a large it is a huge
amount of electron volts. And since we talk about very small numbers in semiconductors
its useful to have this unit of electron volts.
And therefore, the terms potential and potential energy are all in some sense connected,
ok, the terms potential and terms energy are all interrelated because of this neat
connection. And since you have this being the energy what is the wavelength of the
electron? The wavelength of the electron is given by the De Broglie equation and lambda
is h/p which is about 7.2 nanometres in this case. So, you see that the electron fired off
with the velocity 1X105 meters per second has a wavelength of about 7.2 nanometres.
Now, let us take just for just for the sake of argument, ok. So, let us say we have a very
large particle, and I have taken example for tennis ball and a tennis ball is not a single
particle its composed of many little particles, but let us say that is not the case and it is
we are talking about one very large particle. And this particle has got a mass of about 58
grams and it is got a velocity about 50 meters per second, I have used values which are
typically used for a tennis ball and therefore, it is got a momentum of this much. And
what is the wavelength? The wavelength is h/ p and since the momentum is very very
large the wavelength is extremely small its 9.1X10-32 meters.
So, you can compare the wavelength of a large particle or a heavy particle with the
wavelength of a very tiny particle or a light particle. And you see that the difference is
12
quite immense, ok. So, this is an example just to illustrate what mass does to the
wavelength or what momentum does to the wavelength of a object.
So, here I have summarized some useful relations that would that would serve as well in
this course we have the first variable which is k, which we will be using in the context to
the wave vector or a wave number, and which is given by 2𝜋/𝜆 it is got a value of 2 pi or
its got an amplitude of 2𝜋/𝜆 or a magnitude of 2𝜋/𝜆 as the wavelength. Now, k is also,
later on in the course k is also used to define the Boltzmann’s coefficient, and I hope that
this will not be a matter of confusion because the context will be quite clear and I believe
and I hope that this will not create any confusion.
The other important relation is the De Broglie’s relation which is lamp which says that
lambda is equal to the Planck’s constant divided by the momentum. And the momentum
therefore, can be written in terms of the terms of the wavelength as p is equal to h by
lambda. Now, if you were to take this relation and you divide and multiply by 2 pi, you
say h/2𝜋 pi into 2𝜋/𝜆 is equal to your p we see that this is nothing but k, 2k2𝜋/𝜆 is k and
this term which is h/2𝜋 is given a new definition, ok. It is called h bar or its also known
as the reduced Planck’s constant, ok.
So, the term h/2𝜋 is also denoted by h, which is called the reduced Planck’s constant and
it is got a value of about 1X10-34 Joule seconds and therefore, the momentum can also be
expressed as h k.
13
The energy of a photon is already well known which is h into nu which is a Planck’s
constant into nu and since we can write h as h we can say h is h/2 𝜋 and 𝜈 as a and you
multiply 2𝜋 before 𝜈 you find that the energy of photon is also equal to h bar into omega,
by the omega as the angular frequency where nu is in hertz and 𝜔 as in say radians per
second, ok. So, these are some very useful definitions that would serve us in this course.
Moving on another key idea behind in quantum mechanics is something called as the
Heisenberg’s uncertainty principle. And the essence of this idea can be very nicely
understood for a person who is got some experience in signal processing or let us say
who is who understands the idea of a Fourier transform. But just to summarize just to
you know introduce this concept let us say that we are looking at a signal, that is going to
occur or an event that is going to occur at some point in time. So, let us say we have the
time or space domain it really does not matter.
So, first let us look at time and we say that, there is a signal that is going to occur at some
point in time. So, let us say there is a signal that occurred at this point in time if at if
somebody asked us as to when the signal occurred the answer is very very clear we say
that yes the signal occurred at this moment in time that is a very clear question and we
have a very clear answer. But if now, somebody asked us as to what is the frequency of
the signal, what is the wavelength of the signal.
14
Now, that question seems a bit absurd because the signal had a very it was not periodic
firstly, it had it occurred at just one moment in time and that was it. And therefore, the
only way one can determine the wavelength or the frequency of the signal is by taking a
Fourier transform and we find that the Fourier transform is wideband signal, right. It is
got its got frequencies in from minus infinite from frequencies from 0 to infinite, ok. And
therefore, the frequency enhance the wavelength of the signal is not very clear and
therefore, we feel that the question on what the frequency of the signal is not a very clear
question because the signal is a wideband signal.
Now, let us take the example of another signal, and we say that this is a very nice little
sinusoid and that is time. And now, if somebody asked us the same two questions, if
somebody asked us is to vend it the signal occur we find that this is a very bad question
or we find that is a very weird question because it is a sinusoidal signal and there is no
particular moment when the signal occurred the signal has been present in all points in
time.
On the other hand if somebody asked us as to what is the wavelength of the signal this
question seems to be a very nice clear question it is a very clear answer because we can
now, point to the exact wavelength of the signal and therefore, we know the exact
frequency of the signal. So, if you were to take the Fourier transform of this signal, you
would find that it is got a very unique frequency component that corresponds to the
frequency of the sinusoidal signal.
So, this is the key idea behind you know it is a key concept or the key puzzle behind the
Heisenberg’s uncertainty principle, which is if the location in time or in space is very
clear then the message on the frequency or the wavelength is very very unclear. And on
the other hand if the idea of the frequency or the wavelength is very very clear, then the
location in time or in space is very very unclear, and it is this parent it is this balance
between these two errors or the uncertainty that are addressed by the Heisenberg’s
uncertainty principle.
So, in one form what the Heisenberg’s uncertainty principle says is that if you have if
you were to perform an experiment say with the electron, and you were to be measuring
the position of the electron and you were to be measuring the momentum of the electron
simultaneously, ok. And if you were to have an error in the estimate of the position of the
15
electron and an error and the estimate of the momentum of the electron then the product
of these two errors have got a bound and that bound is defined by h bar by 2. And what
this inequality says is that one cannot estimate the position of an electron and the
momentum of an electron to the, to an accuracy which wherein the errors or the product
of the errors lie below h by 2.
Or other words if I were to define the position of the momentum define the position of
the electron to the infinite accuracy, I define the position absolutely accurately such that
delta x is 0. Then I will have a very large inaccuracy in the position or in the error in the
momentum I cannot define the error in the momentum at all there will be an infinite error
possibility and vice versa. So, this is the definition of the Heisenberg’s uncertainty
principle, ok. And I have given you a very interesting you know reasoning behind you
know our experiment on the electron gun over here we will not discuss it here, but if you
have any questions just let me know. It is uses the Heisenberg’s uncertainty principle to
talk about the events of the experiment.
So, now, we have a particle that is also called a wavelike behaviour and I want to now,
start modelling it. So, I can associate a momentum to it, I can associate energy to it, I can
wavelength to it and I want to model this aspect of this wave particle, ok. So, there is
everything is this wave particle and I want to start modelling this entity, ok.
16
And a convenient way to do it is through something called as the wave function, which is
usually denoted by the symbol psi. Now, what is this wave function? Ok, it is for all
practical purposes it is a mathematical function it is a function of space and time and it
contains all the measurable parameters of the particle, ok. And by operating on this wave
function I can extract all the parameters that I would like about this particle and we will
sort of look into this in greater detail.
So, if you say that you are only talking about a 1D x dimensional case you can say the
wave function is a function of x and t which is time, ok. Now, it could be a complex
function. So, this product which is the wave function times its conjugate, is the
probability density function of finding the particle, ok. So, it is basically the it is its
related to the expectation of a particle existing between 2 different between 2 regions,
and that is a very strong point, ok, which means if I know the wave function I can
estimate I can get an estimate of what the probability is of finding a particle between two
regions between two points let us say between x and x plus d x.
If psi is a real number then this is simply psi square and therefore, the probability of
finding the particle between x and x x+dx is Ψ 2 d x, because this is my probability
density and that is Ψ 2 dx is the total probability of finding the particle between x and x+d
x. So, what is this saying? It saying that if I have a wave function that looks like this, for
example if this is my Ψ of x.
Then I can take the magnitude of Ψ and I can take the square of it, ok. So, let us say that
does something like this. So, what this is it is the probability density of finding the
particle which means that according to this wave function the probability of finding the
particle in this region is the largest. And the probability of finding the particle here is not
0, but it is not as large as finding the probability finding the particle in this region, ok.
So, that is the implication of this and just going by these arguments if I need to find what
is the probability of finding the particle between say two locations x equal to a and x
equal to b, then it is simply the summation of these probabilities, which is an integral of
from a to b of Ψ 2 dx, ok. So, that is the probability of finding a particle.
Now, let us look at some more properties of this wave function. It is a very important
concept.
17
So, we said that the wave function contains all measurable parameters of the wave
particle and we said that these measurable parameters could be obtained or extracted
from the wave function by operating on it. So, what are some of the useful operators?
Ok, so we now, look at the momentum operator and the energy operator and therefore,
using these to define something called as the kinetic energy operator. So, let us say I
need to find the momentum of the particle in the x direction, ok. So, this happens to be
the momentum operator and I will give you an argument as to why this is the momentum
operator, ok.
If I need to find the energy of the particle or the wave particle, then this is my energy
operator. So, what this means is if I take this operator and use it to operate on this wave
function psi then I will get the momentum of the particle in x direction times psi. And
similarly this is the energy the particle time Ψ. And if I want to find out what is the
kinetic energy its p2/2m, which is 1/2m into the momentum operator operating it on itself
𝛿2Ψ
and then the wave function which is basically your (1/2mh)2 / I2 which is (-1/2mh)2 /
𝛿𝑥 2
𝛿2Ψ
I2 .
𝛿𝑥 2
So, this is my kinetic energy operator. So, you could imagine this particular element here
to be my kinetic energy operator, when it is operated on the wave function I get the
kinetic energy. So, you have these operators and they tell you what the momentum is the
energy is and the kinetic energies etcetera.
18
So, just to give you a quick feel for why this is the momentum operator, ok. So, let us say
for the sake of argument let us say my (1/2mh)2 / I2 Ψ. is a plane wave, ok. So, let us say
it is 𝑒 𝑖𝑘𝑥−𝜔𝑡 , where 𝜔 is the angular frequency, k is my wave vector. So, this is 2 2𝜋𝜆
this is 2𝜋𝜗 where 𝜗 is the frequency in hertz, t is time and x is the distance and i is
basically the square root of minus 1. So, wherever you see i it is simply your complex
number, ok. So, let us say let us assume this my wave function, ok.
Now, let us use the momentum operator on this. So, what is the momentum operator?
𝛿Ψ
Momentum operator is h/i. , since it is all in one dimension you can just write it as d,
𝛿𝑥
𝑑Ψ
, but nevertheless. So, this is nothing, but h/i and taking the derivative with respect to
𝑑𝑥
x you have 𝑖𝑘𝑒 𝑖𝑘𝑥−𝜔𝑡 . Now, this term here is nothing but your psi itself and therefore,
what this is h bar by i into i into k psi, which is h bar you can cancel the i here and k is
nothing but 2 pi by lambda.
2𝜋 ℎ
So, it is which is equal to which is nothing, but the momentum of the particle. So,
𝜆 𝜆
therefore, we could agree we have checked for ourselves that, this seems to be yielding
the momentum of the particle in the x direction. So, now, let us look at the energy
operator, ok. So, let us use the same kind of argument with this particular wave function
and let us look at the energy.
So, if I were to use this operator, which is minus h bar pi i dou by dou t on the wave
function what do I get? Now, taking the time derivative of the wave function I should
end up with minus h bar by i into minus i omega into psi itself, which is the exponential
term. And this is nothing but h bar omega which is h/2𝜋𝜐, which is my energy, ok. So,
that is why this is an energy operator and it seems like it is all fine. And the momentum
operator just derived from the moment the sorry the kinetic energy operators just derived
in the momentum operator and therefore, this is the kinetic energy operator.
So, the essential message of these two slides was basically you can use the wave function
to define the, what you say the measurable properties or model the measurable properties
of the wave particle. And you can use all these operators on this wave function to extract
information about the position the momentum the energy etcetera.
19
The wave function as we have described now, is a purely mathematical entity, and before
it can be applied to explain the laws of physics it must be constrained and it must be
made to follow a certain set of rules. So, what are these rules? So, these rules are
basically the conditions or constraints it is imposed by the physical laws. And the very
first rule is that the wave function must be a solution to something called as the
Schrodinger’s equation.
The Schrodinger’s equation is nothing but energy balance equation or an energy

conservation equation, which says the kinetic energy plus the potential energy is equal to
the total energy. The second condition is that the wave function must be continuous you
cannot have discontinuities in the probability of finding a particle in one region to now.
The third is that the first derivative of the wave function with respect with in space must
also be continuous because this is connected to the momentum of the particle of the wave
particle.
And the final condition states that the wave function must be normalizable, ok. And what
that essentially means is that is two things the first is since the wave particle or the does
exist in some point at some point in space say between minus infinity and infinite. And
since this is the probability of finding that wave particle in the space that probability
must be equal to 1, ok. And this limitation will imply that the wave function cannot take
a nature that will result in it blowing up.
20
So, for example, and what we mean by blowing up is for example, let us say that the
wave function has got a nature that looks like this which is an e exponential of x, and let
us say x goes from 0 to infinite we are looking at this region of space. So, we see that as
x heads towards infinite the wave function the amplitude of the wave function also heads
towards infinite this cannot be allowed to happen, ok. So, the wave function is limited or
the value or the only possible solutions of the wave function must be limited by these
criteria by this integral.
So, let us look at what is the Schrodinger’s equation. So, as we mentioned the
Schrodinger’s equation is an energy conservation equation, which equates the kinetic
energy to the, which says the kinetic energy plus the potential energy is equal to the total
energy. Now, you have already seen the kinetic energy operator and we found that this
operator operating on the wave function yields the term for the kinetic energy the
potential energy is dependent on the potential terrain or the potential profile in which this
wave particle sets and based on the different experiments or the different models that we
study this potential terrain needs to be identified and defined correctly. And the total
energy which is the summation of the potential and kinetic energy is given by the total
energy operator operating on the wave function. So, this is the total energy.
Now, in this definition we find that the total this is a time dependent Schrodinger’s
equation, because the total energy has got a derivative of the wave function with respect
21
to time but it is also possible to study the time independent Schrodinger’s equation. So,
for example, if we want wave functions or solutions that describe the stationary states,
ok. You can solve something called as the time independent Schrodinger’s equation in
which case the total energy is simply treated as a constant, ok.
So, we lose the time derivative and we write Schrodinger’s equation as the kinetic energy
plus the potential energy is equal to the total energy. And from this point on we will try
to use the Schrodinger’s equation to try and describe some simple models, which will
help us understand the nature of the wave particle and different potential profiles better.
22
Lecture - 02
Quantum Mechanics: Particle in a Box
All right. So, continuing on from you know where we left off in the previous lecture. We
will look at the Schrodinger’s equation in Quantum Mechanics. We basically introduced
it in the previous lecture, but from this point on we will see how to actually use the
Schrodinger’s equation and what can tell us about the behaviour of the wave particle in
certain systems.
So, the first example or the application for our Schrodinger’s equation would be through
an example, which is quite popular and which is something called as a particle in a one
dimensional box, ok. And what we have here is the profile of the potential versus the
distance.
Now, although we keep mentioning the word potential, and I do keep using the term
potential, I would like to emphasize that this term in Schrodinger’s equation which is
−ℎ2 𝛿 2 𝜓
+ 𝑉𝜓 = 𝐸𝜓 This term correspond to the potential energies we are
2𝑚 𝛿𝑥 2
equating energies, ok. So, we are saying that this is the kinetic energy, this is the
23
potential energy and that is the total energy, ok. But since the potential is has an is the
equivalent of the potential energy, I would sometimes use the term potential I while
talking about these concepts.
So, in this case we now have a potential profile that looks like as shown, and the particle
or the wave function exists somewhere in this region from x equal to 0 to x equal to
infinite. But the potential profile is such that between this region of x equal to 0 to x
equal to L the potential has a value of 0.
And in the regions outside x equal to L and below x equal to 0 the potential has a value
of infinite, which implies that the electron or the particle that we are considering has to
have an infinite amount of potential energy in order to be located in these regions, ok. It
has to have an infinite amount of energy to be located in the regions outside this little
valley and this valley is what we call as the box. And as per this definition this valley has
got a length or the size of the box is of length L that the coordinate system being placed x
equal to 0 at this boundary and x equal to L being located at this boundary of the box.
Now, since all these potentials are infinitely large we say that the wave function it is
impossible for the particle to be present located in to be located in those regions and for
this example the particle remains very much confined to the box, ok. So, the magnitude
of the wave function, so since we are talking about the wave particle the magnitude of
the wave functions in these regions is 0 whereas, the magnitude of the wave function or
the made through the wave function inside the box is what is to be determined.
24
So, we can now, write Schrodinger’s equation for this example of a particle in a box and
say that this is the kinetic energy that is the potential energy, and that is the total energy
and for the region that is inside the box the potential energy is 0, ok. And therefore,
Schrodinger’s equation reduces to this simple form and outside the box as we mentioned
it is not possible for the particle to exist since it would require an infinite amount of
energy and we say that the wave function has got a amplitude of 0.
Now, the general solution to an equation or second a differential equation of this kind is
given by defining the wave function as a summation of sines and cosines. So, we say that
the wave function is 𝐴 sin 𝑘𝑥 + 𝐵 cos 𝑘𝑥, where A is nothing but a constant coefficient
and B is a constant coefficient and k is another coefficient that needs to be determined,
ok. So, from this point on they are going to apply the correct boundary conditions to try
and define the wave function of the wave particle inside the box more accurately and in
particular we need to identify the parameters A, B and k. Now, k can also be seen can be
seen to be related to the energy right from Schrodinger’s equation.
So, if we were to rewrite the differential equation as dou square y by dou x square is
𝛿2𝜓 2𝑚𝐸 2𝑚𝐸
equal to − 𝜓, the value of k is related to the energy and it is . But we
𝛿𝑥 2 ℎ2 ℎ2
would now, like to identify a further relation, of k particularly with regards to the
geometry of the box and then use that relation to identify the nature of the energy.
25
Now, since this is a general solution let us apply the property that the wave function must
be continuous, ok. So, this was the second property of the wave function that we
discussed, ok. So, which implies that if I were to consider these two boundaries, which is
there is some wave function here, we do not know what the way we what the function is
but it must be continuous which means that if I were to go to this boundary, ok.
The value of the wave function just inside this edge must equal the value of the wave
function outside this edge, ok. And similarly if you go to the other side the value of the
wave function inside this edge must value must equal the value of the wave function
outside. And since we said that the wave function does not exist in this region, the wave
function is essentially 0, the probability of finding a particle outside this box is 0. So, that
is the statement we made and the argument also the box is infinitely high, ok. So, it is an
in surmount insurmountable cliff. So, you have the wave function being 0 on the outsides
and therefore, the wave function at x equal to 0 must be equal to 0 that is the continuity,
that is the continued continuous made through the wave function.
So, we first apply this condition that 𝜓|𝑥=0 = 0. So𝜓|𝑥=0 = 0, is Asin kx, which is 0
plus B cos kx which is 0, ok. And this turns out to be 0 plus B is equal to the wave
function at x equal to 0. And since this value the wave function takes a value 0 we need
this to be the first condition which implies that my B is 0 therefore, my wave function is
or the form A sin kx. So, for any x since B is 0 the cosine term does not exist. So, this is
26
the wave function. So, you have identify we have reduced the wave function from this
very generic sine and cosine terms to just having A sinusoid term.
So, now let us apply the boundary condition on the other side of the box, which is at x
equal to L. And at x= L again we find that the wave function must be equal to 0 since the
particle cannot exist on or the wave particle cannot exist on the in the regions where x is
greater than L.
Now, applying that boundary condition at x equal to L we say that this is the value or the
magnitude of the wave function is 0 in at sin kL. Now, what are the possible solutions?
They could have A to be equal to 0, but that would simply that would be a trivial solution
as it simply say that the wave function does not exist. So, we will ignore that solution
because we also have the possibility of a solution, where k can be an integer times pi by
L, where n could be positive integers 1, 2, 3 and so on.
𝑛𝜋
And 𝐿n pi by L into L is a integer times pi and sine of an integer times pi is 0.
𝐿
Therefore, this way function has now become more specific we can define the wave
function more clearly by applying the boundary condition at x equal to L and saying that
𝑛𝜋
the wave function is Asin 𝑥. Now, all that remains is to identify A using that value of
𝐿
k.
27
So, we have now identified B to be equal to 0 we have identified k to have the form of,
where n is an integer. So, the wave function has now reduced from this very general
expression of containing sines and cosines its, now reduced to this it has to be other form
𝑛𝜋
of A sin 𝑥. Now, the only thing is left to be determined is A and we do that by using
𝐿
the condition that the particle must exist somewhere, right.
So, if you remember 𝜓 2 was the probability density function for the particle, and psi
square dx is the probability that the particle exists between x and x + dx.
So, therefore, the probability that the particle exists somewhere inside the box, we have
already said that it does not exist outside the box which implies that the particle must be
existing inside the box. So, the probability that the particle exists somewhere inside the
box, is basically the summation that the summation of all these little minuscule
probabilities, which is simply the integral simply stating that the integral from 0 to L of
all these little probabilities is 1, ok. So, this is the probability that the particle exists
somewhere inside the box and that is guaranteed to be 1, according to this experiment.
So, this is the next condition that your general psi must satisfy. So, what does this
𝑛𝜋
become? So, it implies that from 0 to L, if I take 𝐴2 𝑠𝑖𝑛2 𝑥 then this must be equal to
𝐿
1. Now, we write you can write your sine square, ok. So, let us say you have a general
sine square alpha you must note that cos 2𝛼 = 𝑐𝑜𝑠 2 𝛼 − 𝑠𝑖𝑛2 𝛼 or in other words its 1 −
28
2𝑠𝑖𝑛2 𝛼. Therefore, my sine square of any angle alpha is one minus cosine of 1 −
𝛼
cos 2 ok.
2
So, we represent the sine square using this particular relation and what that gives you is
that this integral is nothing but A square 1 minus cosine of 2 n pi by L x the whole thing
divided by 2. And we need this to be equal to 1; we need this to be equal to 1, ok.
𝐴2
Now, this integral is simply , ok. So, this is all with respect to dx sorry I have missed
2𝑥
that out, so here integrating with respect to x everywhere. So, this is the first term is
𝐴2 𝑥 𝐴2
going to be A square into one by 2 dx which is − A square x by 2 minus you have
2 2
A square by 2 into your 2 n pi by L into sine of 2 n pi by L into x, ok. And the boundary
conditions are from 0 to L. So, the first term if I apply these boundary conditions it is this
integral limits sorry these (Refer Time: 14:47) not the boundary condition sorry I mean
the in the limits of integration are from 0 to L.
So, if I apply these limits, what do you get? The first term, the first term becomes A
square by 2 into L and 0 just results in 0. And the second term when I throw in an L there
you should get minus A square by 2 into 2 n pi by L sin of 2 n pi I mean I throw in a 0
there I just get 0, so minus 0, ok. So, this entire term must be equal to 1. Now, you see
that since this is an even integer times pi this in is quite term will always be a 0, ok. So,
therefore, this term also vanishes. And you are only left with the fact that A square by 2
2
into L is 1 or a is equal to √ . So, this is the outcome of applying this relation that the
𝐿
probability of the finding the particle somewhere inside the box is 1, ok.
29
So, we can therefore, describe the wave function, ok. So, now, you have identified A to
be you identified B to be 0 and you have identified k to be n pi by L, where n is an
integer. So, k could take multiple values and they are all integer multiples of pi by L. So,
2 𝑛
therefore, my psi has now become √ sin sine of n pi by L into x. So, this is my wave
𝐿
function. So, now, we have identified we set out on this exercise try and identify A, B
and k we have identified all 3 terms and we now have our wave function.
So, what this says is that. Firstly, note that the wave function is got A sinusoid term and
it must it must always end on the boundaries it must always have a value of 0, ok. So,
you will end up with the wave function looking like this, where at x equal to L by 2 if
you take if you take the case of n equal to 1 firstly, you will have sine of pi by L into x
and at x equal to L by 2 the this value takes a value 1 which is the highest value. So, you
will find the, at n equal to 1 the wave function looks like this.
At n equal to 2 I have sine of 2 pi by L into x, and at x equal to 0 at x equal to L the

value is definitely 0, but it is also 0 at a at x equal to L by 2. So, if I if my n is equal to 2
so this was the n equal to 1 case, if n is equal to 2 at x equal to L by 2 my wave function
again takes a value 0 the amplitude takes a value 0 because sine of 2 pi by L into L by 2
is simply sine of pi which is again 0 but at x equal to L by 4 the amplitude is the highest,
ok. So, you find that the wave function at n equal to 2 takes a takes a shape of this kind
and you can continue drawing this you know for n equal 2, 3, 4, 5 etcetera. So, this is my
30
wave function all right. So, that is the first point to know that the wave function is
sinusoidal inside this box.
Now, let us now, try to calculate the energy of this wave function, and that is going to
tell you a lot of interesting things. How do we calculate energy? So, since my minus x
my Schrodinger’s equation is minus x square by 2 m dou square psi by dou x square is
equal to E psi, all right. And now, I know my psi accurately, so I am going to use this
expression in the left term here.
So, the left hand side will now, become minus h square by 2 m, the first derivative of this
will yield square root of 2 by L into n pi by L of x into n pi by L into cosine of n pi by L
x, ok. So, if you look at what is d psi by dx it is square root of 2 by L, n pi by L cosine of
n pi by L x and my d square psi by dx square is going to be minus square root of 2 by L
into n pi by L whole square sine of n pi by L, ok.
So, we are going to use that term back here which is there is going to be a minus sign
because the derivative of cosine is minus sign and that minus is going to cancel with this.
So, I will put a plus sign it before this term into square root of 2, 2 by L into n pi by L
squared sine of n pi by L into x. So, that is the left hand side and you will notice that this
term here which is square root 2 by L into this sine is nothing but your wave function psi,
ok. So, you might as well just write these two term, get rid of these two terms and just
use, just use psi here and this is equal to E psi, ok.
Now, which means that my total energy is given by this expression here, which is n
square pi square by L square h bar square by 2 m is my energy. For the when n takes a
value where n takes only integer values, ok. So, this is the energy for some particular n.
Now, the first thing which we notice here is that the energy cannot take any arbitrary
value since n can only be integers, ok. So, if my n is 1 my energy is pi square h square by
2 m L square. If n is 2 my energy is 4 pi square h bar square by 2 m L square and so on.
So, the first thing to note is that the energy is quantized, you cannot have you cannot
have this particle having an energy of let us say something in between. So, it cannot have
nearly 1.5 pi square h bar square by 2 m L square because that violates the condition that
n has to be an integer, ok. So, that cannot happen, all right. So, the first thing to note is
that the energies are all quantized, ok. The second thing to note is that the energy is very
strongly dependent on the size of the box it varies as 1 by L square. So, if you were to
31
take a relation you know y is equal to 1 by L square and see how does y vary with L you
will see its quite dramatic as L approaches 0 y climbs up very very large and as L starts
increasing it completes completely comes down you know it reduces very very quickly,
ok.
So, this depends on 1 by L square is quite dramatic. So, which means that if I have a box
that is very narrow then the energy, energy the particle is very very large and the moment
I start increasing the width of my box the energy the particle drops down, ok. So, let us
let us note down these facts very neatly do forgive my reuse of the slide again and again,
ok.
So, the first point we noted was that energy is quantized, ok. The second is that a small
box implies large energy for the particle and a large box implies small energy for the
particle. So, the energy scales is 1 by L square. Now, let us also look at the mass if the
mass of the particle increases then my energy decreases, ok. So, these are all the absolute
values of the energy, ok. But if I have to draw all the different energy levels, like as it is
shown here. So, this is that n equal to 1 this was my wave function and n equal to 2 that
is my energy, at n equal to 3 there is another there is another energy level and so on so
forth, ok.
Now, if I were to draw all the energies and measure the distance or the energy gap
between 2 subsequent levels, which is essentially which is essentially E n minus E n
minus 1, ok. Now, I am looking at the delta E n, which is the energy difference between
2 subsequent energy levels. Now, according to this it should be pi square h bar square by
2 m L square into n square minus n minus 1 squared, ok. And if you were to just look at
this expression it is nothing but 2 n plus 1 I am sorry 2 n minus 1, ok. So, the gaps
between these which is my delta E, scales as 2 n minus 1 into pi square h bar square by 2
m L square.
So, the first thing to notice is as n increases the gaps between these two subsequent
energy levels also increase. So, here you have a smaller gap and there you have a larger
gap and n keeps increasing this gap will keep increasing. Now, the second thing to note
is that as L or m increase the gaps will reduce, ok. So, for a large box not only is their
total energy small the energy gaps between two subsequent energy levels is also small,
for a large box.
32
And similarly the energy gap between two successful successive energy levels for a large
particle the large mass is also small. And that is why when you have large bodies with
large masses, ok, so even though they are composed of many particles if you avoid that if
you just set that aside you know let us say it is all coherent and let us say it is just like
behaves like one discrete large particle.
The spacing between the energies are so small that this massive entity can take almost
any energy it likes, and therefore, you never get the feeling of something being quantized
in your day to day experiences. So, if you were to take a billiard ball or a tennis ball and
throw it you find that depending on the amount of energy you thrown classical
mechanics tells you that it can always take any energy you like depending on how much
energy throw into it, and that is not true in fact.
So, what this experiment is telling you is that if it is a discrete particle if you imagine
even this large mass to be one particle. As the mass increases it would appear it would
give you the illusion that the energies could be almost smooth and continuous, but in
reality it is not and the difference, it is always discrete and that difference between 2
energy levels is given by this particular term here which we looked at this particular
term, ok. And similarly the size of the box starts growing very large the energy levels
become very very fine, ok.
So, let us so this is an important message as well. So, not only is not only the energy
pretty reduced the energy spacing going to reduce with the mass of the particle, but if I
were to increase the size of the box and you look at all these energy levels, you will find
that the energy levels are much more closer to each other here, ok because once again the
difference in 2 consecutive energy levels varies as 1 by L square.
So, if we think about the electron. So, how does this connect with our semiconductor
physics? So, if you were to think about the electron, and let us say it sitting in some
material, and it sitting inside of potential well which cannot resemble this box and in
fact, there does exist such a potential well which is the nucleus of the atom. So, if you
think about the nuclei it is got let us say positive charge and therefore, it is got a certain
electrostatic potential and the only thing is its not shaped like a box, but the electron is in
some sense held inside this potential well.
33
So, if you were to imagine all this as a box, and if you look at the amount the electron is
allowed to move about around this nuclear, and if you sort of imagine an effective length
or that effective size the box to be that then the energy levels electron can take, ok. So, in
a material in which the electron is more or less free it is much more finely spaced as
compared to an electron which is not so free, ok.
Now, let us sort of extend this idea, which sort of generalize this idea to a 2D case, ok.
So, let us say the particle is not sitting inside a one dimensional box, but it is sitting
inside a two dimensional box. And it will still retain the fact that the, you see the particle
see is an infinite cliff, ok. So, you can sort of imagine a 2D box and for the time being I
will draw a rectangular box, for mathematical convenience as I will show you.
So, let us imagine a rectangular box. So, it is a two dimensional drawing here it is a. So,
on the on the z axis you have potential, ok, so let us call that y and let us call this x. So,
initially we only looked at that as a 1D case, where the distance was only in one
direction, but now, you have a two dimensional distance plane, ok. So, this could be L x,
let us call the width the box next direction is L x and the width of the box in the y
direction as L y, ok.
So, it goes from 0 to L y. So, I have to draw the coordinates this would be my coordinate
system. So, that is L x comma 0, and that is 0 comma L y and this is 0 comma 0. So, let
this be my coordinate system. So, I am only looking at the floor and if I were since it is
34
A since my potential is on the y axis the potential of this floor of this box is 0 and the
potential over here which is you know outside in the regions outside of this box, ok.
So, it is a hollow box and there is space all around it and that potential is infinitely tall,
ok. So, this is infinitely high, these are all infinitely tall walls and there is this little
hollow which is sitting at 0 potential and it is got a length x in the x direction L x in the x
direction and length L y in the y direction. So, that is the situation I probably should have
had a nice drawing here, but if you can if you can understand this drawing of mine you
have this situation in place. So, we want to solve Schrodinger’s equation for this case,
ok.
Now, since I chose a nice little rectangular box I can use one simple trick which is I can
say that my psi can be written as a product of two components, the psi in the x direction
and the psi in the y direction, ok. So, let us say my psi my wave function for the particle
can be written as a product of the wave function in the x direction and the product of the
wave function in the y direction. So, if I were use this relation right here my
Schrodinger’s equation which now, has got 2 terms because you now, have to consider
the potential the momentum in the x direction and the momentum in the y direction. So,
it is got the x and y aspects to it and that is my total energy my potential is still 0, ok. So,
it is just Schrodinger’s equation, but written out in the 2D case and it is still the time
independent Schrodinger equation.
Now, if I were to use this particular condition in my Schrodinger’s equation I will end up
with dou square psi x psi y by dou x square plus dou square psi x psi y by dou y square is
equal to E psi x psi y, ok. Now, psi x is purely dependent on x and psi y is purely a
function of y, ok. So, this is not a function of x and that is not a function of y, so which
means that as far as this derivative is concerned psi y is just a constant coefficient, ok.
So, you can put that outside your derivative and as far as this derivative is concerned psi
x is a constant coefficient and you can place it outside, ok. So, that is my differential
equation.
So, now let us mean let me divide throughout by psi x psi y, ok. Now, E is a constant
coefficient and therefore, I end up with minus h square by 2 m and I am going to divide
throughout by psi x psi y 1 by psi x dou square psi x by dou x square plus 1 by psi y dou
square psi y by dou y square is equal to a constant which is energy, ok. So, if you look at
35
this particular term here it is got 1 by psi x and its dou square psi x by dou x square there
is no y component at all, there is no psi y, there is no y, ok. So, this parameter here is not
at all influenced if you play around with the y aspects the wall box.
So, if I say increase or decrease L y this first term here should not be should not, it
should not matter to this term at all. And similarly the second term here is purely a
function of y only, and it should not matter to it as to what is happening in the x
direction. And because this is the case I can rewrite my energy as having two
components, one is the energy in the x and the other is the energy in the y. And I can
split this in such a way that I can rewrite this one equation as two separate equations
because I have got the same message, ok. I can write this as, this is the first term of your,
this is the first term on the left hand side and that is equal to E x and let us say minus x
square by 2 m 1 by psi y tau square psi y by dou y square is equal to E y. So, let us write
this equation it split it into these two terms,
If I were to take this psi x to the other side, I end up with the equations looking that way
I take the psi x and psi y to this side these are my equations. Now, these two are you are
very familiar set of equations which you solve for the particle in a 1D box and therefore,
you can solve these two Schrodinger’s equations independently, ok. So, you can now,
solve these two Schrodinger’s equations independently and say that my psi x has got a
function a wave function that the psi x looks very similar to what you saw before it
should be of this form square root of 2 by L x sine of n x by L x pi by and L x into x.
And psi y will be at the form square root of 2 by L y sine of n pi, pi by n y pi by L y into
y, ok.
So, this is psi x it depends on the length of the box in the x direction and it depends on a
quantum number or integer n x. And on this side you have the psi y depending on the
length of the box in the y direction and another quantum number or an integer n y to
satisfy all the boundary conditions. And the total wave function is simply the product of
these two which can now, be written as psi x into psi y, ok.
So, the total wave function is psi x into psi y which is square root of 2 by L x square root
of 2 by L y into the product of the 2 sinusoid terms, sine of n x pi by L into x sine of n y
pi by L y into y. So, this is the solution or Schrodinger’s equation in a two dimensional
box, ok.
36
So, it was convenient for us to separate these two terms, ok. Now, that may not always
be possible. So, for example, if you have, so here the coordinate system I chose it was a
rectangular box, and in a rectangular box the x and y directions were orthogonal
components. If you look at the basis vectors I have chosen 2 orthogonal components
there. Now, if you had a box say it was shaped as a triangle probably not so obvious, ok.
But nevertheless just to illustrate this point, I think this is a good message.
37
Lecture- 03
Quantum Mechanics: Particle in a Box – Continued, Harmonic Oscillator
So, let us now consider the next case, which is the particle in a one-dimensional box, but
instead of the potential being infinitely large you know outside this box, we will now
lower the potential to a finite value. So, it is exactly the same kind of a diagram or you
know kind of a potential profile, but the only big difference is that these potentials that
you see here, where initially infinite, I have now made it finite, and it is got a value of V
naught ok, and this is no longer infinity.
And you still have the particle, although I have drawn it as a discrete particle we should
remember that it is actually a wave function. And for the sake of simplicity, let us
−𝑥−𝐿 𝐿
actually define this box to be at and ok, so that everything that happens is we will
2 2
watch out for symmetry across x equal to 0 ok, so that is our that is our definition.
38
So, how do we write Schrodinger’s equation. Now, inside the box we firstly, let us let us
−ℎ2 𝛿 2 𝜑
write Schrodinger’s equation in a very general sense, so you have + 𝑉𝜑 = 𝐸𝜑
2𝑚 𝛿𝑥 2
So, this is Schrodinger’s equation. And we always wanted to identify what this potential
is and just like in our previous case inside the box, we can define the reference potential
to be 0 ok. And once again let me correct these axis, because that sort of is consistent
with what I have written later. And inside the box we find that Schrodinger’s equation is
simply this, because my potential term is 0, because the potential is 0 inside this box.
Now, outside the box in the previous case, when we considered an infinitely high
potential barrier, we had assumed that the wave function does not exist right. And if and
we and we went ahead to only solve for the wave function inside the box, but now we
can write Schrodinger’s equation outside the box. And you know using the same idea
here, you will be able to write Schrodinger’s equation outside the box, and instead of the
potential being arbitrary V, it is actually 𝑉0 , because that is the way we have defined our
potential.
39
So, outside the box, the potentials would the Schrodinger’s equation will take a form that
look like this. So, you have a kinetic energy, you have a potential energy that is equal to
the total energy. And inside the box, we have Schrodinger’s equation taking a form like
this. And we will define the coordinate system to be x equal to minus L by 2 x equal to
plus L by 2 on the right side; therefore, the box still got a width of L. And from x equal
to L by 2 all the way till infinite, you have this Schrodinger’s equation; from x equal to
minus L by 2 all the way down till minus infinite, you have this Schrodinger’s equation.
−𝐿
And inside the box that is from for the region of less than equal to x less than equal to
2
𝐿
, you have this Schrodinger’s equation. So, I will call this as region 1; I call this as
2
region 2; and call this as region 3. So, you have three different regions, and you have
written Schrodinger’s equation for each of these regions.
40
So, just to summarize, these are the three wave functions; we have 𝜓1 for region 1, and if
I just draw the box here again, this is 𝜓1 ; and then we have 𝜓2 for region 2; and you
have 𝜓3 for region 3. Now, the first point is that let us assume that E is less than V
naught. Now, this seems to be a very difficult assumption to taken, if you observe this
carefully. So, what are we saying here, by saying that 𝐸 < 𝑉0 . What we are saying is that
yes, the particle does exist somewhere, and it is got a total energy E, and that total energy
E is less than the potential height in region 3. And we have still gone on to write
Schrodinger’s equation in the manner shown here ok.
So, what we are saying is, this is the kinetic energy; this is the potential energy; and the
sum of the kinetic energy and potential energy is equal to the total energy, and that total
energy term is less than V naught, which means the total energy is less than the potential
energy ok. Now, that is that is that sounds a bit odd ok, but it is one of the main features
of quantum mechanics. So, if 𝐸 < 𝑉0 , and if the if I have to describe a wave function
here in this region regions 1 and 3, then what I am trying to tell you is that the kinetic
energy has got a negative value, that is that is that is the implication of it; I mean that is
what it that is what classical mechanics would tell you.
In classical mechanics, if 𝐸 < 𝑉0 , there is no way the particle can have an existence
outside region 2 ok. But, what we are seeing here is that we are able to write the
41
Schrodinger’s equation, it does there is nothing to stop us from writing Schrodinger’s
equation, even outside the box ok. So, let us see where this leads us.
Now, firstly let us get some definitions clear ok. So, we have this to be our Schrodinger’s
equation, I am just going to take the V naught term pull it over to the right side and get
rid of the negative sign here, and I am going to write the same equation in this manner.
So, I have taken h bar squared down here 2 m has gone up there and this is V naught
minus E; and therefore, the negative sign of this term has disappeared ok. And we will
just call this entire term component here, this entire component here to be equal to alpha
square.
So, since my E is less than V naught alpha square, I mean our alpha is greater than 1
greater than 0 sorry. And similarly, for the region for in region 2, we have Schrodinger’s
equation to be the kinetic energy is equal to the total energy. And I can rewrite this as
dou square psi 2 by dou x square is equal to 2 m E by h bar square. And we will define
this term here as minus k square ok. Now, this is this is a this is a situation that is similar
to what we have done before, the only thing is I am trying to bring in consistency
between the way we represent these equations ok.
So, we have minus k square, where k again is greater than 0. So, my k will be square root
of 2 m E by h bar h bar square, so that is my k. And in this case, my alpha is square root
of 2 m V naught minus E by h bar square. Similarly, for region 3 we write Schrodinger’s
42
equation in a manner, which is quite similar to region 1 ok. So, once again we define the
alpha is the same, so we have Schrodinger’s equation and region 3 to be this, so that is
region 1, that is region 2, and this is a region 3. Now, the big difference between region 1
and region 2 is that here alpha is positive, and this entire term is positive, but here since k
is also positive, there is a negative sign here, and therefore this entire term is negative ok,
so that is that is the big difference. And once again in region 3, you have a positive term
on the right hand side.
So, since I have Schrodinger’s equation in that particular manner, I can write an expected
solution for the wave function. So, as I mentioned in this region the right hand side term
is positive ok, and we need to guess the solution for say psi 1, so that when you take the
second derivative, you end up with a form that is got psi 1 in it, and it is also got a
positive term. And that solution could be of this form ok; it will have you have
exponentials of a real number, then you will end up with a possible solution for that.
Similarly, in region 3, you have a similar description, where you have exponentials of a
real term. But, in region 2, we already saw that you know you could describe things as
sines and cosines, when we solved when we solve Schrodinger’s equation for this
equivalence of region 2 in the infinite potential well case; we saw the sines and cosines
appearing in the equation.
43
And equivalently you have if you write it in terms of exponents, you are going to have
exponents of complex quantities that are appearing in region 2. So, and you know by
Euler’s notation, you have this being equivalent to your equivalent to your sines and
cosines. So, these are the three general solutions that are possible. So, phi 1 is expected
to look like that, psi 2 is expected to look like that, and psi 3 is expected to look like this.
Now, how many terms do we need to identify, we need identify A 1, B 1, A 2, B 2, A 3

and B 3 ok. And what are the properties that we have, we have that firstly, psi has to be a
solution Schrodinger’s equation that is all satisfied by defining psi in this particular
manner; that we must have psi to be continuous, you must have d psi by d x also to be
continuous, and you must have psi not blow up, it should be normalizable, it should not
blow up anywhere.
And basically equivalent to this condition is also the other side of it, which is that psi the
way the particle must exist somewhere. And therefore, in this space from minus infinite
to infinite, the particle must exist somewhere, therefore the probability of finding the
particle in some region there is 1 ok. So, we have all these terms to play with.
Now, let us just go through the process of elimination, and trying to bring out some
useful results here. So, first let us take the region 1 ok. So, let me redraw the box here, so
that was my region 1, that is region 2, and that is region 3. And for regions mentioned
before, this is the general solution for the wave function in region 1; this is the solution
44
for the wave function in region 3; and for region 2, you have sins and cosines ok. Now,
let us take region 1 first ok. So, let us let us not worry about region 2 and 3, let us just
take region 1, which is this region.
Now, let us imagine, since the general solution, let us start seeing what happens to
solution. If I start moving towards minus infinity I start walking in this direction and go
from x equal to minus L by 2 to minus infinity ok. At x equal to minus infinity, so as we
start moving in the direction what happens to this term, now alpha is positive. So, as x
starts getting more and more negative, this term starts dying out.
What happens to the second term? Since alpha is positive, and x is negative, this negative
sign disappears, it becomes a positive value here. So, you have e to the power a positive
quantity, and that positive number starts increasing as I head towards x equal to minus
infinity. So, this term starts increasing, as we head towards minus infinity. Now, we
already mentioned that the wave you cannot have any of these components blowing up,
which means you cannot have them head towards infinite. And since, this is doing
exactly that in region 3 in region 1, as I move towards minus infinite, this term cannot
exist you cannot allow this, this might be a general solution, but this term violates this
very basic principle that you cannot have the wave function blow up ok. Therefore, the B
1 term has to be 0 ok.
And if B 1 is 0, I can just write my solution to be just this, psi 1 is equal to A 1 A to the
power alpha x, so that just by a process of simple you know by just using simple
arguments, you have eliminated one of the coefficients, you do not have to identify that.
Now, what about region 3 and you can make the same argument here, but the only thing
is that region 3 exists from plus L by 2 all the way till infinite. So, if you start moving
towards plus infinite ok, you cannot head towards minus infinite, because that is not
region 3 ok. So, psi 3 is not valid for, if you are heading towards minus infinity. So, as
you head towards plus infinite, what happens to these terms, alpha is positive, and you
are heading towards plus infinite. Therefore, this term starts blowing up.
Now, here you have a negative sign, and alpha is positive, and you are increasing x from
L by 2 to infinite, therefore this term starts dying out ok. And since, once again since you
cannot allow the way function to blow up, this component cannot exist you cannot allow
45
that. Therefore, in region 3, the wave function must take a nature that looks like this.
Now, in region 2, you could have sines or you could have cosines.
Now, in the case of an infinite potential well, if you remember the infinite potential well
ok, where may had a solution a general solution of I think we used A and B there, A sin
k x and B cos k x, when we had this general solution. At this at the point of x equal to 0,
when we apply the continuity of the wave function at x equal to 0. Sin of k at x k x at x
equal to 0 was not non-existent, but the cosine term became 1. And therefore, the term B
had to disappear, because the wave function had taken a value of 0 outside the box, so
that was if you just recollect the infinite potential well case that is what happened, so the
cosine never existed in that case. But such a boundary condition need not happen now,
because psi need not be equal to 0. So, you could have the cosine term surviving,
because there is no condition for you know the wave function to be 0, when x is 0 for
example ok, there is no enforcing condition. So, you could have both these cases.
So, if that is the situation, and we only look about look at symmetric solutions of the
wave function so, we will look at solutions that are sort of symmetric around this
midpoint. So, you could have psi taking a sinusoidal expression or it could have a co
sinusoidal expression. So, this would mean that the wave function is even ok. So, it is got
the same sign on both sides of the symmetry and this would mean that the wave function
is odd, it is good find changes as you cross the symmetry. So, we can look at both these
solutions; so you have psi 1, you have psi 3, and you have psi 2, which could be either
this or that for the purpose of symmetry.
46
So, let us go through an example of having these three cases; so that is my psi 1, that is
my psi 3 and let us say we are looking at even wave functions in region 2. And let us say
that is my wave function in region 2. So, firstly, how do we identify all these terms ok,
now the first thing we do is let us say the wave function is continuous. So, the wave
function you have the box here, this was x equal to minus L by 2, that was x equal to
plus L by 2, and that is x equal to 0, now this is region 2, that is region 3, and that is
region 1.
Now, if you look at this boundary here, since the wave function has to be continuous, psi
of psi 2 at this point must be equal to psi 3 at this point ok. So, psi 2 at x equal to L by 2
must be equal to psi 3 at x equal to L by 2 ok. Now, what is psi 2 at x equal to l by 2, you
have C 2 cosine k L by 2 that is my psi 2 ok. So, this is a typo, please pardon me; there
should be a plus here, so that is that is psi 2. And what a psi 3, psi 3 is given by phi 3 is
given by B 3 e to the power minus alpha x ok. So, we saw that only the decaying term is
allowed here, so it is B 3 e to the power minus alpha x that is my psi 3. And once again
there is a typo there should be a minus sign.
And therefore, at x equal to L by 2, phi 3 takes a value of L by 2 e to the power minus

alpha L by 2. So, this is the first condition, which is simply the continuity at L by 2.
Now, since we are looking at symmetric cases, there is not you are not going to learn too
much by applying the same continuity at minus L by 2 ok, so that is not going to give
47
you too much of new information. But, we have you know whatever solution holds here,
you could just directly apply the solution on the other side just taking into consideration
the direction in which we move. So, this is the first condition.
Now, what is the second condition? So, we have these three solutions; we have the
second condition, this is the continuity you know constraint, but this first derivative of
the wave functions to also be continuous right. So, if I have x equal to L by 2 here, I
must also have not only psi being continuous, but I also need d psi by d x to be
continuous. Now, since the wave function has got a form psi 2 in this region, and it is got
psi 3 here, I need d psi 2 by d x at x equal to L by 2 must be equal to d psi 3 by d x at x
equal to L by 2. So, the first derivative of the cosine is going to be minus it is going to
yield minus C 2 k by 2 sin of k L by 2, and this is going to yield this term all right.
So, you now have this condition, and you have this condition. And now, you can go
ahead and apply the condition that the probability of finding the particle somewhere in
all 3 regions has to be 1. But, let us not head there, because let us just investigate what is
needs to be investigated this point, because those details do not really serve our purpose
too much. So, what we are going to do is just examine these two terms. Now, you see
this you have a sinusoid, you have a co sinusoid, you have C 2, C 2 and you have B 3, B
3, and you have the same kind of an exponential term.
48
So, suppose I were to label this equation as equation 2, and this is the equation 1, what I
would like to do is take divide equation 2 by equation 1. So, you will have the negative
signs have already gone here, the 2 will disappear.
And, if you were to divide this term by that term, then you will end up with the condition
that the tangent of k L by 2 is equal to alpha by k. Now, k is an important parameter for
us, because you have k defining the wave number in some sense. Now, if you ask
yourself the question, in the case of an infinite potential well, what were the values k
took. So, in that case you had to meet the condition that k L was equal to 0, and therefore
k could take integer values of pi by L.
Now, in this particular case for a finite potential well, k can only take those values,
where this equation is satisfied ok. So, if you were to plot if you were to plot both these
functions ok, let us plot alpha by k (Refer Time: 23:59) how does alpha by k vary with k
ok, it would probably vary in this manner. And then, you plot how tangent of k L by 2
varies with k ok, you will get another plot.
And if you were to take if you were to take the points at which these two functions meet,
those are the values of k that are those are the solutions of solutions of these particular
equations. And only those values of k are allowed; so no other values of k are allowed.
And then, you could use this k to define your energy, so that is the first point.
49
Now, the second point is if you consider the odd symmetric case that a side to being C 2
sin of a k of k x. You will end up with the condition that the cotangent of k L by 2 must
be equal to minus alpha by k. So, once again k can take only those values that are
permitted by the solution to this equation. So, as I mentioned it really does not serve our
purpose too much to spend a lot of time calculating the exact values of A 1, C 2 and B 3.
But, what is useful to note is that the solutions will have the wave functions will sort of
look like this. So, we have sort of sketched the wave functions here. So, in the case of an
50
infinite potential well, the wave function existed only inside the box. So, we found
solutions like you know if the wave function is to be like this. Now, in the case of a finite
potential barrier, we have already seen that psi 1 and psi 3 can have an exponential
solution. So, if you look at psi 3, it varies as e to the power minus alpha x with alpha
being a positive number, so which means the wave function has got a decay in space in
the x direction.
Now, mathematically that looks fine ok, it is not does not seem to be very difficult,
because the exponential came out from the differential equation. But, just imagine what
it is implying, you have a finite potential barrier ok, and you have got the total energy of
the particle to be less than the potential energy of the barrier ok.
So, class according to classical mechanics, the particle can never climb over the barrier,
but what quantum mechanics is telling you is that by solving Schrodinger’s equation, we
are finding that the wave function exists outside this well, and it exists by decaying
exponentially with the distance ok. So, the wave function exists, and it decays
exponentially with the distance, so that is a very big message. What quantum mechanics
is telling you, is even though e is less than the height of the barrier, the particle can exist
in that region ok.
So, if you imagine if you remember, psi is indicative of phi square is indicate with the
probability density function to find or finding the particle in any given regions. So, size
square d x is the probability of finding the particle between x and x plus d x. The
probability of finding the particle between x equal to L by 2 and infinite is now given by
psi 3 square d x L by 2 to infinite. And what quantum mechanics is telling you is that this
probability is not 0, even though the total energy of the particle is less than V naught ok.
Now, that is a very amazing result, if you think about it ok, but it is so naturally seen if
you simply solve Schrodinger’s equation.
But, this result is important; and it is in fact, used on a day to day basis, it is not that it is
a interesting exotic result that has got no implication. Now, if the particle for example, if
the particle could not exist in this region, many of the devices if you are a since we are
talking about a semiconductor device course, many of the devices and instruments that
you use probably would not work. For example, the injection of carriers at the source
51
drain, the source drain regions are doped, so as to encourage the injection of carriers
through a mechanism like this you know in a MOSFET.
If you have heard of the something called as a scanning tunnelling microscope, that do
works on the principle of trying to measure the electrons that are that have escaped the
well and are found inside a barrier ok. So, this mechanism of finding of getting the
particles to go through the barrier, if you may imagine the wave function is going
through the barrier ok, despite the fact that e is less than V naught. This mechanism is
something called as tunnelling ok; it is a technical term, but if anyone sees the word
tunnelling, this is what they are talking about ok. So, this is what they need to think
about, so that is a very important message of quantum mechanics.
So, if you have to summarize two important things, we have learned so far. The first is
that particles also have a wave, and that wave length is given by the De Broglie’s
equation lambda is equal to h by p. And the second is the fact that the wave function can
exist outside a barrier, even though the total energy is less than V naught. So, these are
two important essential points.
Now, before we wrap up quantum mechanics the topic, let us just quickly go through
another example of using Schrodinger’s equation. And this is an example of something
called as a harmonic oscillator so, we want to see you know what is the nature of the
52
wave function here. So, let us write down Schrodinger’s equation for in a very general
sense, so that is the kinetic energy, that is the potential energy, that is total energy.
And since, we are talking about an oscillator I think the easiest way to imagine one is to
imagine the equations for a spring ok. So, if you remember, the spring if you apply a
certain force to a spring, then it is got a certain spring constant ok, which I have when I
have used a symbol F for it. But, in textbooks you might probably see the symbol k it is
just that we are using k too many times. So, let us say F is the spring constant, then the
force will be given by the spring constant times the displacement (Refer Time: 30:53) so
that is. So, you can imagine you can imagine a little spring a quantum spring if you mean
that we are trying to describe here, now that is the spring constant.
And the potential energy stored in a spring is f x square by 2 that is half the spring
constant into x square, where x is the displacement. And if I were to let the spring go and
oscillate, it would have a natural oscillation frequency of omega, which is equal to the
spring constant divided by the mass per unit length, I am sorry at the mass to the power
half. And the potential energy term would be can be rewritten by writing f in terms of
omega and m as m omega square x square by 2, so that is my potential energy term ok.
So, with this introduction, let us just use this term for the potential term in my harmonic
oscillator ok, so that is Schrodinger’s equation and that is V, e is a constant, and that is
the kinetic energy. So, that would allow us to write Schrodinger’s equation in this
particular manner.
53
Now, how do we solve this? Ok. Now, this equation seems to be a little bit more
complicated, then the ones you encountered in the case of a particle in a box ok. So, how
do we go about solving this? So, we make a good guess that is usually the case for many
situations, we make a good guess. And our guess is psi is equal to e to the power minus
alpha x square by 2 and why is this going to be our guess? If you think about it the right
hand side, here has got a e that is a constant that is a constant here, there is no
dependence on x, there is no x at all on the right hand side. But, on the left hand side,
you have a dependence on x here, and you have a dependence on x in this term.
So, somewhere by solving all this out, that x dependence must cancel off ok and what is
the x dependence, it is very obvious that you have a factor that is multiplied by x square
sitting here. So, I cannot allow that to exist after I have taken the derivative and
cancelled all the terms, this term cannot exist ok, because otherwise if it dot if it did
exist, then I do not see an x square on this side, and therefore it is not a correct solution.
So, this term cannot exist.
So, what is the side that I could use, so that if I take a second derivative, it would end up
with square a term which has an x square in it some say some factor beta into x square,
such that beta would be equal to m omega square by 2, and therefore cancel off with the
potential term here ok. So, the idea is to get eliminate x square from this picture and this
seems to be a good guess for that, because, what does d psi by d x, d psi by d x is minus
54
alpha by 2 e to the power minus alpha x square by 2, and then you take the derivative of
the x squared term which is 2 x, so this is nothing but minus alpha x e to the power
minus alpha x square by 2.
What is d square psi by d x square, d square psi by d x square is minus see you first take
the differential of the x, so its minus alpha e to the power minus alpha x square by 2
minus alpha x and you again differentiate this term once more, and you will end up with
minus alpha by 2 e to the power minus alpha x square by 2 into 2 x. So, this is going to
be a minus alpha so, let us call this psi, because that is what are wave function is. So, it
will be minus and you have you have a minus and minus here that will end up being a
plus alpha square, these two terms would cancel off x square psi.
So, you see that you have a constant times psi, and you have something which is got an x
square component in it and going by argument, this component should cancel off that
component ok. So, for me to have all this work out, we need alpha square x square psi
into minus h bar square by 2 m ok, so minus h square by 2 m must be equal to this
particular term must be equal to your m omega square by 2. So, we do not need to create
in the negative, because it is going to that is the opposite sign it for you to cancel so, the
magnitudes of these two needs to be the same.
So, this m omega square x square by 2. So, this implies that my alpha so, these also got a
psi here so, pardon me. So, this implies that my alpha is going to be m omega by h bar,
so that is my alpha ok.
55
So, here is all this written out quite neatly. So, I need these two terms to be equal, and
therefore my alpha has to be m omega by h bar. If alpha is equal to m omega by h bar,
then I can now use this particular value of alpha to calculate what my e is ok. Now, these
two terms have cancelled off. So, what is left is, on the left hand side is minus h square
by 2 m into minus alpha psi is equal to E psi, implies my energy is minus h bar square by
2 m into minus alpha, which is h bar square by 2 m into alpha. And that using the value
of alpha, you have h bar square by 2 m into m omega by h bar is my total energy, and
which is going to be equivalent to your h bar omega by 2.
56
Now, what we have done here is we have only looked at the so-called ground state
energy. So, it is not a very general solution to Schrodinger’s equation. We just assumed
one particular solution, you could have it could have had constant coefficients for
example, but there are much that solution could be represent a much more general way,
and we have not done that here. But, through these simple arguments, we saw the nature
of the solution, which is the which is that the wave function must have a exponential of
minus alpha x square by 2 component ok, so that is the nature of the wave function. And
we found that the energy scales as h bar omega by 2 ok, now this is a very useful
relation, which we will come back to later on.
57
Lecture - 04
Solids: Formation of Bands, Kronig-Penny Model
So, far we have looked at quantum mechanics, which described the behaviour of the
wave particle via, the wave functions psi. And it described the physics and the behaviour
of very tiny particles such as electrons and said that there was some very interesting
behaviour such as the quantization of energy, such as the aspect of tunnelling through a
barrier etcetera, which could not be explained by classical mechanics.
We are now in a good position to start understanding the behaviour of electrons in solids.
And over the next few lectures, we will describe, what happens we will describe the
nature of electrons in an atom, and we will also describe the construction of a solid by
bringing several atoms closer to each other, and how this actually influences the
behaviour of electrons in a solid.
58
So, to start off let us just look at the atom. And we for an atom, we have a positively
charged nucleus and you have an electron, we have electrons going around this nucleus,
this positively charged nucleus. In fact, you have an electron cloud, because you only
have a certain probability of finding the wave electron at any particular location.
And in order to study, the behaviour of the electron around the nucleus. In your high
school you might have looked into several models, where in you would have equated the
electrostatic forces of attraction, with the centrifugal forces, experienced by the electron.
And also taken into account, the wave nature of the electron by placing a constraint that
there should be an integer times, the wavelength present along each present around each
orbit. And this result this would have resulted in a certain quantum number, which would
have resulted in the quantization of energies.
Now, if we solve sorting this equation for an atom, the potential energy term will have to
take into account electron, electron interactions in the atom, and the electron nuclear
interactions in the atom. And by solving Schrodinger’s equation, we find that for a 3-D
atom, we will end up with several quantum numbers. And you might also be familiar
with these quantum numbers from your high school, and you might have heard of
quantum numbers such as the principle quantum number, the azimuthal quantum
number, the magnetic quantum number, and the spin quantum number.
59
So, therefore, in an atom, we have, we have electrons occupying discrete energy levels
inside the atom. So, what I have shown here is the potential profile for a positively
charged nucleus, and in some sense this entity here represents the atom. So, if an electron
were to be present say infinitely far away from the nucleus, it would not see any of this
potential and we could say that the potential experienced by this electron is 0.
The electron would be at 0 potential and therefore, it would be completely free to move
about, and absolutely free from any forces produced any forces prescribed by this
positively charged nucleus. But as the at electron starts approaching this positively
charged nucleus, it starts experiencing this potential well and it is drawn into this
potential well and it eggs and it can experience many stable states inside deeper inside
this well. So, therefore, these different shells of an atom.
In some sense relate to these energy levels as shown in this figure. So, here you have
these different energy levels, and you have electrons occupying these energy levels, and
the outermost shell is something called as the valence shell. And the electrons occupying
the outermost shell are called as the valence electrons.
Now, these electrons are quite important, because they are the ones that decide or you
know that participate in any interaction between two different atoms. So, if we have two
atoms, it is the valence electrons of these atoms that participate in bonding or interact,
and any interaction between these atoms. So, this is the picture of the electrons in an
atom.
60
Now, what happens when you start bringing several atoms closer to each other. So, here
you have a single atom and you have all these electrons that are sitting closer to the
nuclei, and are more tightly bound to the nuclei. Whereas, the valence electrons are the
father ís I mean other, other farthest away from the nuclei and therefore, they experience
the least amount of binding to the nuclei.
Now, when you start bringing two different atoms closer to each other the electrons in
each of these atoms begin to interact, and what do we mean by that. So, let us take two
atoms A and B ok. So, let us say we have atom A, and we have an exactly identical atom,
which is called atom B ok. And let us say to start with these two atoms are infinitely far
apart ok.
So, the distance between them is infinite, which implies that atom B has no knowledge of
atom A it does not feel atom A at all; and atom A does not know anything about atom B.
So, in these two atoms the electrons would occupy, the energy levels as prescribed by
Schrodinger’s equation, and in fact, both these atoms since they are both identical the
electrons would be occupying very identical energy levels in each of these atoms. But the
moment we start bringing these atoms closer to each other. So, we let us, let us now start
reducing the distance between these atoms. So, the moment we start bringing these atoms
closer to each other, we find that the electrons now begin to feel each other in the sense
that the atoms now learn of each other’s presence.
61
Now, electrons are fermions and they, they follow something called as a Fermi-Dirac
statistics, which we will describe later in this course. And therefore, they have to obey
something called as the Pauli’s exclusion principle. And what Pauli’s exclusion principle
tells us is that it places a constraint ok. It says that note no two electrons can have all
their quantum numbers the same, which is no two electrons can have can occupy the
same exact same energy level or energy state. And, because of this limit, when we start
bringing atoms closer to each other, we find that these electron energy levels have to
start splitting ok, they need to they can both these energy levels cannot remain the same,
and the energy levels start splitting in order to accommodate the extra electrons that are
appearing, because of the two atoms.
So, this interaction between two atoms is in a simple manner described here, and we will
describe it a little better in the next slide. But what you see here is here you had the first
atom, which had it is own potential profile, and then we had the second atom which had
it is own potential profile. And the moment we start bringing these two atoms closer to
each other, these potentials these potential profiles begin to interact, and you start seeing
a potential profile that looks like this.
So, let us look at the splitting of energy levels a little better. So, I here have a very crude
drawing, which shows this aspect. So, let us say the so what we have here is a plot of
energy versus distance between the atoms. So, so this distance is the distance between
62
the two atoms A and B that, we were talking about. So, initially this distance is infinite;
so initially we start off at infinite. And we look at the energy levels of each of occupied
by these electrons.
So, since the atoms are infinitely apart, the energy levels are very are identical, and there
is absolutely no splitting in the energy levels, because the atoms do not feel each other.
But as we start bringing the atoms closer to each other, the valence electrons or the
valence shell first begins to respond the energy levels in the valence shell begin to start
splitting, because they feel the other atom the earliest.
So, as we start reducing the distance between the atoms, these outermost electrons the
energy levels here begin to split and as we bring them closer and close as it start bringing
these atoms closer and closer, the inner shells also begin to feel, the presence of the other
atom. And the energy levels and the inner shells will also begin to split. Now, although I
have just shown two lines here to indicate the splitting that the splitting will involve, so
as you bring say n atoms where n is a very, very large number to start creating a solid.
This flitting will have a large number of finally, separated energy levels ok. So, the
splitting will be very, very fine. So, you have you have the number of energy levels here
being proportional to n. So, as we start bringing in more and more atoms you will have
more, and more energy levels, and the splitting would be very, very fine.
So, if we were to say, if you were to take this particular position, let us say the distance
between the two atoms is this value here. And if we were to look at the energy profile,
because of the system of two atoms, we would find that, this energy level would be quite
discrete, I mean it would be almost single. There has not been much splitting in the
innermost shell; whereas, the outermost shells have experienced significant splitting of
energy levels.
And therefore, the outer shells would be would appear as bands of very finely spaced
energies, very finely spaced energy levels and then we would have, and then we would
have a very large gap, and these are the forbidden energies, because Schrodinger’s
equation does not prescribe solutions to the electrons. And then, we would have see
another little band, another band which is again got very finely spaced energy levels.
So, this would be the energy profile that emerges from the experiment, where and we
will start bringing atoms closer to each other. Now, what do we mean by energy levels,
63
what are these dashes that, we are drawing here. Now, each of these levels ok, so the
energy levels are nothing but solutions to Schrodinger’s equation ok. And the electrons
have to accept one of the solutions to Schrodinger’s equations.
And therefore, these their energies are very specific and they are quantized ok and so
each energy level basically provides it, it says that it is a spot that an electron could
occupy, it is an energy state that an electron could occupy or in other words, it is a
solution, it is a specific solution that they to Schrodinger’s equation that an electron
could take up. And these energy levels are also sometimes called as states or energy
states ok. So, we imply the same thing.
So, now let us look at a different picture ok. So, let us say we now, start we have brought
in several atoms ok, and these are the nuclei the positively charged nuclei of all these
different atoms. And we have arranged them, in a very periodic linear arrangement ok.
So, you have these different atoms, and we have placed them in a nice periodic
arrangement, which is just one-dimensional ok. So, we just have one dimension and
along this dimension x, we have placed these atoms.
Now, the potential energy profile or the potential profile for each nucleus due to the
positive charge of each nucleus would look like this. If I were to draw them all
separately, these potentials would all appear like this and so on ok, so that that is
probably another that is another nucleus here, that I have not shown. So, these it is only
64
at the ends of the crystal so, let us say the crystal has got a length L. So, what do we
mean by a crystal, we will define it more thoroughly later, but it is simply a periodic
arrangement of atoms.
So, let us say this arrangement has got a total length L. So, we have now brought in 5
atoms, and there is a total length L and we find that these the potentials at these edges ok,
as seen by the electron is would be 0, only at these edges. And the electron sitting here
will be completely free from, completely free and would not know anything about this
structure present here. But then as we get inside as we start entering this arrangement so,
as we start looking at the potential profile along this a direction x. We will see that these
potentials different potentials of these at different atoms. Now, interact and you will end
up with a potential profile that looks like this, so you will see the potential starting to
look like this.
So, you have different potential wells, you have a periodic arrangement of potential
wells, and you also have you also have some a kind of a barrier between these different
wells, you also have a periodic barrier between these different wells. And, it is only at
the edge, only at the edge of these crystals that the potential profile expands out in this
particular manner.
So, this is the situation when you have a periodic arrangement of atoms in one
dimension. And you could use, extrapolate this concept to any number of dimensions,
you could say a three-dimensional atom, three-dimensional solid or a two-dimensional
solid as well. And here we represent the potential term by U of x, so that is, that is going
to be the potential that is going to take part into the potential energy term in any
Schrodinger’s equation that we write.
So, now if you look at the energy levels the electrons occupy, so inside let us take this
particular atom ok. So, inside this atom, the electrons are sitting in these discrete energy
levels ok, which have now started experiencing some, some amount of splitting, because
we have started bringing in many atoms closer to each other and the electrons that are
least held by the nuclei ok. So, the so although the although, we have shown a blue circle
there, it is only to indicate the presence of the nuclei it is not, saying that the nuclei is
located there; in fact, the nuclei is located here. So, that is probably something I should
65
have been careful about, so I do apologize, but I hope this is this is this is quite clear all
right.
So, that is the blue circle they are just indicating the location of the nuclei ok, and this is
the nuclei. So, all these energies, so as we head closer to the nuclei, all the electrons that
are occupying energies closer to the nuclei are more tightly held by the nuclear. And it is
only these electrons that are the most weakly held and they are something called as the
valence electrons.
And we have also indicated a red line here, a dashed red line. So, we will not worry
about, the dashed red line for now, but this is something called as the Fermi level, and
we will encounter the Fermi level later in this course. So, it is just something that is
useful to know it is the Fermi level, and that is what this red line is. And we will
encounter this term, and describe it more carefully later in the course.
Now, if an electron sitting in one of these atoms gets excited, let us say it receives a lot
of energy either through either through an interaction of the photon or simply it just
receives a lot of thermal energy, from the from a non-zero temperature in the ambience.
This electron could get excited, and it could jump into energy levels that are higher up
ok. And if these electrons could occupy energy levels that are above, above this Fermi
level say for example, then one could say that the electron is now sitting in a different
66
potential profile, the electron is now free to move out in the solid; and the electron now
sits inside a potential profile that looks like this.
So, what this implies is that the electrons in this periodic arrangement see two kinds of
boxes. So, we looked at particle in a box. So, the electrons here see two kinds of boxes,
they see two kinds of potential wells. The first is the potential well prescribed by each
atom, so this is one box and the second is if they were to escape the clutch of this atom,
and enter into higher energy states; then the electrons would see a much wider box
prescribed by the length of the entire periodic arrangement. So, now we are going to use
all these ideas to develop a more mathematical and a quantitative estimate as with
regards to the formation of energy bands, and energy gaps in a solid and all this is done
via something called as the Kronig Penney Model.
So, what is the Kronig Penney Model? To start off with our starting point is what we
discussed, which is you have a linear ok. So, we say a one-dimensional periodic
arrangement of atoms and these are the locations of all these atoms.
And therefore, you have a potential profile that looks like these green curves shown here,
and as we mentioned the last atom that forms, that forms this linear arrangement would
exhibit a potential profile that looks like this. And you could have electrons either being
either sitting in these potential wells or they could be excited into higher energy states,
and would basically experience a potential well that looks like this ok, which essentially
67
means that the electron is free to move about in the entire potential well or the entire
crystal in this case.
So, now to in order to start quantifying, quantifying this behaviour of electrons, we need
to start solving Schrodinger’s equation and since it is quite complicated to solve
Schrodinger’s equations for this energy, this potential profile, they are going to make a
big assumption. And we are going to say that, we are going to they are going to say that
yes, this is that this is indeed the true potential profile for seen by the electron in this
linear arrangement of atoms.
But we are going to make this assumption, and represent this potential profile by
rectangular potential wells, because we know, we know the methods to solve
Schrodinger’s equation for rectangular potential wells. So, we say that, we say that, we
got it, we are going to take we are going to take this true nature of the potential profile,
and we are going to create a mimic or a model, which looks like this. So, it is a
sequential arrangements of two regions A and B.
Now, A represents is indicative of this potential well here, so that is your A and B is
indicative of this barrier. So, instead of having struck a geometry like this, we now have
a very nice rectangular geometry and we also redefine our potentials. We say that this
floor is now said to be the reference potential, and it is given by the symbol U which is
equal to 0. And the barrier height is said to have another potential, which is given by this
variable U0 naught ok, so we have redefined our potential energies so as to for it to be
applied in Schrodinger’s equation.
And further, we now consider two potential valleys, and we are just going to study these
two valleys by looking at this picture here. So, we now redefine our x coordinates, and
we say that this is the x axis, let us consider a sequential arrangement of A and B. And
we will define the width of this potential well to have to be a naught; and the width of
this barrier to be b naught. And we define the origin to be this location here the x is equal
to 0 and we define, this point here as x is equal to a naught, and we define that point
there as x is equal to minus b naught; and therefore, this point takes a natural value of x
is equal to minus of a naught plus b naught. And this term represents, represents the
periodicity of this potential map, and it is given a very special symbol a, which is called
as the lattice constant. And this periodic arrangement is something called as a one-
68
dimensional crystal ok, we will redefine the crystal much better in a few slides, but you
can just note that, this periodic arrangement of atom in one-dimensional is, is also
referred to as a one-dimensional crystal. It could have been wiser, if I had used a
different symbol, since I have already used a naught here, but this symbol a is quite often
used to define the lattice constant.
So, given this structure, we can now write Schrodinger’s equation. So, we define the
wave function in the different regions. So, we consider this region which is from x going
from 0 to a naught.
So, this particular region when 0 is less than x, less than a naught ok. And we say that we
could say 0 less than equal to. And we could say that the way the wave function in this
region is
Ψ of A. And we then consider this region B here, where x is less than less than equal to
0, less than minus b naught and say that the wave function there is psi of B. So, please
note that we have not defined the wave function in this region to be psi of A, it is not psi
of A. It the wave function takes the form psi of A only in this potential well. And
although this potential well here, is very similar to this potential well to the other to it is
twin. The wave function there is not Ψ of A, but it is related to Ψ of A and we will soon
see, what that relation is.
69
So, here we have these two regions Ψ of A and Ψ of B. And we are going to write
Schrodinger’s equation for this region, and this region. So, since we have already gone
through several such examples, it is quite easy to see that Schrodinger’s equation for
−𝒽2 𝛿 2 Ψ𝐴
region A, this particular region A is given by = E Ψ. The potential term there is
2𝑚 𝛿𝑥 2
0; the potential energy is 0. And therefore, Ψ of A has got a very general solution, which
is of the form A A, that is A in region A sin𝛼𝑥 plus B in region A cos𝛼𝑥 is equal to Ψ of
A.
Now, once again I apologize for not taking choosing these variables more wisely. So,
here A is a constant coefficient, A of A is a constant coefficient; and this represents A in
region A. And B of A is again a constant coefficient that is basically talking about it is
value in region A and the term alpha is nothing but your square root of √(2 m E)/ h. So,
this is the general solution of Schrodinger’s equation describing psi of A.
Now, in region B, in this particular region, we now have a potential term, which is not 0.
So, the potential energy is not 0 and therefore, we now have Schrodinger’s equation,
𝛿 2 Ψ𝐵
which looks like this, which is minus h power square by 2 m dou square psi B by
𝛿𝑥 2
dou x square is equal to E -Uo sin B. And the general solution to this Schrodinger’s
equation is also of this form ok, but with the, but with two different possible cases. So,
here you have the coefficients A of B and B of B that form the pre-factors to the terms of
sine beta times x and cosine beta times x therefore, the solution is A B
sin𝛽𝑥+Bcos𝛽𝑥=Ψ𝐵 of x.
Now, what is beta? If my energy the total energy is greater than the potential energy,
beta is given by this particular expression, and it is a real value, it is got a, it is a real
number. On the other hand, if E is less than U naught, which means the total energy is
less than the potential energy, we see that the wave function can exists, exist even in that
particular region, and that is, that is why the process of tunnelling, and that is something,
which is very unique to quantum mechanics. So, therefore, the wave function can exist
and beta takes the value of a it is a complex number, where i here is square root of minus
1. And, if beta is a complex number, and the sine and cosines of a complex number will
result in exponentials of a real number.
70
Lecture- 05
Solids: Kronig-Penny Model - Continued
So, now in order to make the definitions of these wave functions psi of A and psi of B
much more specific, we need to identify these four coefficients. What are A of A; B of
A; A of B; and B of B? So, the moment we identify these four coefficients, we can
clearly identify Ψ of A and Ψ of B. And how do we identify these four coefficients? We
need four boundary conditions. So, what are these boundary conditions?
71
So, you might be familiar with two of these boundary conditions. The first boundary
condition is that of the continuity of the wave function. So, since if you look at point x
equal to 0, at x equal to 0, the wave function A, the amplitude of the wave function Ψb
Ψa of A, at x=0; must have the same values the amplitude of the wave function Ψb, at x
equal to 0. So, these two amplitudes should be the same. And, therefore ΨB at x equal to
0 must be equal to A at x=0, so that is the first boundary condition.
The second boundary condition is that not only is the wave function continuous, but the
𝑑ΨA
derivatives of the wave functions, but all also be continuous ok. So, therefore ,
𝑑𝑥
𝑑ΨB
| x=0, must have the same value ok, so that is the second boundary condition. Now,
𝑑𝑥
the next two boundary conditions may not be may not be very familiar ok. And the arise
from something called as Bloch’s theorem, and which is applied when you have a
periodic arrangement or a periodic potential profile. So, when you have an infinitely
periodic potential profile, they can use something called as Bloch’s theorem.
And what does Bloch’s theorem? Now, since these two regions that is A and A here ok,
are essentially twins nature does not make any that is not distinguish between, where
these coordinates occur. In other words, see this choice of the origin is completely
arbitrary, we chose this we define this origin to be here. And, we therefore defined psi of
A to exist here, and psi of B to exist here. But if he just translate this entire potential
72
profile by one period, which is by one lattice constant. This potential well would exist
here. And therefore that potential well would then start having this wave function psi of
A ok.
So, therefore the wave function in this potential well must be related to the wave
function in this potential well. And the wave function here is psi of A. And this relation
is brought out by Bloch’s theorem. So, what Bloch’s theorem tells us is let us take the
wave function the value of the wave function at this point at x equal to a naught. So, let
us look at this point x equal to a naught. Now, this edge and this edge are twins. And
therefore the values of the wave function at this point, and the value of the wave function
of this box, at this point must have the same amplitude.
So, let us let say to make this clear. Let us call this as ΨA, and let us call that as say
some psi A naught. So, what Bloch’s theorem tells us is that Ψ(A)| x=a0 that is at this
location is related to the value of the wave function, Ψ(Ao)| x=-b0 at this location, at this
right side boundary by a certain multiplication factor ok. And that multiplication factor is
eika; where a is your lattice constant; and k is a coefficient that represents the wave, and
wave vector ok.
Now, it is not necessary for us to define this new wave function psi A naught, because
the value of Ψ(A0) at x=b0 is exactly the same as the value of psi of B at x equal to
minus b naught, because of the continuity of the wave functions in this edge. And
therefore, we can replace psi A naught at x equal to minus b naught; by psi B at x equal
to minus b naught.
And, therefore the Bloch’s theorem says it provides the relation between the wave
function here; and the wave function here. And it says that Ψ(A)|a0 = Ψ(B)-
|b=0eik(a0+bo)=ΨA0, which is the lattice constant. And this relation also holds true for the
𝑑ΨA 𝑑ΨB
derivatives, | x=0 = | x=-b0 eik(a0+bo). And, now using these 4 boundary conditions,
𝑑𝑥 𝑑𝑥
we can we can now solve for psi A and psi B and obtain a wave function that is much
more accurate. And then we can go ahead and start calculating energies. And all the
other attributes of the wave particle that exists in this periodic potential well.
73
Now, upon solving these linear equations, and after calculating these coefficients, we
will end up with the relation of this sort ok. I have not shown you the details here, but we
will end up with the relation of this particular form. So, you have a large term here, and
this term comprises of terms like 𝛼; 𝛽; 𝑎0 ; 𝑏0 . Now, 𝛼 and 𝛽 are both functions of
energy.
So, if you go back here, we see that alpha is a function of energy, and beta as a function
of energy. So, and beta are functions of energy; a naught is the width of the potential
well; b naught is the width of the potential barrier. And, therefore this entire term in
brackets is a pure function of energy, which is denoted by this function F0(E). So, we
will define a function F0(E), which is basically this expression here and it is a, it is a
function of energy. And, now that function has to be equal to this particular term here, so
we have F naught of E being equal to cosine k a, which is a naught plus b naught, where
k is the wave vector. So, we have this particular relation.
So, we first so what does this tell, so let us first sketch F naught of E versus energy. So,
we have here a plot of F0(E) versus energy E. And that curve would look like there is a
red curve here ok, to trace this particular path. But then the nature imposes a limit on
F0(E), and what is that limit, it says that F naught of E must be equal to this cos k (a0
+b0). And, therefore since a cosine function cannot be greater than 1 or less than minus 1.
74
F0(E)is also bound by these limits, it cannot be greater than 1 or less than minus 1, so this
is the limit imposed by nature.
So, nature says that the electrons in this linear arrangement of atoms must obey this
condition that F0(E) must be less than 1 less than or equal to 1, and must be greater than
or equal to minus 1. And, therefore this purely mathematical function is now bound by
these 2 bounds ok, so it says here is the bound by the bound of 1, and here is minus 1.
And, therefore nature allows the electrons to only exist take up values of F not of E,
which are these values, which are inside these bounds.
And all these places, where F0(E) exceeds 1 or goes below minus 1 are all forbidden ok.
So, the electrons cannot take up values of F naught of E of this kind. Or in other words;
if you look at the energy axis the electrons are forbidden, from taking these particular
energies, because, these particular energies permit F naught of E to go above or below
plus 1 or minus 1 respectively. So, these energies are all forbidden. And the electrons are
only allowed to take these energies, where F naught of E stays within these bounds ok.
So, we see that from a very simple mathematical relation, we see the enforcement of the
discretization or the presence of energy gaps occurring in a, in a linear arrangement of
atoms.
So, if we were to now plot E versus k ok, so we saw that F0(E)=cos k (a0 +b0), and (a0
+b0)=a. And if you were to plot this plot of E versus K, we would end up with a plot
75
looking like this ok. So, here these little blue curves are all a very finely spaced energies,
it’s almost continuous energies ok. And the electrons are only allowed to occupy these
energies. And these little S shaped stitches, if I may use the term are the energy levels the
electron is allowed to occupy. So, then electron is allowed to occupy these energy levels,
the electron is allowed to occupy these energy levels and so on.
But, at these boundaries when k takes a value of an integer times pi by the lattice
constant, at those boundaries, you find that there is a sudden jump in energy, there is a
discontinuity. So, the electron can take all these values, but then beyond this it has to
jump into the next set. So, you see these discontinuities that keep occurring at these
points, and this is the presence of an energy gap or the location of an energy gap, now
and these are the forbidden energies.
So, if you want to imagine this better, let us also plot energy versus F naught of E, the
plot of energy versus F naught of E. So, let us plot let us plot it sideways F naught of E
versus energy. So, we are looking at the same plot that we discussed here, but we just
plotting it sideways. And let us just draw little dotted lines corresponding to these jumps
and energy. And here once again let us place our bounds of plus 1 and minus 1. So, we
find that the places, where these jumps occur are the exact same spots, where F naught of
E crossed these bounds etcetera.
76
So, we found that F naught E, F naught of E was taking all sorts of values, and nature
permitted F naught of E to only exist in within these bounds imposed by minus 1 n 1.
And it is at these energy points or these energy levels that F naught is E crossed these
bounds. And, therefore nature does not permit the electrons to take up those energy
levels. And, therefore you see these gaps occurring at these points in K, which is an
integer times pi by a, so that is the first aspect that comes out from this discussion.
The second detail is that if you look at one of these blue curves ok, they appear to be
very continuous, because, I have drawn it in a very continuous manner. But, in fact they
are composed of very tiny points or tiny energy values that the electrons could take up
ok. And what are these tiny values, these tiny values are the discreteness that came
about, because of the larger box the electrons were occupying.
So, let us go back to our discussion here. So, we saw that the electrons were seeing two
potential wells. So, the electrons that were still held by the atom were seeing this
potential well given by a Kronig Penny model, but the electrons that had escaped the
clutches of the atom, and that were free to roam about in the crystal, were seeing the
potential well imposed by these bounds of the crystal.
So, it is as though you have so let us let us just draw it a little cleaner. So, this was a
Kronig’s penny model. If it take all this and transfer in to your rectangular potential
profiles, it is as though, you have this to be your potential profile. So, this is the length of
77
the entire 1 d crystal, and these are your little potential wells because of the different
atoms. And, therefore the electrons here, experience this potential well, and the electrons
there are sitting in a much wider box. And what happens, when the boxes the length of
the box increases, we saw the distance o0
r the def the difference between the energy levels, drops down significantly, and that is
exactly what this is indicative off.
Therefore, if you think of the difference in k ok, so the difference in k here is the order of
pi by a ok, so that is your Δ K, for the electrons sitting in the inside in the clutch of the
atoms. But, if you look at these finely spaced k points, so what does the order of this K, it
is the order of L, where L is the size of your crystal or the size of this periodic
arrangement. So, if you were to take a crystal, where L is about 1 centimetre, so which is
about 1 e-2 meters, then 1 by L or 𝜋/L ok, so the Δk because of these points is of the
order of a 100 ok. So, you can imagine that the distance between these two is about say
100.
And what is the distance between these two points? It is the order of pi by a and a is
basically a lattice spacing in the crystal, which is say let us say 5 angstroms, so, the order
of 1 e minus 10 meters. And therefore, pi by a so the order of 1 e 10 ok, so meter inverse.
And, therefore while this is all 100, we are talking about 1 e 10 here, so that is what the
order of 1 e 10. So, therefore these points are about 1 e 8 times more closely spaced as
compared to these jumps here. And, therefore it appears like as though it is a continuous
distribution of or continuous line or continuous bunch of states that the electrons could
occupy. So, this is the second point to denote.
78
The third point denote is that you will see that these E k stitches that, we defined have
got a little gentle S shaped nature, they are all looking like little S ok. In other words
more mathematically this lower end here could be described by a parabola, which is of
the form alpha k square, where alpha is greater than 0. So, of course there could be a
constant here, and there could be constant there, but let us not worry about that I am just
describing the general nature.
And this nature this region here the top of this S is another parabola, which is the order
of say minus beta k square, where beta is greater than 0, and therefore you have a
negative sign outside. And the reason for these this kind of a relation will become very
apparent, when we go further down, and discuss the concept of something called as the
effective mass of an electron in a crystal.
79
Now, moving to the very last the other two points or the not the last, but there are two
other aspects that we need to discuss. So, moving on, we find that these are the states the
electrons could occupy, and there are these little discontinuous jumps. But, then the
general envelope or the general trend of this plot is in itself a parabola ok.
And do we expect that? The answer is yes, because if you were to head towards classical
mechanics, you will find that your energy is nothing but your P square by 2 m and P,
which is the momentum is h bar k and h bar k square h bar k whole square by 2 m
provides a parabolic relation between E and k. And, therefore you will find that the
envelope is a parabola. So, as we head towards classical mechanics, these discrete the
discreteness of all this would disappear and we would effectively have an E k relation
that looks like this envelope shown here.
80
Now, let us continue this discussion. Now, the E k diagram that is shown here by these
blue curves; to draw this E k diagram, we arbitrarily defined the origin, we said k equal
to 0 is at this point. But, that need not be the case, nature does not distinguish between
say this point in K, and that point in K. And, therefore if you were to draw the same E k
diagram here, it we could draw another, we could draw a similar kind of a picture, but
just translated along the k axis. So, for example, this is the E k diagram that, we just
studied ok, so that is the picture that we studied.
So, you can see you can see that the origin was chosen here, but we might as well draw
an E k diagram here, which is again got the same kind of an energy profile ok. So, you
can see the little stitches just translated by this period. And once again, you could also
draw an E k diagram here and so on, and you could fill up the entire k space by drawing
such E k diagrams. And when we do that you can see this general bands of energy
beginning to appear.
81
So, let us do that completely, and you will find that, you have these little bands of energy
ok. So, if you were to, if you were to take this apart here, and draw this as a band, you
will find that there is an energy band here, which the electrons are allowed to occupy.
And this these energy bands are now composed of very finely spaced energy levels ok.
And, we already saw that the spacing here is all of the order of you know it depends on
this 1 by L.
So, they are very finely spaced energy levels. And then you have a gap, you have a
discreet gap, because nature does not allow the electrons to take up those energies. And
then you have another band, you have another band that is formed, which is also
composed of discrete energy levels and so on. So, therefore this Kronig Penny model
shows us the formation of energy bands in a more mathematical or quantitative sense.
And, we got here by simply solving Schrodinger’s equation for our simplified model,
where in we simplified the potential profile to the rectangular potential profile.
82
So, just to summarize everything we have looked at, I think this picture is quite
important ok. So, all our discussion so far has told us that, when we bring arranged atoms
closer and form, start forming, constructing, solids, wherein the atoms begin to feel each
other, then we find that the electrons are permitted to occupy bands of energy. So, you
have these different bands that are composed of finely spaced energy levels, and these
bands are themselves separated by large forbidden energy gaps ok.
So, as we move up, so what we have here is the energy axis as the y-axis. So, as we
move down in these energies, we encounter electrons that are occupying these different
energy levels, and that are more tightly held by the nucleus of the atom. And as we move
up an energies, we encounter the electrons present in the outer shells, and those electrons
are less tightly held by the nucleus.
Now, of interest to us in this course are two kinds of energy bands ok. Of course, all
energy bands are important I mean without that, you would not have an atom. But, as far
as this course is concerned, we are interested in 2 kinds of energy balance, and those
energy bands are described by these 2 Blogs that I have shown here. So, there would be a
point in this energy profile that, you will have an energy band, which has got energy
levels, which are big and these energy levels are basically solutions to Schrodinger’s
equation. And, you will find that this band of energy is mostly filled with electrons.
When you say a band is mostly filled it implies that electrons are occupying these energy
83
levels or these electrons have taken up solute those solutions prescribed, where
Schrodinger’s equation that corresponds to these energy states.
So, these energy levels are mostly filled with electrons. And just above that band would
be a forbidden gap; and further above that band, you will encounter, another band of
finally spaced energy levels, but which is mostly empty, which is mostly empty, which
implies that there are solutions that the electron could take up. But, there has been no
electron or very few electrons that have actually taken up those solutions, and are having
an energy level as defined by the energy levels in this band ok.
So, you have this mostly filled band, and you have this mostly empty band. And this
mostly filled band, typically corresponds to the outermost shell of the electron outermost
shell of the atom,. And therefore this is something called as the valence band. And these
energy levels ok, correspond to the larger potential well, which we saw in our Kronig’s
penny model. And these energy levels are something called as the conduction band, and
they are mostly empty. So, it requires the electron sitting in a valence band, which still
held by the nucleus by the clutch of the atom, but is very weakly held by the nucleus ok.
So, you consider an electron there, which corresponds to the electron in the valence
band. The moment this electron would receive some energy, it would climb up, and
across this energy gap; and it would enter something called as this conduction band. And
it enters the conduction band, it essentially sits in a much larger potential well. And,
84
therefore in this much larger potential well, the electron is free to move about in the
crystal.
And therefore, the electrons in the conduction band are free to move about in the crystal.
And therefore the word conduction, which means the electrons now being charged
carriers are free to move about, and they become carriers or they become entities, and
lead to charge transport, and therefore currents in a crystal. And this red line here is our
fermi level. Now the fermi level could lie anywhere, but just for the sake of definition,
just for the sake of connecting you at the previous picture I have located this fermi level.
And we will define the fermi level much more carefully further down the road.
So, now we are in a good, we are in good shape to describe the difference between
metals, insulators, and semiconductors. So, we are slowly heading and making our way
from the general study of solids towards understanding three kinds of solids that are of
great use for us. And these classifications are based on the conductivity of the solid as to
how much, how easily they can conduct electric current ok, and they are called as metals,
insulators, and semiconductors.
Now, metals in the case of metals the conduction band that is the mostly empty states
overlaps with the valence band, which is the mostly filled states. So, essentially what the
electron sees is a continuous distribution of energy levels. There is no energy gap or in
other words the energy gap is 0 ok, so there is no energy gap. And, therefore these states
85
are mostly filled, but it requires the electron to just acquire a very small amount of
energy before it enters the conduction band, and begins to move about in the crystal.
So, it is only at 0 kelvin that it is quite like it is that it is it is possible that, we find no
electrons in the conduction band. But, with any once the temperature starts increasing
above 0 kelvin, and particularly at temperatures, which are close to the room
temperature, which is say about 300 kelvin. It is very, very easy for the electron to enter
these higher energy levels, and occupy states in the conduction band. And, therefore
these electrons ok, as shown here this is the picture of metal, as seen from the point of
view the lattice.
So, these are the different atoms that form this periodic arrangement in the metal. And
these are the electrons in the metal. And these electrons are free to move about, because
they can easily escape the clutch of the atom and enter the conduction band. And
therefore, the metal can be said to have a sea of free electrons. And there and therefore
these free electrons would respond quite actively, and in large numbers to any applied
voltage difference. And therefore the metal would have extremely good conductivity
because of these free electrons.
So, now let us head towards the other end of the spectrum, which is basically your
insulators, so these are insulators. Now, what is an insulator? In case of an insulator, you
have the valence band states, and you have the conduction band states. And these two
86
bands are separated by a very large band gap ok. The forbidden gap is very, very large.
And for the sake of a quick estimate, we have defined the energy gap to be greater than 2
electron volt. But, this does not mean that nature is imposed a strict limit of 2 electron
volts. Nature does not classify these materials say for example a semiconductor, and
insulator by noticing as to whether the energy gap is greater or less than 2 electron volts.
In other words, nature is not sitting there saying that ok, this material has got an energy
gap with 2.1 electron volts. So, let me call this as an insulator.
Whereas, this material has got an energy gap of 1.9 electron volts, so therefore let me
classified as a semiconductor, no, that is not the case. Now, the purpose of these numbers
ok, so you can see that it is quite a vague the difference between an insulator, and a
semiconductor is quite vague. Both of them have an energy gap, both of them have
conduction band, and valence bands described like this. The only difference is the values
that we have assigned.
And what I am telling you is that these are not strict values. For example, you could
have, you could have a semiconductor having an energy gap, which is slightly greater
than 2 electron volts, in fact these are called as wide band gap semiconductors. So, what
tells a what provides this classification of as to what is a semiconductor, and what is an
insulator ok. Now, the answer has got to do with that conductivity at room temperature
ok.
87
And let me explain that a little clearer; let me explain that a little more clearly. So, let us
take an insulator. In the case of an insulator, you have some electrons sitting in the
valence band. Now, at room temperature, which is 300 Kelvin, these electrons would
acquire some amount of energy, and they would now jump to the conduction band. Now,
in the case of an insulator, this gap is so large that at 300 Kelvin, it is very unlikely that
the electrons would get a get the energy required to jump across the gap.
And, therefore at 300 Kelvin while it is possible for us to find a few electrons this count
statistically speaking is not very large. So, one way to look at the electron concentration
in the conduction band, which is the number of electrons sitting in the conduction band
per unit volume of the material. If one were to look at this particular estimate, we will
find that this number is quite low. And, therefore the conductivity or the current carrying
capacity of these carriers in the in this material that we call as an insulator is very low,
and the measurable current is extremely low, and therefore we classify this as an
insulator.
But, in the case of a semiconductor, the band gap is much smaller; there does exist a
band gap, but it is much smaller. And, therefore the electron sitting at 300 kelvin has a
good possibility of acquiring the thermal energy and getting excited to head into the
conduction band. And at room temperature it is possible that, we find a large enough
number of electrons in the conduction band that, we can now have a measurable estimate
88
of the conductivity or the current. So, it is this aspect that defines and distinguishes a
semiconductor from an insulator.
And, therefore in a semiconductor, we will see that the electrons do participate in all
your bonding. So, for example here we have shown something, which is equivalent to
silicon lattice. And you do have electrons participating in bonding, so these are your
valence electrons that are being shared, and that are participating in bonding. But, upon
receiving some amount of energy say thermal energy, it is possible for the electron to
escape the clutch of this atom clutch of the nuclei, and enter into the crystal, and begin to
conduct through the crystal. And this electron, which escapes from this clutch leaves
behind a vacancy ok and that we vacancy is something called as a hole.
89
Lecture - 06
Solids: Electrons and Holes
So, now let us describe this picture in a semiconductor much more clearly. So, we say
that we have a conduction, we have a valence band which is mostly filled states and we
have a conduction band which is mostly empty states. And what do you mean by states?
You have these potential solutions to Schrodinger’s equation that in a valence band is
definitely taken up by electrons; whereas, you have potential solutions to Schrodinger’s
equations that are not yet taken up by electrons in the case in the conduction band.
Now upon receiving some energy; so, these are electrons that are held in the outermost
shell of your atoms. So, upon receiving some energies for example, say thermal energy;
the electron can get excited and can jump across this forbidden gap and into the
conduction band states and occupy a solution to Schrodinger’s equation that defines it to
have an energy level in the conduction band. Now this electron is free to move about in
the crystal and it is something called as a free electron. It is of great interest to us to
count the number of free electrons in the crystal because that is the; it provides a measure
of the current through the crystal.
90
Now, the electron that has left the valence band and moved into the conduction band
leaves behind a vacancy which is called a hole ok and the hole is basically a state that
has now become empty ok. So, there was a state which is occupied by an electron, the
electron left that state and moved into the conduction band and therefore, that state has
now become empty and it is called a hole. So, therefore, there are plenty of holes in the
conduction band or plenty of vacancies in the conduction band, but we are not interested
in those. We are only interested in the vacancies in the valence band because once again
it is these vacancies that can lead to that can add on to charge transport and we will see
how in a minute.
Now it is possible for the electron since these energy levels are all very fine very small.
It is possible for the electrons in the valence band to jump into these vacancies and
therefore, create a new vacancy. So, let us say that these electrons were to jump into
these vacancies. And therefore, this vacancy which is located here would appear to move
through the states in the valence band because, of electrons job occupying it ok.
So, it is possible for the hole to appear like as though it is moving in the valence band
because of these electrons from other states are coming into occupied, thereby leaving
the previous state vacant.
Now, in the case of crystalline silicon; in the case of crystalline silicon which will be in
some sense is our role model of study. Let us say three dimensional crystalline silicon
which is a basically a periodic arrangement of silicon atoms in 3 dimensions. We find
that at room temperature, at 300 kelvin, we do have about 1e10 electrons per CC that is
per centimetre cube being promoted from the valence band and into the conduction band
states ok. So, this is when the silicon atom is pure and it is kept in the dark with no light
no voltages or no currents through it. So, it is that something called as a thermal
equilibrium condition. It is it is possible for electrons to move in and therefore, it leaves
behind holes which have the same concentration.
Now here we also define these three energies the bottom of the conduction band ok. So,
this is all the conduction band states the bottom of the conduction band is given this very
special symbol that we will use throughout the course and it is called as E c. The top of
the valence band is something called as E v and this gap which is the forbidden gap it is
EC-EV is given a symbol Eg.
91
So, let us continue discussing this picture here. So, we saw that it is possible for the
electron to take up energy and jump into the conduction band. Now we are going to have
several definitions.
Now, the first definition is to do with something called as an intrinsic semiconductor. An

intrinsic semiconductors essentially a pure semiconductor that is, it does not have any
impurities such as dopants present in it. The second definition has got to do with
something called as a thermal equilibrium condition which means that you have prepared
the semiconductor and we keep the semiconductor in the dark; that is, interest it is not
exposed to any photons. And we also do not apply any electrical bias say for example,
voltages or currents across the semiconductor.
The semiconductors kept as is in the dark and the only source of energy for the electrons
could be the thermal energy due to their ambient temperature. And this thermal energy is
given by kT where k is the Boltzmann’s coefficient. So, please do not confuse this term
now with the wave vector. It is the Boltzmann’s coefficient and T is the temperature. So,
that is the order of the energy given to the electron due to due to the ambient
temperature.
Now a material in thermal equilibrium is in no way silent ok. It is not a very quiet
material. There are many processes happening in a material in thermal equilibrium.
92
So, for example, an electron in the valence band could acquire the thermal energy and
could jump into the conduction band and you will find many such electrons doing it. But
statistically speaking on an average nothing changes, which means that if there are n
electrons that jump into the conduction band; you have n electrons that are coming back
into the valence band. So, this process of creation is very finely balanced by the process
of destruction. Destruction implies the loss of free carriers which is a more appropriate
term as recombination which we will look at later in the course.
And also you could have electrons moving about in the crystal. So, an electron and the
conduction band few electrons could be moving about in one direction, but this
movement basically which basically constitutes a current is completely balanced by an
equal number of electrons moving about in the opposite direction. So, there is a lot of
randomness and a lot of events happening at thermal equilibrium, but statistically
speaking these are all averaged doubt in time and you find that there are zero currents
and zero shift from the so, called equilibrium position at thermal equilibrium. And this
complete balance is something called as a detailed balance.
Now, if you were to take an intrinsic semiconductor, an intrinsic material. It needs a pure
material and keep it at thermal equilibrium which means that you do not throw light on it
etcetera. And the only source of energy for that is the ambient temperature ok. So, be it
be give it some temperature T and therefore, some thermal energy T and we take an
intrinsic material and keep it at equilibrium. Then we would find that there is a certain
number of electrons that acquired this thermal energy and jumped into the conduction
band states. And this concentration of electrons in the conduction band states which is
basically the number of electrons above E c per unit volume of the material. So, this is
the electron concentration in the conduction band states is given a symbol n. And the
number of holes in the valence band per unit volume; that is a number of holes below E v
per unit volume or the number of vacancies is given a symbol p.
Now in an intrinsic material at thermal equilibrium, this there is a certain value for n.
And for an intrinsic material that value is given a special symbol and it is called as the
intrinsic carrier concentration which is a technical term which you will use quite often
and it is given a special symbol n i.
93
So, yes in a material in general you have n number of electrons per unit volume in the
conduction band, p number of electrons per unit of, p number of holes per unit volume in
the valence band, but for this special condition where you choose an intrinsic material
that is without impurities and you keep it in thermal equilibrium. Then n would be equal
to the special value which is n i and in that case, what would the value of p be equal to?
It would also be n i and, why is that? It is because in such a case, it is only the electrons
that jumped from the valence band to the conduction band that contributed to n i and
these electrons would have left behind an equal number of holes in the valence band.
Therefore, at in this condition your n and p would at thermal equilibrium, you will find
that n=p=n i
Now, in silicon n i is about 110 per CC at 300 kelvin, but n i is clearly temperature
dependent. It is very strongly temperature dependent and we will see this dependence
dependency much more clearly later on. And n i is temperature dependent, why? It is
temperature dependent because if I increase the temperature, more electrons could take
up the ambient energy, the thermal energy and get into the conduction band.
So, just before we conclude this discussion one more final aspect. So, in the case of
silicon for example, the energy gap is of the order of 1.12 electron volt ok. So, that is the
energy gap in 2 dimensional crystalline silicon. Now at 300 kelvin, what does the order
of the thermal energy provided to you know provided to the electron? The order is the
order of k, it is of the order of k T where k is the Boltzmann’s coefficient which is of the
order of 10-23 joule per kelvin and T is the ambient temperature which is about 300
kelvin. And therefore, this product has got a value of about 10 power minus 21 joules
which is of the order of 10 power minus 21 divided by 1 e minus 19 which is 0.01
electron volt. So, does that imply that since the electron has got 0.01 electron volt has
thermal energy, does it imply that the electron that no electrons can climb across the gap
and into the conduction band at 300 kelvin? The answer is no that is not the case.
Now, this value of energy which is k T is an average value and there is a large
distribution. If you were to look at the number of electrons versus the energy they
possess, because of thermal energy this distribution has got a very large tail. And the
average energy possessed is about of the order of k T, but you do have enough electrons
having an energy that is larger than your 1.12 electron volts and it is these electrons that
can actually jump across and into the conduction band. And the number of such electrons
94
per unit volume, in the case of intrinsic silicon at 300 kelvin is your 1.05X1010 per cc per
unit volume that is, particularly per centimetre cube at 300 Kelvin.
So, let us now start to look at the impact of applying an voltage and external voltage on
the electrons in a crystalline semiconductor. So, let us say, we have a semiconductor
crystal and we now attach a battery across this semiconductor crystal. We apply a
voltage v a across the semiconductor crystal. So, let us say that is the positive terminal of
the battery that is the negative terminal. And therefore, this establishes a low electric
field inside the semiconductor crystal. So, we will denote the electric field by this
symbol and we will say the electric field is it is a weak electric field inside the
semiconductor crystal.
So, now in the semiconductor we have already seen that we have the conduction band
states and we have the valence band states. And at some temperature given some
temperature T; so, let us say the T is greater than 0 kelvin and let us say you know for
some practical purposes the T is close to say 300 kelvin. And we see that the electrons
would take in some thermal energy and some of these electrons will be able to jump
from the valence band states into the conduction band states. And in the conduction
band, these electrons are free to move about in the crystal. So, you therefore, have free
electrons in your semiconductor crystal and these electrons are free to move about.
95
So, the moment these electrons which have basically got a negative charge so they have
got a charge of minus 1.6 into 10 power minus 19 coulombs. So, the moment these
electrons see this electric field they will respond to the electric field and they will start
heading towards the positive terminal of the battery. So, all the electrons will start
moving towards the positive terminal of the battery. And this movement of the electrons
is something called as it drifts. So, drift is a very particular term used when the electron
moves because often applied electric field ok. It is a very characteristic charge transport
mechanism which is the response of electrons to an applied electric field and we will
look at this idea in much more greater detail later on in the course.
Now, the movement of the electron in the semiconductor lattice is quite different as from
the way it would move in say vacuum. So, if an electron were to be placed in vacuum
and if an external electric field or if apply if a voltage were to be applied across these
plates and vacuum, we would see that the electron would experience a force and
therefore, it would start begin to accelerate. And we could write Newton’s laws of
motion to say that the mass and the acceleration of the electron is equal to the force that
the electron fields.
Now this mass which we would take for an electron in vacuum is something called as the
rest mass of the electron or you know the mass which is of about 9.1X 10-31 kilograms.
But in the semiconductor crystal the movement of the electron is a lot more complex. So,
firstly, imagine the electron sitting in a lattice ok. So, these are all the lattice atoms. So,
you have you have all the silicon atoms in the semiconductor lattice. And these silicon
atoms have got a positively charged nuclei and it is got an electron cloud which is
negatively charged. And now we have this one free electron that has that is migrating
through this lattice. So, let us say that is our free electron that is responding to the
applied electric field.
Now this free electron has got a negative charge. So, since this free electron has a
negative charge, it will tend to polarize all the nearby atoms of silicon. So, all the nearby
silicon lattice atoms will now get polarized in such a manner that the new positively
charged nucleus is drawn towards the electron and the negatively charged cloud of
electrons is pushed away. So, we will see all the silicon atoms surrounding in the vicinity
of the free electron getting polarized.
96
So, now the movement of the electron is quite unique right. So, you have these positively
charged nuclei closer to it and therefore, this polarization effectively creates a kind of a
trap for these electrons. So, if you want to imagine in the movement of an electron in a
semiconductor crystal in a more poetic fashion, let us say you are asked to run on a very
hard surface ok. So, you have a person running on a very hard surface. So, on a very hard
surface with each step of the foot, there is a good normal reaction and then the person
could actually run very quickly.
But now, let us say a person is asked to run on a very slushy ground. So, you have a
slushy ground when the soil or the surface is very soft. So, with each step of the foot the
soil would actually sink inside and then you will have to pull the foot out and then take
another step in the soil will sink in again. So, it would appear to be you do not appear to
be much more difficult for the person to run on this very slushy ground. Because each
time a person puts his foot down. His foot is stuck in a small gravitational potential well
or you could also imagine a person running on a rubber sheet wherein with each step the
rubber sheet is distorted and you know it is not it is not a very friendly surface to run on;
although it could be a lot of fun.
So, the movement of the electron in the crystalline semiconductor could be imagined to
be exactly that right. So, with each with this polarization, the electron sinks into a little
well which is created by these positively charged nuclei and then, it continues to move
about and then it polarizes another bunch of atoms and so, on and so, forth. And this
entity which constitutes these polarized atoms and this electron is something called as a
polar. And this is something that we will never use in this course, but it is just a
definition that I am giving you or since we are discussing this. So, that is not the only
complexity with regards to the movement of the electron in an in a crystalline lattice.
So, if there is an electric field, the electron almost never goes through the semi-conductor
without interacting with the lattice ions in a much more stronger man. So, the electron;
so, all these lattice atoms are also vibrating about because of thermal energy.
So, you can imagine the silicon atoms to be to be a bunch of you know hard spheres
connected by springs. And there is a certain spring constant dependent depending on the
kind of bonding etcetera. So, these atoms are all jiggling around because of the thermal
energy that is available and this at this movement of the lattice the lattice points or the
97
this movement of these silicon atoms results in a very strong interaction with the free
electron and results in the free electron getting scattered. So, the electron the way it
would move in response to an electric field in a semiconductors if you have an electric
field in this direction the electron would move, it would scatter off from this atom. It
would move away in another direction, it is scatter from through another atom, it would
probably encounter a defect somewhere which could be a impurity or a dopant ion it
could be a charge defect for example, and that would result in some other kind of
scattering and so on, so forth.
So, it is a; so, the electron tries to accelerate due to the application of the field. It does in
this little miniscule length scale, the electron does accelerate. So, you could say that the
electron is trying to accelerate in that little length scale, but then it very quickly scatters
and exchanges momentum and comes to a rest. And then it accelerates once more and
then scatters again and so and so forth. Therefore, the effective movement of the electron
in the crystal would appear to have this periodic bunch of velocity jumps.
So, you will see the velocity increase and then, there will be a scattering and then there
will be a velocity increase and scattering and so on so, forth and therefore, unlike in
vacuum the electron does not completely accelerate throughout it instead exhibits a
constant velocity. And this is something called as the drift velocity of the electron and it
is a technical term and it is a term that describes the velocity of an electron in response to
an applied electric field which we have already defined as the drift transport mechanism.
So, if we want to continue modelling the electron using Newton’s laws of physics. You
know some you know some very simple relations, we need to consider all these
complications with regard to the microscopic phenomena, which is the microscopic
aspects of the electrons interaction with the lattice. So, that is the movement of the
electron with regard with respect to the applied electric field.
But now, let us look at the hole or the vacancy here. So, you also have vacancies in the
valence band. So, you had your you have the electrons in the conduction band that are
moving through the lattice in this particular manner. But then you also have these pots or
these empty states left behind by the electrons that got promoted to the conduction band
and these vacancies were called as holes. And essentially these are solutions to
98
Schrodinger’s equations or these are energy levels that another electron could come in
and occupy.
So, let us say that you have these states all being filled with the electrons. Now these
electrons too would like to move in response to the electric field ok, but they do not have
enough energy to jump across into the conduction band and then move through the
crystal. But instead they could jump from one location in the valence band to another and
because the energy required there is quite small. And therefore, in some sense respond to
the applied electric field.
So, let us say that this electron were to jump into this vacancy which implies that it
would leave behind a vacant, it would occupy the state and leave behind a vacancy here.
And further let us say, this electron now jumps into this vacancy which implies that it
would occupy this state and create a vacancy there. And if this process were to continue,
you will find that the electrons all in the valence band also try to respond to the field and
try to move towards one side. And it would appear like as though the vacancies they are
moving towards the other side in the opposite direction.
So, therefore, in response to an applied electric field; it is not only the it is not only the
electrons in the conduction band that responds to this field, but it also appear that the
holes in the valence band are responding to this applied field and they are moving in the
opposite direction as the as compared to the electrons in the conduction band. So, that
would that that is the picture that is the effective picture, the holes would represent
because as the electrons move here it would appear like as though the holes or these
vacant states are moving in the opposite direction.
Therefore, we can in fact, treat the hole as a particle as a charged particle that can
contribute to charge transport because it does participate in charge transport, but we can
assign to it a positive charge of 1.6X10-13 coulombs. Therefore, we can say that the
electrons ok; so, just to summarize in this discussion, we can say that the electrons in the
conduction band have a charge of minus 1.6X10-19 coulombs and contribute to current.
But the holes in the valence band have a charge of plus 1.6X 10-19 coulombs and also
contribute to current.
Therefore, in a crystal in a semiconductor crystal; if say this is the semiconductor crystal,

we would have a in response to an applied voltage say let us say we have applied a
99
voltage V a. So, that is the positive side and this is the minus negative side. It would
appear like as though there are two kinds of particles in the semiconductor. You have the
free electrons. So, these are the electrons in the conduction band and then you also have
the holes which are the holes in the valence band and these electrons will have a negative
charge and these holes will have a positive charge. Therefore, in response to the electric
field since the electric field direction is in this as shown here, the electrons will all run
towards the positive side and the holes will all run towards the negative side of the
battery ok.
So, this is the effective response of the electrons and holes to an applied electric field.
So, therefore, we could treat the holes as particles of positive charge that are running in
the opposite direction to the electron and therefore, these two currents would add up.
You will have if you look at the effective current, since you have negative charge due to
the electrons moving in this manner, you have a current due to electrons moving from
the positive side of the battery to the negative side of the battery.
And since you have the positively charged holes running from the running in there in this
manner, you will have a current due to the holes also moving from the positive side of
the battery to the negative side of the battery. And therefore, the hole current and
electron current would add up to constitute the total current in response to the applied
electric field. Therefore, we need to have a combination of empty states and carriers in
order for charge transport. So, just to give you some analogies here ok.
So, let us let us play around with this. So, if you if you were as kids you might have
played with this little game which would have which would appear like that. There
would have seen a lot of tiles and these tiles would have had say a certain picture; say
some picture on them that you need that he needed to complete by arranging these tiles
by moving these tiles around and arranging them in proper order or in a proper sequence.
And you might have always seen that there was one empty slot in this puzzle and
therefore, we could move these tiles around and therefore, complete the picture.
Now if I were to ask you as to what would happen as to how quickly would the tiles
moved, if there were no empty slots. So, let us say all the slots had a picture in them and
if you were asked to solve this puzzle by moving these tiles around the, answer is
probably is that you cannot solve the puzzle because the tiles cannot move around. And
100
why cannot they move around? Because there are no empty states or empty places for the
tiles to move about; two tiles cannot occupy the same spot that is exactly the case with
electrons.
On the other hand, if I were to have no tiles at all no tiles at all and give you only empty
spaces and asked you to tell me and asked you to move the tiles around or tell me or
measure the flux of the tiles in this game. The answer is zero because you have only
empty states or empty spots and there are no tiles for you to move around ok. So, that is
equivalent to having only empty states and no cap free carriers. So, again the answer is
zero, but on the other hand if I were to give you a few tiles and if I gave you a lot more
empty slots instead of one, I gave you so, let us say three empty slots. Now, it should be
much easier and you would probably solve the puzzle a lot quicker because, you can now
very quickly move a lot of these tiles around. So, it is a combination of empty states and
free carriers that leads to current.
101
Lecture - 07
Solids: Electrons and Holes – Continued
So, we now do realize that the movement of the electron in the crystal is quite
complicated. The electron scatters through all the lattice atoms and it sort of makes its
way through the crystal and it gives you a feeling that it is moving at a constant velocity
which we said was the drift velocity. And this is in response to the applied electric field
in the crystal.
Now how do we quantify the drift velocity? It so happens that for low electric fields, the
drift velocity is proportional to the electric field for low electric fields and this constant
of proportionality is a quite an important parameter and that is something called as the
mobility.
102
So, let us understand this mobility a little better ok. So, we have now said that the
movement of the electron through the crystal has got all this complex physics of it
polarizing the atoms and its scattering with the lattice and its scattering with any
impurities etcetera. And that movement of the electron in response to the electric field,
allowed it to move at some constant velocity and that velocity was called as a drift
velocity. And that drift velocity was related to the electric field by this parameter
mobility which is mu. So, how does this parameter mobility relate or how does it connect
with the physics of all the scattering ok?
So, we see that the parameter this mobility is a very strongly temperature dependent
parameter ok. It is very temperature dependent and why is that?. So, we now we consider
the different scattering mechanisms or the mechanisms by which the electron interacts
with the lattice and in particular we look at two kinds of scattering mechanisms: one is
the scattering with a lattice and the other is the scattering with the charged impurities in
the lattice. So, you have these silicon atoms in the lattice and as I mentioned these silicon
atoms are all vibrating around ok. So, although I am using the word silicon, it is true for
any semiconductor lattice ok.
So, these atoms are all vibrating around and they are all connected to each other because
of all the bonding and you have this massive spring mass structure that is vibrating
103
because of the thermal energy. And we have this electron that is interacting, that is
scattering with all these vibrations.
Now, since the source of this vibration or the cause of this vibration is temperature as the
temperature increases the vibration becomes more and more vigorous and the scattering
probability also increases significantly. Therefore, it is expected that as temperature
increases, the scattering of the electron with the lattice would also increase. And if one
were to define a mobility or the drift velocity of the carriers in response to the electric
field due to lattice scattering alone to the purely due to the lattice scattering. We will find
that the mobility decreases with the temperature and that relation is approximately given
as mu is to be proportional to T to the power minus 3 by 2. Of course, in reality you want
to extract this coefficient in a more empirical manner.
Now the electrons also interact with charged defects in the lattice and a good example of
charge defects is are ionized dopants and we will look at ionized dopants again in great
detail further down the course. So, let us just assume that you have the lattice and at
some place the lattice atom which is say silicon has been replaced by some other atom
which has actually got a charge. And therefore, the electron is moving through the
crystal and it tries and it interacts and it interacts with this charged ionized impurity. And
that scattering is also temperature dependent ok, but it is got a different argument or a
different kind of relation with temperature.
So, in this case it is quite important to see as to how much time the electron spends in
interacting with the charge defect. If the electron spends a long time interacting with a
defect, the scattering will be larger and the mobility will come down. But as temperature
increases as temperature increases, the velocity the thermal velocity of the electron also
increases. And therefore, since the electron has got a larger energy the electron gets past
this impurity without spending too much of time interacting with it ok. So, you might if
you would like to imagine it, you might want to imagine a very rapid interaction with the
with the impurity as the electrons sort of flies past in its vicinity. And therefore, as the
temperature increases the mobility due to it charged impurity scattering alone. So, if you
were to define a mobility and a drift velocity due to charged impurity scattering alone,
we will find that that mobility increases, because the electron is going to have a larger
velocity and its going to spend lesser time interacting with that impurity.
104
Now clearly the number of such impurities also matter. So, as the number of such
impurities increase or if the impurity concentration increases, it is more likely that the
electron will spend larger amount of time interacting with these impurities. And
therefore, the mobility will go down as the concentration increases. So, the mobility is
said to be due to defect or charged impurity interaction is said to be proportional to T to
the power 3 by 2 divided by, where N is the concentration or the number of impurities
per unit volume of the semiconductor.
Now there could be many other scattering mechanisms. For example, they you could
have something called as a piezoelectric scattering. You could also have electrons
scattering etcetera and although we have not considered every scattering mechanism in
detail. Let us just say that all the other scattering mechanisms affect mobility which is
defined as u in this little expression as u other.
So, let us say there is a mobility due to all the other scattering mechanisms. So, these are
the individual definitions that we gave that is the mobility due to lattice alone is
proportional to T-3/2. The mobility due to defects or charged impurities alone is
proportional to T-3/2/N ok. And then we could have other mechanisms that contribute
you know that define a different physics for the electron interaction with the lattice.
So, given all these different separate impacts on the electron mobility, what is the
effective mobility of the electron in the crystal? So, that effective mobility the
calculation is given by a very simple empirical formula which is got a small logical
argument with regards to you know how the electron interacts with lattice. And it is
given by something called as the Matthiesen’s rule which says that if you have these
different individual mobilities due to different mechanisms, then the effective mobility of
the electron in the semiconductor is given by is related to these different mobilities as 1
by mu effective is equal to 1 by mu L plus 1 by mu D plus 1 by the mobility through any
other mechanisms. And this is something called as Matthiesen’s rule and it says in some
sense it does make some good predictions, but it need not always make accurate
predictions. But it is a useful method or an approach to know.
Now we said that the scattering with a lattice is dependent on acoustic vibrations or is
dependent on thermal it is dependent on thermal energy and it is dependent on the
vibrations of the atoms in the lattice.
105
So, since we are talking about the electron interaction with the lattice and in particular
land the vibrating atoms or the lattice sites, it is useful to define a concept of an entity
called as a Phonon.
Now, it is the vibrations of all the atoms in the lattice that results in the scattering of
electrons and that results in one of the methods that determine the one of the phenomena,
the determine or limit the mobility of the electrons in the semiconductor. Now these
atoms could be thought of or could be modelled as a large spring mass a networked
spring mass system and by looking at you know the vibration and the vibration modes of
all these atoms, one could write a set of this equations and differential equations that
included the displacement of all these springs ok. And you have a n atom periodic a
linear periodic arrangement of n atoms and we could sit down and write out a differential
equation that described this spring mass system.
Now as a consequence of such a differential equation it so, happens that the behaviour of
atoms in a crystal a vibrating due to due to a source of external energy, results in
localized travelling waves in the crystal and these localized travelling waves of atomic
vibration in a crystal are is something called as the phonon ok. And therefore, since the
phonon is directly associated with the lattice vibration it is clearly going to be playing an
important role with regards to the thermal properties of the solid. For example, let us say
106
the thermal conductivity of the solid. So, therefore, the phonon can be thought of as the
particle equivalent of lattice vibrations.
Now, since we are talking about vibrations we can make a direct connection or to the
harmonic oscillator which was the example we considered by looking at Schrodinger’s
equation and we saw that the energies in the harmonic oscillator were all quantized. And
therefore, it can be expected that the lattice vibrational energies in a crystal are also
quantized or in other words the phonons can only take certain discrete energy levels or in
the phonon energies are also quantized.
Now off interest to us in this course is the interaction of the phonon with an electron
which we will look at when we talk about thermal recombination generation
mechanisms. But a clue or you know a good indicator to what this interaction of a
phonon with an electron does is that, a phonon after interacting with electron does not
increase the energy of the electron too much ok. But it is got, but it is it is got the
potential to significantly affect the momentum of the electron ok.
In other words, the energy E, if you look at think of an E K diagram, E K diagram which
we have already seen before an interaction with the phonon has got the possibility for the
electron to move a large distance in K space, but it does not have the ability to transit
much in the E direction. And this plays a very important role while deciding the
generation and recombination mechanisms of electrons in something called as an indirect
band gap semiconductor which he will come back to when we look at the generation of
carriers.
107
So, finally, we will take one more step to try and understand and model the concept of
mobility a little better. So, we have seen the way we have defined mobility so, far is that
we have said that the electron response to an applied electric field by moving by drifting
through the crystal and having a drift velocity v d which is equal to say some mu times
the electric field. And this was true for low electric fields and this mu was defined to be
the mobility. And we said that this mobility is essentially governed by microscopic
phenomena such as the electron interaction with the vibrating lattice, the electron
interaction with charged defects etcetera. But let us say we would let us say that we want
to quantify the mobility and develop relations that are a little bit more suitable for
calculations and for modelling.
So, one approach to doing that is by using a concept called as the effective mass of the
electron and what I mean by this is. Let us say you have you have electron moving
through the crystalline lattice ok. You have all the atoms in the lattice and the electron is
accelerating because the applied electric field, then it quickly scatters through the lattice
and then, it accelerates again quickly scatters and loses its velocity etcetera. So, because
of this very complex movement of the electron in the lattice, it is quite impossible for us
to use Newton’s laws of motion directly. But let us say that we really like to use a simple
model for the movement of the electrons. For example, we would really like to define the
force experienced by the electron as a rate of change is the momentum of the electron.
And try to use Newton’s laws, then we must make some arrangements or we must come
108
up with some models that have got some artefacts in it. And one such artefact one such
modelling parameter is something called as the effective mass.
So, we say that the electron in vacuum has a mass of 9.11X 10-31 kilograms. But the
electron in a crystal does not have this mass. We will not use this mass because we want
to develop a model that is very simple and that can describe the movement of an electron
in a in a crystalline solid in a manner that is very close to Newtonian mechanics ok. And
then we say that let q E be the force experienced by the electron and that force is equal to
the mass which is now not 9.1X 10-31 kilograms, but by this effective mass of an electron
which is a modelling parameter that we define.
So, we say that the effective mass of an electron is m n star n stands for the electron and
m p would stand for the hole and the star there says that it is the effective mass. So, we
say that q into E which is a force is equal to the rate of change of momentum which is
the mass the effective mass of the electron into the drift velocity of the electron divided
by t collision. And what does t collision? It is the mean collision time of the electron.
So, we say that the electron has managed to acquire some velocity and the mean time
before width which it is going to lose that velocity is something called as the mean
collision time t collision. So, it is typically around these expected values of time which is
a statistical estimate it is expected that the electron undergoes a scattering event after
every d collision seconds. And therefore, this equation defines the movement of the
electron as per our simple model ok. So, that is the kind of model that we would like to
develop and if we continue using this model, we see that the drift velocity is nothing, but
mu times the electric field. And therefore, this mu which is the mobility automatically
becomes defined as q.tcollision/m* star where m is the effective mass of the electron.
Now, while performing calculations using such an equation, it is important to note that
you will get the mobility in terms of meter square per volt second which is the units of
mobility. It will be in meter square per volt second. So, what is this effective mass?
109
So, we said that we would like to define or you know create this parameter called as the
effective mass of the electron, in order to write down a simple expression governing the
motion of the electron in the crystal. And therefore, we defined mobility saying that the
mobility is q into the mean collision time divided by the effective mass of the electron.
So, that was the electron mobility.
So, we would like to define the effective mass in a more careful manner. So, how do I
extract, how does one extract? What this effective mass is? What value you should be
used for this effective mass? Is it a guess or is it is a why is it a wild guess that we just
use in to fit any data or is there at least some starting point for this guess ? The answer is
that there is a starting point for this guess and that starting point comes in from our E K
diagram ok.
So, let us say let us go back to the E K diagram that we used when we define the Kronig
Penney Model etcetera. And there we had these s shaped stitches and I had mentioned at
that point that these s shaped stitches the edges of those could be modelled as little
parabolas. So, let us come back to that point.
Now, here is an example of an E K diagram for a certain crystal. You have an clear
energy band gap which is got a this crystal has got a band gape E g. So, those are the
conduction band states and these energies correspond to say those conduction band states
110
and here are the valence band states and these energies correspond to the valence band
states. So, these are all the filled states and those states are mostly empty.
So, if an electron were to jump into one of these conduction band states. Let us say an
electron is to go there ok. So, which means the electron is free to conduct through the
crystal and therefore, it has got a certain mobility and therefore, in order to calculate this
mobility, we need to identify the effective mass or its vice versa. We know the mobility
and we can therefore, calculate the effective mass. So, how does the effective mass
appear in this picture ok? And the answer is given by this little relation here while the
derivation for this has to be a little bit more careful. We will do something that is a bit
more crude.
So, let us say that the energy is nothing, but p square by 2 times the effective mass and p
is nothing, but our momentum which is (hk)2/2 times the effective mass. So, what is
dE/dk ? The answer is it is minus it is its sorry it is (h)2k/m* where m* is the effective
mass. And what is d2E/dk2? The second derivative we say that it is h bar square by the
effective mass. And therefore, the effective mass can also be written as h2d2E/dk2 h bar
square by d2E/dk2 which is the expression given here. So, therefore, if we know if we
have the energy E k diagram for a crystal, we take these energies this E k region which
the electron occupies in the conduction band.
And we say that we will that we would fit the best fitting parabola to this region and let
us say that parabola has a relation such as 𝛼𝑘2 plus say whatever coefficient let us say that
is Eg. So, it is got its got an E k relation that goes as alpha k minus k naught square plus
E g in a very general manner. And once this fit is made the term d square E by d k square
is given by 2 𝛼. And therefore, the effective mass can be said to be h bar square by 2 𝛼
Now, note what happens in the valence bands ok. So, what is important to the valence
bands is that you have holes that are the charge carriers. So, let us say you have empty
states in the valence band. So, let us say we have a hole here ok. Now let us for a second
imagine that the hole is now going to behave like a particle and if you are going to use
the same kinds of laws of motion that is a relation between the force and force and the
effect the sorry the mean collision time and the mobility and the effective mass for a
hole, we can calculate the mass for a hole using a similar argument ok. But since this
parabola is inverted, we find that we end up with negative values for the mass of hole.
111
So, the hole could be imagined to be having a negative charge in response to the that
moves in an opposite direction with regards in response to an applied electric field or it
could be assumed to be having a negative mass which is moving in the opposite direction
in response to the applied electric field.
So, in order to understand the velocity of carriers in response to an electric field better,
we will now look at these 2 plots. Now these plots show the rest of the velocity of the
carriers in response to an electric field for a very large electric field range. So, in
materials like silicon, we will see that the drift velocity of electrons typically follows this
kind of a profile.
At low electric fields, we find that the drift velocity is proportional to the electric field
and that constant of proportionality is the mobility. So, just like how we define an
electron mobility one could you also use the same kind of a relation to define the drift
velocity of holes. So, since both electrons and holes are carriers of current the holes in
the valence band and the electrons in the conduction band are both carriers of current, we
are we can define a drift velocity for both these species and that can be given by this
relation for low electric fields.
Now, as the electric field continues to increase the drift velocity rolls off and begins to
saturate ok. And the mobility and the velocity of these carriers saturates and to a point
which is called as the saturation velocity. So, at very high electric fields the velocity of
112
the carriers and materials like silicon saturate. Now other materials like say for example,
gallium arsenide behave in a slightly more interesting manner. So, here is a plot of the
drift velocity of carriers in gallium arsenide, drift velocity of electrons in gallium
arsenide versus the electric field.
So, at low electric fields we see that the drift velocity is proportional to the applied
electric field and the slope defines the mobility. But as the as the electric field begins to
increase, the drift velocity tends to saturate, but then it suddenly starts to drop low and it
drops down and begins to saturate at a new value. Now this interesting phenomena is
something called as the gun effect and it can be explained by looking at the full E k
diagram for gallium arsenide.
So, in this these plots show the complete energy versus k diagram for different
semiconductors for 3 different semiconductors; germanium, silicon and gallium arsenide.
And these plots have been taken from our reference textbook which is S.M. SZE book on
Semiconductor Physics. And in these plots, what you see are the e k diagrams
considering the fact considering the entire 3 d crystalline material and you find that
depending on the different planes and the different directions you have these range of
curves which constitute the energy k diagram. And our conduction band and the valence
band that we have looked at appear in this particular manner. You look at the lowest
energy of all the states that are mostly empty and that defines our conduction band edge
113
E c and you look at the highest states of all these energies that are mostly filled and that
defines our energy level Ev and this gap is defined as Eg. So, this is the case of
germanium. And here we have the case of silicon where you have the lowest point of the
conduction band conduction bands and the highest point of the valence bands and their
energy gap E g.
Now, the material of interest to us with regards the previous slide is gallium arsenide. So,
here you have the energy k diagram for gallium arsenide and we are particularly
interested in this region that I have shown in red. And therefore, here is the lowest point
which is your E c and that is the highest point of the valence bands that is our E v. Now
at thermal equilibrium, before we apply any voltage, we find that this is the smallest gap
that is your energy gap.
And we find that the electrons from the valence band are promoted to the conduction
band and they begin to occupy this valley here. And this valley has got a certain profile
you have an energy k profile which is say which is say defined as alpha k square and
depending on the value of alpha, we have these electrons are having a certain effective
mass. And based on that effective mass; so in this case, it would be h2 (d2E/dk2 ) which
would be of the order of h2/2𝛼.
Now based on this effective mass the electrons would have a certain mobility which is
given by q into t collision by m star. Now as the electric field increases the n the
momentum of this electrons increases and there they reach they pick up a values of
energy that allow these electrons to cross over to a different valley in the E k diagram.
So, these some of these electrons move over to this valley here.
Now, this valley is along the energy axis, it lies at a higher energy as compared to E g
and therefore, the electrons were normally not populating this valley. But upon the
application of very high electric fields, the electrons could now migrate to this valley
here. And this valley has its own parabola which could be defined as E is let us say
proportional to 𝛽(𝑘 − 𝑘0 ) the whole square where k naught is given by this little
translation in k space. And therefore, our d2E/dk2 ends up as 2 𝛽 and our mobility and is
defined as q into t collision by m* where m* is h22 𝛽 beta.
So, we therefore, had electrons that had an effective mass defined by alpha and now after
movement into another valley their effective mass is defined by beta. Now which of
114
these 2 is greater? This parabola is much sharper as compared to this parabola here. And
therefore, alpha can be said to be much greater than your beta. So, if alpha is much
greater than beta, then the effective mass of the electron sitting in this valley is lower as
compared to the effective mass of the electron sitting in the valley here. Or in other
words the mobility of the electron sitting in this valley is higher as compared to the
mobility of electron sitting in this valley.
So, that is why we see a E k diagram that looks like this. So, the electrons initially had a
high mobility and they were occupying the sharper valley, but then as the electric field
increased electrons migrated to the upper and more shallow E k diagram or the upper
valley that had a more shallow E k diagram. And therefore, it their effective mass
increased and their mobility is came down and therefore, we see a drop in the drift
velocity. So, that is the explanation for the gun effect, in the case of gallium arsenide.
Now, before we wrap up and close this topic, there are a couple of points that I would
like to make with regards to this E k diagram. And these are points that we will come
back to later as well, but since we are already here we might as well just discussed this
aspect. Now in the case of silicon and germanium; so, we note that the valence band that
is the E k diagram corresponding to the valence band and the E k diagram corresponding
to the conduction band have their minima or their maxima or their point of inflection let
me say at different locations in k.
So, we see that this minima and this maxima are located at different points in k. And the
same is true with silicon you see that the minima of the conduction band and the maxima
of the valence band are located at different points in k space. On the other hand, in the
case of germanium or in the case of gallium arsenide, we find that the minima of the
conduction band and the maxima of the valence band are located at almost same point in
k; there is no need for any translation in k space.
So, these materials where there is the minima of the conduction band and the maxima of
the valence band are located at the k same value of k, these materials are materials said
to have a direct band gap and these are called as direct band gap semiconductors. And on
the other hand, materials such as silicon where the maxima of the valence band and the
minima of the conduction band occur at different values in k which would require an
electron making a leap from a valence band to the conduction band having to change its
115
k. These kind of materials are called as indirect band gap semiconductors. And they play
a significant role when it comes to and this different plays a significant role when it
comes to the generation of electron hole pairs and the recombination of electron hole
pairs which is a topic that we will discuss further down the road.
116
Lecture - 08
Solids: Crystals
So, moving on. So, when you start bringing in many atoms close to each other then start
creating solids, you could end up with different kinds of arrangements. And depending
on the if you classify them in terms of their order of arrangements, you end up with
things called as mono crystalline or crystalline solids, poly crystalline solids or
amorphous solids ok. And what are these?
Crystalline solids are solids where the atoms are arranged periodically, throughout which
means that for infinitely long distances there is complete periodicity in the arrangement,
of the complete irregularity in the arrangement of atoms, they perfectly well arranged ok.
So, if you look at the distance between these bonds it is all the same throughout ok.
So, for example, this bond length will be the same between one plane to the other that
bond length will be the same from one plane to the other etcetera. And in fact, they form
such a nice grid of points that we could define a vector system or a coordinate system on
this grid. So, if you if you were to define basis vectors, let us say a one a 2 and a 3 for a 3
dimensional space then any point on the any crystal or lattice point ok. So, this
117
crystalline arrangement is something called as a lattice. And any atom occupying a
lattice any lattice point which is basically an atom in the lattice, can be defined using
these 3 orthogonal vectors ok. So, it is a linear combination of these vectors, where you
have alpha 1, alpha 2, alpha 3 of constant coefficients. And this vector r is basically a
linear combination of these orthogonal vectors. So, it is a very geometrically pleasing
structure.
The poly crystalline structure is a little off, in the sense it is not that perfectly crystalline.
The polycrystalline structure in fact, has got little tiles that are perfectly crystalline
within the tile, and the size of each tile is the order of microns. So, let us say 10 microns.
And you have different tiles, which have got which are within themselves crystalline, but
the type of crystallinity changes. So, for example, the orientation of the crystalline cleans
will be different ok. So, for example, within this little tile, this structure is perfectly
crystalline, but the crystallinity holds true only for this distance of about a few microns.
And then you have another tile that starts and you have a different nature of crystallinity
within the other tile. And these tiles are all separated these tiles are all called grains and
these grains are all separated by grain boundaries ok. So, this is something called as a
poly crystalline material and if you go to the other end of the spectrum. So, if we had
crystalline and then we have poly crystalline which is somewhere in the middle the other
end is a completely disordered structure, which is there is almost no periodicity at any
scale level. And even if there is periodicity it is only of the order of a few atoms long and
this is something called as an amorphous material. So, classic amorphous material is
glass. So, these materials also called glasses.
So, in amorphous materials the atoms have no periodicity what is so ever. So, the bond
lens between the different atoms will be different, their bond angles will be different and
so on ok. So, it is a very extremely defective structure. And the and it is got immense
consequences with regards to the distribution of energy levels, and you know the nature
of band gaps etcetera and in our course we will look at crystalline materials ok.
118
it is helpful for us to study the geometry of these crystals. So, for example, let us study
the distance between 2 crystal planes, which let us say is called d. Now a very powerful
experimental technique to do that is something called as electron diffraction from a
crystal or which is also called as Bragg’s diffraction.
So, what is done is, the crystal is placed inside a system and the system puts out a very
high energy electron beam, and makes it incident on the crystal. Let us say you have 2
rays of this beam ok. So, let me just turn draw this a little better. So, you have an high
energy electron beam that is made incident on this crystal. And 2 rays of this incident
and these of these beam tend to diffract because they are passing through this difference
in the 2 lattice planes through this gap here, which is got a width of d. So, these electron
beams are going to diffract and by placing a screen we would obtain an intensity pattern
due to this diffraction. And by looking at this intensity pattern, one can identify whether
this material is truly crystalline what the crystal spacing is etcetera.
So, firstly, let us just break down this entire experiment. So firstly, why do we need a
high energy electron beam why cannot I just use visible light to study the distance
between crystal plates. So, the crystal the distance between crystal planes the order is of
the order of few angstroms ok. So, let us just say one e minus 10 meters. So, that is the
order of the distance between the planes, but visible light has got a wave length which is
about 4000 angstroms to about 7000 angstroms. So, it is about thousands of times larger
119
than this distance. So, it is not possible for this large wavelength electromagnetic wave to
resolve such a fine distance. And therefore, we need to use an electromagnetic wave
which has got a wavelength of the order of one e minus 10 meters. And how is that done
it is done by using an electron gun, for example, and firing the electrons off at very high
momentum ok.
Because if you add a lot of energy to the electrons you increase the energy implies you
increase the momentum and due to Bragg’s law if an ink the momentum is increased the
wavelength of the electrons will go down. And therefore, we can get very small
wavelengths by having very large momentum in the electron beam. So, you use high
energy electrons to study the crystal plane. So, this high energy electron is made incident
on this plane. So, let us say the angle of incidence here is theta, it is not the angle of
incidence, but it is the angle at which it abeyances off the plane it is theta. And you have
your another ray which comes in like this ok. Now these 2 rays are going to interfere and
create this intensity pattern.
Now, what is the condition for constructive interference? So, if you have let us say one
wave that looks like this, and another that looks like this. So, let us just mark 2 points on
this let us say A and B. So, since the peaks which are defined by A and B are at the same
location, we could say that these in waves would interfere constructively. But now let us
say we translate the second wave be translated by half a wavelength down. So, B now
comes to this point. So, this is down by half a wavelength.
Now clearly the peak of this wave is at the same location as the valley of the other and
therefore, you will have destructive interference. So now, if you translate this wave a
little bit more by another half lambda, the peak of this wave has now shifted by another
lambda by 2. And therefore, the distance the total translation has been one lambda that is
one wavelength, and you will again have constructive interference.
So, you have constructive interference only when the path difference between the 2
waves or the if you want to think in terms of phase difference is also ok, but the path
difference between the 2 waves should be an integer times lambda. It is only then that
you will have constructive interference. So, what is the path difference between these 2
rays? So, let us draw a facet here you know it is sort of perpendicular to this and then
other facet, or another plane here which is perpendicular. So, till this point both these
120
rays travel the same distance exact same distance. And beyond this they are going to
travel the exact same distance to the screen, but it is only this way, this ray has travelled
a little more than the other in this region. So, this is the path difference this is the path
difference between the 2 rays, it is only this little distance here and what is that distance.
So, we just use trigonometry to identify that. So, we have this little structure and this is
the path difference. And the hypotenuse of these 2 triangles is d the crystal plane
difference and you can easily find out that this is theta and therefore, this point this
distance here is d sin theta, and that distance there is also d sin theta therefore, the total
path difference is 2D sin theta. So, the condition for constructive interference is that 2D
sin theta d equal to n times lambda ok.
So, that is that is given here. So, this is the condition for constructive interference and
this is something called as the Bragg’s condition. So, by keeping this crystal inside the
system and by impinging an electron high energy electron beam on it, and by changing
the value of theta one can get an intensity versus theta plot. And from that plot it is
possible to identify whether the material is a crystalline material and you know if it is
then what is the distance between the planes and etcetera. So, it is a very powerful tool to
study crystals.
So now coming to the defining the crystals themselves. So, we will just go through a few
definitions it is not really fundamentally important from the point of view this course, but
121
nevertheless it is useful to know these things. So, let us say you have a 2 dimensional
crystal lattice as shown here ok. So, let me make something that is that is simple in case
it is not visible. So, let us say you have a 2 dimensional crystal lattice. Now what are the
different ways in which I can define this crystal space ok. So, the first thing is to identify
the most fundamental structure that can define this crystal space. And there are 2 ways to
go about it, the first is to define something called as a unit cell ok.
So, what is a unit cell a unit cell is essentially any building block that is composed of
making a polygon let us say out of the out of these lattice points. Such that when this
building block of in this geometrical entity is repeated again and again it can be used to
fill up the entire lattice plane without having any gaps etcetera ok. So, this entity can be
used to fill up entire lattice plane. So, for example, this could be a unit cell this could
also be a unit cell. So, both these are possibilities for a unit cell. So, a unit cell is not
unique you could have many units different kinds of unit cells. So, that is what a unit cell
is. And if you want to locate the number of atoms in the unit cell, you know how many
atoms are present in a unit cell what we have to do is we count the number of atoms and
also watch out for the number of unit cells that are sharing those atoms ok. So, let us do
an example and then it will become quite clear.
So, let us take a unit cell of this kind ok. So, it is you have you have a unit cell which is
got a structure like this. So, this can be periodically used this can be used again and again
to completely cover the entire lattice space. And your lattice space is composed of all
these atoms which are drawn as these little solid circles. So, let us take this unit cell let
me just fill up the space all around. So, that it becomes easier. So, let us take this unit
cell. So, if you need to find out how many atoms are present in that unit cell, we see that
there are 4 atoms of the vertices and there is one atom right here in the middle. So,
therefore, it appears that there are 5 atoms, but then we need to be careful because this
atom is already shared by 4-unit cell. So, this atom is equivalent to one 4th of an atom
because it is shared by 4 unit cells. And how many of these kinds we have we have for
such corner atoms we have this one we have this we have this in here.
So, we have 4 corner atoms vertex atoms and each of them are shared by 4 unit cells.
And therefore, they are effectively one, but there is one in the middle which is shared by
just one cell right, which is which only belongs to this particular unit cell and it is not
shared by any cell. So, it is only belongs to that cell and therefore, there are total of 2
122
atoms in this unit cell ok. So, that is the way you calculate the total count and I think if
you practice this you will obtain it or you will you will get better at this, but it is really
not really important for this course, but nevertheless it is good to know.
Now, the other definition is something called as a primitive cell and what is the primitive
cell a primitive cell is simply the smallest possible unit cell. Or if you want to quantify it
the number of atoms in a primitive cell has to be one ok. So, for example, if you go back
to our lattice we defined our unit cell as that and the number of atoms were 2, because
there were 4 here and there is one in the middle. So, this cannot be a primitive cell a
primitive cell has to be something smaller it has to be the smallest possible unit cell. And
that is possible by having a unit cell that looks like this. For example,. So, if you have a
unit cell that looks like this how many atoms does this unit cell have it has got 4 in the
corner, but these 4 are shared by 4 other cells and therefore, there is a total of one atom
per unit cell. So, this is a good definition of a primitive cell.
So, once again if you were to take a primitive cell, and tile up all the primitive cells next
to each other you should be able to fill up the entire lattice space.
Now, heading towards 3D; so, if you go from 2D to 3D there are different kinds of lattice
structures ok, and they can be classified in to very distinct types and there are basically 5
distinct classes for 2D lattices and there are 4, 14 distinct classes for 3 dimensional
lattices and these are called as the Bravais lattices. So, some simple structures ok, which
123
is useful to know for 3D are something called as a simple cubic the body centered cubic
and the face centered cubic.
So, what is a simple cubic? A simple cubic is simply a cube with the atoms all sitting on
the vertices of the cube. So now, if you have to go through your example and say identify
the number of atoms. So, this is a one this is one kind of a 3 dimensional unit cell. So,
what is the number of atoms in this 3D unit cell? It is you have 8 atoms on the vertices,
but each of these atoms are shared by 8 other unit cells. And therefore, there is one
effective cell per simple cubic unit cell. What about the body centered cubic? A body
centered cubic has got 8 vertices just like the simple cubic structure, but it is also got one
atom right inside the body. And therefore, the total number of atoms are going to be 8
into 1 by 8 plus that one which is not shared by any other cell.
So, which is 2 and the third is a face centered cubic structure, which is basically got your
8 corner atoms just like a simple cubic, but then on the centre of every face you also have
one atom present. So, this is a side face and you have one that is a face on the table that
is another and so on. So, what are the total number of atoms. So, if you think of this face
atom it is going to be shared by 2 unit cells. So, you will have both these unit cells
sharing that atom. So, you have 8 corner atoms shared by 8 unit cells, plus 6 face atoms
shared by 2 unit cells. And therefore, you have a total of 4 atoms per face centered cubic
unit cell.
124
Now, particularly with regard to crystal silicon which is going to be our model of study,
what is the silicon structure like you know just to just to give you an idea. So, here is a
typo. So, you let us ignore that structure for now although the general geometry is
correct, it looks like there are different kinds of atoms there. So, silicon structure only
has silicon atoms and a simple way to draw it is it is basically 2 phase centered cubic
structures which are sort of inter woven ok. And we will I will explain that, but a simple
way to draw it is to simply draw a tetrahedral arrangement of atoms. And then continue
drawing tetrahedral arrangements of atoms. So, you will end up with a diamond structure
and then this will have it is own this will have it is own tetrahedron and so on.
So, that is the way you could you could possibly draw a silicon lattice, but then if you
think of this picture here ok. So, what this picture tells you is you are looking at the
silicon unit cell from the top ok, and these numbers that are written here I pointing out
that there is an atom in that x and y. So, let us let us keep this as the y coordinate and the
x coordinate. So, you are seeing x and y coordinate at the cell and this number here is the
z coordinate of the system it is telling you the height of the atom from the floor. So, if
you had to draw the say z, x and y what this is telling you is first let us look at these look
at these 5 atoms ok. So, it is telling you that you have 5 atoms sitting right on the floor.
And they are like a face centered cubic arrangement. And then let us look at these atoms
which are on the sides.
So now we can construct the side walls that have got a certain height. And there are
atoms sitting at the midpoint on these walls which are there on the side faces and this
roof of the cell is basically the floor of the next cell. So, this structure repeats. So, this is
at 0 1 the roof would be at one and therefore, it is a face centered cubic you see the first
face centered cubic, but there is an also some of the atoms that we have not yet counted.
So now, let us look at these atoms here the 3 by 4 the 1 by 4. If you were to sort of
continue these atoms you will find that you know this arrangement can be extended
throughout, and these are your other interwoven what you say face centered structures
ok. So, you have these 2 interwoven arrangements that form your silicon lattice it is it is
good to just spend some time and imagine it what we talk about.
125
Now, is something called as a packing or the effectiveness of packing in a crystal. And in

particular we define a factor called as a packing fraction in a crystal which tells you
which gives you a ratio of the volume occupied by the atoms to the volume of the unit
cell.
Now, the unit cell depending on the kind of a unit cell taken, and depending on the size
of the atoms and their and the kind of and the kind of packing of these atoms, it is
possible that there are some spaces in the unit cell or the volume of the unit cell might be
larger than the effective volume of the atoms. Or you know the unit cell might be very
efficiently packed. So, how does one determine you know this kind of the packing
efficiency in unit cell? It is given by this packing fraction. So, let us just take an example
and by just solving example you know the message of this packing fraction would be
very, very clear.
So, let us take a face centered cubic. So, what is the face centered cubic you have a cubic
unit cell you have 8 atoms on the vertices and on the centre of each phase you have a
separate atom by itself. So, how many atoms do we have in this unit cell per unit cell,
you have 8 corner atoms and each of these 8 atoms are shared by 8 other unit cells. And
therefore, it is 8 into 1 by 8 plus 6 face atoms and each phase atom is shared by 2 unit
cells and therefore, the 6 into 1 by 2 which is equal to 4 atoms. So, there are a total of 4
atoms per unit cell in the face centered cubic crystal, but now instead of just drawing the
126
atoms as little dots let us actually define a radius for each atoms. Let us say that the each
atom has got a radius which is which is got a radius of r. So, the volume of each atom is
4/3𝜋𝑟 3 . And there are a total of 4 atoms an effect ml for effective number of atoms per
unit cell. And therefore, the total volume of the atoms occupying this unit cell is
16/3𝜋𝑟 3 .
Now, so, let us draw this picture here. So, here is the top face of the cube. So, let us say
the top face of this face centered cubic structure. So, you have one atom there and that is
the next face centered cubic structure, you have another atom here you have yet another
atom there. You have yet another atom here and that is your crystal structure. So, you are
seeing the top of everything. So, if you look at only this unit cell you find that it is
packed in this particular manner you have one face centered atom which has got which
has got a diameter 2r. And you have all the other corner face centered cubic which are
being shared which have all got an effective radius of r.
Now, therefore, given this definition of r, can given that all these atoms are not touching
each other, and they have they have all packed into this unit cell in this particular
manner. Can we define the length of the of this square what is this length a ok? So, what
is a? A can be easily defined in terms of r, and you find that your A is nothing but 2
square root of 2.
127
+2 of r. And therefore, the volume of the unit cell which is basically a cube of each side
having length A is a cube. And a cube can be defined in terms of r because of this
relation. So, you find that the volume of the unit cell is 16 square root 2 r cube and
therefore, the packing density is the total volume of the atoms occupying this unit cell
divided by the total volume of this cube and it turns out to be about 0.74 for a face
centered cubic crystal. So, that is something called the packing density ok.
So, it is again something that is not going to be useful in this course, but having said that
it is quite important because it determines. So, let us say this is the face across which you
are going to have all your electron transport ok. So, let us say this is the face of the
semiconductor on which you are having electron transport. So, this packing fraction
determines the nature of the interaction of the electron with the crystal. So, the surface of
the crystal and the arrangement of atoms on the surface of the crystal where the electron
is going to move across does definitely impact the properties of the charge transport
properties of charge transport and therefore, determines the current etcetera.
.00
Now, we now come to the last bit of the course. And before we just drop off ok, I just
want to define 3 ideas these ideas are again something that we will not be using in the
rest of this course. So, it is all right for the students to completely skip these 3, but I feel
that these are 3 critical ideas that provide a deeper insight. And the first idea is basically
a geometrical construction. And it is something called as a Wigner Seitz cell. So, let us
128
say you have and how do you how do you perform this geometrical constructions let us
say you have a lattice ok. So, let us say these are the atoms of your lattice.
Now let us take any one atom in the lattice. So, let us take this particular atom here we
will first draw lines, as shown by these dotted lines we will draw lines that connect this
particular atom to the nearest neighbours. So, these are the nearest neighbour atoms and
we are simply drawing lines that connect this atom to the nearest neighbours. So, that is
the first step in this geometrical construction. Next what we do is we will draw the
perpendicular bisectors to these lines. So, let us take this particular dotted line. So, what
is the perpendicular bisector it is the line it is another line that divides this line 2 equal
parts and intersects it at 90 degrees. So, we have that to be the perpendicular bisector to
this particular line.
So, we now construct perpendicular bisectors to each of these dotted lines, that connect
the atom that we have chosen to the nearest neighbouring atoms in the lattice. So, we
have all these perpendicular bisectors that are shown here. And these perpendicular
bisectors would all intersect they would all meet each other you see these bisectors. So,
you see this bisector is meeting this one this bisector is meeting this one etcetera and all
these perpendicular bisectors will therefore, enclose a certain space, you see there is
there is if you were to draw this enclosure. So, which I am going to draw this very thick
line here they would enclose this certain space. And they would basically create a cell
ok. So, this enclosure by all these perpendicular bisectors creates this kind of a cell and
this cell is something called as the Wigner Seitz cell.
So, you started off by taking picking an atom drawing lines to the nearest atoms
constructing perpendicular bisectors and taking the inner shell that that is enveloped by
all these perpendicular bisectors. So, this something called as a Wigner Seitz cell it is
simply a construction, a geometrical construction. So, that is the first idea. Ok now this
connects to another idea.
129
So, the second idea is something that is got to do with the reciprocal lattice space a very
simply put you know do not worry about all these this mathematics. You can I will tell
you what that is, but do not worry about it reciprocal lattice space is essentially a
frequency space. So, let us say you have a periodic crystal you have got you have got
atoms arranged in some period. So, it is got a periodic wavelength it is got some spatial
period to it.
Now, what is the frequency it is essentially 1 by the spatial period. So, if it is a 1 by a

could be considered to be the frequency. So, if you were to take the spatial Fourier
transform of the real space you will end up with the in the frequency space and that
frequency space is called as a reciprocal lattice space. So, as essentially a Fourier
transform that connects the real space to the frequency space ok. So, essentially the idea
behind what a reciprocal lattice spaces.
Now if you have a crystal if you have a perfect crystal in the real space and you take a
Fourier transform you will end up with another with another periodic arrangement of
points in the reciprocal lattice space, and that will also be a crystal. Although, it need not
be the same kind of a crystal and the way you translate between real space to reciprocal
lattice space is if you were to define basis vectors in the real space say a 1 a 2 and a 3.
Such that any lattice point in the real space is a linear combination of these 3 orthogonal
basis vectors. Then the reciprocal lattice space will also have basis vectors b 1, b 2, b 3
130
and they will all be connected to the real lattice space vectors in this particular fashion
ok.
So, essentially you will find that in the 3D case it is basically this cross product divided
by the volume of these cells was a very brief introduction to reciprocal lattice space.
If you were to take these 2 ideas, you have the Wigner Seitz cell and you have the
reciprocal lattice space and the reciprocal lattice space. So, Wigner Seitz cell was
construct was a geometrical construct constructed on real lattice space. So, you took the
lattice you had these lattice atoms. So, we are talking about real space which means x is
measured in meters ok.
And we constructed a Wigner Seiz cell by drawing all these perpendicular bisectors in
real lattice space and then we defined something called as a reciprocal lattice space,
which is also a lattice, but it is all in 1 by meter ok. So, it is all the frequency domain it is
the spatial frequency domain it is the dimensions are all 1 by meter and you have
constructed another lattice here which is the reciprocal lattice.
Now, if you were to apply the same geometrical construct in a reciprocal lattice that is
you construct a Wigner Seitz cell in a reciprocal lattice that enclosed boundary is
something called as a Brillouin zone. Now the reason why I am introducing this topic is
that these Brillouin zones you see these are all perpendicular bisectors ok, to these 2 pi
131
by a 2n𝜋/𝑎 lines. And therefore, this Brillouin zone facets occur at points where we are
now in k space right we are in k space, we are all at you know 1 by meter that is the
reciprocal or the frequency spatial frequency domain we are all in k space. So, the
Brillouin zone boundaries occur at points where k is 𝜋/𝑎 2𝜋/𝑎 and so on where integer
times pi by a. So, that is where the Brillouin zones zone boundaries occur I am only
talking about linear spatial thing linear arrangement of atoms.
So, if the Brillouin zone boundaries occur here, have you seen these points anywhere
else? So, we have seen these points somewhere else and those were the points where the
energy gaps occurred. So, when we drew the e k diagram after solving kronig’s penney
model we saw that the energy gaps or the discontinuities in energies occurred at pi by a 2
pi by a etcetera. So, therefore, the discontinuities are all occurring at the Brillouin zone
boundaries. And it so, happens that the entire idea of this energy gap formation can also
be explained by considering the diffraction of electrons at these facets ok. So, electrons
reflect off these zone facets and therefore leads to the formation of energy gap.
132
Semiconductor Fundamentals
Lecture – 09
Density of States (DoS)
We now proceed to the next important topic with regards to Semiconductor

Fundamentals and this is got to do with something called as the Density of States ok.
And the idea behind this are if you remember I see our main goal in this subject is to
calculate the current or the current density let us say and in essence. This current density
is dependent upon the traffic of electrons and holes right. It depends upon the rate of
change of charge ok. To first order this is what we would like to calculate when we apply
you know when we subject the semiconductor device to different bias conditions.
Now this traffic of electrons electron and whole traffic it depends not only on the speed,
but also on the number ok. Now the speed was very nicely characterized by something
called as the mobility of semiconductor of the carriers ok, but you have still not discuss
anything to do with the count on the number of carriers ok. In particular we would like to
calculate the number of electrons or holes per unit volume of the semiconductor and this
is something called as the carrier concentration ok.
133
So, when we say carrier concentration we imply the number of free electrons and holes
per unit volume. So, we are not interested for example, if this is the conduction band
edge and that is the valence band edge ok. We are only interested in the electrons above
the conduct in the in the conduction band and we are interested in the holes in the
valence band we are not interested in the holes in the conduction band there are plenty of
holes in the conduction band plenty of vacant states, but that is not going to contribute
the current.
And similarly there are plenty of electrons you know that are bound to the atoms sitting
inside the valence band and they are not going to contribute to the current. So, we are not
interested in the electrons we are interested in the holes in the valence band and the
electrons in the conduction band. So, we want to get a gauge on the number of electrons
and holes per unit volume again how do we do this. So, this is what our next goal is in
how do we when we know how to calculate speed to some extent ok.
Because we know this parameter called mobility and we know that the drift velocity is
the mobility times the electric field. So, we know how to calculate the velocity, but we
do not know the number we left that we you know we simply use the symbols n and p
and we left those undefined now the way you go about calculating number is to. Firstly,
identify the number of seats or the number of solutions to Schrodinger equation that
electrons and holes can occupy ok. So, if you imagine a classroom which is got which is
got let us say there is a classroom and let us say there are lots of chairs in the classroom.
And the students can come and occupy consider in the classroom, but you make a rule
that a student has to sit on a chair, he cannot stand he or she cannot stand in the
classroom. So, if the student has to sit on a chair and if the students can be imagined to
be electrons and you ask the question you know how many students can you take a guess
as to how many students are present in a classroom? Ok. So, the you need two parts to
answer that question the first is how many chairs are present in that classroom how many
chairs and the second is what is the probability that the chair is occupied?
So, if you know what these two numbers are then you have a good estimate as to the
number of students in the classroom and the density of states is going to give you an
answers to the number of chairs or. In fact, more accurately it is the number of states per
unit energy per unit volume in a 3D material.
134
So, if you are looking at 3D semiconductors ok.
So, 3D semiconductors imply your crystalline lattice is quite large it is much larger than
the size of the atom in all three dimensions ok. So, let us say for example, if you are
interested in a number let us say it is greater than much greater than 20 nanometers I
mean the atom is much smaller than 20 nanometers, but nevertheless as a practical
estimate you can consider all these materials to be three dimensional.
So, you have a lattice that is significantly large in all three dimensions, if you have a
Cartesian coordinate system and you label this as x that is y and this is z then you have a
lattice that is significantly large in x and y and z directions. So, the density of states is
going to give you an answers to the number of chairs. So, that you know the number of
students ok. So, it is an important question to answer if you want to calculate the
concentration.
So, we want to know how many states and what are states? States are basically the
solutions to Schrodinger’s equation ok. How many solutions are there to Schrodinger’s
equation per unit energy per unit volume of the material and if you have this what you
will end up with is if you know the energy you take one unit volume of the
semiconductor and you plot the energy versus number of states. You know the
distribution of states and this energy. So, let us say you plot the distribution of these
states.
135
So, let us say you have the density of states which we will denote by you know N of E
and N as a function of energy or you denote it as G as function of energy or you denote it
as D as a function of energy sometimes depending on you know what the mood is feely
just use a different variable, but you will get the context of it. So, it is it is the density of
states and you have all these states present that are distributed in energy and we want to
know as to what this distribution is it that is it going to be uniform or is it that it is going
to vary you know as various the square root of E or what is it what is the distribution of
the states as a function of energy in a per unit volume of a 3D material ok.
So, let us so, we will do this exercise first for a 3D material and we will see how to go
about making this calculation and then we will do the same calculation for a 2D material
and for a 1D material and finally, answer questions for a 0D material now what do you
mean by a 2D material is like a it is quite large in 2 dimensions, but quite thin in the 3rd
dimension. So, you can imagine a 2D semiconductor which says you might have heard
of for example, say molybdenum disulphide or grapheme etcetera and a 1D material is
probably something that is you know it is just thin and long it is like a little wire. So, any
nano wire it could be you know could be approaching a 1D material and a 0D material is
something which is in a conceptually a point it is it is got no dimension on x y and z.
So, you will be a quantum dot would probably be approaching a 0D material. So, we will
we will look at the density of states for all these examples, but for the time being we are
interested in the density of states for a three dimensional material and we want to know
we want to perform this calculation of how many states per unit energy per unit volume.
So, that you have the distribution and this tells me the number of chairs. So, the number
of places the electrons can occupy and after getting this if I somehow can calculate the
probability that an electron occupies a state at a particular energy I can use these two
answers to get the count of the number of electrons which I will then use to calculate the
current so that is the whole idea ok. So, what we will be doing in this exercise is to find
out how the density of states how does N vary with E ok.
What kind how does it vary with is it proportional to E or what is it how does it vary
with E with E?
136
So, in order to answer this we are going to just ask ourselves five questions and we will
do the same thing for all the cases. Now this is for a 3D material now the first question is
how many solutions of Schrodinger’s equation that is how many states are there per unit
volume in k space ok. So, what does k space k space k is the wave vector in sum and k
space is essentially your reciprocal lattice space.
So, we are not going to be looking at the actual material it is in it is spatial coordinates,
instead we are looking at the reciprocal lattice space that is the frequency the spatial
frequency space if you like and we ask ourselves a question as to how many solutions are
Schrodinger’s equations are there per unit volume of k space and why is this easy to
calculate, because we have already solved for the wave function in these in k space.
137
So, if you remember your particle in a box we always will keep going back to that little
example. So, we solved the wave function for a particle in a one dimensional box and we
found that the wave function was this right. So, we got this by normalizing everything
and we got this little term which is nothing, but your k; k of x and we got this to be a
quantized value and we had to have a integer n and I have used the subscript x to
represent the x direction ok.
So, you have kx is equal to nx𝜋/𝐿𝑥. So, k could only take integer values of pi by 𝜋/Lx
and why should that happen? Because it had to meet the boundary condition of the wave
function not existing outside the box ok so, that was a boundary condition, but
nevertheless k will be quantized no matter what you see you know how you define your
potential.
So, k is a quantized parameter so, n is got all n can only be integers and therefore, if I
were to draw if I were to draw the k x vector or if I look at the k x space then I see it can
only take discrete points right it can have values of 𝜋/Lx , 2 𝜋/Lx, 3 𝜋/Lx and so on and
so forth and you could also have say – 𝜋, −2𝜋 and so on.
So, only those values were allowed and if you solve Schrodinger’s equation for a 2D box
we could we use the technique of separating the two wave functions a separation of
variables which was we split Ψ as Ψ𝑥 and Ψ𝑦 and we found that it is simply a product of
these two wave functions. So, here you have k x and k y and both of them are quantized
and I have defined the quantum versus n x and n y and the lengths of the box are L x and
138
L y and similarly for a three dimensional box you will have the prod you will have the
wave function be a product of Ψ𝑥 Ψ𝑦 Ψ𝑧 .
And therefore, you will have three k, k x k y and k z defining your k space. So, here I
just drawn it out here. So, you have the solutions the Schrodinger’s equation being
discrete points. So, this would be 𝜋/Lx ,2 𝜋/Lx and so on this would be pi by L z 2 pi by
L z and so on this would be 𝜋/Ly , 2 𝜋/Ly and so on. So, you have this three dimensional
grid of points. So, I have not completed this grid, but you can imagine this is a three
dimensional grid of points and each point is a solution to Schrodinger’s equation.
So, it is the electrons can only take these values the electrons can only exist in these
states. So, if you were to imagine this three dimensional grid and now each of these are
solutions. So, what is the answer to this question you know how many solutions are
Schrodinger’s equation are there per unit volume of k space. So, let us you know imagine
this 3D grid that sort of sits like this. So, you have these little cubes cuboids if you would
like now let me just draw it better.
So, you will have these little cuboids and the vertices of these cuboids are essentially the
solutions to Schrodinger’s equation. So, you will have four of them sitting right in front
and you could imagine four at the back you will you will have four at the back. So, you
have this little region and if you imagine the point right in the middle of this it is being
shared by the four cubes in the front and the four cuboids at the back.
So, one solution each solution is shared by 8 unit cells of k space and each unit cell in k
space has got 8 solutions. So, there are 8 solutions each shared by 8 of them. So, that are
effectively 8 into 1 by 8 which is equal to 1 solution of Schrodinger’s equation per unit
volume in case p per unit cell in k space we have not yet calculated the volume. So, this
many solution so, we have 1 solution per unit cell now we need to get the answer as to
how many solutions per unit volume. So, we need to find out what is the volume of what
the volume of each unit cell is.
And the volume is simply this product right it is pi by L x into pi by L y into pi by L z.

So, that is the volume. So, let me just write it a little better my apologies so, the volume
of a unit cell n k space is 𝜋/Lx* 𝜋/Ly* 𝜋/Lz which is 𝜋 3 /Lx Ly Lz. So, there is 1
solution per unit cell and that unit cell has got this volume ok.
139
Therefore how many solutions are Schrodinger’s equation are there per unit volume?
The answer is simply as far as our calculations go so, far the answer is simply 1 solution
divided by the volume of L x L y L z say 𝜋 3 /Lx Ly Lz which is Lx Ly Lz /𝜋 3 .
But this answer is your more or less correct here, but we are still not accounted for
several things we need to a small correction we need to have a correction factor and the
correction factor has got to do with two to read it is got two reasons for it the first is our
integers you know the quantum numbers n x n y and n z could have plus or minus values
ok. So, you could have plus or minus one plus or minus two and both these are you know
the equivalent the same energy.
So, we had double counting for the number of solutions. So, in order to get the correct
number of solutions I probably need to take this answer and divided by 1 by 2 into 2 into
2 you know I need to half it for each of the dimensions, but then there is also another
point the electrons can have plus or I mean the up or down spin it can have two spins. So,
2 electrons can occupy a state ok.
So, we need to account for that by multiplying a factor of 2 therefore, the corrected
answer is essentially this ok it is got it has to have a 4 a term 4 there. So, the correct
answer to the number of solutions are Schrodinger’s equation per unit volume of k space
for a 3D material is Lx Ly Lz /4𝜋 3 . ok. So, this is the answer to the first question.
140
So, now let us proceed ok.
Now, we want to know how many energy states or how many solutions exist between k
and k+dk ok. So, now, imagine you have taken your k space you have your k z, k x and k
y and you have filled it you fill the grid up you have uses zoom out of this. So, we were
looking at the grid in a very microscopic sense and we calculated the number of solutions
per unit volume, but now I am going to zoom out and I am going to see this massive grid
of points right it is just going to be a massive grid of points.
And if you ask yourselves; what is the locus of points having the same k now this k is
essentially you know it is essentially the combination of k x, k y and k z ok. So, if you
ask yourself as to what is the locus of points that have the same value of k in a 3D space
ok. So, you have got this mass of grid of points and what is the locus of keeping the k
constant; the answer is quite obvious it has to be a spherical shell it or it has to be a
sphere. So, if you were to keep k the same and look at all the points that have got the
same value of k the answer is it is going to be a sphere in 3D space ok.
And what is the region between k and k+dk. So, this is k you could have another sphere
which has got a radius of k+dk right and we are interested in the number of energy states
that exists between k and k+dk ok. So, we are interested in the number of energy states
or number of solutions to Schrodinger’s equation that exist in this region here we are still
141
in reciprocal lattice plane we are still in k space and we are looking at the number of
solutions that exist in this region.
So, how do we calculate this? Well I need just need to know how many unit volumes are
there in this space. I know the number of solutions per unit volume which is equal to my
Lx Ly Lz /4𝜋 3 and I just need to know what is the volume ok so, if this is the number of
solutions in 1 volume let us say 1 unit volume how many unit volumes are there in this
region I am interested in ok.
So, what is the volume of the space the volume of that space is the surface of the sphere
into dk so which is 4𝜋𝑘 2 𝑑𝑘. So, that is the volume of the space which I am interested in.
And how many solutions are there in this volume? The answer is simply Lx Ly Lz
Lx Ly Lz
/𝜋 3 . 4𝜋𝑘 2 . . 4𝜋𝑘 2 𝑑𝑘 dk ok which is very neatly written out here.
4𝜋3
So, there are this many energy states in the region between k and k+dk so, which is
where k. So, we are looking at the reciprocal lattice space once more where k where you
have k x, k y and k z and you have all the points that have got the same k to be sphere
because this is this is a massive grid of points and all the points that I have got a radius of
k+dk is another sphere that is very concentric to this and this region has got a volume of
this much and there are this many solutions per unit volume and therefore, there are so
many solutions of Schrodinger’s equation in this region of interest just to repeat
everything ok.
142
So, what is the next question the next question is how many states exist between E and
E+dE ok. So, so far we have been making all our arguments in k space now we want to
convert this information in k space to energy because we are trying to estimate the
density of states which is the number of states per unit volume per unit energy. So, I
know the number of solution number of solutions to Schrodinger’s equations or the
number of states existing between k and k+dk. I know that now can I translate this to the
number of states between E and E+dE ok. So, that is the question ok.
So, if I know this can I calculate that? So, how is k connected to E, k is nothing but your
wave vector so, hk is my momentum and E is p2/2m therefore, E is h2k2 /2m and
therefore, k square is m E by h bar square and k is a square root of this whole thing
which is √2𝑚/ℎ 2 m in the square root by h bar and what is dk, b y dE; dk by dE is
obtained by differentiating this right.
So, you have 2 m to the power half by h bar and the differential of E to the power half
which is 1 by 2 √E power half and therefore, you have dk is equal to this term into d E.
So, we have these four relations with us now what we are going to do is we know that
the number of solutions in this region between k and k+dk is this. Now we are going to
substitute for k2 and d k with these expressions in terms of E and d E.
So, what is k2 square k2 square is 2mE/h2 so instead of so let us keep these terms separate
and instead of k square I am going to substitute 2mE/h2 and instead of dkI am going to
143
substitute this particular expression and after a bit of simplification you cancel off say
common terms you will end up with the number of states between E and E+dE is given
by this quantity here ok.
So, we now know the number of solutions to Schrodinger’s equation sitting between two
energy levels E and E+dE ok. So, now, let us see how to get to the density of states.
Now we ask our fourth question which is how many states exist between E and E plus d
E per unit volume now this is the real volume of the material? So, we are no longer in
reciprocal space. So, what is the volume?
Now, the box had lens Lx, Ly and L z therefore, the volume of the box is L x, L y, L z
and therefore, this numerator here is the number of solutions to Schrodinger’s equation
or the number of states between E and E+dE and this is the volume of the box and
therefore, the number of states between E and E+dE per unit volume is simply that
divided by the volume of the box and that gets rid of the L x, L y, L z terms and you
have these many states sitting per unit volume and between E and E plus d E.
Now, in order to get the density of states I need to know the number of states per unit
volume per unit energy. So, I also need to divided by the; I also need to get a per unit
energy term ok.
144
And how do I do that you simply divided by d E ok. So, you are looking at all the states
between here and between E and E plus d E per unit volume and you get your divided by
d E which is the energy gap with the energy region you are looking at this. Now I know
the that the length of the energy region you are looking at and you end up with the
number of states per unit volume per unit energy to be given as this which is nothing but
the definition of the density of states.
So, this is the density of states for a 3D material ok. So, most importantly you must be
looking at is how does the density of states vary with energy? So, this density of states
for 3D material scales as the square root of energy it is proportional to the square root of
energy. Now since we will be using 3D silicon as the model material in our course here it
is useful to draw this out. So, forgive this very rough sketch which is drawn here.
145
So, for a perfectly crystalline material you have your valence band edge and your
conduction band edge and in between this is something which we defined as the energy
gap. Now for a perfectly crystalline material the energy gap will have absolutely no
states it is only when defects start to appear that the energy gap will start getting filled up
with states. But as far as most of the course is concerned there are no states in the energy
gap. It is only when we talk about disordered material later towards the end of the course
we will we start worrying about states in the gap or even when we talk about doping for
example, but for now there are no states in the gap.
So, your density of states which is proportional to E basically implies this, that there are
states above the conduction band edge which is the bottom of the conduction band and
the valence band edge which is the top of the valence band there are states in these
regions and the way the states vary with energy is as E minus E c to the power half.
So, you have states varying like this in the conduction band and you have states varying
in proportion to Ev-E to the power half. You have states varying like this and the valence
band so that is what it implies. So, this is how the states in so, if you look at the number
of solutions to Schrodinger’s equation right at this point there are 0 states and just
immediately after that you have a little bit and then after that you have a little bit more.
And how does it all how does this envelope scale? It scales as E minus E c to the power
half into this particular term and all this m which you have seen throughout is all the
effective mass should all be in m star it is all the effective mass which we discussed
146
earlier. And similarly the states below the valence band edge so, these are the this is the
distribution if you ask what is N of E what is the density of state distribution and 3D
material perfectly crystalline 3D material.
The answer is it is this when E is greater than equal to E c that is from infinite to E c it is
this when E is less than sorry E<Ev ok, but greater than minus infinity and it is 0 when E
is between Ec and Ev ok which is the energy gap. So, that is the distribution of the states
in a 3D material; now what about a 2D material.
We ask ourselves the same questions, but the only thing is we need to ask ourselves how
many states per unit energy per unit area is in the material, because we are now looking
at a 2D material we just go through the same process again ok. Let's see what it gives us.
147
So, first question is how many solutions of Schrodinger’s equation at that per unit area in
k space. So, in now your wave function is just you know you are just looking at a 2D box
you are just looking at a two dimensional box. So, therefore, there are only k x and k y to
worry about in the k space ok. So, you have this grid of 2D points 2D grid and if you
were to draw if you let us say let us take let us draw these connect these lattice.
If you look at the unit cells you have one unit cell here, one unit cell here, one here and
one here. Now per unit cell how many solutions do you have? You have four solutions
which mark the vertices of this of this rectangle ok, but each solution is shared by 4 unit
cells this solution for example, is shared by this cell by this by this and by this. So, each
solution is shared by 4 unit cells. So, you have 1 solution for unit cell ok. So, that is the
number of solutions per unit cell in k space.
And what is the area of 1 unit cell? Area of 1 unit cell is simply your pi by L x. So, these
points are all 𝜋/Lx , 𝜋2/Lx 2 pi by L x and so on and that is pi by L y 2 pi by L y and so
on so, it is pi by L x into pi by L y. So, that is the area of one cell and therefore, the
number of solutions per unit area is going to be simply one divided by pi square by L x,
L y which is L x, L y by pi square.
148
But once again just like the last time we need to make a correction ok. Now we have to
be a double counting for the plus and minus. So, you have Lx Ly /𝜋 2 the plus and minus
with double counting there are two dimensions. So, we divide by 1 to be half it for each
dimension and there are two spins and therefore, you have a corrected or the correction
factor of 2𝜋 2 which defines the number of solutions for Schrodinger’s equation per unit
area and k space.
And now we ask ourselves next question you know how many energy states between k
and k+dk so, what is the locus of point. So, now, let us take our k space which is a two
dimensional space and fill it up ok. So, you have this massive grid of points if you can
149
imagine a graph paper that you might have used in school you have this massive grid of
points discrete points and these are all the solutions of Schrodinger’s equation and we
want to know how many states between k and k plus dk ok.
So, what is the locus of points on a 2D map in a 2D map in 2D space what is the locus of
points which keeps k constant. Ok the k will now be square root of k x2 square plus k y
square. So, what is the locus of points the answer is it is a circle you want to keep the
radius constant. So, if I keep the radius as k and I draw and I connect all the points that
have a constant radius k it is a circle and what about k plus dk, it is another circle which
has got a slightly larger radius and it is this, these are two concentric circles and we want
to know how many states are there between k and k plus dk.
So, we are interested in the number of states in this little ring here if you remember the
3D case we had spheres because these are three dimensional and we had the spherical
shell that was that was the region of interest, but now since it is 2D space you have this
little ring. So, what is the area of this ring the area of this ring is the perimeter of a circle
into dk. So, that is the area of the ring.
Now, I know that there is L x L y/2𝜋 2 solutions in 1 per unit area one unit area so how
many solutions exist in 2𝜋 1 pi k d k in an area of 2𝜋𝑘𝑑𝑘 ok.
150
So, the answer is simply 2 𝜋𝑘𝑑𝑘 into LxLy/2 pi square. So, this is the number of states
existing in this ring in this region here bet𝜋 2 ween k and k+dk for a 2D material. So, what
is the next step next step is to convert this all to energy ok.
So, the next question is how many states exist between E and E+dE right. So, we know
that between k and k plus dk we have these many states, that is 2 𝜋𝑘𝑑𝑘 Lx Ly by 2𝜋 2 pi
square states. Now how do I convert all this to the number of states between E and E plus
d E we go through the same process again. So, we know E is p2/2m which is h bar square
k square by 2 m and therefore, all these relations follow and now we substitute for k and
d k.
So, these are the number of solutions in terms of E. So, we use this expression and we
use this expression. So, that is my k and that is my d k ok. So, it is 2 𝜋𝑘𝑑𝑘 LxLy/2𝜋 2 the
number of solutions. So, it is this many so when you simplify all this you get this little
neat little answer here and this is the number of solutions or the number of states present
between E and E+dE ok.
151
And now the next question is; obviously, how many states between E and E plus dE per
unit area?
And now the box now we are looking at a 2D box and the area of the box is it is got
dimensions of L x and L y and therefore, the area is simply the product which is an L x L
y there are these many solutions to Schrodinger’s equation in the region between E and
E+dE. And the number of solutions of Schrodinger’s equation between E and E plus d
per unit area is simply that divided by Lx, Ly and that gives you this it is 𝑚/𝜋ℎ2 𝑑𝐸.
So, there are these many states present between E and E plus d E per unit area.
152
And how many states between how many states in the 2D material per unit area per unit
energy which is the density of states for a 2D material the answer is m by pi h bar square
d E divided by d which is m by pi h bar square. So, there are these many states per unit
energy per unit area in a 2D material you have follow the same process as in the case of
3D.
The only thing is we have adjusted terms for the 2D case, but what is fascinating here is
that the density of states does not depend on the energy you do not see an energy term in
the 3D case we saw that the density of states was proportional to E to the power half, but
here there is no it is proportional to E0 it is not dependent on energy at all ok so, that is
interesting. So, you will find that for a 2D material the density of states is constant with
energy.
153
Lecture – 10
Density of States - Continued, Fermi Function
Now, what about a 1D case?
We will quickly run through the same procedure. But, now we need to ask us a question
how many states per unit energy per unit length, in a 1D material, ok. The 1D material
does not have volume or area it is just got a length.
154
And we follow the same set of questions how many solutions are Schrodinger’s equation
per unit length in k space, ok. And now, we have only a 1 D box. We are only interested
in a 1 D box and we only have.
So, let us say this is the x direction. The box has got length Lx and we are only have got a
kx space. There is no ky, there is no kz and there is only a kx space.
155
And the wave function for a 1D box is simply this and that is my k x term which is
quantized with the quantum number n x. And therefore, in this space, I have you know
you have your 𝜋/Lx 2 𝜋/Lx and so on, ok.
And if you look at the number of solutions per unit length of k space first so, these are
my unit cells. Now, since that is 1 unit cell, that is the second unit cell, ok. If you look at
the number of solutions per unit cell, each cell has got 2 solutions. There is one solution
or this end and one solution on that end. There are 2 solutions, but each solution is shared
by 2 unit cells.
So, if you look at this solution, for example, it is shared by this cell as well as this cell,
right. So, each cell is just a line this line segment, ok connecting these 2 discrete points
and these discrete points of the solutions to Schrodinger’s equation.
So, each solution is shared by 2 cells and therefore, there is one solution per unit cell and
what is the length of the unit cell? The length of 1 unit cell is 𝜋/Lx, ok. So, the length of
this unit cell is this length here which is 𝜋/Lx. So, the number of solutions per unit length
is going to be one solution divided by the length of the unit cell which is Lx / 𝜋. But now,
once again we need to make our correction, ok.
156
So, let us deploy our correction terms. You know we have these correction terms in the
case of 3D and 2D. So, Lx/𝜋, but then we are double counting for n ok. For n x could
take plus and minus values.
So, we need to divide by half but then there are also 2 spins, ok snd therefore, this is the
correction term. So, the correction term is simply 1 and therefore, you your initial guess
on the answer was right. So, there are these many solutions to Schrodinger’s equation per
unit length in k space. And now, the next question how many solutions of Schrodinger’s
equation between k and k+dk.
So now, let us draw extend our k x, ok. You have this massive grid of points sitting on
this line it is a 1D case. So, we only have 1 dimension. You have so many points and
then I. So, what is the where is my k; my k is going to be this long and my k+dx is there
ok.
So, what I am interested in is the number of solutions in this region and this region has
got a length of dk. So, how many solutions exist in this region between k and k+dk? So,
that is my question, ok. So, k now is simply k x, there is no k x the k y or k z. So, k is
simply k x and that is k+dx a dk and this is K and the region between them has got a
length of dk and we are trying to find out how many states exist between k and k plus dk.
157
Now, the number of states per unit cell was already this right. It was L x by pi per unit
length. So, there are this many these many solutions per unit length and we want to know
how many solutions in this region of dk, for a length of dk how many solutions exist and
the answer is Lx / 𝜋dk.
So, there are decay into Lx / 𝜋 states or solutions of Schrodinger’s equation solutions in
the region between k and k+dk. So, that is the answer we are looking for. And now, the
next question is of course, convert everything to energy, ok.
158
So now, we know that there are dk Lx / 𝜋 solutions between k and k+dk. Now, how
many solutions exist between E and E+dE. So, we convert all the terms from k space to
E space to energy E and we use the same set of relations ok. The only thing is this time
we only have a decay term ok. That is, so, dk is related to E in this manner.
So, Lx / 𝜋dk is simply Lx / 𝜋 into this particular term here which is your dk. And
therefore, there are these many solutions, ok. So, there are Lx / 𝜋 into we can cancel
some terms of. So, you have m0.5 E-0.5 dE/2𝔥. So, you have these many solutions present
between E and E plus d.
Now, of course, the next question is, we want to get to per unit length therefore, how
many solutions per unit length ok.
So, we have these many solutions between E and E+dE and the pull and the length of the
box is Lx. Therefore, that divided by this is the number of solutions between E and E+dE
per unit length of your semiconductor and finally, to get to the density of states, we need
to get a per energy count, ok.
159
So, how many states are present per unit length per unit energy? So, we once again
divided by d E, we get rid of the dE term and you have this little answer here which is
the density of states for a 1D material. So, there are m 0.5 E-0.5√2 , there is a √2𝔥𝜋 states
per unit length per unit energy in a 1D material.
So, how does the density of states vary with energy for a 1D material? It varies as E-0.5.
So, you see a nice trend here. So, for a 3D material you saw that the density of states was
proportional to E0.5 for a 2D material the density of states was proportional to E0 or in
another sense, in other words, it was not dependent on E. And for a 1D material, the
density of states is proportional to E-0.5, so, that is how it varies from 3D to 2D to 1D.
160
So, what is the density of states for a 0D material. For a 0D material, you have neither a
set of points in kx nor k y nor k z. You only have one solution, one probably just 1 point in
k space. And therefore, your density of states is simply a direct function, ok. You have
states at one particular energy. So, if you wish this could be written as at one particular
energy and you have say 2 spins and therefore, that is permitted, ok.
This is finally, summarize the you know the concepts on the density of states and we
looked at 3D materials; wherein we found that the density of states depended on the E to
161
the power half, that is, it varied with the square root of energy for 2D materials. The
density of states did not vary with energy at all. So, you could call it as it varying as E0.
For 1D materials it varied with E-0.5 and for 0 D materials which are you know similar to
quantum dots it is just a delta function of the energy. There is just 1 or 2 energy states
and the 2 is (Refer Time: 10:18) for the 2 spins.
Now, in this course, we will focus mostly on 3D materials, ok. So, and therefore, this is
the density of states picture that is of great use to us. So, we have our semiconductor
which has got a valence band and it is got a conduction band. There is an energy gap, ok.
And that that is called E g and then you have states in the conduction band above the
conduction band edge and you have states below the valence band edge, ok. And the
density of states which is the number of states per unit volume per unit energy varies as
𝐸 − √𝐸𝑐 to the power half; so, this is like a square root behavior.
Similarly, here too you have a square root behavior with the density of state distribution
and inside the gap as long as the semiconductor is very clean, as long as it is the perfect
crystal, you have no states. So, if you look at this density of states here, it is 0. And you
have a density of states varying as the square root of energy above and below these
conduction band edges. So, this pictures of great use to us throughout this course and it
would sort of help gather some amount of intuition, you know it is a it is helpful to you
know visualize this very clearly.
So moving on so, what we have counted for now is the number of states available and
the electrons can now start occupying these states, ok. And, what we are trying to get at
ultimately is the number of electrons that are available for conduction because,
ultimately we want to measure the current voltage characteristics. Now, we know the
number of states, but we now need to find out as to what the probability of occupancy of
those states are.
162
So, the probability of finding an energy state, so, which is basically our solutions to
Schrodinger a solution Schrodinger’s equation. So, if you have a state at some energy E,
there is a certain probability that this state is occupied with an electron and that
probability is defined by this variable; I mean or the symbol f(E). It depends it is a
function of energy. So, it depends upon the energy and we call it f of E.
So, what is f of E? So, since electrons are fermions which means that, these are particles
that obey the Pauli’s exclusion principle you find that f (E) is defined by something
called as a Fermi Dirac distribution, ok. And it turns out that f of E is defined by this
1
function here which is 𝐸−𝐸𝐹 in this function that you see your K is your
1+𝑒 𝑘𝑇
Boltzmann’s coefficient, T is the temperature and E f is something called as the fermi
level.
So, it is an energy level which has got a special implication in semiconductors and it is
called as the fermi level. So, it is a technical term and we will continue using this and
you will typically use the symbol E f to denote the fermi level. Now, what do you mean
by the fermi level? So, if you were to take this function ok, f of E and plot it here, ok. So,
I have energy on the y axis and I have f(E ) on the x axis.
163
1
So, at T equal to 0, what is this function it is going to be 𝐸−𝐸𝐹 ok. So, this means
1+𝑒 0
this is going to be a very large number, but depending on whether E>E f or E < E f. This
number could be very largely positive or it could be very largely negative, ok. So,
suppose E is greater than E f ok. So, E f is some energy level. So, for now, let us just
mark it somewhere let us just call that E f. So, this is the energy scale and let us say E is
greater than E f which means that this quantity is positive and as the temperature
approaches 0. This number is a very large positive number which starts heading towards
infinity.
So, as T approaches 0, you have 1 by 1 plus a term approaching infinity ok. If E is

greater than E f; so which means that this value is going to be heading towards 0
therefore, for all energies at T=0, for all energies above E f, the value of the fermi
function or f of E is 0. So, you can see that this is going to be the fermi function it is got
a value of 0.
Now, the moment E> E f ok, you have 1 by 1 plus e to the power minus a very large
number, ok. So, this number starts heading towards infinity as T tends to 0. So, this value
now starts heading towards 1 because E to the power minus infinity is going to be 0 and
you have 1 by 1 which is 1.
So, therefore, for all values of energy less than E f at T equal to 0 kelvin, the function
takes a value f of E is equal to 1 and at E=E f. So, when you have E=Ef, ok. So, when you
have E=E f, the fermi function has got a value of 1 by 1 plus E to the power 0 by 0. So, in
some sense it is undefined so, we need to take limits. So, let us say T is heading toward
0.
1
So, you have which is 1 and therefore, this function tends to 1/2, ok. So, at E=E f at
1+𝑒 0
T=0, you have the function actually somehow making the switch from the value of 1 to
the value of 0. And therefore, you can say in the limit the value of this function is half,
ok. So, this is the fermi function, this is the behavior of the fermi function at T equal to 0.
So, what does this say? Since f of E is the probability as to whether an energy state is
occupied by an electron on f of E equal to 1 implies that the electron definitely occupies
that state and f(E )=0 implies that the electron definitely does not occupy that state. So,
164
what this picture tells you is that, at T=0, any state that is below E f has got a guaranteed
chance that an electron will occupy it. So, if an electron if a state exists here, it is
definitely going to be occupied by an electron and at T equal to 0.
What f of E tells you is that if a state exists above E f, it is definitely not going to be
occupied by an electron which means that the state is vacant or in other words that state
is definitely occupied by a hole ok. So, this is the meaning of f of E. Now, as temperature
increases, ok, so, this is what happens at 0, as T approaches 0 Kelvin, right. So, T
approaching 0 Kelvin is T, T equal to 0 Kelvin is not possible, ok. That is, it is against
the laws of thermodynamics.
So, let us say, let us take a more realistic case where you have a temperature T 1 that is
greater than 0 ok.
So now, what happens to f of E. So, you will find that if you were to plot for some
temperature T 1, if you were to plot this function, you will find that it is got this
particular shape. So, it is deviated from it is T equal to 0 shape and it is got a shape that
looks like this.
1
So, what does this say? As E<<E f ok so, 𝐸−𝐸𝐹 becomes a significantly negative
1+𝑒 1
number, ok. So, I let us call it let us call it x for all practical purposes, just for the sake of
165
for the sake of arguments here. So, as x tends to so, x is your E minus E f. So, as E
becomes. So, as E becomes less than E f ok. If E <E f x becomes a negative number.
And if you have an exponent of a very largely negative number, it is going to be much
less than 1 and therefore, this function tends to 1. So, at where as you head very far away
from E f and below E f, the firm value of the function heads towards 1 asymptotically and
equivalently as you had above E f and very far away from E f, ok. So, as you had go to
higher and higher energies, x is a positive number that is continuously increasing and it
this term dominates this term and you will find that 1 by a very large number tends to 0
and therefore, this function here starts heading towards 0.
But at E=Ef, your x is equal to 0 ok. So, x is equal to 0 and which means that this term
here is equal to 1 and therefore, at equal to E f the value of the function is 1/2. So,
irrespective of the temperature, you will find that when my E=Ef the function takes a
value of half or in other words the function crosses this point this coordinate of the x
coordinate being half this plot.
If you plot energy versus f of E, you will find that for any temperature you will always
the function will always cross this coordinate of half comma E f ok, that is E f as being
the energy and half being the probability. So, at temperature T 2, the function deviates
further, ok. It moves further away from the T equal to 0 condition and it sort of start
begins to smoothen out.
166
So, you will see that the occupancy probability above E f begins to increase and the
probability of non-occupancy below E f also begins to increase, ok. So, that is the nature
of f of E and this nature is very important ok. So, we have identified the number of states
defined by the density of states and what we are doing here is we are trying to find out
what the probability is that those states are occupied by an electron, ok.
So, let us look at some of the key properties of this f of this probability, ok ao, which is
called as the probability of occupancy of the electron defined by f of E.
So, what are the key points, ok. So, these are some key elements to remember. So, f of E
is the probability ok. Please do not mistake it for a density function. It is not a density
function, ok.
So, if it were to be a density function then f of E d E would have to be the probability

and which means that this summed over all the states should be equal to 1 which is
clearly not the case. So, please do not mistake it to be a density function f of E is the
probability.
167
And, if it is a probability, what is the random variable? The random variable is as to are
the 2 conditions or the 2 states right, given an energy level is it occupied by an electron
or is it not occupied by an electron, ok.
So, these are the 2 possibilities it is like a coin toss with a head and tail being the 2
outcomes. So, here you have 2 outcomes which is given an energy level is that energy
level occupied by an electron. So, it is let us say occupied is one state and not occupied is
the other state and the probability that it is occupied is given by f of E.
So, which implies that if f of E is the probability that it is occupied, what is the
probability that is not occupied? It has to be 1-f(E ) ok. Because the probability of the
summation of the probabilities of these 2 states occurring has to be equal to 1 it has to be
either this or that they cannot be a third option. So, so that is the second point which is
the random variable is the occupancy or non-occupancy.
Now, what is E f? E f is the fermi energy level as is the fermi level ok. So, it defines an
energy at which f of E takes a value of half and we already discussed that. So, if you look
at the value of this probability of occupancy, it takes a value of half at E=E f.
And we already saw that, the probability of non-occupancy that is given an energy state
at a level E, the probability that is occupied by an electron is f of E. And therefore, the
168
probability is not occupied by an electron is 1-f(E )which is what which is simply saying
that this is the probability that it is occupied by a hole ok, which is a vacancy.
So, f of E is the probability of probability that a state is occupied by an electron and 1

minus f of E is the probability that that state and at energy level e is occupied by a hole.
And then finally, if E-E f ok. So, if this term is much larger than k T ok. So, I will not
say much larger. Let us say it is greater than 3 k T which is basically implying that this
value is much greater than 1.
1
So, if the sum of the denominator can be approximated to 𝐸−𝐸𝐹 , then this becomes the
1+𝑒 𝑘𝑇
Boltzmann distribution ok. So, this is your Boltzmann distribution which is simply
saying that, as the energy goes up, the probability that you will find an electron goes
down exponentially. Now, if E minus E f is approaching values of K T, you cannot what
he say simplify this denominator and you need to take into account the entire fermi
function value.
Similarly if 𝐸 − 𝐸𝐹 ≪ 𝑘𝑇 ok so, this is this was the case when 𝐸 − 𝐸𝐹 ≫ 𝑘𝑇 if 𝐸 −

𝐸𝐹≪ 𝑘𝑇, 1-f (E ) which is basically the probability of occupancy by a whole approach
approaches the Boltzmann distribution which says that the probability that a hole
occupies an energy level E goes down exponentially as E goes much less than E f ok. So,
that is basically these approximations.
And we will use these approximations quite a lot because it helps simplify a lot of the
mathematics and the calculation now. So, this is all we have to say about the fermi
function and the fermi level and the fermi distribution, but the. So, let me tell you a little
bit more, let me give you some intuition about this. So, very nice way to imagine it ok it
may not be the perfect way, but it is an intuitive way to imagine, it is to consider a bucket
of water, ok.
So, let us say you have a bucket and you have got some water you fill these water
molecules 1 by 1. All these water molecules are filled up and you have a water level that
is still this point. Now, suppose I ask you the question so let us say this is all the air and
that is the water as you see it. Now, suppose I ask you this question as to where is the air
water interface, very intuitively somebody might point out to this layer here.
169
But does this mean that there are no water molecules above this. What is the probability
that I will find a water molecule below this level at this level and above this level ok.
Suppose, I ask you this question if the temperature T is greater than 0, you will definitely
find some water molecules above this level. It is just that they are going to be much
sparser as compared to the number of water molecules I will find below this level, ok.
So, this level can be imagined to be what is called as the fermi level ok. It is the in some
sense, the highest energy level that could be occupied by all these species at T equal to 0
ok, where the probability of occupancy is still not 0 ok it is equal to half at equal to 0.
Above this level at some temperature T greater than 0, you will find that the distribution
of water molecules that is a concentration of water goes down exponentially. It is like the
Boltzmann’s distribution ok.
And below this, you have a much larger probability of finding water molecules and if
you imagine the surface ok, let us say these are the water molecules in the surface. You
can see that, it is you know it is shared between the air and water boundaries. So, you
will sort of imagine that to be your you know probability being half, ok. So, it is a very
hand wavy, but very useful analogy to think of what the fermi level is ok.
And we will come back to this bucket of water examples every once in a while. So now,
if this is the fermi level, then what do you mean by the density of states with regards to
this water bucket analogy? So, the best the best parallel to the density of states is the
shape of the bucket, ok.
170
So, if you have a bucket that is shaped like that, you have a different density of states
distribution and the water molecules need to fill all these states and the fermi level set
sets itself at some particular point, ok.
On the other hand, if we have a bucket like this, it is got a very different density of states
and for the same number of molecules the fermi level will be located elsewhere. So, this
analogy cannot be used throughout I mean, it is just a very hand wavy visualization of
what we are talking about, but nevertheless, it is useful. So, that is the key element about
the fermi; fermi function.
Now, some of you might wonder as to where this f of E came about, ok.
171
So, although it is not a part of this course, I can post some notes online, probably I will
give you a link to where you can find these notes and essentially f of E comes about by
finding out the best possible way to distribute say n electrons among s different states.
So, these are n identical particles distributed in say a set of empty say states or seats, ok.
So, that is, if you find the many methods to do it, the best possible way where you can
maximize or you know find the best possible options is what defines your f of E ok. So,
it is a bit of statistical mechanics, all right.
172
So, what do we have? We have the number o-f states which is defined by the density of
states. So, if the density of states tells me how many states are there per unit energy per
unit volume and we have the probability that those states are occupied by an electron. So,
using these 2, we can now begin to count the carriers, ok. So, we are ultimately
interested in counting the number of electrons available for conduction.
Most particularly, if this is the conduction band edge and this is the valence band edge,
we are interested in the electrons in the conduction band and the holes in the valence
band. We are not so interested in the electrons in the valence band. There are many of
those because these are all trapped to these are all bound to all the different silicon
atoms, ok. There are many of many electrons in the valence band and equivalently, there
are many holes in the conduction band it is largely empty.
So, we are not interested in the holes in the conduction band and the electrons in the
valence band. But, we are very interested in the electrons in the conduction band and the
holes in the valence band because, these are the species that are going to you know
constitute my current through these devices and ultimately, I want to get to a point where
I can start to define my current-voltage characteristics and all these devices.
So, always remember, we are interested in the electrons in the conduction band for E
greater than E c and holes in the valence band; that is E less than E v. This is of interest
to us and we want to establish a method to count these, ok. So, how do we count, how do
we identify the number of electrons? So more particularly, the number of electrons per
unit volume so, number of electrons per unit volume which we will call as the electron
concentration, ok. It is not per unit area right now, it is per unit volume and the number
of holes per unit volume that we will call as the hole concentration.
So, how do we identify these 2 ok. So, if you ask yourself this question, what is the
number of electrons per unit volume which we will define by the symbol n of E lying
between E and E+dE. So, I have 2 energy levels ok and this energy level is E that energy
level is E plus d E and we want to know how many electrons are present between these 2
and we will call that as n(E). It is a function of energy. If this E changes, the n will might
change.
So, the answer is very clear. So, you need to first know how many states are there for the
electrons to occupy ok. If there are no states, there can be no electrons because the
173
electrons have to have a state to occupy because these states are essentially the solutions
of Schrodinger’s equation. So, you need to identify the number of states per unit volume
per unit energy or let us say the number of states per unit volume between n E E plus d
E, which is nothing part the density of states in to the density of states is the number of
states per unit volume per unit energy.
But, since we are interested in E between the number of states between E and E plus d E,
so, this energy gap is has got a value dE. So, it is density of states into d E which the
number of states between E and E+dE per unit volume. So, I know the count of the
number of states, but it does not mean that all these states are occupied by electrons.
Because, there is a certain probability that the electron will occupy these states and that
probability as we saw is defined by f(E ).
So, f of E into the number of states between E and E+dE is my electron count, that is n of
E. So, if you were to take a semiconductor, that is extremely clean ok which means that
there are no states in the band gap ok. there are 0 states in the band gap. And let us say, I
pick an energy level between E and E+dE somewhere there at a temperature T greater
than 0.
And we ask the question as to how many electrons are present between E and E plus d E.
Well, the first answer is, the number of states that is the density of states into d E is 0 f of
E is not 0 because f(E ) could be taking a value that is you know nonzero. So, that is a
probability that an electron will occupy that energy level provided. There is a state for
the electron to occupy it, ok.
But, since ours band gap is very clean and there are 0 states, even though f of E is not 0,
the number of electrons to be found is 0. On the other hand, if let us say we want to find
the number of electrons in some energy level lying in the conduction band, ok.
174
So, let us say that is E c so, that is my E c and I want to know the number of electrons
between E and E+dE. So, what do I do? I need to identify the density of states into d E.
So, here I do have a density of state distribution and I know that it varies as the square
root of E minus E c is proportional to this.
So, we do have states. So, E the density of states into dE tells you the number of states
per unit volume in this region and f of E is the probability that these states are occupied
the probability is going to be low. But, it may not be 0 at T greater than 0 ok. At T equal
to 0, it is definitely 0 but at T greater than 0 this probability is non zero.
Therefore, you do have some probability a very small probability that these states are
occupied. And therefore, you can establish a carry account and it is these electrons that
are going to help you in providing a current through the device, all right. So, next, if you
ask the question what is the number of holes between E and (E+dE). So, which means, I
need to identify p of E which is the number of holes per unit volume between E and E
plus d E.
Then again, I get the density of states. I need to find the number of states available, that
is the density of states into d E into the probability that these states are not occupied, f of
E is the probability that states are occupied. But, the probability that the states are not
occupied is 1-f(E ). And therefore, this is my count of the number of holes per unit
175
volume between E and E+dE. So, this is a very useful relation to remember. It is again, it
is good to develop intuition with regards to these concepts, all right.
So now, we are interested to find the current ok. And it is all the electrons above the
conduction band edge that are going to contribute to the current and all the holes below
the valence band edge that are going to contribute to the current. So, we are not only
interested in the number of electrons between some energy level E and E plus d E ok.
But instead, we are interested in the number of electrons from E c that is the band edge,
the conduction band edge, all the way up till infinite, ok. That is all the electrons present
in the conduction band, because, all these are going to aid my conduction.
So, how do I collect or count all the electrons above the conduction band edge? It is
straightforward. It is going to be the answer if obtained the previous example which is n
of E, but summed over all the different energy levels, right. So, this is n of E and I am
going to take a summation which is basically, I am going to take an integral from E c till
infinite of f(E ) into the number of states between E and E+dE.
So, if I were to sum all these electrons present in all these little bands from E to E+dE,
then, I will get my total number of electrons in the conduction band and if I make use of
the Boltzmann approximation and I perform this integral, you will end up with a term
176
that looks like this. After you apply the limits, we will find that the number of electrons
from E c till infinite turns out to be this value in the case of a 3D semiconductor.
So, you see firstly, it depends on E to the power minus E c minus E f ok. So, if I have my
Ec here and if I have my E f here and let us say I have my E v here. This distance is
important and the further the fermi level moves away from E c, the smaller the number
of free electrons and that is clearly understood because, this is the probability ok. The
probability varies like this and if I were to move the fermi level away, the probability is
going to vary.
Because it is going to take a value of half at the fermi level so, my half will appear here.
Instead of a taking a value half there, it will take a value of half at a much lower value of
energy and it is going to take a much smaller probability of occupancy at the energy
levels above E c. So, as Ec -Ef increases, if that has that gap increases, my electron count
is going to start coming down ok. The free electron count is going to start coming down.
Now, this term here is a constant, ok. Now, this pre factor to this exponential is a
constant m of mE* is the effective mass of the electron, k is the Boltzmann’s coefficient,
T is the temperature, h is the Planck’s constant and that particular term is typically
denoted by a constant symbol ok. It is called N c which is the effective density of states in
the conduction band.
So, this is not the density of states, it is not the number of states per unit volume per unit
energy. It is the effective density of states which is basically the number of states per unit
volume in all energy is from Ec to infinite, ok. So, it is analog goes to that measure.
Now, similarly, if you want to find the total number of holes in the valence band because
that is what is going to help me carry current, we find that the number of holes in the
valence band is the summation of all the holes from E between E and E plus d E. But
going from minus infinite to E v so, that is my E v that is my E c. To find the number of
electrons, I needed to find all the electrons sitting from here all the way till infinite and to
find the number of holes in the valence band, I am going to find all the hole sitting
between E v and minus infinite.
177
So, I am going to integrate from minus infinite to E v 1 minus f of E. The density of states
in the valence band times dE is going to give me a measure that looks like this. So, this is
the total hole count in the valence band and what is that?
So, just like in the case of electrons, I have my E v and I have my E f and I have my E c.
We found that as this gap increases the number of electrons in the conduction band
decreases.
Because the probability is got a lot lower. Similarly, if this gap E f minus E v ok so, this
is E f -E v begins to increase, the number of holes available in the valence band will start
to decrease once again; why because, 1-f(E ) will start to get a lot smaller if this gap
starts increasing. So, and all this happens because f(E ) has to take a value of half at E f.
So, if I move E f around, I change the probability of occupancies in the states above E c
and below E v.
Now, just like we defined an effective density of states in the conduction band, we can
define an effective density of states in the valence band which is called as N v. So, the
number of holes available for conduction are the number of holes between minus infinite
to E v which is the effective density of states in the valence band which is essentially this
𝐸𝑣−𝐸𝐹
constant here times 𝑒 𝑘𝑇 . So, that is the way you count the number of carriers available
for conduction.
178
Lecture – 11
Carrier Concentration
So, now, let us sort of visualize this ok. So, let us say this is my fermi level Ef and this
dotted line that you see ok, so, this dotted line is your fermi function. So, this is my fermi
function that is f of E.
Now, at E=Ef, it takes a value of half and below this it starts to approach the value of 1
and above that it starts to approach the value of 0. So, that is your fermi function. So, that
is the probability that an electron occupies the energy level. This line here, this solid line
here, which I mark in red, is the density of states. So, that is the density of states in the
conduction band, which varies as the √𝐸 − 𝐸𝑐 and this is the density of states in the
valence band, which varies as the √𝐸𝑣 − 𝐸 ok.
So, these are the number of states available and that is the probability that these states are
occupied and the product of these two essentially gives you the electron count in the
conduction band and the whole count in the valence band. So, the product is given by
this region, which I am shading here. So, you see that the density of states is large and
179
right at this point the density of states is 0, the fermi level, the fermi f of E is not 0, but at
0 times some non zero value, which is essentially 0. So, you have 0 electrons here, but as
we just, as we just move away from this point as E goes greater than Ec √𝐸 − 𝐸𝑐 is no
longer 0.
So, you do have some density of state and you have the maximum possible fermi level,
fermi f(E ) value the maximum possible value of f of E, sorry not the fermi level, the
maximum possible value of f of E and therefore, the product begins to take a
significantly large number and therefore, you will find that the electron count increases,
but then as you move further up the density of states goes up, but the probability of
occupancy comes down and therefore, the electron count starts to fall off.
So, you find that the electron distribution and energy above the conduction band edge
looks like what is shown in this shaded region here. Similarly, if you look at the whole
population or the whole count in the valence band edge, you have the same story here.
You have the density of states being 0 and 1-f(E ), which is basically, this gap being non
zero ok. So, we are interested in this number.
Now and therefore, the number of holes right at this edge is 0, but then as we start
heading lower below Ev; Ev minus E increases, but at the same time the fermi level
starts approaching 1, which means this value, which is basically 1-f(E ) begins to
decrease and you find that the probability keeps getting down, which, which allows the
number of holes or the whole distribution in energy to take a shape as shown in this
shaded region here ok. So, that is a graphical visualization of how the electrons and holes
are distributed in energy all right.
180
So, now, let us start making some definitions. Now, let us say I have a pure
semiconductor and what I mean by pure is it is not, there are no impurities present inside
the semiconductor and I am bringing this concept up in the context or something which I
am going to talk about very soon, which is something called as doping. So, in a pure
semiconductor, you have your valence band, you have your conduction band and you
initially had a T equal to 0, you had all the electrons being bound by all the atoms of the
semiconductor lattice.
I know you had all the valence electrons all located in the valence band, the conduction
band was largely, it was completely empty at T=0. Now, as we started increasing the
temperature ok, the electrons gained some thermal energy and some of these electrons
were promoted to the conduction band. So, you started having holes or vacancies being
left behind in the valence band and electrons being promoted to the conduction band. So,
this was as we started increasing temperature.
Now, in the case of intrinsic silicon at 300 Kelvin, you will find a count, if you look at
the number of electrons per unit volume, sitting in the conduction band, it would be
about say 110 per cc all right. So, so this is what happens as you start increasing
temperature, which means that the number of electrons present in the conduction band,
must be equal to the number of holes present in the valence band, because it is these,
these holes are the consequence of the electron getting promoted to the conduction band.
181
So, if you were to measure, if you were to take it carry count of the number of electrons
in the conduction band and the number of holes in the, in the valence band. These two
are equal ok, one is led to the other and that concentration has got a special term for an
intrinsic or a pure semiconductor and that is something called as the intrinsic carrier
concentration and we will always denote it by the symbol ni ok.
So, both when I say the number of holes is equal to ni. It means that the holes have the
hole concentration has matched their intrinsic carrier concentration, which is
semiconductors like as though it would have been in a pure state and all this is a
conditions, which is known as thermal equilibrium which is I have not thrown light on
the semiconductor. You have not applied any voltage, you just have the semiconductor
sitting in dark sitting at a temperature at some temperature T and as a consequence of the
temperature you applied.
You have a certain statistical count or you have a certain population of free electrons and
holes and that population is called as the intrinsic carrier concentration, which is ni ok.
Now, if you go back to what we looked at what it implies is that, if you have a density of
states and you have a certain f of E the product of your density of states and f of E
integrated over your band edges that is this count matches the number of electrons. So,
the whole count matches the electron count and we, we always talk in terms of
concentration.
182
So, n is the number of electrons per unit volume p is the number of holes per unit volume
ni is the intrinsic carrier concentration per unit volume ok. So, now we can sort of bring
in some of the concepts we learned earlier. So, we earlier found that n is equal to the
effective density of states in the conduction band into e to the power minus Ec minus Ef
by k T where Ef is the fermi level right and Ec is the conduction band edge
𝐸𝑣 −𝐸𝐹
Similarly, p was equal to and 𝑣𝑒 𝑘𝑇 where nv is the effective density of states in the
valence band. Now, in an intrinsic semiconductor n is equal to p is equal to ni ok. So,
which is the intrinsic carrier concentration, this means that I should be able to equate
these two terms ok. So, what happens? So, n is equal to ni ok. So, ni is essentially that
term that p is also equal to ni.
So, p is equal to. So, ni is also equal to that term there and if I take a product of these two
equations. Let us say this is equation 1 and this is equation 2 and if I multiply these two
𝐸𝑐 −𝐸𝑣
equations, I get ni square on the left side and I get Nc .Nv.e^( ). So, because you
1
𝐸𝑐 −𝐸𝑣
have the Ef term will cancel off and you have Nc Nv𝑒 ( 𝑘𝑇
)
. So, that is Ec that is E V
and this is 𝐸𝑐 − 𝐸𝑣 which is nothing, but your energy gap.
−𝐸𝑔
Therefore, ni square turns out to be Nc.Nv𝑒 ( 𝑘𝑇 ) . Now, what is Nc? Nc is a constant
right, it only depends upon some universal constants and the effective mass of the
electron and v is also a constant, it depends upon several universal constants and the
effective mass of the whole, Eg is again at a given temperature, eg is a fixed parameter
right. You have the energy gap.
So, which means to say that, this ni2 square seems to be very much a constant as long as
you do not vary Nc Nv and Eg, you cannot vary ni square and this is a very powerful
relation, because it tells you a method to count at a given temperature, if I want to know
what is the intrinsic carrier concentration, it is essentially the square root of N c N v into
e to the power minus Eg by 2 k T.
So, as the energy gap increases the ni decreases and that is very intuitive, because as the
energy gap decreases, I need a higher temperature for me to promote my electron from
the valence band to the conduction band and therefore, my intrinsic carrier concentration
decreases. So, this is a very powerful relation. It is a very helpful, a very useful relation.
183
So, let us play around with this a little bit more v. So, now, let us say we have our
intrinsic semiconductor. Now, when we discuss the fermi level, we really did not locate
it ok, we just sort of vaguely, drew it somewhere, between Ec and E v, we did not really
locate it accurately. So, what is the accurate location of the fermi level? So, we know Ec,
we know Ev, because in a semiconductor Ec is defined by the electron affinity ok.
So, there is a certain electron affinity and that defines, my Ec and then I have my energy
gap and therefore, I know Ev with regards to Ec ok, but where is this fermi level located,
because the fermi level location is important, because that is where f of E is going to take
a value of half f of E is equal to Ef is equal to half.
So, what is the location of the fermi level? So, once again let us get back to these
equations, we know that n that is the number of electrons in the conduction band is
−(𝐸𝑐 −𝐸𝐹 ) (𝐸𝑣 −𝐸𝐹 )
𝑁𝑐 𝑒 𝑘𝑇 p is 𝑁𝑣 𝑒 𝑘𝑇 and for an intrinsic semiconductor what we do is we took that
fermi level a special name ok.
So, we want to know where is the Ef for this intrinsic semiconductor we give it a special
name and we call it the intrinsic fermi level ok. So, E i or sometimes you know its Ef i is
also used this Ei or Efi is something called as the intrinsic fermi level and we are trying to
identify, it is location, which is basically, the fermi level location in a pure
semiconductor. So, for a pure for an intrinsic semiconductor I can replace this Ef by E i.
184
So, Ef we will keep it as a more general fermi level, variable and E i, we will use it
especially only to indicate the fermi level of intrinsic semiconductors. So, then in that
−(𝐸𝑐 −𝐸𝐹 )
case your n for a pure semiconductor, you will find n=n i, which is 𝑁𝑐 𝑒 𝑘𝑇 ok. I am
just replacing the variable Ef with the variable E i, because E i has got a special status, it
is the fermi level, location for a pure semiconductor and p is also equal to n i, which is
(𝐸𝑣 −𝐸𝐹 )
𝑁𝑣 𝑒 𝑘𝑇 .
So, that is your p and since both these have the same value, there or both equal to ni, you
can equate the number of electrons with the number of holes with that being the intrinsic
fermi level and just by, by solving this equation, you will find that E i =[(Ec+Ev)/2]
𝑁𝑉
+[kT/2]𝑙𝑛 . So, what is (Ec+Ev)/2, it is the location exactly at the mid gap ok.
𝑁𝑐
So, that is my energy gap, if I were to take a location here, it is Ec plus E v by 2 and
what this is saying is that you reach mid gap and you offset it by this quantity that is kT
by 2 ln of Nv by Nc, if you were to offset yourself from the mid gap by this amount of
energy, you will find the location of the intrinsic fermi level. So, to draw this more
cleanly, the intrinsic fermi level is not generally located right at mid gap, it is slightly
offset from mid gap.
So, let us say this is E c that is E v, let us say that is perfect mid gap Ec plus E v by 2.
Now, if this quantity Nv by Nc is a positive, is, is a say Nv >Nc. So, which means that
this quantity is a positive number then my Ei e location, my fermi level is actually
located there. So, that is my intrinsic fermi level position, which means that my f(E ) ok.
The probability of occupancy will take a value of half at that location and that energy.
So, what; so, when does this term become equal to 0, because if this term had to become
equal to 0, then is then the fermi level, the intrinsic fermi level is located exactly at mid
gap. This term will become 0, if Nv=Nc, because then you have the logarithm of 1,
which is 0.
So, when does Nv become equal to Nc, if you look at the expressions for Nv and Nc, you
will find that you have most of the terms to be constant or most of the terms of universal
constants, except for the effective mass of the holes and the effective mass of the
electrons. So, if the effective mass of the holes matches the effective mass of electrons
then you will find that the intrinsic fermi level lies exactly at mid gap.
185
Otherwise, it is going to be offset by this number and in fact, you could take this 3 by 2
and probably locate it outside the logarithm. So, you can call that 3 by 4 ok. So, that is
your expression for the location of the intrinsic fermi level ok.
So, now, we are going to approach a different topic and we are going to call this as
something called as doping and I will explain what doping means.
Now, we know the location of the intrinsic fermi levels. So, you know this is Ev, you
know that is Ec and I know my intrinsic fermi level location that is E i, that is the fermi
level in a pure semiconductor. Now, I could. So, the pure semiconductor implies I have
my silicon lattice right it is a yeah. So, if you are going to talk about silicon, you have
your silicon lattice, which each silicon atom being bonded to four other silicon atoms ok.
If you were to look at it, in the tetrahedral lattice formation it is essentially this, it is got
four neighbours and these have got four neighbours and so on, the lattice continues that
way. So, if I were to now, create a defect, I get rid of the silicon atom and I bring in some
other species. The silicon has got four electrons in the valence shell; I can bring in some
other species that sort of forces itself to bind and accommodate itself into this lattice but
in order to do.
So, it could change the electron or hole population it could either give away an electron
or take away an electron. So, let us say you are going to add an impurity into the lattice,
186
you are purposefully adding a measured impurity concentration into the lattice and if this
is all done at a suitably high temperature, which means if it is, if the temperature is not 0
kelvin. These impurities could get ionized and they could contribute or change the
electron and hole population in the semiconductor.
So, if the electron and hole population changes the fermi level location in this doped or
impure semiconductor will shift away from Ei ok. So, this is E i it is got a special
relevance ok. So, it is an imaginary fermi level to remember, because that is what your
pure semiconductor look like. We will draw it with a dotted line. Now, if you take
another, if you take the same semiconductor and you dope it, you are going to create
impurities and you could the fermi level, could take up a new position and we will now,
call it Ef ok.
It is no longer Ei and E i is just this dotted imaginary line, which is no longer valid for
this new semiconductor, but it gives you a reference as to how much of impurity you
have added to the semiconductor. So, you could actually artificially move the fermi level
location about by adding impurities, in the semiconductor and this concept of adding
impurities to the semiconductor, in order to relocate the fermi level is called as doping.
Now, why will the fermi level relocate? Firstly, because your n is equal to Nc e to the
power minus a in, in, let us say in an intrinsic semiconductor, in a pure semiconductor. It
was (Ec-Ei) /kT. So, that was your n=ni. Now, I have changed n, now, I made n to be
different from ni ok. So, let us say I made n to be greater than ni. I cannot have this
exponent, have the same value here.
So, now my Ef, it will have a different value for Ef, which is going to be smaller as
compared to Ec minus Ef. So, in general, so, we will get back to the general expression,
which we derived, which we looked at earlier by you know sort of integrating the from 0
to from our Ec to infinite when we integrated the density of states into f of E into DE we
obtained this expression and this expression had the fermi level term in it.
And for an intrinsic case, this fermi level was the special intrinsic fermi level, but in
general we will retain this symbol Ef and increasing n will relocate Ef in order to
decrease, this term decreasing n will relocate Ef to increase this term, because it is E to
the power minus a larger number to in order to reduce n. So, if a semiconductor is doped.
So, as to increase the electron count, you will find that this term has to decrease or the
187
fermi level will move above will move away and above, the intrinsic fermi level and this
semiconductor is called as an n type or an n doped semiconductor.
The doping or the impurity was added to increase the electron count. On the other hand if
the doping or a dopant is added to increase the whole population then it is called a p type
or a p doped semiconductor.
Sorry, if the dopant is added to increase the whole population. It is called a P type or a P
doped semiconductor, where your Eb minus Ef. So, this term has to increase or decrease
in order to accommodate that.
So, you will find that if you are trying to increase the whole population Ef will have to
move below E i in order to, in order to, get closer to the valence band edge. So, that is E
v ok. So, now, how do we; so, let us just talk about more technical terms. So, you have
two kinds of doping ok. So, you have p type doping, where you are trying to increase the
whole population and relocate the fermi level. So, this is the intrinsic fermi level, you are
going to relocate the fermi level to some location here.
So, that to; so, as to bring down, this value of Ef -Ev and this kind of doping was called
as p doping and it is done with something called as acceptors. So, what the concept here
is that you have your silicon lattice. So, let us say, we will just draw these 5 atoms and
initially each silicon atom was bonded to 4 other silicon atoms.
188
So, it had a valence it had 4 valence electrons and now, if I were to remove the silicon
atom and bring in an impurity, which has got only three electrons ok. It is this impurity
has got three electrons and it is going to participate in bonding. So, it is going to bond
with 3 of the silicon atoms, but in order to bond with the fourth. It is going to take away
an electron a free electron, from the lattice and it is going to bond with the 4 silicon
atom.
So, it is going to complete, it is bonding by taking electrons away from the existing
electron population and it is going to essentially leave behind a hole. So, therefore, in
some sense, it is improved the hole to electron ratio. It has increased the hole count with
respect to the electrons and therefore, this dopant, which accepted the electron is called
as an acceptor like dopant and this kind of dopant is called a p type doping.
Now, you can depending on the concentration of dopants. I have added, I can change the
hole and electron population of accordingly, and therefore, it is useful to define a
concentration for your acceptor ions and we typically denoted by the symbol NA. So,
capital NA is the number of acceptor dopants added to the intrinsic semiconductor per
unit volume. So, it is, it is always per unit volume ok. Now, similarly we can add
something called as donor dopants in order to create a n type semiconductor, which is
you have a silicon lattice, where the silicon atoms are all bonded to 4 other silicon atoms.
And now, I am going to bring in another species or different an impurity, which has got 5
electrons in the valence shell. So, we are going to bring in an impurity, which is got 5
electrons, which means it is going to use up 4 electrons to bond itself, to accommodate
itself into the lattice, but it is got this fifth electron that is free and if it is ionized, this
fifth electron is free to move about in the lattice.
So, this is essentially increased the electron population. So, this kind of doping, which
increases the electron population is called an n type doping and it is done with donor
dopants, because this dopant donated an electron as opposed to accepting an electron, it
is called as a donor dopants and what this does is, it essentially moves the fermi level
above Ei in order to decrease the (Ec- Ef ) gap and increase the electron concentration.
189
And once again, it is useful to define a concentration for the donors and we do it with the
symbol N D, which is the number of donor dopants per unit volume ok. So, what does
doping actually mean. So, it means that you have added and ion ok. So, if this dopant is
ionized, what you have essentially done is you took a neutral species as a neutral.
Let us say you are trying to do n type doping, you have taken a neutral species, which is
an atom, which had 5 electrons in the valence shell and several other electrons in the
inner shells and it had some positive charge in the nuclei and you implanted it as a defect
inside your silicon lattice and since one of the electrons could not be used for bonding.
So, these four were used for bonding.
But this electron could not be used and at a significantly high temperature, this electron
got away and became a free electron in the silicon lattice. This is free, which means that
this species now is essentially positively charged, because the charge in the nucleus is
knob not balanced by the electron charge cloud. So, it is fixed, because it is now, bonded
itself to the silicon lattice, it is not a moving charge. This positive charge is now fixed.
But it; so, it is a fixed ion, which has got positive charge and it is left behind, free
moving electron, it has got a negative charged, negative charge. So, essentially by doping
the silicon atom n type, we have maintained charge neutrality, for the entire system,
which is the silicon is charged neutral, the dopant plus this, this ionized dopant plus this
extra electron balances itself out in terms of charge and the entire species charge neutral.
It is not only a charge to the silicon atom ok, it is charge neutral.
All we have done is added an entity, which was charged neutral and ionized it, which
means we have pulled an electron away and left behind a fixed in mobile positive charge.
So, that is what has happened, when we doped it n time. Similarly, when we doped it p
type, we have not changed the charge, we have not disturbed the charge neutrality of the
silicon.
So, we took her neutral species which had three electrons and an equivalent balancing
negative or positive charge inside positive charge is the nucleus and we added this into
the silicon lattice and this species now gained an electron, it accepted an electron in order
to bond with the silicon lattice and by accepting electron, it became an effectively
negatively charged ion, because now the positive charge in the nucleus does not balance
the excess negative charge on the shell.
190
So, it became a negatively charged ion which is an immobile ion, but it left behind a
positively charged mobile species called the hole ok. So, the hole is free to move about
in the silicon lattice and you have an ionized species here which is got a negative charge
and is fixed. So, essentially what you are doing, my doping is for p type doping we are
creating fixed negative charges immobile negative ions in the silicon lattice.
And mobile holes added to the silicon lattice these are mobile and in the case of n type
doping we are creating fixed immobile positive charges and mobile electrons for the
silicon lattice. So, essentially the entire system is charge neutral. Now if all the dopants
were all ionized, if everything that we added were all ionized what is the concentration
of this fixed negative charge, these negative ions in the case of p type doping.
It is an NA, because we added any acceptor like dopants per unit volume all these NA
dopants were ionized and all of them gained a negative charge. So, you have NA fixed
negative charges per unit volume sitting in your p doped silicon. On the other hand, your
donors we added ND donors per unit volume. All of them got ionized and they all
became positive ionized species and we have ND positively charged immobile species
located per unit volume in the silicon.
So, that is the situation when we dope it and clearly, we are affecting the electron and
hole concentration, which means we are affecting the fermi level location. Now it is
possible to counter dope a silicon so you add not only NA or ND we add both positive and
negative or we add both acceptor and donor dopants. So, this is something called as
counter doping now if we add both ND and NA, but your ND turns out to be greater than
NA, then it is equivalent to you doping the semiconductor n type with an effective
dopant concentration of (ND-NA), it is like as though we have added this many donor
atoms
On the other hand, if NA is greater than ND the semiconductor will behave like a p type
with an effective dopant concentration of (NA- ND). Now it is possible to balance NA and
ND and make the semiconductor pure, like as though it is intrinsic, and this is something
called as complete compensation. So, we have completely compensated any dopant with
the counter. So, these things are quite useful, and it would become clearer when we do a
problem session that we will solve some problems, you know just to calculate and get
comfortable with the calculations.
191
Semiconductor Fundamentals
Lecture – 12
Doping
So, here is a better a clearer picture of what I was talking about which is, when you dope
the semiconductor n type, you find that the fermi level E f moves above and towards the
conduction band. So, this is the imaginary intrinsic fermi level position which is no
longer valid because, we have we no longer have a pure semiconductor.
So, this was the fermi level location when the semiconductor was pure and we now
doped with n type; which means, we have increased the electron concentration, which
means we need to now decrease the fermi level, has to pick a place where E c minus E f
has to be smaller. Because, the electron concentration varies as e to the power minus this
gap; there E c -E f gap.
So, the fermi level will have to move closer to the conduction band edge. Similarly, for p
type doping, we have to increased the whole count which means, we need to reduce the
192
gap between E v and E f. And therefore, what you say the hole, the fermi level moves
closer to the valence band edge.
So now, these equations if you remember, well these all based on the Boltzmann’s
approximation. So, we started off with the fermi function which is f ( E) is equal
1
to 𝐸−𝐸𝐹 and we assumed that e minus E f>> k T,.
1+𝑒 𝑘𝑇
So, if you keep doping the semiconductor more and more and let us say the fermi level is
pushed closer and closer towards the fermi level, you will we will end up with the
situation where E f is so close to the conduction band edge that E c minus E f is no longer
greater than k T. Let us say of the order of k T. It is no longer much greater than your k
T. In those cases, we call the semiconductor to be degenerately doped.
And that is a technical term and by definition, the point at which we start calling the
semiconductor degenerately doped is the point at which E c minus E f ok. E c minus E f
this gap the fermi level has gone so, close to the conduction band, that this gap has
become less than 3 times k T ok.
So, at T equal to 300 Kelvin you know, k T is equal to about 0.025 million electron volt
which is about 25 milli electron volts. So, 3 times that is about 0.075 electron volt and
when E c minus E f is less than that value, we say that the semiconductor which is n type
doped is not degenerately doped and a lot of the calculations have to be done more
carefully and we cannot make many of the assumptions we have made so far.
Similarly, if you are going to dope the semiconductor p type and you start pushing the
fermi level closer and closer to the valence band, we will reach a point where E F is
within 3 k T of the valence band edge and once again, we say that the semiconductors
degenerately doped p type and once again we cannot make many of the assumptions we
have made.
So, as far as this course is concerned, we will not look at degenerately doped
semiconductors, we will be very much within this 3 k T threshold and the fermi level
location will be well below these borders. So, if this is the border force, this is say 3 k T,
that is called these 2 energy levels are just to be very clear.
193
As far as the scores and the tests and the assignments are concerned, if this is my E V
and we say this is E V plus 3 k T and that is E C-3 k T. We will only consider cases
where the fermi level despite any amount of doping lies in these regions in within this
band ok. So, that we can still comfortably go ahead and use all these simple calculation
techniques to calculate your electron and hole count.
194
So now, what does doping actually do? So, why is it you know what is the physical
intuition behind the fact that doping significantly improves the electron and hole count?
And what do you mean by doping? Let us look at this with a different perspective.
We already looked at it. As a as a first glance, we know what it does. You know how the
dopant atom bonds with the silicon lattice and contributes an electron and or contributes
a hole. But, let us look at it more closely. So, when you dope, let us take an example of
an n type semiconductor, ok. So, we have a silicon lattice, ok. So, silicon has been a role
model. And therefore, I keep using the word silicon for semiconductor, but that does not
mean that these concepts are not valid for other semiconductors, all right.
So, let us say you had your silicon lattice ok. It had these 4 would we say bonds to the
other silicon atoms and we replaced this silicon with your N type dopant which is a
donor doping which is a species that had 5 electrons.
So, 4 of them participated in bonding and the 5th one was free to move about in the
lattice. And therefore, it left the donor with a single positive charge because there is 1
electron more that should have been with the donor for charge neutrality.
But is no longer at the donor and therefore, there is a single positive charge. So, if you
think about this species here ok, this 1 electron and 1 positive charge ok, what is the
binding energy? You know what does the energy that is held these 2 species together ok.
4
195
One can imagine this to be something like a hydrogen atom. It is not exactly a hydrogen
atom because, there are other electrons present in the donor.
But considering the fact that it is a single positive charge and a single electron that was
initially bound to this donor and which has now moved away from the donor because of
ionization, the energy needed to pull that electron away is the energy required to do. So,
in a hydrogen atom; so, that gives you that gives you a by comparing it to a hydrogen
atom, it gives you a good estimate as to the ballpark number for that energy for that
energy.
So, that is what we are trying to do. We are trying to estimate what does that energy
needed for me to take the electron away from the donor and push it into the conduction
band and by calculating that energy; we can actually calculate the energy level created
by the impurity.
So, what does that mean? So, in a pure crystalline silicon atom, a silicon lattice, sorry
you have your valence band edge you have your conduction band edge and if your
crystalline lattice is perfect, it is perfectly crystalline, no defects, no impurities. Then,
you do not have any states in the band gap. But, the moment you start creating defects,
you start adding impurities, you will start seeing states appear inside the band gap.
And as we move towards more and more disordered materials say you ultimately take it
towards amorphous silicon or a glass, you will find that there is a density of states inside
the gap itself. So, the gap is no longer clean, but we will not go to such extremes at this
point. All you have done now is, we have added a donor, we have added an impurity into
the silicon.
Therefore, there are some traps that have appeared, there have been some states that have
been created inside this gap. And the purpose of this study is to locate those states where
are at what energy are those states sitting. And we are going to do that by identifying the
energy required to take the electron out from this little entity, which is a positively
charged core and electron around it.
Now, if you compare it to a hydrogen atom, the energy required would be this with some
corrections. It would be 13.6 electron volts, but we need to correct for certain elements
196
which is the permittivity of the material is different ok. So, we will we need to bring in a
relative permittivity. You cannot it can no longer be 𝜖0 which is 8.85*10-12. You cannot
have that value. You need to have a relative permittivity correction, ok.
So, you bring in an epsilon r square term and you also need to account for the effective
mass of the electron. You can no longer use the mass of the electron as is used in vacuum
because, you have the electron propagating through the lattice and you have this
interaction with the lattice which results in an effective mass of m e star.
𝑚𝑒∗
So, by taking into account these 2 corrections ok. So, we multiplied by a factor of and
𝑚0
divided by 𝜖𝑟 which is the relative permittivity of the semiconductor epsilon r square

which is a epsilon r is the relative permittivity of the semiconductor and continue using a
13.6 electron volt energy. If you make these corrections taking account, these corrections
and plug in numbers for silicon.
So, any styles will be same order as m naught ok. It is not orders of magnitude different.
It could be slightly more or slightly less depending on the difference semiconductors.
And for silicon, 𝜖𝑟 is the order of 11.7. Say, let us let us call it 10 ok. Just to give you an
easy calculation.
So, the binding energy is about 13.6 electron volt into something which is the order of 1
𝑚𝑒∗
which is your which is about 0.1 give or take 0.1 electron volt. In fact, it is less
100𝑚0
than 0.1 electron volts. Therefore, this trap energy is actually look at the moment you
add a donor, what this is saying is that, the moment you add a donor, you are creating a
trap state or you are creating energy states due to this donor which is located energy level
E d, ok.
197
So, this is something called as the donor energy level or the donor level and this E d is
located at around 0.1 electron volt from the conduction band edge. So, in the case of
silicon your band gap this is about 1.12 electron volts and this donor dopant has led to
the creation of states inside the band gap and those states are located at around 0.1 or in
fact, less than 0.1 electron volt, that is about 10 percent of the band the band gap, 0.1
electron volt from the conduction band edge.
So, what does this mean? It means that, this donor which was initially charged neutral
had a positively charged core and had all it is 5 electrons ok. So, this was the charge
neutral species and this was charge neutral. Let us say temperature at t equal to 0, but the
moment the temperature increased a little these electrons gained enough energy. They
gained this 0.1 electron volt energy to snap away from the donor and to get into the
conduction band of the lattice.
Therefore, these donors that lost their fifth electron all became positively charged. So, in
other words, they were ionized and these electrons which left the donor and moved into
the conduction band were now free to conduct, ok. So, we have increased electron count
in the lattice the free electron count in the lattice. So, we are not when I say electron
count, you know just as a means of speaking; I mean, it is just the it is understood that it
is the free electron count ok.
198
So, we are really not interested in the electrons in the valence band there are plenty of
those. So, the addition of the donor has increased the electron count the free electron
count in the conduction band and it has now become positively charged because of it
getting ionized and since the donor level is located. So, close to the conduction band
edge it really does not take too much of temperature to activate these electrons to ionize
these donors.
So, at 300 Kelvin, in the case of silicon, we are able to promote the electron sitting in the
valence band. If you add it, consider your pure intrinsic semiconductors, pure intrinsic
silicon, our 300 Kelvin we can promote about 1 e 10 per cc electrons from the valence
band to the conduction band and that is at an energy gap of 1.12 electron volt.
Here, I have a gap which is 10 percent of the original energy gap which is 0.1 electron
volts because of these donors. And therefore, it is much easier to ionize these donors and
this can happen at a much lower temperature. Again, we can ionize these donors and
promote all these electrons into the conduction band. And therefore, the addition of a
donor like dopant can significantly improve the electron population,.
The same argument can be made for holes. So, that is that is what the doping, that is
what doping is doing. If you look at it from the perspective of the energy levels and the
creation of these traps inside the gap.
199
So now, let us calculate again ok. So, we know we want to calculate the free electron
concentration and the hole concentration in the valence band after doping right.
So, this calculation is straightforward. It is just N c e to the power minus E c minus E f

by k T. So, that is the calculation, but then there is a much nicer or a much more intuitive
or much more useful relation that you can get now since. So, let us let us draw this that is
E c that is E v and now you have some E f which is different from your E i. E i is just
offset from mid gap. It is close to mid gap. So, I will just locate it there.
So, that was the fermi level for a pure semiconductor. But now, after doping, the fermi
level has shifted to E f and if it is n type doping, it is E f is going to be closer to E c. Now
this expression tells you that the electron count is N c e which is the effective density of
conduction band states k, effective density of states in the conduction band
into𝑒 −(𝐸𝑐−𝐸𝐹 )/𝑘𝑇 , but (𝐸𝑐− 𝐸𝐹 ) = (𝐸𝑐− 𝐸𝑖 ) ok. So, that is (𝐸𝑐− 𝐸𝑖 ) − (𝐸𝐹− 𝐸𝑖 ) E c minus
E i minus E f minus E i.
So, this term is essentially that minus this and we can sort of split that expression by
bringing in the E i term here ok. So, why am I why are we splitting this? Because, this
coefficient here is nothing but the electron concentration in the intrinsic semiconducto
because, your E f has now become E i and for a doped semiconductor this expression this
electron count is essentially this term into the exponent having an energy gap which is
the difference between the new location of the fermi level and the intrinsic fermi level
position
So, this expression is essentially 𝑛𝑖 𝑒 −(𝐸𝐹−𝐸𝑖 )/𝑘𝑇 . So, these 2 expressions are identical. So,
you could count your electrons by saying it is 𝑁𝑐 𝑒 −(𝐸𝑐−𝐸𝐹 )/𝑘𝑇 or we could say that, it is
𝑛𝑖 𝑒 −(𝐸𝐹−𝐸𝑖 )/𝑘𝑇 .So, this expression relates both of these expressions are useful ok. It
depends on the situation.
This expression relates the carrier count by noting down the energy difference between
the conduction band edge and the fermi level. Whereas, this expression relate makes a
carrier count by noting down the energy difference between the new fermi level position
and the intrinsic fermi level position. So, this is the fermi level position for an intrinsic or
pure semiconductor and after doping, it is relocated to E f. And, in this expression can
200
tell you what this new fermi level. You know, how far this new Fermi level moved away
from E i ok.
So, both these expressions are useful and we could you know create a similar expression
for the holes which is the whole population in the valence band is basically N v into e to
the power (𝐸𝑉− 𝐸𝐹 )/𝑘𝑇.
We again bring in the E i and you find that the whole population can also be represented
as 𝑛𝑖 𝑒 −(𝐸𝐹−𝐸𝑖)/𝑘𝑇 . So, this expression; So, which is this expression here for the electron
count and this expression here for the hole count looks more or less similar except for the
fact that the exponential here has got a positive sign and the exponential here has got a
negative sign, ok.
So, that is the that is the major difference. So now, what is n into p in the case of an
intrinsic semiconductor which can be what was the electron and hole count in an intrinsic
semiconductor my 𝐸𝐹 = 𝐸𝑖 . Therefore, this experiment is 1 and therefore, the electron
and hole count were both equal to n i. So, in an intrinsic semiconductor we had n is equal
to p is equal to n i and the product 𝑛𝑝 = 𝑛𝑖2 .
But, what about the doped semiconductor r? What is this product? N p, is now equal to
𝑛𝑖 𝑒 −(𝐸𝐹−𝐸𝑖 )/𝑘𝑇 these 2 terms cancel. And therefore, this product is still equal to n i
10
201
square. So, irrespective of whether the semiconductors doped or intrinsic the electron
count into the whole count is always equal to n i square, provided the semiconductor is at
thermal equilibrium which means that you have taken the semiconductor.
You have either captured intrinsic or doped it you have taken the semiconductor and you
kept it in the dark. You have not applied any light. You are not thrown applied any
voltage. You kept them in the dark as it is with the only influence on the semiconductor
being due to temperature. So, that was thermally that is thermal equilibrium. So, every
process that is going on inside the semiconductor is very nicely balanced by a reverse
process.
So, there is nothing changing. I mean, there is there are many things going on you are
creating a lot of free electrons. You are losing a lot of free electrons. All this is going on,
but it is all balanced in a very nice way, that statistically speaking, then the intrinsic
carrier concentration is fixed.
And, you know the other properties and semiconductor is not really waiting with time
ok. So, that is what you mean by thermal equilibrium and this property that n p is equal
to n i square is very useful. Because, if I know my electron count and I want to find my
whole count, I just need to do 𝑛𝑖2 /𝑛, ok. So, this gives me my whole population. So, this
relation that the n p is equals to n i square something called as a mass action law.
And this might appear to be a very odd looking term. You know what is the meaning
behind mass action law? But, it essentially comes from looking at chemical reaction
rates, ok. So, you can think of it as a chemical reaction that is going on. And what is the
reaction it is got to do with the generation and the recombination or the loss of these free
carriers again? We will talk about that later. So, that is the that is the process that is
going on and just like how you sort of consider reactants in a reaction for a chemical
reaction.
They consider the electron and hole count in this process of thermal generation of
carriers and thermal recombination of carriers which we even talk about soon.
11
202
So, what does doping do. So, we looked at the we now have a good understanding of
what doping does in terms of equations, but it is always nice look at this graphical
picture, ok. So, let us go back to our picture for intrinsic semiconductors, ok. So, this was
my density of states there are many plots drawn on this.
So, please do not get confused. So, that is the density of states. So, that is density of
states versus energy this dotted line is the fermi function. It is f of e versus energy and as
we notice, it takes a value of half when it crosses E is equal to E f and since this is an
intrinsic semiconductor, this E f =E i.
So, that is my intrinsic semiconductor position and we already saw that the electron
count was basically the product of the density of states in the conduction band and the
probability of finding carriers in the conduction in these states. So, that was my electron
count which is this product and that is the whole count.
Now, during by doping, what we have done is if you dope it, n type, ok. So, this is the
intrinsic case where n is equal to p is equal to n I 𝑛𝑝 = 𝑛𝑖2 If you dope it, n type we have
added a donor dopant and let us say this is my imaginary E i, this was E i in the intrinsic
semiconductor. By adding a donor doping, we have increased electron population and we
have effectively moved E f closer to the conduction band edge and it is located above E
i.
12
203
Now, since the fermi level is located at a different location from E i, the fermi function
shifts ok. It is no longer going to remain the same v. It is now going to adjust itself. So,
as to take a value of half in this point, earlier the fermi function was taking a value of
half at E i. But now, it is going to take a value of half at E f. Because, earlier E f was
equal to E i. So, and now E f is not equal to E i.
So, if the fermi level if you if you can imagine the fermi function as moving having
moved up a little ok, along with E f ok. Then, it is very clear to see that the probabilities
here are now suddenly increased the values which are initially very small. So, let us say,
initially the value here was less than half and let us say it was 0.2. But now, since we
have moved the entire f of e curve upward, this value is now much greater than 0.2 ok.
At here and here it was half which is 0.5. So, here it could now become 0.3 that is the
value of f of e which means the probability of me finding an electron in these states has
increased the density of states hasn't changed ok. It is not going to change unless you
dope it very strongly. We are not talking about those levels of dope.
If the density of states hasn’t changed, but the probability of finding an electron in those
states has increased in the case of n type doping [noise.] The probably defining electrons
in the states above E f have increased in the case of n type doping and it is not only that
the electron population is increased. So now, the electron population is increased
significantly, ok. So, you can see that this shaded region has got a larger area as
compared to this region here which was the intrinsic case.
It is not only that the electron population is increased the whole population has also
decreased ok. Why is that because, this is the fermi function. Let us say, this was this is
the value of 1. It is asymptotically heading towards 1. As you go to minus infinity and
the probability of finding hole depends upon this difference, it is 1-f(E ).
Now, initially the probability of m e finding a hole at say some energy level here was 1
minus f(E). Let us call that to say 0.2, but since we have moved shifted the entire fermi f
of e curve upward because E f has gone up. These values are not suddenly decreased. So,
what was 0.2? At some energy level there has now become a much smaller number.
13
204
Let us say it is 0.1. These are just numbers under stuff, these are not calculated values,
these are just numbers I am putting out in order to get the message across. So, what was
the 1 minus f of e was much larger than intrinsic case. But the moment we doped it n
type, 1 minus f of e for energy levels e less than E f have suddenly gone down which
means, the probability of m e finding a hole in these states has suddenly gone down. And
therefore, while the electron count has increased the hole concentration has decreased.
The area under this curve which is symbolizing the whole population is now become less
than the case of the intrinsic semiconductor. So, in an intrinsic semiconductor, the
number of holes were n i, the hole concentration was n i, the electron concentration was
n i in an n type semiconductor. The electron concentration is greater than n i, but the hole
concentration has gone less than n i and how much less.
If n is the electron concentration the hole concentration has to be n i square by n. In the p

type semiconductor, the opposite happens if you have doped semiconductor p type, we
have moved the fermi level down. So, this is my intrinsic fermi level position; that is, E i.
We have now moved the fermi level below E i,.
And, f(E ) has to shift downward in order to take a value of half. So, you can think of
f(E ) as you know sort of pinned to this fermi level position, right. So, if this is the E f
imagine a sort of a pin right there because, it always has to take a value of half. So, as we
move this up and down, this entire curve also has to move up and down it has to move
either up or it has to come down. So, that is f of e responding to the relocation of E f.
So, by now, moving E f downward, f of e is also shifted downward which means that the
probability of m e finding an electron has now gone down drastically. So, it was 0.2.
There it had increased to 0.3 for n type doping. But now, it has become 0.1, that is a 1 the
probability of m e finding a electron in this stage above E f has gone down.
So, although the density of states hasn't changed, the number of electrons in the
conduction band has decreased and it will decrease to a value less than n i which was the
intrinsic case for a p type dopant. For n type doping, the electron count has increased
electron concentration is increased and similarly, if you look at 1 minus f of e, that is this
gap, since the fermi level has now moved down.
14
205
Since the fermi level is moved down and it is brought f of e along with it. This value of 1
minus f(E ) has now increased which means the probability of finding a hole in the states
below E f has now gone up. And therefore, my whole population has now suddenly shot
up the hole count in the valence band states has now gone up. So, here in case of p type
doping my electron count is going to be less than n i and the hole count is going to be
greater than n i.
And in all cases, 𝑛𝑝 = 𝑛𝑖2 . So, if I know my electron count, I can find my hole count as
n𝑛𝑖2 /𝑛 or if I know my hole count, I can find my electron count as 𝑛𝑖2 /𝑝.
15
206
Lecture – 13
Doping - Continued
Now, what we will do is look at the temperature dependence of doping ok. So, first let us
take a condition, where we have added these let us take an N type semiconductor again
ok. We have added these donor dopants right. So, we have populated the semiconductor
with ND per unit volume donor dopants. And it is all created this little bunch of states at
an energy level E d, which we had found that was you know very close to the conduction
band edge, it was about 0.1 electron volts or less close to the conduction band edge. And
all these donor atoms have got five they are all initially neutral, they have they all have
five valence electrons, (Refer time: 01:14) talking about silicon, and they are all
occupying these states.
So, what we have shown here is that these blue circles that you see, these circles are all
electrons. And what this implies by my placing electrons on this, it implies that the fifth
electron is still attached to the donor dopants, so this is still one entire species ok. So, if
you think of the donor as having say a nucleus and it has this many electrons and we will
207
only show the valence electrons, so five valence electrons. So, these four are going to
participate in the bonding with the silicon ok, they are participating in the bonding with
the silicon and this entire thing gather is charged neutral ok.
But, the moment this electron moves away, it is going to leave a positively charged core
behind. So which is this is charge neutral, this is an electron, so there is a negative charge
here, there is a positive charge here, the entire species charge neutral. But then, the
moment you pull that electron away, so let us say you have these four that are
participating in bonding and this is my species. The moment this electron exits, it gets
into the conduction band, you leave behind a positively charged.This entire entity is now
got positively charged, and it is fixed, whereas the negatively charged electron is moved
away.
So, at T equal to 0, what is being shown here is that this electron has still not moved
away, and it is this electron that is being shown in this picture here ok. So it is that
electron we are talking about. By placing an electron in these states, it is this electron
that will eventually get to the conduction band that we are showing. So, all these states
are now charged neutral ok, they have got their fixed ion, but it is still not ionized and
the electron is still with the donor atom. And this happens at T equal to 0 Kelvin, and this
condition is called as freeze out.
So, since there is no thermal energy, and we are at thermal equilibrium, there is no
temperature, there is no thermal energy, the electrons do not have any energy. And
therefore all the electrons are sitting in the valence band. They are not able to get
promoted to the conduction band. And since, it is exactly 0 Kelvin, it is not only the
electrons in the valence band that do not have energy to go to the conduction band, it is
also the electrons present at E d that is with these donors that are not able to escape the
clutch of the donor atoms. So, it is all bound together, and it is all sitting there.
So, the number of electrons in the conduction band is 0. So, if you look at the N count,
which is the electrons at E>Ec, which is what I am interested in for calculating the
currents is 0 ok, so the electron concentration is here. What is the whole concentration in
the valence band? That is also 0; the P is also 0, because there are no vacancies here, all
the states are filled. So, this is the temperature; this is the condition at T equal to 0, and it
208
is something called as freeze out. So, I think this is a technical term, it is called as freeze
out condition. So, doping really does not help you, if your T is equal to 0.
Now, let us say we start increasing the temperature. We will go from this diagram, we
will move over to this diagram here, we start increasing the temperature a little. So, let us
say we pull up the temperature, we cross 10 Kelvin and then head to 50 Kelvin, but we
are still well below 300 Kelvin, and (Refer Time: 05:13) still not reach to room
temperature ok, so what happens. You will find that, since the temperature is now
increased, the electrons do have some thermal energy.
Now, since this energy gap is much smaller than that energy gap, you will find that many
of the electrons that were here have all now moved up. They have taken this little
thermal energy, they have and that is sufficient for them to cross this energy gap. And
therefore they have all populated the conduction band, which means this condition.
These donor atoms, they moved from this state to this state at a temperature T greater
than 0. And this is at T equal to 0 ok. These electrons have now gotten enough energy to
jump into the conduction band. It is these electrons that we are talking about.
But then, is also statistics right, there are still many electrons that did not acquire that
energy. T is greater than 0 Kelvin, a certain population. So let us say this is the energy
distribution for the electrons, so let us say that is the energy, and that is the number of
electrons ok. So, let me draw it let me draw it better. So, let us say it is we are looking at
a tail like this, and this is the threshold. So, it is only these electrons that can jump up
from here to there.
209
So, you still do have some donors that are not ionized, because those electrons did not
statistically collect enough energy to cross this point one electron volt barrier. And
therefore, you do have some of the donor atoms still in this condition with the electrons
still bound to the donor. Whereas, many of the donor atoms are now ionized, they have
left behind positive course. I should have shown the positive core here. They left behind
this positive positive core and promoted the electron into the conduction band. So, these
there are many several of them, where the electron has escaped the clutch of the donor
atom ok, so that is the case at T equal to some temperature greater than 0 Kelvin, but less
than 300.
And since, the temperature is still much less than 300, you do have quite a few of these
donors that are ionized, but then most of the electrons in the valence band are still stuck
to the valence band. You might find one or two that could statistically collect enough
energy to get pass the energy gap, but this gap is you know almost ten times that gap,
and therefore that count is going to be very few. So, we have shown you know say three
of four of them escaping this and maybe just one escaping the valence band. So, but the
numbers are going to be much more different ok, so that is what happens when you start
to increase the temperature slowly. Some of the donors get ionized, and some of them
contribute to the electrons.
210
So, what we should do to correctly show these pictures, we will draw static positive
charge at the locations at the donor sites, where the electrons have left the donor,
whereas these are all still charge neutral these are all charge neutral sites. So, this is the
correct picture. And here we have a vacancy, we have got a hole that is left behind, so
that is going to contribute to your count P.
211
So, as we increase the temperature further, so we let us start approaching 300 Kelvin,
you know we hit we hit say 273 Kelvin, and then we start increasing it, and we start
approaching 290 Kelvin, and we are in and around 300.
So, in and around 300, we have a large number of electrons that have escaped the
valence band, and left behind holes, and these electrons have all migrated to the
conduction band. The temperatures now large enough, but this temperature is much more
than enough for the electrons in these donor levels to escape the clutch of the donor. And
at these temperatures you should find that almost all in fact, all the donors would have
ionized, if the doping is done properly. All the donors would have ionized, they would
all left behind positive ion cores there. And all these electrons would have escaped. You
will find now a lot more electrons from the conduction from the valence band in the
conduction band, and all the electrons from these donor sites now in the conduction
band.
So, at this temperature, all donors ionized; and the semiconductor is now called extrinsic
ok. So, when we say extrinsic semiconductor, it means that as opposed to an intrinsic
semiconductor. That means, it is a pure semiconductor without any doping, as opposed
to a pure semiconductor. The electron population for the case of N type doping, the
electron population is because it is largely, because of the donor contribution, that is all
the donors are ionized, and that has that is what is contributing largely to the electron
6
212
count, and not so much the electrons that have come out from the valence band ok, so
that is what you mean by an extrinsic semiconductor. So, an extrinsic semiconductors
typically a doped semiconductor with all dopants ionized ok, so that is what is happening
at T equal to 300 Kelvin.
Now, let us say we continue increasing the temperature further, you know we push it up,
we go from 300 to 400, and maybe even greater than 400 to 500 Kelvin. So, now the
temperature is so large, that these donors which were initially ionized, have nothing more
to do, there were no more electrons to give up, so they are not going to contribute
anything more beyond your 300 Kelvin. These were they already gave up everything all
the electrons they had at 300 ok, this is the maximum they could do.
But, now the temperature is so large, that the electrons that were sitting in the valence
band, can now significant number can migrate to the conduction band. So, in this
condition, it is the electrons from the silicon lattice that is the valence band electrons that
are contributing most of the electrons to the n count that is the electrons in the
conduction band. The donors have already contributed everything they had, and their
contribution now is minimal.
So, just to give you some numbers ok. So, let us say we put donors of 1 e 15 per cc. The
intrinsic carrier concentration is about 1 e 10 per cc. And the number of silicon atoms in
a silicon lattice is order of 1021 per cc, I might be an order more or more or less, but so
the order of 1021 per cc. So, this is about five times, five orders greater than intrinsic
carrier count, and that is about six orders greater than the donor count donor
concentration.
So, at T equal to 0 Kelvin; N was 0; P was 0; nobody contributed anything. At T equal to

300 Kelvin in a pure semiconductor intrinsic semiconductor, the n count would have
been equal to n i, which is 1 e 10 per cc. But, since we have doped at 1 e 15, the N count
is going to be closer to 1 e 15 that is you know five orders of magnitude greater than
what an intrinsic carrier count would have been. But, now as we keep increasing the
temperature, almost all the silicon atoms that is you know we are talking about 1 e 21 or
122 silicon atoms (Refer Time: 14:11) going to start giving up their electrons. So, let us
say we have 1017 silicon atoms that have given up their electrons ok, so which means that
213
the contribution of the silicon lattice to n is 1017 , and the contribution of the donors has
not changed after 1 e 15, it had only 1015 to give; it is 1 e 15.
So, at lower temperatures, the donor was contributing more ok. At lower temperatures if
you look at the electron count, the donors were giving the bulk of the count towards n,
and the valence band was giving very little. But, at temperatures that are much larger, all
the donors have already given everything they had to give. So, the donor count has not
changed; it is the same ok, it is the same. But, the valence band count is now suddenly
increased, it is become much much larger; and this is the one that is contributing more.
So, it is as though the silicon is behaving like an intrinsic material, because it is the
electrons that are coming into the conduction band now are coming in from the valence
band. And therefore we call this operation at T much much greater than 300 Kelvin as
intrinsic operation. It does not mean of the semiconductors intrinsic; this operation is
intrinsic; I should call it intrinsic operation ok, intrinsic like operation. The
semiconductor is definitely doped; it is not a pure semiconductor. But, if you look at the
electron count in the conduction band, it comes largely from E v, because the donors
have already given up everything they had ok.
And therefore, we call it as intrinsic operation.
214
So, here what we plot is the electron concentration in the conduction band divided by the
donor dopant concentration. And this is to gauge, how this parameter or how this ratio
varies with the temperature. So, as we saw the electron concentration, so at very low
temperature. So, let us call this as 0 Kelvin ok; we do not really approach 0 Kelvin, so
we start from a very low temperature. At this very low temperature, you have the
conduction band edge, you have the valence band edge, and you have your donor levels.
And all these donor levels have the electrons bound with the donor atoms. The
temperature is too low for these donor atoms to ionize.
Now, as the temperature begins to increase, the temperature is still very low for the
electron in the valence band to get to the conduction band ok, so then the electron in the
valence band does not make that jump. But, the since the donor level is located so close
to the conduction band, the donor electrons can easily make this jump ok. So, we are we
are only talking about an n-type semiconductor here, for in this particular plot, and we
have assumed that we have doped at with the donor dopant.
So, now as a temperature begins to increase a little, you find that these electrons can
easily make this jump, and the electron concentration above E c. So, this is the electron
concentration per unit volume at E>E c begins to increase. Now, as the temperature
increases further, both sites start contributing electrons to the conduction band. So, you
have the total number of electrons per unit volume in the conduction band to be n, and it
is receiving electrons from the valence band as well as electrons from the donor sites
from the donor’s donor dopants. So, at T equal to 0, both these were 0. But, as the
temperature begins to increase, since the Ec-Ed is very low, this contribution was
significant. And you find that in all, this region the donor contribution is important.
Now, till a certain point you know since the electron the concentration is very low, we
called this as the freeze out region. In fact, freeze out does technically when everything is
0, but we will call this entire region as freeze out. Now, the donors keep contributing. So,
more and more donors keep getting ionized. So, let us say we had implanted 1 e 15 donor
atoms per cc donor dopants per cc. I would say this temperature let us say 100 donor
dopants contributed were ionized, and therefore they contribute 100 electrons to the
conduction band.
215
And as the temperature increases, (Refer Time: 19:59) this contribution significantly
start increasing. So, let us say here 110 donors contributed, and they will reach a
temperature, where all 115 donors will contribute. So, the temperature here is large
enough, that all 115 donors have ionized, and all these electrons have moved to the
conduction band leaving behind positively charged donor course ok. So, these are
immobile positively charged centers. And all these electrons have now populated the
conduction band.
Now, since the temperature is large enough at this point, you will now start seeing a
significant contribution from the valence band. So, there will be holes left behind, and
there will be some electrons that you know sort of migrated from the valence band to the
conduction band ok. So, this as the temperature increases, this contribution will increase,
but this contribution will increase dramatically, it will be much much larger.
And beyond a certain point, all the donors would have finished contributing, there are no
more donors per unit volume. So, there all the 1 e 15 donors have contributed the 1 e 15
electrons, and the electron count in the conduction band does not increase any further.
And at this point, since the n so, basically what happens at this temperature is all the N D
donors, they have all ionized, they have all become N D plus, and I have contributed an
electron. And therefore, all the n, matches the donor concentration ok.
So, if the intrinsic carrier concentration was say 1 e 10 that is without doping and the
donor concentration is 1 e 15, this n is now going to match N D, because the contribution
from the valence band states at say T equal to 300 Kelvin is 1 e 10 ok. So, 1 e 10 of these
came from valence band states, 1 e 15 came from the donor states, and the summation is
close to 1 e 15. Now, in a 1 e 10 is five orders of magnitude less than the donor
contribution. And therefore, n is approximately ND. So, this is the perfect extrinsic
operation. So, we are in the extrinsic operation region, and therefore n by ND is equal to
1.
And this curve remains flat for a long time, because as the temperature increases, the
valence band contribution is increasing. The donor contribution is still it is fixed. The
donors have all given up everything they have had ok, so it is always 1 e 15. But, the
valence band contribution is increasing, it goes to 1 e 11 per cc, but the total count is still
1 e 15. Then the valence band contribution will increase to say 1 e 12 per cc, but you
10
216
know it is still thousand times less than the donor contribution. So, the total count is still
approximately 115 per cc.
So, this continues for a while, and the curve remains flat, it is completely in this region, it
is completely dictated by the donor contribution, and this is all the extrinsic region. The
valence band contribution is increasing slowly, but it is still not large enough.
Now, there will be a temperature, where the valence band contribution now starts to
match the donor contributions. So, let us say at this temperature, the donor contribution
was 1 e 15 per cc, the valence band contribution will match this contribution. And you
will start seeing the n by N D starting to get greater than 1. So, the n has increased
beyond N D, and the n by N D curve will begin to climb up ok. So, beyond this point, the
valence band contribution becomes greater. So, this is the donor contribution is still 1 e
15 per cc, there are no more donors ok, they have given up everything.
But, now the valence and contribution begins to increase dramatically. So, there are
many more. So, this is a very badly drawn arrow. So, what I will draw is, the valence
band contribution is now going to be much larger than the donor contribution. So, if the
donor contribution is 1 e 15 per cc, the valence band into 1 e 16. And at higher
temperatures, (Refer Time: 24:58) the donor contribution is still 1 e 15, the valence band
will do 1 e 17. And now the count begins to increase. Now, you cannot ignore, you can
11
217
now begin to ignore the donor contribution. So, this is going to be about 1 e 16 per cc,
this is about 1 e 17 per cc and so on. And therefore, n begins to climb once more.
And in this region, since it is the valence band that is contributing to the carrier the
electron count in the conduction above the conduction band, this region is called as the
intrinsic region, intrinsic like operation ok. It is not that the semiconductors intrinsic as I
said, it is the valence band that is contributing to the electron count as opposed to the
dopant ok. And therefore it is like an intrinsic semiconductor.
So, if he had no dopant at all, you know if he had let us say for example, the
semiconductor was not doped at all, what would the electron count looked like ok, so n
by N D does not make any sense. So, let us say we were to plot n, the general shape of n
would so, in that case, we will have to redefine the y-axis, let us just do it for the sake of
exercise.
We will redefine the y-axis in that case to n, and if we plot n versus temperature in the
case of a dopant, this will now become N D. But, in the case of the intrinsic
semiconductor, this would have remained very low. And at a certain point, this would
have begun to increase. I should have crossed the 1 e 10 per cc mark for silicon here, and
it (Refer Time: 26:54) continued to increase and would sort of asymptotically merge with
this curve. So, this would this is what a purely intrinsic semiconductor would have done
12
218
ok. And therefore, in this region, it is as to the doped semiconductor behaves like an
intrinsic semiconductor all right, so that is the temperature dependence of doping.
Now, what we do is, we want to now make a more careful count about a mac more
careful count of the number of ionized donors, and an more exact count of the electron
and hole population ok. And in order to do that, we will start with the charge neutrality
relation. So, you know as we discussed the silicon was initially charged neutral ok. You
have silicon atoms, some of the silicon atoms have given up an electron, and they have
left behind a hole ok, these electrons are left behind a hole. So you do have negative
freely moving negative charge, and you have freely moving effective positive charge
species, which are holes. And therefore, the entire system is neutral, it is not charged.
And what we did now is we added a dopant, if for example, if we added a donor dopant
as we discussed, you have all these silicon atoms that are neutral, we have some free
electrons, we have some holes, and we also have about N D per unit volume of donors,
and if these donors are all ionized. Let us say some of these donors are ionized, these
donors would get a positive charge, and they would contribute an electron.
So, you see this is charge neutral, this donor is not yet ionized, so it is charge neutral, and
the entire system is charged neutral right. So, there is there is no charging anyway.
Therefore, we can write a charge balance equation. All the positive charges must equaled
13
219
all the negative charges. And we will take a very general semiconductor, where and we
have added some donors, we have added also counted up (Refer Time: 29:31) with some
acceptors and so on.
So, all the positive charges are basically the holes that is the mobile positive charge, and
the ionized fixed donor charge, not all the donors, only the ionized donors. So, these are
the donors, this is the donor concentration that has given up electrons. So, for example, if
the semiconductor is not an extrinsic operation, and if the temperature is not large
enough, out of N D; let us say alpha donors alpha N D remains as unionized ok, some
portion is unionized, and the remaining have ionized.
So, this is the charge neutral neutrality condition for the donors. So, this is a small
fraction alpha is less than 1. This is the fraction that has not ionized; this is the fraction
that has ionized. So, when we talk about N D plus we are talking about this fraction. So,
P+ N D+ plus is equal to n that is taken into account, all these electrons plus any other
existing free electrons in the semiconductor is equal to n, which is going to have negative
charge plus the ionized acceptors ok. Just like how we talked about donor ionization, the
acceptors will also have to get ionized, which means that an a neutral acceptor specie
will take up an electron to complete the bonding.
So, let us say it takes up an electron to say let me write this go one step back. An
acceptor will ionize by taking up an electron and giving up a hole, which is free to move
about in the lattice. And let us say you know a fraction beta of these acceptors are not
ionized, and 1 minus beta of these are ionized, and they gave up 1 minus beta holes. So,
just for the sake of argument. So, we are talking about this fraction, when we write N A
minus.
So, these are the ionized acceptor like species, and ionized donor like species that are
charged. And the ionized donors have got positive charge, because they have given up an
electron in the lattice, and the ionized acceptors have got negative charge. And therefore
you have this charge balance relation. So, now the question is at some temperature, what
+
is the ionized donor concentration, you know how much of it is N D at any given
temperature. Can we calculate it? Can we calculated with whatever information we have.
14
220
So, if you think about let us draw this energy level let us draw E c, E v, and E d all right,
so that is where all the donors are sitting. Now, these have electrons, and let us say these
are the ionized species, they have given up electrons. Now, the total concentration per
unit volume is N D that is how much we added out of which a fraction of it has ionized,
so that is N D plus, and that is what is needed for the charge balance.
So, now the question is what is N D+ , you know how much what percentage of N D plus
I mean what percentage of N D is the N D+, the ionized donors. So, instead what we will
do is, we will try to calculate this ratio, you want to make a count, you want to be able to
calculate it. So, we want to specifically calculate how much is N D+ by N D that is what is
the ratio of the ionized donors to the total donor concentration. Now, N D+ by N D if you
think about this diagram ok, these donors are all sitting at some energy level E d.
Now, if you ask the question, what is the probability that an electron occupies these
energy levels E d. Now, you have got states inside the gap, and these states are located at
E d. And how many states do we have. Since we have you know about N D donors per
unit volume, the state sort of scale the number of state scales with this donor
concentration.
So, if you ask yourself the question, as to how many electrons occupy the energy level
the states at energy level E d, the answer is that it is the number of states available. So,
15
221
we will say there are N D states available per unit volume into the probability that an
electron occupies the state, which is nothing but the Fermi function at E is equal to E d.
So, this is the number of electrons occupying number of electrons per unit volume
occupying the states at E d ok, which is effectively these states, you see electrons are
occupying these states.
We are interested in the states that are empty and therefore, charged ok. We are
interested in those states that are positively charged, that is your N D plus. So, we are
interested in the probability that an electron does not occupy the state at E d, which
means it is the probability that a hole occupies the state at energy level E d, that is your
N D+ / N D.
So, what is the probability that hole occupies states at energy level E d, it is 1 minus f of
E, it is 1 minus f of E equal to E d, so that is my f at E equal to E d. And 1 minus that is
essentially this term, we will divide by that factor there, and you end up with this
1
particular ratio. Therefore, my N D+ is equal to this term here, which is −(𝐸𝑑 −𝐸𝐹 ) 𝑁𝐷 , so
1+𝑒 𝑘𝑇
that is my number of ionized donors. So, it is this fraction. So, when we earlier
mentioned 𝛼 alpha and 1- 𝛼, so this ratio 1 by 1 plus e to the power minus E d minus E f
by k T is your 1 minus alpha (Refer Time: 37:05).
So, although this is the number of ionized donors, we need to be a little careful ok, we
need have to make a small correction ok, and that correction is something called as the
donor degeneracy factor. And this degeneracy factor has got has got little relation to the
fact that whether it is a degenerate semiconductor or not. It need not it need not be a
degenerate semiconductor ok. We are not interested in degenerate semiconductors in this
course I mean de degenerately doped semiconductors. We are only talking about E c
minus E f being greater than 3 k T.
But, despite that, while making a calculation of ND+/ ND, we need to attach a correction
factor to this term ok. Why do we need that correction factor, we need that correction
factor, because if I were to use this argument for any energy level let us say that is some
energy level located inside your inside your lattice, I mean inside this energy gap
because of some defect ok.
16
222
Now, two electrons could have occupied that energy level ok; you could have plus half
spin and minus half spin. But, since this is a donor, and the donor allowed only one
electron occupancy, and it gave up that one electron occupancy. We have to have a
correction factor of 2. And therefore, we add this correction factor by taking this answer,
which is come out from pure logic ok, which is the ratio of N D plus by N D is nothing
but the probability that the hole occupies the states at energy level E d, which is equal to
this term, which is a logically good argument. But, we attach this little correction factor
called g d, and you sort of insert that g d in there ok, and that g d takes a value of 2.
And (Refer Time: 39:19) point is that as this as if you have a trap state, that is sort of
located. If you have this these states instead of being located so close to the conduction
band edge, if they were to be locate the little closer to mid gap, you do not have to worry
too much about g d, you can take a value of 1, and you are quite safe with that value ok,
so that is the count of the ND+.
Now, what about NA- So, let us say we have counter dope to the semiconductor, we
trying to keep things very general right. You have a semiconductor, you have added
some N D out of which some of them have ionized, and you have also added some N A
out of which a few of them have ionized. So, what is this NA- count? So, let us make a
logical argument there once again. So, if you have NA-, they are going to sit close to the
valence band edge, you know just like how we made in we made a calculation about how
17
223
E d is very close to the conduction band edge. The acceptor level energy levels are also
very close to the valence band edge, so they too will be about say 0.1 electron (Refer
Time: 40:36) ok. So, you have these acceptors are also sitting very close to the valence
band edge
And these acceptors, when they are charged neutral, they are actually empty, they only
had three electrons the valence shell. And by and they get ionized by taking an electron
from the lattice. So, by taking an electron from the lattice, these acceptors are now
ionized, and they have a negative charge, whereas these are all still charged neutral. So,
it is these acceptors that are N A minus ok. And there are a total of N A acceptors per
unit volume.
So, we now ask the question what is this ratio NA-/ NA. So, going by the same argument,
NA-/ NA is the probability that an electron occupies these states at energy level E a,
where E a is the acceptor energy level. And that is given by your Fermi function that is
1
your 𝐸𝐴 −𝐸𝐹 . But, just like in the case of donors, we have a correction factor called the
1+𝑒 𝑘𝑇
acceptor degeneracy factor ok, and that is given by the symbol g a.
Now, logically one might say that g a has to be 2. But, since there are two E-k diagrams
for holes, you have as you know these holes have got two different effective masses, and
they are called light holes and heavy holes. We use a factor of two times (Refer Time:
42:22) which is g a equal to 4. And we have this to be the ratio of NA- to NA. So, we have
now identified what is ND+, and what is NA- for any given temperature T ok. Now, as T
keeps increasing, your ND+ will eventually merge with ND, and NA- 0will eventually
merge with NA ok. But, for any arbitrary T, this is the ratio.
18
224
So, now we are all set to write a complete charge balance equation. So, what is p, p is
𝐸𝑉 −𝐸𝐹
equal to your 𝑁𝑣 𝑒 𝑘𝑇 , we have already discussed that. N D plus is your N D plus 1 by
g d e to the power minus E d minus E f by k T, where g d is the correction factor. N is
nothing but N c e to the power minus E c minus E f by k T; and N A minus is this term.
So, what we have done here is just bring all the terms to the left hand side, and say that
the left hand side is equal to 0. So, you have the positive charges and you have these
negative charges, they must balance out each other to made a charge neutrality.
Now in the case of extrinsic operation that is we are talking about this region of the
curve, we have gone past freeze out, and we are an extrinsic operates operation, and we
have not yet reached intrinsic operation. So, we are talking about this region, where all
the donors have ionized and all the acceptors are ionized. Now, in this case, your ND plus
is equal to ND, because this term has become 1. The temperature is so high, the
temperature is so large that this term is negligible, and it is mostly 1, and here to the
temperature so large that this term is negligible, and it is mostly 1.
So, in extrinsic operation, all the donors have ionized. So, there is no ratio anymore ok.
So, for example, sorry: so if I were to draw the conduction band edge and the valence
bandage, and let us locate all the donors levels. Initially all the electrons were here
during freeze out, then as a temperature increase, some of these electrons moved up and
19
225
there were some ionized donors that was my N D plus, and that was my N D, and that
was the ratio that was N D plus as compared to N D.
But, as temperature increases, all the donors get ionized. And N D plus approaches N D,
it becomes equal to N D. So, ND+ plus is equal to N D, N A- minus equals to N A, and the
charge neutrality relation instead of having N D + N A, you might as well just retain N D
and N A ok. So, the charge neutrality relation or the charge balance becomes this.
Now, if you use this condition, that n p is equal to n i square, and we are operating an
extrinsic region operation that is N D plus is equal to N A N D and N A minus is equal to
N A. Then this relation can be rewritten as this, because I am now representing my p as n
i square by n. So, now we can establish a quadratic equation with n, we can solve this
quadratic equation, and you will end up with the exact electron concentration per unit
volume to be equal to this term, and the exact hole concentration per unit volume to be
equal to this term ok. So, you just solve this quadratic equation, and I will end up with
these two terms.
So, what assumptions have you made here, we have made the assumption the first
assumption is all donors have ionized, all acceptors have ionized, cannot happen at low
temperatures. This has to be at significantly high temperatures ok, need not be too high,
room temperature sufficiently is a sufficiently good case for this assumption. The second
20
226
assumption is this mass action law holds. And this will hold, when you are in thermal
equilibrium, and in the case of non degenerate doping ok. So, this equation will hold. So,
under these assumptions, you can say that this is my electron concentration in the
conduction band, this is my hole concentration in the valence band.
Now, we can make one more assumption if these concentrations are quite large, which is
if in the case of n-type semiconductors, your effective donor doping is so large that it
goes much greater than n i. So, n i is in the order of the case of silicon you know it is the
order of 1 e 10 per cc. If let us say we are doing donor doping, where N A is 0, no
acceptor ions, and N D is say 1 e 15, this is a reasonably good case to assume that N D
minus N A, which is basically your N D is much greater than 1 e 10 ok.
So, in that case, if you look at this quadratic equation, if you just go back, if you look at
these terms. So, let us say let us just take this particular equation there, let us say N A
was 0 or essentially N D let us just say ND - NA is very very large. So, N D minus N A is
very large, which means that if you look at these two terms ND - NA is much larger than n
i square. Therefore in this term in the square root the n i square is negligible ok. We
ignore that term. And you end up with n is equal to (ND - NA/2 +√ND − NA/2)2, which
is basically again ND - NA/2, which is equal to your ND - NA, so that is the assumption we
are making here.
21
227
If an n-type semiconductors if N D minus N A is much greater than n i, then n is
approximately ND - NA. And p can be calculated from the mass action law, n p is equal to
n i square, n is equal to ND - NA, and therefore p is equal to 𝑛𝑖2 /𝑛 n i square by n, which
𝑛𝑖2
is . And if you are if there are only one type of doping, so you know there is no
𝑁𝐷− 𝑁𝐴
counter doping that is your N A is 0. Then it is simply that n is equal to N D, p is equal to

𝑛𝑖2
. And equivalently for holes, if you have accept like doping, if p is approximately NA,
𝑁𝐷
𝑛𝑖2
n is approximately . . So, these are very useful relations for very quick calculations.
𝑁𝐴
And this will be very useful for you during these during the assignments or tests in this
course ok.
So, with that we will conclude doping. And we will start a new topic, which has got to do
with the mechanism by which carriers are generated. The mechanism by which
temperature influences care in the free electron and hole generation by thermal
excitation.
22
228
Lecture – 14
Recombination and Generation
We will now look at a very important concept with regards to semiconductors which is
the mechanism of recombination and generation of carriers. So, far we have been able to
establish a carrier count by while considering the semiconductor to be an thermal
equilibrium. So, we kept the semiconductor in the dark where the electrons had only the
ambient temperatures the source of energy. And we could calculate the density of states
in the semiconductor we established something called as the Fermi function which gave
us the probability that an electron occupy these states. And these two quantities we could
measure the number of electrons and the number of holes in both an intrinsic
semiconductor as well as in a doped semiconductor.
Now, what we are interested in this topic is to look at what happens when the
semiconductor is taken out of equilibrium. You know what is the dynamics of the change
in carrier count when the semiconductor shifted out of equilibrium. And even in
equilibrium what are the mechanics or the mechanisms of you know of generation of
carriers, and the recombination of carriers.
229
So, firstly, what do we mean by generation and recombination? So, by definition
generation implies the mechanism by which electron hole pairs are created. So, you have
a neutral specie and I you know it gets excited maybe through a photon, and you create
an electron hole pair. And recombination is the process by which electron hole pairs are
and highlighted. So, which means the electron and the hole they sort of meet each other,
and they recombine. And they basically the electron essentially occupies the vacancy and
there is no more free l that is no longer a free electron and a hole. So, we are you know
as we mentioned before we are interested in the electrons in the conduction band and the
holes in the valence band.
So, generation essentially increases n and p because you have increased you have created
an additional electron and a hole in the conduction and valence bands respectively.
Whereas, recombination lowers and n p because you are now taking carriers that were
free, and now you are recombining them to form a neutral species, and therefore loop,
the therefore, you are losing then you are in fact, dropping down in the carrier count. So,
let us look at some of the basic mechanisms of carrier generation and recombination.
The first is something called as band to band generation and recombination. So, what
does this mean? Band to band simply implies that you have your valence band, you have
electrons here. So, I represent the electrons by these blue circles and I represent holes by
these hollow circles. So, you have electrons here which are bound to the silicon atom.
230
And you have these electrons get some energy because of either thermal energy or
perhaps a photon you know interacted with these electrons. And this electron is now
excited and has enough energy to get into the conduction band, so that is something
called as the band to band generation. So, you have created an electron in the conduction
band and this electron has left behind a hole in the valence band.
Now, what is band to band recombination? It is the exactly opposite process. So, you
have an electron in the conduction band and this electron most free I mean it was it was
contributing to any current. But now due to some reason this electron has now dropped
its lost energy and it has dropped down into the valence band and it has recombined with
this hole. And this process this loss of energy typically results in some radiation. So,
probably this electro might release a photon when it loses this energy. So, this
mechanism is called as a recombination, and it is the band to band recombination. So, it
is directly from the conduction band to the valence band.
So, the next mechanism is something called as R-G center based generation and
recombination. So, this mechanism is quite important from the point of view of this
topic. And we will look at this mechanism in a bit of detail, but what this essentially
states is that you have certain trap states located inside your band gap. So, these trap
states we already looked at creating traps by doping a semiconductor, and those traps
were you know the donor level traps or the acceptor level traps in the conduct created
231
very close to the bandages. But one could in principle create trap states deeper inside the
gap by probably doping it at some other specie or by or by just having defects in the
semiconductor.
So, these trap states E it as a stepping stone or it is like a ladder for the recombination
and generation processes. So, therefore, an electron in the valence band can first climb
on to this state, can first occupy this state and then get excited into the conduction band.
So, this may this two step process need not happen at the same time. And this is very
strong because the energy needed for this jump is much smaller than the energy needed
for a direct band to band generation. Therefore, this could lead to a significant increase in
carrier count.
The R-G center based recombination mechanism is the exact opposite. So, you have an
electron in the conduction band. And this electron falls into this trap state releases a
small amount of energy, and then it drops down into the valence band or it could simply
recombine at the location of the trap state. So, these are the generation and
recombination mechanisms involving a R-G centre. So, R-G centres are simply
recombination generation centres or we can also call them as traps. So, and we these
traps are all sitting at this energy level E T.
The next mechanism of recombination generation is something is a process that uses an

entity called as an exciton. So, we say generation via excitons; and generation and
232
recombination via excitons. So, what does the generation via excitons mean? So, an
exciton is an entity there in you do have an electron that is free you do have a whole, but
this electron and hole are sort of coupled to each other. So, you can think of it as a you
know in the very hand wavy manner you can think of it as a hydrogen atom all right. So,
you have this electron and hole that are bound to each other. So, they are not completely
free from each other that is it, but this species such can actually diffuse through the
medium it can actually move through the semiconductor.
So, excitons are quite common in materials that have got a very low relative permittivity
for example, say organic semiconductors. So, excitons are quite common and they are
not so common in probably materials that have got very high permittivity except under
certain circumstances. So, you have different kinds of excitons you have the Frenkel
exciton you have, the Mott exciton and you have something called as a charge transfect
exciton. And what is common in organic semiconductors are due to their low relative
permittivity is something called as the Frenkel exciton. It is a, very tight, it is a very
tightly bound electron hole pair with a very small radii.
So, the generation via exciton so simply a mechanism wherein the electrons in the
valence band where excited because of some energy they received. And they did become
free from the clutch of the atoms in the lattice, but then they are still bound to the whole,
and this species called as the exciton. And this exciton can be sort of energized or you
know it can be it can lead to a pure electron hole pair separation probably due to this due
to in the presence of an electric field.
The other mechanism is recombination via excitons, wherein you have a free electron
hole pair and this electron hole pair they recombine, but do not recombine fully they
instead form an exciton state. So, you they are no longer a completely free electron hole
pair, they instead form an exciton. So, the exciton does not have any charge. So,
although it can move it does not have any charge. So, therefore, this is an effective loss
in terms of charge carriers.
233
The next mechanism of generation and recombination is via something called is impact
ionization. So, what is the impact ionization means you have a very fast moving electron
and this electron collides with the atom in the lattice and that collision is so powerful that
the atom gives up an electron and it not electron leaves behind a hole. So, here you have
a high velocity electron, it is got a high velocity probably because of the presence of a
very high electric field. And this high velocity electron loses energy after collision and
that collision leads to the formation of a new electron and a whole pair.
Now, the exact opposite process is responsible for recombination. So, s, this is something
called as Auger recombination. And what happens here is you have 2 electrons in the
conduction band, and let us say they have a momentum exchange. So, they collide, and
one of the electrons gains energy the other loses energy and recombines. So, one of the
electrons gains energy it goes up to higher energy levels, whereas the other one loses
energy and falls down to the valence band and recombines with the hole. So, this is
something called as the Auger recombination process.
234
Now, so these are the you know I would say the basic generation recombination
mechanisms. Now, let us consider band to band generation recombination. So, band to
band implies you need to have a direct transition of an electron from the valence band
into the conduction band thereby leaving a hole behind. Now, the efficiency of a band to
band generation mechanism strongly depends upon the kind of semiconductor used.
And what we specifically mean by that is if we look at the E k diagram of the

semiconductors what will what is the E k diagram, it is the energy versus wave vector
diagram. Essentially the E k diagram is also energy momentum diagram because k is the
wave vector which is 2𝜋/𝜆. And h bar k is nothing but h/2𝜋 *2𝜋/𝜆 which is h/𝜆 which is
the De-Broglie equation for momentum. So, p essentially relates to k.
So, what we are looking at is an energy momentum diagram. And you can have two
kinds of semiconductors. So, you could have a semiconductor where if this is the
conduction band edge and that is the valence band edge, the maximum of the valence
band and the minima of the conduction band lie at the same value of k. So, they both lie
at the same value of k, and these semiconductors are called as direct band gap
semiconductors, you could also have a semiconductor wherein the maxima of the
valence band and the minima of the conduction band lie at different values of k. And
these are called as indirect band gap semiconductors. Now, how does a semiconductor
being direct or indirect effect the generation process?
235
So, in the case of a direct bandgap semiconductor for an electron sitting in the valence
band to be excited into the conduction band via band to band generation, all that is
needed is an increase in energy. If the energy of the electron increases, you have you
have a generation process. On the other hand, if I have an indirect bandgap
semiconductor for the electron to go from the valence band to the state in the conduction
band, we not only need an increase in energy, but we also need a change in momentum.
So, it is both E and k have to change and the mechanisms or the species involved in
providing a change in energy and change in momentum are quite different. So, we need
to have both the photon and phonon interaction for in an indirect band gap
semiconductor for a generation process, whereas for a direct band gap semiconductor we
need to have only a photon interaction. So, why do we say photon and phonon.
236
So, let us look at these two species. So, if you take a typical lattice. So, let us say you
have all these atoms in the lattice and the typical spacing between these atoms is given
by d which is a few angstroms and which for the sake of argument we will just say 5
angstroms. Now, let us take a photon. So, let us consider a full photon that is that is got
an energy sort of less than the visible band. So, let us say lambda is 10 power 4
angstrom.
So, what is the energy of the photon? The energy the photon is head c by lambda where c
is the velocity of light. So, let us say c is the order of 3 into 10 power 8 meters per
second, and your 𝜆 is 104 angstrom. And for this particular wavelength, the energy turns
out to be about 1 electron would. And considering the fact that silicon has got a band gap
above band gap of about 1.1 electron volt, the energy provided by the photon of this
particular wavelength is of the order of the band gap of silicon.
But what is the momentum in a photon, the momentum in a photon is h by lambda,

which is h/104 angstroms. So, although I have to used to symbol equal to we should just
say that it scales or it is proportional, because k is not the momentum k is the wave
vector. And as we state the momentum is related to k by the Planck’s constant, so that is
that is a little typo there. So, we have we have the photon having a momentum that is h
by 10 power 4 Angstroms. So, the that order but what is the momentum that we are
talking about when we look at this generation process.
237
So, when we want a shift in k what is the order of that, it is the order of h by d where d is
the spacing between the lattice. So, we need something that is equivalent to your lattice
vibration or you know of phonons or momentum. So, therefore, since d is just 5
angstrom as compared to lambda being 10 power 4 angstrom the momentum provided by
a photon is very very small it is extremely small, it is about 1000 times less.
Therefore, a photon can provide good E, it can provide a large amount of energy, but a
very small shift in k, very very small. But let us think about a phonon next a photon is
the unit of your vibration right, it is the energy carrier for vibrations. So, if you think
about the energy of a phonon if you look if you remember the harmonic oscillator the
energy is the order of h omega which is about say 10 to 50 milli electron volts, and that
is the order of the thermal energy that is what is responsible for your lattice vibrations.
So, the energy in a phonon is very very small compared to a photon. The energy in a
phonon is about 1000 times less than the energy in a photon. So, the energy is very
small. But the momentum is of the order of 2 pi binding into h by 2 pi, so that is the
momentum. So, the momentum is quite large. So, this is the momentum of a the
momentum of phonon is of this order, whereas the momentum provided by a photon is
very small. So, therefore, a phonon interaction can give a large shift in k and a small shift
in E, whereas an interaction with the photon can give you a large shift in E, but a small
shift in k.
238
So, going to this picture what this means is that in order for us to take the electron from
the top of the valence band to the conduction band, we need an increase in E which
comes through an interaction with the photon. So, just one interaction with the photon is
sufficient to push this carry out. But here we need interaction with two different species
you need an interaction of the photon to push the carrier up in E, but a photon interaction
cannot give you too much of shift in k; instead we need a interaction with the phonon to
give you a shift in k and interaction of the phonon cannot give you too much of a shift in
E.
So, we need to have both, you need to have an interaction with the photon and a phonon
for us to generate a carrier. And this is a much rarer process as compared to carrier
generation in a direct band gap semiconductor.
Therefore, when we talk about indirect band gap semiconductors, for example, silicon,
the R-G center based recombination generation mechanisms are quite important they are
the ones that are going to create enough carriers. So, you need so that was the we look at
band to band recombination generation and we also looked at R-G center based
recombination generation. It is the R-G center based recombination generation that is
going to lead to a significant change in electron and hole count.
So, what we are going to do from this point on is we are going to get involved with some
amount of calculation and some amount of mathematics that would sort of enable us to
239
understand in a more analytical, and a more quantitative manner the impact of R-G
center based generation and recombination. And in particular, we are trying to calculate
the end goal is to calculate the recombination rate; the rate at which excess carriers
recombine in a semiconductor because that is going to be of use to us further down the
road. So, from this point on we are going involve a bit of mathematics, but we will just
go through all this, and try to understand the key ideas behind the different concepts.
Now, if you look at the electron and hole count, if you look at the total number of free
carriers, so which are n the electrons in the conduction band and the holes in the valence
band. So, what we mean by n is the electron concentration, and the hole concentration is
p. So, if you look at the total count and you know we are now interested in trying to play
around with this count, we are trying to create a dynamics, we are trying to study the
dynamics of the change in n and p via generation and recombination mechanisms.
So, the if you take any little volume in a semiconductor, and you look at the n and p
values in that, how are these n and p values affected. They are affected by any currents
coming in or going out of the semiconductor. And they are affected by any mechanisms
of generation inside the semiconductor or they are affected by recombination
mechanisms in these semiconductor. So, and that could be say other processes depending
on the specific case that affects the population count.
240
So, the dn/dt, the effective total variation of free electrons, the free electron
concentration and the hole concentration is dependent on the rate of change the electron
concentration due to currents due to the R-G mechanisms and due to any other processes.
And similarly for p for the hole count, it is dependent the total the effective variation is
dependent on any currents that are coming in and going out, it depends upon any R-G
mechanisms and any other processes.
And at equilibrium conditions when you say equilibrium what that means is each one of
these mechanisms is balanced out, which means that each one of these derivatives are 0.
So, you have at equilibrium you have no changes due to current, you have no changes
due to R-G mechanisms, you have no changes due to any other process processes and
you have no change in the total effective count.
So, each one of these terms is equal to 0. So, everything is balanced out, internally this is
something called as detailed balance. And when we move out of equilibrium, these
parameters need not be 0. So, when you start throwing light on a semiconductor and start
creating a method of generation or you apply a voltage and start establishing currents
through a semiconductor. These rates need not be 0, they can then shift away from 0 and
we are no longer in equilibrium. Now, it could be that we are not in equilibrium, but we
are in steady state. So, this is the condition for equilibrium.
241
Now, suppose we are in non equilibrium, we have gone out of equilibrium which means
that this is not 0, this need not be 0, all these need not be 0, but we have reached steady
state condition. So, when we say we are not in equilibrium, but we are in steady state
what we imply is that these two parameters the total effective variations are all 0,
whereas, these internal variations need not be 0. So, these are just some fine definitions
that could you know help us understand all the analysis a bit more clear.
So, we could be an equilibrium where each and every one of these processes 0, and we
could be in non equilibrium where these individual processes move out of 0, but we
could be in non-equilibrium steady state at the same time wherein this effective count
does not change with time while although these individual rates are all nonzero. So, there
is balance between the processes.
So, using these ideas, so this is the big picture; so using these ideas, we will now focus
only on the recombination generation rates. We will not worry about any other processes,
we are not worried about currents, we are not worried about anything else that is going
on, we are only interested in the recombination generation mechanisms, so that is going
to be the focus of the next few slides that is just this. So, this is of importance to us.
242
Lecture - 15
Recombination and Generation – Continued
In order to establish the analysis, we will set up a few definitions ok. So, we are
interested in R G center based recombination generation mechanisms. So, the first
definition is let us say that the electron concentration, the free electron concentration is n
and the hole concentration is p and therefore, d n by d t due to R G mechanisms is given
by the time derivative here and the rate of change of the hole concentration due to R G
mechanisms alone is given by the rate term here, the now we have we are only looking at
R G center based recombination generation.
So, let us say at energy level E T we have a we have a certain density of states, density of
trap states and let us say that the total number of trap states per unit volume is given by
capital NT. So, that is a total number of trap states per unit volume that is the trap
concentration. Now, of this capital NT some of the traps are occupied by electrons and
some of the traps are empty. The traps that are occupied by electrons, we will say that it
has got a concentration of small n T ok.
243
So, this, these are traps that are occupied by electrons. It is a number of traps per unit
volume located an E T that are occupied by electrons and the number of empty states or
you can say that the empty traps or traps that are occupied by holes are P T, ok.
So, there are P T traps per unit volume that are empty and since n T and p T have to sum
to capital N T we have a fixed condition here, where in capital N T is a constant while
whereas, n T and p T are more dynamic ok. So, small n T and p T, the small these counts
could change with time ok. So, for example, if you are not in steady state these
concentrations could change in time, but the sum of these two cannot change time capital
N T is fixed ok. So, these are our little definitions and you will use these definitions to
define the dynamics of R G mechanisms.
So, what is the rate of change of electrons due to R G processes? Any change would
depend upon the rate at which electrons are emitted minus the rate at which the electrons
are captured. So, you have, you have your conduction band edge, you have your valence
band edge and you have some trap states here. Some of these trap states are occupied by
electrons and you have some free electrons there.
Now, what is the rate at which electrons are emitted. The rate at which electrons are
emitted, the emission can only happen from the trap states ok. We are not talking about
band to band generation here. So, emission implies generation. So, this by this we
actually implied generation, and use the word emitted because these electrons are going
244
to be generated from the trap states which is equivalent to saying that the electrons are
emitted from the trap states.
So, what is the rate at which the electrons are emitted, they depend upon the number of
electrons present in the trap states. If none of the trap states have any electrons the rate at
which electrons are emitted is 0 because there are no electrons to be emitted ok.
Therefore, it is only when the electrons are present in these trap states that an emission is
possible and therefore, the rate at which electrons are emitted is proportional to n T
because n T is the concentration of trap states that are filled with electrons. And you will
define a proportionality constant called e n which is to say the emission of n type or the
emission of electrons and say that the rate of electron emission is equal to e n into n T.
Now, what is the rate at which electrons are captured? The electron capture involves a
process wherein the electron that is sitting in the conduction band and which was free is
now going to be captured by a trap state ok. So, there is an empty state here and this free
electron is going to drop down and it is going to occupy this empty state. So, that is the
capture of an electron. So, what does that mean, what does that depend upon. Firstly, it
has to depend upon the number of electrons or the concentration of electrons in the
conduction band. If there are no electrons in the conduction band then the rate of electron
capture 0 because there are no electrons to be captured.
So, definitely the rate of electron capture has to be proportional to n. So, let us say there
are electrons to be captured. What else is it proportional to? These electrons can be
captured and for them to capture, they need to find an empty state ok. If there are no
empty states at E T, they the electron cannot be captured because there is no state for the
electron to occupy. So, remember the states are nothing, but solutions to Schrodinger’s
equation.
So, you need to have 2 conditions, you need to have electrons that are there to be
captured and you need to have empty states for these captured electrons to occupy. So, it
is only when you have these 2 together that you can have a rate of electron capture. So,
therefore, the rate of electron captures proportional to n and it is also proportional to the
concentration of empty states and therefore, it is proportional to P T into n where P T is
the concentration of empty states or the number of empty states per unit volume are
located at E T.
245
So, rate of an electron capture can be defined to be equal to C n P T n where this term C
n is a proportionality constant which relates to capture of electrons, ok. So, we have
defined two constants of proportionality e n and c n. So, now, let us talk about the holes.
Just like in the case of electrons, what is the rate of change of the whole population? The
whole population depends upon the rate at which the elect the holes are emitted from the
trap, trap states and the rate at which the holes are captured into the trap states.
Just like in the case of electrons, so, all the empty states are essentially holes right. So,
just like in the case of electrons, the rate of hole emission. So, I am sorry this is a type on
this should be holes, the rate of hole emission, has to be proportional to P T because if
there are no holes in these states (Refer Time: 0:8:12) all the states are filled with
electrons there can be no holes emitted. So, it is only when you have holes located in
these states that you can have an emission of a hole to the valence band.
So, therefore, the rate of hole emission is equal to e p into p T where e p is again another
proportionality constant relating to the emission of holes. And the rate of hole capture by
the same arguments be used for electrons is proportional to n T into p with c p being a
constant proportionality which is the capture of holes. So, we now have 4 rates, the rate
at which electrons are generated and the rates rate at which electrons are captured, the
rate at which holes are emitted and the rate at which holes are captured.
So, based on that we can set up 2 need differential equations ok, one pertaining to the
electrons and one pertaining to the holes and this basically tells you that the rate at which
the electron population changes because of R G mechanisms.
246
So, remember we are not talking about the total. If you go back a few slides, this is the
total change of the electron population in any given volume. We are only talking about
the R G mechanisms. So, the rate at which the electron and hole count changes because
of R G mechanisms is the rate at which electrons are emitted minus rate at which
electrons are captured and the rate at which holes are affected due to R G processes is
due to the rate at which is dependent upon the rate at which holes are emitted minus rate
at which holes are captured. So, this is our starting point for any analysis.
So, at non equilibrium so we are now out of equilibrium because we are having a lot of
dynamics and there are there is generation and recombination mechanisms going on.
247
At non equilibrium we have been using these variables, e n for the proportionality
constant determining the emission of electrons, cn for the capture of electrons, e p for the
emission of holes, cp for the capture of holes, nT which is for the number of occupied trap
states. P T for the number of unoccupied trap states and all per unit volume and for the
free electron concentration and P for the hole concentration.
Now, this is at non equilibrium. Now at equilibrium denote these variables with a special
symbol and we want to say that at equilibrium these values have got a special value
which is given by the subscript O attached to all these parameters. So, at equilibrium all
these parameters take these values and this is a special condition because everything is at
equilibrium.
Now, why do we want to consider equilibrium condition because we can make a lot of
calculations at equilibrium conditions, we can easily calculate electron and hole counts at
equilibrium and we want to take advantage of that to see what happens when we push the
material slightly out of equilibrium and into the non equilibrium condition.
So, at equilibrium you have d n by d t R G has to be equal to 0 because everything is

balanced out. So, which means that if you go back to these 2 equations and you rename
or you know change these variables to their equilibrium values. So, en becomes eno, nT
becomes nTo-cnopTono that is this equation at equilibrium.
248
And this equation becomes epoPTo-CpnToPo. So, if you were to consider the equilibrium
condition we know that these rates have to be 0 and using that we can develop a relation
between e n o and C n o. So, were trying to get rid of this variable e n o we will not just
represent all our equations with Cno and Cpo.
So, for electrons your e n o is nothing but this term here and your e p o is nothing but this
term here. So, this is all at equilibrium.
Now we make a small assumption. We will say that as we move the semiconductor, so,
this is my equilibrium condition. So, I had all these parameters n, p, n T, p T this is my
sorry we had a this is my equilibrium condition where we had the parameters n o, P o, n
T o and P T o and we moved this semiconductors very gently out of equilibrium to the
non equilibrium condition and here my parameters became n p n T and P T.
Now, we will say that, yes these parameters, these 4 parameters are different but the rates
or the proportionality constants are not changing too much. So, we are going to make
their assumption ok. So, we will say that and in the non equilibrium condition, the
proportionality constants that relates these rates have not changed too much, ok.
So, therefore, e n is the same as e n o, C n is the same as C n o, e p is the same as e p o

and c p is a same as c p o, but although the proportionality constants are being assumed
to be the same, we are not assuming that n p n T and p T are not have not changed
249
because a that is a bad assumption because then there is no difference between non
equilibrium and equilibrium condition.
So, yes n, p, n T and p T have changed from their equilibrium positions have changed
from their equilibrium values but the rates, the proportionality rates the dictate electron
hole capture an electron and hole emission have not changed. Therefore, if you make that
assumption based on this connection or this relation between the emission rates and the
capture rates they can say that e n can be written as C n times this little parameter here
and e p can be written as C p times this parameter here ok. So, essentially what we have
done is say that e n is equal to e n o and C n is equal to C n o.
Now, if we make this, if you take advantage of this assumption the differential equations
that we had for electron and hole change for R through R G mechanisms is now given by
these 2 terms, which is now simplified to writing everything in terms of the capture rates.
So, we write e n in terms of c n and e p in terms of c p and say that these differential
equations are now essentially the same as these 2 terms here, and we call this very
important equation, we give it a label and call it equation 1.
So, now let us just go back to the big picture ok. So, this is the total change in electron
and hole count in our volume of interest and this is the change in electron hole count due
to R G mechanisms and that is what we have been discussing so far. Now, at equilibrium
250
these parameters were both equal to 0 but at steady state, these 2 need not be at, at sorry,
at non equilibrium these 2 parameters need not be equal to 0.
However, at steady state condition, so, we are now out of equilibrium but we have
established a steady state condition. So, what does that mean? So, let me just give you an
example in terms of electrical circuits. So, let us say you have a, say a resistor capacitor
circuit, say something of something like this and you have a voltage source that is
driving a current through this resistor and capacitor.
Equilibrium condition means that this voltage source is 0 which means there is no
voltage there is no current through the circuits everything is balanced out. In fact, there
are there is a lot of movement of electrons inside your resistor, you can see all the
electrons moving in both directions, but the movement the flux is so well balanced out
that there are no currents in any of these elements.
So, that is the equilibrium condition. So, what steadies, what non equilibrium implies is I
have now applied a voltage ok. So, let us say I have applied 1 volt, now the moment I
apply 1 volt because there is a certain time constant to the circuit it is going to take some
time for all these voltages at all these nodes to establish. So, you will see some dynamics
there ok.
So, that is the non equilibrium transient state ok, everything is moving with time. So, all
these voltages are all shifting with time, they are all changing, the currents are all
changing with time. And then after some time these transients will all die out and the
system will achieve something called a steady state. So, these voltages would now have
settled or there would be very little variation.
In fact, ideally this busy there would be let us say 0 variation in steady state ok. So, in
steady state these transients would all die out and you will have no variation and you will
establish a fixed current, a time invariant current through the circuit and so on so forth,
ok. So, this condition is called as being a non equilibrium but being in steady state and
whereas, this condition is called being a non equilibrium, but in transient ok. So, that is
the, that is the best analogy I can give to this (Refer Time: 20:16) to what we are looking
at.
251
So, in non equilibrium these rates need not be 0, they are not 0 and in the non
equilibrium transient condition all these parameters need not be 0 but at steady state and
non equilibrium, these total effective variations in n and p have to be equal to 0 whereas,
these individual, you know rates need not be 0 ok. So, now, let us consider the case
where we are at non equilibrium, we are looking at R G mechanisms, we are at non
equilibrium and we are at steady state ok.
So, which means that these parameters are not 0 but there should be something within
the perspective of the R G process that has achieved steady state condition and that is
basically your count of say the trap states that are occupied or the count of the trap states
that are unoccupied. So, these could be assumed to be fixed then time in variant at steady
state within the perspective, within the study of your R G mechanisms.
So, therefore, with the R G processes alone, we can say that both d n by d t and d p by d t
are not 0 individually when it non equilibrium steady state but due to the fact that it is
steady state this is balanced out by any changes in the whole concentration and therefore,
d n by d t is equal to dp/dt in non equilibrium steady state condition.
So, we will now take this general equation which we have labeled as equation 1 and try
to look at these steady state implications on this equation and the steady state
implications are that this rate must be equal to this rate.
252
So, using equation 1 and establishing this condition we can now solve the equation and
estimate what n T is. So, that is the number of occupied states it is just a bit of Algebra
that is wherein you equate these 2 terms. And using that value of n t we can also get P T
because we know nT we can calculate pT as n T minus, capital N T minus small n T
which is your small p T and using n T and p T and in plugging it into any 1 of these
equations into any one of these, we can get the recombination rate which is basically
your d n by d t. The rate at which we are losing electrons in the rate at which we are
losing holes so we can calculate our recombination rate.
So, in steady state we find that the recombination rate is given by this equation here and
this is a very- very useful equation, we will give it a special label and call it equation 2.
So, what it tells you is that the rate at which the carriers, any free carriers recombine. So,
which means what, which means at equilibrium I had some carriers, I had some electrons
and I had some holes now when I took it out of equilibrium in the non equilibrium
condition I added some more electrons. So, these are suddenly appeared into the material
and we have added some more holes.
So, this is a non equilibrium condition and with time these if the non equilibrium
condition is not forced, if it is not maintained, if the semiconductor is simply taken out of
non out of equal equilibrium and left alone it will try to find its way back to equilibrium,
253
and the rate at which these carriers would recombine would be given by this rate here,
ok.
So, if there are no more mechanisms to generate carriers and there is no other process to
take it keep it in, non in the non equilibrium condition, the semiconductor will head back
to equilibrium and it will head back to equilibrium by sort of an highlighting these
carriers, these excess carriers because this is the equilibrium carrier concentration, ok.
So, in order to for the semiconductor to get back to equilibrium these carriers will have
to recombine, this accessory carriers will have to recombine and that rate can be
calculated by calculating this recombination rate. Although we will be looking at a more
general case wherein there is no necessity for the semiconductor to stop being forced in
non equilibrium. So, this is just an example what I am stating here, all right.
So, in this equation what do we see we see that n i is the intrinsic carrier concentration
which is so, n i square is nothing but your n o p o because your n o was your equilibrium
electron concentration p o is the equilibrium hole concentration and n o p is n i square, n
and p are the non equilibrium electron and hole concentrations it is a product of the non
equilibrium electron hole concentration. And n p is not equal to n i square because they
are in non equilibrium condition and this rate depends upon how much we have shifted
away from the mass action law.
So, if we wear an equilibrium, if we wear in equilibrium n p would have been equal to n

o p o and n o p o minus n i square would have been 0, all this numerator would have
been 0 and there is no need for any recombination rate. On the other hand if you are a
non equilibrium n p is not equal to n i square and that difference that deviation from n i
square is what is determining the recombination rate.
The next two points are, look at these two terms. So, capital N T is the total number of
trap states per unit volume which is related to R G centres and is c p’s are the capture
coefficients for I am sorry the proportionality constants for the capture of holes in the
capture of electrons and this term 1/cp nT has got a very special significance. It is given a
symbol tau p and it is called the minority carrier lifetime of holes. What does that mean?
It means that if you take a semiconductor that is doped. So, let us say a semiconductors n
doped. So, it is an n type semiconductor.
254
So, it is an n type semiconductor, you have a large number of electrons and a small
number of holes, the whole population is n i square by n in equilibrium. Now, you add
some excess you take this in the semiconductor out of equilibrium and you add some
excess holes inside the semiconductor ok.
Now there are many electrons and you have just injected, as you have injected some
extra holes inside the semiconductor you have shift into semiconductor out of
equilibrium. Now the typical time a measure of the time these holes would last in this n
type semiconductor is the minority carrier lifetime of holes, why minority carrier because
when n type semiconductor the holes are the minority they are the minority carriers and
the electrons are in majority they are the majority carriers.
So, when I inject holes into an n type semiconductor, they would last for typically tau p
seconds before they recombine. So, that is the, that is the lifetime. Similarly, if we inject
electrons into a P type semiconductor. So, in a P type semiconductor holes are the
majority carriers electrons are the minority carriers. If we inject electrons into a P type
semiconductor then they would last for a life for a life time for about tau n seconds
before they recombine.
So, that is called as the minority carrier lifetime of electrons. So, there are, so, it is
important that we are not interested in the lifetime of holes in a P type material. So, that
is not the minority carrier lifetime, we are interested in the lifetime of holes in an n type
material and the lifetime of electrons in a P type material.
255
So, we will now replace these 2 terms with tau p and tau n, but before that let us take a
look at what these 2 terms are, ok. So, what are that? What are those terms, what are p T
o n o by n T o and n T o p o by p T o, ok, so will get a good gauge on this. So, p T o n o
by n T o are all equilibrium values, they all got the subscript o around it but P T o is the
number of trap states per unit volume that was empty ok. So, this is P T o and therefore,
P T o nothing but NT-nT o, it is a total number minus those that are filled and you just
divide n T o here and you end up with this expression.
So, this expression is the same as that. Now let us take the reciprocal of this term which
is n To by capital NT. What is nTo by capital NT? So, you have the trap states here, you
have some of them that are filled with electrons and this concentration is n T o at
equilibrium and this total number is capital N.
So, what is nTo/ NT, it is the probability that the states at E T are filled with electrons ok,
it is a measure of that probability which basically implies that we are talking about the
Fermi function, right. So, Fermi function is the probability, gives you the probability that
the energy states at energy level E are occupied by electrons.
So, f(E ) which we discussed earlier is that probability. So, we are talking about the same
case and we had looked at this condition even in the case of doping, when we looked at
the number of donor sites or the dopants that are ionized as compared to the total dopant
concentration. So, it is exactly the same argument.
256
So, nTo/ NT is essentially the Fermi function with the small correction factor g T and
therefore, this term which is the reciprocal of this Fermi function. So, you take the
reciprocal of this which is 1 plus g T to the power E T minus E f by k T minus 1 is
essentially this. So, this term here is that term within the brackets and n o is the
equilibrium hole concentration which is nothing, but n i e to the power E f minus E i by k
T.
So, the product of these 2 will result in the E f disappearing ok. So, PTono by n T o is
nothing, but this term and this product is nothing but g T e to the power E T minus E i by
k T. Now, if these trap locations are close to mid gap. So, we have not said anything
about where E T is.
So, that is Ec that is Ev, let us say this is my intrinsic Fermi level position. If E T, if these
traps the R G centres are all located close to mid gap. So, let us say E T is approaching E
i, Ei is slightly off mid gap by a factor of k T by 2 logarithm of v n v and n c ratio and if
ET approaches E I, E T - E i will approach 0. So, this term will become 0 and this g T can
nicely be approximated as 1 ok. So, g T will take a value of 1.
So, therefore, in the condition that E T lies in the mid gap area, we can safely say that g
T is 1 and therefore, the exponent disappears because E T minus E i is 0, g T disappears
and therefore, this entire term is approximately n i ok. So, that is the order of magnitude
of that term.
257
So, that term is approximately equal to 1 e 10 per c c. And similarly n T o p o by P T o

by same arguments, can be shown to be approximately n i if E i is approximately close to
E T. So, we now got some simplified results to plug into this big recombination rate
expression here. So, we know that this is we have given this is a separate symbol Tau p
and 𝜏𝑛 and we know that these two are approximately n i provided E T lies close to v i.
258
And therefore, we can write a much simpler expression for R which is n p minus n i
square by tau p into n plus n i plus tau n into p plus n i and we call this as equation 3 and
it is something that is extremely useful in making a quick computation.
So, what this expression tells you is that gives you a gauge on the recombination rate that
is the rate at which the electrons and holes are being an highlighted in non equilibrium
ok. So, n and p are non equilibrium values we have taken a semiconductor out of
equilibrium what is the rate at which the n and p are being annihilated or R recombining.
So, that is what R is. So, let us solve a few examples.
So, we will make use of equation 3 and let us solve a few examples. So, we will take 2
conditions, two special cases because these are going to be useful to us later down the
road and we can simply refer back to these arguments.
So, the first special case is something called as Low Level injection. So, we are in non
equilibrium and we are in steady state and we have a condition called as Low Level
injection. So, what does low level injection mean it means that the excess carriers added.
So, let us say n o and p o are the equilibrium values of the electrons and holes and in non
equilibrium we have added a small number of electrons and a small number of holes
delta n and delta p and the key to low level injection is a these values are really small.
259
So, what do you mean by small? So, firstly delta n is the order of delta p and delta n
delta p is less than the majority carriers, is much smaller than the majority carriers in a
semi conductor. So, ∆n∆p is much smaller than n o in n type semiconductors and delta
and delta p is much smaller than p o in p type semiconductors, again sorry another typo
here it is in p type semiconductors.
So, just to give you some numbers ok. So, let us say, let us say we have silicon. So, n i is
110 per c c and we dope the silicon with donor dopants of 115 per c c, which means that
the doped semiconductor at equilibrium will have let us say it is purely extrinsic which
means all the dopants are ionized which are. So, in equilibrium my n o will be close to n
d which is approximately 1 e 15 per c c, my p o is n i square by n o which is
approximately 15 per c c and my delta n and delta p are let us say about 1 e 8 per c c.
So, 18 is definitely much smaller than 1 e 15 and these arguments could be applied to that
condition note that 1 e 8 may not be smaller than 1 e 5, that is we are not talking about
we are not comparing the low level injection to the minority carriers, we are comparing
to the majority carriers. So, that is the only condition we need. So, that is what we mean
by Low Level injection.
260
So, now, under these conditions let us go ahead and use our equation 3. So, just for the
𝑛𝑖2
sake of clarity, I will write it down again it was it said that, it said that 𝑅 = 𝑛𝑝 − ∗𝑛+
𝜏𝑝
𝑛𝑖+ 𝜏𝑝∗ (𝑛 + 𝑛𝑖 + 𝜏𝑛∗𝑝 + 𝑛𝑖 ). So, that was equation 3.
So, now what does my n, n is nothing, but n the equilibrium value plus delta n and p is
nothing, but p0+∆𝑝. So, we plug in those values for n n p everywhere and we open out
this term ok.
261
So, what happens when we open out this term. So, you will end up with n o p o, the
𝑛𝑖2
numerator will be 𝑛𝑝 − ∗ 𝑛 + 𝑛𝑖+ 𝜏𝑝∗ (𝑛 + 𝑛𝑖 + 𝜏𝑛∗𝑝 + 𝑛𝑖 + ∆𝑛 + 𝑛𝑖 + 𝜏𝑝 𝜏𝑛 ∗ 𝑝 +
𝜏𝑝
∆𝑝 + 𝑛𝑖 ) .
Now, 𝑛𝑜 𝑝𝑜 is nothing, but 𝑛𝑖2 n i square. So, this is n i square and these 2 terms can
cancel. Now, between these 3 terms since delta n and delta p are much smaller than n o
and since my n o is much greater than p o in an n type semiconductor. So, let us say we
are looking at n type semiconductors. So, in an n type semiconductor this term dominates
and the other terms are all negligible and the numerator will simply be delta p n o.
And in the denominator n o is going to be much greater than n I, it is going to be much

greater than delta n, it is going to be much greater than p o, it is going to be much greater
than delta p and therefore, only this term dominates in the denominator and therefore, my
recombination rate for low level injection at steady state for an n type material is going
to be the excess minority carriers injected divided by the minority carrier lifetime. So, in
this case it is going to be the excess holes injected divided by the minority carrier
lifetime of holes.
Similarly in the case of p type semiconductor, so if you retain these expressions, p o is

much greater than no and ∆np. So, these 2 terms disappear and the numerator becomes
delta p, I am sorry ∆npo delta n into p o and from the same arguments this term is what
262
dominates and your denominator will be 𝜏𝑛𝑝𝑜 and the recombination rate is the is delta n
by Tau n, that is the excess electrons injected in the p type material divided by the
minority carrier lifetime of electrons.
So, the second example and the final example, we will look at is the steady state, the
recon trying to calculate the recombination rate in steady state in a depletion region. So,
we have not yet discussed what a depletion region is, we will do so very quickly but a
depletion region in a semiconductor is essentially a region and a semiconductor which is
sort of engineered to have very little free carriers. So, that the n and p count are going to
be extremely low, it is depleted of free carriers.
So, which basically means that n is much less than n i p is much less than n i. So, if you
use these 2 conditions in this equation your n p is going to be much less than n i square.
So, the numerator simply n i square, your denominator the n is not important compared
to compared to n i. So, here this is a typo again it should have been p, but that is not
important compared to n i and therefore, you end up with the recombination rate being -n
i y 𝜏𝑛 /𝜏𝑛 + 𝜏𝑝 and this negative sign tells you that the recombination rate is negative
which means you are actually having generation.
So, in a depletion region there is carriers are being generated, there is an encouragement
towards generating carriers and that generation rate is 𝑛𝑖 /𝜏𝑛 + 𝜏𝑝 .
263
Lecture - 16
Recombination and Generation - Continued, Charge Transport
We have two more topics to discuss in this with regards to Recombination and
Generation. And the first is something true with the Capture Cross Section. So, this is an
order is not a new concept that is going to change things that we have learnt, but it is a
different way of viewing the R-G capture and it is a very geometrical method of viewing
the capture of electrons in R-G centers and the argument for this goes like this.
So, let us say you have a volume in a semiconductor ok. So, were looking at the actual
physical picture in a semiconductor and this volume is going to cross section A and an
electron is sort of traveling through this cross section it is going through this ok. So, let
me draw that better. So, the electron is actually traveling through that.
And let us say the electron has got a thermal velocity V t h and V t h is what it is k T by
2 because it is just one dimension here and m V square by 2. So, that is the relation that
is dictating what is V t h; it is got some thermal velocity. And therefore, in time t the
electron would have travelled a distance of V t h by t.
264
So, let us look at this volume which is got an area of cross section it is a cuboid which
has got an area of cross section A and it is got a thickness or you know the depth is V t h
into t. So, the volume of this cuboid is V th * t*A and inside this region of space, you
have some R-G centers.
So, these are all R-G k centers that can capture electrons and these centers have got a
capture radius. So, if an electron where to move inside here, it is definitely captured; it is
trapped and if it does not, if it can escape all these centers then the electron is not trapped
ok.
Therefore, the since these centers for capture have to be empty states ok. So, we have the
R-G states some of them are filled; some of them are empty and it is only these states
that can capture electrons ok.
265
So, the number of capture sites inside this volume is P T which is the number of capture
sites per unit volume into the total volume under consideration. So, the total volume
under consideration is V t h into t into A and that volume occupied by the capture sites is
V t h *t*A into P T. P T is a number per unit volume.
Now, let us project this entire three d structure onto a 2 D plane. So, we sort of squeeze
in we flattened out all the R-G centers. So, you have the R-G centers here, they might
have been to say behind one another and you know these are all regions occupied by the
R-G centers and this entire area has got an area of A and the total area, the total projected
area as seen through all these R-G centers is the number of R-G centers to some extent
into the cross section area of each R-G centre ok.
So, we will define this cross section area; this is the area within which if an electron were
to enter, the electrons were to be captured. We will define this cross section area as
sigma. This is the number of R-G centers and if these R-G centers are seen
independently that is there is no intersection between them.
The projected area of all the R-G centers is essentially this and this sigma can sort of take
an effective value. So, if there is a lot of intersection between these R-G centers as we
project them. So, let us say it is a very nice geometrical concept. So, let us say there were
2 R-G centers just one behind the other and they appear like this in this projection. So,
sigma could take an effective value ok; need not it can be defined the little differently.
266
So, essentially what we are talking about here is that this area represents the shaded
region which are all the capture zones and A is the total area and an electron is going to
go through this region and it is going to go through at a random point.
So, what is the probability that the electron is captured? It is a probability that the
electron lands on one of these shaded regions which is V t h t into A p T sigma by A
which is this term here. What is the rate, the probability rate that the electron is captured?
It is the time derivative of that which is simply this term. Now, the probability rate times
the number of electrons per unit volume is essentially a measure of d n by d t ok.
So, d n by d t; then becomes V t h into p T into sigma into n which according to our old
definition was c n n* p T. So, if you remember of many slides back, we said that the rate
at which electrons are captured. So, this way this is only capture. So, we are not saying d
n by d t, there is only capture of electrons rated which they are captured or trapped.
We had said that that was proportional to n*p T and we gave a proportionality constant
called c n for the electron capture. So, we are talking about the same derivative here.
Therefore, these two terms are the same which implies that we get a geometrical insight
into what c n is. So, you see n and p T can be cancelled and therefore, your c n is nothing
but the thermal velocity into sigma. So, it is the same concept, but it is a very
geometrical way of viewing the same idea. Now, the final topic in on that we will study
that we will study under Recombination and Generation ok.
267
So, this is a typo, there is no thermal that I am sorry quite a bit of quite a few typos and
this topic; please do bear with me.
The final topic that we will study in recombination generation is to do with the concept
of Quasi Fermi Levels ok. As this is a concept that we will not be using quite a bit, we
will not we will not really use Quasi Fermi Levels through the course. We will manage it
without using this concept, but it is a concept that you need to be aware of and I think its
best discussed with the example, with a with an example. So, let us say you have an N
type semiconductor equilibrium ok.
So, N d is 1e15 that is the; it is N type semiconductor. Therefore, the donor concentration
is 1e15 per cc and they are asked to locate E f minus E i. So, we had an intrinsic
semiconductor that was E i, the intrinsic fermi level and after doping it N type, the fermi
level moved closer to the conduction band and we want to calculate what this is.
Now, there are two methods to calculate it because I know that n is equal to n i e to the
power E f minus Ei/kT and P is equal to n i e to the power minus E f minus E i by k T.
Therefore, if I use this approach my E f minus E i is equal to k T ln n by n i and if I use
this approach my E f minus E i is equal to minus k T ln P by n i. So, let us see what both
these approach is giving?
Now, n i let us say is about 1 e 10 which means my n since its doped and since now the
dopant concentration is much larger than 1 e 10, we will say n is approximately 1 e 15
per cc and P according to the Mass action law it is all at equilibrium. So, therefore,
according to the Mass action law, it is n i square by n which is 1 e 5.
So, if we were to plug in n/ni. So, n by n i is 1e15 by 1e5 1e10 which is 1 e 5 and what is
n i by what is p/ni? p/ni is 1e5/1e10 which is minus, I am sorry 1e-
5. So, you plug in these two values here, we will find that both these options give you the
same answer and you will find that E f minus E i is approximately 0.29 electron volts.
268
Now, let us take this semiconductor out of equilibrium ok. So, let us say we throw some
light on the semiconductor and we generate 1 e 12 electron pairs electron hole pairs.
So, what is happened to the carrier count now? Now, n was initially 1e15 at equilibrium,
we have now added 1e12 carriers and therefore, 1e15 plus 1e12 is approximately 1e15
because the amount of injected carriers about is about 1000 times less than the
equilibrium carrier concentration. On the other hand, p was only 1e5 per cc in
equilibrium and now you have added 1e12. So, therefore, we have made a significant
change in the count for p. So, p has now become 1e12 per cc. So, although n remained
the same, P has suddenly changed.
Now, let us try to calculate E f -Ei. We will use both these approaches. Now clearly you
can see that there is going to be a problem because the E f -Ei calculation sorry. E f
minus E i calculation using the electron count has not changed because n is still 1e15, but
the E f -Ei calculation using holes has to change significantly because the hole has now
gone from 1e5 to 1e12. So, this used to be 1e5 for cc, but now because of light being
through on it, it will become 1e12 and you will end up with a very different terms; so,
you will end up with 0.17 electron volt. So, which of these is correct ok? So, there is this
seems you were paradoxical.
So, therefore, when we discuss non equilibrium conditions ok, we cannot allow a
common fermi level for electrons and holes. It cannot be the same. Therefore, the
269
electrons now take a fermi level called E f n which is the quasi fermi level for electrons
and holes tickle fermi level called E f p which is the fermi level for holes and therefore,
we must rewrite these equations we cannot use one common E f.
Therefore, Efn- Ei is 0.29 electron volts which means the quasi fermi level for electrons is
still very much the same as the fermi level at equilibrium. But the quasi fermi level for
holes has shifted from the equilibrium fermi level value ok. It is now only one 0.17
electron volts above your E i or you know yeah ok.
So, this is the idea of quasi fermi levels and this is something that we will not use, we
will not be using this concept, but although it is useful to know, it can be a very powerful
tool while performing calculations, but we will try to avoid it ok. So, with this, we come
to the end of the study on Recombination and Generation and we now have the tools to
understand how d n by d t is affected by recombination generation mechanisms ok.
And we will now embark upon establishing a concept or an idea which is called as the
continuity equation or a complete budget of the carrier counts in the volume which is the
big picture we have been discussing and we will use that to study semiconductors in
various conditions.
So, we now consider mechanisms of charge transport ok. So, far if you make us if you
take a summary of everything we have done, we have established method to count the
270
number of carriers in the conduction band and the number of holes in the valence band.
So, we can count them and in order to count them, we established what the density of
states was; we established what the fermi function was and with these two, we could
evaluate the values of n n p. Then, in the generation recombination processes, we saw
how the count through the dynamics of the electrons in the conduction band and the hole
holes in the valence band is influenced by the various features.
So, now we are trying to establish the concepts of charge transport. So, these are the
mechanisms by which the electron and hole concentrations could vary and now based on
these variations or based on these counts of electrons and hole concentration, we want to
establish what the currents would be in a semiconductor and the different mechanisms of
charge transport in a semiconductor. So, we are gradually heading towards something
called as the Continuity equation which is a very very useful concept and a useful idea in
order to solve out the reaction of the semiconductor for various situations. So, let us
focus on two most important methods of charge transport ok.
So, we have already come across one phenomenon which was called as Tunneling ok.
So, we will skip that for now. We have we are not interested in tunneling in this
particular discussion. We are interested in two particular modes of charge transport and
they are called as Drift and Diffusion ok. So, Drift and Diffusion and in a very little
while, I will explain to you what these two terms imply ok.
271
Now of course, there are other methods of charge transport and I think the most
prominent one is a tunneling which was a you have the wave function of the electron on
one side of a barrier and the wave function does exist on the other side, simply because
there is a non-zero probability that the wave function can pass this barrier, despite the
energy of the electron the energy of the carrier being less than the barrier height.
We have already looked at this in a brief discussion and quantum mechanics and we will
use we will use the concept of tunneling later on, but it is not the focus of this particular
study.
So, first let us look at the Drift current ok.
Now, what is Drift current mean? Drift current simply means that your electrons and
holes are moving through the semiconductor because of the presence of an electric field.
So, you have a semiconductor, you have your let us say you have your conduction band,
you have your valence band and you have free electrons in the basically the electrons in
the conduction band and you have holes in the valence band.
And the moment you apply an electric field, what would happen is these bands would
actually bend ok. I have shown here and all the electrons would run downhill and the
holes would run uphill ok. So, that is the way we the energy band diagrams are drawn to
define the lower the lower side if you might allow as having the higher potential. So, all
272
the electrons would run towards the higher potential side and all the holes will run
towards the lower potential side.
So, this is the response to the electrons and holes to an applied electric field. So,
essentially what this requires is you need to have free electrons in the conduction band
and there are empty states and these electron sort of move through these states; whereas,
the electrons in the valence band move through these vacancies which essentially gives
you this picture of the holes moving towards the other side ok. So, that is what is
happening when you applying electric field through a across a semiconductor.
So, in order to get a quantitative estimate as to what the drift current is; if you think of
the semiconductor ok. So, you have applied an electric field, let us say the area of cross
section is A and let us say the semiconductors got a length capital L and let us say we
have applied and a voltage V a cross this ok.
Now, as long as the electric field, we which we will sort of write down as V a by L ok.
So, may not be perfectly correct, but for all practical purposes for the time being we will
just say it is Va /L. It is a very uniform electric field there is nothing else inside and we
have a field of Va /L. So, the electric field in the semiconductor alarms the electrons and
holes to move to the opposite electrodes.
273
Now, the electron flux ok. So, if you look at the current density and we will say that the
drift current density has got this what is a subscript d r as this written here; the drift
current density, the total drift current density is the drift current due to electrons and the
drift current due to holes.
Now, do these two current sum or take a difference ok. So, if you apply electric fields,
the holes which are positively charged run in the opposite direction as that of the
electrons and therefore, these two currents sum up to give you the total drift current.
Now, what is the drift current due to electrons? It is the total charge per unit volume. So,
we are talking the drift current density. So, this is the current per unit area ok. So, we are
talking about I/A.
So, this is nothing but the total charge of free electrons per unit volume. So, n is the
number of electrons per unit volume that is in the conduction band. So, which is a free
electron concentration; q is the electronic charge and if a q n is the total charge of the
electrons that are responding to the applied field into the drift velocity of the electrons.
So, this gives you this gives you a measure of the distance covered by the electrons per
unit time and therefore, these two terms together is nothing but the rate of change of
charge per unit area which is nothing but d q by d t, where this all taken as a per unit area
estimate ok. So, that is the drift current due to electrons ok. So, you have q and V d n and
similarly for holes you have q p V d p; where, V d p is the drift current due to the applied
electric field.
Now, we have already looked at the concept of mobility of carriers and we did notice
that in the presence of a low electric field. So, you have you have the electric field and
you have the drift velocity of the carriers; at low electric fields the drift velocity is
linearly proportional to the electric field and then, it sort of saturated out in some of the
semiconductors.
So, at low electric field, the relation between the drift velocity and the applied electric
field is given by V d s proportional to the electric field with a constant of proportionality
being the mobility of the semiconductor. So, therefore, they can rewrite the drift velocity
of the electrons and holes as q n mu times the electric field or mu n times electric field;
where, n mu n is the mobility of the electrons and q p mu p into electric field; where, mu
p is the mobility of the holes.
274
So, the summation of these two terms gives you the total current or the total current
density due to drift and this is a very useful relation keep in mind, where in the electrons
and holes contribute to the current together. So, certain points to note here; so, one the
first thing is you might wonder what happened to the electron charge being negative ok.
So, well that is already taken into account because the electrons and holes are moving in
the opposite direction. So, if you are going to look at the conventional current ok. So, we
are talking about the current in the opposite direction of the electrons.
So, the electron flux is one way, the current is always the opposite side that is the way
current is defined. It gives the it is from the positively charged electrode to the negatively
charged electrode and therefore, the current is in the same direction as out of the holes
and since, the electrons have got negative charge and are moving in the opposite
direction it is equivalent to having a current of plus q and mu an electric field in the
opposite in the direction of the current ok. So, this is the relation between the drift
current density of electrons and the drift current density of holes and the summation
being the total drift current.
Now, this essentially leads to the formulation of Ohms law ok. So, for example, let us let
us look at the total current ok.
So, we have the total current through the semiconductor being J times the area and we
already know that the current per the drift current density ok. So, we are only interested
275
in the current due to the applied, current due to the applied voltage and the drift current
dense the drift current density is nothing but q and mu n plus p mu p into the electric
field into the area gives you the total current.
Now, the electric field is nothing but your d v by d x ok; you can throw in a negative
sign, but we are interested in the magnitude of the current because in the in the case of in
order to derive Ohms law. I really would not establish a relation between the voltage the
current and the resistance of a resistor. And therefore, the direction of the current is not
really a matter of concern right now ok. So, therefore, we will not worry about the
electric field being minus dv/d x; we will just say dv/dx and instead use the modulus of
the current ok.
So, this happens to be the current through the device. Now if you just solve this
differential equation. So, as so instead of Va/L which is the very liberal way of defining
electric field, we have defined the electric field more carefully as d v by d x ok. So, B by
L is a very liberal way of defining the electric field, particularly when a semiconductors
involved; as you will soon see ok. So, that is that is probably true for an insulator, but not
so true for a semiconductor. So,.
So, let us define let us define the current versus voltage in this particular manner. So, if
you solve this differential equation and you integrate, you integrate the current with
respect to d x and as you go from one end of the semiconductor to the other. So, let us
say you go from 0 to L the voltage varies from 0 to V a and by just solving this
differential equation, you find that your V a is nothing but 1 by q n mu n plus q p mu p
into L by A into I ok. So, this is the relation between the applied voltage and the deep
current through the device.
So, what does this remind us off? So, this term here is the resistance. So firstly, it tells us
a thick the current voltage have a linear relation. So, the voltage is equal to I into R and
this R has got a term which is equivalent to the resistivity. So, this particular term is the
𝜌𝐿
resistivity, where R is a rho . L and A defines the geometrical parameters and most
𝐴
importantly rho is the resistivity that is 1 by q n 𝜇n +p𝜇p.
So, the resistivity of a material is taken to be a constant ok, but now we know what that
constant depends upon. It depends upon the it depends upon the mobility of electrons in
276
the mobility of holes and the total electron concentration per unit volume and the hole
concentration per that is the number of holes per unit volume. So, in a metal, you do not
have holes and therefore, the resistivity is simply 1/ qn𝜇nmu m all right. So, that is Drift
current. So, it is a response of the carriers to the applied electric field which is obtained
by applying a voltage across the device.
277
Lecture- 17
Charge Transport - Continued
Now, the second mechanism of Charge Transport is something called as diffusion. Now,
what do you mean by diffusion? When the while the driving force for what you said drift
based conduction is the electric field a driving force for diffusion based conduction is the
concentration gradient ok. So, if you were to pass by on a bakery for example, and if it is
a really nice bakery and as a warm day probably you will find that you know you get a
lot of sweet scents that are sort of in the even though you are not at the bakery, you do
get a whiff of all the sweet scents in the air and that is got to do with all the molecules
from all the sugars and everything else that is used in the bakery. So, diffusing through
the atmosphere and ending up being sensed by a nose of course there is a big difference
those molecules are all neutral species and their diffusion is defined by a slightly
different parameters.
But what we are interested in is a diffusion of charged species which are electrons and
holes and the diffusion of these species is got is basically driven or they know the flux of
these the flux due to diffusion is because there exists a difference in concentration. So, if
278
you go back to your, if you go back to a drift, we saw that you know the drift is because
of a potential difference right, but you can imagine the electrons ok. So, if you have an
electron concentration which is very large and say this area, you have lots of electrons
here, and you got very little electrons there. Now, it is very a there is a tendency for these
electrons to move or migrate towards this low concentration region and that is what
diffusion is.
But, if you think about it a presence of a lot of charged particles here implies this region
has got low voltage, lower potential is compared to this region. And therefore, in the case
of charged entities the diffusion can be thought of some kind of a self-induced drift ok, it
might be a wrong way of thinking about it when it comes to other matters, but it does
help the understanding if you if that in some sense ok. But essentially the driving force
for diffusion is the concentration gradient. So, let us try to derive the diffusion current
ok.
The diffusion current so let is to start off let us say we only talk about electrons a let us
say with distance x ok, so that is my its just one-dimensional case that is x increasing.
And you have a concentration of electrons that is varying its space you know we have
just drawn this little squiggly line which is to indicate the electron concentration that is
the number of electrons per unit volume that is varying with x. So, your d n by d x is
greater than 0, which means as x increases my electron concentration is increasing,
279
because this is the direction of increasing x. Now, these electrons in the semiconductor,
so let us say that does exists the situation of this kind.
Now, if you look at all these electrons, here you got lots of electrons here, so that is
basically all the electrons there and you have got few a electrons here ok. Now, these
electrons have got some thermal velocity. We have not applied any voltage ok so, there
is 0 voltage or 0 electric field being applied these electrons are simply have thermal
velocity ok and what is thermal velocity in one dimension your thermal energy by 2 ok.
So, it is only a 1D case. So, kT/2 is equal to your half effective mass of the electron into
V2 and your V is your thermal velocity which is √(kT/me*). So, all the electrons have a
velocity of this value.
And this velocity is in an arbitrary direction. So, if it is a one-dimensional case, some of

the electrons are moving this way, some of these electrons are moving this way. And
similarly on this side, you have few electrons moving this way and few electrons moving
that way. So, this is you know this is the situation we are looking at.
So, our interest is to look at a very specific region and as these electrons are moving
through they are colliding with the lattice. So, there is a mean collision time. And we had
defined this mean collision time when we define mobility right. So, the electrons collide
with the lattice, and therefore it is only within this mean collision time that the electrons
actually see an acceleration ok. So, we think about the when we had looked at mobility
280
when we had looking mobility we defined the mobility in terms of the mean collision
time.
So, this is the time between any scattering event ok, between any two consecutive
scattering events statistically speaking. So, we are interested in a boundary that is defined
by this value L. So, x equal to L which is equal to the thermal velocity the electrons into
the mean collision. So, the electron moving with a certain thermal velocity travels this
distance before it scatters ok, so that is the boundary we are looking at. And what we are
interested in identifying is the total number of carriers that are going to effectively move
from left to right and from right to left and cross the plain of x equal to 0. So, you want
to make a get a count of these carriers ok.
So, since half the carriers are moving left. So, if we take the x equal to let us start with x
equal to minus L, let us start with this boundary the leftmost boundary. So, if we get a
count of how many carriers are moving in there on towards the right side that is towards
x equal to 0, it is half the number ok. So, we will say that n by 2 carriers are moving
towards the right and they all have a velocity of v th, therefore the effective flux or the
number of carriers per unit area that are per unit area that are moving in a given time
interval from left to right is n into n at x equal to minus L into v th by 2 ok, so that is the
approximate average number of carriers moving from this boundary to the that.
And the since they are not going to have any scattering events, all these carriers are
going to cross this x equal to 0 plane right. Now, what about the number of carriers
moving from the right to the left ok, so that is going to be n the carrier concentration at x
equal to L*vth/2 because half of them are going to be moving towards the left. And once
again everything as starts moving towards the left is going to cross the x=0 plane because
there are no scattering events and that was the purpose of taking the mean collision time
into this definition.
So, therefore, the effective flux from left to right is basically this particular term which is
this flux which will call it flux left to right, and this is flux from right to left ok. So, the
flux from left to right effectively is the flux from left to right minus the flux of carriers
from right to left so that is the effective number of electrons moving from left to right.
281
And that is given by n at –L-n (L)*vth/ 2.
Now, if we know the carrier count at n at x = 0, so let us say so this is the carrier gradient
right and these are two boundaries and that is your x equal to 0 plane. Now, if we know n
at x equal to 0 which we call n of 0, the value of the carrier concentration at n at of n the
carrier concentration n of L is simply n(x)
equal to 0 plus the slope into this distance L right. So, it is n(x) equal to 0 by dn/dx into
L or if you want you can think of it is a Taylor expansion.
282
Similarly, the carrier concentration here n of minus L is n at 0-dn/dx into L ok, because
the way we have drawn at dn/dx is positive and therefore, these equations fit the physical
picture. So, this is the carrier concentration at minus L n l written in terms of the carrier
concentration at x=0. If you simply add these up these two terms disappear and you end
up with this particular term defining the effective flux from left to right. And we will
now replace this L with v th*t collision to get you this relation. So, this relation tells us
that there is an effective flux from left to right, which is dependent on d n/d x and the
square of the thermal velocity which is a constant and t the mean collision time which is
also a constant, therefore, the driving force is d n/d x.
Now, to understand this better, so that is the effective flux right, so that is the flux, what
is the total that is a rate of change that is that is number of electrons moving from left to
right to unit time because you taken into account v th, so it is a flux per unit time. So, the
total current simply the charge transiting per unit time which is q times that flux ok. So,
it is simply q times d n by d x into v th square into t collision.
Now, what is v th square, we had already defined v th square as since it is a since it is a

or it is a 1D case your half mean cool m e v th square was equal to k T by 2 or v th
square is equal to kT/m, where the m is the effective mass of the electron or m star,
where m star is effective mass of electron. So, we will replace v th square by this term k
283
T by m star. And therefore, we have the total current due to the concentration gradient
from left to right being equal to this particular term.
Now, note what happened to the negative sign, the negative sign, so if you look at the
previous page there was a negative sign here that was to do with the effective flux,
because there are few electrons here and more electrons there, therefore it is
understandable that the effective flux was in this direction. But then the electron charge
is also negative q, it is minus 1.6 e minus 19 coulombs which is equal to minus q ok. So,
we have removed the negative sign by having a minus q into this negative flux which is
basically telling you that the current is positive. So, you have all the electrons drifting
this way which means that there is a positive current going on from left to right.
Now, to this term we will add in we will try to replace these two terms of m and t
collision by the parameter mu. So, if we remember you know how does how does mu
connect to t collision and the effective mass of the electron ok.
So, if we just step back ok, and look at what is 𝜇, 𝜇 is nothing but the relay it is the
constant the proportionality for the drift velocity and the electric field. So, the drift
velocity is mu times the electric field right. So, all the way discussing diffusion we need
to step back and understand what mu is an how it is connected to m star and t collision.
So, it does not mean that this process is dependent on an applied electric field, there is no
284
need for an applied electric field, it is only to understand knew better that we need to sort
of step back.
And when if you apply an electric field the total force experienced by the electron is this
which you can say is the rate of change of momentum of the electron. So, it is m star into
v drift by t collision and therefore, v drift was q.d collision by m star into electric field
where this term is the mobility ok, so that is the definition of mu. So, if I were to apply a
field to the semiconductor the carriers would move with this drift velocity, but right now
we are not applying a field, but nevertheless this relation still holds in the sense that t
collision by m star is nothing but mu time mu divided by q ok, so that is what we are
going to use that relation and replace this ratio of t collision by m*with 𝜇/q. So, we
replace t collision by m* with relation mu by q in order to get this particular expression.
Now, this term in bracket has got a very special significance and it is called the diffusion
coefficient and it is denoted by the symbol capital D. And since we are talking about
electrons here this is the diffusion coefficient of electrons ok we will call it D n. So, you
could also call that mu n. So, the so therefore, the diffusion current can be written as qDn
dn n/dx ok. So, this is D n is nothing but the capital D subscript n, which is that diffusion
coefficient for electrons. Similarly, we could define a diffusion coefficient for holes and
therefore, define a diffusion current for holes.
285
Now, if I were to have a similar profile if I had to have a picture of P, and if I have a flux
if I have a concentration gradient you will find that all the holes are moving from right to
left because that is because they have got a higher concentration on the right and lower
concentration on the left. And therefore, the current from left to right will be negative ok.
Now the now the electron current the negative sign disappeared because the charge was
negative, but in the case of holes you have a positive charge, and therefore the if the flux
is in this way the current from left to right is a negative current which is essentially
telling you that the current is in the same direction as the migration of holes.
So, this is the diffusion current density for electrons. And we define the diffusion current
density as J d f comma n electrons and J d f comma p for holes. And D n is the diffusion
coefficient for electrons, and D p is the diffusion coefficient for holes. And we will and it
is useful to note this relation and we can sort of derive this, and I will show that to you in
a little while. So, D n is kt𝜇/q ok. So, this relation is something called as the Einstein’s
relation. In order to understand how Einstein’s relation comes about, all you need to do is
take any situation it is not just limited to carriers in a semi conductor in any situation
where the carrier count follows a Boltzmann distribution.
286
In the sense that n scales is proportional to e to the power minus a potential, so let us say
q𝜑/k t or let us say it is a minus phi by phi naught. So, if you we have a Boltzmann slight
distribution where n say some n naught e to the power minus phi by phi naught, where
phi is a potential you will end up with the Einstein relation being valid just to derive it.
So, if you take if we take on charge carriers itself or carriers the electrons and holes in a
semiconductor.
The when left alone at thermal equilibrium, you need to have if there exists any drift
current it must be completely balanced out by the diffusion current ok. So, at thermal
equilibrium, you cannot have currents. And if there is a drift current it must be balance
start by diffusion current. And using that as a starting point you need to have your q, let
us only take electrons is an example let us say electrons the only carriers. So, you need to
have q n mu electric field matching your diffusion current which is qDn dn/dx. Now, if n
follows a Boltzmann relation say n is equal to some no 𝑒 −𝜑 by let us say instead of
thermal voltage we will just call it some 𝜑𝑜 which is a constant your d n by d x is equal
to no 𝑒 −𝜑 /𝜑𝑜 all right into minus 𝑑𝜑/𝑑𝑥 ok. And your electric field is equal to −𝑑𝜑/
𝑑𝑥. So, you substitute these two in this particular expression here.
287
And you will end up with relation qn mu d phi by d x ok, let us get it to the negative sign
is equal to q.
Dnno𝑒 −𝜑 /𝜑𝑜 is again n, so n 𝑑𝜑/𝑑𝑥. So, the
term n 𝑑𝜑/𝑑𝑥 cancel out ok, and sorry I missed out a 1 by phi naught term here ok and
so you take into account the 1/𝜑𝑜 ok. So, you end up with Dn being equal to let us just
bring that bring that n right here being equal to 𝜇𝜑𝑜 . Now, a 𝜑𝑜 is a thermal voltage
which is kT/q, we will end up with your Dn 𝜇𝑘𝑇/𝑞 which is your Einstein’s relation, so
that is the way you know this relation comes about ok. So, the diffusion coefficient is
very much connected to the mobility of carriers.
288
So, now let us look at the total current ok. So, we have identified the two mechanisms
you have drift and you have diffusion. In the case of drift, we found that the electron and
hole currents add up ok, the electron currents and the hole currents add up to give you
the total current, but in the case of diffusion for the same profile the effective current was
in the opposite direction. So, if we will if we want to define the total electron current ok,
it is the current due to drift plus diffusion of electrons the current due to so let us take
this profile here ok.
So, you have dn/dx greater than 0 and let us say we apply an electric field in this
direction ok. So, all the electrons will move from right to left because of the applied
electric field ok, and all the electrons will move from right to left because of the
concentration gradient, so which means that the drift current where the driving force is
the electric field and the diffusion current where the driving force or the concentration
gradient both exist in the same direction and they both contribute to the curve. So, J drift
plus J diffusion for the electrons is going to be the summation of q n, mu n the electric
field plus qDndn/dx to give you the total electron current density. But for holes if you
have the same situation you have dp/dx greater than 0, you have the electric field which
is the driving force for drift pointing in the same direction.
But what happens is that the whole flux due to the concentration gradient is towards the
left, because there are more holes here as compared to here. And the whole flux due to
289
the field is going to be along the direction of the field. So, it is going to move from right
to left and therefore, these two are against each other. And therefore, the total hole
current is going to be the total drift current due to holes plus the diffusion current due to
holes, but the diffusion current has got a negative sign. And therefore, you have q p mu p
into electric field minus qDp. dp/dx with the total effective current due to holes, so that
is a quick summary of the different current mechanisms.
290
Lecture - 18
Continuity Equation
So, now we will begin our discussion on a very useful idea which is a something called
as a Continuity Equation. And students are familiar in with other areas of engineering
you know such a fluid mechanics etcetera, might have come across this term. And the
idea is very similar which is to draw the complete budget or a complete balance of
charge and mass and make sure that we consider all processes that influence charge in
mass and establish some understanding of the behavior of the device.
So, to start with I think a figure here might be very useful, which is we always looked at
this big picture wherein we considered a volume of a semiconductor. And we wanted to
have a count or a measure of the number of electrons per unit volume inside this element
of interest ok. So, let us say that the element of interest has a cross section A and that
anything that is of interest is only happening in one direction, it means in one dimension
which is the dimension x ok. And inside this volume had many mechanisms which could
change the population of n and p.
291
So, we had processes that are called as generation mechanisms, which created electron
hole pairs ok. We had a mechanisms of recombination which and highlighted electron
hole pairs, and we discussed these mechanisms. And we also discussed mechanisms
which result in current which is your drift and diffusion currents; so you have currents
coming in to this region and currents going out of this region.
So, specifically we will define a little volume element to have these values let us say this
boundary is x, this boundary is x+dx ok. And it is got an area of cross section A and the
current coming in is the J at x and the current that is going out is the J at x plus d x. So,
these are all the processes that are going on. And therefore, you have currents coming in
and going out you have R G mechanisms and any other processes which is affecting your
total electron and hole count.
Now, if you consider only electrons ok, so let us not worry about holes first we will
represent these currents as J n that is coming in and going out. You will find that your
total so n is the number of carriers per unit volume; it is the count number per volume ok.
J is the current density it is the rate of change of rate at which the charge is changing per
unit area; so in some sense it is per unit area. Therefore, and your generation and
recombination rates they are all talking about the rate of change of the number of
electrons or holes per unit volume in this element. So, that is what your G and R.
So, let us just establish everything in terms of the total number ok, let us not worry about
densities because here you have per unit area there you have per unit volume and
therefore, things can get a bit confusing. So, we will just establish everything in terms of
the total number. So, in this element of study so, this element has got a volume of A
times d x so, that is the volume of this element. And the cross section area is A. So, dn/d
t.Ad x is the total count of the total rate of change of electrons in this volume. That is
equal to that is dependent on all the current that is coming in, minus the current that is
going out, plus all the generation mechanisms minus the recombination mechanisms.
So, the current that is coming in brings in electrons in the current that is going out takes
electrons out the generation mechanisms increases the electron population and the
recombination mechanism destroys the electron population. So, since these are currents
and n is a number, we are not talking about this charge it is a number right. And these are
currents which I have already got the parameter charge embedded in it.
292
So, we are going to divide that by the charge of the electron which is minus q in order to
get this in terms of a total number. So, this is the number of electrons that is what we are
talking about here. And similarly we can write a similar expression for holes ok, but the
only thing is the hole charge the holes is a positive quantity ok.
So, this is a very useful equation ok, it is a complete balance of everything that is going
on and it is something called as your continuity equation. So now, we are going to work
with these equations and bring them to a form that is much more usable ok. So, first let
us look at what is Jn -J at x- J at x+d x.
So, let us first only worry about the electrons. So if you write; if you think of your J at
x+dx has simply Jn at x plus the slope into d x, then Jn of x- J n at( x+dx) is simply equal
to minus d Jn/dx.dx ok. So therefore, this term this numerator is going to be minus
dJn/dx into dx. And using that in that numerator your continuity equation now becomes
dn/dt.A.dx has got this term, which is based essentially this plus your generation
recombination mechanisms.
Now, you will notice that this minus sign here has been cancelled using the minus sign at
the electron charge, so there was a minus q because the charge is electron is negative. So,
that is that minus sign has gone away and you have a plus in that location. So now, we
can very easily clean out all these volume elements Ad x can be cancelled everywhere
and we have that equation to be appearing in a very simplified form as shown here.
293
Now, let us think about the currents a little bit more so, we had discussed earlier that you
know the electron you have the drift currents and you have the diffusion currents and the
in the case of electrons. The drift and diffusion currents were supporting each other, and
therefore you have the drift current of electrons plus the diffusion current due to
electrons contributing to the total current density ok.
So, we substitute for J n using this term, but what we need is the rate at the variation in J
n with x so, dJn/dx into 1/q. So, this term 1/q dJn/dx is going to be is going to result in us
taking a spatial derivative of the drift component and a spatial derivative of the diffusion
component.
So, let us say that n so the spatial derivative the drift component is going to have two
possible variation, two possible elements that could be varying in space that are that is
basically your n and the electric field. So, you have dE/dx and you also have dn/dx so
that is the drift contribution. And the diffusion contribution simply leads to your d2n/dx
square appearing because of a second derivative on the concentration gradient.
So therefore, the continuity equation for electrons is given by this relation here. So, dn/dt
has got components due to drift, components due to the diffusion currents, components
due to the generation of electrons and components due to the recombination of electrons.
And we have already looked at how to compute the recombination term for different
cases, when we studied generation recombination mechanisms.
294
Now, the same exercise can be performed for holes as well except for a few minor
differences. The first difference is that this charge is positive for holes and the second
difference is that the drift and diffusion currents; you know as we discussed earlier are
opposing each other in the case of holes so, you have a negative sign. So, you allow for
these two corrections and you will find that the continuity equation for holes is given by
this expression here.
So, now we can run through some exercises ok. So, we will take the continuity equation
it is a very useful expression, because its telling us the dynamics, it is giving us a
description on the dynamics of the electron and hole population in any given volume.
And we will use it for a special case and the special case is that. Let us only worry about
the continuity equation for minority carriers with low level injection. And we are
interested in the special case because this is something we will be using later down the
line ok.
So, when we say; minority carriers with low level injection they are going to make a
further approximations ok, we are interested in this very very special case, we are going
to apply this general relation. So, this relation is general it is very useful and we are
going to use this relation for this very very special condition which is minority carriers,
low level injection, no electric fields so, your E is 0 ok. And the condition where your
295
equilibrium concentrations do not change with space; that means, your n o and p o are
independent of space ok, they do not change its space ok.
So, using these conditions we will first establish that n is your n o plus delta n. So, this is
the low level injected carriers. And since these are minority carriers we are talking about
the electrons in a p type material so, this is a p type material and we are talking about
electrons in a p type material. So, dn/dt since, n0 is the equilibrium carrier concentration,
dn0/dt is a constant. Therefore, dn/dt is similar to d∆n/dt dn/dx is same as d∆n/dx
because, n0 is not going to vary with x right.
And if you take your general continuity equations so, this is my continued equation with
all the terms in it ok. That is a general equation and we are going to apply all these
approximations and all these conditions that we have. So, my dn/dt becomes d∆n/dt, my
electric field is 0. Therefore, I get a 0 here and I get a 0 here as well. My diffusion
current component does exist, but the n can be replaced with ∆n simply because of this
relation.
And I have a some generation mechanism and my recombination, is the recombination of

minority carriers which is given by this term. So, if you remember the R G mechanisms;
we studied it, we solved for a very special case ok, which was the recombination rate for
electrons in a p type material. And we had found that that recombination rate was delta n
by tau n. So, that is the recombination rate that we are using so, this is the special case
that we want to study.
So, now we have a new continuity equation, which is essentially got only these terms.
So, there is no electric field and there is a it is we are looking at minority carriers with
low level injection. So, let us see what this continuity equation tells us how do we solve
for different conditions we just go through several examples.
296
So, example one so, let us say we are at steady state and there is no light being thrown on
the semiconductor. So, this is my continuity equation ok. This is all assuming everything
if we discussed in the previous slide which is low level injection, no electric field nopo
are not varying with space. They are constant in x and with all these conditions and we
are talking about minority carriers. So, it is be received we be obtained from the general
continuity equation we obtained this new continuity equation under these conditions. So,
that is what we are going to use now.
297
And we are going to use this under the condition where in the first example we are going
to say we are at steady state and there is no light. So, steady state implies this derivative
has to be 0. So, now we are looking at the total change in carrier concentration so steady
state implies this derivative is 0.
And diffusion we have not made any comments on the diffusion current so, therefore, we
will retain this term there is no light. So, therefore, they cannot be any generation
mechanism and there does exist a recombination mechanism. So, this is the modified
continuity equation for this particular example and you can solve this. So, this is a very
useful relation because we will use it when we discuss our p n junction diodes ok.
And the solution to this equation so what we have here is an equation that looks like this.
And this is a differential equation that we have solved several times before you know we
look at these differential equations when we looked at the case of particle in a box in
quantum mechanics for example, ok. And the general solution takes up this form you got
a you got Ae-. So, suppose this was some parameter A say some parameter y, then it
would the general solution remain x/√y+Be^(x/√y) ok.
298
So, so here we have this product of Dn and 𝜏𝑛 and therefore, this happens to be the
general solution for this differential equation. Now this product the square root of D n
tau n is a very useful parameter, it is something called as a diffusion length. So, what is it
saying? It is saying that the electrons which are the minority carriers and p type material
can diffuse, they are diffusing you have some high concentration here they are trying to
diffuse out ok.
And they are going to last for about tau n seconds, that is the time they are going to last
before they recombine. So, it is telling us it is giving us a measure of the spread,
diffusive spread of the electrons before the electrons have recombined. So, it is a length
which is called as the diffusion length of the electrons. And similarly if you have 𝜏𝑝 it is
the diffusion length of holes, when the holes are the minority carriers, which is holes in a
p in a n type material.
So, this is the general equation and its it is worthwhile to remember this and by applying
different boundary conditions based on your situation you can calculate what B and A
are ok. So, what this equation tells us is for this particular example; delta n is spread out
in space in this particular manner.
299
So, let us take another example, let us take the example of no concentration gradient and
no light ok.
So, this is our general equation, if you have no concentration gradient here is no
diffusion. So, that term is 0. If there is no light your G n is 0, but we have not made any
conditions any assumptions on whether steady state or not. So, we will keep a time
derivative there and that is your recombination rate. So, the solution to this is simply this
so, you can see how the continuity equation can be used to obtain quantitative estimates
of what is happening to the minority carriers in space and time for different conditions on
the semiconductor.
So, just as another example, just to complete things we will assume a case of steady state
and no concentration gradient. So, steady state implies the time derivative here is 0,
concentration gradient there is no concentration gradient implies there is no diffusion.
And therefore, you only have generation and recombination which means that such a
situation is sustained only delta n is the generation rate into the lifetime of these carriers.
So, that is essentially what so this, these words actually imply this quantitative result ok.
So, with that we will conclude our discussion on the continuity equation. And we are
now in a position where and we can estimate the behavior of the semiconductor in
equilibrium and out of equilibrium. And, we will now use this knowledge as the basis for
all our discussions henceforth.
300
Lecture – 19
Junctions
We will now look at the next unit as far as this course is concerned, and this unit is called
Junctions ok. And by junctions we mean the merger of 2 different materials ok; which
could be a semiconductor and a metal or a metal, and a metal or 2 different
semiconductors ok. And, in particular we want to look at what happens; we will use all
the concepts we have you know established so far in the fundamentals of
semiconductors. And we will use those ideas to see how a material should behave when
they are adjoined to another material ok. And it gives you a lot of interesting physics and
some interesting ideas for engineering.
So, in high school I think several of you might have heard of p n junctions. So, p n
junctions is one kind of a junction. But we will be looking at metal semiconductor
junctions, p n junctions, junctions’ calls to create the other devices or such as transistors
etcetera et cetera. And the purpose of this lecture is to establish some very key ideas
some very key concepts that we will be using time and again a when we want to analyze
and understand how junctions behave ok.
301
So, the first concept has got to do with a drawing something called as a band bending
diagram. Now this is a very useful idea ok, it is really not compulsory or necessary, but it
is a very nice graphical representation of what happens near a junction ok. And how the
semiconductors are the materials involved will be influenced by the creation of such a
junction. Again, we will we will try to understand what band bending means, we will we
will work out some examples and we will get those concepts clear, but before that let us
try to look at some of the properties of different materials.
So, let us say we are trying to create a junction between say semiconductor one and
semiconductor 2 ok. So, we have we have these 2 semiconductors separately initially ok.
We know how the behavior thermal equilibrium, we know how they behave they throw
if we throw light on it, we know how behave if you establish a current if you have
establish a voltage across it etcetera.
So, you have studied all this with respect to a semiconductor and thermal equilibrium
and in non-equilibrium. But now what we are trying to do is, see what happens when you
bring these 2 semiconductors in contact with each other. And this is what is called as the
junction between the 2 semiconductors. We want to see what happens with what is the
physics in and around the junction.
So, before we establish all these concepts, let us look at some of the key properties ok.
So, there are some pointers ok. So, I have just labeled these out as key pointers, and we
will just go through these aspects one by one and be these aspects will get more and
more you will familiarize yourself with these aspects as we go through this course as you
solve what more examples.
So, the first thing is let us take semiconductor.
302
So, you are all familiar with you know the conduction band edge, the valence band edge
and the location of a Fermi level. The semiconductor will also have an intrinsic Fermi
level position which is the Fermi level when the semiconductors pure. But here as you
can see the Fermi level is offset and it is closer to the conduction band which is
indicative of the fact that the semiconductor has been doped n-type ok. So, nevertheless
for any semiconductor you have your valence band, you have your Fermi level and you
have your conduction band.
Now, we define a reference energy level called as E vacuum the vacuum energy level ok.
It is just a reference energy level. And we establish 2 terms ok, the first is called as a
work function which is the term which many of you might have heard in with respect to
metals and with respect to the photoelectric effect. It is the same work function that we
are talking about. And something called as the electron affinity ok, which I think
students from chemistry might be more familiar with.
So, we have the work function as being the difference between the vacuum energy level
and the Fermi level of the semiconductor ok. And we typically represent the work
function by this symbol phi, the Greek alphabet phi, and for a semiconductor I will use
the subscript S and for a metal I will use the subscript M. Now phi is the potential is the
potential, and the energy is simply q𝜑 ok. So, if you know this potential as say so, if you
know the energy in terms of let us say the q𝜑 is equal to say 0.1 electron volt; it simply
303
means 𝜑 is 0.1 volt ok. So, the energy and potential are easily related in any picture like
this by simply multiplying the potential with the charge.
So, the energy level, the energy difference between the vacuum energy level and the
Fermi level is essentially the work function of the semiconductor. And the difference
between the vacuum level and the conduction band edge is the electron affinity of the
semiconductor. And we will represent the electron affinity by this symbol, I think it is
called chi and for a semiconductor we will say with a subscript S. So now, for a metal so,
in the case of a semiconductor, you must always remember that the electron affinity is
always a constant, it is a material property.
So, you can engineer the device in any manner you like you can apply whatever voltage
you like, you can you can we can play around with the engineering as much as we want.
But the electron affinity is always a constant. So, you cannot really change this energy
difference between the vacuum energy level and the conduction bandage. So, that is a
useful pointer to remember which is that the electron affinity for a semiconductor is
fixed.
Now, in the case of a metal so just before we head to a metal; what about the work
function of a semiconductor? Is that fixed; clearly the work function of a semiconductor
is quite variable, because it depends upon the doping, right. If I dope it p-type, I have
changed my work function. And if I dope it slightly more n-type I have moved the Fermi
304
level closer to the conduction bandage and my work function has changed ok. So, the
work function for a semiconductor really isn’t a constant, whereas the electron affinity is
a constant ok.
Now, for a metal, the metal does not have a conduction bandage and a valence bandage,
instead a metal just has the Fermi level.
And the work function for a metal is the energy required to take the electron from the
Fermi level and into vacuum ok. So, is again the difference between the vacuum energy
level and the Fermi level of the metal, but for a metal the work function of a metal is a
constant ok. So, in other words for a semiconductor the electron affinity is a constant.
And for a metal the work function is a constant. The work function for the
semiconductor depends upon the doping. So, these pictures are clear then you are one
step closer to drawing a band diagram. Can I, and I will show you what a band diagram
is and how we can draw it.
So, essentially what do you mean by band diagram? So, what we are trying to see is if
you bring 2 different materials closer and create a junction between these 2 materials. So,
you create a junction between these 2 materials it is very possible that you know you
have carrier movement across the junction simply because the Fermi levels are not
aligned ok. So, you can think of it this way. So, if you remember a while back had used
305
the water bucket analogy ok and I told you the energies helpful, but it is not really going
to take you all the way ok. But let us get back to that it is going to help us a bit here.
So, let us say you have 2 buckets ok. And you have water fill different levels in these 2
buckets. Now we had already discussed you know the water molecules above this level
do have a Boltzmann’s distribution. And therefore, we are quite you know we are quite
close to what the Fermi level in a semiconductor is you know, they are using this little
picture here.
So, you have a lot of water molecules here, you have plenty of water molecules below
the Fermi level here, and this was the Fermi level for this bucket and this is the Fermi
level for the other bucket. Now while creating a junction what we do is we bring these 2
buckets close to each other and we simply remove the edge between them ok.
306
There is no wall between these 2 buckets. So, what is what is going to happen? So, what
will happen is the water molecules will not like to see these 2 different levels that is
clearly a concentration gradient here. And therefore, you will establish a diffusion of
electrons, or in this case water molecules from the bucket that is got a higher water level
to the bucket that is got a lower water level. So, the electrons or the water molecules are
flow out till the water level is the same in both buckets ok.
So, in this case water molecules have transferred from there to here in order to align, or
in order to make the water level the same. You will not end up in equilibrium; you will
not end up in a situation where there is a difference in the water level within space ok.
So, the water will flow and it will ensure that the level which was equivalent to a Fermi
level is constant throughout this region ok. A similar thing happens here ok. So, here too
we will have a Fermi level alignment. And if you create a junction between 2 different
materials, the Fermi level tells you is an indicator of the energy at which the probability
of you finding an electron is half; electron occupying a state is half ok.
So, that probability is half at this point for this material and that probability is half for
that point in that material, which is which in some sense implies that there are larger
number of electrons sitting here as compared to this material. And just like in the case of
a bucket with water, if you bring these 2 materials together and you create a good
junction; which means, very nicely getting rid of any edge between them ok.
307
So, do not take do not take that will previous statement extremely literally. So, what the
way junctions are created is through a very different process, and it is not equivalent to
you removing the wall between 2 bucket. So, please do not take that statement too
literally. So, we were to create a good junction, then the electrons will flow from the
material having a higher Fermi level to the material having a lower Fermi level which
means the material having a lower work function to a material having a higher work
function. So, what do you mean by band bending diagram?
It is that we want to see what happens to the conduction band, valence band and Fermi
levels of the 2 materials when you create a junction.
So, when I create a junction we will have electrons moving back and forth and you know
electrons moving from one material to another, and you know these bands might change
their location ok. They may not want to be where they are right now ok. And that is what
we wanted that is what we want to look at in a more quantitative manner. So, that is what
we mean by band bending. And I will soon bring up an example. So, those things will
become clear. So, the first point in order to understand what happens to the conduction
band edge valence band edge and the Fermi level, is to note that the Fermi levels will
align the moment you create a junction. That is the first point. The second point is
something that we have already discussed which is the work function for a metal will
always be a constant.
308
And the electron affinity for a semiconductor will always be a constant ok. So, these 2
will not change so, remember that. And the next thing is very far away from the junction.
So, let us say you have a very long piece of material 1 and you have another long piece
of material 2, and you created a junction ok. So, these are long and we can soon define
the length scales you know more accurately, but we create a junction between these 2
materials. Now you have a lot of interesting things happening in and around the junction
ok, you have the bands moving shifting away from where they were, you have electrons
moving across it stretched.
But very far away from the junction ok, the material really does not see any of these any
of these phenomena. And very far away from the junction the material could be said to
live in it is own world, and you know retain the properties, that it had if it were to be in
equilibrium. If the entire system is in equilibrium that the material really does not see
any change ok. And very far from the junction nothing interesting happens.
So, then while drawing a band bending diagram, we watch for how the electrons move
about. So, this is exactly what I described with the water bucket analogy ok; which is to
see you know which material transfers electrons ok. And that will help us gauge the
bending of the bands in a semiconductor. Finally, in order to analyze you know the
degree of bending and the shape of bending, and you know what the electric fields are
and what the potentials are inside the junction, we solve Gauss’s law.
309
And you know we solve the differential form of Gauss’s law which is something called
as the Poisson’s equation ok. So, these are several pointers that we have I have thrown
out quite early ok. Any is a things to remember and we will discuss each of these in the
next few minutes. And we will soon see how to you know how these can be used to draw
a band diagram ok.
So, let us go to the very first point, which was the alignment of the Fermi level ok.
So, I stated that you know I gave this analogy of the water bucket ok, the bucket carrying
water. So, one bucket had a different level the other bucket had a had water at a much
shallower level ok. And when we created a junction we just removed the wall separating
these 2, and we said that the level of the water would align. And this was sort of they sort
of built upon the use of this analogy a while back when we were discussing you know
the Fermi function and the Boltzmann’s distribution etcetera. But now let us try to
mathematically prove this, the alignment of the Fermi level ok.
So, let us take 2 different materials.
310
Let us say there is material 1 and that is material 2, and let us say material 1 has got some
density of states which is given by g 1 and material 2 is got a density of states given by g
2, and depending upon the location of the Fermi level ok. So, this material 1 has got a
Fermi level located at E f 1 and material 2 is got a Fermi level located at E f 2. And
depending upon the location of this Fermi level, material 1 could have a Fermi function
called f 1, and material 2 will have a Fermi function f 2.
1 1
So, f 1 would be something like 𝐸−𝐸𝐹1 . And f 2 would be of the form 𝐸−𝐸𝐹2 . And to
1+𝑒 𝑘𝑇 1+𝑒 𝑘𝑇
continue on, let us say there are n 1 free electrons and material 1, and n 2 free electrons
and material 2, and p 1 vacancies in material 1 and p 2 vacancies in material 2. And now
what we do is we are trying to bring these 2 we try to bring these 2 materials together
and form a junction.
So, before we do that let us try to get a gauge on what are n 1 and p 1. So, what is n 1? N
1 is calculated from the probability of occupancy of all these states in material 1 so, it is f
1g1, and n2 is f2g2. And what is the probability of non-occupancy? That is the
probability of finding holes it is, 1- f1g1 and in p in material 2 it is 1-f2g2. So, these are
the parameters that we have in material 1 and material 2. So now, let us say that we try to
create a junction between these 2. Now we bring material 1 and material 2 together and
we have established a junction.
311
Now, at thermal equilibrium so, we created this junction and we left it in dark, we did
not apply a voltage, we did not apply light. We left these 2 material and these 2 materials
with the junction in the dark ok. And what would happen? In the dark there cannot be
any current ok, you cannot have current. If you have a current you can establish, you can
do work and you have probably solved the energy crisis; so in the dark, any flux from
any flux across the junctions. So, let us call that the junction ok. So, between these 2
materials that is one that is 2. Any flux from left to right across this junction should be
matched by a flux of carriers from right to left across this junction ok. These 2 should be
the same otherwise you will have an effective current.
So, what does the flux from left to right depend on? It depends upon the number of
electrons I have a material 1, and the number of vacancies I have in material 2, because if
there are no electrons in material 1, there is no current from material 1 to material 2. And
even though there are electrons material 1, if there are no vacancies in material 2, the
electrons cannot fill any states in material 2. And therefore, there will not be any current
ok. So, the current from 1 to 2 depends upon n 1 and n 2 in fact, it is the product.
So, if I want to match the flux, the flux from left to right depends upon this product n 1 p
2. And the flux from right to left depends upon the product n 2 p 1. So, you need to have
electrons from the second material trying to occupy the vacancies in the first material.
So, at thermal equilibrium, they cannot be a current and therefore, these 2 flux, flux is
must balance each other, and n1p2=n2p1.
So, let us write out n1 and p2 in terms of the density of states and the Fermi functions
using these terms that we described here. So, n 1 is f 1 g 1 and p 2 is 1-f2g2 and so on so
forth. And if you were to solve this equation out you will find that it says that f 1 is equal
to f 2. And f1 =f2 automatically implies that at thermal equilibrium Ef1 will be equal to E
f 2. So, in other words, if you were to create a junction between these 2 materials, and
keep this system at thermal equilibrium, the Fermi levels will align and Ef1 will have to
be equal to Ef2 ok. So, this is the little hand wavy proof that the Fermi levels of these 2
materials will align when you create a junction.
312
So now let us start creating a band bending diagram ok. So, this is a very useful concept
as I told you. And I think it is best done we take help of an example. And I will create a
band diagram for 2 material 2 I will consider 2 examples, because we will very soon be
using these examples. So, you might as well just take this opportunity to consider these 2
cases ok.
So, the first band bending diagram will be between will be for a junction created with a
metal and an n-type semiconductor. So, you have a metal that is going to form a junction
with an n-type semiconductor ok. And we will choose this particular condition, that the
work function of the metal, so q𝜑𝑚 phi m. So, it is all if you consider everything in terms
of energy. So, it is all got a q there and phi M is simply the potential. So, q phi M is the
work function of the metal is greater than q phi S. What does that mean? It means that
the Fermi level of the metal is lying further away from the vacuum energy level as
compared to the Fermi level of the semiconductor.
So, the Fermi level of the semiconductor is closer to the vacuum energy as compared to
the Fermi level of metal. Or in other words, the semiconductor has got a Fermi level that
will lies above the Fermi level of the metal ok. And it is a semiconductor that is going to
donate electrons to the metal, if we create a junction because that is what we looked at in
the water analogy ok. So, this is the condition.
313
So, the first thing we need to make sure is if we create a junction the Fermi levels will
align. And the second thing to make sure is that the work function of the metal does not
change and the electron affinity of the semiconductor does not change ok.
So, let us say we create a junction between these 2 materials. So, the first thing we do is
let us align the Fermi level, we keep the Fermi level flat ok. So, you can see that the
metal and the semiconductor now have the same Fermi levels. So, we are all at thermal
equilibrium so, all at thermal equilibrium. So, if we go out of equilibrium the Fermi
levels need not be aligned. Because then there is no reason for the flux from the left to
the right to be matching the flux from the right to the left. So, is all at thermal
equilibrium.
So, the moment you have a metal n-type semiconductor junction with phi M greater than
phi S, the Fermi level line, the vacuum level will adjust itself; so as to ensure that the
electron affinity in the semiconductor and the work function of the metal is always a
constant. So, you can see the work function of the metal has never changed the work
function is the same ok. The difference between the Fermi level in the metal and the
vacuum level has is constant throughout. But since the semiconductor had a higher Fermi
level as compared to the metal, the semiconductor will now donate electrons to the metal
ok.
314
So, you will have in the metal semiconductor junction with phi M greater than phi S ok.
So, you had an n-type semiconductor you have a metal here, and you have an n-type
semiconductor, and it had a Fermi level that was located higher as compared to that of
the that of the metal, and electrons will effectively flow from the semiconductor to the
metal upon the creation of a junction. So, what do you mean by electrons are going to
flow from the semiconductor to the metal?
So, let us look at close to the junction, let us look at this region close to the junction. So,
the semiconductor has now lost electrons, in the case of this it initially had the
equilibrium electron concentration, you know, it had some n o which was approximately
ND, and now the semiconductor has lost electrons because it is donated it to the metal in
order to align the Fermi levels. So, if you have to if the semiconductors lost electrons, it
implies that the conduction band has moved away from the Fermi level position, because
what is the electron count the electron the electron concentration is n c e to the power
minus E c minus E f by k T.
So now, as this electron concentration goes down, the E c minus E f must increase ok.
Therefore, since this region has got fewer electrons as compared to the bulk ok; so this is
the bulk region, very far away from the interface, the conduction band edge will move
away from the Fermi level ok. So, it is no longer going to retain it is position here. Close
to this is going to move far away and then just a little after that it is going to be located
there and there and then therefore, you have a gradual bending of the conduction band
edge with respect to this Fermi level.
Now, the energy gap in the semiconductor hasn’t changed ok. So, therefore, the valence
band edge will also move in accordance. Now very far from the junction; so, let us look
at this region, there is no change here. It is as though you had a pristine semiconductor
right it you look at this condition here. And this condition here you still have an n-type
semiconductor. So, this is the Ec-Ef, it is n doped because the E f is lying above mid gap,
that is your band gap, that is your E v E c E f and that is your E vacuum with the electron
affinity being constant. So, nothing has happened. This region far away from this
junction hasn’t really seen anything interesting, but it is only close to the junction that we
have lost some electrons.
315
And as you move deeper, as you start walking from the junction and into the
semiconductor, as you get deeper and deeper, you will find that the movement of the
conduction band away from the Fermi level is sort of gradually decreased, till you reach
a point where you are in the bulk ok. So, this region here is called as the bulk, it is a
region that is not seen anything interesting, because it is too far away from the junction.
And it is only this region here that is influenced by the creation the formation of a
junction.
So, there are some simple ideas we will again come back to drawing a band bending
diagram when we talk about metal semiconductor junctions. And throughout the vacuum
level, I know as I mentioned it adjusts itself ok. So, as to keep the electron affinity
constant and the work function of the metal constant, but note here that the work
function of the semiconductor need not be constant, again it is not required ok.
So, this is the way you construct a band bending diagram. So, essentially what we do? Is
you first align the Fermi levels and you just mark out which of the materials is going to
donate electrons, where is electron going to flow from? Is it from the right to the left or
left to the right?
316
And depending upon which material is lost electrons and which material is gained
electrons, we adjust the band position the E c minus E f location in these 2 materials ok.
And therefore, you get a bending of the bands in space. So, this is a energy distance
diagram. So, the x axis is distance the y axis is energy, and it is also indicative of
potential. So, if you want to locate the potential difference between these 2 points, it is
simply the energy difference between those 2 points divided by q. So, that is the potential
difference ok.
Now, several other points, you know the band bending diagram is always drawn with
respect to electrons ok. And what I mean by that? Is if you see a slope here. So, for
example, if you see for example, let us take this region here ok, you see this bending of
the bands. You see the bending of the conduction band. Essentially, what this means is
that the region that is lower located at lower energies have got a higher potential ok, as
compared to the regions located at higher energies.
So, if I have a Fermi level and I increase the potential of that so, let us say there is a
metal and it is a Fermi level here. And I increase the potential of the metal, I increase the
voltage the Fermi level will start moving downward. And therefore, since this is the case
if I have an electron and a hole ok, the electrons would like to run down the hill, when
you see a bending and the holes would like to run up the hill ok. So, this gives you a very
intuitive feel for what the electrons will do and what the holes would do ok.
317
Now, since it is an n-type semiconductor there are not many holes, there is quite a very,
very few, and whatever majority electrons whether I have all moved in here into the
metal and essentially any electron coming in here will try to run down the hill and away
from the junction. So, they are going to be very few electrons here and very few holes
here. And therefore, in some sense this junction can be said to be sort of depleted of free
carriers ok. So, we will come back to these points later. So, these are the this is the way
you draw a band bending diagram, that is the way you look at the different aspects of it.
So, just to complete this let us go to another example.
So, let us take a p-type semiconductor and an n-type semiconductor. Now let us try to
create a junction. So, this is your very well-known p n junction ok, it is the same
semiconductor. So, here that is my electron affinity that has to be constant. Again if you
want the energies it is q𝜒, and the Fermi level for a p-type semiconductor is located
closer to the valence band edge. And the Fermi level for an n-type semiconductor is
located close to the conduction band edge, and here you have your 2 different
semiconductors. And now we try to create a junction.
So, let us try to think about what will happen before, we actually see we see the next
figure. The first thing is the Fermi levels will align ok. So, let us align the Fermi levels.
Now which material do you think will donate the electrons to the other? So, the Fermi
318
level here is higher than the Fermi level in a p-type ok. So, it is very likely that the
electrons will flow from the n-type into the p-type once we create a junction.
So, the electrons flow from the n-type and into the p-type, what do you think will happen
to the conduction band edge, and the with respect to the Fermi level in the n-type
material. So, since the region near the junction will lose electrons, the conduction band
will have to move away from the Fermi level position ok.
So, in the n-type material: so, this is the Fermi level that is the junction it is expected that
the conduction band edge move away from the Fermi level position. And in the p-type
material, the p-type material has now received electrons ok. It had any a small number of
electrons initially, but now it is received a lot of electrons from the n-type material.
These electrons have all diffused in from the n-type material. And therefore, this region
near the junction has now got excess electrons and therefore, the conduction band would
like to move closer to the Fermi level in order to reduce the E c minus E f gap. And
therefore, the p-type material will see a band bending that looks like that; where the
Fermi level sorts of bends downward where the where the conduction band bends
downward to meet the Fermi level in the p-type material.
319
Therefore, we can construct a band diagram that looks like this.
So, you can see that the conduction band edges moved away from the Fermi level in the
n-type and it has moved towards the Fermi level in the p-type. Therefore, giving you a
band bending that looks like what is shown. So, this is the method to construct the band
diagram. So, throughout you will find that you know the electron affinity is always
constant.
320
So, the final point you know before we you know stop this study of this overview of
junctions, and actually start looking at some certain very specific junctions; is to
understand something called as a Poisson’s equation ok. Now we did draw all this band
bending. So, let us say it was a p-type and n-type semiconductor we said that you know
the band bending would look like it would look like that. The question now is how do we
exactly and you know very quantitatively determine the shape of this bending you know,
what is this ok. So, that is x so, what is the relation between the potential and the electric
field etcetera with respect to as a function of x.
So, you want to establish this quantitatively. We want to know the exact shape of this
bending, because these quantitative results can help us estimate many things such as
currents and charge concentrations etcetera in a more accurate manner. So, in order to
establish this, get this quantitative estimate of the amount of bending. We need to solve
Gauss’s law ok. And we want to particularly solve the differential form of Gauss’s law
ok, where we use the differential form of Gauss’s law which something called as a
Poisson’s equation.
And here is a very brief understanding of that. So, Poisson’s equation simply says that
the change of the electric field in space. So, this is the one dimensional Poisson’s
equation. It is simply equal to the charge concentration per unit volume ok. So, this is the
charge per unit volume ok. It is not per unit area; it is not the total charge the charge per
unit volume divided by the permittivity of the material in which we are doing all this
analysis.
So, in our case it is mostly going to be the permittivity of the semiconductor. So, that is
the electric field that is the distance that is the spatial coordinate x, rho is the charge
concentration per unit volume and epsilon is the permittivity. And since the electric field
is minus -d𝜙/𝑑𝑥 which is; where phi is the potential ok; we can also rewrite Poisson’s
equation as d square phi by d x square is equal to minus rho by epsilon. And this is a
relation that we will use and we will look at some examples soon on in order to establish
the correct nature of the bending of bands ok.
With that we will close our overview on you know how to create you know what a
junction does, and how to create a band bending diagram, and you know, you know what
the Poisson’s equation is. And we will now start looking at very specific junctions and of
321
interest are the metal semiconductor junction and the p n junction. And we will start
looking at these junctions in order to establish our understanding of the different
semiconductor devices.
322
Lecture – 20
Metal Semiconductor: Junctions
So, so far we have looked at creating junctions, and you know some of the basic ideas
behind how to draw a band diagrams. So, we will now start a formal approach to the
metal semiconductor junction. So, a metal semiconductor junction is quite important
because if you think of any device.
So, let us say you are building a transistor or you are trying to build a device wherein you
will not, you know have these metal contacts and you want to actually apply a voltage or
you know send a current through this device. So, this is your semiconductor device.
Now, how do the properties of this metal contact and the semiconductor affect the
current voltage characteristics, ok? Ideally what we would like is we want to think of
these contacts is very ideal contacts, they should offer no impedance and they should
very faithfully transmit any voltage applied to the metal and on to the semiconductor, but
that is not really true. And in fact, the depending on the kind of a contact you make
between the metal and semiconductor various things could happen.
323
So therefore, the study of having a metal and interface to the semiconductor, that is to
create a metal semiconductor junction is quite important. And what we will do here is
first look at and get some qualitative estimate as to what the different metal
semiconductor junctions might look like depending upon the work function differences
between the metal and semiconductor, and draw the band diagrams for these kind of
contacts.
And then we will get into a bit of detailed analytical calculations on estimating a
impedance and the particularly the capacitance of a metal semiconductor junction, and
also the current voltage characteristics in a metal semiconductor junction, ok.
So, first let us start with, you know trying to draw some band diagrams ok. So, first let us
take the example of a metal and an n-type semiconductor where the Work function of the
metal is greater than the Work function of the semiconductor. So, when we looked at the
different materials before we noted the fact that the electron affinity of a semiconductor
never changes that is always fixed whereas, the work function of a metal is also a fixed
parameter. So, irrespective of the kind of contact you make the work function of a metal
and the electron affinity of the semiconductor will not change, at any location in space.
So, now let us try to redo this problem. You know we, we discussed we used this as an
example when we discuss the, brief study of junctions but let us come back to this
problem. Which is we have a metal which has got a work function of q phi M and we
324
have got a semiconductor which is n-type. So, it is an n-type semiconductor and it is got
a Fermi level slightly above the mid gap slightly above E i and it is got a work function
of q𝜙𝑠 phi S. And the key is in this particular case your phi m is greater than 𝜙𝑠 which
means that the Fermi level of the metal lies a further away from the vacuum energy level
as compared to the Fermi level of the semiconductor.
So, let us, see what happens when we make a contact between these two. Now using our
old analogy which is that of water in a bucket for whatever use that analogy is we do
realize that since the Fermi level of the semiconductors located above the Fermi level of
the metal its very likely that the semiconductor is going to transfer electrons to the metal
and the second thing is the Fermi levels will align as long as we are in thermal
equilibrium.
So, once you make this contact and approve and keep it thermal equilibrium the Fermi
levels align. So, that is the common Fermi level for metal and semiconductor. Far away
from the contact of the semiconductor does not see anything interesting, it is exactly the
way it would be and this is called as the bulk region, ok; this is called as a bulk and this
is called as the interface. And since the semiconductor gave up electrons, since the
semiconductor gave up electrons to the metal, the conduction band has to move away
from the Fermi level near the interface.
325
So, this region of the semiconductor has lost a lot of electrons. And therefore, the
conduction band edge at this region of the semiconductor moves away from the Fermi
level. Since n is equal to capital N c e to the power minus E c minus E f by k T. If when
n goes down it implies that this term increases because there is a negative sign this term
has to increase.
So, Ec-Ef has to increase and therefore, the conduction band bends away from the Fermi
level. So, the band bending diagram looks like this. So, you have the metal Fermi level
and at this point at the junction there is a massive barrier, ok and we call that barrier
height. That energy of this barrier is denoted by a special symbol which is called phi B
and that is the potential and q𝜙𝐵 phi B is the energy. And after this, there is bands bend
away. So, you can see this gradual bending of the band and to a point beyond which the
semiconductor does not see any band bending.
So, the region of the semiconductor which does not see any band bending where, where
in the semiconductor does not even realize there is a metal semiconductor junction it is
called as the bulk semi bulk of the semiconductor ok. So, when we say bulk we
automatically imply the, the region of the semiconductor far away from the junction and
if you look at, if you look at this contact ok. So, let us, let us think of it let us assume that
we have established this kind of a contact in order to study a semiconductor, in order to
electrically excited semi conductor by sending in some a current and some voltage. Now
how good is this contact?
So, think about the majority carriers the semiconductor. So, we have electrons in the
semiconductor there is an n-type semiconductor, we have plenty of electrons and very
few holes ok. And you think of the electrons in the metal ok. So, let us say the electron in
the metal has to get passed in get into the semiconductor, what does the electron see, it
sees a massive barrier. So, as we discussed earlier the band bending diagrams are drawn
such that the higher potentials lie at lower points ok.
So, what I mean by that is let us say you have a, you have a Fermi level ok. So, let us say
you have a metal and you got a Fermi level and you start increasing the voltage of this
metal. So, what would happen is that the Fermi levels start moving downward ok. So, if
you see two energy levels in an energy band diagram then the level that is lower has got
a higher electrical potential.
326
So, that is just the way it is drawn. So, that is this convention its drawn thinking of the
electron as rolling downhill, ok. So, if you have electrons the electrons would like to roll
downhill towards the higher potential and the holes would like to roll uphill towards the
lower potential ok. So, you can think of electrons trying to roll down here.
Now, if the electrons, the electrons in the metal they see a massive barrier here, it is like
a huge cliff, it is not very friendly for an electron that wants to get across ok. So,
therefore, it is not going to help the conduction, it is not going to be very easy to have
straight forward drift conduction. On the other hand what about the electrons in the
semiconductor the electrons in the semiconductor also see a barrier right. So, not as steep
as that of the metal, but they do see a barrier because they have to now roll up the hill in
order to get into the metal.
So, therefore, this kind of a contact is not going to give you an Ohmic nature in the sense
that an ideal contact. So, you would have a metal and you would have a semiconductor
and if you want to model this ideal contact, we would at most want to see a resistor that.
So, that is about; so we would like a contact to be like this. So, that any voltage I apply in
the metal is going to have and let us say I send a current through this contact, there is
going to be a small voltage drop across this contact but generally the I V characteristics
is linear, it is going to behave like a small resistor. So, that is the contact we want but
here that is not the contact we are going to get ok. In fact, this contact has a rectifying
nature ok.
So, I will tell you what that means when we look at the current voltage characteristics it
is, it is got a Rectifying nature and it is something called as a Schottky contact or a and
this barrier is something called as the Schottky barrier. So, it is not a contact that is going
to give you an ohm, ohmic relation between current and voltage, ok but nevertheless that
is the contact you get.
Now, what happens is in reality becomes a semiconductor metal any interface with the
semi conductors got a lot of trap states because that is a place where you have cleaved
off the semiconductor and or that is the place where you have impurities and therefore,
you have a lot of, energy levels sitting inside the band gap.
327
So, because of that the Fermi levels are not free to move around the Fermi levels cannot
move around in the, as easily as described by the mechanics, we discussed for the band
bending. And therefore, you have something called as a Fermi level being pinned and
due do that you have a lot of interesting effects that come up and it so happens that more
likely than not you will end up with a contact that might be rectifying, ok. So, in many
practical cases this is a kind of contact you might actually get and that is the best you can
probably do.
So, how do we build good devices with such a contact because we already, we have
already seen that this kind of a contact is not going to be very friendly for electron
transport ok. So, you know because of that we will pay special attention to the Schottky
contact and look at the current voltage characteristics of this contact and study what the
impedances etcetera and think about mechanisms by which we can improve the nature of
this contact; so that we can use such a band bending to our advantage so but anyway.
So, now let us proceed and try to look at all the other possible combinations. So, this is
the n-type semiconductor with the metal with phi m greater than phi s. So, that gave you
this kind of a band bending diagram. So, let us take a look at the band bending again. So,
the bands the E c the conduction band edge moves away from the Fermi level and it
slowly bends back downwards and approaches the bulk condition right.
So, till this point we could say that there is band bending and beyond this, the
semiconductor does not see anything interesting. Now if you think about this region here,
this region in the semiconductor here, what does it have? The region initially had a lot of
electrons, because it is an N Doped semiconductor but after the semiconductor
transferred electrons to the metal during the contact formation. And after the creation of
this band bending, you will see that this region does not encourage or does not have too
many electrons, at least particularly very close to the interface. So, the electron count
here is now sparse, the hole count is anyway very sparse, because it is an n-type
semiconductor, ok.
Therefore this region is in some sense void of free carriers ok. Now void is very
powerful word, it does not mean that there are absolutely no free carriers, what it means
is that the free carrier count is much smaller compared to the static charges present in
this region? So, what are the static charges?
328
So, since it is an n-type semiconductor, we had initially doped the semiconductor with a
lot of donor dopants right. And at 300 Kelvin or the semiconductors extrinsic and all the
dopants are ionized and what we have is fixed positive charges in this n-type
semiconductor. Now these charges cannot move they are a part of the lattice, and they
are all fixed and these ionized donors they gave up an electron which was all which had
become a part of your free carrier concentration.
Now, this donor dopant concentration, these ionized charges are how many of them are
there we have N d per unit volume of this ok. So, that is the number of ionized charges
that we have. And since the semiconductor gave up some electrons to the metal and since
the bands are bent in a manner that encouraged the electrons to stay away from this
region.
This region has got very few electrons as compared to the bulk, ok. Now the bulk has as
many electrons as Nd and this region has got much smaller number of electrons as
compared to N d. And in order to quantify the number of electrons what we need to do is
identify the E c minus E f at every location and we will know the exact count of the
electrons.
Therefore, since this region has got very few free carriers ok, holes are anyway low, this
region is said to be depleted and this region is called as the depletion region of a metal
semiconductor Schottky contact. So, this is a technical term and it is something that
people use quite often. Now if you make the assumption that this region has only fixed
ions, because of the donors, the ionized donors and it has got no or it has got very
negligible N and P. Then the assumption that is made is called as the depletion
approximation or full depletion approximation. Now how good is that approximation?
Near the interface it is probably a reasonably good approximation but as you start
moving towards the bulk, ok.
So, let us look at this region here look at the E c minus E f and let us look at this region
here look at the Ec- Ef, the Ec- Ef, here is say much larger than the Ec- Ef, in the bulk and
since it is an exponential dependence of on Ec- Ef, for the carrier count, we can say that
here the N is really very low and it is much lower than the N in the bulk, but as we start
approaching as the bands bend and the bending gets more and more gentle and as it start
329
approaching the bulk this approximation begins to fall off for example, we cannot
definitely make that assumption at this interface because we are, we are at the bulk.
So, you need to be very careful as to when and where these approximations are made.
Now these kinds of approximations are very helpful in sort of simplifying the
mathematics. Now in this course, we will consider both aspects we will first simplify the
mathematics to get a gauge on how the device behaves etcetera. We will make use of
such assumptions ok, but then we will also try to be very careful and point out how we
can calculate all these answers out with all the great details, right.
So, let us say we again have a metal semiconductor junction. And let us say the
semiconductors still an n-type semiconductor, but this time we will let the work function
of the metal will be less than the work function of the semiconductor which means phi M
is less than phi S, ok.
So, in the previous case we had 𝜙𝑀 greater than 𝜙𝑠 , now since 𝜙𝑀 < 𝜙𝑠 , the metal will
start providing electrons to the semiconductor ok. So, it is like two buckets with water
level higher here as compared to the other one, that is a semiconductor bucket and this is
the metal ok. So, I think this is the last time I will try to use the bucket analogy, we need
to refrain ourselves from heading towards that, ok.
330
So, if you create this junction, the metal will give the electrons to the semiconductor.
Once again the Fermi levels will align the metal will give the electrons to the
semiconductor. And since this interface is now got more electrons as compared to the
bulk, the conduction band will bend towards the Fermi level instead of away, ok.
So, the band bending diagram will look like this ok. So, you have the metal the Fermi
level is aligned in the bulk the semiconductor does not know what is going on it sees no
change or it is not seen anything specifically interesting, and therefore everything is as it
should be. And this is the bulk region but close to the interface since the metal pushed
carriers push the electrons into the semiconductor. The conduction band bends towards
the semiconductor, towards the Fermi level, sorry and the valence band follows that
bending because the energy gap is fixed.
Now, this means that this region has got excess, is got more electrons as compared to the
bulk ok. So, this is the electron count here is greater than the bulk. So, most definitely
this is the exact opposite of your depletion region, right. This is definitely not a depleted
region it is got a lot of free carriers and more interestingly you see that the bending is
now favorable for electrons to move through ok. The electrons can easily roll down the
hill and you know they do not see much of a barrier in either direction. And you have an
easy migration of electrons across from the metal to this semiconductor. This is very
much against you know what we saw, this is very much the opposite of what we saw in
331
the case of a Schottky contact wherein we saw a barrier of this kind, the metal the
electrons in the metal saw this cliff and the electrons in the semiconductor had to roll up
the hill ok, but here it is a very different scenario and this is the kind of contact that we
would ideally like ok.
So, if you were to make a contact between a metal and a semiconductor. And let us say
this is an n-type semiconductor and you make a contact with the metal whose got a work
function less than that of a semiconductor, then you should ideally end up with
something called as an Ohmic contact, because here the electrons can easily move across
through from the metal to the semiconductor and vice versa.
Now, let us go ahead and consider some more examples. Now you will again consider
metal semiconductor junction but we will consider a P type semiconductor.
So, P type semiconductor has got a Fermi level that is located closer to the valence band,
that is the major difference and let us take the case of phi M being greater than phi S. So,
this is your q𝜙𝑠 that is the work function of the semiconductor, this is the work function
of the metal. And since q𝜙𝑀 is greater than q𝜙𝑠 , the Fermi level of the metal lies further
away from the vacuum energy level as compared to the semiconductor.
332
So, now what would happen? So, this is a P type semiconductor, the majority carriers
here are the holes ok. There are very few electrons, very very few electron and a large
number of holes ok. So, you have heavily, it is a large holes of the majority carriers here.
So, if you were to create a junction, the first thing is of course, a Fermi level will build a
line, but then where do the electrons go. Since the Fermi level of the semiconductor is
higher than the Fermi level the metal the semiconductor will give electrons to the metal
ok. There are not very many of them, but the semiconductor will try to give electrons to
the metal.
So, what would happen in that case if the semiconductor gives electrons to the metal, E c
minus E f has to increase. And therefore, the conduction and way a conduction band
edge will try to move away from the Fermi level.
So, the band bending would look like this ok. So, you have your E f, the E f is aligned
but since the semiconductor gave electrons to the metal, the conduction band edge moves
away from the Fermi level and the valence band edge follows.
Now, what kind of a contact that is going to aid or charge transport or not, ok. Now if
these were to be an n-type material and you saw this kind of a barrier it is definitely a
Schottky contact but here the majority carriers are the holes. So, we should not worry so
much about the electron movement across this junction but rather look at the movement
333
of holes and the holes run up the hill and they see a very nice gentle slope, where and
they can easily move and migrate into the metal. So, what about the metal? So, the
movement of holes into the metal simply implies that the electrons at this interface
recombine therefore, leading to an effective electron flux in this direction.
So, there is good charge transport between the semiconductor and the metal and this is
once again an Ohmic contact. So, this is again a kind of contact you would like to have.
So, if you have a P type semiconductor and you want to make an Ohmic contact with a
metal you must choose a metal whose got a work which has got a work function that is
greater than the semiconductor all right.
So, this is an Ohmic contact with a P type semiconductor and far away in the bulk, you
see that it is still a P type semiconductor the Ef is close to Ev and the semiconductor sees
nothing interesting.
Finally, we will take the fourth of the cases ok. It is a metal P type semiconductor
junction but 𝜙𝑀 is now less than 𝜙𝑠 which means that you have your P type
semiconductor ok.
So, this is all these are all energies (Refer Time: 25:23) be q there you have a P type
semiconductor that is your Fermi level, it is close to the valence band edge, that is my q
334
phi s and phi M less than phi S implies at the Fermi level of the metal lies closer to the
valence band edge. And therefore, this is the situation.
Now if you create a contact you will align the Fermi levels as usual, but now it is going
to be the metal that transfers electrons to the semiconductor because the metal has got a
higher Fermi level as compared to the semiconductor. And therefore, the bands in the
semiconductor will now the conduction band, we will now bend towards the Fermi level,
because the semiconductors accepted electrons and there is a large electron population
here.
So, the band bending will look like this. So, you have the Fermi levels that are aligned
deep in the bulk, there is nothing interesting it is still your regular P type material but at
the interface the semiconductors increased to electron population. And you find that the
conduction band bends towards the Fermi level because of this increased electron count
but then what kind of a contact is this. So, since holes are the majority carriers let us look
at the, let us look at what the holes experience. The holes now see a barrier they, they
have to run down, they have to run down the hill ok.
So, in the previous case the holes had to run could run up the hill and into the metal and
it was a nice Ohmic contact, but now the holes have been asked to run down the hill they
do not like it and they see a barrier. And therefore, this is very much like your barrier or
a Schottky contact which we discussed earlier but the only thing is that the barrier is now
335
formed in the between the Fermi level of the metal and the valence band edge and that is
your barrier there, ok.
So, just to summarize, we considered four kinds of contacts. We took a metal with an n-
type semiconductor and we considered the case of phi m greater than phi s and phi m less
than phi s. In the case of a metal, in the case of phi m greater than phi s, we found that
we ended up with a Schottky contact or a rectifying contact.
In the case of phi m less than phi s, the metal gave electrons to the semiconductor and we
ended up with an Ohmic contact. In the case of a metal P type semiconductor junction
when 𝜙𝑀 was greater than 𝜙𝑆 we had an Ohmic contact, because we are now looking at
what the holes are doing at the valence band and when phi m is less than phi s we ended
up with a Schottky contact.
So, that is a summary of the four kinds of contacts that we have established, and what we
will do from this point on is to study the electrostatics and the current voltage
characteristics in each of these contacts.
336
Lecture - 21
Schottky Contact : Electrostatics
So, now let us continue on and look at the Schottky Contact. So, we saw that there are
four there are two kinds of contacts that are formed in metal semiconductor junctions.
The first is the Schottky contact, which is you know which we call the rectifying contact,
but we really did not say why it is rectifying. And the second is the Ohmic contact. And
what we will do now is first take the Schottky contact as our object of study ok. And we
will always consider the Schottky and ohmic contacts for the n-type n-type
semiconductor.
And what we would like to do is firstly, study the electrostatics ok. So, we will study the
electrostatics of the junction. And then look at the current voltage characteristics, we will
then take the material out of equilibrium and study the current voltage characteristics.
And from these two you will effectively try to develop a small signal impedance model
for the device ok.
So, this is the first of the semiconductor devices and by going through this analysis we
will be in a position to then repeat this analysis for other devices. So, in order to; so just
337
to revise you know just to look at the Schottky contact again, you have a metal and you
have a n-type semiconductor so, this is your n-type semiconductor.
And we already saw that the Fermi levels will align, the semiconductor will give
electrons to the metal because your phi m is greater than phi s and this is an n-type
semiconductor. So, the electrons will move into the metal, and the conduction band will
bend away from the Fermi levels, and this will be the band bending diagram.
So, what we are trying to find out here is we are trying to solve out for the exact nature
of the variation of this bending. You know how does the energy bands with in space?
You know what are the electric fields in this region? What is the potential in this region?
What is the charge in this region etcetera, etcetera? So, that is the objective of our study
and that is what we mean by the study of electrostatics.
So, if you think of this region, we already discussed that you know the electrons are not
encouraged to exist here, because you know this region has given up some electrons to
the metal plus. And it has resulted in the bands bending in this particular manner which
encourages electrons to roll downhill and stay away and keep away from this region.
So, if you want to think about it in terms of looking at 𝐸𝑐 − 𝐸𝐹 which is basically that
difference. Then 𝐸𝑐 − 𝐸𝐹 here is very large and therefore, the electron count which is
338
𝐸𝑐 −𝐸𝐹
n=Nc𝑒 𝑘𝑇 is very small. And therefore, the electron population in this region very close
to the junction is quite small ok.
So since, this 𝐸𝑐 − 𝐸𝐹 is greater than the E c minus E f in the bulk, the electron
population here is much smaller than the electron population in the bulk. So, you should
remember that the population scales exponentially with the energies. So, small
differences in energies lead to large differences in carrier concentrations. And K T is not
a very large number. So, I mean it is KT/q is of the order of 25 milli volts and K T is
there for about 25 milli electron volts. So, it is not a very large number.
So, this region has got very few holes, because in a holes of the minority carriers in the
n-type semiconductor. And it is gone now got very few electrons.
And therefore, we said that this region is depleted of all mobile charges ok. But what are
the charges present here. We discussed that it is all the fixed ions due to the donor
doping. The donor dopants have all ionized and then they are all have their fixed positive
charge that is sitting in here ok.
So, this is the condition and we want to look at the electrostatics of this problem. So,
when we introduced the concept of junctions we said that in order to study the
electrostatics. We will need to solve the differential form of Gauss’s law, which is
339
something called as the Poisson’s equation ok. And that is exactly what we are going to
do here.
We are going to solve the Poisson’s equation which have sort of rewritten out here.
Which says that and this is a 1 dimensional Poisson equation which says that the dE/dx,
where E is the electric field ok. So, I am calling E, but it is actually the Greek alphabet
Ψ.
𝜌
So, dE/dx is equal to , where rho is the charge concentration per unit volume ok. In the
𝜖
space in the region of the semiconductor in which, we are trying to analyze electrostatics
so that is the charge concentration per unit volume. Epsilon s is the permittivity of the
material which in our case is the n-type semiconductor that we are looking at and for
silicon and for silicon the relative permittivity is of the order of 11.
Now, in the depletion region we are going to make very some might say drastic, but let
us say it is an approximation which is something called as a full depletion approximation
ok. So, the depletion approximation basically says that the charge concentration here. So,
what are the charges present in the depletion region in this region? You have the fixed
donor ions you have the positive charge.
So, you have plus q into N d per unit volume so, these are all the ionized donor dopants.
Then you have minus q n where n is the electron concentration and plus q p ok. So, that
is the complete count of all the carriers or all the charges present in the in this region the
depletion region. And if we have counter doped this material if you have added both N D
and NA and kept ND to be greater than NA. We will also account for the ionized acceptors
by writing -qNA.
But here let us just keep it simple let us say there is no counter doping it is just these
three components ok. You do have holes, you do have electrons and you do have donor
dopants, but the electron count you see, the electron count at the interface is much
smaller than the electron count in the bulks.
340
So, if we had defined this region her which is very far from the junction as a bulk and the
electron concentration of the bulk is approximately your N D which is the ionized donor
dopants. And therefore, your interface electron count is much less than N D and that is
why we said this region is depleted. And P in n-type semiconductor is again very very
small so, this is also small.
And therefore, we say that in the full depletion approximation we say that n and p are so
small that they can be neglected ok. And therefore, the charge concentration per unit
volume is simply qND. And therefore, we replace the rho which is a charge concentration
per unit volume with qND.
And we rewrite our Poisson’s equation as dE/dx is equal to qND/𝜖𝑠 . Now you know now
this is a very simple experiment this is simple differential equation to solve and your
electric field turns out to be varying linearly with distance. So, that is what it tells us
what is the solution of Poisson’s equation under the full depletion approximation says
that the electric field varies linearly with distance and there is an integration constant
here ok.
Now, how do we determine that constant ok? We say that see what does electric field?
The electric field is there because there is a bending, you see this bending is also the
potential right. So, we are also we already defined we already said that your energies are
essentially the q times the potential ok.
341
So, q times the potential is equivalent to your energies. So, the band bending in the
energy band, the energy band bending ok, I am sorry he band bending of these energy
levels is equivalent to the existence of a potential that varies with distance ok.
So, the potential is varying with distance so, the potential is a function of x. And
therefore, there is an electric field so, whenever you see a bent band it means that there is
an electric field in that region. So, very far from the junction, you know there is once you
cross this point there is no more band bending ok.
342
So, you have you had band bending to a certain extent and then the band slowly flattened
out and then it is completely flat. And beyond this region it looks like the semiconductor
does not know about the existence of a junction at all ok. And everything here was the
depletion region. Now, we define this width, as the width of the depletion region and
give it a symbol x subscript d, the d stands for the depletion width. And beyond this point
there is no band bending. So, any electric fields that exist in the material due to the
formation of this junction it is all existing between x equal to 0 and x equal to x d ok.
And therefore, the electric field from this point onward so, the electric field from this
point onward is 0 all right. So, that is the boundary condition that we are going to use.
So, we are going to say that at x equal to x d, the electric field is 0 and of course, the
electric field is 0 even for x greater than x d.
Now, applying that boundary condition and using it. We can calculate this integration
constant and we find that the electric field, the expression for the electric field is given
by this term here ok. So, the electric field is qND/𝜖𝑠 (𝑥 − 𝑥𝑑 ). So, let look at this picture a
little bit more carefully ok, so, let us look at this map ok.
So, this I have drawn these two dotted lines to sort of mark out all the events in the
depletion region. So, that is my x equal to 0 that is my x equal to x d and I have shown
the x axis here ok. So, that line there is x equal to x 0 and that is x equal to x d. So, that is
the width of the depletion region which is x d. Now all the band bending happens within
this region and beyond this the bands are all flat and this was my depletion region.
So, what are the charges? According to the full depletion approximation the charge
concentration in the depletion region is simply qND it is constant, because the doping is
constant through the entire semiconductor. If the doping was varying with space then this
would vary, but right now it is a constant. And we are not considering electrons and
holes. And therefore, which is basically the full depletion approximation and therefore,
that is not included into this picture.
So, you have all the positive charges here and these positive charges are balanced by
negative charges in the metal which forms a very fine sheet. So, this picture is very
exaggerated. So, this sheet is going to be very very thin ok, it is going to be a screen of
electrons that balance out all the fields. And beyond this region there is no field in the
metal and beyond this region the bands are all flat, and there is no field in the
343
semiconductor. So, that is the situation and we have calculated out the electric field to
vary as qND/𝜖𝑠 (𝑥 − 𝑥𝑑 ).
So, if you have to plot this electric field so, in the semiconductor, so again it is this plot
is there only in the semiconductor. So, we are looking at only the semiconductor at x
equal to 0. So, if you were to take this expression here at x equal to 0, the electric field is
qND/𝜖𝑠 (−𝑥𝑑 ).
So, that is the electric field at x=0 and at x equal to x d the electric field is 0, which was
our boundary condition. And between these two points electric field varies linearly with
x. So, now how do we calculate the potential so, what is the potential variation? The
potential variation is essentially it is a measure of the band bending.
344
So, if you can find the variation of the potential in the semiconductor with x you have
essentially found the nature of the variation of the conduction band bending with x, the
nature of the variation of the valence band bending with x etcetera. So, how do we
calculate the potential variation? Now the electric field is nothing, but minus d phi by d x
ok.
So, where 𝜙 is the potential and it is important so all though I am not shown it here, it is
important to know how we have defined the potential. So, this is the band bending right.
345
So, let us look at this as the distance x that is the energies and that is my conduction band
so, we just look at the conduction band.
Now, we have to keep a reference for potentials. Potential at any arbitrary point does not
make any sense it is with respect to a certain reference. So, we need to have a reference
potential and then measure the different locations in x. So, 𝜙x1 phi 𝜙x2 etcetera is all
measured with respect to that reference. So, we will set that reference at this point we
will say that at x=0, my potential is 0. So, that is my ground ok, if you want to think of it
that way that is my measurement ground.
And, as we go into the semiconductor, this gap starts increasing and therefore, my
potential starts increasing so that is my potential variation. So, that is the way we are
going to define this phi and the electric field is minus d phi by d x. Now it need not be I
mean you are free to choose any reference you like ok.
So, for example, an equally valid reference is you know choosing the potential in the
bulk as a reference. So, if this is the band bending and say that is x equal to x d. We
could say that that is the reference potential that is phi equal to 0; no problem. As long as
you apply the correct boundary conditions it is absolutely no problem as to what you take
as your reference potential.
346
But now for this analysis, I have to choose one of these two and I have chosen this as my
choice of reference. So, at x equal to 0 my phi is 0 and I want to know what is phi of x
because, if I know what is phi of x I know the band bending; I know what is E of E c of x
because, E c of x is going to scale or it is going to be proportional to q times phi of x plus
some offsets. So, since the electric field is minus d phi by d x so minus d phi by d x is
equal to this term here, which was the electric field which we obtained from solving
Poisson’s equation.
And therefore, we solve this is the second differential equation and we find that phi of x
varies as the square of the distance ok. So, it is a quadratic variation with distance plus an
integration constant. Now, this integration constant really does not matter because we are
going to measure 𝜙 (x) with reference to some other location. But then we could use that
reference point to identify that constant and we say that we have chosen our reference to
be x equal to 0 is where 𝜙 =0.
So, if we had chosen the reference are to be different if we had chosen the reference to be
phi equal to 0 at x=xd. Then we will have a different integration constant, but now since
we have chosen this as our reference.
347
We find that the potential that is phi at x d ok. So, this is my potential variation in space
ok. So, that is my phi of x and x keeps increasing and at this point we have your phi of x
d. And beyond this the potential remains the same.
So, phi of x d minus this point that is phi that is the potential at this location minus this
reference potential is given by this term here ok. And this has got a special meaning it is
something called as the built in potential of your device and we will denote this with a
special symbol 𝜙bi. So, b i stands for the built-in potential, because from the edge of this
junction till this point.
So, from this point till this location here we have a potential difference of phi b i ok. And
phi b i is very nicely connected to x d and x d you know you can inverse this relation to
obtain x d in terms of 𝜙𝐵𝑖 as the √2𝜖𝑠𝜙𝑏𝑖 /𝑞𝑁𝐷 . So, that is your depletion width. So, if
you know the built in potential you can calculate a depletion width and on the other hand
if you know you the depletion width you can calculate your built in potential.
So, that is the straightforward I mean that is the simplest analysis that one could do to
determine the electrostatics of the Schottky contact ok. So, it started off with the full
depletion approximation and we find that we get a reasonably good answer to what they
are depletion width is and how it depends upon 𝜙bi. Now all this is happening at thermal
equilibrium, we have not applied any voltages. We have created the junction left the
348
device in the dark and we find that this is the electrostatics. So now let us say we decide
not to make the full depletion approximation ok.
So, we are not happy with the full depletion approximation we want to solve things a bit
more accurately how do we solve it ok? So, let me just introduce it and I will introduce
the tools and the techniques do it and this is something that will be helpful even in the
future ok.
So, let us again come back to this you know a nice drawing here, I do not have to use my
marker on this slide because it is all shown very neatly. And we have got a lot of labels
on this drawing so, this is the band bending. So, that is your constant Fermi level, so that
is a Schottky contact and you have your barrier here ok. So, the Schottky barrier
formation and then that is your conduction band the conduction band bends and then
flattens out after some x=xd.
So, that is the picture we are looking at so let us just mark that point x equal to x d or
somewhere there and everything in the middle here is the depletion region. So, what are
all these definitions so that is the barrier height which is given by 𝜙𝑏𝑖 ok, that is the
potential that we had defined so that is our reference potential so, phi is 𝜙1 =0 at x=0.
This is my x coordinate ok. So, that is x d that is x equal to 0 my 𝜙 is 0 at x equal to 0
and this is the potential that is varying with space because of the bending.
349
So, that is q 𝜙(x) which we had identified to be a quadratic term under the assumption of
the full depletion approximation. And this term here is the built in potential. So, this is
your q of 𝜙 at x d which is your q times the built in potential, which we found you know
to various xd2. Now since E c is varying with E f, in this region this gap here that is
between the conduction band and the Fermi level is essentially E c minus E f. However,
to be more accurate you can say Ec(x)- Ef.
So, that Ec- Ef term varies so at this location what is the Ec- Ef it is equal to q𝜙B. And at
this location what is the Ec- Ef it is equal to whatever it is in the bulk. So, we define q𝜙B
q phi B and the bulk E c minus E f so, this is what the semiconductor hand before the
creation of the junction ok. So, we give it a special symbol because everything in this
region is the bulk region and this is all, the depletion region.
So, the Ec- Ef in the bulk is given is clearly identified by defining the subscript bulk that
and the Ec- Ef here is varying with in space ok. So, once this picture is clear it helps with
the lot of the analysis.
So, now let us say we want to do a more accurate study. So, we will say the Poisson’s
equation now I have written it directly in terms of the potential.
350
So, the way this comes about is let us say d E by d x is equal to rho by epsilon s where
𝑑𝜙 𝑑2𝜙
rho is the charge concentration. Now clearly, since 𝐸 = − you have is equal to
𝑑𝑥 𝑑𝑥 2
minus rho by epsilon s. So, you need to have a minus sign in place that.
So, in the full depletion approximation we had said that despite the charges being q N D
-qn+qp in the full depletion approximation we had said that these two are negligible terms
and we had we had gone ahead with just that ok. But now we want to do something a
little bit more accurate so, we will say that ok.
351
We will say that you know getting rid of n and p is not really necessary because we
really do not know the numbers particularly as you approach the bulk ok. So, over here it
might be a great assumption to make, but as you start approaching the bulk it seems that
it is quite logical to expect that the n the number of electrons keeps increasing. And
therefore, if you are trying to perform an analysis in this region it is definitely not very
advisable to get rid of the end ok.
So, what we are going to do? Is we are going to perform the study in a region of the
depletion in the depletion region, but closer to the bulk as compared to the interface ok.
So, we need a more accurate Poisson’s equation and we say that although we could
retain p we say that p is negligible, but there is no need because we have no information.
And in fact, it actually makes the mathematics a bit more symmetrical if you keep the p,
but just for the sake of an example I have not retained the p. So, we say that the p is still
negligible, because an n-type semi conductor. And in the depletion region as I approach
the bulk I have this to be my charge carriers.
So, that is my Poisson’s equation. So, you have a negative sign because of the potential
you are using a potential term, but then you have q N D minus these are the positive
donor ions minus the electrons in the bulk. So now, how do we go about solving this so,
we need to identify. So, this is a constant it is not a problem.
352
But this is not a constant; it depends upon the potential because it depends upon E c
minus E f ok. So, let us identify n bit more clearly. And I will just get rid of these marks
here so, that slide is more clear. So, what is n? N is the effective density of states in the
Ec− Ef
conduction band into 𝑒 𝑘𝑇 .
And we know that in the depletion region E c minus E f is a function of x ok. So, what is
I need to represent Ec- Ef as a function of the potential. I want to get everything in terms
of the potential because the left hand side has got a potential term. I want to sow I want
353
to be able to construct a differential equation that I can solve ok. So, I do know, I can see
that it is connected to the potential, but let me sort of make it very very clear.
So, what is Ec- Ef

is nothing, but q times the barrier height minus the potential so think about it. So, this is
Ec- Ef and that is the barrier height and that is the definition of the potential. So, this
minus that so 1 minus 2 gives you 3 so that is exactly what we have written here. And
what is the barrier height itself, the barrier height is connected to the bulk levels right the
ball barrier height can be written as this plus this so let us call it 3 and 4.
So, the barrier height so, 1 is equal to so repeated 3. So, that is 1 2 3 and let us call that 4
and 5 4 and 5. So, 1 is equal to 2 plus 3 I mean please do not take it as numbers. So,
maybe I should say A B and C do forgive me for D and E ok. So, A is equal to B plus C
and A=D+E you could use either one of these definitions now they just give you
Ec− Ef
different table they give you a different picture. So, we could say that n=𝑁𝑐 𝑒 𝑘𝑇 and
we use this expression here and write enhance this term here.
So, what we are saying is the n at the interface. So, the expression for n at any location in
the depletion region so, if you say this is the depletion region.
And I want to get an expression for n at say some x. What we are saying is? It is the nth
the interface that is n at x equal to 0 times e to the power q phi by k T; which means as
354
phi increases n increases, which is expected because the interface has got very little n
and as phi keeps increasing n increases till it approaches the bulk concentration.
So, now I can use this expression or I could write another expression, I could say that I
could go one step further. I can say that I want to define the interface term more clearly.
𝑞𝜙𝐵 𝑞𝜙𝐵
. .
So, I say that n is equal to 𝑁𝑐 𝑒 𝑘𝑇 𝑒 𝑘𝑇 which is same as that. And then I can rewrite
the barrier height 𝜙𝐵 in terms of the bulk Ec-Ef and 𝜙𝐵𝑖 using this expression here.
So, then if you look at this term N c into e to the power E c minus E f bulk by k t what is
that? That is basically the electron count in the semiconductor before the creation of the
junction. So, that is what is going on in the bulk. So, that is the electron count in the bulk
which is approximately equal to your N D e ok. So, this is another expression for n, you
could use either of these expressions.
Now, what we have done is we have taken this variable n and expressed it in terms of phi
which is the potential and we have done that in order to solve this differential equation
ok. So, we are just going through this example because it is quite. Now I think it is quite
illustrative with regards to the techniques; we will use in future did it illustrate a lot of
these techniques. So, now you have your n in terms of phi, we will have n as a function
of phi and we make that substitution.
355
So that is my Poisson’s equation so, that is the equation I want to solve to get a more
accurate answer ok. And how do we solve this? Now this looks like a tricky equation
right you have a d square phi by d x square term on the left hand side and you have got a
function of 𝜙 on the right hand side. Essentially you have got an equation which says d
square y by d x square is some function of y ok.
So, let us take a general x equation like this. So, here is one trick that you will find is
very useful in order to solve an equation of this kind. And simply what you do is you
simply multiply dy/dx on both sides ok. Now what is this term this is nothing, but the
derivative of (dy/dx)2 whole square with respect to d x, because if you were to expand
that you will first differentiate this term and then you will take the second order
derivative that is exactly the same as this so that is your left hand side.
And your right hand side is simply d y by d x. So, you can get rid of these d x and now
you have a very nice differential equation to solve. If you were to integrate both sides
you will end up with d y by d x square being equal to the integral of f of y d y. So, that is
one that is a technique to solve and that is what we will be using. So, what we do here is
we multiply both sides by 2 d𝜙/dx. So, you have that term that essentially gives you this
term here and that is equal to 𝑑𝜙/dx square is nothing, but the square of the electric field
right.
356
So, we now have an expression which says d electric field square is equal to all this term
on the right hand side and you just integrate both sides, the left side will be the electric
field square and after you integrate the right hand side term, you will get some function
of phi and we now know the electric field to be the square root of that particular term.
So, this is very much solvable and in fact, we will go ahead and solve it in great detail
later on. Of course, from this point on particularly this topic onwards the course does get
a bit mathematical in nature. There is there is going to be there are going to be a lot of
concepts to learn, but it is also important to learn how to calculate. So, we will go
through a lot of derivations it might appear quite painful to some students, but
nevertheless it is my duty to go through all these derivations ok. Because so, that you
have an understanding as to where everything came from.
But having said that in your exams and in your examples we will not be quizzing you on
how to derive a certain expressions so, that is not the point of the course. The point of
this course is to you know in a sort of communicate certain concepts and quiz you on the
concepts and yet show you methods and techniques to calculate and also show you how
these tools came about ok. So, that is so that is why we need to go through all these this
bit of mathematics, it is show you all the tools and the techniques and the methods to
perform calculations ok.
So, do bear with me if the mathematics gets a bit too dull. So, that is the electrostatics of
the Schottky contact we will not worry about the electrostatics anymore. And now we
will; what we will do is; we will try to take the Schottky contact out of equilibrium.
357
And apply a bias voltage and start observing whether there are any currents. So, we are
all still working on the Schottky contact ok. So, it is still we look at the electrostatics and
now we are going to look at the current voltage characteristics ok.
358
Lecture - 22
Schottky Contact: Current-Voltage (IV) Characteristics
So, here you have a metal-semiconductor junction. This was my n-type semiconductor.
And I have applied a voltage Va. And the voltage is applied such as you see the ground
or the reference is kept at the semiconductor. The positive sign is sitting at the metal. So,
when Va is greater than 0, it implies that the potential of the metal is greater than the
potential in the bulk of the semiconductor, so that is what we mean by this circuit
configuration right.
Now, we will call Va> 0 as forward bias ok. Now, you might have all heard this contact
this term in context of diodes, and in fact, the Schottky junction does behave like a diode,
so that is why although we have not formally discussed diode some pretty. If the students
have heard of diodes before, then you might have heard the term forward bias in the
context of diodes, but nevertheless. Let us let us just go ahead and call this condition as
forward bias ok. And when Va <0, it is something called as reverse bias ok.
359
So, now what happens to the band diagram, when I apply a potential. So, let us look at
these pictures here. So, this is the condition of equilibrium. So, this is the equilibrium
condition, that we are very familiar with and on which we performed all the electrostatics
analysis right, so that is my conduction band edge, that is the valence band edge, that is
the Fermi level, that is the built in potential, that we calculated out, and that is the
depletion region, which is got a depletion width of xd ok. And that was the relation
between so x d turned out to be √(2 ∈𝑠 ∅𝑏𝑖 )/𝑞𝑁𝐷 ok, so there was the relation that we
obtained.
360
Now, the moment you take it out of equilibrium by applying a bias by applying a
potential V a, what happens is you start the Fermi level need not be aligned anymore.
You are now creating you are trying to change the flux, now why did the Fermi level
aligned. So, if you remember the calculations, when we looked at the fundamental
junctions. We said that the Fermi levels aligned, because the flux of carriers from the left
to right across the junction was the same as the flux of carriers from right to left, and that
is true only at thermal equilibrium right.
Now, the moment you take it out of thermal equilibrium, but the moment you apply a
bias voltage. You are attempting to favour one direction as compared to the others. Now,
when you apply a forward bias, when your V a is greater than 0.
361
The Fermi level of this semiconductor is now split away from the familiar of the metal.
The two Fermi levels are no longer aligned, and you have to split away by the amount of
voltage you applied. And since, we are showing everything in terms of energy it is q into
V a. Now, because of this the built in potential changes case, the built in potential is no
longer q phi bi, the built in potential two is lowered by an amount of by the same amount
that is V a.
362
So, essentially what this means is that if you look at this picture here, on the left hand
side that bottom left, so that is my equilibrium condition, when my V a was 0. So, we are
looking at only the conduction band ok. So, you are trying to compare put these two
plots on the same figure here, so that is my conduction band at V a equal to 0. The
moment we apply V a to be greater than 0, the band bending becomes a little bit more
gentle ok. So, this bending becomes a little bit more gentle. And the built in potential
goes less than the built in potential and equilibrium ok, so that is what happens in
forward bias.
But note, what has happened to the barrier height, the barrier height remains unchanged.
And why is it, because the barrier height ok. If you think about it, let us say there is the
vacuum level there ok, that is the vacuum level. This is the electron affinity. So, as we if
you mark it from the semiconductor side, that is the electron affinity of the
semiconductor. .
And from the metal side just if you just slip it down to the metal side, that is the work
function of the metal. And both these are material properties that are constant. And
therefore this barrier height is essentially q times phi m minus chi s ok. And therefore
that the barrier height remains unchanged ok. So, you are not you do not influencing the
barrier height in any manner. So, the barrier height remains the same, it is only the
bending of the bands that becomes more shallow.
363
Now, when you apply reverse bias, the opposite happens. The work function of the I am
sorry the Fermi level of the semiconductor, now lies below the Fermi level of the metal.
So, by applying a positive voltage by applying a negative voltage to the metal we have in
fact made the metal Fermi level go higher as compared to the Fermi level of the
semiconductor.
And therefore, the bands now bend a little bit more steeper ok. And you should
remember that Va is negative, so that is why q into phi bi minus V a is actually going to
be greater in reverse bias as compared to equilibrium ok. So, V a < 0 in this case. In this
case Va >0, sorry should have label that, I just realized that these labels are missing, so
that is the case of V a greater than 0. This is the case of V a less than 0. And that is the
case of Va equal to 0. So, this case corresponds to this drawing here that is in black. This
case corresponds to this drawing here that is in blue. And this case corresponds to this
drawing here that is in red.
364
So, now let us look at, what happens to the depletion width. At equilibrium, the depletion
width was simply given by this term, but V a being equal to 0. There was no V a, so that
was the depletion width. This is what we had calculated from the solution of Poisson’s
equation. Now, after we apply a bias, the depletion width changes to having a term of
∅𝑏𝑖 − 𝑉𝑎 . When V a is greater than 0 that is when you are in forward bias, the depletion
width goes smaller, because phi b i minus V a is going to be smaller as compared to ∅𝑏𝑖 .
And the depletion width is now less than the equilibrium depletion width.
And when the device is reverse biased, ∅𝑏𝑖 − 𝑉𝑎 is now greater than ∅𝑏𝑖 . And the
depletion width goes to be greater than the equilibrium depletion width ok. So, therefore,
the depletion width of the device in forward bias is smaller than the depletion width of
the device in reverse bias, but the depletion width an equilibrium lying in between these
two conditions ok.
365
So, now let us look at the capacitance of this junction ok. So, we are trying to we are
trying to head towards the calculation of the effective small signal impedance right. So,
we want to know, what happens you know what if I have a Schottky junction, what the
impedance does it offer, when I apply a voltage across this. And what is the kind of
current I would get across, yes I capacitance is one of the aspects of the impedance
calculations, so what is the capacitance.
366
So, if you think about the Schottky junction, you have metal on one side. Then you have
a depletion region here. And you have a n-type semiconductor. So, the n-type
semiconductor the bulk can be thought of as a resistor, which is what we have looked at,
when we study the mobility of carriers n-types, the metal also is a very low resistance
resistor.
The depletion region of however, has got space charge right. It is gone just these fixed
ions. It is got no mobile carriers, there are no electrons. I should not say no mobile
carriers got very few mobile carriers ok. So, these two are negligible. So, it is as though
you have these two resistors. And there is some kind of a depleted or you know non-
conducting insulator in the middle. So, you one can imagine it to be that way. So,
therefore, this is like a capacitor in some sense. And what we would like to do now is to
calculate the capacitance of this capacitor.
Now, if you think about it this in this manner, the capacitance of a what we need to
calculate is the capacitance of the parallel plate capacitor right. So, the device had some
cross section area A ok. This is equivalent to having a capacitor, which is got two plates
with cross section area A with a insulator in the middle, which has got a thickness of x d,
which is a depletion width. So, what is the capacitance, it is going to be epsilon s A by x
d that is going to be the total capacitance of this junction.
But if we want to the per unit area capacitance ok, so we are interested in finding the per
unit area depletion capacitance, capacitance of the depletion region. So, per unit area
∈𝑠
implies I will just remove the A term there, and just have . Now, I know my x d is this,
𝑥𝑑
∈𝑠
and I divide in order to get, this to be the capacitance of the depletion region.
𝑥𝑑
Now, this may not be the most general method of thinking of capacitance ok. So, it
worked out very nicely here, because we could draw a nice analogy to a parallel plate
capacitor. But, from henceforth it would be good to think of the capacitance across two
terminals in a slightly different manner, let me describe that ok. So, what we want to
calculate is something called as the small signal capacitance ok. Now, what do what do
we mean by that.
367
So, let us say there is some device; it is sitting inside a black box ok. And you have these
two terminals across ok. And we want to measure the small signal capacitance small
signal capacitance. Now, what is a capacitance, how do you define capacitance,
capacitances so suppose you had your familiar parallel plate capacitor. And you were to
apply a voltage V across this capacitor, you will store a charge Q on this capacitor, and
your capacitance of this capacitor is Q/V. So, it is the amount of charge stored on the
capacitor for a given voltage V.
Now, what we mean by small signal capacitance is this. We now apply a small
fluctuating signal ok. We will apply a small ac signal. So, let us say, it is a tiny
fluctuating sine wave, if you like at some frequency omega. So, essentially it is going to
be the input, that we are going to give to the circuit is going to be say some V naught sin
of we just write it a little bigger. It is going to be V is equal to V naught sin of omega t
ok.
So, let us say we giving a small, this is a very small signal. You know, and what you
mean by small is a great question, it simply means that we have it is small enough that
the device can be considered to be linear ok. And we will understand these concepts a
little better as we go down. So, now if I were to apply a small voltage fluctuation here,
this device has got some charge ok. It is all we cannot see it, but we can sort of measure
it ok.
368
And this charge is going to fluctuate in response to this applied voltage, so because of
this applied voltage. You going to see an effect of fluctuation in the charge inside the
device ok, so that is d V applied, and that is the d Q the fluctuation in the charge. And we
know, what we are applying here, and we can measure this fluctuation. And therefore,
the small signal capacitance is simply dQ/dV. And in our case and so the depletion
capacitance is dQ/dVa.
So, if you were to perform the same experiment on your Schottky junction. So, you have
that to be the junction, this is the depletion region, and that was xd all right. And we
already know that xd is a function of the applied voltage. It varies its proportional to
square root of ∅𝑏𝑖 − 𝑉𝑎 . And this region has got all the positively charged donors right.
It is got some charge that is balancing any charge that is applied on the other side.
Now, if I were to fluctuate 𝑉𝑎 , so let us say I set a certain bias value for 𝑉𝑎 . And then, on
top of that dc bias we are going to fluctuate 𝑉𝑎 , a little ok. We are going to have a small
fluctuation in 𝑉𝑎 . So, how would the depletion region respond, the depletion region
[res/respond] will respond by changing the position of xd right. So, this position the x d is
going to fluctuate, it is going to move up and down.
Let us say, we are doing these measurements into reverse bias, so V a is less than 0. And
then, you are having the small signal fluctuation here. So, as 𝑉𝑎 goes a little higher x d
reduces a bit. And a 𝑉𝑎 goes a little lower xd increases a bit ok. So, therefore as V a
369
fluctuates your x d fluctuates. So, essentially it covers and uncovers either more or less,
it exposes more doughnut charges fixed ions, when V a goes up. And it covers up some
of the doughnut charges, when V a goes down.
So, by fluctuating x d you expose and cover up these doughnut charges. And therefore, it
is a response of charge to the applied voltage, so that is how your Schottky diode
behaves in the Schottky contact behaves in response to the a small signal variation in the
applied voltage. So, your dQd/dV a is essentially your capacitance. And you will find
that, if you were to perform this calculation, you have x d in as a function of V a, you
take it differentiate this with respect to V a. You will find that, you end up with the same
expression as our little picture of a parallel plate capacitor.
Henceforth, it is more useful to use this approach to calculate capacitance, because it is a

more generic approach, it is a approach that is compatible with performing an
experimental measurement ok. So, we will keep this in mind. And whenever, we discuss
the other devices let us say a p n junction diode or any transistor, we will use this
approach to calculate or measure what the capacitances.
If you look at the capacitance, the expression for the capacitance ok, so C is given by,
this term here. So, if I were to plot, if I will take 1 by C square ok. What is that 1 by C
square is nothing but 2(∅𝐵𝑖 −𝑉𝑎 )/𝑞𝑁𝐷 ∈𝑠 . So, if I were to plot, 1/C2 as a function of V
a, when V a is negative ok. So, there is a reason, why I am picking reverse bias, because
370
in forward bias, we are going to see soon that we are going to have other aspects of the
impedance coming up, because you are going to have currents significant currents.
If you were to perform a measurement, where you plot 1/C0 square versus V a, so you
measure you take an experimental setup. And you go ahead put your device through the
setup, measure out the capacitance for different negative biases ok. You are going to set
V a as minus 1 volt apply a small signal measure the capacitance, and mark a point. Set
V a as minus 2 volts perform a measurement mark another point. And do this for several
negative biases. And you will end up with the, you should end up with a straight line,
because 1 by C square is linearly related to V a, so this is a straight line. And the slope of
the straight line has to be −2/𝑞𝑁𝐷 ∈𝑠 .
So, therefore if I know my if I have my capacitance measurement, if I have this

measurement, if I have this experimental result, I can extract what my N D is ok. So,
even though ND was decided at the time of fabricating your device, we can measure out
what ND is, so it is a useful point remember.
So, now let us look at the current voltage characteristics through a Schottky contact. So,
we have seen that by applying a bias voltage to the metal-semiconductor junction. We
can change, we can alter the band bending by varying your applied voltage V a. So,
when V a is greater than 0, the band bending becomes more gentle ok, so the band
371
bending is gentle. And V a is less than 0, the band bending gets more steep as compared
to the equilibrium value.
So, equilibrium lies on band the middle, you just call it as band bending is normal, so
because of this change in the bands. I mean the band structure; we are going to see some
mechanics you are going to see things move out of equilibrium ok. And this essentially is
going to lead to the effective change in the change in the effective flux of carriers across
the junction, which implies that we are going to start seeing charge moving in one
particular direction as compared you know more charge moving in one particular
direction as compared to the other. And therefore, we are going to see currents across the
junction. .
So, we are interested in looking at the current voltage characteristics in a Schottky

contact, we are still talking about Schottky contacts ok. So, metallic semiconductor
contacts had two kinds, they were Schottky contacts, Ohmic contacts. We are still talking
about Schottky contacts. We looked at the electrostatics, and we looked at the non-
equilibrium case, we have looked at the measurement of capacitance. And now, we are
looking at the current voltage characteristics. .
So, there are three general mechanisms for current that lead to currents in our Schottky
contact ok. Now, the first two mechanisms are in some sense, you know the general
behaviour is quite similar, but although they imply a slightly different physics ok. And
they are called as diffusion and thermionic emission ok. And they are both represented
by this arrow here ok, so this arrow represents both these ideas. Again, I will I will
explain what those are.
And the third mechanism is something that we have looked at, when we studied quantum
mechanics, and it is called try it is called tunnelling ok. So, what are diffusion and
thermionic emission ok. Now this is what happens ok, so you have you have your you
know some electron population yeah. So, let us say this region is all depleted. .
Now, these electrons are at equilibrium, these electrons are trying to diffuse over to the
other side. And what do you mean by diffusion, you have a barrier ok. There is a
concentration gradient here and this concentration gradient is driving this current across
this junction, but that current is balanced at equilibrium that columns current was
372
balanced by carriers that are moving across from the metal to the semiconductor ok. You
have an equal an opposite flux from the metal to the semiconductor.
Now, this concentration gradient ok, this d n by d x is very dependent upon the built in
potential. And why is that, it is because the electrons all have a Boltzmann distribution.
𝑞𝑏𝑖
So, though all vary as you know they are all proportional to your 𝑒 𝑘𝑇 ok. So, as the built
in potential changes your electron concentration would be affected ok.
And in case you are wondering, how this came about you can you know you go back to
the general study. I am just going to go back a few slides ok. So, you can see that you
know in this expression here ok. Here this is where we you know sort of wrote out all the
quantities for your exact solution of Poisson’s equation. So, I am going back a few
slides, because it is illustrative. And you can see that the carrier count depends upon all
these potentials ok. So, the built in potential plays a role in the carrier counter.
373
So, now if I change the built in potential, I therefore can influence the carrier count ok.
And I can therefore influence the carrier concentration and that can lead to an effective
diffusion current, so that is the diffusion current mechanism ok. And we will look at it in
more detail. So, essentially it is the diffusion of the carriers that have got a count of
𝑞𝑏𝑖
𝑒 𝑘𝑇
𝑒 𝑘𝑇 . So, it is this effective number that is going to have create the concentration
gradient that is diffusion.
So, diffusion is very strongly it depends upon the nature of the band bending. And you
know what the built in potential is and you know what the shape of this bending is
etcetera. So, it is very dependent upon all these features ok, so that is that is diffusion.
Now, thermionic emission is quite is a little different ok. So, what thermionic emission
says is that irrespective of what this band bending looks like ok, you know we really do
not care about the band bending.
We only know that the barrier, that have we need to cross from the semiconductor to the
metal side; we have to cross this barrier. And for the metal-semiconductor side, we have
to cross that barrier. And we only look at those electrons that have got a large enough
energy and large enough velocity ok. Basically if you think about it, the electrons sitting
here let us take an electron at the conduction band edge, it has just climbed and climbed
up from the valence band to the conduction band, and it is called 0 velocity ok.
374
Now, as the electron climbs in energy, its velocity increases ok. And be after it crosses
this particular line here, its velocity is increased to such an extent that its kinetic energy
ok. The kinetic energy of the electron is now greater than say of phi ∅𝑏𝑖 − 𝑉𝑎 . If you are
talking about, this in a very general science so, these electrons have got enough energy,
they are sitting here ok. They have got large enough velocity, these electrons are located
there. They really do not worry about the barrier ok. They do not really worry about the
band bending.
Now, this electron is got so much of energy. And if you just measure the effective flux
and one direction, it is the current it is the current contribution of that electron ok. So,
thermionic emission takes a look at all the electrons that have got a velocity large enough
that is got a high enough energy, that can simply you know walk past the barrier like as
though there was no barrier and existing there ok.
375
So, we just see how the flux from right to the left is affected by the applied voltage, and
how the flux from left to right is affected by the applied voltage. And the effective flux is
the difference of these two. And if not effective flux is [def]- effected by the applied
voltage, then definitely you will have a current, and that current is something called as
thermionic emission ok.
Finally, tunnelling is quite different. So, in the case of tunnelling even though there is a
barrier ok despite the presence of a barrier, you have electrons sitting here. They do not
have enough energy to you know you know walk over the barrier, but they can definitely
tunnel through this barrier ok. So, the electrons actually the wave function the probability
of finding the electron on the other side of the barrier is non-zero ok.
376
So, we have we are familiar with looking at rectangular barrier. So, when we look at
quantum mechanics we you know, when we said the particle in a box, we said the it was
really a box. So, we drew diagrams like this the potential energy diagram looked like this
with the electron having a you know a sinusoidal wave function here. And there an
exponential decay in the wave function outside the box, and that was the barrier we
always discussed.
But here, the barrier if you think, about this band diagram the barrier is not the rectangle
ok. It is got this little it is got a straight edge on one side and this quadratic appearance
on the other side ok. So, these are the three mechanisms that we will discuss the
diffusion, the thermionic emission, and the tunnelling ok.
377
And what we are going to do now is go through a lot of mathematics ok. And calculate
out, what the current due to diffusion is, what the current due to thermionic emission is,
and what the current due to tunnelling is ok. So, once again please do bear with me, I
understand that lecturing on a lot of equations may not be the most interesting thing to
do, but, nevertheless it is important that you know how to calculate, and see where all
these you know how to put down, this physics and this concept, and you know choke out
the equations required for calculating these currents ok.
But, as I mentioned, I know you will not be quizzed on nobody is going to ask you to
derive these different aspects in your tests and exams ok, the exam should be purely
based on your understanding and concepts.
378
Module-05
Lecture - 23
Schottky Contact: IV Characteristics – Continued
So, in order to you will first look at the diffusion current, so and we will see why it is
called diffusion and where that comes about. So, we are going to make a few
assumptions, you know without these assumptions the mathematics can get very difficult
and therefore, we have to make assumptions in order to look at some answers, these
assumptions might sometimes in general in semiconductor physics when you make
assumptions you always do. So, in order to get a glance into what the non equilibrium
condition is.
It is very hard to calculate the non equilibrium condition, while making 0 assumptions.
And it is only to get a glance as to into the behaviour of the device in non equilibrium or
may be very close to equilibrium that you make certain assumptions proceed with the
analysis and develop an understanding on the behaviour.
So, here the assumptions that we are going to make are the barrier height is much greater
than k T. So, first assumption that we are making is this barrier height q∅B is much
379
greater than k T ok, the second assumption that we will make is that even though we are
applying a voltage and taking the device out of equilibrium.
We are applying a voltage checking the device out of equilibrium, yet the carrier
concentrations at these 2 nodes that is x equal to 0 and x equal to x d have their
equilibrium values that is the second assumption that we will make, they have not
changed too much, they are just slightly out of equilibrium. We just going to take the
system slightly out of equilibrium, we are going to apply a very small voltage v a and
you are going to assume that these concentrations do not change from that equilibrium
value.
And the third assumption we make is that it is a non degenerate semiconductor; that
means, that even though it is an n type material the doping is not so high that E c minus
E f in the bulk is less than 3kT, ok. So, it is most definitely the case than E c minus E f in
the bulk is greater than 3kT, all right. So, with these 3, let us write out a general
expression for the current which is got the drift, it is called the diffusion term and that is
my current density.
380
Now, you are going to make a few substitutions here. Now this electric field that you see
here, you are going to say that the electric field is minus d phi by d x where phi is a
potential and we substitute minus d phi by d x in terms of in, in the place of E, the
second substitution is that if you think about the diffusion coefficient we had earlier seen
that the diffusion coefficient when we looked at the derivation of the diffusion equation
we found that the diffusion coefficient was related to the mobility of my Einstein’s
relation right kT mu n by q, you are going to say that this relation holds and your mu n
is simply q D n by k T.
So, you are going to make this assumption and instead of mu and we are going to throw
in this particular term, instead of electric field we are going to throw in this particular
term and we will end up with an expression for the current density to be this. What we
going to do is, we are going to take this expression and they are going to multiply this
little factor which is e to the power minus q phi by k T on both sides and we are doing
this for a certain reason and it is because you will find that the expression on the right
hand side is equal to this term here.
So, if you make these changes, if you were to make all these substitutions and multiply
this factor you will end up with seeing the diff, the current across the junction the
current, the current density to be equal to this term here.
And now this is nothing, but your expression for a diffusion current. So, if you imagine
that to be some carrier concentration n 1, the current is nothing, but qDndn1/dx which is
381
nothing but your diffusion term. So, that is why we are talking about the diffusion of
carriers across the in a Schottky contact and phi is my potential. So, there is a potential
dependence here.
So, with this being the current across the junction we just need to solve this differential
equation. So, you have you have your J e to the power minus q phi by k T to be equal to
this term and we need to solve this differential equation in order to obtain the current
density.
382
So, the current density is simply, if you just integrate on both sides you take the d x from
the denominator put it on the left side and integrate, and then rewrite the expression you
know keep J out, you will find that the current density is simply given by this factor here.
−𝑞∅
−𝑞∅
𝑑𝑒 𝑘𝑇
We had seen that e𝐽𝑒 𝑘𝑇 = 𝑞𝐷𝑛 .
𝑑𝑥
−𝑞∅
So, we are taking the d x to the other side. So, we are saying that 𝐽𝑒 𝑘𝑇 𝑑𝑥 =
−𝑞∅
𝑑𝑒 𝑘𝑇
𝑞𝐷𝑛 𝑑𝑥 J and we now integrate both sides, the J is common so, the J comes out
𝑑𝑥
and this term here the remaining term that integral is shifted over down to the
denominator, that is how you end up with this expression.
So, the upper integral is easy. So, this is integration from 0 to x d sorry that is not very
clear, but it is 0 to x d. So, the numerator is quite straight forward because, there are no
terms here that depend on this particular factor and therefore, the numerator is simply q
D n into n e to the power minus q phi by k T with the limits going from 0 to x d. The
denominator is something we will look at, little later. So, that is that expression.
Now, if you look at the numerator, we are going to apply these limits. So, all we are
going to do is, you are going to take the n at x d into e to the power minus q phi at x d by
−𝑞∅
𝑘𝑇 − 𝑛|𝑥 = 0 . 𝑒 𝑘𝑇 e to the power minus q phi at x 0 by k T, you could apply these
limits, and the denominator has got this term phi, the phi is the potential and we know
the potential varies, you know as the quadratically with x right. So, that is the potential.
So, that is what you obtain by solving Poisson’s equation, assuming the full depletion
approximation.
383
So, we are going to substitute that phi. So, because phi is a function of x and that is that
function. So, we are going to make that substitution and rename our denominator to this
term yeah. So, now, what we need to do is identify what is the carrier concentration at x
d, the potential at x d the carrier concentration x equal to 0 and the potential at x equal to
0.
And if you identify these 4 terms, you have your numerator. So, the potential at x d so,
this is the band bending, we had earlier referenced the potential at x equal to 0 to be 0.
384
So, that was our reference and at x equal to x d we had our built in potential minus v a,
because we are now applied a bias voltage v a. So, it is phi bi minus V a.
So, the potential at x d is phi bi minus V a, the potential at 0 is 0 and the charge
concentration at x d is the same as we had said its going to be the same as the
equilibrium concentrations that is one of the assumptions we have made. So, the
−𝑞∅𝐵
concentration at x d is 𝑁𝑐 𝑒 𝑘𝑇 N c . So, how does that come about. So, would have
been good to have a band bending diagram here, but we can just draw it.
So, this is the case right. So, you have your conduction band that is the Fermi level and
that is q times the barrier height and that is let us say x d and that is your built in
potential minus V a and your Ec-Ef. So, if you look at this term here Ec – E , it is the
barrier height minus this particular value.
So, it is 𝑞∅𝐵 − 𝑞∅𝑏𝑖 , if it was the same as equilibrium. So, if this is the equilibrium
case 𝑞∅𝐵 − 𝑞∅𝑏𝑖 is your Ec-Ef in the bulk. So, that is basically your Ec-Ef at x=xd. So,
−(𝐸𝐶 −𝐸𝐹 ) −𝑞∅𝐵 𝑞∅𝐵
𝑁𝑐 𝑒 𝑘𝑇 at equilibrium is nothing, but 𝑁𝑐 𝑒 𝑘𝑇 . 𝑒 𝑘𝑇 . It is a basically this term
minus this term and there is a negative sign then therefore, you end up with this
expression.
385
So, that is the carrier concentration at x t in the at x t at equilibrium and we are going to
say that in the non equilibrium case that carrier concentration has not changed much and
at n equal to 0 once again the carrier concentration has not changed much from
−𝑞∅𝐵
equilibrium case and your (Ec-Ef ) remains, remains to be your 𝑁𝑐 𝑒 𝑘𝑇 So, that is your
E c, that is your E f and that is the difference which is nothing but your barrier height by
k T.
So, we now have, a n at x d we have n at 0, we have ∅|x=d and ∅|0 and we now
substitute these 4 parameters in your expression in your in the previous slide, if you look
at the expression here, we have to calculate these 4 parameters, we now have defined all
4 of them we substitute them and we will end up with the numerator having a term like
this.
Now let us worry about the denominator. So, that is the numerator. So, that is done ok.
So, now let us look at the denominator. So, this is definitely a typo here so, that is not
complete. So, if you go to the previous slide, if you look at the denominator, you had a
−𝑞∅𝐵 −𝑞∅𝐵
𝑁𝐷
term which was 𝑒 𝑘𝑇 e to the power minus q in q square𝑁𝐷 𝑒 𝑘𝑇 𝑥 2/𝑘𝑇 , all right.
∈𝑆 𝐷𝑥
So, now we are going to say that, if x is really small because we are talking about regions
that are here. So, we let us look at x being mx is much greater than x square by 2.
386
So, I am going to simplify that expression to this term. So, you are going to write so the,
this is a, there is a typo here. So, I am going to rewrite that expression, I am going to say
𝑁𝐷
it is 0 to x d e to the power. So, let me just make sure I transfer it correctly 𝑞 2 𝑥𝑑𝑥/
∈𝑠
𝑘𝑇.
So, that should be the expression here, and then you perform the integral with respect to
x. So, the integral of this term with the respect to x and after you apply the limits will
give you this relation that you see here, all right and if you further, if you make 1 further
2
𝑁𝐷 𝑥𝑇
assumption that is e to the power 𝑒^(−𝑞 2 ) is quite small as compared to 1 and you
∈𝑠 𝑘𝑇
replace this term by 1, you will end up with the current density to be equal to this
expression in a, we shown in the blue box here, wherein you have a, a big pre factor
here, ok.
Now the key points of this pre factor term are since it is the diffusion current definitely
there is a dependence on the diffusion coefficient. There is a significant dependence on
the barrier height which basically means that if the barrier height is very large, the
current is going to drop quite a lot most interestingly is dependence on V a.
Now, when V a increases what happens you will find that as V a increases you have an
exponential dependence on V a so, V a increases. So, if the bias, if the device goes to
forward bias conditions let us say V a is greater than 0 and it continues to increase, this
387
term is going to shoot up quite significantly but then you also have a phi b i minus V a,
but since it is a square root dependence on phi b i minus V a and an exponential
dependence on q v a by k T, this term is going to dominate the current expression and the
current will increase.
So, though it I have not shown it, it is worthwhile to imagine what the current voltage
characteristics looks like. So, if you forgive this expression or you know if you sort of
ignore this dependence on V a and say that even a V a changes a little this is not going to
change significantly as compared to this particular term.
If you go about with that understanding, you can say that J is as though you have a J
naught of course, it is a light dependence on V a but you know as we discussed you can
𝑞𝑉𝑎
ignore that 𝐽0 𝑒 𝑘𝑇 − 1.
So, what is this characteristics characters look like. So, we have to draw the current
voltage characteristics here, you will find that this is the current, that is the voltage and at
V a equal to 0, the current is 0 which is, which makes sense because you do not want to
have a current at V a equal to 0 because that would mean it is a current at thermal
equilibrium. So, that is V a equal to 0.
388
So, that is the thermal equilibrium point and then as V a increases, your current goes up
quite a bit, it increases exponentially and when V a is less than 0, the current decays, but
not it is not that.
So, you have a curve that looks like this and therefore, it is a device that allows current to
flow in 1 direction as compared to another. So, if I were to applies a plus 1 volt and
minus 1 volt, I would have a lot more current flowing through the device at plus 1 volt as
compared to the current flowing through the device at minus 1 volt. So, you could
substitute the values here and you can calculate that. Therefore, it is a device that allows
current in 1 direction and therefore, it behaves like a Diode.
So, this is a technical term we will, I think some of you might be familiar with p n
Junction Diodes but these devices are something called as the Schottky Diode and they
have a rectifying nature and that is why, it is called as a rectifying contact and what do
you mean by rectifying, Now, that the term rectifying comes from the use of Diodes in a
processing sinusoidal or ac waveforms, but essentially it means that it allows current in
one direction it aids current in 1 direction and does not permit current in the other
direction.
So, let us just leave it at that. So, it is a Diode it is called a rectifying nature therefore, if
you think of if you somebody says a metal semiconductor contact with a rectifying
property then they talking about the Schottky contact and it is a Schottky Diode. So, just
like you can buy p n junction diodes in a market, you can also buy Schottky Diodes and
Schottky Diodes have got some very interesting features. So, the first point is since we
talked about electron transport. So, you have a n type semiconductor when we are talking
about the transport of the electrons across the junction.
So, we are talking about the majority carriers. So, these Diodes are using their majority
carriers for charge transport. So, this is a point remember and I mean discuss p n junction
diodes will come back to this point.
389
So, Schottky Diodes have a current voltage that is dependent on the majority carriers
carrying the current, say it is a Schottky Diode based with a based on a metal p
semiconductor then it is going to be the holes that are going to be responsible for the
current. So, majority carriers are responsible for the current. So, that is a key point to
remember, there are other features that we that will become more clear when we discuss
p n junction Diodes but this is a key point to remember.
So, the built in potential so, we will see that when it comes to p n junction diodes the
built in potential in a Schottky Diode can be made much lower as compared to the built
in potential in a p n Diode but we will save that point full lead. So, this is the, this is the
derivation of the diffusion current through the Schottky contact. So, just on the few
features. So, just to think of it very intuitively.
So, why is it that you have this exponential dependence on the voltage. If you think
about it in a very intuitive manner, if you just forget the mathematics and you say that
you know I just want to birds eye view or but it is a very simple understanding of what
the diffusion current is.
390
So, you had your equilibrium condition, right. So, let us look at the carrier flux or the
current going from the semiconductor to the metal and the current going from the metal
to the semiconductor. So, the current going from the metal to the semiconductor is this
and the current going from the semiconductor to the metal is that. So, this is the current
so, which means that J m s corresponds to the electron transport electron flux from the
semiconductor to the metal and J s m corresponds to the electron flux from the metal to
the semiconductor.
So, let us look at the electron flux. So, at equilibrium you have to have these 2 currents
being equal you cannot have currents at equilibrium. So, which means that the electrons
are trying to cross this barrier phi B from the metal to the semiconductor side and the
electrons from the semiconductor to the metal side are trying to cross this barrier phi bi
which is the built in potential.
Now, when we applied a positive bias and these 2 were equal. So, this flux balanced that
flux there was no current. Now when we applied a positive bias, this barrier became a
little bit shallower. So, the built in potential is now phi b i minus V a. So, your V a has
changed the barrier height, I am not talking about this side, I am talking with the barrier
from the semiconductor to the metal ok, the V a has changed the barrier seen by the
electrons trying to move from the semiconductor to the metal.
391
But the electrons moving from the metal to the semiconductor, so no influence on their
barrier height, it is still phi B and why is it still phi B, because it is a constant right this is
nothing but your q phi m minus q i s, right it is a constant. So, you can even though we
have applied a voltage V a the metal to the semiconductor current it has not changed
because the electrons moving from the metal to the semiconductor remain the same.
But on the other hand the electrons coming in from the semiconductor to the metal
suddenly see a mass, a much lower barrier and therefore, there flux increases
dramatically. So, this is the reason that applying a forward bias suddenly allows a lot of
electrons to migrate from the semiconductor to the metal while not having any influence
on the flux of electrons from the metal to the semiconductor. So, if a lot of electrons flow
from the semiconductor to the metal, it implies that we have established a large current
from the metal to the semiconductor,
So, we want to think of this p n junction, now sorry if you want to think of this Schottky
junction. So, you have a metal semiconductor contact, that is your positive, that is your
negative, it is like as though you may have a Diode. So, that is a symbol for a Diode and
it is got this rectifying or you know a Diode like nature. It allows current in this direction
while not permitting too much of current in the opposite direction.
So, that is, that is basically a good overview of the current due to diffusion in the
junction Diode in the metal semiconductor Schottky junction Diode and these are called
as Schottky Diodes.
392
So, we will now look at another mechanism and this mechanism is something called as
Thermionic Emission. This too, this mechanism too gives you; you know shows you
gives you insights into the rectifying nature of this edge it shows you that the Schottky
contact behaves like a Diode.
So, we will see that towards the end of this discussion this mechanism also supports the
diffusion current expression, but it is just that the terms are a little different and therefore
these 2 mechanisms are sometimes treated together and under something called the
thermionic diffusion theory, you can think of both these mechanisms the same, you can
model them in a more in a better manner but we will just talk about the thermionic
emission separately.
393
So, what was the Thermionic Emission you know, as we discussed Thermionic Emission
takes into account all the carriers that I have got a high enough velocity or you know
high enough energy. So, you have an electron there and that is got a high enough energy
that it really does not care about the nature of the barrier. So, it is simply this electron
moving from the semiconductor to the metal that is going to create a current from the
metal to the semiconductor.
So, let us see in a let us take a count on the number of electrons and see what their
velocity is. So, that should be the approach to calculating the thermionic emission or the
thermionic current. So, it is called thermionic because its temperature dependent because
the velocity of the carrier is dependent you know it is it comes from the temperature.
394
(Refer Slide Time: 27:27).
So, let us now define 3 directions. So, we need to first define an axis before we talking
about there is all this mathematics. So, let us define this to be the x axis, you know as we
have always been using, we will define the y axis to be say out of the plane ok. So,
which I will just draw in this manner and the z axis is you know going upward there the
y and z are interchangeable because you will see that it is really irrelevant as to what you
pick as your y and what you pick as your z axis.
So, the current we are interested in is basically the flux of the carriers right it is
essentially n q into the velocity. So, that is the current, the only thing is this velocity is
not because of an applied electric field if it was due to the applied electric field we would
call this a drift current and then it would have been new into the electric field, but that is
not the case. So, this velocity is simply because of the energy possessed by the carriers,
you know due to which they are sitting well above the conduction bandage ok.
So, we will just and we are interested only in the velocity in the x direction, we really do
not care if a carrier sitting on the semiconductor side is moving in the y direction or the z
direction because that is not contributing to the current because the semiconductor has to
cross the junction, I am sorry the electron has to cross the junction in order to, contribute
to the current and therefore, we are only interested in the velocity in the x direction.
So, n into q into the velocity in the x direction gives me the current but I need to sum up
all the electrons or the electron count all the free electron free electrons sitting from E c
395
to infinite. So, therefore, I need to perform an integral form here today and at equilibrium
this current is equal to the current from the metal to the semiconductor side.
Now what we will do is we will take this d n and rewrite it as dn/de into de because we
know what dn/de is dn/de is a number of electrons per unit volume per unit energy right
and that is nothing but the density of states into the Fermi function. So, we will just write
dn/de is equal to the density of states into the Fermi function and that is the density of
states there and this is the Fermi function here which is approximated by the Boltzmann
distribution here, ok.
1
So, I, we have rewritten 𝐸−𝐸𝐹 because we have assumed that this term is greater than 1
1+𝑒 𝑘𝑇
that is the Boltzmann approximation for f of e. And furthermore, we have split you know
for the sake of clarity in the analysis we have split E minus E f.
You have rewritten that term as E-Ec and I have retained this term here and that term
there.
396
So, why are we making the splits? So, this is the conduction band edge, we are interested
in looking at the velocity in the x direction of all these carriers and these carriers have
got some energy E has got, have got some energy E and this energy here is E c.
So, the carrier sitting at E c has got 0 velocity and E-E c, this gap is specially important
for me because that is the energy, that is given to the carrier in the form of kinetic
energy. So, that is why we are trying to rewrite E-Ef as E-E c, we are trying to get a
gauge on what that energy is plus Ec-Ef or Ec-Ef is our 2 well known quantities and
therefore, that is a constant, ok.
Now, E minus E c is essentially the kinetic energy of the carriers. So, here we are
making an assumption that all the energy of the carriers above the conduction band is the
kinetic energy and we are doing so in order to s get a, get a measure of this velocity ok.
So, E minus E c is mv2/2. So, me is the effective mass of the electrons and here the v is
not v x v y v z it is the V vector ok.
397
So, it is, if that is your x y and z you can think of v as in all 3, it is got all 3 components
in it. So, essentially your V square is, V is Vx2+ Vy2+ Vz2 because that is the way the
energy is going to distribute itself. So, E minus E c is m v square by 2 and if you take the
derivative, if you take what is de by dv de by d v is nothing but m e v into 2 by 2. So, it
is m e into v ok. So, a m e star into v and therefore, my d is nothing but m e v into d v
and what we are going to do is now we are going to substitute these 2 terms and we are
going to try to estimate what that quantity is. So, I am trying to estimate calculate that
integral and that in and I know my dn/de is nothing but ge into f of e so.
Now, I have dn/de into de which is basically my d n has nothing but this term here into e
to the power Ec-Ef into all these terms put together. So, you can see that the whole
expression is now written out in terms of the velocity of the carriers which was our goal
we are trying to get an estimate on the velocity.
So, now all this is in you do not see the vx vy vz components because we need to convert
this to the Cartesian system, you need to split out the V in into these 3 components.
398
Now, v square is v x square plus v y square plus v z square and this volume element is
simply this volume element here. So, we now rewrite this expression d n by d e into d e
in terms of v x v y and v z because now we need to, we have a measure of the velocity,
but in particular we want to get a measure of the velocity in the x direction because that
is what is going to cross, that is what those are the carriers that are going to cross the
junction and that those are the carriers that are going to contribute to the current.
So, that turns out to be, I will not read out this expression but it turns out to be this big
term here, ok. So, here you can see or the summation of the 3 components. So, that
would have been v2 /2kT. So, that is now split into these 3 exponential components.
So, now, we can write out the current expression right. So, the current was nothing, but
this term here. So, that is why we started and we now have estimated this component in
terms of the velocities. So, that is what all this you know all these equations were about.
399
We wanted to get this entire expression as a function of v x and we have done that we
have got a function of vx vy and vz.
So, now we substitute this expression for this quantity here and we rewrite the integral in
this form in the y direction we really do not care about what the velocity is. So, we
consider all velocities from minus infinite ok. So, that is the velocities in either direction
z direction. We again do not care about what the velocities are, but in the x direction we
are only interested in those carriers.
400
So, let us say I have a carrier sitting here at the conduction band edge, let us take another
example of a carrier sitting here and let us take a carrier sitting here. So, let us say we
have these 3 carriers which we call as a electron A, electron B and electron C.
Now, the velocity of the carrier here, this a is 0, 0 velocity. So, that is not going to
contribute to any thermionic currents, the velocity of carrier b is not 0 but it is still not
high enough, it is not going to, it is not something that is going to walk pass this barrier
without you know really caring about the barrier. The electron B really depends upon the
nature of the barrier and you know how all these changes and you know the width of the
barrier here etcetera in order to contribute to any current.
So, we are not interested in carrier b as well, but carrier c has got enough velocity ok. So,
essentially it is above this dotted line and it is got enough velocity to just walk fast, this
barrier without really worrying about the nature of the barrier and that is the essence of
the thermionic current. So, it is carrier c that we are interested in.
So, it is only those carriers with a velocity corresponding to this energy level, this dotted
line and above that are going to contribute to any thermionic emission. So, we are only
interested in those velocities.
So, if you think about this entire integral from minus infinity to infinity with everything
in there the everything that is below v min does not contribute any current. So, that entire
current is essentially 0 and it is only that which is which is got a minimum velocity of v
min and above that is going to contribute to any current.
401
So, it is these carriers at this dotted line are said to have a velocity of v minimum. That is
the minimum velocity of the carrier for it to contribute to a thermionic current and we
only consider the integral from that minimum velocity to above and what is that
minimum velocity so, what, what velocity are we talking about, if A the carrier at A has
got 0 velocity, it is this energy which corresponds to v minimum.
So, v minimum, so if you think about the kinetic energy of the carrier here m e square m
v minimum square by 2 is essentially your phi bi minus va it is this. Now, here for the
first time you see the impact of va all right. So, far we never saw v a appear anywhere but
v a determines what the minimum velocity is.
So, if my va is. So, let us say, that is my equilibrium condition and if my v a is greater
than 0, the requirement for my v minimum has changed because the carrier here now has
got 0 velocity and I only need a very small amount of energy because I only need to
cross phi bi minus ∅𝐵 − 𝑉𝑎 in order to achieve the V, minimum required to get past the
barrier to contribute to thermionic currents.
So, b minimum is dependent on phi bi minus v a and we rewrite phi bi minus v a in terms
of all these quantities then ok. We can we can rewrite phi bi minus v a in terms of those
quantities.
402
So, now how do we perform this integral. So, look at the these 2 terms ok, look at these 2
terms here. They are an integral of say some e to the power minus a x square d x and I
will not say, I will not use the symbol x. So, let us say a something else. So, let us say a
alpha square d alpha ok. So, that is, that is the integral we are talking about.
Now, what is integral e to the power minus alpha square d alpha? It is nothing but the
Gaussian integral which is square root of pi. So, I have put a note here and in fact, this
note is a bit confusing because I am using x for something else. So, I would not want this
need to be confusing. So, this x is not related to anything else it is just to indicate think of
it as a purely mathematical note here, it is just indicate that integral of minus infinity to
infinity of e to the power minus x square d x is square root of pi this is something called
as a Gaussian integral and we are going to use that relation in that you know and
therefore, we can easily identify these derivatives these integrals.
So, these 2 are easily calculated they are constants they are error they are independent of
v y, they are just constants, but this integral is special because this integral is not the
−𝑚𝑒 𝑣𝑥
Gaussian integral, it is got a vx𝑒 2𝑘𝑇𝑑𝑣𝑥 .
403
So, this integral is going to give you a different answer. So, you solve for these 3
integrals and you will end up with an expression for the current from the semiconductor
to the metal is going to be this term here and we will find that the current from the
semiconductor to the metal is dependent on v a because the b minimum now requirement
is dependent on v a and therefore, this is the current from d. So, I should say that, this is
the electron flux. So, the electron flux from the semiconductor to the metal depends upon
v a.
And the electron flux from the metal to the semiconductor does not depend on v a
because of this barrier height and therefore, the effective current. So, electron flux of the
semiconductor to the metal is the same as the metal to semiconductor current and this is
nothing but the semiconductor to metal current. So, the effective current across the
barrier is Jms minus J semiconductor to metal.
So, it is this current going from the metal to the semiconductor minus the current coming
from the semiconductor to their metal and that is given by this expression here. So, this
is the final expression for the thermionic emission current.
404
So, once again if you think of all these terms as a constant ok. So, this entire thing if you
say its J 1 ok, the current is J 1 e to the power q v a by k T minus 1 which is again very
similar to your Diode like behaviour which we saw on the diff in the case of a diffusion
current, the current is going to have a current voltage characteristic level, that has not
changed.
But what is different is to look at is the pre factor term. So, here once again another
aspect that is common with the diffusion current expression is that the exponential
dependence on the barrier height seen from the metal to semiconductor.
405
So, the larger the barrier, larger the cliff the smaller the current, so that is also the same,
but if you look at the remaining terms you find that there is a dip strong dependence on
the temperature. So, as the temperature increases this current goes up dramatically. So,
therefore, there is a very strong dependence in the temperature and there is a strong
dependence in the barrier height and an exponential dependence on the applied voltage
and this pre factor here is something called as a Richardson’s coefficient.
So, that is about Thermionic Emission. So, once again it, it, it agrees with the fact that
the current has got an exponential dependence on the voltage.
406
And that the metal semiconductor contact behaves like a rectifying contact. So, now let
us come to the last bit which is the current voltage characteristics with the Schottky
contact but why are the process of tunnelling. So, what do we mean ok so the as I
described very briefly. So, far we have looked at tunnelling through a rectangular barrier.
So, all right so, it is always been tunnelling through this barrier the carrier even though
does not have enough energy.
407
So, let us say this is the potential energy that is the total energy is e and despite e being
less than v the carrier is able to get past this barrier and have a non 0 probability on this
side.
So, that is tunnelling and this essentially is a transport mechanism, the carrier has gone
from this side to the other side with some probability. So, what we need to find out is
where is the tunnelling. So, if you have your band diagram, let us say, that is the
conduction band all right. So, the electrons from the semiconductor to the metal will
tunnel through this barrier here and that is the tunnelling that we want to, what do you
say capture that is, that is the current that we want to capture.
And what we want to do is we want to identify what is the probability of finding an

electron at w ok. So, what is w? So, the first thing we are going to do is this is a very
difficult potential profile right.
So, this is a quadratic dependence on distance. So, it depends on x square. So, we are
going to approximate, this to our triangular barrier ok. So, this is a, this is the
approximation you are going to make, we are going to say it is not a rectangular barrier
but at the same time we do not want to use a quadratic barrier. So, you will give it some
distance w and we will say that the tunnelling probability is like is equivalent to the
probability of getting the carriers to tunnel across a barrier which is got this triangular
shape.
408
So, this is my potential profile right. So, this is my potential well if you want to call it.
So, that is potential energy and that is the barrier q phi b and let us say at some distance
w the potential profile hits 0. So, that is my potential profile, it goes like that it comes
down like that and it is there. So, that is my potential profile. So, which means that V of
x minus E. So, all this is offset you know. So, V of x minus E is this potential. So, it is q
𝜙𝐵 into 1 minus x by w. So, that is, that is defining this line here.
And therefore, if you remember the tunnelling through a rectangular barrier, you found
that the probability had an exponential dependence on Vx-E on V-E right. So, we are
going to borrow that idea and say that you know if I know the Wave function at Ψ of x,
the wave function at Ψ at x+d x is simply going to be the Wave function at psi of x into e
to the power minus square root of 2 m star to v(x) Q-E/h dx ok.
So, that is essentially saying that this is, this can be imagined to be a rectangular barrier
of width d x and I know the wave function, I know the magnitude of the wave function
there and therefore, I can find the wave function here because it is got this dependence
ok.
So, you are going to take an, we are going to take an infinitesimal width d x and say that
that is a rectangular barrier and then proceed to that. Now, if you were to, now continue
this you know you take Ψ(x)+2dx= Ψ (x)+dx into e to the power so on so forth and go
all the way till w you essentially end up with this relation the relation between the wave
409
function at x equal to 0 and the wave function at x equal to w is simply e to the power the
integral from 0 to w of this particular term that,.
So, one can, if you are familiar with something called as a Beer Lamberts law when it
comes to in optics when you throw a light let us say a light of certain intensity is thrown
onto a material, let us say some liquid which is stored in a container in a transparent
container you throw light on this material in you measure the intensity of light on the
other side, you will find that the intensity decays, you know sort of exponentially with
the distance depending upon the absorption coefficient.
So, this is a very similar analogy here. So, in some sense, the probability is absorbed ok,
this is the absorption coefficient if you want to think about it ok. So, what is the
probability of finding the electron on the other side?
It is we know that, if the wave function is given to us the probability is simply Ψ 2 ok.
So, it is the square of magnitude of the wave function. So, the probability the ratio or you
know, if you say if we create a ratio which is the probability of finding an electron at w
was is the probability of finding electron and 0, it is simply the ratio of psi of w by psi of
psi of 0 the whole square and we already know the relation between psi of w and psi of
0.
410
And therefore, this probability is simply this term here. So, this is the probability. So,
given n electrons at x equal to 0, if I put n electrons at x equal to 0, there is a probability
that n alpha electrons. So, this is my term alpha right. So, n alpha electrons appear at x
equal to w and that probability is given by this and we solve this integral and we end up
with alpha being equal to this term here.
Now, this little coefficient here is, it is got the same dimensions as the electric field ok.
So, this is not the electric field in the semiconductor it is not it is just phi b by w and the
problem is we have not defined our w very clearly. So, this is some in some sense a very,
it is a parameter that needs to be modelled very carefully phi B is a constant, but our w is
you know arbitrarily take chosen we have chosen it in order to get this very nice
triangular barrier that through which we have push the carriers through a lot of the
carriers to tunnel through.
So, therefore, we retain all the other parameters but alpha could be modelled very
carefully by modelling this term, this denominator here. So, typically the way see, if you
think about it you know you could cancel these 2 terms and as 5 b power half, but then
we do not know what w is and that cannot really it measured.
So, we leave it as this as some kind of an effective electric field which needs to be
modelled better in order to get a good estimate for alpha. So, this is the probability of
these carriers tunnelling through and we could very simply write an expression of say of
411
calling the current density as I am sure these things can be modelled much better but a
simple expression is n q into the velocity which normally be there which be the current
normally for n carriers but since only alpha and carriers got across this would be the
current through the metal semiconductor Schottky contact via the process of tunnelling
all right.
So, with that we will complete our discussions on the current voltage characteristics of
Schottky contacts.
412
Lecture - 24
Schottky Contact: Small Signal Impedance
So, we have seen in the Schottky Contact we have looked at the basic pan bending
diagram you have looked at the electrostatics, we have derived the current voltage
characteristics when we apply a certain bias. And we will now try to define the
impedance, the small signal impedance the Schottky contact based on all our studies.
413
So, the small signal impedance basically has got 2 terms. So, you can think of so, we are
all talking about low frequency aspects here.
So, in forward bias the Schottky contact behaves like a diode. And, you know, it can be
imagined to have a small signal impedance. So, just like we measured small signal
capacitance, you know, we are interested in the small signal parameters; which means
you have your metal semiconductor contact, you have got these 2 terminals. And we
have applied a small signal voltage V a. And in order to measure the capacitance, small
signal we wanted to measure the fluctuations in the charge with due to the applied
voltage. And in order to measure there is something called as a small signal resistance
ok, we measure the fluctuations in the current with respect to the applied voltage.
So, therefore, this these 2 terms give you the small signal impedance parameters for the
diode. So, the capacitance per unit area small signal is dQ/dV a which we have already
derived and found to be that. And since the current voltage characteristics due to
𝑞𝑉𝑎
diffusion and thermionic emission, you know, could be a written as sum /𝐽0 𝑒 𝑘𝑇 − 1, dJ
𝑞𝑉𝑎
by dV a is the per unit area conductance. And that is given by 𝐽0 𝑒 𝑘𝑇 . And if it is in
forward bias, this term would dominate and therefore, in forward bias this would be
dJ/dVa; which implies a dVa/dJ, that is the which is basically the resistance term is
𝑞𝑉𝑎
approximately kT/q that is the thermal voltage divided by 𝐽0 𝑒 𝑘𝑇 which is approximately
your J.
414
So, therefore, we just say it is of this order. It is a diode essentially, it is a diode and we
are looking at the small signal resistance of the diode. So, this is the small signal
resistance. And of course, there are many, there are lot more details to this, this is got we
look to the diffusion current, we looked at the thermionic current, you look to the
tunneling current and this does not take into account the tunneling current. But based on
the diffusion and thermionic currents, this𝐽0 J naught can be this factor Jo can be defined
slightly differently.
So, that is with that we conclude our discussions on the Schottky contact. We say that,
the Schottky contact is a metal semiconductor contact which has a diode like or
rectifying nature, simply because of this kind of a relation. I mean now proceed to
discussing the ohmic contact. Now ohmic contact is, you know, the impedance of an
ideal ohmic contact is quite simple because it has to be just a resistance.
So, therefore, we will not spend time discussing the current voltage characteristics of an
ohmic contact; instead, we will look at the electrostatics of an ohmic contact. And you
know, look at the, you know, what the Poisson’s equation is and, you know, how you
solve it, how do you go about solving that.
And this is something we have already it is it is familiar territory, because there is there
is there is nothing new here; is just that you need to be careful in looking at the ohmic
contact.
415
So, if you this is the band bending for an ohmic contact, right so, if you have an N type
semiconductor and you have a metal. And the metal N type semiconductor was such that
the work function the metal is now less than the work function of the semiconductor;
which means that you have q phi m there, and q phi s here. And you create this contact.
The metal will donate electrons to the semiconductor, and the bands in the
semiconductor will bend downward, you know, the conduction band will get closer to
the fermi level and the electron population here is greater than the electron population in
the bulk.
Where the electron population the bulk is about N D; my apologies from writing very
small. So, this is the electron population at the interface. So, that is going to be greater
than n bulk. So, clearly we cannot make the depletion approximation here, absolutely
not, because we know that this region is filled with a lot of mobile carriers. And
therefore, this is a more accurate representation of our Poisson’s equation. So, you have
your rho is equal to minus q n, because n is the number of electrons per unit volume and
q is the negative charge in the electron and therefore, a Poisson’s equation takes this
form.
Now, n is definitely dependent on phi and how does it depend on phi we have already
gone through this exercise once before, when we try to write down the Poisson’s
equations more correctly for the Schottky contact. So, it is a same exercise. So, n is
𝐸𝑐−𝐸𝐹
nothing but 𝑁𝑐 𝑒 𝑘𝑇 and therefore, all I need to do is identify what is this is E f that is E
c and identify what E c minus E f is and that is that that depends upon x in this region.
So, this is going to be my, let us define that as my phi. So, E c minus E f at any location
is the Ec –Ef in the bulk −∅ right. So, if I if I call that as A B and C, C is equal to A
minus B ok. So, therefore, E c minus E f in the region where the band is bending, I will
call that as E c as a function of x. So, that is what I have written there, E c minus E f in
the region where the band is bending is E c minus E f in the bulk minus q phi. So, that is
q phi, and we use that expression and plug it in to your expression for n, and say that q
by epsilon N c e to the power, you know, all these terms is your Poisson’s equation.
So, what we have done now is we have represented n in terms of phi. And this is nothing
but this expression here. So, what this is, is that. So, that is phi so, at this location at that
location phi is 0. So, if you think about it be now redefined our reference point so, that is
416
that is a point to note, there is no barrier height. So, it is a new problem, we have
redefined our reference point such that my potential is now 0 at in the bulk. So, that is
my 0 potential and that is my potential, it is a little different from what we did before.
But as I told you can choose an appropriate reference so, here I am choosing a different
reference point.
So, phi is 0 here and as we head towards the interface phi increases, and it is maximum
at the interface. So, that is the way I have defined my potential. So, therefore, as we
approach the interface so, at phi equal to ∅ =0, this term is one and you can see that the
electron concentration is same as the bulk electron concentration. But as we approach the
interface phi increases, and the electron concentration increases; which is as expected in
the case of an ohmic contact. And we are now familiar with the tools to solve an
expression of this kind a differential equation of this kind right. So, we multiply both
sides by d∅/𝑑𝑥 and we will end up with a differential equation that we need to solve and
then obtain your electric field.
So, this is the there is an introduction to the ohmic contact and this is the electrostatic
this is how you approach electrostatics of this case. The current voltage characteristics is
clearly defined by, you know, just a resistance that is a good it is a good approximation
to define the current voltage characteristics of an ohmic contact.
And as and by definition ohmic contact can be represented by a small signal model of it
just being a resistor on the region of operation, all right. One thing which I would like to
define before we close this topic is; the concept of something called as a Debye length,
ok. Now when you have an electric field and you have say carriers or some kind
screening this electric field off.
So, based on the electric field that you have applied, the charges respond to the electric
field leading to a charge distribution inside. And essentially the point is after a certain
length this electric field needs to be screened off, the electric field beyond a certain
length is 0. Now in order to measure this, in order to get an estimate on the screening
distance, we have a parameter called as the Debye length. It is an estimate of the
screening distance; it is not the complete screening distance. It is an estimate ok, and that
length is given by this particular term ok, and you can see that most of the parameters are
constant except for the charges available.
417
So, in this particular case, if n if the interface electron concentration is n interface, that is
the electron concentration, it is these carriers that are going to be available for screening,
and if I increase this carrier concentration I can reduce the screening length. So, which
means that if I have a large carrier concentration, the electric field will be screened off
much sooner, there will be no electric field seen beyond a certain point.
But if I reduce this carrier concentration, then I need to penetrate the electric field
penetrates deeper into the material and the screening will happen after a much larger
distance. So, it is a measure of the screening length of a electric field and that is
something called as a Debye length,. With this we will close our discussions on the metal
semiconductor junctions, and from this point on we will look at another junction that is
quite popular which is something called as the P-N junction.
418
Lecture - 25
PN Junctions: Electrostatics
So the next topic we will get into discussing is something called as a PN junction. And it
is also a junction that we sort of looked at earlier on when we just you know when we
discuss the concepts behind how to create band diagrams etcetera. As I mentioned the
PN junction is actually a junction created between a p type semiconductor and an n type
semiconductor. Now if the semiconductor is the same you will end up with two these
two materials having band diagrams of this kind. So, you have your conduction band
edge, you have your valence band edge and since it is a p type material you have a Fermi
level that is located slightly below the intrinsic Fermi level and this depends upon the
doping density.
And in the case of an n type semiconductor you have your valence band, your
conduction band and you have a Fermi level that is above the intrinsic Fermi level
position. Now since the semiconductor is the same your electron affinity is also the
same. So, if you were to create a junction between these two semiconductors if you if
419
you remember if you recollect the idea behind drawing band binding diagrams we need
to understand how the electrons would flow.
So, in this case the Fermi level here for the n type semiconductor is higher than the
Fermi level position for a p type semiconductor. And therefore, which basically implies
that the work function of the n type semiconductor is smaller than the work function of a
p type semiconductor therefore, the electrons will flow from the n side to the p side.
And therefore, if you have to construct this junction, what we will see is that in the n
type material near the junction; the junction is lost electrons and therefore, the
conduction band edge moves away from the Fermi level. Whereas, the valence band
edge moves towards the Fermi level whereas, in the p side the interface is gained
electrons. So, you have added electrons to the p side and therefore, the conduction band
edge moves towards the Fermi level while the valence bandage moves away from the
Fermi level.
Now throughout this band bending diagram your energy gap is more or less constant,
your electron affinity is constant and far away from the interface that is very far from the
junction the semiconductor really does not notice any effect of the other type of
semiconductor or the effect of the junction. And therefore, you have your typical n type
bulk material and your n and your p type bulk material. So, this is the PN junction
420
diagram the PN junction band bending diagram. So, the Fermi level is aligned as is
required for at thermal equilibrium.
So, this is true only at thermal equilibrium and you have this kind of a band bending
diagram and just as in the case of metal semiconductor junctions, we can define the
electric field in this region, we can define something called as a built in potential that is
effectively the potential difference between due to this band bending ok.
So, that is the built in potential and now we will we will try to use Poissons equation, to
try and understand what the electric field is and what the built in potential is. And that
will be the first you know exercise we will go through and we will also define some
coordinates because these regions near the interface ok. If you look at this, let me just
draw a dotted line here and a dotted line here. Now if you think about this region on the
inside it has lost electrons and on the and it did not have many holes to begin with.
So, in some sense this region is depleted of free carriers and on the p side it has gained
electrons which would have recombined with all the holes. Because you have it had a lot
of holes, but since these electrons have come in, the holes and the whole population is
also cantina ok. If you want you could think of it as the holes have been diffused to the n
side and the electron population was not very large to begin with. Because here the
𝑛𝑖2
whole population was approximately and there the like the electron population was
𝑁𝐷
421
𝑛𝑖2
. And therefore, those numbers not very large to begin with and therefore, this region
𝑁𝐴
is also depleted.
So, everywhere you see a band bending you could say that the region is devoid of free
carriers. Now in reality it is not completely depleted you do have free carrier
concentration and the electron and hole concentration can be identified by looking at the
difference between the Fermi level and the conduction bandage or the Fermi level and
the valence bandage respectively. But if you assume its depleted it makes you know your
solution for the electrostatics a lot simpler and we will start with that assumption ok.
But in the metal semiconductor junction when we already looked at a very general
Poisson equation wherein we considered the electrons we consider the whole population,
we considered the dopant ions and we did solve Poissons equations you are already
familiar with what to do if you have to consider everything. But for the case of for this
analysis, so we will just assume it is a depleted region.
So, if it is depleted what is the charge concentration on the n side you have all the
positively charged fixed donor ions ok. So, these had donated electrons to the n type
material and those electrons have all diffused onto the p side and therefore, these are
fixed donor ions. And on the p side you have the negatively charged fixed acceptor ions.
So, these are the acceptor dopants that gave up or that took an a electron and essentially
422
gave up a hole and those holes have now diffuse to the n side and therefore, you have
exposed acceptor ions.
So, this is the situation you have a depleted region where you have a charged
concentration here, which is basically plus qND where ND is the number of dopants per
unit volume. And the charge concentration in the p region is -qNA where NA is the
acceptor number of acceptor ions per unit volume. So, if we have to write Poissons
equation for the n and p side what would it be. Poissons equation tells you that d by we
looking at only one dimensional cases here. So, dE/dx where this is the electric field is
equal to your charge concentration by epsilon s where this is the permittivity of this
semiconductor.
So, the charge for the n side your Poissons equation is going to be de n by dx the
subscript n is to denote the n side is equal to your charge concentration on the n side
1
𝑞𝑁𝐷
which is . And for the p side you have a Poisson equation which is dEp/dx is equal
∈𝑠
1
−𝑞𝑁𝐴
to . So, this is the these are the equations that we need to solve to understand the
∈𝑠
electrostatics.
And what we will do is we will solve for the n and p side separately. So, on the n side
you have your Poisson equation as this, which is quite straightforward because this is a
423
constant ok. And it is a constant because we have not assumed we are not consider the n
and p the free carriers. And therefore, the electric field is simply given by this term here
and you will already recognized that the boundary condition has been applied ok. And
what was the boundary condition; since on the n side if you only look at the n side we
can draw the band bending diagram as this ok.
So, that is only the n side of it and that is your Fermi level. And we say that at x equal to
xd n all the band bending has stopped and the bands have now become flat. And
therefore, at this edge the electric field is 0. And its only within this region from say x
equal to 0, which has been defined at the junction to x=xdn that you have any electric
field.
And if you say electric field is 0 at x equal to xd n then you have apply is boundary
condition to this to the solution and you end up with the electric field being this
particular term here. And the potential if you look at the potential, since the since the
potential since the electric field is nothing, but the derivative of the potential is a
negative of the derivative of the potential you can now find the potential function as well
which is which is going to give you a quadratic relation of this kind.
Then on the p side you have your Poisson equation to be defined by this expression
where this is the charge concentration NA is all negatively charged ions and therefore,
you have a negative sign. And you apply the boundary condition if you look at the p side
you sorry you have a band bending that looks like; that looks like this ok.
Let me just be a bit more accurate that looks like that and on the p side since this is my x
equal to 0, we have band bending till a point of x is equal to minus x dp ok. So, this side
of the axis is all negative values of x and that is all positive values of x. And you apply
that boundary condition and you will end up with this solution to your electric field and
now you can go ahead and calculate the potential. So, the expression for the potential
would be another quadratic which is telling you what is the potential on the p side.
424
So, since there has to be a continuity in the electric field now we have solved for the
electric field on the n side and the p side separately right On the n side we got an
expression which was q ND by epsilon s into x minus x tn which means at x equal to 0
−𝑞𝑁𝐷 𝑥𝑑𝑛
my electric field is negative. So, en is negative and it says it is . So, that is my
∈𝑠
electric field on the n side at x equal to 0. And if you use the expression for the electric
field in the p side if you substitute x equal to 0, you end up with an expression of ep at x
−𝑞𝑁𝐴 𝑥𝑑𝑝
equal to 0 is .
∈𝑠
Now, we are talking about the same field for the field has to be continuous. So, you have
an electric field that varies like this and varies in this manner with that being your xdn
and that being your -xdp. So, since at x equal to 0, en has to be continuous you have en
and x equal to 0 being your ep at x equal to 0, which gives us a very important relation
that your NDxdn=NA xdp.
So, what is this relation mean intuitively it means that, if my charge if my dopant
concentration is large let us say NA>>ND it says that xdp is going to be much smaller
than xdn, which means it takes a lot less distance for me to shield all the field ok.
Because my charge concentration is very large, I just need a smaller amount of charge
exposed before all the electric field is nullified.
425
On the other hand if my charge concentration is very low I need to expose a lot more
charge before the electric field is nullified. So, that is essentially what this expression
intuitively implies. And now we also know the potential at in the n side and on the p side
from our Poissons equation. See if we set a reference potential so, let us say we set a
reference potential at x equal to minus xdp you could set a reference potential anywhere,
but we say that this point has got a potential of 0. And therefore, this entire potential
difference between here and x equal to xdn is my built in potential.
So, that potential is my phi n at xdn minus phi p at minus xdp which is my reference, so
that is my reference potential. And you substitute for these two values of x, you say x
equal to xdn there and you say x equal to minus xdp here and you solve for the potential
and you calculate this difference and you will end up with this expression for the built in
potential. So, it depends on the square of xdn and xdp ok. So, this is a bit about the
electrostatics of your PN junction which is obtained from your Poissons equation.
Now, taking this one step further, so that is my potential term right, but we already
know. So, let us make use of point number one in this expression that is NDxdn=NAxdp
so let us make use of that.
Since we already know that we can write your xdn in terms of xdp alone or vice versa.
And using that in your potential expression using that in your expression for phi bi we
can estimate what xdn and xdp are in terms of phi bi you will find that your xdn is got
426
this dependence on the built in potential, and your xdp has got this dependence on the
built in potential. And therefore, the total depletion width that is taking into account the p
and n side the total depletion width that we call as x d is equal to xdn plus xdp. So, that is
your xdn that is xdp and this sum is simply the sum of these two terms and you end up
with this expression.
So, these are all very useful expressions for you to perform quick calculations ok. So,
even though we have not made we have assumed that these regions are deplete devoid of
free carriers we did not take into account n and p because if we do take this into account
these will all be functions of the potential. And therefore, your Poissons equation will
have a nature of this kind it will be you know some function of the potential.
So, just to summarize what we looked at in the electrostatics do forgive me if this writing
is too small, we saw that this is the n side and that was your p side. And once the
junction was created you have a band bending diagram that looks like this and if you
look at the charge profile versus distance you have all the exposed donor ions exposed
there and you have all the acceptor ions exposed here. And since our donor dopant
concentration was constant right it was not varying with space this is constant with space
and that is constant with space.
And the electric field profile looked like this, we saw that it had a relation of x minus xdn
on the n side and is proportional to x plus xdp on the p side with x being a negative term.
427
And if you looked at the built in potential, we had set this as a reference of 0 and the
built in potential slowly increase as you went from the p to the n side. So, this is a
summary of the electrostatics of a PN junction. So, now, let us look at the current voltage
characteristics and before we do that, we need to see what happens when we take the
system out of equilibrium. So, this is all at equilibrium it is only at equilibrium the Fermi
level aligns when you create a junction.
So, when we take it out of equilibrium let us do so by applying a voltage ok, so we are
going to now apply a voltage across the PN junction. And the voltage is being applied in
a manner where the plus is toward the p side is at a higher potential as compared to the n
side; if the voltage on the p side or you know as we have drawn this picture if my Va is
greater than 0.
428
That is the voltage on the p side is greater than that on the n side then the device is
supposed to be in forward bias, and if the voltage is less than 0 then the device is said to
be in reverse bias, and if the voltage is equal to 0 then you are at equilibrium.
So, everything we have discussed so far is all at equilibrium there was no voltage applied
or in other words you know Va was equal to 0. But now when you start applying a
forward bias or a reverse bias that you would start taking Va to be nonzero then your
built in potential changes a little. And we have already looked at it in the metal
semiconductor junction you see these relations are quite familiar a quite similar to your
metal semiconductor junction relations except for the fact that you have here you have
both an NA and ND doping etcetera, but you will see that all the other terms are quite they
are quite familiar ok.
So, just like in that case you can take an intuitive leap by noting that your built in
potential term will now at equilibrium instead of using the equilibrium value we replace
that with phi bi minus Va in order to calculate your new xd n and your new xdp and your
new xd. So, instead of just using phi bi instead of just using phi bi we replace that with
phi bi minus Va where Va is the applied voltage in order to get your out of equilibrium
values.
So, which means what? Which means that when my Va is positive my xd decreases, both
my depletion widths the xdn and xdp decrease and my total xd decreases. And in reverse
429
bias when my Va is negative if Va is less than 0 then ∅𝐵 −Va is going to be a larger
number and my depletion widths would increase.
So, to give you an intuitive feel again just to summarize what non equilibrium does, so
let us say let me sort of draw this much more clearly. So, let us say that was my
equilibrium case that is the band bending in equilibrium. So, that was the value of xdn,
so let us say my x equal to 0 is somewhere here and that is my x=-xdp and that was the
built in potential right there so this is all at equilibrium. Now when we shifted out of
equilibrium the band bending changes a little. So, you firstly make the bands a little bit
more shallow, because the built in potential will now become ∅𝐵𝑖 −Va.
So, in forward bias the band bending will be a little shallower and the depletion region is
only going to be extending from here to here it is not going to be as large as before its
going to be much smaller. And on the other hand if it is a reverse bias, if its reverse bias
the band bending will be a little bit more steeper, because my ∅𝐵𝑖 is increased and it will
also extend to a larger distance ok. So, that is going to be the reverse bias case that is
going to be the equilibrium there is going to be the forward bias case and that is going to
be the equilibrium case.
So, that is sort of intuitively summarized and if you look at the electric field profile in
forward bias it is going to be something like this. Because the depletion widths are now
reduced and the built in potential and forward biases of course, going to be less, it is
430
going to be phi bi minus Va and in reverse bias it is going to be larger because your Va is
negative.
So, now let us start heading towards understanding the current voltage characteristics.
So, what defines the current voltage characteristics in a PN junction, so at equilibrium so
let me just draw the equilibrium band bending diagram.
So, if you think about the n side it has a lot of electrons so that is my built in potential
there has a lot of electrons and I will just, in a very poetic manner let me just show you
all these electrons is a big heap of electrons there and it is got very few holes ok. So,
these are all just holes, but in a lot more electrons and these are all sitting there on the
other hand the p side has got a lot of holes we need to mark these differently. So, the
electrons are all marked as solid solid objects and the holes are all these hollow circles it
has got a lot of holes on the p side has got a lot of holes and very few electrons.
Now, if you look at the electrons in the n side they are trying to climb up the hill ok. If
you look at this if you look at the band bending, if you recollect the band bending
diagram is all drawn in a manner, where you know it is easy to understand the motion of
the electron and hole the electrons like to run downhill the holes like to run up the hills
and you know if you increase the potential the energy levels go down and if you decrease
the potential energy level goes up ok.
431
So, that is the way you imagine the band diagram. So, these electrons on the n side have
to overcome this barrier they don’t like they don’t they have to overcome this barrier in
order to get to the p side on the other hand the electrons in the p side will just try to run
down the hill towards the n side. So, the electrons in the p side will try to drift down
whereas, the electrons on the n side have to diffuse across and they can do. So, only if
there; there are larger number of electrons you know sitting above this barrier very
intuitively, and it is a similar idea on the pisa or if you look at the picture of the holes.
So, essentially at equilibrium there is no current the electron current is 0 the hole current
is 0 and the total current is 0 that is because the electrons that are drifting towards from
the p to the n, and the holes that are drifting from the n to the p are balanced by any
diffusion of electrons from the n to the p and any diffusion of holes from the p to the n.
So, all these currents are essentially 0, but the moment I reduce the barrier height, so let
us say we apply a forward bias and we reduce their barrier height.
So, let us sort of redraw this picture the moment the band bending shifts to a shallower
built in potential. So, by applying forward bias this heap of electrons now protrudes
significantly above this much lower barrier height and a lot of electrons start diffusing
from the n to the p side.
On the other hand the drift of electrons from the p to the n side does change, but the
change is not that significant and therefore, the current is heavily dominated by the
diffusion of electrons from the end to the p side and the diffusion of holes from the p to
the n side. So, just like the barrier height was lower than the conduction band that barrier
height is also lowered on the valence band and you have a new picture of this kind.
So, you have a lot these holes now have to diffuse across this much lower barrier and
therefore, the diffusion of the holes increases ok. So, you will find that in forward bias
the electron current goes to greater than 0 and this electron current is composed of
diffusion and drift and it is a diffusion component that dominates.
Similarly, the hole current goes to be greater than 0 and it is the diffusion of hole
diffusion of holes that dominates the drift has not changed too much. And therefore, the
in effect what this implies is the as a huge amount of transfer of electrons from the n to
the p side and holes from the p to the n side. And since we had bias star PN junction in
this manner wherein the voltage is positive here and its negative is in the n side which
432
means its forward biased. You have a lot of electrons moving here you have a lot of
holes moving this side which means you have an effective large effective current going
from the positive terminal of the battery through the PN junction to the negative terminal
of the battery.
So, it is a very large current that goes through and a small increase in Va reduce the
barrier height a little, but since the diffusion component or the you know the electron
populations exponentially dependent on the voltage. You find that the electron
population scales are much large it scales up very rapidly as you change the value of Va
therefore, the diffusion is quite a powerful component in the current. Now on the other
hand if you say let us not apply forward bias lets apply reverse bias which means that my
Va is negative and the n side is got a larger potential is compared to the p side.
Then in that case the band bending will become much more steeper.
433
And there will be only few electrons peeping up only few electrons refusing from the n
to the p side and very few holes diffusing from the p to the n side.
Whereas the drift once again has not changed too much, but it has definitely been
impacted and you will find that the drift current dominates.
The magnitude of Jn is still greater than 0, but it is the drift component that is important
now the diffusion component is not so important. And similarly for the p side the drift
component is important, but the diffusion component is not important because you have
now increased the barrier height by applying a reverse bias. So, what does electron drift
do since the electric field is now become larger because of the because of the reverse
bias.
434
You will feel you will find that the electrons drift from the p to the n side and the holes
drift from the n to the p side and that is the current that is important. So, it is this is the
scenario you have your PN junction, you have a voltage source that is now negative. So,
let me put a negative sign than positive or maybe I should just retain the old convention
and just say Va is less than 0 ok, which effectively implies that side is a negative and this
side is a positive ok.
Or once again you know let me just make it just to make things very clear let me apply a
battery like that ok. So, that is holes everything and you find that the electrons drift from
the p to the n side. So, that is my current and I have holes drifting from n to the p side
which means that you have a current flowing this way from the n to the p ok.
And that is my effective current and the PN junction and this current is not very large it
is not as prominent. So, if I were to apply the same amount of forward and reverse bias,
if I say that the magnitude of Va is same and in one case I apply let us say just for the
sake of example, let us say I apply plus 1 volt, the other case I apply minus 1 volt yes I
will have currents in the forward and reverse direction.
But the forward direction current will be much larger because the electron population
scales exponentially whereas, here it is only the field that is determining the current it is
a drift current. And therefore, the forward bias current at 1 volt will be much larger than
the reverse bias current at minus 1 volt ok.
435
And we were sort of mathematically now you know sort of bring an all these arguments
all these qualitative arguments in the more mathematical description. In a manner which
is quite similar to what we did for metal semiconductor junctions, but we will just look at
some of the important differences between the metal semiconductor schottky junction
and the PN junction diode ok. So, 1 point which is obvious from what I told you, but
this, but needs to be mention this.
Since the current in forward bias is going to be much larger than the current in reverse
bias you do have a rectifying nature for the PN junction. So, you will find that the current
voltage characteristics for a PN junction looks like this wherein my bias goes to be
greater than 0. The current scales quite rapidly it is an exponential dependence on the
voltage and when it is less than 0, it is still an exponential dependence the voltage, but its
exponential of a negative negative term.
So, it is going to be dependent on this term there and dependent on this term here and we
will see this come out in the mathematics of it ok. So, therefore, the diode rectifies it is
got a rectifying property it allows current in one direction, but does not encourage
current in the other direction while although it does allow that current ok.
436
Lecture - 26
PN Junctions: IV Characteristics
So, let us go ahead and start looking at the proper derivation of the current voltage
characteristics of a P-N junction. But before we get to this point we need to understand a
bit of what is going on either sides of the junction.
437
So, this is our P-N junction. So, let us let this picture be there, but I will I will use this.
And we have a depletion region so, let us make all that as the x equal to minus x d p
point and that is the x is equal to x d n point.
So, we have a P-N junction that is got there is got this depletion region in the middle.
And let us say we have now forward bias to this P-N junction so, V a is going to be
positive. So, as I mentioned a lot of holes are going to diffuse from the p to the n side,
you are going to have lot of holes moving in there. You are going to have a lot of
electrons diffusing from the this is the electron diffusion from the n to the p side that is
the hole diffusion from the p to the n side. And you are going to have a small drift, which
we are going to ignore it is a small drift of electrons from the p to the n side, and a small
drift of holes from the n to the p side which we will ignore.
So, for all practical purposes the diffusion component is the current component. And
since these holes are being injected so all this region is depleted. So, if you look at this
boundary of the P-N junction so, at x equal to x d n if you just look at this interface here
what is happening here? It is initially got a very small hole population at V a equal to 0
the whole population here is very, very small, right it is it is n i square by N D. So, just to
give you an example if n i is say for silicon let us say it is about 1 e 10 per cc. Capital N
D let us say we have doped it to 1 e 15 per cc. The n bulk region had about these many
electrons if all the donors are ionized. And it has, because of the mass action law which
438
is applicable at equilibrium, it has about ni2 square by 1e15 which is about 1e5 holes per
cc.
So, that is the equilibrium condition right. So, if you were to not apply any voltage, and
if I were to just deplete the semiconductor, then this region just after x=xdn; which is
x=xdn+ s. At that interface, we had these many holes and that many electrons. But the
moment I apply forward bias, you are going to have a massive influx of holes from the p
to the n side. They are going to move through the depletion region, and they are all going
to appear here at this interface.
So, suddenly the whole population there which was initially 1e5 is going to spike, it is
going to spike all the way up to, you know a very large number. Several orders of
magnitude larger and that number depends on the bias voltage and we will see what that
number is. And because of recombination, this whole population is going to immediately
start declining in the semiconductor.
So, it is as though you see a huge injection of minority carriers, the holes of the minority
carriers in the N type material. So, there is an injection of minority carriers. And this
mind this minority carrier population is going to decrease with distance because of
recombination. And we have already looked at this phenomenon, when we looked at the
continuity equation and recombination generation mechanisms.
So, if you remember, if you look at the special case so, something called as a low level
injection, the already looked at what is going to happen when you have a few minority
carriers present in a material. And you know how the recombination controls the
population density of these minority carriers with distance. We have already looked at
that phenomenon, and we are going to be using that relation, here we are going to be
using that special case the continuity equation are to analyze your P-N junction.
Now, this is going to be the overall idea, they are going to first. So, strategically I think it
is useful to define the plan for deriving the current voltage characteristics. And I should
have probably put a slide on that, but nevertheless I will write it out quite neatly. So, the
plan for the current voltage characteristics for deriving the I-V characteristics of a P-N
junction; so, what is the plan? The plan is first identify the minority carrier concentration
on either sides; that means, it is the injected holes in the n side and injected electrons on
the p side. We are going to identify the minority carrier concentration as a function of x
439
as a function of distance. We are going to find out what is this function, what is f of x?
The determines how the population of holes. So, let me call this as delta p on the n side
which is the excess holes on the n side.
How does the excess holes on the n side vary with distance? We need to identify this
point first. How do the excess electrons on the p side vary with distance? You are going
to identify this. And now since they are going to vary with the distance, there is going to
be a diffusion current, there is a diffusion current. So, how do we identify the diffusion
current? It is proportional it is qDnd∆np/dx. That is the diffusion of electrons and
qDpd∆pn/dx is the diffusion of holes. We are going to then identify the diffusion current
components and then add these 2 current components to get the total current. So, a very
simple idea. First identify the minority carrier distribution n with the distance. And then
use that function to differentiate that function and get the diffusion current of electrons
and holes.
So, that is our strategy; that is our game plan to identify the current voltage
characteristics. And we are going to do this by using continuity equation, and then
identify the correct boundary conditions and identify the correct terms for these 2
expressions for these 2 and then use that to identify the current. So, the first part of it is
to solve the continuity equation. Now, what we have to do is, we are going to shift the
coordinate system a little. Because we are very comfortable is defining an x equal to 0 at
the point of injection. So, what I am going to do is, our initial coordinate system we
defined x equal to 0 here at the junction, and then we had at x equal to x d n we had the
boundary of the depletion region on the n side, and at x equal to x at x equal to minus x d
p, we had the boundary of the depletion region on the p side.
So, at x=xdn is the boundary the depletion region on the n side, and at x equal to minus x
d p we have the boundary of the depletion region on the p side. So, this was the initial
coordinate systems, that is my x. But now I want to change it, I am going to translate I
am going to create a new coordinate system, just to make my mathematics a bit simpler.
And in that new coordinate system instead of x I am going to use x prime or x dash if
you like. And I am going to mark the x x dash equal to 0 the origin point at x equal to x d
n. So, x equal to x d n corresponds to x=x’ equal to 0. And this is the coordinate system
are going to use only for the n side. I am not going to use this for the p side.
440
For the p side, I am going to use a different coordinate system; which is x equal to
double prime. And it is heading off this way, it is heading off in the minus x direction.
So, the positive values of x double prime are the negative values of x. And x double
prime has got an origin located at your x equal to -x d p. When x is equal to -xdpx double
prime is equal to 0.
And when x is equal to xdnx’ is equal to 0. So, I am going to define 2 new coordinate
systems of this curve so that I can write my continuity equation in a manner that I am
very familiar with and it is not going to change the physics that is just going to simplify
the math a lot. Instead of using the x coordinate system and using a translation of
coordinates I am just going to use, something I am going to use this transformation and
we will end up the continuity equation that we are very, very familiar with.
So, you will find that the continuity equation for the, that describes the whole population,
the injected whole population on the n side. Which is the delta p n on the n side is this.
So, if you are wondering where this expression came from, I would strongly urge you to
go back to the that the points where we discussed the continuity equation, and you will
find that this was one of the special case is discussed which is the minority carrier low
level injection of minority carriers. So, here you have your diffusion coefficient for
holes. You have your minority carrier lifetime for holes, and delta p n is the excess
minority carrier concentration. In some sense what does that mean? It means that let us
say p n represents the number of holes on the n side per unit volume. P n o is the
equilibrium number of holes on the n side per unit volume. Delta p, now we have shifted
this device out of equilibrium and we now have p n minus p n o which is the excess
minority carriers per unit volume on the n side being your delta p n. So, that is your ∆pn.
And clearly that spatial derivative of this term is nothing but the spatial derivative of this
term. And therefore, we can write everything every expression in terms of delta p n of x
dash. So, this is the continuity equation, and we know that the solution to that is given by
this particular expression here. Now, the key now is to identify the correct boundary
conditions. Now the first boundary condition we will use is that x and x x dash is equal
to infinite; that is, if I have an infinitely long P-N junction diode or if I have a very large
P-N junction diode. And these are my bound depletion regions. So, my x dash coordinate
system goes like this. At x star is equal to infinite, there are no minority carriers present;
441
which means that all the minority carriers have recombined by then and nothing exists.
So, Pn is equal to Pno; that means, these 2 values are exactly the same; these 2 values are
exactly the same and delta ∆pn ∆pn is 0 that is very far from the junction. That is a very
reasonable assumption because there is no band bending, the semiconductor is going to
be like the bulk semiconductor has got it has got no idea as to what the junction is doing
etcetera. And this assumption is something called is a long base diode approximation.
And this term base comes and it will make a lots lot of sense when we discuss bats, but,
but you can you can you can note this down as the long days diode approximation.
So, if you say that this is the first boundary condition and applied to that expression, you
will find that the term a automatically has to be 0, because at x dash is equal to infinite
this term blows up. It becomes infinite, but my delta p n is 0 and that can happen only if
A is 0. So, I cannot allow this term to exist. So, this term has to disappear and that
happens only when A is 0. So, remember we are using we are looking at the n side, we
are looking at the n side this is not these expressions are not V a lid for the p side. We are
looking at the holes injected into the n side and are using the x dash coordinate systems.
We are only looking at the n side of the junction. So, that is written here, but just in case
that is not clear.
So, what is the next boundary condition? I need to identify this coefficient B, and once I
identify B I have my expression for delta p n, the next boundary condition is to identify
the whole population at x dash is equal to 0.
442
And here I have given you a short derivation, but the in short the answer is this. The
whole population so, we have we have your x equal to x d n or x dash is equal to 0
coordinate system sitting here and that is my x dash. So, we have some excess holes
injected there and it all tapers down to 0. So, let me just write this very clearly. So, this
point here is my x dash is equal to 0, which corresponds to x equal to x d n. And on the y
axis I am drawing sketching the delta p n; which is the excess holes injected into the n
side because of the application of the forward bias. And these holes are recombining and
they have got a minority carrier lifetime of tau p and therefore, you see this exponential.
Now, what is the whole population at this interface? What is delta p n at x dash is equal
to 0. How many holes were injected into the n type semiconductor? Clearly they have to
depend upon V a, because it is the applied voltage. And clearly they have to depend upon
the exponential of q V a, because that is what the Boltzmann statistics tells you. But you
will find that the expression is exactly given by this particular term here. And there is a
short derivation it you end up with an expression by simply defining the built in potential
as an integral of the electric field. That is of course, with some near equilibrium
approximations. But what this expression tells you is that, my delta p n which is the
excess electrons on the n side; excess holes injected into the n side at x dash equal to 0
which is at x=xdn is equal to p n o which is the equilibrium hole concentration on the n
𝑞𝑉𝐴
side into (𝑒 𝑘𝑇 − 1).
443
𝑞𝑉𝐴
Which means at Va equal to 0 at equilibrium is 1, 1 minus 1 is 0 and therefore, there
𝑘𝑇
is no excess holes injected into the n side which makes a lot of sense. But then if I were
to make my V a positive, I find that my excess electron, excess holes injected into the n
side increases exponentially with the increase in applied voltage. So, this is the second
boundary condition that we will apply.
And that second boundary condition defines what my B is. So, if you remember my delta
p n of x dash had 2 terms, it had an A e to the power x dash by D p tau p, and it had a B e
to the power minus x dash by d p tau p and square root of D p tau p, sorry.
And we saw that A has to be 0, because of the long base approximation. But we need to
find out what B is, and finding B we are going to use a condition, that an x equal to x x
𝑞𝑉𝐴
dash is equal to 0 my ∆pn is equal to 𝑝𝑛𝑜 (𝑒 𝑘𝑇 − 1). So, that is my value. So, I substitute
x’=0, that makes this term one and therefore, B is equal to this particular term here. That
is exactly what is mentioned in that equation. And therefore, my complete expression for
delta p n x dash is this particular term which is this.
Now, here I forgot to mention, I forgot to mention one key component. You will find that
this D p tau p has become L p square. So, that is that is quite an important definition. D p
is the diffusion coefficient of holes tau p is the minority carrier lifetime; which means it
is the average time hole which is a minority carrier would last as a free carrier on and in
444
an ns n doped semiconductor because it is going to recombine eventually. So, it is the
lifetime of the minority carrier. In which case, in this case it is p hole. And this product
in this product of D𝜏𝑝 p tau p, it is quite significant, if you take the square root to this
product. We define that square root as a term called as L P; which is the diffusion length
of the holes.
So, this is something called as a diffusion length, and what it tells you is that it gives you
an indicator. So, Dp is telling you how quickly the species would diffuse and tau p is
telling you for how much time they would diffuse. And therefore, in some sense the
speed of diffusion and the time of diffusion is cut a length component to it. So, but since
D p is not exactly velocity it is a Dp 𝜏𝑝 is exact is equal to Lp2 , it is a square of this
diffusion length. And this diffusion length is an indicator to the distance through which if
the just like you have an a time constant defining the, you know, the typical response or a
measure of the response of a RC circuit, you have L p defining the measure of the length
that t minority carriers are diffused into the material.
So, you have Lp being a measure of the speed of diffusion and the minority carrier
lifetime. So, I forgot to mention that explicitly. So, you find that this is the expression for
the excess holes injected into the N type material and a P-N junction diode in forward
bias. So, you have this term here which is your B and you have that term here which tells
you an exponential dependence with distance. And L p is your diffusion length where L
p is equal to square root of Dp 𝜏𝑝 . And what is D p? D p is approximately your kT𝜇/𝑞
from your Einstein’s relation.
But of course, do you remember that we are not at equilibrium? We are just off
equilibrium. So, you are fine with using these estimates, all right. So now, we have the
concentration variation. So, that is what we have been looking for we have delta p n as a
function of x’. And therefore, since we know the concentration variation, we can now
easily find the diffusion current. And the diffusion current is simply q D p into this
derivative and that is a negative sign because we are talking about holes. And that
answer turns out to be this. So, this is the diffusion current, you see that the diffusion
current is dependent on x dash and rightly so, because the concentration profile is
varying with x dash.
445
So, what we should be looking at is the easiest way to compute the current in the P-N
junction. So, if that is the junction, that is the x d n and that is the minus x d p. So, this is
the n side that is the p side that is x equal to minus x d p; and that is x equal to x d n or x
dash equal to 0. We need to compute the diffusion current here at x dash equal to 0
because after this point, it becomes a little tricky to compute the total current. Because
you have a diffusion component and you have the minority carriers moving in etcetera,
etcetera.
So, we want to compute the diffusion current at this edge due to the holes. We want to
compute the diffusion current at this edge due to the electrons, and sum these 2 diffusion
current components to give you the total current. So, at x equal to x dash equal to 0, my
diffusion current simply becomes this particular term. So, what have we done? Just to
summarize the strategy, we have first identified so, because we applied a forward bias,
we saw that holes were injected into the n side and electrons were injected into the p
side.
And we firstly, calculated the distribution of this excess injected holes on the n side. And
then we use that by solving the continuity equation, and we use that estimate to calculate
the diffusion current due to holes in the n side. And in particular we are interested in the
diffusion current value at x dash equal to 0 that is at the depletion boundary in the n side
material. And with that we have completed all our analysis of the current due to holes.
We have still not done anything with regards to electrons, but the electrons follow an
exact you know process, and it is easy to calculate the electron current on the p side as
well.
446
So, once again you solve the continuity equation to calculate the injected electrons into
the p side. You find this profile, and then you find your diffusion current due to the
electrons. And you will find the diffusion current due to the electrons at the depletion
boundary or the p side is given by this expression, very symmetric solutions.
And therefore, the total current in the P-N junction is the summation of the diffusion
current due to the electrons and the current due to the holes and that turns out to be this
expression here; which is got an exponential dependence with the applied voltage. We
saw that to be true even in the case of a metal semiconductor junction, but the key
difference between a Schottky junction and the P-N junction; is that in a Schottky
junction it was the majority carriers that were carrying the current, and in the P-N
junction it is the minority carriers that are carrying the current. Quite a significant
difference, which puts out the difference in the operating speeds of these 2 devices.
But both of these can operate as a diode or can have a rectifying nature in the current. So,
this entire term can be treated as a constant right. So, this is what the charge in an
electron; that is the diffusion coefficient of electrons, the diffusion coefficient of holes L
n; which is basically square root of Dn D𝜏𝑛 and L p which is √(tp𝜏𝑝 ). And pno which is
the equilibrium concentration of holes in the n type material n p o which is the
equilibrium concentration of N in the P type material; which is basically depend on the
doping. And therefore, this entire term is treated as a constant, and we will denote that as
447
J naught not I naught we are talking about current densities here J naught. So, that is a
𝑞𝑉𝐴
typo there it denotes. So, j is equal to 𝐽𝑜 (𝑒 𝑘𝑇 − 1), and this is an expression that most of
you will be very familiar with from your high school studies. So, this tells you that the
current and the P-N junction varies like this. It is an exponential relation and this is the
derivation of the current voltage relation.
448
Lecture - 27
PN Junctions: Small Signal Impedance
Now, let us move on to another topic which is to try and identify the effective Impedance
of a PN junction. The case, just summarize the current voltage characteristics, we found
that in a PN junction, you had a depletion region; we derived the electrostatics for it. And
we found that when you apply a forward bias, we had excess holes being injected in the
n type and excess electrons being injected into the p type and these excess holes would
recombine and therefore, you had a concentration gradient and similar likewise with the
and injected electrons.
And this concentration gradient gave rise to a diffusion current for holes and a diffusion
current for electrons. And we computed the diffusion current at this boundary at equal to
minus xdp and at x=xdn and therefore, we identify the complete current voltage
characteristics for a PN junction diode so, that is what we did.
449
So, now we know the current voltage characteristics and we found that J was you know
𝑞𝑉𝐴
some constant into 𝑝𝑛𝑜 (𝑒 𝑘𝑇 − 1) and just to recollect, what this constant was it had all
these terms in it ok.
It had q D n L n n p o D p L p and P n o ok, L p is the diffusion length, D p is the

diffusion coefficients. So, this is the current voltage characteristics, now however, there
are some finer points to all these discussions ok. And we will sort of discuss these finer
points by looking at the calculation of the impedance of a PN junction ok. So, if we have
450
a PN junction in a circuit, what is the effective representation for this device? So, this
very strongly depends on whether you are operating in reverse bias or in forward bias ok.
So, in reverse bias so, if you recollect your metal semiconductor junctions, we had a
depletion capacitance and this PN junction is also going to offer an effective resistance
ok. So, what we are interested in is something called as the small signal impedance and
we have discussed this before and what it implies is that if you have this device in a
black box.
So, let us say here you have your PN junction diode in a black box and you have these
two terminals and you apply a you apply maybe so, you say apply a DC bias voltage
firstly and on top of that bias voltage you apply a small signal AC ok. So, you have a DC
voltage and on top of that you have a small signal acc AC voltage and around the small
signal behavior you can linearize this device.
And therefore, the small signal impedance that, we are looking for is basically your
response of the device to this small signal, which is how does the current through the
device vary in correspondence with the applied small signal voltage. So, this is the small
signal impedance, which were interested in. So, if you think about the PN junction
reverse bias so, if this is the band bending diagram, you have a junction here and you
have your depletion region boundaries here, so, that is a p side that is the n side so, let us
say this is a equilibrium condition.
So, in reverse bias the depletion region increases a little right and the band bending
becomes a little steeper. So, if I were to fluctuate, if I were to keep my device in reverse
bias so, which means that I apply large negative voltage. So, V a is less than 0 and on top
of that I apply a small signal AC voltage, what is the effective small signal impedance of
this PN junction? So, that is the question asked on the slide ok. And the answer is there
are two components here, you have a heavily doped n side and you have heavily doped p
side and at the interface you have a depleted region ok.
451
So, you have your p you have your p, you have your n and this depleted region has got
no free carriers. But, this these two regions are almost metallic right, you have a large
number of electrons, you have a large number of holes and this region is almost metallic,
but here in the middle you have a depleted region. So, when I say almost metallic it is
not to be taken in the sense of it being treated as a metal, what I mean is the resistance is
very low. So, you have a device that looks like a capacitor ok and we have already
looked at this in the metal semiconductor junction.
Wherein you have these two regions and then you have this depleted region and because
of the small signal fluctuation the depletion region widths fluctuate ok. So, if the voltage
goes more negative then so, let us say we are this is our reverse bias. So, let us say we
have applied say minus 5 volts and let us say we have a small signal AC here.
So, if the voltage is in this region it is low it is more reverse bias than 5 volts and
therefore, the depletion region increases and uncovers more fixed ions. And then as it
goes to 5 volts it reaches it is you know mean depletion width and as it goes to above
slightly above 5 volts, then it covers some of the ions and it reduces in size and it
performs little fluctuation. And it is this fluctuation that results in an effective fluctuation
in the charge and since there is an effective fluctuation in the charge, we can calculate d
Q by d V a and say that this is your effective small signal capacitance.
452
So, that is what we did in the case of a metal semiconductor junction. So, what is dQ/dV
a Q is nothing, but N D into x d n on the n side and NA into x d p on the p side. Since, x
d n and x d p are changing with respect to the bias ok, you have your dxdn/dV a plus dx
dp by d V a giving you this capacitive component. Now, that is one method to calculate
the capacitance, the other method to calculate the capacitance is to simply think of it as a
parallel plate capacitor in a very crude manner ok.
Because, you know these two regions have got low resistance and this region here is
depleted. And therefore, if the junction has got an effective cross section area of A my
capacitance is simply epsilon A/x d which is x d n plus x d p. And since x d n and x d p
are dependent on the built in potential and the applied voltage in this manner, this is my
total small signal capacitance ok.
And we saw the similar argument even in the case of a capacitance of a schottky junction
a metal semiconductor schottky junction. So, that is the small signal capacitance of a
diode in reverse bias ok, in forward bias there is something more interesting that happens
we will come to that. What about the resistance of a diode in reverse bias we already
know the current voltage characteristics.
453
And therefore my small signal resistance is d V a by d I and that is easily obtained by

taking the derivative of d I by d V a and then taking the inverse of that. And you will find
that the small signal resistance is k T by q into I plus I naught, where what is I naught I
naught is nothing, but your J naught times the cross section area and I is nothing, but the
current density into the cross section area ok. And we have already seen what J naught is
all the parameters it depends upon. So, in the in reverse bias this is the small signal
capacitance and that is the small signal resistance ok.
454
But in forward bias ok, let us sorry that is a repeat of the slide in forward bias, you do
have the same capacitor component ok. That capacitance which we discussed is
something called as a depletion capacitance, you do have a depletion capacitance, you
once again do have a small signal resistance.
But you have an additional capacitance term which is called as the diffusion capacitance,
and I will tell you what this is. And another parameter that now starts playing an
important role, because the currents are now larger is simply the series resistance due to
the P and N junctions. This was there even in the case of a reverse bias, but for practical
purposes it need not be considered, because the currents were not very large, but if the
currents grow very large then this series resistance can cause a huge significant, it can
cause a significant voltage drop across the PN junction.
And that series resistance is simply because, the doping of your n n p type. And due to
any metal contacts that you may have or any external circuit now this adds in a series
resistance ok. But apart from that you have your depletion capacitance, you have your
small signal resistance should be the diode behavior and you also have something called
as a diffusion capacitance.
Now, we have already looked at depletion capacitance, we will not worry about that we
have already calculated what d V a by d I is which is your diode resistance, we will not
worry about that. The series resistance is quite simple to calculate, because the resistivity
is simply q1/q𝜇𝑛 on the inside and in the case of p side is 1/q𝜇𝑝 on the p side 𝜇𝑛 into n
where n is the electron concentration and 𝜇𝑝 into pvp is the hole concentration ok.
So, that is your series resistance, if you only consider the diode alone, but what we have
never discussed so, far is this component called as diffusive capacitance diffusion
capacitance. So, what is the diffusion capacitance very intuitively ok?
455
Now, in forward bias what we have done is we have injected the minority carriers all
right. So, we now have a minority carrier injection, because the large number of large
amount of carriers diffusing across the junction. So, if we have to apply if we were to
change and we already saw, you already noticed that this term was P n naught into e to
the power q V a by k T minus 1.
And this function that was also my component which is B and this function was given by
𝑥′
−
𝐿𝑝
𝐵𝑒 ok, just to recollect. So, obviously, this minority carrier concentration depends
upon Va and the concentration at every point in space after that depends upon Va ok. So,
if you take a very broad definition of what capacitance means, it is if I were to put this
diode in a black box ok so, if I were to put this diode in a black box.
456
And I vary the voltage across and the amount of charge so, as the voltage varies if the
voltage small signal fluctuates, the total charge content inside this black box fluctuates
ok. And it is this ratio dQ/dV a which is my capacitance that is the that is my small
signal capacitance. In reverse bias there was no minority carriers being injected, I mean
it was it is not considered at all ok, because the current is purely drift paced and we really
never worried about the injected minority carriers.
And therefore, the only capacitance we considered in reverse bias was the depletion
capacitance; we never bothered about this new component here the diffusion
capacitance. And therefore, this dQ was only due to the movement of xdnx dn and xd p.
So, any voltage I apply the way that, the black box would respond is by uncovering or
covering the fixed ions. It would uncover or cover the fixed ions and that was the only
response the device gave and therefore, we could model it as a parallel plate capacitor
and get away with everything.
But now, we have a rather tricky component here ok, but in forward bias not only do we
have the depletion capacitance due to the depletion width change, which is already
accounted for, but this minority carrier concentration will also change. So, if I were to
apply my small signal if the small signal so, now, we are in forward bias let us say we
have got a forward bias, let us say its plus one volt now drop of that I have a small signal
fluctuation.
457
So, as I as the small signal goes above 1 volt, this minority carrier concentration will
increase and as it goes below one volt it will decrease ok. So, therefore, any fluctuation
here will be seen like a wave like response in the minority carrier concentration, the
injected minority carriers. So, as this applied voltage fluctuates, not only would the
diffusion a depletion width fluctuate, but also the injected minority carrier concentration
would also fluctuate. And this fluctuation in charge was never accounted for in the
calculation of the depletion capacitance and that fluctuation in charge is accounted for in
what we call as a diffusion capacitance ok.
So, now that is a very qualitative understanding of what diffusion capacitances, but how
do we quantify this ok, we want to write a nice mathematical expression for the diffusion
capacitance. And what we will do is we will quickly look at how to go about calculating
the diffusion capacitance. So, clearly it arises from the concentration variation the
transience of the injected minority carriers right so, that is what is causing the diffusion
capacitance.
So, therefore, we are starting point must be in order to derive the diffusion capacitance
starting point must be the continuity equation, but not in steady state, we must take into
account that transient ok. So, we are no longer interested if it was steady state, we could
simply say that you know this component was equal to your d delta P n by d delta P n by
458
I am sorry Δ𝑝𝑛/ 𝜏𝑝 and just get away with it. But now we want to look at the transient
response ok.
So, transient response of the injected carriers and therefore, we have a time dependent
derivative term on the left side, it is not steady state and therefore, if you recollect you
continuity equation you will find that this becomes your continuity equation ok. Now,
this is the total injected this is the difference between the total injected charge and the
equilibrium charge. So, that is my Δ𝑝𝑛 , now because of the DC bias is called a DC
component and because of the applied AC signal it is got an AC component also. So, this
is the component that we are interested in is fluctuating.
The DC component is already been accounted for calculating the DC current ok so, we
looked at this already. So, in general this delta P and the injected carriers has got a DC
part and it is got an AC part and I will define this AC part with this little tilde that at the
top ok. And the DC part with just an over bar or just a straight line there ok, so, this v v
line on top of delta P and describes the AC part.
So, now, I can split this continuity equation for the DC and AC part. So, if the DC part
this time dependence is all 0 and we are talking about steady state. And therefore, this is
my continuity equation that we have already looked at and solved, that is what we used
for the current there was no fluctuating voltage component there.
459
But, now since we have a fluctuating voltage the AC part alone that is if you get rid of
the DC component of it. And we take only the AC component we find that, this is the
continuity equation of interest, there is a transient, there is a diffusive term and there is
this recombination term ok.
So, now how do we solve this continuity equation, it is got a time and it is got a time and
space dependence ok. So, usually what we do that is the best approach for such things is
to use separation of these two variables ok. So, we will say that if these two are truly
independent and it so, happens so, that is a it is a good enough assumption and we can
get away with it ok. We can truly separate the time and spatial component and get away
with it and therefore, we will do that now I will tell you why very intuitively ok.
So, this is the AC component of the injected charge. And we will say that that is equal to
Δ𝑝𝑛∗ ok, which is the magnitude which is got only the spatial part of it into 𝑒 𝑗𝜔𝑡 , which is
got the time dependence ok. So, very intuitively you can think of it is a wave right it is a
it is, your fluctuating it is like somebody holding a string, which is tied to one end and it
is tied to one end implies it is like, if you take the long base approximation at x equal to
infinite at x dash equal to infinite you have that end tied to delta P n equal to 0 that is not
moving ok.
And then on the other side they are fluctuating the voltage, in other hand somebody is
pulling string up and down or you can think of it as water ok, either way. And you have
460
this little you know variation of Δ𝑝𝑛1 in space and in time and it is got this wave like
behavior. And therefore, very intuitively we will just say that this time dependence is
simply 𝑒 𝑗𝜔𝑡 ok, we getting a wave without explaining a lot of things, but we are using
strong intuition here and we are actually fine with these approximations.
So, now we will just substitute this expression, for delta P n the AC component of delta P
n is given by this and we will substitute this approximation inside our continuity
equation. And when you make the substitution, you see that this time derivative becomes
Δ𝑝𝑛∗ 𝑒 𝑗𝜔𝑡 , this term here will give you this component it is that is Δ𝑝𝑛∗ there ok.
And this component is simply a direct substitution here and you find that all the 𝑒 𝑗𝜔𝑡
terms conveniently cancel away and you will end up with this expression ok, which is
neatly written out over here ok. So, that is what the continuity equation boils down to if
you were to take only the AC component and use a separation of variables.
Now, if you look at this term ok, inside the bracket inside the square bracket, it is a
𝜏𝑝
complex number it is 1+𝜔 . Now, it is a complex number, but think just think about
𝐷𝑝 𝜏𝑝
it a little omega tau p is a dimensionless term and 1 plus j omega tau p is some complex
number that has got no dimensions ok.
461
So, we will just say it is some complex number 𝐷𝑝 𝜏𝑝 is nothing, but your Lp2 square ok.
So, what you have here is Δ𝑝𝑛∗ /Lp2 into a let us into this complex number 1+𝜔𝜏𝑝 . Now,
we will take these two terms a complex number divided by Lp2.
And we will say that this entire thing is equivalent to 1 by L p star square, where L p star
is some kind of a complex diffusion length ok. So, just a mathematical construct here, it
is a complex diffusion length; it is a Lp2 divided by 1 by divided by 1+𝜔𝜏𝑝 ok. So, that is
our mathematical construct and we are doing it in order to get a differential equation that
is very familiar, which is d square delta P n by dx square is equal to Δ𝑝𝑛∗ /𝐿2𝑝 . So, , L p
star is a complex number it is in fact, the square root of this complex number if you wish
square root of this particular term.
462
So, now we already know the solution to that equation, let me just rewrite that equation
𝑑 2 Δ𝑝𝑛 Δ𝑝𝑛∗
here so, you have your =( ) 2. Now, it is this come to my attention that, I have
𝑑𝑥 2 𝐿2𝑝
missed out mentioning something here ok.
So, throughout I have been using this variable x ok. So, please do remember this x
corresponds to x dash ok, we are still interested in the translated coordinate system ok.
So, if you have any query I am sure all of you understand this by now, but if you have
any questions you can just get back to me on that.
So, x so we are talking about x’ ok. So, if you still use our old PN junction diode
diagrams, but nevertheless I think you understand the context of what we are talking
about. And the solution to this differential equation is again given by this particular term
here, where B you know is your injected carrier minority carrier hole concentration at the
depletion boundary on the n side. And that is equal to Pnoe to the power in the previous
𝑞𝑉𝐴
case, we saw it is
𝑘𝑇
, but now it is going to be q V a plus the small signal Va/kT the
small signal V a has got a time based component ok.
So, d turns out to be this particular term it is going to be Pnoe to the power q into the DC
voltage plus small signal voltage by kT-1 ok. So, that is the expression so, there is a
minus 1 term here which is missing ok.
463
And we could now approximate this ok, we can approximate this and say that this
particular term is nothing, but P noe to the power will just take out the large signal or the
DC components out k T and you leave behind only the small signal part, but we will
approximate that small signal part ok. And we will say that it is we will use a Taylor
1
expansion on that and say that it is 𝑞𝑉
𝐴
− 1 ok so, there is a minus 1 sorry about all
𝑘𝑇
these typos.
And therefore, the small signal component the small signal part alone becomes just this
ok. So, all these cancels away and this is your small signal component. So, what we have
done here is we have gotten rid of the exponential with regards to the small signal
component; we have just trying to linearize this and say it is dependent on V a ok. So,
that is the whole point of small signal analysis which is to which is the ability to use a
Taylor expansion and linearize things and we have done that here.
And therefore, Δ𝑝𝑛 at x=0 is going to be B e to the power e to the power minus x by L p
star and at x equal to 0 it is simply equal to B. And therefore, , Δ𝑝𝑛 at x=0 delta P n star
at x equal to 0 is simply given by this expression wherein, we have used Taylor
expansion and linearized the V a term ok.
464
So, in summary this is this quite important this is the expression that we will use. And
now what is the current component, the current is because of the diffusion term all right,
it is q into a area of cross section into D p into D delta P n star by dx and that turns out to
−𝑥
𝐿∗𝑝
be this, because we now know what is B and we now know what is B𝑒 . And
therefore, we know what this derivative is and we can now find the current at x dash
equal to 0 and you will end up with this expression here.
So, that is the current that is the small signal current due to the fluctuation, due to the
voltage fluctuation the small signal voltage fluctuation, set at a DC bias of capital V a.
And therefore, the impedance term a general complex impedance is d V a by d i because
of the I being that small signal current and that will end up being a number like this ok.
And since it is got this term L p star, this is definitely a complex number therefore, it is
got a real term add a complex term and the imaginary part of that is the diffusion
coefficient.
So, it is quite understandable if the last bit was quite confusing, but it is just a
mathematical follow up of everything that we have done. So, once again if there is any
questions on this feel free to send in a message ok.
465
But that is a very intuitive analysis of what they how did the way to go about calculating
the diffusion capacitance ok, it is because the key message here is, because it is because
of the fluctuations and the injected minority carriers, which is quite significant in
forward bias. And therefore, the diffusion capacitance is quite important when you are
talking about a PN junction in forward bias.
So, therefore, to summarize the PN junction forward biased has got different terms, it is
𝛿𝑉𝐴
got it is own small signal resistance that is it is got it is depletion capacitance which
𝛿𝐼
is say 𝜀 as A/xd. And then it is got its diffusion capacitance, which is defined by this
analysis here and it is got a series resistance which is important because of the large
currents. And therefore, more significant voltage drop since, it is not advisable to ignore
the series resistance because, the doping profile is not defined the metal semiconductor
junction is not defined etcetera.
So, if this is the equivalent circuit of the PN junction in forward bias. So, to summarize
everything you have done so far we looked at what the PN junction is we looked at the
electrostatics, we looked at the current voltage characteristics. And the key message was
that the current voltage characteristics has got a rectifying nature to it and it is, because
the diffusion of minority carriers that you have a current voltage characteristics, defined
by an exponential function. And then finally, we can used all these information to
calculate the impedance of the PN junction diode in forward and reverse bias.
466
Lecture - 28
PN Junctions : Non-Idealities
So, before we close to Schottky P-N junction diodes let us just look at some of the non-
idealities ok. So, it is not, if you look at a perfectly ideal P-N junction diode, it would be
the way we defined it to be but there are some non-idealities ok. First we look at the non-
idealities in reverse bias ok. I am not sure if that is all very clear, but I will write it out
so, reverse bias non-idealities.
So, the first one is something called as the avalanche break, which is to say that it is not
that the P-N junction can keep operating with the same current voltage characteristics for
an infinitely large negative voltage. There is a large enough there is a voltage, which is
so this is my, this is when I apply an increasing reverse bias V a keeps getting more and
more negative. So, far what our equations tell us is that the P-N junction would continue
having a current that is exponentially dependent and continue increasing in magnitude in
a very small manner based on our function but that is not really true.
You will reach a point where the electric field due to the band bending. See, the electric
field the bending keeps increasing as you apply the reverse bias right, it gets that band
467
bending gets steeper and steeper. And there will be a point where the electric field is so
large, that the electron drift becomes so large that the electrons can now start colliding
with the other silicon atoms and pull free electrons out from those silicon atoms ok. So,
this is and this leads to a multiplication effect. You have 1 electron you know getting
another electron out and then these would these were in turn, start generate pulling out
more and more electrons because this very high electric field. And therefore, you have
this avalanche of electrons that start adding to the current and you will find that there is a
massive surge in the current in the device simply cannot take that surge and it simply
breaks down ok.
So, this is a purely drift based mechanism, but you have a large your electron population
increases dramatically because of this breakdown phenomena. And therefore, the
breakdown voltage, the voltage at which this happens, what is the V a at which this
happens? The Va at which that happens is defined in terms of the electric field that is
needed to break the semiconductor term.
So, at Va =Vbr you find that you know using your regular expression of electric field in,
if you write your electric field in terms the built in potential you have this expression.
And therefore, your breakdown voltage is simply given by this term where the electric
field is now large enough to create a breakdown. And therefore, if you write these
expressions in this manner you find that the breakdown voltage is inversely proportional
to the dopant concentrations. So, if I dope it very large the breakdown voltage decreases.
468
The 3rd point is something called as is a third non-ideality is defined is related to the
generation of carriers in the depletion region ok. So, in a reverse biased P-N junction
your depletion region is quite large and it is quite strongly depleted, with the there being
no diffusion current and there being a drift current due to the electrons and holes. And in
the in this depletion region your if you look at the equilibrium condition in a
semiconductor your np=𝑛𝑖2 so, that is your mass action law.
But since this region is not depleted, your np is less than ni square, which means that if
you go back to the recombination generation mechanisms we looked at. If we look at the
rate of recombination and generation this implies that this implies that, it this implies that
it is going to prefer generating carriers as compared to recombining carriers.
In some sense you have altered by depleting this a P-N junction you have altered this
balance of generation and recombination of carriers and you have encouraged generation
a bit more than recombination. That is the carriers will effectively you will be generating
carriers in the heavily depleted P-N, heavily reverse biased P-N junction diode. And
whatever carriers are generated here, so let us say you generate an electron and an hole,
this electron would quickly run downhill to the other side and the hole will quickly run
uphill to the other side.
And this mechanism is the key between the is key to the operation of something called as
a photodiode ok. So, if you want to sense light, particularly low levels of light what we
469
do is be reverse bias the P-N junction diode ok. So, you apply a very huge reverse bias
voltage and when light falls on this junction you create electron hole pairs and this
electron hole pair creation is encouraged and the recombination is now going to be quite
low. And these electron hole pairs now quickly drift to the other side because of this
large electric field and therefore, you can collect these electrons and holes and you have
a current that depends upon the amount of light you throw in ok.
So, if you were to actually throw in light the curve would actually shift you will offset
the P-N junction curve a little by adding more by but because the generation of carriers.
And therefore, this generation of carriers, if it has a certain generation rate G, it would
create an effective current and that effective current is simply G dx integrated from
minus x dn to minus x dp to x dn which is basically telling you that it is the total
generated carriers in this region times the area, times q which is basically a drift
component to the excess carriers generated.
So, that is the excess current that you would create by because of generation. And in a
heavily depleted region we already know what the generation rate is. If you look at the if
you go back to the topic on recombinational generation you will find that we look at one
special case which is recombination generation mechanisms in a heavily depleted region
and we find that the generation rate is given by this particular term.
And therefore, the total current in a P-N junction diode after taking into account you
know your generation recombination the generation of carriers in the depletion region. In
reverse bias if you include this non-ideality the total current is not only the current due to
you know due to the exponential dependence of the voltage, but it is also due to the
generated carriers and this current adds on to the current due to the diffusive component
of the current ok.
470
So, now what are the non-idealities in forward bias? In forward bias the first thing to
notice is so the currents are going to be quite large, the currents are quite large. And
because the currents are large you do have an exponential climb initially but, then
because of the series resistance or because of any voltage drops in the P-N junction diode
ok. If you remember the equivalent circuit or the P-N junction it had a diffusion
capacitance, a depletion capacitance and a diode resistance as well as a series resistance
due to the bulk regions and the contacts they make with the metal etcetera.
And if the currents are quite large it can lead to a voltage drop across the series resistance
which will effectively implied the diode is seeing a much smaller voltage and that
phenomena has to be taken into account and it will lead to lead to a saturation in the
diode characteristics. So, the diode current would look a bit like this, this is a logarithm
of the current versus the voltage.
So, initially you would see a slope of one and then slowly the slope will start tailoring
down because of these effects and this is all taken care of by adding something called as
an ideality factor to the equation. So, ideally your P-N junction relation had a term like
this it was (qVa/kT)- 1.
But then in order to account for all this we sort of tune down this exponential by adding a
factor here ok. So, this is a real number that you just earned the denominator and if this n
is something called as an ideality factor. It can be used to empirically model any such
471
behavior in the P-N junction. So, if n is equal to 1 then its behaving like a perfectly ideal
diode if n is greater than one then there are other effects that are sort of damping the
current in the diode.
Now, the second non-ideality in forward bias is clearly the opposite of what happens in
the depletion region as in reverse bias ok. So, it is the opposite of that which is in reverse
bias the depletion region saw an encouraging of generation mechanisms as compared to
recombination. But in forward bias the np product in the depletion region is much greater
than 𝑛𝑖2 and therefore, recombination is encouraged as compared to carrier generation ok.
And therefore, if you take into account the recombination this will contribute to a current
and the recombination current is essentially given by this term here ok. You have a
recombination rate and then you have your continued diffusive component. And since
you have excess recombination a diode and forward bias is something that is used to
build your light emitting diodes for example, because you have excess recombination
and therefore, you have emission of light if you choose the materials carefully.
The final non-ideality which we have not treated at all in this course is as to what
happens under very large forward bias. When V a is extremely large you have a large
number of carriers diffusing through and the continuity equation we used is not sufficient
to handle that ok. So, the continuity equation we used was delta pn by dx square is equal
to delta pn by tau ok.
So, this equation is valid only under something called as a low level injection
approximation. We have a very large Va, we have a high level of injection and therefore,
these equations need to be modified or you need to have a more thorough analysis of the
situation. And therefore, you cannot assume any things that we assumed in this case.
So, particularly this particular term for recombination cannot hold true if you have very
high level injection. So, these are some non-idealities that you must be aware of on over
all the P-N junction theory we have looked at in this class. With that we will conclude
our analysis of P-N junction.
472
Lecture - 29
Bipolar Junction Transistors (BJT)
So, far we have, you know, just to summarize, you know, our journey so far in this
course, we initially started off with basic ideas of quantum mechanics. And then we
extended this to develop ideas on solid state physics ok which is basically what happens
as you go from one electron sitting in a potential well, to an electron in a solid, where in
you have a periodic potential value because of all the atoms.
And then from here, we brought in basic ideas of semiconductor physics ok, and we
spent a lot of time discussing concepts of mobility, effective mass, you know,
recombination generation, the different charge transport mechanisms and the continuity
equation. And then we continued with our journey towards looking at junctions. So, this
was a very important step when and we looked at 2 kinds of junctions the metal
semiconductor junction and the P-N junction.
And we saw how we saw the tools and techniques right off using solving for
electrostatics using ideas like the Poisson’s equation, and then trying to use the
continuity equation to identify the current voltage characteristics etcetera. And now we
473
are going to take a next big leap in our journey, and we are going to start talking about
transistors ok.
So, these are so, we have now reached a point where the flavor of the course will slowly
shift from more fundamental physics and device physics towards more of circuit design,
ok. So, we are not here there, but the course will start taking that taking on that flavor.
With regards to transistors, you know, we you know firstly, we will just look at what are
transistors.
So, let us say, you know, just to just to sort of introduce this idea, these are essentially in
a simple language we saw the building blocks of all your modern day integrated circuits
right, along with all the diodes etcetera. So, the idea of a transistor is to construct a
switch or a resistor that you could control whose resistance you can control
electronically.
So, let us say you have you have a point A and, you know, you have a wire and you want
to send a signal or you want to send a message to point B. Now one way to do it is to
have your regular old switch which is probably used for controlling the lights at in your
homes. And you have a signal and whenever there is a signal that needs to be transmitted
to this side, you have somebody push this switch down and the switch is closed, and you
have you have your signal whatever your signal is being transmitted to this point B, ok.
474
And then we do not want the signal to be transferred you keep the switch open ok. So,
here the switch is closed and here the switch is open, and your message at point A will
now not be transferred to point B. So, how do I make the operation of the switch very
fast? If I have a mechanically operated switch, it is going to be a it is going to be a very
slow process; because your now dependent on that mechanical time constants ok, which
are which are quite slow ok.
So, the idea of a transistor is to develop a switch which can be controlled electronically
ok, an electronically controlled switch; which means that this control of the switch going
from on to off is done electronically, and your signals heading from A to B are all
electronic. And it is a very nice way to interface the real world with the world of
electronics, by having sensors and then having a bunch of switches performing different
operations. To take this one take this idea one step further, what we need here is; so, this
switch isn’t either it is got 2 states he is either got a state of on or off. And ideally what
do you mean by on is it is called a 0 resistance between A and B, and what do you mean
by off is that it is got an infinite resistance between A and B. So, that is that is our idea of
the switch.
But let us say that we would like this resistance to have any arbitrary value, and that we
would will not control this value also electronically. So, what we are talking about is
now a device that is a variable resistor that sits between A and B, and we would like to
control, control the resistance of this resistor electronically ok. So, maybe by applying a
small voltage or a small current etcetera and that is the idea of a transistor. So, this
device that we are talking about is the transistor. So, essentially the transistor should
have 3 points right. So, corresponding to the point A here, corresponding to the point B
here and corresponding a terminal which corresponds to where the control signal is
given, ok.
So, that is an between A and B you have this variable resistor. So, that is the idea of a
transistor ok, and that is what the transistor is. Now there can be different kinds of
transistors, but all of them would have certain properties. Since we have plenty use the
transistor for, you know, we imagine the transistor to be used in such an application.
What are the ideal properties of a transistor ok? So, the first one is that if I am using it in
on and off state, let us say I am using it as a bi stable device ok. It is either conducting or
non-conducting. What we would like is that, the on resistance of the transistor is 0 and
475
the off resistance of the transistor is infinite ok. That is what we would like. And then if
it is on, we would like the transistor to be able to carry an infinite amount of current.
We do not want any limitations on the current density current density. And we would
like it to operate at, you know, all kinds of voltages, you know, you should be able to
take very large voltages or very small voltages as and how it desire, ok. So, this a bit
arbitrary, but you will understand the meaning of this later. And if you look at the
voltage across if you if you look at the off current ok, if when transistors off we would
like it to have 0 current which is basically a consequence of the resistance being and off
state resistance and it should take a value of 0 amps at any voltage. So, we want no
matter how large the off voltage goes we would like this current to be 0 amps ok. So, by
V off here I mean the voltage difference between A and B.
So, even if V AB is infinite volts I would like my device to carry 0 amps of current ok.
That is probably asking for a lot, but that is what we want. And the other thing is that we
would not like the control signal. So, here you want to send a signal a message between
A and B, between from A to B right. You do not want any signal that you used to control
the transistor to influence that message. So, essentially what we want is that if there is an
if there is a resistor here ok. So, let me call that is R in, we would ideally want that
resistance to be infinite; or let us say if it has an impedance we would like that
impedance to be infinite ok. We do not want the control signal to impact the operation or
476
the message it is not corrupt or of influence the message being transferred between A and
B.
And we can continue listening a lot of these properties, but the other one is that we
would like the switch or the transient between on to off state, or the time taken to go
from on 2 off and back to on to be nearly 0. We would like all this to happen
instantaneously. So, we want all this operation to be having a 0 time constant, 0 second
time constant ok. So, these are some of the points that, you know, or some of the dream
assets; so, the dream properties that a transistor must have ok. But realities quite different
and depending on how you construct a transistor, you get some advantages and you lose
and you lose some of the benefits ok. So, what we will talk about is we will focus our
attention not only on these very general features of a transistor.
But we will try to look at, you know, how do you design some kinds of transistors, and
what are particularly the once that are very commonly used. And we will look at the
device physics of these devices, we will use all the tools and techniques we had so far
and look at the device physics of all these devices, and then extract the current voltage
characteristics. And finally, try to build in equivalent circuit ok. So, we are going to
follow the same protocol, we have already done this through metal semiconductor
junctions and P-N junctions we are going to firstly construct a transistor
477
Then use all our tools of device physics on this construct and then from here extract from
here extract your IV characteristics. So, the current voltage characteristics, and see
whether it matches these properties that we desire ok. And if it does not we will try to
improve upon it, or try to find the reasons as to why it does not ok; so, dones the exercise
that we are going to pursue ok.
Now, once again all these topics are going to involve a lot of mathematics, there is going
to be a lot of derivations around. Because as the device gets more and more complicated
your device physics is also going to get more and more involved as it going to be more
involving ok. But the point of the courses not nobody is going to test you on whether you
can remember the derivation, that is not the point of the course. The point to the course
is, you know, how to approach the device physics ok.
So, that is why we will go through this painful probably some people it is painful
excursion of me talking about the derivations and me writing down the equations. But at
the same time keeping in mind as to what the concept is, and what the physical intuition
behind all that is ok. So, it is in some sense necessary it is a part of this course, but
nobody’s going to test you on whether you actually can derive this particular model for a
particular device ok.
So, the point is to give you the toolkits and show you how it is done; so, that you will be
able to build your own device in future and perform this kind of an analysis on that
particular device ok. Now there are many kinds of transistors ok.
478
You will hear of many and I am probably I have listed a few of them here, but I am sure
there are many others. Again this is something the design and development of transistors
in the materials used for the transistors is very much the forefront of research in
semiconducted devices. But these are some of the ideas behind transistor constructs that
some of you might be familiar with, but in some sense are also very important.
So, you have bipolar junction transistors or BJTs. You have field effect transistors, and
there are many kinds of field effect transistors, you have metal oxide semiconductor field
effect transistors which are MOSFETs. And in MOSFETs you have something called as
a depletion mode MOSFET or an enhancement mode MOSFET. You have something
called as a MESFET which is a metal semiconductor field effect transistor. You have
thin film transistors, you have tunnel FETS, you have junction field effect transistors
etcetera. So, there are many kinds of transistors.
479
Let us start looking at the BJT ok. So, many of you might be familiar with this because
the it is it is a it is a widely taught topic in your undergraduate studies, but nevertheless
we will go through this entire exercise again ok. Particularly focusing on the device
physics, and as I mentioned whether you like it or not a lot of the mathematics in the
derivation. So, the BJT is got a construct like this. The geometry of it is simply it is it is
there are 2 kinds ok, let us call it as a pnp BJT and an npn BJT. And the idea behind that
is you have a construct where in you have a heavily doped P type semiconductor a
moderately doped N type semiconductor and moderately doped P type semiconductor.
So, you have a sequence of these 3 layers for a pnp BJT.
And for an npn BJT you have a heavily doped N plus region, you have a moderately
doped P type and a moderately doped N type region. So, in some sense, it might look
like these are 2 diodes right. It might look like these are 2 diodes. So, and look like this.
But there is a significant difference ok. So, this region which is in the middle ok, where
which is the N doped region here in the P doped region here has to be extremely small or
extremely narrow. I will say not small, but extremely narrow for you would have a
properly working BJT.
So, in some sense, just blindly using a circuit equivalent of 2 diodes is not probably the
right representation of a BJT. So, these 3 regions are given very specific names. The
heavily doped region is called as the emitter. The region in the middle is called as the
480
base, and the region the remaining semiconductor bit is called as the collector. So, once
again here for the npn you have the emitter, the base and the collector. And the key
construct for the BJT is at the width of the base has to be very narrow. And what do you
mean by very narrow? What we mean by very narrow is if it is a pnp BJT we expect. So,
the idea of the BJT is to get the majority carriers from the emitter to migrate through the
base and get into the collector ok.
So, in this case we are talking about getting the holes from the emitter to go through the
N doped base and get on to the collector. So, the key is for this to happen very
successfully, there should not be too much of recombination in the base ok. And
therefore, the base the width of the base has to be made narrow and it should be less than
the diffusion length of the carriers we are talking about.
So, for a pnp BJT we want holes from the emitter to get to the collector and therefore,
the width of the base should be less than the diffusion length of the holes. And if you
remember if you have forgotten what diffusion length is it is the square root of the
diffusion coefficient into the minority carrier lifetime, a square root of this product ok.
We have we have looked at it when we studied P-N junction diodes.
Similarly, for an npn BJT we would like the electron from the emitter to successfully get
through the base and approach the collector. And therefore, the width of the collector
should be less than should be less than the diffusion length of the electrons ok and why
less than the diffusion length? Because the diffusion length is a measure of the distance a
minority carrier can migrate before it is recombined. So, if you remember we just like in
an RC circuit, you have you have the time constant being represented by the product of
RC which is which is a measure of how quickly the circuit responds. Just like that you
have a diffusion length it is a measure of the distance traveled by the minority carrier
before it recombines ok. So, that is the idea of a diffusion length.
So, with this understanding of the geometry and the construct let us just go ahead.
481
So, there are some circuit symbols ok. So, we had a circuit symbol for the diode as this.
So, this is the circuit symbol for the BJT, you have for a npn BJT the key is you watch
out for this arrow. The arrow is pointing towards the migration or the current direction
and it is always pointed I mean it is always between the base and the emitter.
So, for the npn BJT you have the collector the base and the emitter and you have the
collector current, the emitter current and a base current. And similarly for the pnp BJT
you have the emitter current the collector current and the base current, and in either case
you will have this relation holding true ok. We are probably left into this a little bit early,
but it is just for me to introduce the symbol the circuit symbol for the BJT, all right.
482
But now let us let us head off an talk about something a bit more familiar; which is to
draw the band bending diagram. So, here you have a P-N junction and you have another
N-P junction. So, what is the equilibrium band bending diagram for this? Ok, so, for a P-
N junction diode if you remember; so, when we looked at a P-N junction diode, at
equilibrium the first thing that happens is the fermi level a lines ok, and then since the N
region has given electrons to the P, they will find that the conduction band edge bend in
the P bends towards the fermi level near the junction. And it bends away from the fermi
level in the on the N side. And similarly the valence band follows ok. So, there was the
band bending diagram for a P-N junction.
So, it is the same diagram that we need to draw for these 2 P-N junctions ok. So, here
you have a P plus region and here you have a moderately doped N. So, these dashed lines
indicate the depletion boundaries. So, this is the total depletion width so, that is the
depletion region. So, similarly this is the depletion region for this junction. And if you
watch carefully since this is heavily doped, the depletion width here is expected to be
much lesser than the depletion width on the N side.
And why is that? Because for a P-N junction we had this condition, because of the
continuity of the electric field at the junction if you recollect we had the condition that
xdnND=xdpNA; which meant that if your as the doping concentration increases your x d p
has to come down. And therefore, your depletion width here is smaller as compared to
483
the depletion width here because the doping concentration here is larger as compared to
the doping concentration here.
So the, I have tried to represent that as accurately as possible, all right. So, here you have
the depletion region on the emitter base junction. So, this is called as the emitter base
junction. And then similarly you have the depletion region on the collector base junction.
So, since both of them are moderately doped I have sort of represented both these
depletion widths as almost equal, and that is the total depletion width of the collector
base junction.
So, in this region you have your P-N junction your P plus N junction giving you a band
bending, and then you have your N-P junction which gives you a band bending of this
kind and therefore, this is your equilibrium state band bending diagram, with the fermi
level being aligned and shown by this blue line here.
So, that is what happens in equilibrium. And since all of you are now familiar with
solving Poisson’s equation, we will not go through that exercise again, since it is just
solving Poisson’s equation for 2 P-N 2 different P-N junctions. With the with the only
exception that though that is not really required for the electrostatics with the exception
that this N region is very narrow, and N particular this region which is called as the
width of the base is very, very narrow it has to be less than the diffusion length of the
minority carriers, all right.
484
So, the BJT operates in several modes right. So, here let me just redraw the circuit, let
me redraw this device structure. So, we considered a pnp, but all this is equally true for
an npn ok. And what? We have several different modes of active of operation. So, the
first is something called as the active mode ok, which is a more appropriately called the
forward active mode of operation.
And this mode of operation is used in amplifier circuits, which we where use the BJT as
an amplifier in a and in any analog circuit you will probably be using this mode of
operation quite a bit ok. And in this mode of operation the emitter base junction is
forward biased ok. So, you want to forward bias this diode, you want to forward bias this
diode, and you want to reverse bias the collector base junction.
I mean we will see intuitively as to why this is needed ok, what happens in this mode ok.
So, you want to forward bias the emitter base junction and reverse biased the collector
base junction. And since we have talked about the forward active we can also talk about
the reverse active ok, which is for the inverse active if you like. In which case you just
flip these 2, which is the emitter base junction is sorry, it is a typo here the emitter base
junction is a reverse biased, and the collector base junction is forward biased. Apologies
for this typo, but please do correct it ok. So, the idea is since the m if not for this x is
doping the emitter and collect this device structures completely symmetric. So, for
example, let us say we did not excessively dope them it let us say it is pnp.
The device structure completely symmetric; so, whether you forward biases and reverse
bias that or reverse biases and forward bias that it is the same ok. But the difference
comes about because of this excess doping ok. And that is why in the forward active
region, we are very careful to forward bias this diode which is the excess doped P and
the moderately doped N. And reverse bias the moderately doped P-N the N. But in
reverse active we do the opposite; which is P forward bias this diode and reverse bias
this diode ok. So, that is the only difference between active and inverse active, all right.
485
In saturation mode operation the forward bias both sides, both junctions. So, this diode is
forward biased and that diode is forward biased; which means that this voltage is going
to be higher than that, that voltage is going to be higher than this and in this case this is
going to have a larger voltage is compared to the other 2 sides. And in the cutoff mode
so, before I move on the saturation mode is the mode that is used as the on state in any
digital circuit. So, if you want to completely close a switch, we are talking about the
saturation mode of operation. And the cutoff mode is the off state in any digital circuit.
So, if you are building digital circuits, you will probably be using the saturation and
cutoff mode a lot, while maybe also using the active mode for certain special purposes.
So, the cutoff mode is where both junctions are reverse biased and do not permit any
current ok. If you just reverse bias both junctions, you do not permit any current at all
through the circuit, because these are the different modes of operation. And in this
particular lecture, I will we will derive the general current voltage characteristics for the
BJT from which you can actually then go in and specifically apply these conditions of
making V eb large or V cb negative etcetera, etcetera, and derive your current voltage
characteristics for the active saturation or cutoff or reverse active as an when you desire.
But I will pay special attention to the active mode of operation. Because this is
something that we might use in the future towards the end of this course, all right. So,
these are the different modes of operation of the BJT.
486
So, particularly let us look at the active mode ok. So, I am going to pay some attention to
the active mode of operation.
So, what is the active mode mean? The active mode means that the emitter base junction
is forward biased and the collector base junction is reverse biased; which means that this
P is at a higher potential is compared to N, and this P is at a lower potential compared to
the N region. So, since this diode is now forward biased, the depletion width will reduce
a little and the band bending will become a little bit more shallower ok.
So, which is sort of represented here. So, the dark line shows the active mode of
operation, while the dashed line is the equilibrium condition. So, you can see that, you
know, this built in potential is reduced as compared to that. So, that was phi b i and this
become ∅𝐵 − 𝑉𝐸𝐵 .
And on the other hand since this diode is now reverse biased this built in potential is
increased.
487
The equilibrium let us see that was phi b i it is now increased to ∅𝐵𝑖 − 𝑉𝐶𝐵 phi b i minus
V cb where V cb< 0. So, that is that is the change in the band bending diagram. So, based
on your understanding of P-N junction theory, what do you expect would happen here?
Since this diode is now forward biased, you will have a huge diffusion of carriers from
this P to this N.
And these carriers would now start in your, in your P-N junction diode theory, we saw
that there was a depletion region ok. And you had you had your P you had your N and
488
when the junction was forward biased, you had a huge migration of the holes from the P
to the N a migration of the electrons from the N to the P. And the holes became the
minority carriers on the N side and you had a concentration gradient which went like
this, which we saw from solving the, which we found to be exponentially dependent
from on distance from the solution of the continuity equation. So, that was the P-N
junction theory.
It is the same thing here, but the only difference is, on this N side all these carriers are
not recombined, ok. Because the diffusion length is now the base width is now less than
the diffusion length. So, therefore, you have a lot of these my carriers coming in, but
now all of them are not recombined your base width is very, very small, you recombine
just a few of them. And most of these carriers now get into this depletion region. And
you will see a massive flux of these carriers so, the other side ok. So, let us to look at that
in more detail.
So, in active mode we are talking about the active mode of operation, this P-N junction is
forward biased. And this P-N junction is reverse biased ok which means that there is a
huge electric field here ok. This electric field is going encourage all the holes if the hole
gets in here is going to very rapidly move up and if there is an electron here it is going to
very rapidly move down help.
489
And because this junction is forward biased you have a large hole diffusion from the P
plus to the N side. And you also have a significant electron diffusion from the N to the P,
but the hole diffusion is much larger than this electron diffusion simply because this P
junction this P side is more heavily doped than the N. So, you do have a large electron
migration or the diffusion of electrons from the N to the P, and from the emitter to the
base because of this forward bias, and just at the whole movement from the emitter to the
base is going to be much larger it is going to dominate the current total current. Now all
these electrons are going to enter the base and in the base. So, what is the base made of
the base is N doped?
So, you now let us let us focus on the base.
490
So, here you have the base region, and let us say that those are the depletion boundaries.
So, you have all these holes that are now diffused into the base, you have a large number
of these holes now this base is made of a lot of it is got a excess number of electrons are
the majority carriers in the base and the holes are the minority carriers. So, this excess
whole population is going to start recombining with the electrons in the base. But and
they are also going to diffuse from the higher concentration to the lower concentration
end. So, they are diffusing, but they are also recombining.
Now, if the base width was very large, then they would all recombine and very little will
get on to the other side. But the base has been specially designed to have a length less
than L p and therefore, all these holes will now make it to a across the base and get into
this depletion region at the base collector junction. So, all the holes will successfully
make it through perhaps not all, but most of them would successfully make it through,
and get into this depletion region.
And once they are in this depletion region they see a very large field because this
junction is now reverse biased. So, they see a very large field and all these holes will run
into the collector. So, we have successfully transferred a large number of holes from the
emitter through the base and to the collector. And that is the idea of active mode
operation, and that is why you forward bias one junction and reverse bias the other.
491
So, just to summarize, you know, I have just written down these points hole, in active
mode of operation holes diffuse from the emitter to the base because the forward base
bias diode, in the base there is some recombination, but not much since the base width is
going to be much less than L p. Most holes get across to the base collector junction
where the electric field drifts the large hole concentration to the collector.
So, all the holes here are now drift to the collector and essentially the collector collects a
large a whole population from the emitter. But to maintain all this operation in steady
state, you know, if you want to keep this phenomena going continuously, you need to
compensate for all the electrons lost in the base.
Whereas, the base lost electrons, a lot of electrons have migrated from the base to the P
side because of the diffusion. And a lot of the electrons have recombined.
Now, if you do not replace this electron, this mechanism would stop, you know, it will
reach a new equilibrium and everything would stop. So, in order to keep this going you
need to provide a steady base current, you need to keep injecting electrons into the base.
If you think of the current in the conventional sense, you have a base current that flows
outward.
So, that is why in your so, we are discussing a pnp junction, and that is why if you look
at the circuit diagram, you will say see that for a pnp the base current is drawn this way.
492
So, here you have the holes moving from the you have hole migration, you have holes
moving here, and you have electrons moving in to compensate for the electrons that are
recombined that have diffused to the emitter and therefore, the base current moves here.
So, you have a emitter current, you have a collector current, and you have a base current,
with this relation being followed. So, that is the essence or the qualitative discussion of
the working of the BJT. So, we have looked at the pnp BJT, the npn BJT is the same
argument. So, he want to draw this properly so, let us take the npn junction. So, you have
the N plus P N n the band bending is now going to look like that, ok. So, that is the
heavily reverse biased P-N junction and this is the forward bias P-N junction. And if you
are going to have a large number of electrons that are going to head to the base, some of
these electrons will recombine, and a large number of these electrons will now start
drifting away into the collector and will be collected by the collector.
So, in order to recompensate for these holes in the base to keep things moving for an npn
junction, you need to supply a you need to add more holes; which is being which is in
which implies that you need to add a base current that is directed this way ok; so, that is
the npn operation. So, we discussed both the pnp as well as the npn operation, in active
mode.
So, we are only looking at active mode at equilibrium we know what the band bending
diagram is I am being special attention to active mode operation although we will derive
the general current voltage characteristics. So, it is this mode of operation that is used in
analog circuits. And the idea is that a small fluctuation in your base voltage will result in
a small fluctuation in the base current; which would in turn result in a large amplification
of this current because it the collector current is going to it is going to be quite large. So,
that is the idea that is the idea of the amplification. And we will get to this when we talk
about circuits, we will come to this later on.
493
Lecture - 30
BJT : IV Characteristics
So, now let us get into the Derivation of the Current Voltage Characteristics for a BJT
and the starting point for that for this current voltage characteristics is going to be
something that you are already familiar with, which is in the case of your pn junctions
what was our strategy to derive the current voltage characteristic. So, we had your P side
and you had your N side and you had a depletion region ok, you had a depletion region,
what was the strategy for the current voltage characteristics. We saw that when we
forward biased this diode or when we apply any voltage Va whether its forward or
reverse bias. We had holes diffusing through ok. So, let us say forward bias we had holes
diffusing through and you had this excess minority carriers coming in to the N side right.
So, we had an equilibrium population of holes on the N side, which is the number of
holes on the N side and that stands for the equilibrium value that was (Refer Time:
01:26) notation and by applying a forward bias this became pn ok, which was greater
than the equilibrium population and the excess injected carriers was delta pn which was
basically your pn minus Pno. So, that is equilibrium that is non equilibrium and this is
494
the excess due to the non equilibrium and we said that we will first identify delta pn as a
function of x.
So, we said that on the N side we are going to have this coordinate system which was x
dash which had a starting point at x dash equal to 0 which corresponding to x equal to
xdn that was x=xdn and we have solved the continuity equation and we first found that
this delta pn varied exponentially with distance ok, to sort of proportional to e to the
power minus x dash by lp ok.
And then after identifying this gray gradient we then said we will now derive the current
voltage characteristics by using that because the current is now diffusion current we said
that the current was due to the p was q minus q Dp d delta P on the N side by dx and that
was the current density. We are going to do the same thing here the only problem is that
we have two junctions and we have to solve for both the junctions so; we need to have
three sets of coordinate systems ok. So, we had if you remember the pn junction for the
N side we had this coordinate system x dash and for the P side we had another coordinate
system which pointed in that direction and it made things easy and it is we are going to
do the same thing here ok.
Now, we are going to watch the minority carriers in each segment and the emitter. What
are the minority carriers? Minority carriers are the electrons in the emitter which is p
doped ok. So, the symbols here let me just spend a minute addressing the symbols when
we say npe it implies that we are talking about electrons in the P type in a P doped
semiconductor which means it is the minority carriers and this e stands for the emitter
and when we say if we say Ppe we are talking about the holes in the P types doped
emitter ok, which is the majority carriers.
And when we say np0 it is the equilibrium concentration of this whereas, the npe is what
happens when you apply a bias which is the non equilibrium concentration. So, npoe is
basically the equilibrium concentration it is npo in the emitter. So, thatis that is the
symbol here and if this e is replaced by C we are talking about the collector and if its
replace by b its talk we are talking about the base ok. So, that is the notation used it is
quite intuitive if there are any difficulties do feel free to send me a message.
So, what we are plotting here is the minority carrier concentration which is the electron
so, the Emitter is P times the minority carriers electrons. So, electrons in the p type
495
emitter and these are the electrons that have been injected from the base into the emitter
because of the forward bias action of the diode and they have got a variation that looks
like this ok, and in the base the minority carriers are the holes, that have been injected
from the emitter into the base due to the forward bias action and they have got a
concentration that looks like that looks like this because of the recombination also little
recombination in the base and that is P N in the base ok. And in the collector the
minority carriers are again the electrons and the collectors reverse biased and therefore,
you have an electron concentration that varies like this.
So, you have N in the p type collector and you have a electron concentration that varies
like this remember the base collector junction is a reverse biased. So, that is why you see
the concentration variation in the manner shown and this junction is forward biased ok.
So, that is why the concentrations are as shown because if you recollect we will again be
making use of this idea the concentration at this interface, which we use as a boundary
condition is basically the equilibrium concentration into e to the power q times the
voltage applied across this junction ok, in this case is the base emitter voltage I am sorry
it is the emitter base voltage because that is the pn by kt minus 1 and by just
understanding the nature whether this is positive and negative you can identify what the
concentration profile is ok. You can understand the point the starting point of the
concentration profile all right. So, that is the profile.
496
Now, all we have to do is sit down and solve the continuity equation in the collector the
base and the emitter and you will have your current voltage characteristics ok. Now
before we do that one thing which I have not marked out here is the coordinate system.
So, you can see three different coordinate systems. So, this coordinate system I will just
call it as x which means x equal to 0 is at the is defined at this depletion boundary ok, the
base the depletion boundary due to the base emitter junction and the depletion boundary
which is sitting in the base very specific ok.
So, that is my x equal to 0 point and if I am correct the this coordinate system is defined
as x dash I would like to call x dash equal to 0 as this point here and this coordinate
system is x double dash with x double dash pointing in the opposite direction and with x
double dash equal to 0 define at this interface and the reason we do it is simply makes
the mathematics a lot more easier because you can independently treat these three
differential equations without worrying about any translation of the variable etcetera ok,
just makes it just simplifies the mathematics a lot more and we can use the well
established solutions again and again ok.
Because we are going to treat these three independently and just bring in the effective
idea or the impact of these three solutions at the last step. So, we have these three
coordinate systems and we have concentration profiles of the minority carriers and we
are going to identify. So, here I have just qualitatively sketch the profile, but you want
the exact mathematical variation of the profile you want to be solving the continuity
equation for these three regions ok. Now the continuity equation for these three regions
are written using several assumptions the first is low level injection. So, that I can very
conveniently use my recombination rate, which is basically my delta npE or whatever
delta np by tau N, so, I can use the recombination rate of that kind.
Then there is no generation; that means, we are not thrown light on the BJT ok, there is
no new generation of no photo generation. So, let me just say no photo generation of
carriers, let me very specific and we are talking only about steady states. So, we are not
talking about transients so, the continuity equation will not have a time component here
ok. So, we are in steady state low level injection condition and here in the continuity
equation if you and I want to understand what this means it means this G is 0 the G is 0
this means I can use my very convenient delta ∆𝑁/𝜏𝑁 N by tau N or ∆𝑃/𝜏𝑝
497
recombination rate and what this means is that any derivatives with time is 0 ok. So,
these are the three conditions and we will throw it into the continuity equation.
And of course, electric field component are not important. So, that is another aspect so,
the continuity equation is something that you are already familiar with ok, we have used
this for the pn junction diodes. So, I am not going to elaborate a lot on this I made one
leap right away which is I have represented this term here as the diffusion length of the
electron. So, where did this Ln2 come from the actual continuity equation is it is going to
be Dn d square delta N p in the emitter by dx double dash square is equal to delta npE by
tau N and therefore, this becomes d2∆npE by d double x square is equal to d delta N delta
npE by tau N D N which is nothing, but your Ln in the emitter squared ok. So, that is that
is the leap we have taken here.
And similarly for the collector so, that is for the emitter right. So, that is the continuity
equation in the emitter and in the collector we have another continuity equation because
they are the minority carriers are n. So, d square delta N pc by dx dash square because
that is a new coordinate system in the collector is equal to delta N pc as a function of x
dash by Lnc2 ok. So, this is the emitter and collector and N stands for the diffusion length
of the electrons.
So, these are things that are already familiar with, but for the base there definitely is
recombination, but for now we will just assume that there is absolutely no recombination
the BJT is working in a manner that we dream of, which is the base width is so narrow
that there is no recombination that is that never really happens ok, but it is a big
assumption, but nevertheless an assumption that will simplify lot of things for us and
which will give us a very intuitive glance as to the working of the BJT. So, let us at this
point assume that there is no recombination in the base, which means that this term see
this term comes appears because of the recombination so, that is a recombination rate
right.
So, what this is saying is that there is no recombination, which mean that term is 0. So,
this is the differential equation in the base. Now this is very easily solved right so, the
solution to that is simply it is got a linear dependence it asked for a linear dependence in
of the minority carriers with respect to x and the way you find out these two constants is
that you identify what is the minority carrier concentrate, which is the holes in the base at
498
x equal to 0 and at x=w. So, what is a x equal to 0 and x equal to w we are asking for
what is the concentration here and what is the concentration here and if you know these
two points you identify these two coefficients and therefore, this is the general solution
for delta P n in the base ok.
Now, what are the boundary conditions? We have listed three different differential
equations and I am sure all of you know how to solve these it is quite obvious as to what
the solutions are this will have a form of whatever your x double dash by LnE + Be to the
power minus x double dash by LnE and similarly for this one with different a and b ok.
We already we have already seen these solutions in the in the case of a pn junction we
will do that again, but what are the boundary conditions? The boundary conditions are
exactly what we used for a pn junction. So, if you have understood pn junction theory
well BJT is quite straight forward because we are just using the same ideas again and
again.
So, what are the boundary conditions? So, what we are looking at is np in the emitter at x
double dash equal to 0 so, that is my x double dash that is x and that is x dash at this
point. So, we are looking at this and that value is given by boundary condition 1 ok, and
we have already seen we have already applied this boundary condition at the in the pn
junction diode npoE is the equilibrium concentration it e to the power (qVeb/kT) -1. Veb is
the if you think of the base emitter junction that is the base emitter junction we are
499
looking at a pnp BJT. So, that is my base emitter junction and this is the applied voltage
ok. So, Va is the same as your Veb in this case, but on the other hand if it was a npn
junction if that was the npn base emitter junction then that would be the way we had
defined the applied voltage in the pn junction theory and therefore, my Va would be
equal to Vb ok. So, since we are talking about pnp junctions you have Veb.
So, delta npE at x double dash equal to 0. So, that is x*” = 0 is given by this term ok, the
same boundary condition we used in the pn junction diode delta npC at x dash equal to 0
that is basically this population here is equal to this that is your second boundary
condition ok, that is my x dash equal to 0 point delta Pnb at x equal to 0 that is basically
this is given by this delta ∆Pnb at x=w ok.
So, that is that is my let us say x =w and what is the population here that is given by the
collector base junction because that that boundary is now at the collector base junction
interface it is at this interface and therefore, that is my x equal to w and since that
junction is reverse biased you see that you know this value takes this particular nature.
Finally, what are the values very far off at x”=∞ ok. So, that is x double dash equal to
infinite asymptotically and that is x dash equal to infinite. So, what are the values there
in the case of ∆npe x double dash equal to infinite is 0 and delta ∆npc at x double dash
equal to 0. So, what this means is that if you look at nPE minus the equilibrium
concentration nPOE there is not much difference and; that means, nPE approaches the
equilibrium concentration as we head towards infinite and npC also approaches the
equilibrium concentration as we head towards infinite.
So, the in the limit of x double dash heading to infinite and in the limit of x x dash
heading to infinite nPE-nPOE is 0 and npC- npoC is 0. So, that is what these two boundary
conditions mean. So, you apply so, you have three equations you have three second you
have three equations and therefore, you have six unknowns ok, you will have A B let us
say let us call it as A of the emitter, B for the emitter then A for the collector, B for the
collector and you have your C 1 and C 2 as I have shown you here you have six
unknowns and therefore, you have six boundary conditions with which you can identify
every unknown and therefore, you have the exact solution for the nature of the minority
carrier variation with distance ok.
500
So, what we are going to do now is simply solve the differential equations apply the
boundary conditions get the variations of delta np as a function of x etcetera and then
find the current the diffusion current ok, just like we did in the pn junction.
So, first let us do the emitter side ok, it is all the same I mean the methodology is the
same. So, that is going to be the general solution I have used the same C 1 and C 2 please
do not mistake this for the same coefficients as used in the base in the case of the base.
So, we have used the same variable C 1 and C 2, but there is just a matter of habit. So, if
you if you would like you can call it C 1 e and C 2 e, but this C 1 and C 2 is different
from the coefficients we used for the base ok. So, please note that ok. It is different from
what was used in the base ok. It is just a matter of habit you to which I have used C 1 and
C 2 again.
So, you have delta nPE is going to have this particular form ok, you have an exponential
of a minus x double dash by LnE and plus x double dash by LnE. So, we use all the
boundary conditions you will find that at x double dash tending to infinite, what happens
as x double dash tends to infinite this carrier concentration goes to 0 where it as x double
dash tends to infinity this term blows up therefore, we cannot allow C 2. So, C 2 has to
be 0 and therefore, this is the only solution and we identify C 1 by using the condition of
delta nPE at x double dash equal to 0 ok, which was this boundary condition here and
that you can identify your concentration gradient the excess electron concentration in the
501
p doped emitter to vary as this with the x double dash coordinate system pointing in that
direction it goes from the base to the emitter and all the way till infinity.
So, that is my concentration gradient of N in the p type emitter. So, what is the current?
The current is the diffusion current so, here we have written the total current and now the
current density therefore, we bring in the area of cross section. So, you have the q Dn E
which is the diffusion coefficient of the electrons in the emitter into d delta nPE by dx
double dash and multiplied by the area of cross section to give you the total current and
just like in the case of pn junction diodes the way we want to extract the current is by
looking at the current at the depletion boundaries we want to measure the diffusion
current here you see the diffusion current is going to vary with space, but if I were to
measure the diffusion current here I must also account for the other current mechanisms
right.
So, I would rather prefer to get the diffusion current right at this interface and therefore,
we calculate the diffusion current at x double dash equal to 0 and we find that that is the
diffusion current of electrons in the emitter. What about the holes from the emitter? The
holes from the emitter are moving to the base so, you have what we have now calculated
is the electrons pushed by the base electrons going in from the base to the emitter, what
about the larger contribution of the holes coming in from the base to from the emitter to
the base. So, for that we need to solve the differential equation of the minority carriers in
the base which is my IPE at the emitter the whole contribution of the emitter current this
is the electron contribution of the emitter current ok.
The whole contribution of the emitter current located at x equal to 0. So, we are going to
measure that at x equal to 0 we are going to do that calculation here. So, we have already
calculated for this is done you are now going to calculate for this. So, that is nothing, but
qAD p B into d delta n p in the N type base by dx and we have already found that since
there is no recombination in the base we found that our ∆npB had this particular form. So,
if you remember if you recollect when we solve the differential equation for the base we
said there is no recombination and therefore, that was the solution. So we are going to
use that solution now ok, there is a by W there which probably is not visible on the
screen ok.
502
So, we are going to use that solution now and we find that the current due to the holes
from in from the emitter is given by this particular expression ok. So, that is the current
that is the electrons moving in from the that is the electrons from the base to the emitter
you have a forward bias base emitter junction, that is the electrons from the base to the
emitter this is the holes from the emitter to the base and the total current is therefore, the
sum of these two ok.
So, we are now talking about the emitter base collector we are talking about this total
current. What is IE? It is got two components it is got the electron current from the base
to the emitter and it is got the hole current from the emitter to the base and the sum of
these two is the total IE, which is given by the sum of IPE plus INE it is basically
summing these two expression is that we derived. So, this is the total current from the
emitter so, that is the emitter current.
So, we have not. So, remember this is being done in any mode right we have not said that
Veb is forward biased was greater than 0 or Vcb is less than 0 this is true for saturation
cut off active it does not matter we will later on adapt the values for this and see which
terms dominate. So, this is a very general derivation.
503
So, now let us do it on the collector side. So, we know what IE is what is the collector
current? The collector current ok, now how do we go about the collector current. So,
collectors the collector base junction is a reverse biased right. So, this is the argument we
make.
We say that there are no loss of carriers in the base ok. So, we have the emitter base
collector we know what IpE is ok. So, we say that there is this IpE came into the base and
as per our assumption there was no recombination in the base therefore, there is no loss
of Iep therefore, I am sorry IpE is that I do not know whether what I mean is the whole
current from the emitter so, IpE ok. So, that current is the entire entering the collector
without any loss of current because the base is not doing any recombination. So, there is
no loss of carriers and therefore, the all of IpE is going to go into the collector.
So, IpC that is the holes entering the collector is simply equal to IpE at x equal to 0, which
is equal to IpE is at x equal to w ok. So, what this means is just you reiterate so, you had
your x equal to 0 boundary that is my x equal to 0 that is my x equal to w that is the base
width ok, this is the base and you have your depletion boundaries here this junction is
forward biased that is the emitter base junction is forward biased. So, you have a lot of
holes that came in and we calculated the hole current here at x equal to 0, what we are
saying is that we had assumed that the base did not recombine any of these holes
504
therefore, the whole current here is equal to the whole current here is equal to the whole
current here ok.
So, IpC is nothing but IpE at x equal to 0 which is equal to IpE at x equal to w if there was
recombination then this greater this is not equal to this and you cannot make this
argument ok, because these holes would not have entered the base collector junction.
Now to identify the electron contribution in the collector what do we do we solve the
continuity equation at the collector and we go through the same exercise which is which
is to try and identify the solution for the differential equation.
So, all this differential equation for the collector and apply these boundary conditions
this one here and this one here you apply those two boundary conditions and you will go
through the same process like what we did for the case of the emitter and we will find
that the current in the electron current in the collector is given by that the hole current in
the collector is given by this term that we have already calculated and therefore.
The total current in the collector is InE+IpC which is given by this term. So, that is my
collector current. So, IC that is my collector current. So, you have now calculated IeE and
I C.
505
So, finally the base current. What is the base current? We will simply use Kirchhoff’s
law we will say that there was a current through emit from the emitter there is some
current with the collector remain current remaining current is the base current. So, if in
case of the pnp there was IE there is some IB and there is some IC ok. So, in other words
IE is equal to your IB+IC. So, therefore, IB is simply the difference of the emitter and the
collector currents and that is simply given by this term here. So, that is the general
derivation of the emitter collector and base currents using the continuity equation and
looking at the minority carrier distribution profile in the three in the emitter base and
collector and then solving for the diffusion current in the three region. So, this is the
general current voltage characteristics.
506
So, just to put them all together I have written down the emitter current, the collector
current in the base current and I have taken I have gone one step further, which is since
we know the this is the equilibrium concentration. So, it is got these a nPoE is the
equilibrium concentration of electrons in the p typed emitter right. So, we have my we
have the pnp emitter base collector npoE is the number of electrons at equilibrium which
is ni square by the doping etcetera like doping concentration of the emitter. So, this is
quite large so, therefore, npoE is expected to be quite small.
Similarly, PnoB is ni square by NDB, PnoC=ni2 square by NAC and therefore, we just
substitute that and we get expressions of this kind here ok. So, it is just it is just
substituting that and sort of summarizing the collector emitter and base currents.
507
Now, we have some other notations that many of you would be familiar with ok. So, we
can define several terms there is the beta factor for a BJT. So, when you have a BJT
particularly if your interest in building circuits or forming circuit analysis there is a beta
factor, which is basically your IC by IB and then there is an alpha factor which is IC by IE.
Now, what is IC? IC it is got the whole contribution and the electron contribution IE is got
the whole contribution and electron contribution, there is something called as a base
transport factor which only looks at the hole concentrations ok, what is the what is the
whole population that the collector collected divided by the total whole population the
emitter emitted ok, it does not take into account the entire IC in IE, but it is a base
transport factor you know how efficient was the base and transporting the holes in the
case of a pnp junction in the case of an npn we will be interested in the electron
population and here you have gamma which is the emitter injection efficiency you know
how efficient is the emitter in injecting holes in the case of a pnp.
So, all this is for the pnp junction that we analyzed so, the emitter. So, what is the so,
you want to push a lot of holes and you know not have too many electrons come in from
the base and all those depends upon the doping of the emitter.
So, you have all these terminologies most of you might be familiar with beta and alpha,
but since you have already derive the general IV characteristics these are now you know
just terms that you could use for doing circuit design, if you prefer ok. So, you now
508
know what is what is hidden behind these terms and since IC IE and IB are also related
by this relation beta and alpha are also related. So, generally if you want a good active
mode operation you would like the BJT to have small IB large IC and therefore, very
large value for beta and a close to one value for alpha, which is all the electrons from the
emit all the holes from the emitter where collected by the collector with very little
recombination in the base. So, you that is what you would like.
509
Lecture - 31
BJT: Non-Idealities and Equivalent Circuit Modeling
So, far we have seen the general IV characteristics for the BJT. So, we have not made
any specific assumptions on whether it is active mode or saturation mode or cutoff mode,
but we managed to derive the very general expressions for the current-voltage
characteristic of the BJT.
510
And we did this by simply solving the continuity equation and identifying the diffusion
currents. So, now, we can take these very general equations and you know make it more
specific. So, let us say we are interested in defining the IV characteristics in active mode
operation. So, how do we go about doing that?
So, what we do is if you just make use of the different conditions right. So, you have an
active mode of operation you have the emitter base junction as is forward biased and
collector base junction is reverse biased which means that we can say that V eb is greater
511
than 0 V cb is going to be less than 0, because it is a pnp device that we have been
looking at. So, you have your p, you have your n and you have your p that is your emitter
base or p plus emitter base and collector. If my emitter base junction has to be forward
biased, then my emitter base voltage should be greater than 0; if my collector base
junction has to be reverse biased, then my collector to base voltage has to be less than 0.
So, if that is my positive, negative, that is a positive and negative.
So, the moment you make these assumptions watch out for what is happening in to V eb
and V cb they all appear in terms of exponentials ok. So, the next obvious step is to say
that since this is greater than 0 the exponential of qVeb /kT is much greater than 1. And in
the case of Vcb less than 0 you can say qVcb/kT my k T is much less than 1 and so on so
forth so that is that is the idea. So, we are going to we are going to make use of these
different conditions and identify whether these emitter base voltages and collector base
voltages are greater than 0 or less than 0 and then start making start throwing in these
assumptions into your equations. And you will get a good flavor for what happens in
each mode of operation.
So, on the other hand for saturation mode you have both of them to be greater than 0.
And therefore, both the exponents are going to be much greater than 1. And in the case
of cut off the exponents are going to be less than 1 and so on and so forth.
512
So, for active mode is something we are particularly interested in because we will later
use this for doing circuits. So, you say that V eb is greater than 0 Vcb is less than 0
because it is reverse biased. And therefore, we have these two assumptions and therefore,
the emitter current becomes this. So, by the time the emitter current become this, if you
go back to the general equations what we have done is since exponential of qVeb/kT is
much greater than 1, we have ignored the 1. And since exponential of qVcb /kT is much
less than one we have ignored the exponential term ok, so which means that is the
emitter current. And the collector current will have this dependence all right; so which
means so these are all approximations, which means that in active mode of operation the
emitter current and collector current are more strongly controlled by the base to emitter
voltage on the emitter base voltage ok.
In saturation mode ok, so what is the saturation mode of operation? You say that all these
exponentials are going to be much greater than 1 and therefore you have all the
exponential terms being present here ok.
513
And similarly for cutoff we just make use of the fact that all the exponentials are going to
be much less than 1 and therefore that is your emitter current that is the collect base
current and that is the collector current ok, so that is there is a negative sign there of the
emitter so that summarizes you know the three different equations and all this can be
plotted off into a nice chart ok, where which gives you this IV characteristics of your
BJT. You can take this as a homework and try to build this chart for yourself.
514
So, you find that you have the saturation mode of operation here because this is the plot
of IC versus VEC that is the emitter to collector voltage. So, you have taken into account
both the base to emitter voltage as well as the emitter to collector voltage. So, we are
plotting the emitter to collector voltage versus the collector current ok. And this is being
plotted for different base current values. So, you see that this curve is for base current of
0. So, here my base current is 0. And everything within this region is the cutoff state ok,
there is a no base current flowing.
And as the base current increases, the collector current increases. And for large VEC
essentially what we are saying is that the emitter base voltage becomes positive; and the
base collector voltage becomes positive which means that the junction is now reverse
biased and therefore, you start having a characteristics of this kind. Now, the inverse
active operates like the active, but it is going to be much weaker the currents are going to
be much smaller because we do not have a p+n junction anymore instead we have a pn
junction. So, this is a summary of all the current-voltage characteristics. And as a
homework you could take this ahead and go and plot it for yourselves and convince
yourselves as to what the current-voltage characteristics are.
Now, so far we have been assuming a very ideal BJT ok. And there are several factors
that we sort of ignored or we did not really consider when we derive the current-voltage
characteristics. And some of these factors are listed here. The very first and the very
515
obvious one is recombination in the base. We said that there is no recombination in the
base and we very conveniently wrote our what we say our equation for the holes in the n
type base we conveniently said that was equal to 0, whereas that is not really true
because you do have recombination in the base so that is the first non ideality.
The second is something called as a base width modulation and we will explain what that
is this is also called as the early effect ok. The third is again got to do the width of the
base and something called as a punch through in a BJT I will talk about that. And the
fourth is the avalanche breakdown which we discussed in the case of pn junction diodes,
but in this particular case in the case of the BJT, it is the reverse heavily reverse bias
collector base junction that could cause that could experience an avalanche breakdown.
And then in the case of the pn junction diode we looked at non idealities like the
recombination and generation mechanisms in the depletion region that is valid in the case
of BJT as well and we have ignored that. And finally, in the case of the pn junction when
we built the equivalent circuit you know we said that there is a depletion capacitance,
there is a diode resistance and then there is the bulk resistance in forward bias operation
which is due to the resistance of the doped regions and the contact and that is another
factor that we have ignored in the case of a BJT.
So, first let us look at the recombination in the base. So, we had assumed that there is no
recombination and therefore, we had gone ahead and used this equation, but in reality
516
this equation is not completely correct. There is a small amount of recombination and
therefore that the correct differential equation should involve a component like this ok.
This is similar to the differential equation we used for the collector and the emitter side.
And therefore, we already know the solution it will have two exponents, but the key here
is in the case of the collector emitter we could get rid of one of the coefficients because
of the condition that as x” went to infinite or when x dash went to infinite you know the
excess carriers became 0. But in this case that is not true because you are talking about
the base and therefore, we need to have two boundary conditions of p and b at x equal to
0 and delta ∆p and b at x equal to w.
So, we need to have these two boundary conditions, we need to solve these two to get an
exact solution. And what is delta ∆pn be at x equal to 0, it is going to be this it is going to
depend on the emitter to base voltage. So, in case that is not that writing is not clear it is
the emitter to base voltage. And in the case of ∆ p and b at x equal to w, it is the same
term, but it is going to have the collector to base voltage which is a reverse biased
junction ok. So, in this case the collector to base voltage is going to be negative which
means it is going to be very small and therefore you are going to have a term that
actually goes below this particular value.
The other non ideality is something called as a base width modulation this quite
important from the point of view of circuits and it is got another term it is called as the
517
early effect it is called as the early effect named after the person who discovered it. And
this particular effect has got a counterpart in MOSFETs. So, when we talk about
MOSFETs, we will discuss something called as the channel length modulation ok. Now,
both these effects are talking about a phenomena wherein the current the current through
the device, in this case it is the collector current is dependent on the collector to emitter
voltage or in the case of the MOSFET, it will be dependent on the drain to source
voltage. We will come to that a little later.
But the effect of these phenomena on circuits is that it leads to the reduction of the output
impedance of your device and that impacts your amplifier design and the amplifier gain
it is it ok. So, just keep this in mind it is quite important. And the reason the idea behind
the base width modulation is actually quite simple. So, what is the width of the base? So,
you have let us say here we have taken an npn device ok. And the reason I have taken an
npn device is because this plot corresponds to an npn device and I have taken I borrowed
this plot from an online website.
So, in the case of a in the case of a BJT, you have your emitter base and collector. And
you have these depletion regions right, you have a depletion width there, so that is the
depletion region because of the base emitter. And then you have a depletion region
because of the base and collector these regions are all depleted. The base width is this
region here that is your x equal to 0 at x equal to w, so that is the base width.
518
Now, the base width has to be less than the depletion diffusion length of the minority
carriers. So, in this case, it has been less than Ln. Now, the base width clearly depends
upon the width of these depletion regions. If these depletion regions become wider, if
they become wider, the base width has to decrease; and if they are narrower the base
width increases. So, therefore, the width of the base is dependent on the depletion widths
which in turn are depend on the bias voltages.
So, if I make my region here more forward biased, then this would decrease; and if I
make this region more reverse bias then that would increase. And therefore, essentially
we are controlling or modulating the width of the base. So, since it is dependent on
effectively these two depletion widths, the width of the base is quite dependent on the
collector to emitter voltage ok. Now, ideally in an active mode operation we saw that the
collector was not doing anything, it was ideally you should had a collector current that
really did not depend too much of the collector voltage.
But because of this width modulation, you will find that the current does depend on the
collector voltage ok. It does have a dependent the slope is not 0, it has got some finite
value. And therefore, this is the true nature of the IV characteristics. And by taking a
tangent to these lines and projecting them backward, you will end up with all of these
519
lines ideally intersecting at a point. And this voltage this collector to emitter voltage is
called as the early voltage.
And why does it modulate the collector current, because if we go back and look at the
expression for the collector current. So, let us say active mode operation, so that is the
collector current, where is the width you find that the width is there in the denominator
right here ok. And therefore if the width changes it is going to quite significantly impact
the collector current, so that is the base width modulation or the early effect. And we will
I will refer to this once again when we discuss the channel length modulation effect in
MOSFETs. So, they are quite important from the circuit’s point of view.
The third non ideality is something called as punch through ok. So, it is not so forgive
me for the poor handwriting, but this is something called as punch through. And what
this implies is that it is things of it discuss the phenomena when the width of the base
starts tending to 0 ok. So, let us say ideally you have let us say your pnp BJT. So, you
have a BJT that looks like that ok, so that is the base width that is the width of the base.
But now as the width of the base starts heading towards 0 probably due to the way we
have doped it or due to the geometrical construct or due to the biasing circuit etcetera for
whatever reason let us say that starts heading to 0. You will end up with structure like
this, so that becomes your band bending diagram ok.
520
And what happens now is if I further reduce this you know as we start if I as I start
approaching W toward 0, we cannot handle these two junctions separately. So, far we
handle the base collector junction, base emitter junction separately, we solved for all the
minority carrier concentrations there. We handle the base collector junction separately
and solved for everything there and we simply summed up all the currents.
So, we cannot do that anymore, because now the emitter is directly interacting with the
collector because the base width does not exist anymore ok. So, this is the phenomena
where you will start having a significant transport of carriers across the base because the
width is approaching 0 and this effect is something called as the punch through
phenomena. So, but from the point of view of the analysis you cannot these two are
electro statically coupled.
And finally, the last three points which is the avalanche breakdown in the reverse bias pn
junction at the collector base junction, the recombination and generation of the depletion
and the bulk contact resistance were already discussed in the case of pn junctions and
they are the same ideas here. So, what this means is that if you look at the collector base
junction, you have the base, you have the collector. So, in the case of pnp, you have a
heavy reverse bias, you have a large field. If I keep reverse biasing this stronger and
stronger the field could get so large that the electrons migrating through this junction
could have enough energy to create an avalanche of electrons that is pull electrons or rip
521
the other silicon atoms apart and pull the electrons out of them thereby creating
avalanche of current ok, so that is that is the breakdown phenomena there and that is
because the reverse biased collector base junction that you might see in the case of active
mode operation.
Then recombination and generation and the depletion region; so, once again just in the
case of pn junction, since you have a depletion region, so if the depletion if the devices
and reverse bias your np is less than ni square and therefore, this encourages generation.
On the other hand, if it is forward biased, then it np would be it would encourage
recombination mechanisms. And finally, you have the bulk and contact resistance which
you have already discussed which is due to the doped regions and due to the metal
semiconductor contacts. But on that note since we will head back to circuits, it is useful
to talk about an equivalent circuit model or an equivalent circuit representation for a
BJT.
And we have a simple model for that which is something called as the hybrid pi model.
So, if you were to open a textbook and look at an equivalent circuit for a BJT this is what
you would see. And you know as I mentioned there now you know the course is good to
start having a circuits flavor to it ok. And I will just discuss very briefly as to what the
522
different aspects of these models are ok. So, this is the ideal model and what we have
here is a non ideal ok.
So, let us focus on the ideal model first. So, what you see here is the BJT has got three
terminals, you have the base, you have the collector and you have the emitter. So, the
way we have drawn it here it is probably indicated npn in junction, but it does not matter.
The base is going to have some resistance, some input resistance which means that the
control signal for the transistor is going to impact the signal from the collector to the
emitter. So, if you head back to our original discussion on what constitutes a transistor,
you will have an implication of what that you will have an understanding you will
appreciate what that means.
And you correct your collector to emitter current is dependent on the base to emitter
voltage ok. So, let us say that is your V be ok. So, this is going to depend linearly on the
base to emitter voltage provided, we have linearized the circuit which means that these
are all small signal equivalent circuits. I should have probably mentioned that right at the
start. It is a small signal equivalent circuit. So, we have already understood the meaning
of small signal which means we are only talking about small fluctuations and V be, V ce
etcetera, we are not talking about the large applied bias voltages.
So, under these small fluctuations, your collector current, for example, let us say varies
with the base emitter current. So, we have already chosen our dc voltages; and around
these dc voltages we are going to have a small signal fluctuation and that fluctuation is
going to impact the current. And at this point we can linearize this curve and say that it is
got a tangent which has got a slope of g m and therefore, my ic is gm times Veb which in
turn decides the current between the collector to the emitter, so that is the message this
equivalent circuit is trying to portray. It is all right if you do not understand this now
because we will come back to this a little later. So, that is that is the base to emitter
voltage.
And therefore, you have a collector current that is that is g m times the voltage across the
base and emitter. And gm is something called as a trans-conductance of the device ok, it
is a technical term. And rpi it is the pi is got no special significance it is simply because
the name of the model, it is called the hybrid pi model probably because this looks like
an inverted Greek alphabet pi ok. But the point is that r pi indicates the input impedance
523
of the BJT. So, it is basically 𝜏1 that dIB/dVeb, so that is my small signal input resistance.
So, if I fluctuate Veb of the BJT a little how much of base current do I push it, so that is
that is the question that is determined by rpi.
Now, ideally the output impedance of the BJT should ideally be infinite, but we already
saw that there is an early effect there is a base width modulation and therefore, we have a
non ideal circuit which is got the rpi, it is got the current collector emitter current, but it is
also got this output impedance which is dou VEC by 𝛿Ic. Now, since ideally I c was
independent of VEC Ic did not dependent on VEC and therefore, ro was infinite, but since Ic
does dependent does depend on VEC a little because of the base width modulation we
give a special symbol r o which is the output impedance of the BJT ok.
And finally, you have r mu which is another non ideality which is the VEC by dou 𝜕IB
that is how does the collector to emitter voltage impact the base current ok. So, again
ideally it should be in infinite, but in reality you have a finite value for that quite large,
but finite. So, in that is the summary of the BJT that is that is all we are going to discuss.
So, we looked at the geometrical construct, we looked at the device physics and we
identified the general IV characteristics and then we looked at the special IV
characteristics for the three different regions. It is a lot of derivations not particularly
going to be tested in those details, but definitely it is an indicator as to how to approach
the understanding of the device physics of a device. And finally, we looked at the
equivalent circuit diagram.
524
Indian Institute of Sciences, Bangalore
Lecture – 32
Metal Oxide Semiconductor Capacitor (MOSCAP)
We will begin our study towards looking at this very important device which is the Metal
Oxide Semiconductor Field Effect Transistor and this device is quite important because
it is the building block of all modern day integrated circuits ok and before we get to
understanding all the details regarding MOSFETs. We will initiate our study on
something called as the MOS capacitor that is the Metal Oxide Semiconductor capacitor.
Now, MOSFETS itself I mean there are generally speaking there are two kinds you could
say ok. So, you have two modes of operation one is the Depletion MOSFET ok, which is
built a little differently from something called as the Enhancement mode MOSFET. The
focus of our study is going to be the Enhancement mode MOSFET and before we start
this we will look at the MOS capacitor.
525
So, the MOS capacitor has got a structure of as shown here. So, you have a Gate Metal
you have an Insulator or an Oxide you have a Semiconductor and you have a contact to
the semiconductor. So, this gate metal is your control terminal of your MOSFET, when
you talk about sending a signal from A to B ok. So, we have not yet built the MOSFET
here. So, this is just the MOS capacitor. So, what you see is it just a capacitor circuit.
So, this gate metal is given a gate voltage V which is measured with respect to the
reference then the reference contact is the back contact. So, here we accept the reference
contact ground and this VG is with respect to ground. So, VG is the GATE VOLTAGE
provided to the gate metal and then the gate metal sandwiches an insulator sandwiched
between the gate metal and the semiconductor ok. So, this is a very key difference. So,
which means that if this insulator is an ideal insulator there is no current flowing from
the gate to the semiconductor so, this current is ideally 0 at low frequencies ok. So, you
have a gate metal and you have an insulator.
Now, since this insulator in the case of silicon technology since the most commonly used
insulator is silicon oxide or some other oxide the insulators also called as an oxide layer.
So, that is why you have this main metal oxide semiconductor a good equivalent name is
metal insulator semiconductor and the insulator sits on top of a semiconductor and this
semiconductor is likely doped either P or N type ok, but in this case for our example and
for this study we will consider a P doped semiconductor.
526
So, you have the gate insulator semiconductor and back contact or the reference terminal,
which provides not only the voltage reference, but which also supplies this
semiconductor layer with an infinite source of carriers ok. So, this is the structure that we
will be looking at and we let people analyze for some time. Now, as usual let us start
looking at the band bending diagram ok. So, let us say be a good it take the a metal the
gate metal we will we will not worry about this back contact we will just assume it is an
ideal Ohmic contact not worrying so much about it, because everything that is interesting
with regards to this device and with regards to the study happens here at the insulator
semiconductor interface. So, this is the region of interest ok because that is where the
junction is ok.
So, we are basically creating a junction since we have looked at junctions of typically all
our junctions have been a metal semiconductor junction or a P N junction, but here you
are looking at a gate insulator and an insulator semiconductor junction ok. So, since this
is our region of interest all our focus will lie in this particular region.
So, us draw the band diagram. So, I have just thrown in some numbers just to give you
a feel for some typical values. So, we have three different structures. So, first let us say
this is the GATE and this is the GATE METAL. So, we have chosen a metal whose work
function is 4.21 electron volts this is the insulator or the oxide and we have chosen an
oxide or an insulator whose band gap is about 8 electron volt, so, it is a large band gap.
527
So, that is the conduction band that is the millions band and that is a large band gap in
the insulator and the electron affinity of the insulator ok.
So, the electron affinity for the insulator is about 0.95 electron volt. So, the electron
affinity has to be a constant and is irrespective of what kind of junction you create or
what bias you apply etcetera and similarly the work function for a metal is a constant
once you pick a metal you pick the work function and then finally, you have a
semiconductor. So, we have not yet created the junction in this diagram. So, we have the
semiconductor and here you have some band gap.
So, for example, let us say the band gap is about if you take silicon it is about 1.12
electron volt, ok. We have some band gap and the semiconductors doped P type which
means that the Fermi level of the semiconductor lies a little below mid gap ok, it is closer
to the valence band edge is compared to the conduction band edge because it is doped P
type and the work function of the semiconductor is the difference between the vacuum
energy level and the Fermi level of the semiconductor and that is about 4.9 electron volt,
so that is that is what we have chosen. So, these numbers are just examples I mean these
are not the only numbers that you have to use I hope that is obvious and the electron
affinity of the semiconductors again a constant. So, once you pick the semiconductor
your picked electron affinity and let us say the electron affinity is about 4.05 electron
units.
And now we are going to create a junction. So, when you create the junction the work
function of the metal will not change, the electron affinity the insulator will not change
and the electron affinity the semiconductor will not change and if it is a thermal
equilibrium the Fermi level will align ok.
So, let us say we have now created the junction, we have brought these three pieces
together and we have now created a junction and that junction is shown in this figure
here. So, what happens? So, here you have the metal. Firstly, the work or the Fermi
levels of a line the metal Fermi level is in the same position as the semiconductor Fermi
level ok and since there was some work function difference between the metal and the
semiconductor in order to align these two Fermi levels there has to be some residual
voltages ok, because charge has been transferred and in this case charge has been
transferred from the metal to the semiconductor.
528
So, here is an interesting question. So, far we have when whenever we did a band
diagram at thermal equilibrium let us say it is a p n junction or a metal semiconductor
junction. We said that the Fermi levels align at thermal equilibrium and the Fermi levels
align because of the charge transfer and very early on we looked at you know some very
poetic analogies like you know water in a bucket. So, you have these two buckets and
you have less water here and more water here and this correspond to a Fermi level that
was higher or you know work function that was lower whereas, this corresponds to a
work function that is higher and when you merge these two buckets the water flows from
that here there by aligning the Fermi levels ok.
So, we looked at all these analogies and we know it had a mathematical derivation of you
know why the Fermi levels should align at thermal equilibrium because the flux, so the
carriers have to be balanced, but here is the question. So, for the Fermi levels to align
you need to have charge transfer like for example, in this case electrons are moved from
the metal to the semiconductor, but you now have an insulator sitting in the middle. So,
the question is how did the charge transfer ok.
So, there is something for you to think about, but let us just say that the charge did
transfer and the Fermi levels have aligned and this is a reasonably good assumption ok.
So, the metal has transferred electrons to the semiconductor which means there is going
to be some band bending in the semiconductor, but since there is also an insulator the
driving force for any band bending which is the metal semiconductor work function
difference leads to not only a built in potential in the semiconductor, but also a potential
across the insulator or the oxide.
So, this difference in the metal and semiconductor work functions is the driving force for
the band bending and this difference is split between the built in potential of the
semiconductor and the voltage drop across the oxide, but then what else do you see here.
So, since at this edge now let us come to this particular edge lets come to the met the
gate metal insulator interface, so, you see this barrier here.
Now, since at this point the electron affinity of the insulator was about 0.95 electron volt
and the electron affinity of the semiconductor of the gate of the work function of the gate
metal at this point is 4.1 electron volt. So, that is 4. 1 electron volt and therefore, since
529
these two are constants this barrier height has to have a value that is the difference of
these two energies and that is going to be about 3.15 electron volt.
So, similarly you can do the same thing for the semi conductor insulator interface the
semi conductor. So, if you look at this particular point here with where the conduction
band is meeting the insulator at that particular point the semiconductor has got a well has
got an electron affinity of 4.05 electron volt the insulator has an electron affinity of 0.95
electron volt and therefore, this barrier is going to be the difference of those two energies
and it is going to be about 3.1 electron volt. And this work function difference that is if
you think of it in voltages, so we have removed the q here. So, we are looking at the
voltages now this work function difference which is say 4.1 4.9 minus 4.1 ok.
So, that work function difference. So, in this case the work function of the semiconductor
is larger. So, let us say well let us just instead of using negative signs let us just say its
∅𝑀𝑆 let us just take the modulus of all this ok. So, this work function difference has led
to a drop of V ox across the oxide. So, that is why you see that there is an electric field in
the oxide ok and the remaining which is 0.8 electron volt minus q V ox is the energy
difference between the bulk conduction band and the point, where the conduction and the
point near the interface the conduction band energy near the interface ok. So, this is what
happens when you form the junction, when you form the metal insulator semiconductor
and junction.
530
So, this is all at thermal equilibrium. So, we have formed the junction and it is all sitting
at thermal equilibrium with the Fermi levels aligned and we see that simply because of
the work function difference between the metal and the semiconductor, you have all this
band bending, you have an oxide that is got a built in field, you have a semiconductor
with a built in field and we can now solve Poisson’s equation here and estimate the field
etcetera.
Now, the question is how do I not have any band bending ok, let us say I want to make
all these bands flat again ok. So, which means that I want to get the bands back to the
way they were before the junction was formed, but I want to do it when after the junction
has been formed ok. So, I do not know if I am clear enough, but the point is in order for
me to get these bands to be flat again artificially, what I need to do is I need to apply a
voltage on the gate ok, I need to apply a voltage on the gate and by applying a voltage in
the gate what we are essentially doing is we are going to recreate the initial Fermi level
difference.
So, that is the metal work function, that is the semiconductor work function and we want
to recreate the difference of 0.8 electron volt because the only way we are going to get
the bands to be flat is if you go back to if you get the metal and semiconductor to have a
work function difference that is 0.8 electron volt.
So, you want the metal to you want the metal Fermi level to go up relative to the
semiconductor Fermi level and this has to be done by applying a gate voltage. So, when
you since we are looking at in all these band diagrams, we draw it is all drawn such that
the electrons run downhill and the holes like to run uphill, which implies that if I apply a
positive voltage to the metal all the energy levels will go down and if I apply a negative
voltage to the metal the energy levels will all start moving upward ok.
So, essentially what we need to do in this case is apply a negative voltage to the metal
and shift the metals Fermi level upward by about 0.8 electron volt. So, if you can create
this difference of 0.8 electron volt between the metal and the semiconductor Fermi level
then we can get all the bands to be flat because we have removed the cause that created
these built in potentials.
So, you can see that the insulator the bands are all flat here, the semiconductor does not
have any band bending and all the band bending has disappeared, but of course, we are
531
now not at thermal equilibrium because we are applying a gate voltage and therefore, the
Fermi levels are not aligned at this point and the moment you remove the gate voltage
the material head back towards equilibrium with the metal transferring electrons to the
semiconductor and therefore, the conduction band edge near this interface will start
bending towards the Fermi level in the semiconductor.
Now, this gate voltage needed to create this flat band condition ok, something called as a
flat band voltage ok. So, this is a technical term ok. So, at thermal equilibrium we had
some situation it need not always be this is the situation for the numbers we have chosen,
the bands could be bent the other way it really depends on what the work function
differences are what the work function difference is between the metal and the
semiconductor, but in our particular case the bands were bent downward and in order to
achieve flat band condition we needed to apply this gate voltage and the gate voltage at
which the bands become flat that is the gate voltage at which there is no potential drop
across the oxide and there is no built in potential in the semiconductor is something
called as the flat band voltage and we will give it a special symbol we will call it V f b.
So, when my Vg is equal to Vf b all my bands are flat ok.
So, in this particular case we need our Vfb is definitely your ∅𝑀𝑆 phi m s which is the
work function difference between the metal and the semiconductor. It is not the work
function sorry the work function difference divided by q. So, if you are looking at the
potential here, so Vfb is the potential it is not energy all right.
532
So, now let us say what happened? Let us start playing with this system ok. So, this is
our flat band condition, so let us look at the figure on the left hand side first. So, this is
our flat band condition you see there is no banding there is no bending of bands in the
semiconductor, there is no band bending in the insulator and that is because we have
artificially applied we have applied a gate voltage Vg which is equal to Vfb which is the
flat band voltage.
So, now let us say we start making Vg further we start making V g less than V f b ok. So,
we have already applied a negative voltage we start making it more and more negative
ok. So, let us say in this case Vfb was what V f b was minus 0.8 volts because phi m was
phi m is 4.1 electron volt, phi s is 4.9 electron volt and ∅𝑀𝑆 which is ∅𝑀 − ∅𝑆 is going to
be minus 0.8 volts. So, that was our flat band voltage.
And now we are going to make Vg less than Vfb. So, which means that let us say we start
making V g equal to minus 1 volt and then minus 2 volts etcetera. So, what are we
doing? So, we are further increasing this difference between the metal work function and
the semiconductor work function. So, the metal Fermi level will start moving further
away from the semiconductor Fermi level and it will move it will increase this energy
gap between energy difference between these two levels.
So, this means that at Vg equal to Vfb. We were at flat band and if I make it further
negative, then we are again going to drift away from flat band, but in the opposite
533
direction ok. So, since my Vg has now become minus 2 volts you will find that the bands
start bending the other way ok. So, the electric field in the oxide is going to change
direction and you are going to have band bending in this manner and the bands and the
semiconductor will start tilting upward ok, as compared to the downward tilt it had at
thermal equilibrium it will now start bending the other way because we are making V g
less than V f b.
And when we start making Vg <Vfb this band bending is going to encourage all the
majority carriers since remember this is a P type semiconductor and it is got a lot of
holes ok. So, for example, if its doped with the doped concentration of N A and if all
these dopants are ionized then the hole concentration is about N A if n a is much greater
than the intrinsic carrier concentration. So, plenty of holes and these holes are going to
start running up the hill and they are going to accumulate at the interface. So, as I make
V g as I continue making V g less than V f b they are going to encourage holes to appear
at the interface.
Now, let us go the other way around. So, let us say we head back to a flat band condition
and the way we go back is we start bringing this voltage higher and we make it equal to
the Vfb. So, let us say we are back at flat band and now we start taking V g the other way
around. So, we make Vg >Vfb ok, now all as you start making V g greater than Vfb the
Fermi level of the metal will start moving downward ok and you know we will pass. So,
here we have crossed the thermal equilibrium point when these two fermi levels were
aligned. So, as you start pushing the metal work function metal Fermi level below, you
will reach the point where it was you know you will artificially recreate the thermal
equilibrium point because you are to set Vg equal to 0. So, you will make you are
moving from Vg equal to minus 0.8 volts through 0 which is your equilibrium state and
now beyond 0.
So, let us say we make it 0.1 0.2 etcetera. So, we start increasing V g ok, so, when V g is
greater than Vfb the bands start bending the other way. So, because we are now pushing
the Fermi level of the metal below the Fermi level of the semiconductor and what
happens at this stage. So, we are now started adding positive charge on the gate. So,
when Vg was less than Vfb we were adding negative charge on the gate right, we were
applying a negative gate voltage and all this negative charge in the gate was balanced by
the positive positively charged holes at the interface.
534
But now by making Vg greater than Vfb you started adding positive charge on the gate.
So, how is this positive charge in the gate balanced? This positive charge in the gate is
balanced by two mechanisms the first is that the semiconductor starts getting depleted
near the semi conductor insulator interface, so that is the first reaction. So, what does
depletion do it exposes all the fixed negatively charged acceptor ions.
So, it is a P doped semiconductor it has got negatively charged acceptor ions and free at
which have given up holes to the semiconductor and these negatively charged acceptor
ions or the space charge starts getting exposed. So, this region is depleted, so that, firstly
compensates for any positive charge that you are applying on the gate and this bending
of the bands begins to encourage the negatively charged electrons and encourages them
to appear at the interface. So, you have two potential mechanisms the negatively charged
electrons and the negatively charged space charge, which is the negatively fixed
negatively charged fixed acceptor ions in the semiconductor that can balance the positive
charge on the gate.
Now, the electrons will not appear till Ef- Ei ok, is going to go greater than 0 ok. So,
what does that mean? So, we have also shown the intrinsic Fermi level, so this is the
Fermi level after doping. So, you have the semiconductor it initially had an intrinsic
Fermi level position and then after doping we created that we shifted the Fermi level to a
both P type semiconductor and we are using the P type to create your MOS capacitor, but
having said that I have just indicated the intrinsic Fermi level.
So, this is the Fermi level after doping and this is the aligned Fermi level and we have
also indicated the intrinsic Fermi level here ok and that is an it is a very good indicator
because if you recollect your electron concentration is dependent on the exponential of
𝐸𝐹 −𝐸𝑖
ok.
𝑘𝑇
So, it is only when Ef is greater than E i when the electrons start appearing. So, as you
are bending the bands so, for a very low positive gate voltage, so, very so, let us say V g
is just slightly above Vfb the bands would not have bent that to that extent. So, let us say
the bending was only so, let us say this is the Fermi level and let us say the intrinsic
Fermi level has just bent this much.
535
So, 𝐸𝐹 − 𝐸𝑖 is still negative in this case ok, e f and that is 𝐸𝐹 − 𝐸𝑖 i is still negative
and therefore, you do not have so many electrons at the interface, but all the charge all
the positive charge in the gate is being compensated by the exposed acceptor ions ok. So,
that is the depletion region, but the moment you make the gate voltage more and more
positive E f at some point will start bending more and more Ei start bending more and
more and it will ensure that E f minus E i becomes positive near the interface.
So, that is when the electrons start participating ok. So, in this case we have just made V
g slightly greater than Vfb and it is mostly the acceptor like ions that are being depleted
the semiconductors depleted and expose the acceptor like ions.
But then as I mentioned if I you make Vg much greater than v f b. So, when I say much it
just simply means; that, we have bent we have managed to make the band bending so
large or so severe that if I were to imagine, what the intrinsic Fermi level would be doing
it would have gone past the Fermi level of the semiconductor. So, E i has; so, if you look
at E i minus E f in the bulk it is negative, but −(𝐸𝐹 − 𝐸𝑖 ) in the bulk is positive and
−(𝐸𝐹 − 𝐸𝑖 ) in the at the interface has now become negative ok, or in other words
𝐸𝐹 − 𝐸𝑖 at the interface has become positive.
So, at this point we will start encouraging electrons to start appearing at the interface ok,
because the band bending is. So, severe now where are these electrons coming from?
536
There are no electrons in the P type material ok, so the type material has got very few
electrons its only got 𝑛𝑖2 /𝑁 𝐴 . So, these electrons are not you know there are migrating
ok, definitely not these are electrons that are thermally generated in the depletion region.
So, we had this depletion region before heading here. So, we have these exposed
acceptor ions ok, you still have the depletion region here, but beyond a certain point
electrons also start participating and these electrons are thermally generated in the
depletion region ok.
Since your region is depleted it encourages generation of carriers and these are thermally
generated carriers and our depletion region and is those electrons that start appearing at
the interface. And now to compensate for the heavily positively charged gate because
your V g is now very large. So, you have a strong positive charge on the gate metal, this
positive charge is now compensated by not only the exposed acceptor ions ok. So, these
acceptor ions the space charge but also by the negatively charged electrons. So, both q
NA- minus and q n start compensating for the positive charge on the gate.
Whereas in the previous case when V g was just slightly greater than V fb and Ei was not
yet bent so much. It was mostly it was only the acceptor ions that was compensating for
the gate charge. Now the bending the bending could become more and more severe and
we reach a point we could reach a point at some particular gate voltage, we reach a point
where the Ei- Ef in the bulk matches or becomes equal to the 𝐸𝐹 − 𝐸𝑖 at the interface
ok. So, what does that mean? So, let us say we define we define the E i this energy
difference ok.
537
So, let us say we say E i minus E f in the bulk of the semiconductor bulk means very far
from the interface. So, this region here it knows nothing about the junction formation ok,
the semiconductors as it works. So, you have the Fermi level there you have the valence
band there and you have the conduction band and we this dashed line here there are two
dashed lines ok, but what the black one if you can see it is E i ok. So, that is E i all right
and I will draw E f as a continuous line just for clarity.
So, E i minus E f in the bulk is let us say we will give it a special symbol we will call it
phi F ok. So, we will say that this difference is that is a q phi F q phi F. Now at the
interface the band bending has become so, severe that E i is bent to such an extent that E
i has gone below E f, but there is a special point where then E i is as much below E f near
the interface as it is above E f in the bulk ok. So, if you look at this difference let us call
this energy difference also equal to q phi F.
So, let us say we have applied a gate voltage at some particular gate voltage the band
bending is such that, this difference between E i and E f this modulus of this difference is
same as the difference between 𝐸𝑖 − 𝐸𝐹 in the bulk. So, let us say this is 𝑞∅𝐹 and at
some particular gate voltage this difference also becomes 𝑞∅𝐹 ok. So, at this gate
voltage we say that the MOSFET has reached threshold voltage or threshold operation
ok. So, this gate voltage, so, just like how we define the flat band voltage and we said
that the gate voltage, when the gate voltage equals the flat band voltage is the bands
538
become flat. So, them and V g was equal to V f b the bands became flat and when V g is
equal to the threshold voltage you would have this condition ok and the threshold voltage
is given by a special symbol V t and it is an important modeling parameter ok.
And we say that when we build MOSFETS we say that the MOSFET is considered to be
on or you know in or it has got a significant current through it, when the only after V G
crosses VT ok. So, but VT is not some kind of a magical boundary it is not that the
electrons are waiting in the background and saying that you know let us just wait and V g
still not reached VT that is definitely not the case, it is just a very useful modeling
parameter ok. So, for example, even when the band bending is here for example.
So, let us say E f minus E i at the interface is greater than 0, but it is still not equal to phi
F you still do have electrons ok and this region is supposed to be you still do have a
small number of electrons at the interface ok and how many electrons do we have at the
interface? It is quite easy to calculate it is simply n i exponential of (𝐸𝐹 − 𝐸𝑖 )/𝑘𝑇.

So, you have that many electrons at the interface. Now how many holes do you have in
the bulk ok? So, that is the; so, let me just write this out neatly because it is ok.
So, if you look at the so, let us say Vg =VT ok.
539
So, let us say VG is equal to VT at Vg=VT the number of electrons at the interface. So,
electrons per unit volume at the interface is n i which is the intrinsic carrier concentration
𝑞∅𝐹
into exponential of 𝐸𝐹 − 𝐸𝑖 ok. So, is going to be basically your . So, that is the
𝑘𝑇
number of electrons at the interface the number of electrons per unit volume.
What is the number of holes in the bulk? So, P in the bulk ok. So, how many holes are
𝑞∅𝐹
present here? It is 𝑛𝑖 𝑒 𝑘𝑇 again because𝐸𝑖 − 𝐸𝐹 is 𝑞∅𝐹 and VG=VT we said that
this difference is also q phi F and hence the electron population illness. So, another way
of defining threshold voltage is the gate voltage at which the electron concentration at
the interface, the electron or the minority carrier concentration at the interface matches
the majority carrier concentration in the bulk. So, that is another definition of threshold
voltage ok.
So, here we have a p type semiconductor therefore, the majority carriers are the holes the
minority carriers are the electrons. So, when the minority carrier concentration is the
interface which means the electron concentration at the interface matches the hole
concentration in the bulk, then you say that the that VG is equal to VT ok. I hope this is
clear and what we have done now is by making VG much greater than Vfb we have
brought in electrons at the interface.
540
So, you look at the previous case. So, here when Vg < Vfb we had holes at the interface
and holes are the majority carriers they were accumulated at the interface and by making
Vg much greater than Vfb we have brought in the minority carriers at the interface and
therefore, the semiconductor interface semi conductor insulator interface is said to be
inverted ok. So, there is a technical term inverted and what this means is, that we have
brought the minority carriers of the semiconductor close to the interface. The interface
become populated with the minority carriers ok and that is and we say when that happens
we say that the semi conductor insulator interfaces inverted or the MOS capacitor is an
inversion regime.
541
Lecture – 33
MOSCAP – Continued
So, we defined something called the flat band voltage. So, at equilibrium the bands need
not be flat it will be flat only when phi m is equal to phi s, but that need not be the case.
And this ∅𝑀 − ∅𝑆 . is the driving force for any band bending at equilibrium, but now
when we start applying a gate voltage we take the semiconductor we take the MOS
capacitor out of equilibrium and when we make VG =Vfb.
542
We achieve flat band all the bands become flat. So, that is like a starting point for all our
experiments all the bands of flat. And we need to create this kind of an offset voltage to
make all the bands flat when we make V G less than V fb we bring the majority carriers
at the interface. So, in this case; so here we are all talking about only a p type
semiconductor ok. So, the same arguments hold for an n type semiconductor the only
thing is you need to watch for the electrons being the majority carriers and holes being
the minority carriers ok, but here we are talking about a p type semiconductor
So, when VG is less than Vfb we bring holes to the interface ok. So, the hole start
appearing at the interface. And this mode of operation is called as an accumulation mode
of operation; so is again a technical term. So, when we say that the MOS capacitor is an
accumulation mode operation it means that the gate voltage is such that the majority
carriers in the semiconductor are at the interface.
Now, let us say we start making VG greater than Vfb, but we are still less than VT you
know VT is our threshold voltage. So, VG is sitting somewhere between Vfb and VT, but
what happens in this case is that the gate metal now has positive charge ok. And
therefore, it has to be balanced by the negative from the p type semiconductor and that is
done by depleting the semiconductor. So, this is the depletion this region is all depleted.
So, ideally speaking there are no free carriers in that region. So, the insulator
543
semiconductor interface is depleted. And in this case all the negatively charged acceptor
ions are exposed and they are balancing all the positive charge on gate metal.
And this mode of operation is called as the depletion mode of operation depletion mode
of operation. And then when you start increasing VG further and further you first you
will reach a point where the Ei starts going below Ef Ei starts going below E f at interface
and electrons will start appearing at the interface. So, when E i is just started going
below E f it is something called as weak inversion weak inversion. And when E i has
gone well below Ef well below E f and when the when the electrons at the interface when
the electron concentration at the interface matches the hole concentration in the bulk. We
say that the semiconductors the MOS capacitor strongly inverted strongly inverted
strongly inverted.
And this happens at VG equal to VT. And when you make V G greater than V G all the
excess positive charge in the gate is now mostly compensated by the electrons rushing at
the interface and not so much by the depletion not so much by the depletion charge. The
depletion width stops increasing. So, initially this is a story. So, initially you had flat
band when you start making V G greater than flat band first the first thing that happens is
that depletion width.
So, let us say that is the interface the depletion width starts increasing ok. It uncovers
more and more positive charge, but the moment electrons start appearing at the interface
this increase in the depletion width more or less stops. Because a small increase in the
gate voltage will exponentially increase the number of electrons at the interface. And
therefore, you have a lot of electrons coming in there to balance any positive charge.
So, that is the story of everything that is going on in a MOS capacitor.
544
So, go back and look at the band bending picture again just to make things clearer. So,
the moment you hit inversion, so you see this bending. So, initially it was just bent like
that. So, that was the depletion operation. And the moment it started bending like this
you have an equal number of holes and electrons at the interface ok. And then, when you
start bending it further down you have more electrons at the interface than the holes. And
finally, reach a point when the electron concentration with the interface matches the hole
concentration in the bulk and that is what the definition of our threshold voltages.
Now, if you still increase the gate voltage further, this band will not move too much
more. It is not going to bend this much, because now the electrons have started appearing
at the interface in large numbers. So, beyond at this point this band bending is not going
to increase to much more. I mean you need to really make your gate voltage very very
large for you to see any significant movement in the band bending that, and that is
because for any small movement they export the electron population increases
exponentially. And therefore, you have a lot of electrons coming in that easily balance all
the charge and the gate.
So, you really need to make a significant amount you need to add a signature amount of
positive charge in the gate in order for you to see any movement in the band bending.
And therefore, after inversion or after threshold or after strong inversion you could say
that the band bending is the bands are almost pinned ok. So, they are almost pinned at
545
this point they are not going to move too much. And another technical term which I
should have probably mentioned is that is to define this potential.
So, let us say I create this voltage this built in potential as a reference ok.
So, let us say that is the reference potential. This potential which is the total amount of
band bending in the semiconductor has got a special term, and it is called as the surface
potential, phi s is the surface potential ok. So, that is the complete band bending in the
semiconductor.
So, at threshold voltage what is the surface potential? The surface potential is going to be
2𝜙𝐹 phi f, because it is going to be 𝜙𝐹 here and 𝜙𝐹 there ok. So, meant surface potential
is equal to 𝜙𝐹 your VG is equal to your threshold voltage. So, this is a very useful term
to remember, because we will keep using reference in this particular term. And since, we
have defined the surface potential and if you define the voltage drop across the oxide,
you can now the summation of the voltage drop across the oxide plus the surface
potential has to be equal to VG-Vfb ok.
546
So now how do we define threshold voltage? So, let us now start coming to a more
quantitative definition of threshold voltage. Because it is such a useful in modeling
parameter, but you must remember as I mentioned the semiconductors the electrons are
not waiting for the threshold voltage to you know for the gate voltage across the
threshold voltage before; V before they invert. I mean there is some weak inversion
happening even before the surface potential reaches to phi f ok. So, the moment the
surface potential goes increases beyond phi f you have electrons appearing quite rapidly
at the interface. And these are all the thermally generated electrons here.
So now let us say how to define threshold voltage. There are 3 milestones right you need
to cross these 3 milestones. So, how do I define V G in a more quantitative manner? So,
VT looks good, but what is VT exactly you know what does it depend on. So, there are 3
milestones. Firstly, let us make the bands flat. We do not know what the equilibrium
gives you. So, let us first make the bands flat. So, VG has to first cross. So, VG must first
equal the flat band voltage ok.
After that we keep increasing VG you start depleting the interface. You know the bands
you have E f initially that was E i it was all flat band ok. So, that is let me just draw this
clearly. That was E i it is flat band and now you start depleting the interface the E i starts
bending a little and you start having depletion. And then you keep bending a bit more
you still have depletion. So, you then have to cross the second hurdle which is not only
547
have built put added a flat band voltage, but you start depleting the semiconductor and
this total depletion charge by Cox or the insulator capacitance is the effective voltage.
That you need to add, and then you still start keep adding more and more positive
charges still not enough. You have still not reach threshold you keep add adding more
and more positive charge.
And now you will see E go below EF. And now the electrons have started appearing, but
we have still not reached threshold voltage. We reach threshold voltage only when the
surface potential equals 2𝜙𝐹 ok. So, the third milestone is to make Ei bend the E i minus
Ef at the interface have the same magnitude as the 𝐸𝑖 − 𝐸𝐹 in the bulk. Or in other words
make the electron concentration of the interface match the hole concentration in the bulk.
So, the hole concentration in the bulk and the electron concentration at the interface are
the same. At this point the surface potential is 2𝜙𝐹 and your you have added enough
charge to make VG equal to Vfb plus the depletion charge plus 2𝜙𝐹 phi f. And when you
have added enough charge to make all this happen then you can say that you have
reached threshold voltage.
And therefore, the threshold voltage has got these 3 components, it is got V fb which is
the flat band voltage it is got q depletion by C ox and I put a minus sign here, because in
this case it is all the acceptor ions. You will see because the q depletion has got a
negative charge it is negatively charged acceptor ions. So, you first reach V fb, it is the
548
flat band voltage then you deplete the semiconductor. And finally, you make the surface
potential reach 2𝜙𝐹 and it is only when these 3 happen that you could say that that is
enough charge in the gate that from this point on the semiconductor strongly inverted or
the most capacitor strongly inverted and that voltage is the threshold voltage.
So, in some sense this is the definition of the threshold voltage. Now let us start getting
into the details of these little terms here. So, what is Vfb, what is q depletion, what is the
C ox, what is phi f, ok? So, what are these things? C ox is the gate insulator capacitance
per unit area ok. So, it is the gate capacity gate insulator capacitance per unit area. So, it
is not the total capacitance.
So, since it is per unit area what does it depend on. So, let us say the insulator has got an
permittivity of epsilon ox, ox for the oxide. And it is got a thickness of t ox the total
𝐴𝜖𝑜𝑥
capacitance would be 𝐶= where A is the total area, but since we are looking at
𝑡𝑜𝑥
the per unit area capacitance C ox is simply epsilon ox by t ox. So, that is your C ox ok.
So, we understand that in term. What does q depletion? So, this is something you are
already familiar with you have solved personas equation enough number of times that
you know that in a p type semiconductor which is depleted. You have q NA as the
charge minus q NA minus acceptor ions are all negatively charged minus q NA is the
charge per unit volume in the semiconductor into the depletion width right. So, you have
that is x d and you have this minus q NA is the total charge per unit volume. And there is
some area, but since we were looking at charge per unit area we do not really worry
about that is a q NAxd ok.
If you had some area qNAAxt would be the total charge present, because this charge per
unit volume there is part of it which are you know is not visible on the display here.
549
So, minus qNA a into x d is the total charge present, because this is charge per unit
volume it is the area and that is the depletion width. So, the charge per unit area is simply
this stuff. So, that is your depletion charge. So, that is the amount of charge you need to
put on the gate that is completely balanced by the depletion in the semiconductor ok.
So, you understand what this is you understand what this is what is Vfb? Vfb is simply a
𝜙𝑀𝑆 that is what is driving.
550
All this it is a 𝜙𝑀 − 𝜙𝑆 . Ok we call that as 𝜙𝑀𝑆 . So, you understand what Vfb is, now V
fb can have more complications we will come to these things later on, but for now it is
phi m s, and what are these complications. These complications are that you could have
charges trapped in the oxide you know you could have. So, that is that is one
complication that we will look at. So, that will change the flat band voltage
But for now your V f p is simplify m s that is the driving force for all the band bending at
equilibrium. And what is phi f 2𝜙𝐹 phi f is the surface potential at the interface when you
have hit threshold voltage, but what is 𝜙𝐹 is the 𝐸𝑖 − 𝐸𝐹 by q in the bulk right. And how
do you know what is 𝐸𝑖 − 𝐸𝐹 by q in the bulk. So, if you know the doping concentration,
if you know NA you can say NA is approximately your p and p by n I is equal to the
exponential of q𝜙𝐹 /𝑘𝑇. And therefore, 𝜙𝐹 is nothing, but kT/q into logarithm of your
NA/ni. So, that is your that is your 𝜙𝐹 , and we have already done these kind of
calculations before.
So, if you know the doping concentration you know the intrinsic carrier concentration
you know what𝜙𝐹 is ok. So, this is the threshold voltage. Now we are still not happy
with this definition right. So, what is x d?
We have not defined xt. And how do you define x t? You solve Poisson’s equation. So,
we have done this many times you just write you assume full depletion you make the
assumption of full depletion that is there are no free carriers. Therefore, Poisson’s
551
equation is simply this term here and therefore, your x d your potential turns out to be
this term. So, I will not go through these details or I think all of you are, now at this point
familiar with all this that is the potential and if you say that the bulk is at 0 potential the
interface. So, the bulk is that 0 potential the interface will have a potential of this term
here and that is the surface potential.
So, that is the surface potential at the interface. And what is x d: x d can be written in
terms of the surface potential as this. So, what is the surface potential again? It is you are
imagining the intrinsic Fermi level bending this way. So, if you keep your bulk as the
reference let us say that is reference or ground. It is this difference. So, that is your
surface potential. So, the surface potential equals 2𝜙𝐹 . Then we have reached threshold
voltage. It could be less than to 5, but when you reach 2𝜙𝐹 it is threshold voltage and
when surface potential is greater than 2𝜙𝐹 you are above threshold ok.
But as I mentioned that after inversion the surface potential is more or less pinned it is
not bands are not going to bend too much more. I think the surface potential is more or
less pinned and therefore, you can now say the threshold voltage is V fb which is now
written out as phi m s. This is q depletion by C ox, because now we have taken into
account x d ok. So, you multiply. So, q depletion now becomes q i have taken care of the
negative signs we have removed all that; so qNA xd ok. Let us q depletion is the charge in
depletion charge per unit area q NA into x d and for x d you substitute this term here and
you will end up with this being the value for q depletion. And that is C ox we already
know what Cox is it is the capacitance the oxide capacitance the insulated capacitance
per unit area plus phi s and at threshold 𝜙𝑆 is 2𝜙𝐹 ok.
So, that is the threshold voltage. So, if you know the doping concentration you basically
know phi f. And so, here again here the surface potential term inside has been.
552
Two 5 f have been substituted for the surface potential term. So, that is why you have phi
f there if you know the doping concentration, you know phi f you know this you know if
you know 𝜖𝑜𝑥 and 𝑡𝑜𝑥 no C ox that is the permittivity of the semiconductor and that is
the work function difference between the metal and the semiconductor. So, this work
function difference divided by cube. So, this is the potential term, it is a potential. And
you end up with the threshold voltage which is a very useful modeling parameter.
553
So, here are some very important or useful relations, we will sort of summarize all this
by these equations here; so VGS -Vfb. So, let us say. So, let us do pardon this usage of s
ok. So, that is just a force of habit. So, you can say s is the reference potential here. So,
once we learn MOSFETs you will understand what the VGS is, but just replace all s with
just VG ok. So, you have VGS -Vfb is the driving force.
That is the x. So, we have VG=Vfp you have flat band there are no voltage drops
anywhere, but when VG is different from V fb you start having a you have start having
electric fields in the oxide and in the semiconductor. And therefore, you have some
voltage drop across the oxide and the semiconductor.
So, the first key relation is that VGS -Vfb that voltage has to be split between the oxide
and the semiconductor. And in the semiconductor it will be it will appear as the surface
potential. So, let us say that is the drop across the oxide the remainder of it will be seen
as the built in on the surface potential in the semiconductor. So, Vox +𝜙𝑆 phi s is equal to
your V G minus V fb, but what is V ox? Think of the semiconductor like this ok. So, you
have you have the gate metal you have the insulator and you remember this is a MOS
capacitor right. And you have the semiconductor and all the charge here is being
compensated by some all the negative charges in the semiconductor b depletion charge
or b it electrons. So, this is your most capacitor in action right.
554
So, what is this? If you think of it as a fuse your old, I mean your conventional capacitor
notations. That is Cox that is the oxide capacitance per unit area, into V ox is the voltage
across this capacitor right as the voltage across the insulator right. That is the voltage
difference voltage drop across the insulator that is V ox. So, C ox into V ox has to be the
charge stored on the gate ok. And the charge stored on the gate is the charge in the
semiconductor ok. And what are the charge in the semiconductor, that it is a summation
of the depletion charge and the inversion charge ok.
So, C ox is all per unit area here. So therefore, your C ox into V ox is equal to the
depletion charge plus the inversion charge. And I put a negative sign because all this is a
negative charge for this example. And therefore, your Vox is simply the depletion charge
per unit area plus the inversion charge by inversion charge you mean that you mean the
electrons that have appeared at the interface the inversion the minority carriers ok.
So, you have the depletion charge, you have the inversion charge divided by C ox is your
V ox. And now, we will substitute for this V ox here and what do you get you get VG
minus Vfb is Vox+𝜙𝑆 , but you have already defined the threshold voltage isn’t it. So, the
threshold voltage is Vfb it has the q depletion by C ox and it has got the phi s term in it.
So, if you replace these 3 terms with the one term which is the threshold voltage, what
we are essentially saying from all these equations is that VG-VT that is the excess gate
voltage above the threshold voltage ok.
555
And that is something called as the overdrive voltage it is helpful to know this term in
case you hear this in the context of MOSFETs what they are saying is it is VG -VT ok.
So, again let us just replace all this as V G. So, VG -VT, because I have now gotten taken
care of these 3 terms is simply your q inversion by Cox. So, what you are saying is that.
If you want to know how many electrons are there the semiconductor insulator interface,
you want to calculate you essentially want to calculate q inversion. What you need to do
is simply say that q inversion is equal to C ox into V G s minus V T ok, the magnitude of
q. So, VGS -VT is the voltage access voltage above the threshold voltage because still
threshold voltage you did not have inversion charge.
So, you do not have a lot of it, but above the threshold voltage you start having
inversion. And therefore, the excess voltage everything that go all the V G s what this is
saying is that all the VG above the threshold voltage is creating electrons at the interface.
And you can further write q inversion as q into n ok. Where n is the number of electrons
per unit area and therefore, here you have the number of electrons if you want to know
how many electrons are there per unit area.
556
At the interface that is the answer ok. So, that those are some useful relations which will
help you imagine the situation a little better.
So, that you are not surprised when we just start writing these equations. So, it all comes
from this basic understanding or this description of your device so far. The other relation
now how do you find out what phi s because you have one equation, but I still cannot
calculate phi I know what VGS because I am applying V G. We can find out what V fb is
by looking up the metal and semiconductor work functions, but we still not still do not
know what phi s ok. Because we do not know what e ox’s here, we ox’s V ox all this
term, but we cannot calculate it without calculating phi s because your q depletion is also
dependent on phi s ok.
So, you see it is a quadratic relation your q depletion does depend on phi s and you also
have a phi s there. So, we need another equation. Now the other equation is equating the
displacements at the insulator semiconductor interface. So, what this equation says is the
epsilon ox epsilon ox 𝜖𝑆 G -VT, electric field in the oxide this product at this interface
must be must be continuous, which means it must match epsilon s into the electric field
in the semiconductor.
And you can calculate the electric field in the semiconductor by solving Poisson's
equation. And that calculation will have an x d or a phi s dependence on it. And by using
this relation you can estimate what 𝜙𝑆 ox and what is epsilon ox it is simply going to be
557
your 𝑉𝐺 − 𝜙𝑆 /𝑡𝑜𝑥 in a very crude sense if there are no trapped charges in the oxide that
is it. So therefore, you have 2 relations with fires in it and you solve for fires. And these
are not easy equations to solve ok.
So, with that we complete the basic understanding of the MOS capacitor, you know how
the MOS capacitor works the operation of the MOS capacitor. And from this point on we
will start looking at some electrical properties, particularly what is the capacitance or the
MOS capacitor. We will start quantifying the capacitance. And this piece of information
this knowledge is quite useful when we start talking about MOSFETs.
558
Lecture – 34
MOSCAP: CV Characteristics
So now let us look at the MOS capacitor, but start looking at the capacitance voltage
characteristics of the MOS capacitor. The essential idea of this experiment is to see how
does; see measure the capacitance of the small signal capacitance of the MOS capacitor.
And see how the MOS capacity the capacitance varies when the V G or the gate voltage
varies.
So, we have already done these kinds of exercises before. You know, if you look at if
you remember the metal semiconductor junction, we built an equivalent circuit. We said
that the metal semiconductor junction, you know it had rectifying nature, if it was a
Schottky junction it had a depletion capacitance, etcetera. In the case of a diode we said
yes, in reverse bias it had a depletion capacitance and in forward bias it had a depletion
and diffusion capacitance.
559
So, this is another such exercise, but with the metal oxide semiconductor capacitance ok.
So, we have been calling it a capacitor. So, let us start measuring the capacitance, what is
the capacitance of this capacitor. So, here is the experimental setup ok. So, how do you
measure capacitance? In order to measure capacitance essentially you need to give an
AC signal, and measure the current through the AC current due to this AC signals.
Because a capacitor would have an equivalent impedance of about one by j omega c, if C

is the capacitance the impedance is about 1 by j omega c; where omega is the frequency
which is angular frequency which is 2 pi f, with f being the frequency in hertz, C being
the capacitance and j the complex number square root of minus 1 ok.
And if you apply a sinusoidal or a time varying voltage, it is only a time varying voltage
that is going to create a current in the capacitor. And therefore, the time varying voltage
is going to create a current of V by this impedance; which is going to be the voltage
applied into j𝜔𝐶 . And by measuring this current and by knowing this voltage and
knowing the frequency, you can extract the capacitance.
So, there are many equipments or instruments that will measure the capacitance for you
in a very accurate manner. And our investigation here is not the development of the
instrumentation to measure the capacitance, but let us assume that we have an accurate
estimate of the capacitance. And we are trying to see or guess what the capacitance will
be based on our understanding of the MOS capacitor.
So, this is our experimental setup ok. So, we have our gate voltage, but the gate voltage
is now variable can varied from say a very negative voltage to positive voltage. So, we
have a variable gate voltage. And on top of this gate voltage, so we are going to apply a
gate voltage. And on top of this gate voltage we are having a small signal that is our
capacitance measurement signal, that is applied a small signal AC that is applied along
the large signal gate voltage. The gate voltage determines the DC value and the small
signal determines the AC value, ok.
So, it is an AC signal, and all operations are linearized around this AC measurement
right. And using these 2 in combination, we are going to estimate what the capacitances,
we are going to see what the behavior of the devices. But as before you know, what is
560
small signal capacitance? So, let us say you have a device sitting in a black box and, let
us say it is got 2 terminals outside. The way we want to define capacitance is we will met
we will apply this small signal fluctuation, the voltage fluctuation. And we are going to
estimate what the charge fluctuation because of this small signal is ok.
So, how does the charge in the device under test respond to our small signal fluctuation?
That is the measurement of this small signal capacitance. So, it is dQ by d applied
voltage. So, that is the small signal capacitance ok; so that is. So, we are going to what
we are going to do in this experiment? Is you are going to have this AC signal we are
going to set the frequency of the AC signal we have a knob to set the frequency. The
magnitude is typically kept very low we will not worry about the magnitude of the AC
signal; we will keep that constant. But we are very interested in the frequency of the AC
signal and we are interested in varying the gate voltage; and also the rate at which we
vary the gate voltage. And I will let you know about nodded at some point ok.
So, first let us say let us run our first experiment ok. So, let us not worry about the MOS
capacitor to start with. Let us start with a very traditional metal insulator metal parallel
plate capacitor, that all of you are very familiar with. So, let us say we were trying to
build the mosque capacitor, but we forgot to put the semi conductor ok.
561
So unfortunate, but we ended up with this structure here. So, you have the gate metal,
you have the insulator and you have a rather back contact. And the experiment we are
going to run is we are going to vary the gate voltage. We are going to vary V G from a
very negative value through 0 to a very positive value or let us say through flat band, but
here flat band does not make any sense; so from negative to 0 through positive value.
And we are going to see what the capacitance per unit area is, what is the capacitance I
am recording from this measurement. So, what do you expect for a metal insulator
metal? So, let us say the thickness of the insulator is t ox and it is got a permittivity of
epsilon ox ok. So, what is the capacitance per unit area of this capacitor? It is simply it is
parallel plate capacitor; so therefore, epsilon ox by t ox ok. So, do you do we expect the
𝜀𝑜𝑥
to vary with the applied voltage? The answer is no, because the insulator is ideal,
𝑡𝑜𝑥
there are no trap charges in the insulator and we do not expect any variation, ok.
So therefore, there is all that we expect to see is this. So, if you were to sweep V G as
you go from negative values to positive values, this capacitance will be constant ideally,
𝜀𝑜𝑥
and the value of this is going to be . So, that is the value it is going to take ok
𝑡𝑜𝑥
562
Now, we can also try to vary the frequency of the AC signal, we can try to vary the speed
at which we sweep etcetera. But ideally, for an ideal capacitor all this should not make
any difference, you can measure the capacitance at any frequency and you should see the
same behavior. But of course, in reality it is not going to be. So, because that every
capacitors also got a say a resistor in parallel etcetera. So, there are many leakage
mechanisms instead of it is not going to allow you to see this response. But let us just
talk about the ideal case for now ok.
So now let us say we went back and we constructed our MOS capacitor more correctly
ok. We now a brought in our semiconductor, and we are going to do the same
experiment, but with the metal insulator semiconductor metal structure. So, what is going
to happen? So now, as you vary the V G our semiconductor is going to react to it, and the
structure is going to behave very differently, and why is it different?
So, let us say we are in accumulation mode; that is, V G is less than V fb we are in
accumulation mode. And what happens in accumulation mode? All the majority carriers
appear at the interface. And in this case the majority carriers in the holes; and you find
that all the holes have appeared at the interface. The band bending is now like that, and
you have all the holes appearing at the interface. So, for VG <Vfb if I were to provide a
563
small signal fluctuation; I have now set my gate DC voltage right. So, let me pin my DC
voltage further.
So, we have now set V G equal to minus V G minus V fb equal to minus 1 volt and left it
that ok. So, the DC point is set. And now we are going to apply the small signal AC ok.
So, I have forgotten to add the AC source everywhere. So, we apply the small signal AC
and we are now going to measure the response of the carriers in the MOS capacitor. So,
what is going to happen if the gate metal fluctuates a little? You will find that since
though there is a charge fluctuation on the gate, since the negative charge and the gate is
fluctuating a little.
So, you have some negative charge on the gate, because of the DC voltage, but now
because the AC signal this negative charge is fluctuating. This small fluctuation has to be
reflected by the whole population. And the holes are all the number of holes is also going
to fluctuate, in response to the fluctuation here, ok.
So, it is as though you now have a capacitor which is got a gate metal, you have the
insulator and this fluctuating whole population, there are plenty of holes in the
semiconductor there is no dearth of holes. It is a p-type semiconductor, the majority
carriers. This fluctuation of the holes is like equal to another metal plate if you imagine.
So, you can imagine a metal plate an imaginary metal plate here and therefore, it is like,
you know, you have carriers fluctuating ok.
If you want to imagine it that way, but more properly what we are looking at is the
charge fluctuation in the device under test in response to the applied signal. And what
does the charge response? It is the hole fluctuation in the semiconductor insulate
interface.
And therefore, our dQ by d applied the small signal applied voltage is simply going to
be, what is the whole population going to what is the fluctuation there? It is simply C ox
into Vox right; so Cox into v, so that is the voltage difference. And therefore, our
capacitance is going to end up being just C ox, but now let us say we increase the gate
voltage, we reach the point where V G is greater than V fb, but it is still less than V T.
We have still not reach threshold voltage, but we are greater than the flat band voltage.
564
So, in this case the interface is not depleted. So, we are in the depletion mode of
operation. So, what is this capacitor like? You know, you can now imagine this back
metal with an ohmic contact and is doped semiconductor width which is basically like a
resistor ok, and you now have a gate metal here. And in between you have 2 capacitors
sandwiched in series. You have the insulated capacitor and then you have the depletion
capacitance ok. And we all know how to calculate the depletion capacitance by solving
Possion’s equation.
So, this region is depleted does not have free carriers, you can imagine it to be an
insulator another insulator ok, if you wish. So, you have a series combination of these 2
insulator; so the capacitance now decreases. It becomes a series combination of not only
it no it is not only Cox, but it is a series combination of C ox and C depletion in depletion
mode operation. So, that is the capacitance that you see. And if you want to imagine the
response to the charge, how is the charge responding?
So, let us say you have applied your AC signal; there is a small fluctuation on the gate
ok. So, we have now set V G is equal to is greater than V fb, but it is less than V T. So,
let us say that is our V T point so, that is V T. And this we are plotting V G minus V fb.
So, we have somewhere there. Just slightly above V fb. So, you have depleted the
semiconductor and now you are fluctuating this voltage. So, there is a fluctuation in this
voltage right.
So, we are moving this voltage back and forth with a small signal AC. So, how does the
semi conductor respond? It responds by increasing or decreasing the depletion width.
So, the x d increases or decreases in response to the applied voltage. And by increasing
or decreasing the depletion width, you are either covering or uncovering all the fixed
ions. So, if the depletion boundary is there you have uncovered all these and if the
voltage on the gate increases the little, you need to add more negative charge and
therefore, that goes away and the x d increases a little bit more. You have more negative
charge being uncovered. So, that is the response of the semiconductor to the applied AC
signal in depletion mode operation. And therefore, it is a series combination of Cox and C
depletion ok.
565
And what about the inversion mode: inversion mode operation; so, if my VG becomes
greater than VT ok, greater than or equal to VT you have a lot of electrons in the interface
ok. So, electrons have appeared at the interface. So, I have still shown the depletion
region here, but the electrons are now sitting there. They are probably I should have
redrawn this picture, but this picture is not entirely accurate.
Because you can still see the depletion region, the depletion region does exist, but they I
am I am worried about the way I have located the electrons. So, these electrons must be
located at the semi conductor insulator interface. So, the electrons are located there, not
on the other side of the depletion region. So, this is a mistake so, do I apologize and it is
a mistake. So, it is actually quite a serious mistake, I would say me because it changes
the situation of it.
So, the electrons are at the semi conductor insulator interface, you still do have depletion
regions, but the electrons are between the depletion region and the insulator ok. So, they
are at the interface ok. So, please do make this correction. And since the electrons are at
the interface, any increase or in any fluctuation of charge for V G greater than V T. So,
we are now operating there. So, we have crossed V T, we are now operating here, and
any fluctuation in the gate voltage is going to respond to the electron population
increasing or decreasing ok. So, this situation is exactly like, you know if you think of it
as a met if you want to imagine things as a metal insulator metal plate, this electron sheet
is acting like your pseudo metal plate.
So, it is like exactly like what the holes were doing, the electrons are doing the same
thing now. Or in other words, all the response to the applied charge in the gate is being
met by the fluctuation in the electron charge at the semi conductor insulator interface.
The depletion width need not increase or decrease anymore. So, the depletion width is
more or less static now.
So, that is not the response anymore ok. So, here it was a depletion width that was
responding, here it is not, here it is the electron count that is responding. And therefore,
our capacitance once again becomes same as Cox ok. So, the capacitance is C ox, then it
decreases, then it again increases and comes back to C. But this is all under a specific
566
condition and we will explain that conditions. And the condition has got to do with the
frequency of the AC signal that we are using to make this measurement, ok.
So, if so, here in the accumulation mode we had holes appearing at the interface. And
there are plenty of holes in the semiconductor absolutely no problem, they can you have
enough holes present for you for them to respond to your applied gate voltage. In the
depletion mode, if the frequency changes, the frequency of AC signal changes, will the
covering and uncovering of the ions respond to an increase in frequency? The answer is
yes, because it is a response of an electric field, right it is the electric field whether the
bands are bending fast enough. So, of course, they can respond.
What about the appearance of electrons? The minority carriers they interface. So, these
electrons are very few in the bulk material. So, where are these electrons coming from?
These electrons are coming from these are thermally generated electrons in the depletion
region, we are thermally generated, they there is a certain rate of thermal generation ok.
So, there is a certain rate at which they are being generated. And if the AC signal
frequency is much higher than this rate, then the electrons cannot respond to that AC
signal frequency. And if they elect AC signal frequency is much lower than this rate,
then no problem, the electrons will respond in the manner which we described ok.
Therefore, the signal frequency becomes very important when we look at the response,
the capacitance of the MOS capacitor in inversion mode, when the minority carriers
come into the interface.
567
00Semiconductor Devices and Circuits
Lecture – 35
MOSCAP: CV Characteristics – Continued
So now let us divide the capacitance voltage characteristics into 3 parts, and we will just
go one by one. So, first is something called the low frequency characteristics; what does
this mean? This means that I am sweeping my VG very slowly. So, the DC voltage ok.
So, let us say we change the DC voltage and then you do a measurement with an AC
signal.
So, that VG is being changed very slowly and the AC signal being used to test the
capacitance is also very low. The frequency is low. And what do you mean by it is low?
It is low implies that the electrons in inversion we are only interested in version when we
talk about the frequency in inversion there is enough time for the electrons to be
thermally generated, and for them to appear at the interface in order to respond to the
fluctuating AC signal.
So, you have an inversion layer. So, that is your insulator you have. Firstly, formed an
inversion layer in inversion and the frequency is low enough that when this charge goes
568
up thermally generated electrons can appear at the interface. So, the electrons are being
are appearing and disappearing at a certain rate ok.
At the generation and recombination rates and the generation rate is fast enough for the
electrons to respond to this low frequency AC signal. So, that is the low frequency
characteristics. And based on our discussion earlier in accumulation mode if you were to
plot the capacitance versus voltage the gate voltage and sweep the gate voltage.
So, let us say let us mark our VFB somewhere or easier to just mark VG minus Vfb, but
nevertheless. So, in accumulation mode you have holes of the interface the capacitance is
𝜀𝑜𝑥
equal to C ox which is , but the moment the VG starts going above VFB ok. So, let us
𝑡𝑜𝑥
say you start going above VFB ok.
The capacitor the MOS capacitance starts entering the depletion mode of operation. So,
this is all the depletion mode and in depletion mode the C ox appears in series with the
depletion capacitance. And therefore, the capacitance is now lower to capacitors and
series the capacitance becomes the effective capacitance becomes this. And it becomes
smaller than C ox and you will find that the capacitance voltage curve dips in the
depletion mode, but then you will start reaching a point where E i starts going below E f
and the electrons start appearing at the interface. So, soon once the electrons start
appearing at the interface the electrons also contribute to the response. And you will find
that you enter something called as the weak inversion mode ok
So, we have not discussed this mode in full detail, because for that we need to solve
poisons equation accurately. So, you can; so you quickly go and get into inversion mode,
where you have electrons a lot of electrons in the interface and the capacitance once
again becomes Cox, because now the electrons can easily respond to this small low
frequency electric AC voltage.
569
Now, what about the high frequency characteristics ok. So, what is the experiment we
are doing here? So, in this experiment you have 2 things to vary, you have VG and you
have your e C signal we can pick your AC signal frequency. The VG is varying very
slowly you set this you make a measurement then you set this and you make
measurement you give the device enough times. So, the VG is varying very slowly ok.
So, let us say you are manually turning a knob, but this frequency the AC signal
frequency has now become very very large, you have ramped up that AC signal
frequency it is not low it is become very large and it is become. So, large that it is greater
than the generation rate the thermal generation rate of the electrons. So, in this case what
happens?
So, when the MOS is in accumulation mode the holes have to respond to this high
frequency AC signal. And there are plenty of holes they are abundantly available and
there is a band bending which encourages whole step here at the interface and the holes
easily respond to this charge. So, the whole pub the accumulation mode capacitance will
not change. It will be exactly like your low frequency capacitance curve and that will be
your C ox which is epsilon ox by t ox. And then in depletion mode the depletion width
starts increasing or decreasing in response to the applied AC field and that too can
happen quite quickly it is.
570
But now when you hit the inversion mode the thermal generation rate of electrons is not
fast enough for it to for the electrons to respond in time to the AC signal. So, the
inversion layer is present ok. So, it is not that the inversion layer is not present the
inversion layer is present because you have made the VG very positive. So, VG is large.
So, you are first created the inversion layer no doubt the inversion layer is present.
But now to on top of that V, DC you have an AC signal and for that AC signal the
electrons have to either more electrons have to come in or some of the electrons have to
go away as. And when the signal goes up and down that is the response we are trying to
measure that is the small signal capacitor. And this AC signal frequency is so, large that
this response cannot happen this response is not seen the thermal generation rate is too
slow to match that AC signal response.
So, although there is an inversion layer this fluctuation and electron charge or the
electron concentration does not happen and therefore, the measurement device thinks
that the MOS is still in depletion mode operation. You can continue increasing a V G,
but it sees no response of the inversion charge no small signal response of the inversion
charge. And therefore, the value remains pinned at this value here ok. And that value
there is approximately your 1 by 1+Cox +1/Cdep.
On the other hand in the low frequency characteristics the electrons thermally generated
electrons could respond and we saw that the curve looked like that ok. That was the low
frequency characteristics. And this is the high frequency characteristics of the of high
frequency C V characteristics of the MOS capacitor
Now, we have one more case.
571
And this case has got to do with the DC voltage and the AC voltage. So, we know not
only use a very high frequency AC signal, but the rate at which we are sweeping the DC
voltage is also very large ok. So, instead of you turning the knob very slowly imagine the
MOS capacitor experiences a rapid pulse, let us say the VG went from 0 to whatever 10
volts and or you know minus 5 to 5 volts within a few picoseconds or a few nanoseconds
a very sharp pulse appeared at the gate.
So, the DC voltage itself was very very varied very rapidly and the AC frequency is very
large. So, what happens in this case? The holes will respond the depletion charge will
also respond, but then since the voltage went up. So, quickly it was too quick for the
electrons to even form an inversion layer ok.
So, you have your metal insulator semiconductor interface the DC voltage went up. So,
rapidly that the thermally generated electrons are the thermal generation rate is too slow
for it to even match the DC response. Therefore, an inversion layer never formed the
electrons never got to the interface ok.
So, how is the positive charge on the metal plate compensated? Now it has to be
compensated by the depletion charge, because there is no inversion charge anymore. The
generation rate is too small. So, you have the depletion width here normally when the
DC rate was slow beyond this width the inversion layer would have formed and after that
should have been the electrons that were responding to the applied gate voltage.
572
But now, the electrons have not formed and therefore, the only way the excess charge in
the gate is compensated is by increasing the depletion width. So, the depletion width
continues to increase like as though it is still in depletion mode. And therefore, the
depletion capacitance becomes smaller and smaller, because x d is increasing say the
depletion capacitance is what it is epsilon s by x d so that becomes smaller and smaller,
because the x d keeps increasing.
And therefore, the series combination of C ox and C depletion also become smaller. And
smaller and therefore, the curve looks like this ok. So, it is those drawn in a very non
artistic manner it is not really accurate, but it portrays the message. The curve looks like
that ok. The capacitance keeps going keeps heading low it does not, it does not flatten
because there is no electrons present.
So, to summarize these 3 characteristics: so what is the difference between high

frequency and deep depletion?
So, this character is called the deep depletion characteristics ok. I have no I do not know
whether I mentioned that and what is the difference between high frequency and deep
depletion. In high frequency that DC voltage is varied very slowly, but the AC signal has
got a high frequency. In deep depletion not only does the AC signal have high frequency,
but the DC voltage is also varied very rapidly ok. So, that is the deep depletion
characteristics.
573
So to summarize all these 3, I have taken this nice picture from an online source and
what we are plotting here is C by the C is the effective capacitance of your MOS
capacitor C by C ox. So, first is the low frequency. So, we have the low frequency
characteristics the high frequency characteristics and the fast sweep or the deep depletion
characteristics.
So, the low frequency characteristics you have the 3 different regions of operation you
have the accumulation the depletion the inversion the weak inversion what we have been
calling as the weak inversion; and the strong inversion which means, which is well above
threshold voltage. We have still not discussed this portion very clearly because we have
not solved Poisson's equation accurately.
So, in accumulation mode do you have C effective being equal to C ox or C by C ox is

one because it is holes appearing at the interface and then in depletion mode you have
the series combination of C ox and C depletion and the capacitance goes down. So, this
happens after V is equal to VFB.
Now, once you hit threshold voltage the device approach is strong inversion. And you
have electrons at the interface and they can easily respond to the low frequency signal
and you have your low frequency characteristics. And then in the case of the high
frequency characteristics these 2 regions are the same the accumulation mode and the
depletion mode is the same. But the electrons can now no longer respond to the small
signal capacitance cannot reflect the response of the electrons to the high frequency AC
signal, because the thermal generation rate is too low to match the high frequency signal.
And therefore, the capacitance appears to be flat it appears to be the same as the
capacitance at over here which is basically the series combination of your depletion
capacitance and oxide capacitance. And finally, we have the deep depletion
characteristics where the DC voltage was itself varied very fast and the inversion layer
itself did not formed.
And therefore, the only way to compensate for the charge and the gate was by increasing
the depletion width and then you have characteristics that reflect the deep depletion
behavior of the MOS capacitor. So, this is the qualitative understanding. So, what we are
going to do now is very quickly quantify these results.
574
So, a very simple model what is the model for accumulation mode the capacitance is
simply C ox in depletion mode this is our capacitance it is a series combination of C ox
and C depletion. And C depletion is epsilon s by x d and we know what x d is in terms of
the surface potential. So, these are, I am not going to go we are not going to spend too
much of time in these relations this is. So, that is the depletion mode capacitance.
And for inversion mode you should know whether you are in low frequency or high
frequency. And low frequency it is going to be C ox and in high frequency is going to
take the same value as the depletion mode capacitance ok. The only thing is you need to
be use the same capacitance at different VG values all the VG values.
So, this is a simple model. In order to get a more accurate model you need to solve
Poisson’s equation more accurately ok.
575
So, the very general Poisson’s equation is simply this it is d square psi by d square phi by
d x square is q by epsilon into any acceptor ions donor ions, electrons, and holes you
cannot get more general than this for a for our case. And the way we are going to define
the potential what is this ∅.
The way we are going to define the potential is this is E i that is E f and that is the band
bending this is the reference potential. And the way we have defined ∅ is it is this
potential ok. So, it is basically phi is a function of x. So, if that is the bulk that is the flat
band and during band bending the Ei bent this way. So, that becomes phi of x. As x
𝐸𝐹 −𝐸𝑖
varies your phi varies and what is Ef- Ei is an indicator of what n is n is n 𝑒 i 𝑘𝑇 times
e to the power (Ef-Ei )/kT and 𝐸𝐹 − 𝐸𝑖 can be written in terms of phi.
Similarly, p can be written in terms of ∅𝑁𝐷 is nothing but 𝐸𝐹 − 𝐸𝑖 in the bulk ok. So, that
can be written in terms of ∅𝐹 and NA can be written in terms of ∅𝐹 ok. So, you define all
these 4 variables NA ND np and n and p, in terms of∅𝑛 ∅𝐹 ok. So, these things should be
quite clear because we have already discussed such a case in the case of metal
semiconductor junctions.
576
And you solve this differential equation ok, and how do you solve this. So, this becomes
your differential equation how do you solve this differential equation. You multiply both
sides by 2𝑑∅𝐹 /𝑑𝑥 and the left hand side simply becomes this term here ok.
And what is this there is nothing but the electric field squared. And you solve this
differential equation.
You will end up with the electric field you will end up with an expression for the electric
field ok. So, I have just written out the solution for your reference this part of it is neither
577
for any test and it, but although it is quite important, because it gives you the exact
solution.
So, that is the electric field and in some sense this is the low frequency electric field and
why is it a low frequency, because we have assumed electrons are present at the
interface. So, this is all this is not the deep depletion case. It is definitely either low
frequency or high frequency characteristics, but the electrons are present at the interface
the inversion layer is formed ok. And then you use this condition ok. So, this is the
electric field at what gate voltage is this electric field present you need to calculate what
phi s is and you use these 2 conditions to calculate to solve for ∅𝑠 .
So, you know ∅𝑠 and therefore, you know the electric field. So, this is the exact solution
to Poisson’s equation.
And that solution can also give you a indicator as to what the charge is at the interface,
the charge concentration at the interface. So, this plot is borrowed from the textbook of S
M sze and what you are seeing here is essentially what happens when you start varying
phi s ok. So, do not worry about too many details of this plot, but this probably this label
has gone away from your screen. So, this x axis is ∅𝑠 is being varied ok.
So, initially you are in accumulation mode. And you find that the hole concentration is
basically this is the hole concentration. So, you are in accumulation mode and you have
578
holes that are varying as exponential of phi s. And then you enter depletion and your
carrier and your charge in the semiconductor is basically the depletion charge and how
does the charge vary it is q and q NDxd and how does xd vary with ∅𝑠 .
It varies as the √∅𝑠 and that is what is being reflected in these characteristics here. So,
you see a √∅𝑠 dependence. And then finally, you hit inversion wherein the electrons
start appearing at the interface and you have an exponential increase with respect to ∅𝑠 .
So, it is a nice summary of all the charges in the semiconductor I should say
semiconductor rather than the interface.
So now we can solve for the exact capacitance voltage characteristics ok. So, these
equations are borrowed from this source; so what is our low frequency characteristics in
depletion? It is 1/1+Cox+1/C s and not in d depletion. In fact, this is in general case
because we will you can solve it for all voltages.
So, in general this is the capacitance and I have replaced C depletion with C s because it
is a general expression. So, it is AC semiconductor and what is Cs? It is simply the
derivative of the displacement to the electric field. So, it is epsilon s electric field by d
phi s. That is basically your dQs/d∅𝑠 right. So, that is your Cs.
And since you have already solved for the electric field in Poisson’s equation, we have
solved for the electric field. We take the derivative with respect to ∅𝑠 ok. And you end
579
up with this expression here it is all again got the electric field term there. So, that is your
C s. And this is the value of C s that corresponds to this value of gate voltage.
So, phi s you know at the electric field, and you have this term here and what is this;
what is this particular term it is representing your V ox right. It is since your epsilon s in
the electric field in the semiconductor is equal to epsilon ox and the electric field in the
𝜀𝑠
oxide your epsilon oxide is written as . 𝜀𝑠 and 𝜀𝑜𝑥 . 𝑡𝑜𝑥 ox is basically your V ox. And
𝜀𝑜𝑥
therefore, this term represents V ox ok. So, that is V ox.
So, you have VG that corresponds to this. So, this Cs this capacitance corresponds to this
VG. So, using different values of phi s you can plot your C V characteristics in a very
accurate manner, you end up with the low frequency characteristics.
For high frequency characteristics what you do before, we go to high frequency for deep
depletion what do we do. For deep depletion this Poisson equation is not correct because
in deep depletion there is no formation of an inversion layer.
So, we solve a different Poisson equation. For deep depletion we solved this Poisson
equation which does not take into account the inversion layer and this will give you a
different electric field. So, that is the electric field I put a subscript of deep depletion
there. And that in turn will give you a different capacitance value. It is once again the
580
deep depletion capacitance is epsilon s electric field deep depletion d of this by d phi s.
So, that is the capacitance.
So, that will give you a different capacitance, but this capacitance corresponds to this
particular voltage. It is VFB ∅𝑠 𝑡𝑜𝑥 into the electric field you use here is the deep
depletion electric field. So, that is the difference ok. And if you plot this C, C effective
deep depletion versus VG you will get the deep depletion characteristics of your
capacitance voltage characteristics.
And how do we get high frequency for high frequency what you do is you use the same
capacitance as the deep depletion ok. Because it tells you that the electrons are not
present in the AC response, but for the DC part, because the DC is still varying very
slowly you make this electric capacitance correspond to a different gate voltage, and
what is the gate voltage you make it correspond to the low frequency gate voltage which
is VFB +∅𝑠 you are using the low frequency electric field ok. So, I hope that is clear.
So, just to summarize, for this is for low frequency you have Cs corresponding to this
gate voltage where this is the low frequency electric field obtained from this Poisson
equation. For deep depletion we solve a new Poisson equation which is permits the
inversion layer. And you find the new electric field you find a new capacitance and that
capacitance corresponds to this gate voltage which uses the deep depletion electric fields
not the low frequency electric field.
581
So, both the capacitance and the electric field correspond to deep depletion, but for high
frequency you use the deep depletion capacitance, but you use the make it correspond to
the low frequency gate voltage, which is obtained from the first Poisson equation. The
one we solved here. So of course, these are details that are not a part of your course, but
nevertheless I felt it is important to mention ok.
With this we come to a conclusion we conclude our MOS capacitors ok. We have looked
at the structure of the MOS capacitor the operation. We defined terms like the gate
voltage the flat band voltage the threshold voltage. You know what is accumulation
mode depletion mode and inversion mode. You know how the capacitance voltage
characteristic of the MOS capacitor varies in low frequency high frequency and in deep
depletion region. You know simple model as well as the detailed model for the
capacitance voltage characteristics ok.
So with that, we conclude the MOS capacitor discussions on the MOS capacitor.
582
Departments of Instrumentation and Applied Physics
Lecture -36
MOSFET: Introduction
The enhancement mode MOSFET looks like this. So, you are already familiar with this
bit which is the MOS capacitor ok. So, if we look at this cross section if you forget
everything else that is present beyond this, we forget all that what you have here is the
gate, insulator, semiconductor and the back contact and this was your MOS capacitor.
But, what the MOSFET does is it makes use of this device, but it adds some more
terminals ok.
So, now you have the gate which is provided a gate voltage VG. And then you have the
back contact which is got a voltage VB and this back contact is now called in the context
of the MOSFET it is now called the body or the bulk contact because, it connects to the
body of the semiconductor ok. So, it is the body or the bulk contact and then that is your
familiar gate contact that is connecting to the gate metal. You have your insulator and the
insulator could be an oxide and it is got some thickness t ox and it is got the permittivity
𝜖𝑜𝑥
epsilon ox and therefore, the insulator capacitance per unit area is .
𝑡𝑜𝑥
583
And then you have your semiconductor and let us stay with the example of a P doped
semiconductor. Now, in order to create your A and B that is the two terminals through
which the signal has to transfer, the signal has to transfer from A to B and in order to
have those two terminals you have your source and you have a drain ok.
So, these are two additional terminals in your, that are added to your MOS capacitor to
create your MOSFET. Now what does the source in drain have, now the first thing is that
the source and drain let us look at the way it is created ok. So, it is not just metal placed
on the semiconductor and we will see why it should not be. Essentially what we need is a
electrical contact and the current will actually transfer from source to drain or the carriers
will move from the source to drain or vice versa depending on the type of carriers and
depending on the potential.
And the first thing about the source and drain is normally they are absolutely
symmetrical. There is nothing to say as to there is nothing in the design or in the
fabrication to say that one of them is the source and the other is the drain. The source and
drain are purely defined based on the voltages that you apply, the bias voltages that you
apply ok. So, this is this structure is absolutely symmetry.
So, give it a MOSFET like this it is equally possible to you treat this terminal as your
source and that terminal as your drain it does not matter, but it all depends on the
voltages that you apply. So, we will just label that as your source and it is given a source
voltage VS and this is the drain and this is given a drain voltage VD. Now all these
potentials that is VB VD the gate are all taken with respect to the source. So, this is the
reference potential for your MOSFET structure. So, what is important in the MOSFET is
the gate to source voltage ok.
So, it is not just VG with respect to some arbitrary ground or some universal ground that
really you need to really extract out what is the potential seen or what are the fields seen
within the device and therefore, all the voltages are measured with respect to the source.
So, you have the gate to source voltage which is basically VG -VS, you have the body to
source voltage which is basically your VB -VS and you have the drain to source voltage
which is VD -VS ok, but for now let us just say that the source is at VS and the gate is at
VG the body is at VB and that drain is at VD. So, how are the source and drain terminals
created.
584
So, you have a metal contact for us to have interconnects on the outside, but then the
metal interfaces directly with an N doped semi conduct very highly N doped
semiconductor; I should probably put an N plus that ok. So, look at the structure you
have a P doped semiconductor as the basic body ok. So, that is you substrate in some
sense and then when you are going to create the metal contacts you have strongly N
doped regions and then you have your metal contact ok.
So, essentially you have created a diode here, you create a diode here, and you create a
diode here and then you have your metal contact ok. Now what is the purpose of this
source and drain? So, the idea of the MOSFET is to operate an inversion. So, the
MOSFET if you think of the MOSFET as an on and off device ok; so, when it is off let
us say the flat band voltage is 0. So, we are familiar with the MOS capacitor and the flat
band voltage. So, let us say the flat band voltage is 0 ok.
And theoretically when my VG is at the flat band voltage or generally we will speaking
when VG is less than VT there should be no electrons here and therefore, you are in
depletion mode operation or your in you are switched off ok. So, there is nothing there is
no way carriers from the source can actually move to that drain because this region is
completely depleted. But, when VG becomes greater than VT the MOSFET enters the
inversion region which means that the, which means that minority carriers start
appearing at the interface.
585
So, we have already looked at this in the in the discussion on the MOS capacitor. And
the idea is that if you have an N doped region not these minority carriers which are
electrons, not only appear from the not only appear from the thermally generated carriers
from the P doped bulk, but are also drawn in from the source side; provided that there is
a drain the source voltage. So, the source also contributes to these.
So, the essential idea so, this is the first role of the N plus doped source and drain which
is to act as a carrier source or the minority carrier source in their device. And, essentially
create this is electron N plus pathway or this conduit because everything around this is
doped is all depleted sorry it is a PN junction device therefore, these all depleted you
have a depletion region throughout here ok. So, all this is depleted, but when inversion
happens you have this entire region being depleted and you have an electrically
conductive path from this N doped region through these electrons and through that N
doped region into the drain. So, that is the way the MOSFET intends to operate ok. So,
that is one point of having a N plus region.
The other key purpose of these N plus dope, wide are we insisting on a heavily doped
drain and source contact ok. So, this is your drain source contact, the idea is we want to
inject the current right from this metal and into that metal which means that we want this
metal semiconductor junction to be ohmic ok. So, if you remember your metal
semiconductor junctions we could had two kinds of junctions; one is where we had a
586
short key barrier and the other is where we had a ohmic contact ok. So, which contact
would you like if you want to have current flowing through both ways, we would ideally
I like an ohmic contact.
So, by simply choosing a metal and by doping the semiconductor correctly we should
have an ohmic contact, but unfortunately that is not it is not that trivial. And, it is not that
trivial because of a lot of issues particularly because of the Fermi level gets pinned,
because of interfacial states etcetera et cetera which means there are trap states here;
electrons get trapped and electrons cannot move about very easily. So, it is very likely
that we will end up creating a short key barrier or a short key contact and this is not what
you need. So, how given the fact that you can do nothing other than create a short key
contact.
So, let us say we have this problem we can only create a short key contact and we want
to get electrons in and out of this from the metal to the end to the doped semiconductor
which is the source we want to get electrons in and out of this; how do we go about
doing that ok. So, let us look at the band bending of a metal semiconductor junction short
key junction.
587
So, let us say that is your band bending ok. Now what happens if I increase my dope
and concentration in the semiconductor? If you increase the dope and concentration in
the semiconductor the band bending will become a lot sharper, will be the bending will
be much more narrower right. All the bending will finish quite early the x d becomes
small, the bending becomes the bending is very the electric fields are quite large etcetera
etcetera.
On the other hand if you reduce the doped and concentration the bending is more relaxed
the x t is much larger the electric fields of weaker and so on. But, the advantage of
having a bending that is very sharp is that if you remember the different mechanisms of
charge transport through a metal semiconductor junction we had thermionic emission,
we had diffusion, but we also had tunneling through the barrier through this triangular
barrier we had this WKB approximation etcetera etcetera. So, the point of having this
heavily N plus doped region is to create a sharp triangular barrier through which
electrons can easily tunnel in and out ok.
And therefore, give you a very low resistance contact between the metal and the
semiconductor. So, that is the reason for having this N plus doped region apart from very
obvious reasons it is good to have a strongly doped strongly doped source and drain
contact. So, these two together that is this metal and this doped region together make
your source. So, essentially the potential here is the potential here. So, all this is short
588
circuited ok. And these two together make your drain ok. So, now let us look at some of
the geometrical aspects of the MOSFET architecture.
So, we have. So, this is the MOS structure. So, you have the back contact which is that
VB the source which is at VS the drain which is at VD the gate which is at VG they have
an insulator between the gate and the semiconductor and all potentials are taken with
respect to the source. So, what is important is VD’s VG’s and VB’s. So, what is the
geometrical aspects now this region between from where the source ends to where the
drain begins ok. So, in that sense; so, this region the length of this region is quite
important and it is called the channel length it is a technical term ok. So, when we said
channel length we are talking about this distance from the drain to the source that is the
length the electrons the carriers must travel if they want to go from the source to the
drain.
Now, the channel length is quite an important parameters when you hear about Moore’s
law and when you hear about scaling when you hear about the technology the MOSFETs
and the transistors becoming smaller and smaller it is the channel length that they are
trying to reduce. Because, if you reduce channel length the speed of the MOSFET is
expected to increase of course, they are reducing all the other parameters also which we
will look at when we when we talk about scaling. But, it is a channel length that if you
reduce the channel length you generally tend to improve the speed of the MOSFET. The
other parameter is as to how wide this devices right. So, you have the same cross section
through the page.
589
So, as you look inside the page you have a certain width if you look at the perspective
view the devices has got a certain width ok. So, that will be gate and so on. And so for
so, that will be your contact and that is your semiconductor etcetera. So, I am not drawn
it in perspective. So, you are looking at a cross section to the device has got a certain
width and that width is called as a channel width ok.
So, that is the width of the channel and that is given a symbol W. So, when we say W
and L we mean the channel width and the channel length and a special ratio which is
particularly useful when we talk about currents etcetera in the MOSFET is the ratio of
the channel width to the channel length ok.
590
So, this ratio so the ratio which is basically W/L right; W/L is the channel with of the
MOSFET divided by the channel length. And this ratio is particularly important in terms
in the context of device design or circuit design and it is given a special name it is called
as the aspect ratio of the device. So, when we say aspect ratio when we say the aspect
ratio is 1, it implies the channel width is equal to the channel length and we say aspect
ratios 2 it implies the width is twice that of the length and. So, on now that is the channel
length and channel width now apart from that another important parameter with regards
to frequency response in particular is the overlap length.
So, the source and the drain they extend below the gate. So, the gate can impact a certain
region of the source and the drain. So, if you look at this cross section you can see that
the gate metal is overlapping the source and the drain of course; there is an insulator in
between. So, there is there is no direct current unless there are leakage currents there is
no direct current from the gate to the source and drain or vice versa, but there will be a
field impact the any voltage in the gate we will try to influence are these regions at this
source and drain. And, this length of this overlap is something called as the overlap
length or the gauge to source overlap length and the gate to drain overlap length if these
two are symmetrical will just say both of them are L overlap ok.
So, this essentially becomes a little capacitor right. So, what is the capacitance of this
∈𝐴
capacitor. So, it is , but what is epsilon it is epsilon ox what is the area of cross section
𝐷
591
or what is the overlap cross section. It is L into W. So, L overlap into W is that is overlap
length into the width of the device it is the area of cross section divided by the thickness
of the oxide t ox.
So, that is the overlap capacitance which is the capacitance between the drain and the
source other a many other capacitances because we have to PN junction diodes. So,
therefore, there is a diffusion capacitance or a depletion capacitance etcetera etcetera, but
we are into we are talking about the capacitance of this device this structure here. So,
that is to do with the general architecture the design of the MOSFET there are many
other small features, but mean these are the basic points alright.
So, this structure is got what it is a going to P doped semiconductor and it is got N doped
contacts right strongly N doped contacts and the idea is when you have inversion you
will have electrons which are the minority carriers between the source and drain. And
you will have an electron channel going in from electron you will have a electrons being
the current carrying current carriers between source and drain. So, this device is called an
N channel enhancement mode MOSFET N channel means it is the electrons that are
going to form the channel. So, this regional is called as the channel ok. So, that is the
channel which is the conduit of charge transport. So, it is an N channel enhancement
mode MOSFET we will simply call it as NMOS device or N type MOSFET or N
channel MOSFET.
592
Similarly you could have the counterpart the complementary device which is the P
channel MOSFET and for the P channel MOSFET you have to have a MOS capacitor
with N doped semi conductor and you have P doped source and drain contacts.
And, the idea is any inversion you will have holes at the interface which are the minority
carriers and you will have holes being the carrier of the current we will have a hole
current going from the source to the drain or from the high potential to the low potential.
And, everything else remains the same you have an you have a channel width you have a
overlapped length and you have a channel length.
593
Now, there are different kinds of field effect transistors. So, I have taken this from an
online source and these are the circuit equivalents. So, if you want to represent them as a
circuit. So, these are the ways you would represent these transistors, but we are interested
in the enhancement mode MOSFET right. So, we are interested in these 6 you have any
of these options ok. So, you can use the any symbol you like, but if you are using let us
say you want to represent the body contact or the bulk contact and you want to set the
body a bulk contact it a different potential from that of the source.
So, normally we connect the body and the bulk to the source, but it can be may it can be
used in a more clever manner by creating a potential difference between the body and the
source. So, if you want to really use it that way you want probably using this symbol it
shows the gate, it shows the drain, it shows the source and it shows the body here ok. So,
that is the body terminal and here what it is telling you is that the body is connected to
the source, but what I will be using is typically this.
So, this is just as a force of by the force of habit I tend to represent end channel devices
like this and P channel MOSFETs with a circle there and that circle is probably comes
from the use of the knot logic or the inverse logic. So, that becomes a P MOS device that
is your NMOS device you cannot see the bulk contact in this ok. So, that could be your
drain source and gate and these are your two devices ok. And, this is yet another
594
representation for the PMOS and the NMOS; but let us just stick to one we will just stick
to this particular representation.
So, here we have the NMOS and here you have the PMOS ok. So, in the NMOS we have
the drain terminal, you have the source terminal and you have the gate terminal the body
contact is not shown here. Now, the gate is at potential VG with respect to some let us
say universal ground, the source is at a potential VS with respect to the same ground and
this is at a potential VD. Now, what is important with regards to the MOSFET are these
potentials because that is what determines the electric fields inside the device.
So, it is the gate to source potential which is VG minus VS ok. So, when we say VG s it
means it is VG -VS or automatically the gate with respect to the source. And when we say
VDS it means it is the drain minus the source potential so, that is VDS. Similarly for a
PMOS you have the source the drain and the gate and the reason I am relocating the
source, here is if you think of it from the point of view of a circuit ok. So, I am trying to
sort of establish that thread right now, which is typically for an NMOS you will find that
if this is the construct you will find that this is at a higher potential everything is with the
respect to the source.
So, this is at a potential that is greater than the source the gate is at a potential is greater
than the source. And, in the case of a PMOS you will find that if you want to continue
using the same equations without changing the variables from gate to source to source to
595
gate etcetera et cetera. You want to be so, I am sorry gate to source to gate to drain
etcetera et cetera. You want to sort of relocate these variables and in the manner where
the source is at a higher potential, the drain is at a lower potential and the gate is at a
potential that is greater than the drain ok. So, you will probably gather this intuition as
we move along, but nevertheless that is the reason why I am just relocating the source
here, it is basically the opposite kind of carriers ok.
So, it is the holes, it is the holes which are carrying current and the hole current is
moving from the higher potential to the lower potential. In this case it is the electrons
that are carrying current. And, the electron current is moving from the lower potential to
the higher potential. So, in both cases the source is sourcing the carriers and the drain is
accepting the carriers. So, if you want to think of it that way then that is the another way
to look at this ok.
So, that is the source this is a drain and that is the gate and what we are interested in is in
the gate to source voltage and the drain to source voltage. So, if you recollect if I tell you
that the source is at a higher potential than the gate and drain, you will know that the VGS
and VDS are negative in value ok. But, these are the important voltages and these are the
circuit symbols we will be using for the NMOS and PMOS; the big difference in the
PMOS is this is little circle and you see at the input ok. So, that is the difference.
596
Department of Instrumentation and Applied Physics.
Lecture - 37
MOSFET: I-V characteristics
The now let us go ahead and develop the current voltage characteristics. The
electrostatics of the MOSFET, it has already been in some sense covered when we
discussed the MOS capacitor ok. So, what we will do now is, look at the current voltage
characteristics of the MOSFET. Of course, the electrostatics in the MOS capacitor the
electrostatics did not include the influence of the source and drain, but that is quite easy
to include, and it is just a matter of changing the potential of the semiconductor. What
we will look at now is, once you have created the inversion channel what is the current,
how what is the current voltage characteristics in particular, what current am I looking
for? We are not interested in the gate current, because ideally there should be no gate
current ok.
So, we are not interested in the gate current, IG should be 0 and that is the whole point of
the MOSFET or the field effect device. If IG is 0 it is, because you have an insulator than
the therefore, the input impedance is infinite ok. So, ideally that is what we are trying to
achieve. We are trying to achieve a variable resistor with an infinite input impedance
597
variable channel resistance, but the gate input impedance is infinite ok. So, that is the
ideal drain. Now what we are trying to look at is this current ok. What is the drain to
source current?.
So, if you talking about P channel device, if you talking about the N channel device or an
NMOS you have a P type semiconductor and you have electrons there and the electrons
and the drain is going to be at a higher potential.
598
So, let us say at 2 volts, the source is with respect to the source and therefore, the electric
field is in this direction and therefore, the electrons will all come in from the source and
they will all migrate to the drain which means the current is going to be in this direction.
You going to have I from the drain to the source and we want to find out what is the
drain to source current all right. So, how do we go about this? And before we continue,
this is equal true for an N channel device ok. So, I for a P channel device. So, if it is a P
channel device you will have if it is a PMOS, you will have your n type semiconductor
and you will have holes of the interface after inversion and these holes that that the
source will be at a higher potential with respect to the drain.
So, for example, let us say look at this is being 2 volts and that is ground. So, the holes
will be entering from the source they will migrate to the drain and the direction of the
movement of the holes will be this and the current direction will be this ok. So, you will
have an ids that is negative VDS that is negative etcetera et cetera ok, but the idea of the
physics behind deriving the current voltage characteristics is the same. Now since we
have discussed the P type body and since we have discussed the P type MOS capacitor.
599
Let us just take with the P type semiconductor and then N MOS device or an electron
channel being formed at the interface during inversion.
So, how does this device work? Now when VGS is not 0 when VGS is equal to is less than
VT, there is no channel technically because its solved depleted. But once VGS becomes
greater than the threshold voltage, you now have inversion and you have a channel that
forms. So, the channel it is indicated in this picture and it is indicated as this blue
regions; that is the electron channel from source to drain. If we have IA take a voltmeter
and head from the source to the drain or for the, from the source to the drain and I were
to measure the potential of the channel at every point ok. What is the potential?
600
So, let us call that potential as VC which is the channel potential. So, let us give it a name
for this particular potential which is VC ok, which is the channel potential which means
we have created N channel, we have created an electron interface ok. So, you have
electrons here, but then these electrons, we are going to go about go around with the
voltmeter and measure the potential of this channel of electrons from the source to the
drain and we are going to measure that potential with respect to the source and we are
going to call that potential VC. What is that potential?
So, right here so, you can think of this whole path of electrons as a long resistor right. It
is got some resistance. So, right at this edge what is the potential? It is 0 it is the same as
the source ok. The potential here is VS and this point is V s. So, V c is basically the
potential that you measure minus is all with respect to the source and therefore, the V c is
0 ok. So, let us give this a direction. So, we are calling this the y direction and therefore,
at y equal to 0 V c is 0.
601
Now let us move ahead, now at some other point the potential difference between these
two will be V c of y because its varying with y and as you go to the drain as you
approach the drain side the potential here is going to be the same as the drain potential
and therefore, at y equal to L which is the channel length V c is going to take a value of
VDS which is the drain to source voltage because all these potentials are measured with
respect to the source.
So, you are going to keep one probe there another probe here and this potential
difference is VDS all right. So, that is your channel potential. So, in order to imagine in
order to derive this current, we see that it is the VDS that is providing the electric field
and these are in response to the electric field these electrons are going to drift through
from the source to the drain
602
So, what is the drift current density? So, that is J which is the drift current density the
current per unit area it is qnVD where VD is the drift velocity of the electrons, n is the
number of electrons per unit volume and q is basically your 1.6*10-19 coulombs.
So, that is your drift current density. Now what is the total current? It is the current
density times the area of cross section and you want to identify the area of cross section
what we are going to do is we are going to define this electron channel as having some
thickness h ok. Now that is that is an odd definition because all of you are already
familiar the fact that the electrons are all wave particles. So, do you define them as
having a certain thickness h is odd, but were going to go ahead and treat these electrons
as particles and define the sheet as having a thickness h and this sheet of electrons has
got a width which is equal to the channel width of your device ok. You have got you
have got this sheet charge which is all migrating towards the, from the source to the,
from the source to the drain.
603
So, all migrating from the source to the drain; so, what is the total current? The total
current which is your IDS is basically this current density into the area of cross section
and what is the area of cross section? It is basically W*h where W is the channel width
and h is your thickness of this electron sheet ok.
So, we can write this little equation down, which is ids by W h is the current density and
that is equal to qn into the drift velocity, but what is drift velocity? The drift velocity is
nothing, but the mobility times the electric field. So, we can rewrite that equation as qn
604
mu into the electric field where mu is the mobility of the electrons and furthermore what
is the electric field?
So, if I were to now take a little section we already know what the channel potential is
right. So, if I were to take a little section, let us take a small section here. So, that is our y
direction and let us say the section has got a length dy and we say that the voltage drop
across the section is basically dVc that is a delta of the channel potential.
So, you have a section of length dy and the voltage drop across the section is basically
dVc. So, what is the electric field across the section it is dVc /dy. So, we replace this
electric field term by dVc/dy.
605
So, you have qnµdVc /dy as your current density and that is equal to your IDS /Wh ok.
Now in order to identify IDS we need to now calculate what is n ok.
So, how do we find out what is n? n is the number of electrons per unit volume sitting
near the channel. So, if you recollect your MOS capacitor we had already identified. So,
if you remember the MOS capacitor let us just go back a little bit.
606
So, we had the MOS capacitor which was a metal insulator semi conductor structure and
we had in inversion we had an electron sheet and we had defined that inversion charge is
Qinv that was a inversion charge per unit area.
So, the MOS capacitor had a gate voltage V G it had and this also with respect to this
grounded back contact and it had it had generated this inversion layer and if you go back
you will recollect that this inversion layer was basically qN where N was the number of
electrons per unit area which was nothing, but Cox (VG-V T) where VT was the threshold
voltage. So, that is the way we defined that the number of electrons per unit area because
Cox is per unit area. Now what we want now is the number of electrons per unit volume
which is given by small n. So, if this is the number of electrons per unit area, what is the
number of electrons per unit volume? It is this number divided by the thickness of the
channel which is are h.
So, this is the number of electrons per unit volume at the interface provided the potential
in the semiconductor or this channel potential is 0 because in the case of a MOS
capacitor what is the difference, what is the voltage difference between the metal and the
semiconductor? It is VG minus 0, but now if I were to take this little section here of
length dy. So, let us say that is my insulator and that is my semiconductor the potential
here is Vc of y and the potential drop across the section is dVc and now my gate metal ok.
607
So, let us let us look at the section that we are analyzing here. So, we have, we have to
consider this little section you have taken the section all the way through let us cut this
little section all the way through you have the gate, you have the insulator, you have the
semiconductor, the length of the section is dy the potential drop.
So, you have the channel here, you have the channel here the potential drop across this
channel is dVc and therefore, the electric field in that section from here to there is dVc by
dy ok. dVc let me just write it clearly dVc /dy that is the electric field and this potential of
608
the channel is some Vc of y, because you know its sitting depending on where the y is
how far are you from the source depending on what your y is. So, that is that is the
source and this distance is y right. So, that is the situation. So, now, if you think of it as
MOS capacitor it is a MOS capacitor, but this in the semiconductor is not at ground, but
it is some Vc and therefore, the gate to semiconductor voltage difference is not V G, but
it is VG - VC.
So, that difference is so just to tidy up this place a little this difference is VG - VC of y
that is the gate to semiconductor. So, what I am going to do is we are going to replace
this little term here instead of saying V G minus ground we are going to say its (VG -VC -
VT )*Cox/h is your q*f. So, we now have this equation here it says that q*N where N is
the number of electrons per unit volume is Cox *(VG -VT -VC) of y/h and what we are
going to do is when are now going to substitute this qn in this expression for the current.
So, we now have Ids we have taken the Wh over to the other side is qnµdVc /dVy. So,
that is the current density into Wh that is the area and qn is nothing, but Cox* V gs minus
(Vt -Vc )/h ok. So, it is now the potential with respect to the source into µdVc/ dy *Wh
and see the h cancels off the h was not required. So, even though our assumption on what
h is and how we can how we treat h is all was all arbitrary we really do not need it, it just
served to explain the situation and therefore, you have this little differential equation that
needs to be solved.
609
You have Ids is equal to this particular term. So, you have mu WCox(Vgs–Vt-Vc) in a
dVc/dVy dy. So, you take dy to the other side and you solve this differential equation. So,
you integrate y from 0 to L. This does not is not clear, but this is L. So, as you integrate
from 0 to L what are we doing we are moving from the source to the drain and therefore,
Vc will vary from 0 to Vds
So, if when we did our experiment by moving the voltmeter along the channel we saw
that as we moved from the source to the drain L varied x y varied from 0 to L and the
610
channel potential varied from 0 to Vds. So, you apply these limits and solve this equation
and you will end up with your current voltage characteristics of the MOSFET ok
So, that is the derivation of the simplest model for the current voltage characteristics of
the MOSFET, but there are problems there are a lot of problems in this ok. So, many
assumptions made, many implicit assumptions made. Now what is the first assumption
that we have made here? The first assumption is the fact that we have actually used this
expression for drift velocity. When is this expression valid? This expression is valid only
under low electric fields.
So, that is the first assumption you have assumed low electric fields, which implies low
Vds or large L, but I tell you why it implies low Vds in this case. This expression is valid
for low V ds because we have made another assumption we have said that this channel
thickness remains h throughout from source to drain I mean irrespective of Vc varying a
lot ok. Vc is varying from 0 to V ds. We have said that the channel thickness remains h.
Normally as we approach as V ds becomes larger and larger as Vc becomes larger and
larger, you see qn becomes smaller and smaller ok. So, therefore, this channel should
actually taper off as we head towards V ds
611
So, we have made this implicit assumption of low Vds and we have assumed we have
derived the current voltage characteristics for that condition ok. Now so this region of
operation is something called as linear mode of operation ok. So, it is a technical term.
So, when we say the MOSFET is operating in linear mode. What it implies is the Vds is
very low and in particular Vds is less than Vgs -VT and I will tell you why.
612
So, why is this, a linear model clearly? This is non-linear right you have a (Vds)2/2, but
if the Vds is very very small the square compared to Vgs is the V ds is much smaller then
Vgs-VT then this term can be ignored, and you have only this particular term. So, what
you are saying is that you have this device and you have a V ds which is a voltage across
this device and you have a current that is more or less proportional to V ds which means
that this device despite it being non-linear at low V ds it is behaving like a resistor you
see that is the conductance of this resistor. So, you have Ids is equal to the conductance
into V ds or inverse of this term is the resistance of the resistor. So, that is why we say, it
is the linear mode of operation ok. Now what happens if we start increasing Vds. So, this
model is great, but you know it is not really general because we do not know what
happens when you start increasing Vds ok.
613
So, when you start increasing V ds the channel or the thickness the so called thickness of
the channel will not be constant. My media why is it not constant? So, let us look at two
MOS capacitors, let us take a section here and let us take a section here ok. So, let us
take so that is that is a very wide section. So, let us take a little thin section here and let
us take a thin section here. So, what is the MOS capacitor in this region first? So, that is
what we used that is the one we used for all the analysis you have a gate insulator semi
conductor. What is the channel potential or what is the potential? Here it is some V c and
the V c is approximately 0.
And we know that Vg -Vc is greater than VT or in other words let us take let us say Vgs-
Vc is greater than VT or in other words Vgs -VT is greater than Vc or effectively this gate
voltage is strong enough to overcome the channel potential. So, this is large enough and
therefore, it is strong enough to pull electrons near the interface, but what happens if my
channel potential becomes larger then Vgs -VT .
614
So, let us take a situation where you have gate insulator semiconductor, but the channel
potential is very large. Again when can this happen? If this can happen when the V ds is
very large the channel potential is very large and now you look at the capacitor you have
Vg what is the voltage difference between the gate.
And the semiconductor its Vg-Vc now Vg-Vc is negative or its in its let us say 0 or in fact,
we should be looking at the potential above V T because we have to cross V T to get
this.
615
So, when you have a condition of V gs minus V T being less than the channel potential
you cannot have an inversion layer there are no electrons there and therefore, as we start
heading towards the drain the channel should actually taper off and the moment V gs
becomes Vgs -VT becomes equal to V sds that is when that become when the channel
potential heads becomes equal to the channel potentials sorry becomes equal to the
channel potential this thing should there should be no more electrons and the channel
would simply taper off, it would taper off ok.
616
You will not see electrons the anymore and this point where the channel tapers off is
called as the pinch off point ok, it is a technical term, it is called pinch off ok. So, I do
not know if I was clear enough. So, I am going to repeat this concept. So, the point
where so we have a channel you have got a drain, you have got a source and you have
got a channel potential that is varying from source to drain. Now the drain voltage is very
high, there is a possibility the channel potential is very large particularly near the drain
side that is the channel potential is close to 0 near the source side. So, you will have V gs
being greater than Vgs -VT being greater than Vc near the source side.
So, you will have electrons there, but there will be a point where the V c becomes equal
to Vgs -Vt.
617
So, let us say that happens at some location where V c becomes equal to Vgs -VT and as
you move between that location and the drain the Vc is always going to be greater than
Vgs-VT. So, throughout that region you cannot have an inversion and the channel is said
to be pinched off. So, you have electrons here, but suddenly beyond this pinch off point
beyond this point, there are no electrons because inversion has not happened. Now the let
us do this experiment its thought experiment, let us say we are running the transistor in
linear operation, in linear mode of operation; that means, the V ds was very low.
618
And we now start tweaking up the Vds, we start increasing the drain to source voltage the
drain to source voltage at which. So, as you are increase the drain to source the channel
potential is increasing. So, let us watch the channel potential near the drain, let us only
focus on this section here.
So, this channel potential is increasing as you increase the drain to source voltage. This is
in fact equal to VdVds right. So, the drain voltage at which the channel potential
becomes equal to Vgs-VT or you know that particular drain voltage is given a special
symbol it is called the Vdsat and I will tell you what that is at that particular drain voltage
the channel is just pinched off near the drain end ok.
There are electrons every other region is inverted, but just this end near the drain is you
do not find electrons because the Vgs-VT is now no longer greater than the channel
potential. Your Vgs-VT minus channel potential has become equal to 0 and therefore, qn
has become equal to 0. So, this end is now pinched off.
And we say that the transistor has entered has gone away from linear mode of operation
because now our model is no longer valid. It is going away from linear mode operation
and it has entered something called as saturation mode operation.
619
And now if we keep increasing V ds further the transistor we will continue to remain in
saturation mode operation. Now in saturation mode operation so what saturates we find
that the current through the transistor saturates and I will explain what happens.
So, you have you think of this channel as a resistor this entire region between source and
drain as a resistor. So, let us say we have hit the pinch off ok. So, the drain is sitting
there, the source is sitting here and there were electrons all the way till this point, but
then the channel potential at this point is now greater than V gs minus V T. And
therefore, this region is all pinched off, all this is the pinch off point and there is no
electrons from here all the way till the drain. So, let us imagine this to be a large resistor
between source and drain.
So, how would the resist what would the resistance look like? You have a lot of electrons
between from the source to the pinch off point I am sorry I have run out of space there is
a lot of explanation here, but there is not much space for me to write anything. So, do
bear with me. So, let us say you have this as a resistor you have electrons all the way and
then you hit the pinch off point and beyond this you have a large resistor because you do
not have an electron channel anymore. So, now, when you have a drain to source
voltage, most of the voltage drops across the large resistor in the region that is actually
depleted, all those region is not depleted the region is beyond the pinch off point is all
depleted right. So, you have a large resistor here and you have a much smaller resistor
620
here. So, you have most of the voltage drop across this region. So, what happens to the
electron current? You have a large number of electrons.
So, qn is good; it is larger and then qn becomes 0 beyond this point. So, you have a large
number of electrons, look at the electron traffic, you have a large number of electrons
moving at low electric field all the way till the pinch off point and then after that the
electric field is very large, but there are not many electrons. So, you have a few electrons
that are shooting fast that are racing fast this depleted region that are racing fast this
region because of the high electric field. So, you have fast electrons and you have slow,
but larger number of electrons ok. So, that is what is going on.
621
So, you have slow electrons, but a large number of them. So, which means that the entire
flux has got low velocity, but large number beyond this the number becomes very small,
but the velocity becomes very large. So, never the less the current is maintained, but the
current does not increase very significantly if you increase V ds because it is not it is a
non-linear device in some sense ok. So, the V ds does not impact current beyond the
pinch off operation and therefore, the current has increased all the way in the linear mode
in the linear region the current kept increasing, but then after that the V ds stopped
impacting the current.
So, we say that the current has saturated, the current is not going to be impact impacted
by the Vds anymore and what is the equation for saturation we simply say that the
moment this point is pinched off the current is the same as the last value of the linear
current ok. So, which is what which is when Vds becomes equal to Vgs-VT, but in other
words when the channel potential at this location becomes equal to Vds V gs -V T then the
device current becomes constant it stalls it does not increase anymore with me. So, what
we do is we take this equation, which is the linear equation and we substitute it in order
to find out what this current is we just substitute we look at the last point on the linear
region we substitute Vds is equal to Vg-Vgs-VT and we use that as the current for all
values of V ds which are greater than V gs minus V T. So, we just say when V ds is V gs
minus V T what happens the current becomes (V gs -V T )2/ 2.
622
Lecture – 38
MOSFET: I-V characteristics – Continued
Ok so therefore, we can now write the current voltage characteristics of the MOSFET for
2 conditions, when Vds is less than VGS-VT which is the linear operation even though this
equation is non-linear it is it is linear when Vds is very, very small because you can
ignore this term. And when V ds is greater than or equal to V gs minus V T we have
saturation operation, and this becomes the current voltage characteristics you see that this
current does not depend on Vds it is independent of Vds.
So, if you were to plot these 2 together you would end up with a plot like this you will
find that if you plot VDS versus IDS you see this is a parabola. In fact, initially in the early
regions it is linear, but in fact, a parabola right so it is something that I will do this, but
just when Vds is equal to (Vgs -VT) the current saturates. And it does not depend on V ds
anymore ok.
So, that is what these 2 equations telling you? So, these 2 equations are a good enough
first order jump as to what the current voltage characteristics in a MOSFET are, but then
623
we need one more detail ok, which is does Vds really not impact the saturation mode
current ok.
So you think about it so let us take the let us say we have Vd, we have just hit the V DS
let us say V DS is equal to (VGS –VT). At this point the channel only this channel potential
here, only that region has got the channel potential being equal to V GS minus V T.
And therefore my q n which is Cox* (Vgs -VT-VC) is now 0 right. It is only that point
which has got 0 electrons, everywhere else the channel potential is less than (Vgs -VT)
because the maximum value it could take is the VDS equal to (Vgs -VT). But let us say we
increase VDS further, and we keep increasing VDS you keep increasing the drain to source
voltage.
624
So, the drain to source voltage is now great much greater than V GS minus VT which
means that if you look at the channel potential if you were to probe V C there is a point
here where V C is already become equal to V GS minus V T. And beyond this V C is
greater than (VGS-VT), and at this point my V C is equal to V DS which is of course,
greater than (VGS-VT).
So, it is the pinch off point now relocates to a different region and as I keep increasing V
DS, the pinch off point was start drifting towards the source side right. So, therefore, we
are actually altering this resistor combination. So, if I told you there is a resistor beyond
pinch off in the resistor below pinch off by actually moving the pinch off point I am
changing the values of these resistances. So, technically speaking my V DS must
influence my saturation mode current, but that is not what the equations told us ok.
So, what we are going to do is you are going to add an empirical correction to the current
voltage characteristics. Now this fact that the VDS that when a transistor is in saturation
the fact that the VDS can actually influence the current voltage characteristics is
something called as channel length modulation.
625
Which means that the effective channel length is being the behaviour of the channel is
being modulated by the VDS, even though the device is in saturation. Now this is a very
important phenomena because it directly impacts your circuit design particularly if you
want to build amplifiers channel length modulation plays an important role in
determining the gain etcetera.
Now, from the point of view of a circuit designer this channel length modulation a
MOSFET has got a equivalent equivalence. If you remember in the case of a BJT we had
the Base Width Modulation. And the base width modulation was you know you had your
BJT it was operating an active mode operation, and in active mode operation your V CE
was not supposed to influence the current a lot, but even then we saw that the V C does
influence the current because the width of the base changes because the depletion
thickness is changed.
And we saw that the current versus VCE did have a small trend even in active mode
operation. It is exactly the same situation here, the physics is different completely
different, but from the point of view of a circuit designer these 2 have the same impact
ok. They determine the output impedance of your MOSFET or of the transistor BJT or
MOSFET.
626
So, the channel length modulation impacts D, it makes the V DS influence the drain
current. Not by too much by a little, but by a little, but it is important enough that it
impacts the design of your amplifiers.
So now given these 2 equations ok so these are the current voltage characteristics of
MOSFET the first order; given these 2 equations what are the plots looks like?
627
So, we have 2 plots we have the drain to source current versus V GS, which is something
called as the transfer characteristics of the MOSFET. And we have the drain to source
current versus V DS which is the output characteristics of the MOSFET.
So, what we do, we take the MOSFET. So, let us say I want to plot the transfer
characteristics, we have the source, we have the drain, we have the gate the transfer
characteristic I DS versus V GS. We set a value for the V DS we set this and then we
sweep the V G. And we measure the current and then you change the V DS to a different
value, and again sweep the V G and measure the current. So, let us take this particular V
DS so here the V DS is let us say 1 volt. And now you sweep the gate voltage and as you
sweep the gate voltage you find that the current increases. Now shift the V DS to 2 volts
and now as you sweep the gate voltage you get a new curve and so on.
So, you develop that transfer characteristics of their MOSFET. And you can go back
here and you can actually see whether you can fit these models into the transfer
characteristics. Now for the output characteristics of the MOSFET what do we do we
take the MOSFET.
We fix our V GS we say that V GS is fixed we fix this voltage and we sweep the V 2 and
you measure the current through the device.
628
So, here let us take a particular value of V GS, and now we sweep the V DS and you
trace this particular curve. And then you take another value of V GS and sweep the V DS
and you get another characteristics. Now this particular characteristics both of them tell
you a lot of things, but let us just look at this particular cap the output characteristics are
little.
When V DS is very small you can see that the I DS and V DS relation are almost linear. So,
that is the linear here the MOSFET behaves like a resistor, but you can keep increasing.
So, we have picked a certain value of V GS isn’t it? So, let us say V DS keeps increasing
and let us say it reaches a point where the V DS is equal to V GS minus V T. So, let us
say that is that point that corresponds to this x coordinate. So, till here the MOSFET is in
the linear mode of operation and beyond this point the MOSFET enters a saturation
mode of operation.
So, you can see that in saturation the V DS does not change the current a lot. The current
is more or less constant, but still because of channel length modulation there is some
small increase in the current. So, if you look at it the current does increase a little with
increasing V DS, but not by too much ok. And that is sort of empirically added in by this
little linear relation.
Where this lambda is something called as the channel length modulation parameter. Now
if somebody were to ask you as to what is the resistance of the MOSFET or the small
629
signal resistance of the MOSFET, what is the answer ok. So, think let us think of the
MOSFET let us take the MOSFET and what we want to know is what is this resistance
the small signal resistance here.
So, in other words we want to find out what is d V DS /d I DS, which is if I have a small
signal fluctuation in the drain to source voltage, how does the current in the MOSFET
fluctuate ok. So now, this strongly depends on whether the MOSFET is in linear or in
saturation. So, let us say that the MOSFET is in linear mode operation.
So, what is dVDS/dIDS for a MOSFET in linear. So, if the MOSFET is in deep linear let
us ignore this term because V DS is let us say V DS is much much smaller than V GS
minus V T so we can just ignore this term. So, what we are asking is; what is for a
MOSFET in linear what is d V DS by mu W over L so, d of ids which is mu W/L Cox
(VGS-VT) VDS.
So, you just take this derivative and take the inverse. So, you take you calculate dIDS by
d V DS and take the inverse of that, and you will end up with the resistance being small
signal resistance; of the MOSFET in linear in linear mode operation being 1/Cox W/L
into VGS minus VT. So, that is the small signal resistance of the MOSFET which is
operating in linear mode operation.
630
So, if you increase the mobility of the carriers as the resistance go up or down it
definitely it makes a lot of sense that it goes down. If you improve the field effect by
increasing Cox the resistance of the MOSFET goes down. If you make the channel width
much larger than the channel length and increase the aspect ratio of the MOSFET of
course, the distance travelled by the electrons from the source to the drain decreases and
the sheet charge area increases and therefore, the channel length resistance will come
down, if you increase the aspect ratio.
And if you increase the gate over drive voltage which is the gate to source voltage minus
VT if you increase that definitely you have created more inversion more electrons and
therefore, the resistance comes down. So, this is the resistance of the MOSFET in linear
operation. Now what about saturation? So, what is dVDS/dIDS in saturation.
So, let us say the MOSFET does not have channel length modulation let us ignore that
ok. So, you have this, now since this current does not depend on so let us write the linear
equation first let us just keep it aside. So, that is 1 by mu C ox W over L into V GS
minus VT that is the linear small signal ok. Now for saturation they are using a saturation
mode current, but we are not going to we assume that it is an ideal MOSFET and there is
no channel length modulation.
So, what is this since there is no dependence on VDS in the current this is infinite ok.
Because dIDS/dVDS now becomes 0 in the current does not vary with the voltage it
631
becomes 0. We can vary this voltage as much as you want in saturation ok. So, let us say
VDS increases in saturation is greater than (VGS-VT) and you increase or decreases
voltage as much as you want in this region. The current in this will never change ok, it
does not depend if there is no channel on modulation.
So, the MOSFET behaves like constant current source, you can change the voltage
across it as much as you want and it behaves like a constant current source, and in a
constant current source the output impedance is infinite and the MOSFET is telling you
that it is got an infinite output impedance, but if you do include channel length
modulation ok. So, if you do include this then this is no longer infinite you have a V DS
term and therefore, you have a 𝜆 into this particular parameter appearing in as the
channel resistance ok.
But ideally speaking the small signal resistance of the MOSFET in saturation is very
high and ideally infinite ok.
So, if you look at these characteristics here if you take the slope you find that there is a
resistance it is a finite resistance, but here you see that there is no change in current
irrespective of what the change in V DS is and the resistance is infinite, the small signal
resistance is infinite ok. We are not talking about what is VDS/IDS because you do have a
large enough I DS. And that is I am we are not talking about the large signal resistor we
632
are talking about the small signal resistance, which is if there is a fluctuation here then
what is the fluctuation in the current an ideally it should be 0.
Now what are the transfer characteristics tell you; the transfer characteristics tell you that
till a certain point there is no increase in current so we could define a threshold voltage
somewhere.
So, if you take a very low V DS if V DS is much smaller than any V GS that you are
applying, then that characteristics is almost linear which is what is expected it is almost
633
like a straight line. And if you were to project the straight line all the way back you
would intercept the x axis at the point of VT that will be the threshold voltage.
So, we will have a small module on parameter extraction where we can show how to use
these characteristics to extract the parameters, but just to look at that just look at that
point again. Let us take the linear mode operation of the MOSFET right so here is the
MOSFET in linear model.
So, you have I DS is we will say it is deep linear where VDS is much smaller than V GS.
So, you have mu 𝜇lCox.(VGS-VT)VDS. So now, we are plotting the transfer characteristics,
how does IDS depend on V GS; it is a straight line. As below VT as long as V GS is below
V T the current is 0 we will develop finer models which will tell us exactly what the
current should be, but for now it is 0 because there is no inversion layer and after V T the
current should linearly increase. And what is the slope of this the slope is simply
𝜇WLCoxVDS.
So, that is the slope of the current voltage characteristics current the transfer
characteristics when the MOSFET is in linear mode operation. And what is the x
intercept of this characteristics; it will be V T. So, you can work out and see what the x
intercept of this characteristics is. So, this becomes a very useful method to extract both
the mobility as well as the threshold voltage. So, it becomes a very powerful tool and
you want to extract the parameters of a MOSFET.
634
635
Lecture – 38
MOSFET: I-V characteristics – Continued
Ok so therefore, we can now write the current voltage characteristics of the MOSFET for
2 conditions, when VDS < (VGS -VT) which is the linear operation even though this
equation is non-linear it is it is linear when VDS is very, very small because you can
ignore this term. And when V ds is greater than or equal to (VGS -VT) we have saturation
operation, and this becomes the current voltage characteristics you see that this current
does not depend on V DS it is independent of V DS.
So, if you were to plot these 2 together you would end up with a plot like this you will
find that if you plot V DS versus I DS you see this is a parabola. In fact, initially in the
early regions it is linear, but in fact, a parabola right so it is something that I will do this,
but just when V DS= (VGS -VT) the current saturates. And it does not depend on V ds
anymore ok.
So, that is what these 2 equations telling you? So, these 2 equations are a good enough
first order jump as to what the current voltage characteristics in a MOSFET are, but then
636
we need one more detail ok, which is does VDS really not impact the saturation mode
current ok.
So you think about it so let us take the let us say we have Vd, we have just hit the V DS
let us say V DS is equal to (VGS -VT) . At this point the channel only this channel
potential here, only that region has got the channel potential being equal to (VGS -VT)
And therefore my qn which is Cox(VGS -VT-VC) is now 0 right. It is only that point which
has got 0 electrons, everywhere else the channel potential is less than V GS minus V T
because the maximum value it could take is the VDS =(VGS-VT). But let us say we
increase V DS further, and we keep increasing V DS you keep increasing the drain to
source voltage.
637
So, the drain to source voltage is now great much greater than (VGS -VT) which means
that if you look at the channel potential if you were to probe VC there is a point here
where VC is already become equal to (VGS -VT) . And beyond this VC is greater than
(VGS -VT) , and at this point my V C =VDS which is of course, greater than (VGS -VT).
So, it is the pinch off point now relocates to a different region and as I keep increasing V
DS, the pinch off point was start drifting towards the source side right. So, therefore, we
are actually altering this resistor combination. So, if I told you there is a resistor beyond
pinch off in the resistor below pinch off by actually moving the pinch off point I am
changing the values of these resistances. So, technically speaking my VDS must influence
my saturation mode current, but that is not what the equations told us ok.
So, what we are going to do is you are going to add an empirical correction to the current
voltage characteristics. Now this fact that the V DS that when a transistor is in saturation
the fact that the VDS can actually influence the current voltage characteristics is
something called as channel length modulation.
638
Which means that the effective channel length is being the behaviour of the channel is
being modulated by the V DS, even though the device is in saturation. Now this is a very
important phenomena because it directly impacts your circuit design particularly if you
want to build amplifiers channel length modulation plays an important role in
determining the gain etcetera.
Now, from the point of view of a circuit designer this channel length modulation a
MOSFET has got a equivalent equivalence. If you remember in the case of a BJT we had
the Base Width Modulation. And the base width modulation was you know you had your
BJT it was operating an active mode operation, and in active mode operation your V CE
was not supposed to influence the current a lot, but even then we saw that the V C does
influence the current because the width of the base changes because the depletion
thickness is changed.
And we saw that the current versus VCE did have a small trend even in active mode
operation. It is exactly the same situation here, the physics is different completely
different, but from the point of view of a circuit designer these 2 have the same impact
ok. They determine the output impedance of your MOSFET or of the transistor BJT or
MOSFET.
639
So, the channel length modulation impacts D, it makes the V DS influence the drain
current. Not by too much by a little, but by a little, but it is important enough that it
impacts the design of your amplifiers.
So now given these 2 equations ok so these are the current voltage characteristics of
MOSFET the first order; given these 2 equations what are the plots looks like?
640
So, we have 2 plots we have the drain to source current versus V GS, which is something
called as the transfer characteristics of the MOSFET. And we have the drain to source
current versus V DS which is the output characteristics of the MOSFET.
So, what we do, we take the MOSFET. So, let us say I want to plot the transfer
characteristics, we have the source, we have the drain, we have the gate the transfer
characteristic I DS versus V GS. We set a value for the V DS we set this and then we
sweep the V G. And we measure the current and then you change the V DS to a different
value, and again sweep the V G and measure the current. So, let us take this particular V
DS so here the V DS is let us say 1 volt. And now you sweep the gate voltage and as you
sweep the gate voltage you find that the current increases. Now shift the V DS to 2 volts
and now as you sweep the gate voltage you get a new curve and so on.
So, you develop that transfer characteristics of their MOSFET. And you can go back
here and you can actually see whether you can fit these models into the transfer
characteristics. Now for the output characteristics of the MOSFET what do we do we
take the MOSFET.
We fix our V GS we say that V GS is fixed we fix this voltage and we sweep the V DS and
you measure the current through the device.
641
So, here let us take a particular value of V GS, and now we sweep the V DS and you trace
this particular curve. And then you take another value of V GS and sweep the V DS and
you get another characteristics. Now this particular characteristics both of them tell you a
lot of things, but let us just look at this particular cap the output characteristics are little.
When V DS is very small you can see that the I DS and V DS relation are almost linear.
So, that is the linear here the MOSFET behaves like a resistor, but you can keep
increasing. So, we have picked a certain value of V GS isn’t it? So, let us say V DS keeps
increasing and let us say it reaches a point where the V DS is equal to (VGS -VT). So, let
us say that is that point that corresponds to this x coordinate. So, till here the MOSFET is
in the linear mode of operation and beyond this point the MOSFET enters a saturation
mode of operation.
So, you can see that in saturation the V DS does not change the current a lot. The current
is more or less constant, but still because of channel length modulation there is some
small increase in the current. So, if you look at it the current does increase a little with
increasing V DS, but not by too much ok. And that is sort of empirically added in by this
little linear relation.
Where this 𝜆 is something called as the channel length modulation parameter. Now if
somebody were to ask you as to what is the resistance of the MOSFET or the small
signal resistance of the MOSFET, what is the answer ok. So, think let us think of the
642
MOSFET let us take the MOSFET and what we want to know is what is this resistance
the small signal resistance here.
So, in other words we want to find out what is dVDS /dIDS, which is if I have a small
signal fluctuation in the drain to source voltage, how does the current in the MOSFET
fluctuate ok. So now, this strongly depends on whether the MOSFET is in linear or in
saturation. So, let us say that the MOSFET is in linear mode operation.
So, what is dVDS /dIDS for a MOSFET in linear. So, if the MOSFET is in deep linear let
us ignore this term because V DS is let us say V DS is much much smaller than dVDS /dIDS
(VGS -VT) so we can just ignore this term. So, what we are asking is; what is for a
MOSFET in linear what is d V DS by mu W over L so, d of ids which is mu W over L C
ox into (VGS -VT) VDS.
So, you just take this derivative and take the inverse. So, you take you calculate d I DS
by dVDS and take the inverse of that, and you will end up with the resistance being small
signal resistance; of the MOSFET in linear in linear mode operation being 1/ 𝜇Cox W/L
into (VGS -VT). So, that is the small signal resistance of the MOSFET which is operating
in linear mode operation.
So, if you increase the mobility of the carriers as the resistance go up or down it
definitely it makes a lot of sense that it goes down. If you improve the field effect by
643
increasing Cox the resistance of the MOSFET goes down. If you make the channel width
much larger than the channel length and increase the aspect ratio of the MOSFET of
course, the distance travelled by the electrons from the source to the drain decreases and
the sheet charge area increases and therefore, the channel length resistance will come
down, if you increase the aspect ratio.
And if you increase the gate over drive voltage which is the gate to source voltage minus
VT if you increase that definitely you have created more inversion more electrons and
therefore, the resistance comes down. So, this is the resistance of the MOSFET in linear
operation. Now what about saturation? So, what is dVDS/dIDS in saturation.
So, let us say the MOSFET does not have channel length modulation let us ignore that
ok. So, you have this, now since this current does not depend on so let us write the linear
equation first let us just keep it aside. So, that is 1/𝜇Cox W/L(VGS -VT) that is the linear
small signal ok. Now for saturation they are using a saturation mode current, but we are
not going to we assume that it is an ideal MOSFET and there is no channel length
modulation.
So, what is this since there is no dependence on V DS in the current this is infinite ok.
Because dIDS/dVDS now becomes 0 in the current does not vary with the voltage it
becomes 0. We can vary this voltage as much as you want in saturation ok. So, let us say
VDS increases in saturation is greater than (VGS -VT) and you increase or decreases
644
voltage as much as you want in this region. The current in this will never change ok, it
does not depend if there is no channel on modulation.
So, the MOSFET behaves like constant current source, you can change the voltage
across it as much as you want and it behaves like a constant current source, and in a
constant current source the output impedance is infinite and the MOSFET is telling you
that it is got an infinite output impedance, but if you do include channel length
modulation ok. So, if you do include this then this is no longer infinite you have a V DS
term and therefore, you have a lambda into this particular parameter appearing in as the
channel resistance ok.
But ideally speaking the small signal resistance of the MOSFET in saturation is very
high and ideally infinite ok.
So, if you look at these characteristics here if you take the slope you find that there is a
resistance it is a finite resistance, but here you see that there is no change in current
irrespective of what the change in VDS is and the resistance is infinite, the small signal
resistance is infinite ok. We are not talking about what is VDS /IDS because you do have a
large enough IDS. And that is I am we are not talking about the large signal resistor we
are talking about the small signal resistance, which is if there is a fluctuation here then
what is the fluctuation in the current an ideally it should be 0.
645
Now what are the transfer characteristics tell you; the transfer characteristics tell you that
till a certain point there is no increase in current so we could define a threshold voltage
somewhere.
So, if you take a very low V DS if V DS is much smaller than any V GS that you are
applying, then that characteristics is almost linear which is what is expected it is almost
like a straight line. And if you were to project the straight line all the way back you
would intercept the x axis at the point of V T that will be the threshold voltage.
646
So, we will have a small module on parameter extraction where we can show how to use
these characteristics to extract the parameters, but just to look at that just look at that
point again. Let us take the linear mode operation of the MOSFET right so here is the
MOSFET in linear model.
So, you have I DS is we will say it is deep linear where V DS is much smaller than V GS.
So, you have mu 𝜇lCox(VGS -VT)VDS. So now, we are plotting the transfer characteristics,
how does IDS depend on VGS; it is a straight line. As below V T as long as V GS is below
VT the current is 0 we will develop finer models which will tell us exactly what the
current should be, but for now it is 0 because there is no inversion layer and after VT the
current should linearly increase. And what is the slope of this the slope is simply mu 𝜇W
LCox.V DS.
So, that is the slope of the current voltage characteristics current the transfer
characteristics when the MOSFET is in linear mode operation. And what is the x
intercept of this characteristics; it will be VT. So, you can work out and see what the x
intercept of this characteristics is. So, this becomes a very useful method to extract both
the mobility as well as the threshold voltage. So, it becomes a very powerful tool and
you want to extract the parameters of a MOSFET.
647
Semiconducter Devices and Circuits
Lecture - 40
Subthreshold swing, Additional concepts
So, the way this region is defined or the way the transition is defined is through a term
called as a Subthreshold Swing or in fact, the inverse of it which is the Subthreshold
Slope ok. And the subthreshold swing or is measured in terms of the volts per decade and
I will tell you what; that means, it means that if we plot if you plot the current voltage
characteristics on a log scale.
So, we plot the log IDS ok. So, we say to the base 10 let us say log I DS to the base 10
versus VGS I am sorry V GS the gate to source voltage. So, till a certain point it is all in a
basically half and then from here from say V fb to VT well it is not off very far below VG
V fb just to keep things a little accurate.
So, let us say it is off around this region and from V fb to VT is start having diffusion
defining a current and since the current is proportional to the exponential of the surface
potential which in turn is related to the gate voltage, you will find that the characteristics
takes a linear relation on a log ID versus VGS scale and then beyond V T you start entering
the square law dependence of the current.
648
So, this region is our subthreshold region of operation, now for us to climb one decade.
So, it is a log 10 right. So, let us say this is 1e- 7 and then that is 1e-6 and 1e- 5 and so,
on so many amps of current. So, for us to get from one to one decade that is from 1e
minus that is to climbed one order of magnitude ok.
So, 1e- 7 to 1e- 6 we need to apply a certain amount of gate voltage. So, we have gone
from say V GS1 to V GS2 for us to get past one decade. This voltage difference divided by
this one decade is essentially the idea behind your subthreshold swing in, how many
volts does it take for us to get the current past one decade ok. So, it is the number of volts
per decade.
On the other hand the slope is just the inverse right it is the reciprocal of this parameter
ok. So, it is good to have a better estimate of what the subthreshold swing or
subthreshold slope right.
So, subthreshold swing is defined by dVgs/dlog10 Ids which we will rewrite as dVgs/dfires
because V gs is nicely connected to the surface potential into d𝜙s/d natural logarithm of
Ids and will make a correction for the natural log to base 10 log conversion ok. So, we
have redefined this differentially this derivative as with these 2 terms and this
coefficient.
Now, how is V gs related to fires, Vgs so, we know this relation. So, all the gate voltage
that has been applied is equal to the flat band voltage, plus the depletion charge, plus phi
s and when 𝜙s heads here 2 𝜙s is called as the threshold voltage, like a so, we are now at
a phi s below the threshold voltage that is the sub thresholds volt.
So, that is the subthreshold region. So, we know that V gs is equal to V gs minus V fb is
equal to phi s plus Qdepletion/Cox. Now what we will do is, we will define an artefact or
parameter which is called as a depletion capacitance ok. So, I would not call it an artefact
it is it truly does exist, but the way we will define Qdepletion we say depletion capacitance
effective and the way we will define Qdepletion is by saying that it is the surface potential
into C depletion ok. So, what does this mean?
So, this means that ok, in you have your insulator, you have your semiconductor and you
now have your depletion layer, because of all the applied gate voltage and the electron
concentration is very small they just started to appear ok. So, the way we have defined Q
649
depletion is we say that, this is the depletion this is the band bending and therefore, that is
my surface potential or in other words the surface potential is basically the potential drop
across this region that is the surface potential. And we say that all the charge here is
equal to this potential into this effective capacitance of this region it is be treating this as
a capacitor.
So, we define Q depletion as 𝜙s C depletion and make the substitution here ok. So, you get you
get this relation V gs =Vfb + 𝜙s(1+Cdepletion/Cox) and therefore, the dVgs /d 𝜙s becomes this
term, it is (1+Cdepletion/Cox)..
So, we have solved for this part of or subthreshold swing, now in order to get this
derivative let us go back to this relation.
So here you have your dependence of I ds on 𝜙s right. So, this is the only dependence on
𝜙s everything else is independent of 𝜙s right. So, all this is independent of 𝜙s. So, let us
just say that all this is equal to some a times exponent of q 𝜙s/kT that is your IDS ok. So,
my writing at there does not make it appear on the screen. So, we just say I DS is some
constant times exp(q 𝜙s/Kt), which means which implies that my natural log of I DS is the
natural log of A q 𝜙s/kT and therefore, d natural log of I DS/d𝜙 s is simply this is a
constant and therefore, this is simply q /kT and the inverse of that is kT/q. So, we can get
the second part of this definition which is d𝜙s/d ln of I ds is equal to kT/q.
650
So, therefore, the subthreshold swing is given by this term here. So, it takes this many
volts sorry it takes this many volts to get the current past one decade. So, what does this
mean how do we improve the subthreshold swing, we can improve the subthreshold
swing by we cannot do much with the temperature and we cannot do much with the
Boltzmann coefficient and you cannot do much with q since we will have to change the
universe we live in order to do that. Therefore, the only things that are under our control
are the depletion charge and the oxide charge.
Now we have although you still not defined these things you will realize that as more
parasitic starts developing in the device. So, for example, let us say the interface between
the semiconductor and insulator will have interfacial trap charge. So, all that will add on
because all those regions will trap charge in them and they will all add on to the charge
present in the semiconductor.
So, any parasitic capacitance that traps charge or contains charge at the semi conductor
insulator interface and which lies in the at the semiconductor in insulator interface will
actually add on to these terms in the numerator and all such parasitics will effectively
start increasing your subthreshold swing and decreasing your subthreshold slope.
So, the subthreshold slope probably this is a more popular definition is simply the
inverse of the subthreshold swing for and therefore, the subthreshold swing will increase
which means you need a larger voltage to head towards one decade ok. So, that is a
lower subthreshold swing and that is a higher subthreshold swing. So, the subthreshold
swing will increase as we have more and more parasitics and the sub thresholds slope
will decrease as we have more and more parasitics.
So, this parameter the subthreshold swing is a very key parameter which determines the
rate at which the MOSFET goes from the off to the on state and you can see there is got
a very fundamental limit. So, let us say C depletion is 0 or you know let us say the no
parasitic, somehow you know you built an architecture where in all this is minimal and
you can almost ignore this term compared to one. Even then you have a fundamental
limit of ln 10 kT/ q being your subthreshold swing, but if let us say C depletion plus C ox
is equal to Cox ok.
So, you could say that this is a basic limit for us to what is say, then there is a limitation
of this order in the subthreshold swing, you cannot do too much better than this using
651
this present MOSFET configuration. So, therefore, in order to overcome this barrier we
need to reengineer the MOSFET architecture and you have many such architectures
which try to overcome this barrier although we will not be talking about them in this
course.
So, just to summarize the subthreshold characteristics, we have we saw that your I DS
versus let me just use the space a little better. So, we let us draw 3 characteristics here.
So, we will show I DS versus VGS and we saw that. So, let us say this is the VT. So, we
saw that there is a exponential dependence on V GS and then you have an almost square
law dependence on V GS after that ok. So, you will find that the characteristics of a
MOSFET all look like this as you plot the I DS versus V GS. So, this is the transfer
characteristics of the MOSFET for different V
and as V DS increases the characteristics gets more and more into the saturation region.
So, for very low V DS you will find that you remember the subthreshold current also
depends on the exponential of the V DS. So, for very low V DS it is going be all most the
characteristics would look almost like this because you are very strongly in linear.
Now you could plot the same characteristics on a log. So, this is the subthreshold region
right. So, technically let us say from Vfb to VT. So, you could plot the same
characteristics on a log I DS versus VGS (Refer Time: 13:26) and that is quite useful. So,
we will have a complete a lecture on parameter extractions. So, we will actually look at
652
all these things in good detail there, but what are the useful piece of information that you
can get from this curve.
From the slope of this curve at very low VDS or even at any other VDS if you know the V
DS it is possible to extract this parameter mu 𝜇Cox.W/L and from that we know what mu
is ok, if you know the insulator thickness we can guess what Coxs, but in order to get C ox
we can always do a capacitance voltage curve and you will find that you know what C
oxs.
So, if you know C ox and you know the aspect ratio because we have designed the
transistor and we know the geometry of it we can calculate what mu is from the slope of
this curve. But from the log I DS it is the same data right you take the transfer
characteristics the same experimental data sets and you plot it on a log I DS versus V GS
curve and what you will find is that, you will find that the characteristics shows this
almost linear relation and then it bends off and this region is the subthreshold region.
And that region is the linear or saturation region of operation that is all above threshold.
So, in the subthreshold region you are technically expected to get a straight line because
of the purely exponential dependence, but having said that when we talk about non
crystalline semiconductors we will see that there are some variations in the subthreshold
slope itself because of the existence of states inside the band gap. But for now for this
lecture what this lecture tells you is that this should be a straight line because the current
is exponentially dependent on the surface potential or there the V GS and then you have
a entry into the above threshold region and this mu of the curve sort of defines what your
threshold voltages.
So, from this curve we can estimate the subthreshold swing, which tells you which gives
you a lot of information on see depletion by C ox and which you can actually correlate
with this CV measurement and iterate for this measurement to make sure that you have
the right values and therefore, calculate if there are any differential capacitance is or
etcetera, etcetera.
Now this log characteristics also gives you another information if you look at the current
here which is below V fb. So, have to say if you look at the current here you will
understand what the leakage current in the MOSFET. So, we have still not approached
that boundary so, we will do that, but a little later. It tells you what is the current between
653
the source in the drain then we think that the MOSFET is completely turned off which is
when we assume the MOSFET is switched off that is there is no V GS present ok.
So, there is still there is V DS, but there is no V GS and what is the current through the
MOSFET or the V GS below V fb so, what is the current through the mosfet. So, we will
talk about the leakage mechanisms in the MOSFET a little later. And the third plot that
we looked at was the output characteristics, which is very useful when we design circuits
because it tells us a lot of things about the resistances.
So, this is I DS versus V DS which is the output characteristics or the MOSFET and that
had a form like this and as you increased V GS, you found that you can get these different
family of curves, you can get these different curves and as we increased V DS we went
from the linear region to the saturation region ok. And also we looked at channel length
modulation which did not allow these curves to become perfectly flat, but instead gave at
dependence on the saturation provided at dependence of V DS on the for the saturation
current.
So, this is summary a quick summary of our current state of knowledge in the with
regards to the MOSFET current voltage characteristics. Now so, we at the beginning of
this lecture we said that there are 2 aspects that we have still not studied, one is the
currents below V T, which we have now you know understood a little better because
these are all the subthreshold currents and they are all related, they are all governed by
the diffusion of electrons from the source to the drain, the diffusion of minority carriers
from one of the electrodes to the other. And second point that we said that we were not
we had not covered completely was a variation in the depletion width, which we thought
might affect the threshold voltage.
654
And we are going to now address that point. So, what are we doing we have just we have
this very basic model for the current voltage characteristics the MOSFET and we are
simply improving this model you know with these details. So, the model the first model
we had was from V T till above a 2 gate voltages above V T, which is all the purely
strong inversion characteristics where in we looked at linear and saturation regions of
operation. And then we defined the regions the current below V T which is the
subthreshold currents and we defined the subthreshold swing which was an important
parameter.
And now, we are defining the current voltage relation a little more accurately ok.
Although we will just introduce this introduce the ideas behind this, but not actually use
this for many other purposes, but this is a more accurate model. And this model is
something called as a bulk charge theory and it takes into count the variation in the
depletion width.
So, as we move from the source to the drain the initial model I mean just to re restate the
introduction, as we move from the source to the drain not only it is the channel charge
decrease, but that depletion width is also supposed to vary. And because that is because
the P-N junction here is this P-N junction is more strongly reverse biased as compared to
this P-N junction. And this depletion width variation creates an inaccuracy in the way we
defined the channel charge.
655
So, in the previous case we had assumed that the V DS is very small. So, therefore, we got
away with many things we said the V DS is very small and therefore, you know the
depletion width does not change too much and that is the channel charge and that qn is
simply equal to your C ox into (V GS -V T -V c) /h or you know the charge per unit area is
C ox (V S -V T - V c), where the V T was treated like as it is a constant, but in reality as if
you take a section here and if you take a section here, since the V DS is different or since
the V c is different the depletion width is also different ok.
So, it becomes more prominent in this case if I were to take a section here and a section
here, the depletion width is suddenly increased right. So, the depletion width here is x d 2,
there is a depletion width here is x d1 and x d2 is clearly greater than x d 1. And since the
depletion charge depends upon the x d, the threshold voltage in this region has to be
greater than the threshold voltage here and that variation and threshold voltage is
completely missed out in our study, because we went from very low V DS and directly
into saturation by simply substituting V DS is equal to V GS -V T ok.
So, that is the little gap in our model and what we will do now is, we will completely we
will try to see how that gap can be corrected for. So, the V T is got 3 milestones it is the
flat band you first reach the flat band voltage and then you do Q depletion /C ox you deplete
the region and then you bring the surface potential to equal to 2
𝜙F. Now the problem is with Q depletion right it is not with phi F is purely dependent on
the acceptor on concentration V fb is completely dependent on phi ms and it is only the Q
depletion that depends on the value of xd and it is qNAxd/Cox.
So, as xd varies VT must vary. So, near the source side there is at xd at y=0 this is our x d
that is the depletion width and near the drain side you have a larger what you say
depletion width and that is given by this term. You are all now already very familiar with
you know how in how to define depletion width in terms of the surface potential. And at
any point y this is the depletion width it depends upon the potential of the channel there.
656
Therefore, we can correct for our relation of qn in this particular manner. So, we will
define. So, this should we will try to define this should be qnh ok, it is the concentrate
that free carrier concentration per unit area ok. So, particularly I am looking at all of
been say capital N here ok. It is the inversion carrier concentration per unit area and not
per unit volume and that is equal to Cox(VGS-V T -VC) just like the way we defined, but
the V T is computed at the source which means using an x d value that is present near the
source which is close to the equilibrium x d value of the P-N junctions.
So, we will use that V T, but then we will come to we will correct for it by saying that
instead of you should have we should have actually been using (VGS-V T -VC), but we
will say that this is equal to Cox(VGS-V T ) at y equal to 0 minus V c minus qN A into x d
of y minus x d of 0. Because all the other terms the phi in the phi ms and the 2 phi F
terms do not change it is only this term that we will correct form ok.
So, essentially we are subtracting qNAxdo and instead adding qNAxdy to this relation here
to the VT here. And by making this correction we can now rewrite the differential
equation for the drain to source current. We say that IDS is qn𝜇 electric field into area,
where this area is your you know once again it is the Wh where the h cancels off and we
say that the current voltage relation or the differential equation that you need to solve this
IDS is equal to W𝜇 into this particular term into dVc/dV y ok. And what does this term
657
this term is nothing, but your corrected channel charge per unit area, upper unit volume
because the h has cancelled off.
So, this is the h and that is cancelled off so, that is the channel charge ok. And you saw
now you if you see the Vc, the Vc term appears in your differential equation here as well
as here. So, earlier we did not have this term and therefore, we ended up with a square
law model, but now with this term being present and you have a Vc to the power half and
a V c term. The model is going to be slightly more complicated than a square law, but it
is going to be more accurate because it takes into account the variation in x d.
So, this theory is something called as a bulk charge theory and it generally predicts a
lower amount of current ok. So, if that was the square law model the bulk charge theory
predicts the current that is slightly lower. And the second point is that the way you define
the saturation VDS that is in the square law model the V DS saturation that is the V DS of
beyond which the transistor entered saturation was defined as simply any V DS greater
than or equal to (VGS –VT ).
Then definition will change a bit because your current voltage characteristics has now
changed, but it is up to you to solve for this exact differential equation as a homework.
Although in one of the assignments I will put out the exact solution we will have this is
an assignment question due to which I and it is due to that reason that I have not given
our the answers here, but we will discuss the exact answer when the solutions to the
assignments are presented.
So, these are the 2 key points the Subthreshold Slope the understanding of the
subthreshold current and the Bulk Charge Theory model with sort of refines the current
voltage characteristics that we developed.
658
Lecture – 41
Trapped charge, Body-bias
So, let us now come towards a more practical case and that is the case of having charges
trapped inside the dielectric ok. And so, fine our discussions with the MOSFET or the
MOS capacitor we always assume that the insulator was ideal, and it had no charges
trapped inside which implies that the insulator; all the insulator did was to prevent any
current from the gate into the semiconductor. And it did that by allowing for a voltage
drop across it and by acting like a perfect dielectric with a high band gap.
But in reality the insulator could have charges trapped inside ok. And there are many
possible mechanisms by which charges can enter the insulator. And the insulator need
not be a perfectly crystalline material, it could be it could have defects in its structure
which could result in it is energy band diagram having you know states inside. And all
these states as well as any trapped charges inside make the insulator non ideal. And these
energy levels or these trap states inside insulator also acts as a home for any free carriers
that are floating around.
659
So, let us now consider what happens you know what happens when you handle an
insulator of this kind. So, far in our discussions we always assume that the insulator was
ideal ok. So, now, what we are going to do is we are going to improve the model for the
MOS capacitor and a MOSFET, by considering the possibility of charges being trapped
inside the insulator ok.
And this could happen due to several reasons and we will address all those reasons. But
before we do that, let us consider what happens in an ideal insulator ok. Let us look at
the ideal insulator little better and I am sure most of you are familiar with this, but we
just doing it to establish continuity between this discussion and the next.
So, if the insulators ideal let us say we build let us keep the MOSFET or the MOS
capacitor side and let us just build a parallel plate capacitor. So, you have a metal plate
here, and you have an another metal plate here, and you have applied a voltage V a
across these metal plates. And you have a perfect insulator, that is an insulator with no
defects and no charges inside sitting inside between these two metal plates so it is a
perfect capacitor. And you also have the insulator having a thickness of tox.
Now, if the charge concentration in the insulator 0; then you have a Poisson equation
which is given by de /dx is equal to 0 which means that the electric field and the
insulator has to be a constant. Moreover your potential, the potential term in the insulator
660
will depend linearly with x. So, you have d∅/dx=-Co. And therefore, ∅ is equal to this
constant times x, plus another constant which is C1.
And by applying these two boundary conditions there is at x equal to 0 the potential is Va
and at x equal to tox the potential is 0, we can establish what C1 and C0. And therefore,
you end up with your electric field being a constant which is Va/tox and you end up with
the potential that you know that dk is linearly in space and it is basically your Va by tox
into x where this is nothing, but your electric field in the oxide into x. So, this is the
situation in an ideal insulator.
But the moment the insulator gets some charges inside ok, rho is no longer 0. And here is
a nice picture taken from one of the reference techs ok. So, I will edit this in edit the
correct reference in and this picture has been borrowed from that text. And it shows the
shows a more practical scenario of the insulator or in this case a silicon oxide sitting in a
MOSFET. So, you have the metal you have the silicon oxide and you have the silicon
and you have the insulator in the MOSFET. Now this is a very busy diagram you see a
lot of species here.
So, ideally we would like to keep the insulator very clean, but in some practical scenarios
it may not always be possible. So, in that case your charge concentration in the insulator
is not 0, and your Poisson’s equation becomes this, where you have a certain charge
concentration depending upon all the species present in the insulator. And the electric
661
field is simply an integral of this and the potential is therefore, the double integral of this
term with the integration constant or with the with the integration constant over here
appearing providing a linear dependence on x to the potential. So, this becomes your
potential function ok.
And therefore, if you were to take the potential difference between the case where the
insulator has charge trapped inside, and the insulate in the case where the insulator does
not have any charge trapped inside. In this case you have this term and in the case of no
charge trapped you had this term. So, of course, we have ignored the any constant
potential applied across the insulator.
662
And therefore, that the difference between these two terms the difference between these
two terms is this particular entity that crops up here. So, therefore, you do have a
potential difference that can vary based on whether they depending on the amount of
charge is trapped in the insulator. So, how does all this affect our MOSFET and how
does it affect the MOS performance ok? The key is that if you have charges trapped
inside you are going to change your flat band voltage, at why does the flat band voltage
change. Let us let us consider a simple example ok.
So, let us say you have your ideal MOSFET ok, you have your gate, you have an
insulator which is ideal, no trapped charges inside, and you have your semiconductor, so
663
I am not shown the source train etcetera. So, if I were to apply some positive charge on
the gate, then the semiconductor responds by bringing in negative charges close to the
semi conductor insulator interface ok. So, this is the only response we looked at. Now
these charges could be the depletion or it could be the inversion carriers and depletion,
but nevertheless it is all happening in the semi conductor.
But now let us say with the insulator also has some trapped charges; so, in order for me
to bring the same amount of charge concentration at this in the semiconductor interface. I
need to apply a larger gate voltage because, I need to first compensate for this insulator
charge and then bring in the field to establish my inversion layer or the depletion charge.
Therefore, it is equivalent to us saying that the flat band voltage is increased or flat band
voltages decrease depending on what the charges are in the insulator. And therefore, the
V t which depends on the flat band voltage also changes.
So, therefore, charges trapped in the insulator can affect your flat band voltage and can
affect the threshold voltage of your transistor. So, what kinds of charges are trapped in
the insulator. So, you have you have the you have different kinds ok, you have charges in
the insulator and you have lot of charges at the semi conductor insulator interface. So, let
us just go through you know what these are and you know how they occur etcetera.
So, the first ones the charges in the insulator you have mobile ions. Now mobile ions are
typically you know your sodium and potassium ions basically these are salts that have
664
ionized. And they occur because of poor handling of the samples or you know use of
poor glassware etcetera. So, for example, you know you the ions present on your skin
could get into the dielectric if you do not have proper clean room procedures ok.
And these mobile ions once they are inside the oxide they can start drifting inside based
on the field is based on the field applied on the through in the oxide. So, if the gate
voltage is positive these ions will start drifting towards the semi conductance latent
interface, where they will have a larger impact on the threshold voltage. And one way to
neutralize these ions is by having a ion trap layer present right at the gate insulator
interface.
The other kinds of charges in the insulator are something called as the oxide trapped
charges. These are basically the electrons and holes from your channel, that had very
high energy and that got into the oxide. And once inside the oxide they are trapped and
there mobile because they can move at high temperatures and high fields they can move
around. So, these are oxide trapped charges essentially refer to electrons and holes as
opposed to mobile ions.
665
So, these are electrons and holes that are trapped inside the insulator. So, let us let us
view the situation little better. So, if you if you remember, you have your insulator and
you have the band bending in the semiconductor and you have these electrons here. And
these electrons are all basically you know they are all they all have their own wave
function.
And therefore, there is a probability of finding an electron in the insulator because these
electrons are now trapped in this potential barrier. So, if there are if there are defects in
the insulator, the electrons the high energy electrons can actually get in and occupy these
defect states ok. And then they could hop around from one state to another and thereby
migrate around in the insulator.
666
The other kinds of charges on the charges of the semi conductor insulator interface. So,
you have your gate metal, you have your insulator and you have your semi conductor
and there are these are charges that are located at this interface. They are not inside the
insulator, but they are at this interface. The most common and the hardest get rid of is
something called as the interfacial trapped charge. So, essentially this is because of
dangling bonds. So, you have your silicon lattice or in fact any lattice for that matter ok.
And at the point where we build the insulator, this lattice terminates this lattice structure
which was let us say you have a perfectly crystalline lattice. This perfect crystal
terminates at this boundary. And now you have a different crystal in this case it is the
silicon oxide crystal ok. And this oxide crystal has to now match with the silicon lattice
ok. I mean these two have to start sharing electrons to create bonds. Now, depending on
the insulate and the semiconductor that is not always very efficiently possible. And what
happens is at this interface we are left with a lot of dangling bonds ok.
So, several low the bonds are shared, but then this atom could not find another atom
share this bond width ok. So, you end up with a lot of dangling bonds and these dangling
bonds are essentially traps. They are they have got a certain energy and they are looking
for an electron to trap. And they not only reduce the electron concentration at the
interface they act as traps, but they also effectively add on to something called as the
interfacial capacitance ok. So, if you recollect the point on sub thresholds swing.
667
We saw that a key parameter was (1+Cdepletion/Cox). And at that time I had mentioned that
if there are trapped charges then it could increases sub threshold swing and therefore,
reduce the sub threshold slope. So, these trapped charges can effectively contribute to
slowing down the turn on and turn off of the device. And these trap charges also change
the flat band voltage. The next kind of charges are something called as fixed charges, and
these are charges which are just inside the insulator, but very close to the insulator
semiconductor interface ok.
And these are these have got these charges are curved simply because of you know the
way for orientation, whether it is a 1 0 0 plane at the surface or 1 1 1 etcetera or the
process flow and the temperature at which we had the process flow etcetera. So, these are
fixed charges they are not moving anywhere, but they are very close to the semi
conductor insulator interface and they also contribute to the trapped charges in the
insulator. So, these are different categories and classes of the different kinds of trapped
charges you might see in the insulator.
Now, what is the effect of all this the effect of all this is that it is going to change the flat
band voltage. So, far the flat band voltage in the insulator was purely because of the
metal semiconductor work function difference. So, when we drew the band diagram at
the start of the MOS theory. We said that there is a volt Fermi level difference between
the metal and semiconductor at thermal equilibrium these Fermi levels will align.
668
And therefore, you start having you start having bands bending even at equilibrium. And
we said that the way to get the flat band voltage was to apply a voltage so as to positive
or negative voltage; so, as to get all the bands flat ok. Now, therefore, ∅MS, was the only
voltage we need it to apply, that was the gate voltage that we need it to apply. But now if
you have trapped charges in the insulator you also need to compensate for the trapped
charges because if even after applying phi MS.
There might still be band bending because there is already a field because of these
trapped charges. So, even these could cause a band bending. So, we need to overcome
even these charges ok
669
And therefore, the total flat band voltage will now be a summation of phi MS plus Q
trapped by C ox into a parameter called 𝛾. And I will tell you what 𝛾 is? Now 𝛾 is trying
to indicate the location of these trap charges. Since these not all charges are present at the
insulator interface ok. You have you have charges that could either be located at the semi
conductor insulator interface or somewhere in the middle or maybe at near the gate
insulator interface etcetera.
So, and this distribution is important because it directly impacts a Poisson equation right.
Because your Poisson equation says it is 𝜌(x)/𝜀=dE/dx. So, unless you know the
distribution of all these charges you cannot solve Poisson’s equation accurately. So, good
estimate for all these charges is to get the average or the expected value of the location of
these charges ok.
670
And that expected value is defined by this parameter gamma and this gamma is a
dimensionless parameter. And the way it is defined it can take a value between 0 to 1 and
the way it is defined is it is. So, in case you cannot see these it is the integral of 0 to t ox
of x𝜌dx ok. If you think of 𝜌dx as some kind of a distribution a density function, the next
times that is the expected location of this charge ok. So, people are familiar with
probability theory can think of it is a as a definition of the expectation, divided by tox
into integral of 0 to tox of rho dx.
So, this is telling you, so if this so what this is telling you if this is x equal to 0; if the
gate metal insulator interfaces x equal to 0 and if the insulator semiconductor interface at
x equal to tox. What this is telling you is, it is trying to estimate a mean position for all
the charges. So, if you say that the mean position from here to there let us call that as x’.
So, if x’ the mean position then gamma is equal to x’/tox. So, which means that if the
charges are located if most of the charges are located near the insulator semiconductor
interface 𝛾 takes a value of 1. And if the charges are located near the gate insulator
interface 𝛾 takes a value of 0. And if the charges are located right in the middle gamma
takes a value of half ok.
So, this is near the insulator semiconductor interface, and that is sorry this is near the
insulator semiconductor interface and that is near the gate insulator interface. So, if
gamma 0 then the flat band voltage is not affected ok. So, you can think of it as a part of
671
the gate metal and it is part of the gate charges. But if 𝛾 is 1; that means, if the charges
are located closer to the semi conductor insulator interface, then your flat band voltage is
more significantly affected ok.
So, in this case the VT is significantly affected; whereas, in case of gamma equal to 0 VT
is not significantly affected. So, this is a simple model it is taking making use of the
average location the expected location of all these charges in order to make an estimate
on the flat band voltage. So, this is how you handle trapped charges in the insulator. So,
this is just our little excursion into handling this non ideality in the MOSFET.
Now, that is not the only thing that can affect the threshold voltage ok. So, now, another
key parameter is the body voltage. So, far in our discussion of the MOSFET we said that
all potentials are taken with respect to the source. And therefore, what is important is the
gate to source potential, drain to source potential, and the body to source potential, which
so far we had kept at 0.
So, so far if you remember the body was always at ground we never worried about the
body ok. But now let us dislodge this ok, let us not keep it at ground. Let us say that the
body has got it is has got a potential that is different from the source. So, how does this
affect the performance of your MOSFET.
672
Now this is a very although we spend a very short amount of time on it is simply because
of the time constraints on this course, there is actually a very important tool for the
circuit designer.
Because, you can control the threshold voltage of the MOSFET by influencing the body
to source voltage; so, you can have different MOSFET’s having different threshold
voltages and we will see why it happens. So, this in the structure the MOSFET so this is
the P type body, and you have this N type, N doped source and drain contact. Now, if the
673
body and source are at a different potential then if we increase the body to source voltage
or let us say we decrease the body to source voltage we make the body voltage less than
the source. So, if the source is at ground let us say the body voltage is𝜙 -2 volts. So,
essentially we reverse bias this PN junction diode ok. If you look at this PN junction
diode it is got it is got the source and it is got the body here and that was the barrier ok.
So, if we apply a body to source voltage we now start shifting playing around with this
barrier. See if you reverse bias set or you forward bias set you can change this barrier
right.
Now, if you reverse bias this PN junction diode, essentially you have made this depletion
region have a very high electric field. And therefore, any charges that are trying to
appear at the interface will all run runaway because of this high electric field. They will
all migrate to the source and drain electrons.
So, in other words so far we were we were getting inversion when my phi s was equal to
2𝜙F where 𝜙F is nothing, but the work function insulated difference in the bulk. So, that
is your 𝜙F so, when the only when the surface potential hit 2 𝜙F did we have did we
have inversion. But now the surface potential must be equal to 2 𝜙F -VBS.
674
So, the body the body voltage basically changes the back potential of your MOS
capacitor. So, your MOS capacitor had a gate voltage it is always with respect to this
body which was grounded. But now this is this is offset that potential so, the condition
for inversion is now different.
So, how does this impact the threshold voltage? So, if you have to look at the threshold
voltage. So, threshold voltage is basically the gate to source voltage at which we achieve
turn on right. And this happens in my surface potential hits 2 𝜙F -VBS now.
675
But what is the VGS? Now, the body is decoupled the VGS can be now written as VG to
body plus V body to source ok which means it is basically VG-Vbody+Vbody-Vsource. So,
this is my gate to source voltage. And what is the gate to body voltage? Now, the gate to
body voltage has now become this term here. And what does it have? It has your phi MS
because of the metal semiconductor work function difference.
It has your depletion region by Cox depletion charge by Cox, but the depletion charge has
to now take into account this extra potential here. So, it is not only 2 𝜙F so this is V gb
during turn on so this is during turn on. And why is it during turn on, because we have
said that the 𝜙F is equal to 2 𝜙F. So, this is the depletion charge by Cox.
676
And it is also got the 𝜙s term, which is now 2 𝜙F -VBS which is the new 𝜙s needed for
turn on. And therefore, Vgs turn on is nothing, but the VT that is become VGB+VBS which
is basically this entire term plus VBS ok. And these two terms cancel off and you end up
with this being the expression for threshold voltage.
So, if the body to source voltage is less than 0 which means if you reverse bias your p-n
junction then the threshold voltage will have to increase. And if the body to source
voltage is greater than 0, which means your forward bias your p-n junction, then your
threshold voltage will decrease. And you can see that happening here. So, if you reduce
if you make VBS positive then this terms becomes lesser and the threshold voltage
decreases.
677
On the other hand if VBS becomes negative and the this term becomes greater than 2 𝜙F
and therefore, the threshold voltage increases ok. So, the body to source voltage can also
be used to control the threshold voltage of the device. So, you can see the different things
that impact the threshold voltage, particularly through the depletion charge right.
So, you had your you have your bulk charge theory, which said that the bulk charge
theory which said that the VT will change along the channel simply because xd is
changing. Then you have your trapped charges in the insulator and then now, we have
seen the body to source voltage also affect threshold voltage ok. Now, we are going to
see more examples particularly when we head towards scaling of the MOSFET.
678
Lecture – 42
Scaling of MOSFETs
So, now let us start off, on a new topic, which is got to do the Scaling of the MOSFETs, and
this is a particular interest as it is the symbol, or you know, it is very symbolic of the
advancement of your semiconductor technology. So, it is quite, famous particularly in the
form of Moore's law. Although surprisingly, I am never referred to that here, which basically,
says which defines a pathway or a roadmap to scaling, which says that you know the number
of transistors in any system or any chip, doubles every so many years; but just to give you an
example.
So, what is scaling essentially before we start talking about scaling, what does what does
scaling mean. So, we have a MOSFET here which is, you know drawn in a very simple
manner, that is your source, that is the drain and that is the gate and that is the insulator, that
is got some thickness tox. There is a channel length between source and drain and the
MOSFET has got some channel width.
679
So, what do you mean by scaling? Scaling simply means, reducing the dimensions so, what
we want to do is, when we say we are going to scale the MOSFET down by a factor of 𝛼,
what we mean is, we are going to reduce the channel length by a factor of 𝛼. We are going to
reduce the channel width by a factor of 𝛼 and we are going to reduce the thickness by a factor
of 𝛼.
So, that is what scaling implies and why do we need to do scaling; so, that you can make the
transistor smaller and so that, you can have more transistors in a given layout area and by
scaling down the transistor you improved, the speed and the performance and I mean, the
speed and performance in the power. We change the power consumption, depending on how
you scale, of course, and we will look at that and the most importantly you can fit in a larger
number of transistors in a given layout area.
So, there are lots of benefits. These kind of benefits to scaling and that is what is led to you
know, the microprocessors, in the in the computers becoming faster and, you know, it is led
to, mobile technology etcetera etcetera. So, it is quite an important feature of any, technology
has to how rapidly and you know how effectively you can scale down to improve
performance.
So, just to give you an example, I hope, I am right on all these, but, if I am not, I should, I
would probably correct it, but in the 1970s you know. So, Intel 4004 processor had about
2000 transistors in it ok. And in 2015, that is what, 45 years ok. So, in half centuries time, the
number of transistors, has increased to these many.
That is a massive increase and in the 1970s going back to the 1970 the, there is a number of
transistors. The transistors had a typical channel length of 10 microns and in 2015, there the
channel length of the transistors about 40 nanometers ok. So, that is going, that is going by 3
orders of magnitude lower in terms of channel lengths. So, these kinds of advancements are
possible by scaling down MOSFETs, in a reliable manner.
So, what we are going to do here is, we are going to look at, what scaling is introduce it. Talk
about two different kinds of scaling strategies and what its effect is, ok. Some of the key
features and the parasitic that rise along with scaling that come along with scaling etcetera.
680
So firstly, how do we scale down ok. So, as I told you, you reduce LW and tox by a factor of
alpha ok. It does not matter. So, there are two strategies, one is something called as constant
field scaling and the other is something called as constant voltage scaling and of course, you
can have a hybrid of these two, but we will just introduce these two strategies. In the case of
constant fields scaling the electric fields, inside the device are kept constant ok, and what
does that mean? So, let us say, we have scaled down, we have scaled down the MOSFET
from L to L/ 𝛼.
So, there were there was a certain electric field in this MOSFET ok. So, there was a there was
a gate field, there is a source to drain field etcetera and we want to keep the magnitude or
these field strengths the same even though the transistors have been made smaller. And how
do we achieve that? We achieve that by simultaneously scaling down the voltages ok. So, you
want to keep the fields the same and therefore, the voltages have to go down along with your
dimensions.
681
The other idea is something called as constant voltage scale which, in this case even despite
your scaling down of lengths the voltages remain constant. In order for you, to continue
having compatibility with regards to the power supply required for operation etcetera and;
obviously, in this case, the electric fields are going to increase as you scale down the
MOSFETs.
So, you have a quick table here. So, I will just be prepared easy table, for see as to what
happens when you use constant fields scaling and constant voltage scaling. So, here we have
682
scaling down by a factor of k. So, let us say, we say k is equal to 4, what that means is, my
channel and let us say, my initial channel length was 8 microns, what that means is, if I scale
down by a factor of 4, my new channel length has become 2 microns ok, and my new channel
width has become the old channel width by 4. The new tox has become the old tox by 4
etcetera-etcetera. So, that is what scaling down by a factor of k implies.
Now, how does it affect the simple parameters? Channel width of course, goes down by a
factor of k, whether its constant voltages are constant field. Channel length goes down by a
factor of k, with its constant voltage a constant field, oxide thickness goes down by a factor
of k, with a channel length, whether it is constant field a constant voltage.
The area is basically, your W into L essentially of course, there are lot more things to the
area, because you have to have your pads, you have to have your interconnects, but to first
order the size of the transistor. If you are familiar with transistor layout, then you have a
certain channel length, you have a certain channel width and therefore, the area simply scales
as W over L plus of course, there are many other, aspects to the area.
So that, we will go down as 1 by k square since both my W and L reduced by a factor of k;

so, that is area and the electric field in constant field scaling. The electric field is not
supposed to change, that is the whole idea. So, that does not scale and remains constant at 1
and if you want to keep the electric field constant, you have to reduce your voltage and the
683
electric field is essentially, the voltages divided by all the lengths, and since the lengths have
scaled down by a factor of k, we also need to scale down the voltages by a factor of k.
So, the voltage in the case of constant field scaling will scale down by a factor of k. On the
other hand, the voltage in a constant voltage scaling case will not change. It has to be
constant and therefore, the electric fields all increase by a factor of k.
So, what about the insulator capacitance per unit area; so, that is Cox that is the capacitance
per unit area is nothing, but epsilon ox by t ox and our t ox scaled down by a factor of k and
684
therefore, the insulator capacitance per unit area has to increase by a factor of k and that is
true, whether this constant field or constant voltage scaling.
Now, what about the total gate capacitance and what do you mean by the total gate
capacitance? So, this is the total this is the insulated capacitance per unit area and therefore,
the gate capacitance is the total capacitance; so, CoxW L which is the area of the device. So,
since C ox increased by a factor of k and W decreased by a factor of k and L decreased by a
factor of k irrespective, whether it’s constant voltage or constant field scaling the total gate
capacitance reduces by a factor of k, what about the current in the devices.
Now, whether it is a linear mode or saturation mode the current has a square dependence on
the voltage. So, if it is a linear you say its 𝜇Cox W/L. So, let me write this properly; so, in
case of linear, we say that it is mu 𝜇CoxW/L, into (V GS -VT -VDS/2 )VDS. So, that is the
current ok. So, you can see that there is a voltage into voltage term. So, that is a V2, the
voltage square dependence and in saturation of course, its mu 𝜇CoxW/(L/2), into VGS-VT the
whole square.
685
So, again it is a voltage square dependence ok. So, what happens if you scale down? So, what
are the parameters here that scale, the mobility not so much W and L scale, but W / L does
not scale. That is the aspect ratio since W scales by k and L scales down by k. W over L does
not scale C ox does scale. So, Cox how does, C ox scale, in the case of constant field scaling, it
C ox increases by a factor of k.
So, Cox becomes k time cox and how does voltage change, with in case of constant field
scaling, the voltage reduces has to reduce by a factor of k and therefore, in the case of
686
constant field scaling, the voltage reduces by a factor of k and therefore, in the case of in the
case of constant field scaling. So, whether it is here or here it is the same. So, in the case of
constant field scaling the current has to reduce by a factor of k.
So, you can see its V2/k2 square into k Cox and therefore, the current reduces by a factor of k
on the other hand, in the case of constant voltage scaling. So, let us just write, I just keep
things simple. We just write the current, it is current scale, scalable the scaling aspects of the
current is dependent on Cox into V square we, because the rest of the parameters do not
change. So, in the case of constant voltage scaling V does not scale.
So, that remains 1 whereas, C ox has increased by a factor of k. So therefore, the current
increases by a factor of k ok. So, you can see that, in a constant field scaling case, the current
in the transistor decreases by a factor of k, in the constant voltage scaling case. The current in
the transistor increases by a factor of k and what about power consumption power is nothing,
but your current into voltage IV. And, since all the voltage is reduced by a factor of k, in
constant field scaling and the current reduced by a factor of k, the power reduces by a factor
of k square. Whereas, in the case of constant voltage scaling, the current increased by a factor
of k, the voltage remained invariant and therefore, the product scales up by a factor of k, ok.
687
So, here the power increases, here the power comes down ok, here the current increases here
the current comes down. So, this is how you handle scaling relations of course, there are
many other parameters, one can consider you know delay etcetera-etcetera, but we have just
talked about some other basic or the key parameters. So, what is the impact of scaling ok; so,
the first impact of scaling down is got to do with the threshold voltage ok. So, we will talk
about this first. So, is the threshold voltage influenced by scaling of the MOSFET ok.
So, what we are going to do is, we are going to start with a regular long channel device. So, it
is a long channel device and then we are going to head, towards a short channel device,
which is the scaled down version short channel device, and we are going to see what has to,
what happens to the insulator. So, before we head that, how could the threshold; we are going
to see what happens to the threshold voltage. Sorry I am probably getting 2, to use to saying
these terms and I am, making mistakes, but nevertheless, let us just carry on.
688
So, we have the threshold voltage and what is the threshold voltage depend on. So, once
again let me repeat, it has got these three milestones. You need to first get your flat band
voltage which you now know that, it depends on phi ms plus any trapped charge in the
insulator. Then you have to have your Q depletion /C ox ok, and this depends on the varying x d,
it depends on, the body to source voltage etcetera and then you need to achieve your 2 phi F
ok, but here again it depends on the body to source voltage, which is your phi S has to be
equal to 2𝜙 F. So, this is when you hit threshold voltage ok.
689
Now, let us look at this depletion term. So, what exactly are we depleting in a MOSFET. So,
here you already have a NP-junction. So, this is a p-type body; So, you have a n-type source a
n-type drain and you have a P-type body and you already have a depletion layer here, because
of this diode.
So, this region is already depleted for you. The gate has to deplete this region of the
semiconductor ok, what about this region here, that region is shared between the gate and the
PN-junction. So, this PN-junction at equilibrium will help the gate and depleting that. So, let
us say, half of that region is depleted by the PN-junction and the remaining half is the job of
the gate. So, the gate gets some help in that region, in these regions. It is only the PN-junction
and that region is only the gate and in this region the PN-junction helps the gate depleted, but
in a long channel device ok.
So, let me just write this down neatly, in case my; so, this is the job of the gate. The gate
needs to deplete this region. Half of this region is also the job of the gate, because this region
is shared between the PN-junction and the gate and everything else the remainder half and all
these regions are all depleted by this PN-junction between source and drain and bulk. So, all
these are jobs of the PN-junction.
So, the gate essentially, all the gate charges go in to deplete that particular region. Now, if it
is a long channel device, these regions here are as negligible portion of the entire area that
690
gate has to deplete. So yes, if you have a very long channel device and you reduce the
channel length by 5 percent, the gate is not going to see too much of, in effect ok.
So, what do you mean by long channel device? Firstly, long channel device implies that, the
distance between these channels ok. The channel lengths, sorry the distance between these
two electrons, the channel length is much larger as compared to all these depletion regions,
much larger than these ok. And in that case, it really does not make too much of a difference,
whether the gate depleted that much or whether depleted this much, because this portion is a
small percentage of the entire channel area ok.
691
So, the gate can be said to be depleting, the entire area, which is about W into L ok. So, that
is the area, the gate has to depleted.
Now, what happens; if start reducing the device, if you start reducing the device ok. So, here
I had a nice picture and I did not use it. This dark region shows, the portion of the, portion
depleted by the PN-junction, whereas, the gray region shows that, portion depleted by the
gate.
692
Now, if you start scaling down the channel length, so now, the channel length is reduced
quite a bit ok. So now, this depletion regions are not that small, we cannot really ignore them.
They have gotten very close to each other.
693
So, now this region suddenly forms a significant part of the total depletion charge, right, and
now, we need to start taking the count. This area, quite carefully in order to calculate, all are
make calculations on the threshold voltage, and since the source this PN-junction now, takes
up a larger percentage or the larger share of the workload, in trying to deplete the gate has to
do a smaller job and therefore, the gate depletes that region faster at lower Vgs, as compared
to the long channel device. So, in other words, as we reduce the channel length, the PN-
junctions between that is between the source drain and body the PN-junctions in case. In fact,
the other way, it is the body and source or drain ok. Since, we are talking about NMOS we
are always discussing the n MOSFET we have never discussed the p MOSFET.
So, far; The PN-junction helps the gate significantly ok. So, that is the key significantly, and
it depletes, it performs half of the workload it depletes it helps the gate deplete a lot of the
semiconductor. And therefore, the gate can a smaller gate voltage a smaller gate voltage
depletes the remained a portion of the semiconductor and takes the semiconductor to
threshold and threshold voltage is reached faster. So, in other words, you need to add a lot
less charge on the gate, in order to completely deplete this entire engine and head towards
inversion ok. So, therefore, as the channel length reduces the threshold voltage reduces and if
you want to see this mathematically, we just take the small analysis here.
694
So, let us just take the depletion region of the gate the region of the semiconductor depleted
by the gate. So, that is where your source drain would be and that is how the depletion region
would be around. Your source drain and this part of it is all taking care. It is all depleted by
the source and drain and it is only this region. Here, that is depleted by there gate and that is
the insulator ok. So, that is the picture.
Now, what is this area? You know, how much of this volume, this gate depleting what is the
total depletion charge. What is the total amount of charge? I must add to the gate, in order for
695
me to get a total depletion charge over here and by total I mean in coulombs. So, we are not
talking about per unit area or per unit volume total employs in coulombs ok.
(Refer Slide time: 22:59)
So, not per area or volume not this; so, what is the total charge? So, let us take a difference let
us say. Now, what is this area? You know, how much of this volume, this gate depleting what
is the total depletion charge, what is the threshold voltage of the long channel device minus
the threshold voltage of the short channel device, in the case of a long channel device ok.
This, it is almost the total charge is almost it is going to be qNAxd, that is a charge per unit
696
area into the total area which is all L has not changed too much. It is only a small percentage
of L, that has changed. We can also be more accurate. Let us also do that, but let us just finish
this first.
In the case of a short channel device these differences, if have the trapezium here and with
one side as being L dash, the other side is being L. Now, L’ is significantly less than L
whereas, in a long channel device L’ was almost the same as L. So, that is the key difference.
697
Now, since L dash is much less than L, because a larger percentage of this region is depleted
by the PN-junction or this region forms a larger percentage of the, total length. Here, you
have the area to be considered as the area of the trapezium. So, that qNAxd (WL+L’/2), in that
is the area of the trapezium, divided by, I mean L+L’/2 is basically, your total charge, in the
depletion region in the short channel device.
So, what is the difference between the Vt of a long channel device and a short channel
device. It is a difference, in this depletion charge minus this depletion charge divided by the
total capacitance. So, its capacitors per unit area into WL ok, and this minus sign is there,
because of this minus sign is present, because of the all the charge is being negative.
698
So, this total depletion charge is qNA/Cox into WL x d divided by WL. So, that is this term
and q total depletion in the short channel device divided by the total gate capacitance is given
by this term. So, essentially this reduces to an expression of this kind and if L’ is less than L,
this term is greater than 0, which means the V t of the long channel device is greater than the
V t of the short channel device or in other words, if you reduce the channel length the V t
reduces ok.
Now, if you want to be more accurate, let us just go through that exercise, as well although it
is not written here, we will just work it out. Let us say that, you want to say, you want to
consider this trapezium even in a long channel device, that is true, because although here, the
trapezium is more pronounced. It also does exist in a long channel device. It is just that, these
regions are a much smaller percentage of the total length total channel length and that is the
primary difference right.
699
So, what is let us call this as L double dash and capital L, and what is the depletion charge. In
this device it will basically, be your q N A I am just taking the magnitude q N A x d into W
into the area of this trapezium it should be L plus L double dash by 2, and that divided by the
total gate capacitance is essentially going to be. Let us get rid of the WS its essentially going
to give you qNA xd /Cox(1+2Cox/ L).
700
So, that is the q depletion of the long channel device in a accurate manner accurately ok. So,
L say, L is 10 microns and this is say of the order of 1 e or 0. 1 microns and therefore, L
double dash is 9. 9. 8 microns. So, this number here is basically, 0. 0. 9 8. Whereas, if my L
was much smaller ok. So, if on the on the other hand, if my L approaches, say 1 micron and
this still remains to be 0. 1 micron. Because, this doping concentration has not changed and
my L dash for a short channel device approaches 1 minus 0. 2 which is, 0. 8 micron. Then my
L dash by L is a smaller number and therefore, it is going to have a larger impact and further
if we further reduce L.
So, let us say, we make L as 0. 4 microns, as we, we are beginning to approach the depletion
thicknesses, then my L’ should now become 0.2 microns. So, it will be L minus 0. 2, because
each of these regions is 0. 1 and therefore, my L dash by L has now become half, so much
more significant. So, you can see, that as L decreases the impact of this term begins to show
up to a greater extent ok, and therefore, the therefore, the, threshold voltage starts becoming
smaller and smaller as my L decreases.
701
So now, what about the impact of scaling on the of the channel width ok; so, we have your
we have the MOSFET which is got the it is the gate insulator semiconductor and you now
have, we have discussed the impact of channel length, but what about reducing channel
width. So, for reducing channel length, we say it is a long channel device to a short channel
device, but for channel width, we say it is a wide device to a narrow device.
Now, let us take the cross section of a wide device ok. We are going to look at this device
from this side ok. We have always been looking at the device inside the plane. We have
always looked at this cross section, but now let us start looking at that cross section ok. So,
that cross section is shown here ok. So, here I have taken a cross section right, in the centre of
the device. So, you cannot see the source and drain. We have a bypassed area somewhere, in
the middle of the channel and you have your gate metal you have the insulator, that is a semi
conductor and that is the depletion region and you have oxides on all these side ok. On the
side walls it is all insulator and the source is somewhere, outside the screen and the drain is
somewhere inside the screen.
So, we are looking the channel the channel lies in and out of this plane. So, we are looking
through the device in this direction and therefore, this is the channel width of the device or
this is the intended channel width of the device. Now, what is the impact of reducing this
channel width? As we scale down, impact of reducing the channel width on the threshold
voltage of the MOSFET, so, if the channel width; so, let us look at the depletion profile.
702
So, once again the answer has got to do with the depletion charge. So, total depletion charge.
So, if you look at this depletion profile, yes, most of this depletion, if you look at the total
depletion charge all depletion charge spans over a, length of W and of course, the channel
length to say, L and therefore, WL is what is depleted, but then, because of the fringing in the
field ok.
So, you do have some field lines ok, heading towards the sides and therefore, due to this
fringing of the field you are going to end up, whether you like it or not we are going to end up
depleting some excess charge on the sides and this excess charge is not really contributing
too much to the behavior of the MOSFET. But, it is simply that if you add if you add some
charge in the gate you are going to end up not only depleting this region, but you are also
going to unintentionally deplete the regions on the sides, because of the fringing of the fields.
703
Now, for a very wide device this extra work done, really does not make any difference ok. It
is a very small percentage of this total channel width. So, this W’ is a very small percentage
of W, but for a short channel device it becomes a very significant portion ok.
So, let us say let us say if, let us say, we intended to add some positive charge on the gate and
quickly deplete this region and after that, when we add some more extra charge, we expect
inversion ok, or the definition of V t, but now since the extra charge so now, since, whatever
charge is being added there is also compensated for, by these regions on the sides right.
704
So, I still need to add more charge and once again, that is also compensated for this and only
when you add a lot more charge well the depletion region finally have been fully depleted
and then, allow inversion. So, you need to add a lot more charge in the gate, because these
regions on the sides are helping or, helping the semi conductor respond to this applied charge.
And therefore, you need to do a lot more work, before you get inversion and therefore,
reducing the channel width increases the V t and it is again a percentage question right.
So, this is my channel width W, these regions have now become significant compared to this
W and therefore, the V t actually increases. And, if you want to see the mathematics of it, you
can say that, let us say, what is the difference between a wide channel device and a narrow
channel device the wide channel device is got what is a total depletion charge. Again, we are
going to measure the charge in coulombs and not per unit area or volume.
705
The total depletion charge in a wide channel device the total depletion charge is your q NA xd
into WL and that divided by the total gate capacitance is nothing, but C ox W L, what is the
total depletion charge in a narrow channel device. Now, the side regions cannot be ignored.
So, even here the side regions are present, but W plus anything on the side is approximately
W. This is this excess, charge is very-very miniscule compared to the total charge in the gate,
but now it is not ok.
706
So, in this case, it is going to be qNAxdCox WL into W Lx d, which is this region which is
your regular which is the regular region that you have already, that you need to deplete. But
we are now depleting some extra regions, and we are going to model it model this extra
region not like this little triangles, but you know model, it along these fringing lines the
fringing electric field ok. So, that is it is going to have a more of a circular profile. So, you
are going to have the circular arcs. So, although I have shown it as, although I have shown it,
as you know, straight edges just for this just, because it is easier to draw.
707
You have a gate you have an insulator here and this insulators are sharp if you look at this
edge of this insulator. It is a sharp a corner and this corner is going to have a very strong
fringing field. So, that is the way the field lines are going to be present it is going to have a
very strong field. And therefore, this is the depletion region, it is you can say it is more like a
sector of a circle and then here you have this region that is depleted.
So, that corresponds to this and this, corresponds to this. So, if you take that little accuracy if
bringing, that little accuracy in the picture you will find that, we use the area of that little
sector to define this excess area being depleted that divided by W L, that divided by C ox into
WL is the is this term. So, what is the difference between these two terms? The difference
will now be a negative number, which implies that my threshold voltage for the wide channel
device is going to be less than the threshold voltage for a narrow channel device.
So, in summary as we scaled down L the V T also decreased is the transistors, turning on
faster and why is that, because the PN-junction is helping is aiding the depletion. But, if you
scale down W the V T increases and why is that it is, because you have to do some extra work
in depleting some regions on the sides of the channel ok. So, I hope this qualitative and
quantitative explanation helps out, but if there are any questions do feel free to send me an
email.
708
lkjSemiconductor Devices and Circuits
Lecture – 43
Scaling of MOSFETs - Continued, Leakage currents in MOSFETs
So now, what are the impact of scaling hap? So, now, let us look at some parasitics carry
a multiplication effect ok. Now what happens? So, this is a very interesting phenomena,
which is a there is a possibility for positive feedback here, which can actually destroy
your device. So, if you make the channel length of the device, very small the electric
field between the source and the drain starts increasing ok. So, this electric field becomes
larger and larger.
So, let us say we have used constant voltage scaling and we have reduced the channel
length of the device ok so, V is a constant voltage scale. Now this high electric field
makes these electrons travel at a very high velocity. So, in fact, the electrons will now be
traveling at the saturation velocity.
709
And due to this very high velocity there is a possibility that these electrons can impact
other silicon atoms and can free electrons and holes ok. So, you can have impact
ionization they can knock off electrons and holes from silicon atoms ok. So, now if an
electron and hole pair is created if an electron hole pair is created here, the electron
would quickly run away migrate through this and enter the drain ok, it will just simply
contribute to the IDS.
But, what does a hole do? The hole sees a pn junction here ok, it sees sorry sees a p type
bulk and an n type drain that is bulk and let us say the hole was created here is going to
run upward and into the bulk ok. So, the hole is going to migrate away into the bulk. So,
if the hole migrates into the bulk ok, if let us say, there are several holes and electron
pairs created and a lot of these holes start migrating into the bulk.
710
They are essentially going to create a hole current here. So, essentially you are going to
have a current, flowing inside the bulk and this current is because of the holes. And the
bulk has a certain resistance ok, because it is doped n type it is doped p type. So, it is
going to have a resistance of resistivity of 1/q𝜇pNA and depending on the geometries is
going to have a certain resistance.
So, there is a certain resistor here and you are going to have a current flowing through,
this resistor and what does this current do? It is going to increase this potential here so,
the larger the current the larger the potential ok.
So, essentially we have a whole current flowing through the bulk. The bulk has some
resistance and therefore, this potential of the bulk begins to increase if this whole current
increases. So, if V B, which is the potential the bulk starts to increase the VBS starts to
increase and what happens VBS starts to increase? VB starts to increase, this junction gets
forward biased and it is going to reduce the threshold voltage.
711
If VBS starts to increase is going to reduce the threshold voltage and if you reduce the
threshold voltage, you are going to encourage more carriers, more electrons to get in
from the source. And this is going to further create more electron hole pairs and this
positive feedback mechanism can simply lead to an instability, which will eventually
destroy the device.
So, there is a possibility of this happening, if the device is not very carefully engineered
by using correct scaling principles. So, the next parasitic due to scaling is something
called as drain induced barrier lowering and the idea here, is you know, if you have a say
long channel MOSFET. So, here were looking at band diagrams for long channel
MOSFETs ok. So, the drain and source are far apart and the case you are looking here,
the case 1, we are looking at is here, you have a source electrode, there is the body and
here is the drain electrode.
712
And if you think of this pn junction, you have a pn junction here and you have another
pn junction here and in order to inject carriers, this is at equilibrium, this is when there is
no VDS applied. And there is no VGS applied and when we have to, when we increase the
drain to source voltage, all that happens is that all these barrier heights remain the same,
but the drain to source voltage increases and therefore, you see that, this pn junction is
now strongly reverse biased and when we apply a gate voltage.
So, this is the case, when VGS is still 0, but when VDS is greater than 0 and now when we
apply a gate voltage. So, when we bring in the gate voltage, when we make VGS greater
than 0 and VDS greater than 0, we lower the gate voltage helps to lower this barrier and
therefore, encourages electrons to come in from the source. So, that is the normal
operation of a MOSFET. Now when we head towards short channel MOSFETs ok, the
drain starts getting very close to the source. So in fact, if this is the depletion layer of the
source, then you can see that drain depletion starts immediately and that is the band
bending ok.
So, the band bending at the condition when VGS is equal to 0 and VDS equal to 0 for a
short channel device would look, somewhat like this of course, I have used straight lines
simply because, it is easy to draw, but you could imagine a very strong interaction
between the depletion region at the drain side and the depletion region at the source side.
And without the application of VGS so, this is all at VGS equal to 0 and VDS equal to 0, if
VGS is still kept at 0 and VDS is made greater than 0, then what we would expect? Is a
band bending of this kind.
Now, as we see in long channel devices, we do not want VDS to actually influences
barrier height near the source side. But since these 2 depletion regions are so, strongly
coupled because of the short channel, you see that when VDS is greater than 0, the VDS
itself, the VDS field itself influences, this barrier height. And this barrier height is
lowered because of VDS, without VGS being applied and that is a problem because, now
what we are doing is? We are taking control away from the gate, the VDS itself controls
the barrier height. And this is something called as drain induced barrier lowering, where
the drain to source voltage lowers the barrier for carrier injection and you start seeing
electrons get in even, when VGS is 0.
So, this is a very significant you know in some sense, it is a leakage mechanism, it can
lead to leakage because, if you look at the IV characteristics. So, here is an IV
characteristics for a very low VDS and the moment VDS is increased to something that is
quite significant. So, let us say here, the VDS is increased to a larger value, you find that
there is significant change in the currents, the device turns on at a different point and it
also produces a larger amount of current.
So, this is something called as drain induced barrier lowering and in order to alleviate
this problem, what happens is you know the how do we you know get the gate to have
more control? So, you want to not only have the gate having an interface with just one
surface layer, but you know you come up with architecture such as surround gate or
double gate MOSFETs etcetera, to sort of reduce this impact of the drain and allow the
gauge to have a significantly larger control on the carrier injection.
713
So, with that we you know cover these topics on the impact of scaling on different
parasitics and we will now look at our studies on the impact of scaling on mobility and
then head towards the other topics on MOSFETs. Now, with regards to mobility there
are 2 influences right. So, you have two fields in a MOSFET that control the carrier
movement, you have the gate and then you have the source and drain voltages.
Now, if the gate voltage is increased, you increase the interaction of these carriers of
these electrons with the insulator. And this becomes particularly strong, when the
insulator thickness is reduced during scaling and the drain to source voltage also;
obviously, impacts the I know the movement of the charge transport of the carriers and at
low fields the velocity, the drift velocity of the carriers is defined by the mobility. So, if
you remember the plot of drift velocity versus electric field at low electric fields, you
have this almost linear relation with the electric field with the drift velocity being mu
times the electric field and this was the relation that, we used to derive the current
voltage characteristics of the MOSFET.
But, in short channel MOSFETs it is very likely particularly, if say constant field scaling
is not deployed and if you say, if you use constant voltage scaling it is very likely that the
velocity saturates as because, the field is very large. So, the current voltage
characteristics is, then determined by the saturated drift velocity, that is Vd sat and not.
So, much by the mobility and. In fact, the current voltage relation here, this is drain to
source colored is given by basically, the total charge present in the channel into Vd sat
ok. So, this is the charge per unit area into W into length per unit time will determine the
current through the drain and source.
Now, if the channel is reduced to such an extent. So, if you take your drain and source
terminals and you start bringing them closer and closer ok. So, you reduce L to such an
extent that L becomes smaller than the mean collision time, mean collision length of the
carriers, then we enter into a very different physics of operation. So, what we are saying
is that normally, these electrons would when you when you keep an electron in vacuum,
it experiences a force and therefore, it accelerates. But in a semiconductor lattice the
electron undergoes all these scattering events, it call as it scatters with the lattice defects,
714
it scatters with the lattice atoms, it scatters with the I mean you have a electron scattering
etcetera.
And therefore, because of all these scattering events, we saw that the electron
experiences a affective drift velocity and as the electric field is increased this drift
velocity saturates, because the electron is still undergoing these scattering events. But, if
you reduce the channel length and these scattering events, you could say that they have a
mean collision time of t col and a mean collision length of l collision. If the channel
length of the MOSFET, that is if this distance between drain and source is reduced to less
than, the mean collision length then, it is very likely that a large number of electrons, do
not on average experience any scattering and these electrons would simply shoot past
without experiencing any scattering.
So, we have brought in the channel lens to within you know these mean scattering
lengths. And therefore, in these regions the electrons could accelerate and in this case,
the transport is called as ballistic transport. So, under very high fields electrons could
move so, quickly that they can also enter the oxide. So, you have very high velocity
electrons, they have got very high energy and one consequence of that is you have an
insulator here and you have your semi conductor and these high velocity electrons are
got enough energy to get into the oxide and they essentially become your oxide trapped
charges.
So, these are all the consequences of having a having a combination of short channel
effect as well as having high fields and their impacts on the mobility of carriers. So, as a
final topic with regards to the physics of MOSFETs, let us just look, you know just sort
of let us very quickly look at the different mechanisms of leakage currents in MOSFETs
ok. So, leakage currents in MOSFETs is an exhaustive lecture by in itself and because,
we constrained with regards to time we cannot talk about this so, exhaustively but
although, we will touch upon certain topics particularly things, like space charge limited
current, we will have a special lecture on that. But as such we will only spend this one
slide on understanding the different mechanisms of leakage currents in MOSFET.
715
Now firstly, why are leakage currents, why the study of leakage current so important? It
is important because, when you have MOSFET and when the MOSFET, when let us say,
you have a N channel MOSFET, when VGS is greater than VT, the MOSFET is said to be
on and you are supposed to have a current through the MOSFET. And when VGS is less
than VT, we technically want the MOSFET to be turned off, we want all the currents in
the MOSFET to be as low as possible.
But in reality that is not the case ok, we do have leakage mechanisms, we do have
leakage of charge through between the drain to source even, when VGS is less than VT
and these mechanisms contribute to leakage currents. So, what are the different leakage
mechanisms? So, first let us look at leakage through the gate ok.
Now if you look at gates if you look at the insulator, the insulator was in place to make
sure that the input resistance to the MOSFET was theoretically infinite at least at low
frequencies ok. But in reality you do have the insulator leaking charge; you do have a
small amount of current from the gate to the semiconductor.
So, what are the mechanisms by which a current can exist through this insulate? The first
mechanism is something called as space charge limited currents. It is a drift driven
charge transport, but it is called space charge limited, because depending on your
structure ok. So, depending on the trap state density inside the inside the insulator or you
know depending on the quality of the insulator, you could have different mechanisms of
transport, where the transport is limited by space charge which is in this case, the
electrons themselves ok.
So, the idea is this, you have an electric field across say 2 plates and you inject a lot of
electrons inside ok. Now, what is the electric field felt by an electron here, and what is
the electric field felt by an electron here? Now if there was only 1 electron present in this
region, we would say that the electron experience, a certain field and therefore, it is got a
certain drift velocity. So, if this is a semiconductor, it is got a certain drift velocity which
is mu times the electric field and this contributes to the current.
716
But, on the other hand, if there are several electrons present and this electron these
electrons from a very sparse pathway between these two terminals, then you need to
solve Possion’s equation here ok, we need to write Possion’s equation, in that region to
say that, the electric field distribution in that region is defined by the charge distribution
in that region. And therefore, note all electrons will see the same field and therefore, this
current is determined by the space charge, which is basically this charge distribution in
this region and therefore, it is something called as space charge limited current.
Now, we will have a special lecture that identifies different means or mechanisms and
models for space charge limited currents, but this is one of the methods of charge
transport through the insulator. The other method that you are familiar with is tunneling.
So as we saw since the insulator is a barrier ok. So, let us say you have your metal
semiconductor let us say, we have turned off the MOSFET so, the gate voltage is 0 or
717
less than 0. And the MOSFET is supposed to be technically off ok, but you still have
electrons, these electron wave functions can exist inside through can penetrate this
barrier and can these electrons can appear on the other side and therefore, this leads to a
current through the insulator and that is due to tunneling.
So, this is another mechanism of charge transport, through the insulator the next
something called as hopping and this is again something a topic that, we will briefly talk
about when we when we have a lecture on disordered semiconductors.
So, the idea is you have an insulator and let us say the insulators defective. So, you
actually have energy states inside the band gap of the insulator and these energy states
can act as little homes for electrons of any carriers and the carriers can hop between
these energy states. And this hopping can be encouraged by the field and it is determined
by the energy level differences between these states as well as the spatial distances
between these states.
So, there are many mechanisms that determine hopping ok. So, that is another
mechanism of charge transport through the insulator and finally, you have mobile ions
which can also contribute to gate currents. So, these are the different mechanisms by
which through which you have currents through the insulator.
718
Now when we scale down the MOSFET the idea is to scale down. W, L as well as t ox
and when the thickness of the oxide is scaled down by a factor of k mechanism such as
tunneling and space charge limited currents are all supposed to increase and they will
increase quite significantly ok.
So, what is the option? So, we want to improve field effect, which means if you look at
your current voltage relation, where you have mu 𝜇CoxW/L into (VGS-VT)2 , we would
like to improve this parameter here, this Cox. And we are attempting to do that by
performing scaling so, we have C ox is 𝜀 ox by tox and you scale down tox and therefore,
scale up Cox, but scaling down t ox will result in these leakages and how do we overcome
that we say that we do not scale down t ox, all the way we instead pick a dielectric, which
is got pick an insulator, which is got a high permittivity and we improve 𝜀ox and
therefore, improve Cox. While trying to keep tox same, the same or while trying to scale
719
down t ox, just a little bit and in this way we can still improve, your field effect and at
the same time not given to all these gate leakage mechanisms. .
Now, the gate leakage mechanisms effectively reduce the input impedance, they do not
keep the input impedance infinite and they make it quite finite and this is got
implications, when we design circuits.
Now what about leakage mechanisms between the drain and source? So, you have the
drain and source electrodes in a MOSFET and when the MOSFET is off ok, when the
gate voltage is supposed to be less than VT the idea is you do not have inversion layer
formed. So, there is no inversion layer, when the MOSFET is off and theoretically, we
would like that the current be 0 or in other words, if you remember the the switch or
sending a signal from A to B, when the switch is open, we would like the resistance of
this MOSFET of the switch to be ideally infinite ok.
720
So, which means that you do have a drain to source voltage, there is a VDS that is
greater than 0, if VDS is 0 and a VGS is 0, there should be no current in the MOSFET, that
is nothing driving any current mechanism. But a VDS is greater than 0 and VGS is less
than VT or you know if you keep the VGS at a point, where the MOSFET is supposed to
be turned off then we do hope that the current is 0, but that is definitely not the case and
that is because of many leakage mechanisms through the transistor between drain and
source.
So, what are the different leakage mechanisms? The very first and in fact, the most
significant is sub threshold leakage. So, we have already seen that, when VGS lies in this
region V fb less than VGS less than VT, we have a few carriers injected in the near the
source side and these carriers diffuse over to the drain side. And we have something
721
called as the sub threshold conduction and we looked at this in some detail and we
defined the sub threshold slope etcetera.
So, that becomes one of the leakage mechanisms ok, you do have sub threshold currents.
Now apart from that, let us take a look at the drain bulk interface ok. So, you have the
drain, you have the body of the semiconductor and you have this pn junction, which has
got a band bending of this one. So, all leakage mechanisms can this pn junction is
reverse biased and near the source side, it is not depending on the body to source
potential.
Now, all mechanisms that contribute to currents in a pn junction also contribute to

currents in a MOSFET, that is turned off and what are those mechanisms? You do have
diffusion across from the t type body and the N type drain So, you do have diffusion
currents and in the depletion region, you will have a generation of electrons and holes,
because the depletion region encourages generation in a reverse bias pn junction diode.
And therefore, these electrons will now drift over to the drain side and these holes will
drift over to the body and then you do have something called as you have something
called as band to band tunneling, which is essentially the tunneling of carriers from the p
side to the n side ok.
So, these are that is another mechanism and finally, you have something called as gate
induced drained leakage ok. Now this is something that, you are probably hearing for the
first time ok. So, what is gate induced drain leakage? So, let us take the condition when
Vg is less than 0 and Vd is greater than 0 and let us now focus on the gate drain overlap
region.
722
So, you have your P type body here and the gate is negative or you know it is turned off
let us say so, the gate voltage is kept at less than VT and essentially it is much lower than
the drain voltage and you have a VD that is greater than 0. And the hope is that there are
no leakage currents through the source and drain, but, if you watch the gate insulator
metal or a gate insulator n type semiconductor region what you have is a metal insulator
semiconductor contact and with the metal having a lower potential. So, this would lead
to several depletion region.
So firstly, you have the depletion region on the p side and the N side here, because of
this PN junction diode between the body and the N doped drain and then you will now
also have a depletion region that exists in below this gate. And any carriers generated or
any carriers present here, will drift through all these electric fields ok. So, you will have
all the electrons ok. So, if you look at this depletion region it looks like that. So, all the
electrons will now start running away to the drain side and all the holes will now start
running away to the body and this causes an another leakage mechanisms.
These are basically carriers that are present under the gate overlap ok. So, it is the same
basically, the same mechanisms that would contribute to depletion the currents due to
depletion regions in this pn junction diode, but the only thing is this depletion, we are
talking about is now big under the gate overlap region and it is because the gate drain
field ok. So, this is something called as gate induced drain leakage and that too leads to a
small leakage current.
So, normally without gate induced drain leakage your currents should be somewhere else
and with gate induced drain leakage, your currents are a little higher so, you are basically
having leakage. Now why are all these why I am so worried about leakage? So firstly, it
completes our MOSFET models right. So, we looked at let us draw the IDS VGS
characteristics of a MOSFET. So, we now have a characteristic is actually very well
drawn here.
So, if so if you look at this characteristic, we initially had defined models for only the
above threshold region, which is all your square law characteristics. We then refine the
723
model to define the sub threshold region, where you have this exponential dependence
on V0 and that is why a log ID. So, if you look at this is a log scale the log ID versus VGS
characteristic, show a straight line in the sub threshold region it is a straight line. And
now we are going one step further and refining the model by defining all the leakage
mechanisms on the MOSFET.
So, you have leakage through the gate at gate and insulator and you also have leakage
from source to drain and all these leakage mechanisms are important, because they
contribute to power you know the loss of battery life.
So for example, if you are using these MOSFETs and circuits and let us say, you have a
circuit and this MOSFET is supposed to be turned off ok. So, we think the MOSFET is
turned off and there is supposed to be 0 current or we hope that there is 0 current despite,
there being despite the MOSFET being connected to a power supply. And therefore, we
expect that the power consumption of the MOSFET, when it is turned off is the VDD into
the current through the MOSFET, which is 0.
But because, you now have a leakage current since I is not 0 and you have all these
leakage mechanisms; you have static power consumption in a MOSFET circuit. And this
power consumption exists even, when all the MOSFETs are turned off and this is going
to drain the battery charge because, it is going to continuously drain this much of power
from the battery, it is going to drain this much of charge from the this much of energy
from the battery at this power. So, therefore, leakage is important and as I said this is all,
we will talk about with regards to leakage because this itself could form an entire chapter
or you know detailed set of lectures by itself but, we are short of short on time.
724
Lecture – 44
MOSFET characterization: Parameter extraction
So, the final bit with regards the very last topic with regards to MOSFETs is with regards
to characterizing MOSFETs. So, I felt that this would be a very useful module to have
which is the characterization of MOSFETs because in most of your since this course is
being offered to students from various disciplines. It is very likely that you are probably
making MOSFETs you are fabricating MOSFETs using new materials or new process
techniques etcetera. And in order for you to characterize these MOSFETs; we have gone
through several ideas through this course, but I do not think it is quite obvious as to how
to use those ideas to characterize MOSFETs, ok.
So, these few slides, we will just give you some insights as to what you might need to
look out for when you have you know built a MOSFET and you know what you might
want to test for after you have built a MOSFET. So, that it gives you some very
elementary picture on some very useful or important parameters of the MOSFET that
you could use for your research or for your for modeling.
725
So, essentially let us say you have made a MOSFET, ok. The very first characteristic
characterization technique is the most essential, which is using your IV and CV
information to extract key parameters of the MOSFET.
So, what is the point of parameter extraction? So, you know, if we want to extract
parameters right and we want to extract parameters via experiments. And what is the
point of extracting parameters by experiments and what are these parameters? So, ideally
the point of parameter extraction is to be able to define your current voltage
characteristics in static and in dynamic that is you know low frequency and high
frequency characteristics of a MOSFET as accurately as possible, ok.
And we essentially want to identify all these parameters mu Cox W over [laughter], V T
so, that we can then start writing equations for the MOSFET and start defining the
behavior of the MOSFET. Now after we have built a MOSFET what are the experiments
that need to be performed in order to extract these parameters and start using these
parameters to build models so, that we can then predict the behavior of MOSFET circuits
or behavior of different, you know MOSFETs with different you know mobilities or
oxide capacitances, etcetera by simply using these ideas, ok.
So, what I mean is you can extrapolate the behavior of other devices that you make after
you have characterized your fabrication process flow, ok. So, that is the whole point of
parameter extraction, you can build compact models and you can predict the behavior of
MOSFETs. So, what are the experiments? So, the very first one is what we will talk
about is the IV and CV measurements and I think this itself would be quite useful for
most of you.
So, you have made your MOSFET, ok. So, we will just show three terminals, we will not
worry about the body. And this analysis is true for you know any other kind of fet
structure you know we have not talked about say for example, thin film transistors yet,
we will cover that. But these ideas are valid for any other tft stuff, any of the MOSFETs
that is what and the, the key experiments to perform the very first set would be your IV
and CV measurements.
So, you have your MOSFET, you got a drain, source and gate. Now, we want to first do
experiments to extract the transfer characteristics out and the output characteristics. And
then perform experiments to extract the CV characteristics of the MOSFET. Now, how
726
do we do that? For transfer characteristics, we want to obtain a plot of IDS versus Vgs.
And this is done by setting by setting the drain to source voltage to a certain value and
then connecting, you take the MOSFET to a probe station you add in your probes, ok.
So, make sure that these probes are all connected to high impedance you know, if you
are going to measure voltages they need to be connect it to high impedance nodes and we
are going to measure currents you want it to have low impedance. So, you typically want
to be connecting these probes to a good semiconductor parameter extraction system, or
you know very good quality source measure units. And since we are going to be
measuring very low currents in the MOSFET, you probably want to be using Tri axial
cables for current measurement, ok.
So, after making this connection, you want to be varying the gate to source voltage after
setting for a certain drain to source voltage. And we will end up with a characteristic that
looks like this. And then we would vary the drain to source voltage and again perform
the same experiment so as to get another characteristic. And we can now obtain a family
of curves for different drain to source voltages. So, this is the transfer characteristic.
Now you can take all this data and plot it on a log IDS versus Vgs scale. So, you should
end up with a bunch of curves that look like this. The next IV characteristics is the output
characteristics ,which is the IDS versus VDS curve. So, we would now take the MOSFET,
set the VGS to some particular value say greater than VT, we still do not know what VTS,
but you set it to a particular value and we are going to vary that gate voltage and you are
going to sweep the drain to source voltage.
So, when the drain to source voltage is swept, you now have a characteristic that looks
like this. If the VGS is below VT, you will probably have a characteristics that is does not
show too much of current and for larger gate voltages, we will obtain a family of curves
which form the output characteristics of the MOSFET, ok. And finally, we will perform
a capacitance voltage measurement, ok. And how do we do that?
So, ideally you want to we would like to do that on a MOS capacitor structure that is a
metal, insulator, semiconductor and back contact. But if we have a MOSFET, we make
use of the overlap capacitances. So, you have a gate, you have the drain source and we
have a small overlap capacitance between the drain and source.
727
So, if you consider this structure here, we have the gate insulator. So, how do we
perform the CV characteristics on a MOSFET? Now, ideally we would like to perform
the CV characteristics using a metal insulator semi conductor architecture. But we can
obtain such an architecture by simply short circuiting the drain and source making sure
that there is no current flowing and then using the moss capacitor in the semi conductor,
in the MOSFET.
So, at the end of it, we should obtain a family of curves that look like this. So, here I do
not, I have not shown a family of curves, but we are we are seeing just one set of one
data set, ok. So, here we have the transfer characteristics, we have the transfer
characteristics plotted on a log scale. Here we have the output characteristics and here we
have the C V characteristics of the MOSFET. So, how do we extract all the parameters
required by using this IV and CV measurement?
So, the first point note is let us write down our equations, these our theoretical models
help us a long way, ok. So, in linear, in deep linear mode operation so, we have the
MOSFET characteristics defined as Mu Cox W over L. Let me just make use of the
entire slide here. So, we have the current voltage characteristics and linear operation to
be Mu 𝜇Cox W/L into Vgs-VT-V ds /2 .Vds.
Now, if Vds is made much, much smaller than (VGS -VT) then we can ignore this term
and say that this is our current voltage characteristics. Now, the first thing you could do
728
is once you have these characteristics, you can simply take all these data points, note the
appropriate voltages plug them into these curves and just solve those equations to obtain
the unknown parameters. And what are the unknown parameters? The unknown
parameters are the mobility, Cox and threshold voltage. Vgs is known because we are
applying the gate to source voltage, the aspect ratio is known because we have designed
the MOSFET, ok.
So, we know the channel width and we know the channel length what remains and Vds is
known because we are performing the characteristics by setting it to a certain VDS. So,
what remains is 𝜇Cox and VT. And we are going to use all these characteristics to define
all these plots to define these terms and we will use it to cross check and reconfirm these
definitions. Now, as I said you could take each point here and plug it in to the equation
and obtain the parameters. But then an easier way to do it, is to perform the VDSI VGS I
mean that perform the IDS VGS characteristics at very low VDS and make use of this
equation here.
So, the slope at any point on this characteristics is simply 𝜇CoxW/L into Vds, ok. And if
you were to extrapolate, if you were to take a tangent if the VDS is really low, we will
find that this characteristics is quite linear, you will not see too much of non-linearity. If
you take a tangent, let say some point here you know why we are picking a point here
because that is where Vgs is the largest and that is were its most likely that Vds /2 is going
to be much less than (VGS -VT).
So, you take a tangent at that point and you extrapolate the tangent all the way to the x
axis, ok. And the point at which it meets the x axis is a very likely value for the threshold
voltage, why is that? If you set your Ids as 0, you find that that can happen when your
Vgs is equal to VT. So, that is the x intercept. So, from this characteristics what we can
obtain is this product which is Mu into C ox and this value which is a likely value for the
threshold voltage.
Now, what we need to identify 𝜇Cox separately is the CV characteristics or the MOSFET.
So, let us now perform a CV characteristics of the MOSFET and it could be low
frequency or high frequency CV characteristics, it does not matter, but the point should
note is unlike in the case of a mos capacitor, ok. The high frequency CV characteristics
of a MOSFET will occur at a higher frequency. So, what I mean by that statement is that
729
if let us take a CV characteristic for MOS capacitor. So, we have the capacitance and
voltage curve and what we have done is we first perform the low frequency measurement
and we obtained characteristics that looks like this.
So, let us say we performed another measurement, but it is at a higher frequency and at
some point let us say when the frequency was say 1 megahertz we obtained the high
frequency characteristics of a MOS capacitor. But it is very likely that if you were to take
a MOSFET, ok. And perform the high frequency CV characteristics on the moss
capacitor and the MOSFET, it is very likely that at the 1 megahertz point that we still
obtain the low frequency characteristics. And why is that? It is because the high
frequency characteristics flattens out in this region, simply because the electrons
thermally generated electrons in the bulk are not able to respond to the change in the
applied signal.
The generation rate of electrons is not fast enough and see a MOS capacitor has to rely
only on the thermal generation rate, because all the electrons are coming in only from
this generation mechanism. But on the other hand in the case of a MOSFET the source
can also inject electrons. So, you have the source and drain, the source is also it is also
possible for electrons from the source to enter the channel.
And therefore, it is likely that these electrons will appear at the interface at on time and
that these electrons can respond to the applied AC signal. And therefore, it is possible
that you still see a low frequency measurement in a MOSFET at the same frequency
corresponding to a high frequency curve of a MOS capacitor. Nevertheless, so, let us say
we have a MOS capacitor or a MOSFET and we have performed the CV measurement,
ok.
So, we end up with the CV characteristics that looks like this now where do we obtain
our information of Cox? Clearly the C ox can be obtained in accumulation mode in this
case. So, that is our Cox, ok. Now, you should be very careful yeah because the CV
measurement as performed on a semiconductor parameter analyzer we will give you the
complete capacitance, ok, it is the capacitance in farads. It is not C ox it is not the
capacitance per unit area.
So, in order to obtain C ox, we need to take this capacitance and divide by the area of the
MOS capacitor used and then we obtain Cox. So, in this case, let us say the MOS
730
capacitor used had a 100 microns by 100 Micron length lengths. And therefore, the area
is 100 Micron into 100 microns which is 1e4 into 1 e-8 centimeter square, ok. And
therefore, Cox in this case would be 20 Pico Farad that is 1 e -12 divided by 1 e minus 4,
which is about 1 e -8 Farads per centimeter square that is about 10 Nano Farads per
centimeter square.
So, that is the value of Cox in this example. So, we need to be careful with how we use
the CV data. So, now, we have Cox and we have the product of Cox𝜇 and we can extract
𝜇. So, just as a complete process, you know let us add numbers in a little later, but let us
just go through the techniques first, ok.
So, we now know what is 𝜇, we know what is Cox and we know what is VT. And all this
has been obtained from these two characteristics. Now, what else can the CV
characteristics tell you ? The CV character is a very powerful tool, it can be used to
identify through a sequence of experiments, you can use to identify craft charges in the
dielectric, ok. We will talk about an example of that in the next slide you can use it
identify Cox, ok. But you can also use it to identify the typical flat band voltage.
So, we know that the flat band voltages in and around at this point here, ok. And it is also
used to identify the semiconductor capacitance, ok. And what do we mean by that? So,
this is the depletion and the interfacial capacitances combined, ok. And how do we do
that at the low at the high frequency characteristics here, we know that this is a series
731
combination of Cox and the semiconductor capacitance. And that is basically the
capacitance here which is 10 Pico Farad, divided by the area is and we said let us say it is
a 100 micron by 100 Micron device.
So, this divided by your 1 e- 4, 1 e- 8 is nothing, but oh sorry, 1 e-4 is nothing, but the
series combination of C ox and Cs and what is Cs? Cs is nothing, but your depletion
capacitance plus interfacial trap capacitance. And if interfacial trap capacitance does not
exist then it is purely your depletion capacitance, ok.
So, at high in high frequency characteristics, it is very likely that the interfacial trap
capacitances will the interfacial trap charge will not respond. And therefore, it is very
likely that this is simply the depletion capacitance. And if you know the depletion
capacitance, we essentially know Q we essentially know epsilon 𝜖 S by x d. So, that is the
depletion capacitance per unit area and if 𝜖 S is known we essentially know X d which is a
depletion width in the semiconductor and since depletion width depends upon 2𝜙 f. And
among several other parameters, we actually know the doping concentration of the
semiconductor.
So, CV characters is quite a powerful tool, you can make very good estimates of all these
parameters, right.
732
So, now we, we have obtained VT, Mu and C ox and we can now plot the same transfer
characteristics on a log I D versus VG scale in order to obtain look at the sub threshold
behavior of the MOSFET. So, we know VT, so, we know that in and around let us say
about 2 Volts, ok. So, after 2 Volts, the MOSFET enters above threshold operation. So,
here it is all given by a square law characteristics and below this below this we know that
we are in sub threshold region.
So, here you can see that even though there are a few data points, it is an almost linear
dependence. And we can now calculate the the sub threshold swing of the MOSFET and
what is the sub threshold swing, it is the Volts per decade. So, how many Volts did it
require? What was the gate voltage required for the MOSFET to travel 1 decade up in
current, ok.
So, this is 2 decades up in current and the gate voltage required there is about like if we
do not have a gate voltage scale well we have a scale here. So, the gate voltage required
is about 1 volt for the current to travel up 2 decades. And therefore, it is 1 volt divided by
2 which is 0.5 volts per decade as the sub threshold swing, ok.
So, that is how you calculate the sub threshold swing and sub threshold swing is also
related to 1 plus C depletion by Cox and if there is no interface strip capacitance, we can
calculate C depletion here. And we can cross check with the C depletion extracted from
this place. And if there is a mismatch, it is possible that you have interfacial trap
capacitance. And therefore, it is also an indicator for it is also means to combine these
two characteristics to extract the interfacial state capacitor.
Because the interfacial trap capacitance will not respond at high frequencies, it is less
likely to respond at high frequencies. So, this is one useful parameter, you can get from
the log ID versus Vgs characteristics. And the other useful parameter. Since, you are
looking at the log scale it is possible to identify the leakage current and by plotting a
family of curves at different Vds, here we can identify the gate induced drain leakage
drain induced barrier lowering and these phenomena, ok. You can identify the different
leakage mechanisms and we know the impact of Vds on leakage and it becomes a very
useful tool to study the mechanism of leakage.
So, we know the leakage current. So, in this case for this one characteristics, it is difficult
to identify the mechanism, but we can say that the leakage current is 1 nano Amps. So,
733
that is the leakage current through the MOSFET. So, this is when the MOSFET is
supposed to be off and yet it is having a current that is not 0 and that leakage is 1 Nano
Amps, ok. So, now, finally, we will plot the output characteristics. So, since we already
know 𝜇Cox and VT, we can cross check to see whether the output characteristics in linear
mode matches the use of these parameters. And you can iterate between all these plots in
order to see whether we have the right parameter set.
But N saturation mode, we observe for two other features in the saturation
characteristics, I mean the in the output characteristics is quite powerful for a circuit
designer.
So, in saturation mode we firstly, identify the point at which saturation occurred. So,
technically this should be (VGS-VT), but it could be offset by some amount because if you
remember the bulk charge theory predicts a different saturation voltage as compared to
your square law the simple model.
So, this could give you indicators as to those to those corrections involved, and it is also
cross check between this VT and the VT obtained in the transfer characteristics because at
it only at (VGS-VT), that you enter saturation. The next point is to look at the variation in
current due to VDS in saturation, ok. Now, this variation in current due to VDS is
theoretically not supposed to exist in an ideal MOSFET.
734
An ideal MOSFET the current supposed to be constant with Vds, but it is because of
channel length modulation that you actually have a variation in current in the MOSFET.
And therefore, this is a useful plot to identify your channel length modulation parameter
and the effective output resistance of the MOSFET, ok.
So, what do you mean by the output resistance of the MOSFET? It is the output small
signal resistance of the MOSFET which means that if I were to fluctuate, if I were to set
this as my Dc operating point and if I were to fluctuate Vds around this point, what is the
fluctuation in the current. And that slope or the inverse slope of that tells you the output
impedance of the MOSFET which is to say that if you have an ideal MOSFET with no
channel length modulation in saturation operation there is supposed to be no current in
this MOSFET.
But we then use this ideal MOSFET and we then add a parallel resistor a resistor in
parallel. So, that when the MOSFET operates in saturation, the current here is 0. The
small signal current is 0 whereas, the small signal current in the resistor here is not 0.
And it is in fact, defined by the current variation in saturation mode, ok. So, just as an
example so, since we have already performed a few calculations, just as an example, if
ever run through these numbers a bit let us just quickly do that. So, these are the, these
are very powerful experimental techniques to extract all these parameters.
735
So, let us just run through it and through in a few numbers in place, ok. So, this is quite
important from the point of view of example problems or you know your assignments,
because you will be given assignments with regards to this kind of parameter extracting
parameters.
So, firstly, let us say we performed the most capacitances on MOS capacitor having 100
microns by 100 Micron. Those are the dimensions, which means that, that was 100
microns that is 100 microns and you know that was the MOS capacitor that we used, ok.
So, as a metal insulator semi conductor back contact. So, what is 100 microns by 100
microns, we would like to represent everything in terms of centimeter square.
So, what is 100 microns into 100 microns? It is 1 e 4 into 1 e- 8 centimeter square which
is about 1 e- 4 centimeter, sorry, 1 centimeter square. and the capacitance acts, they come
to the total capacitance measured here there is a Cox in accumulation mode is 20 Pico
Farad.
So, this is all in Pico Farads that is basically your 20 e -20 farads. So, therefore, my Cox is
essentially this capacitance divided by the area which is 20 e minus 12 divided by 1 e
minus 4 which is 20 e minus 8 farads per centimeter square, ok.
And the capacitance here that is basically your Cs in parallel with C depletion is much
lower, it is of the order of 10 or maybe less than that, it is about 8 Pico Farads. And
736
therefore, my Cox in series with C depletion is 8 Pico Farad by 1 e minus 4 which is
about 8 e minus 8 farads per centimeter square, it is about 80 Nano Farads per centimeter
square, ok.
So, do forgive me, if I am making typos here, but I think it is right. So, it is this is about
200 Nano Farads. So, I think, when we had the discussion on CV measurement earlier, I
made a mistake, it was one order lesser, but what this tells you is Cox is 200 Nano Farad
per centimeter square.
So, now, let us come to the slope. So, what is the slope here? So, let us say this point
here is about 1.5 e minus 4 that is the current, that is a 10 volts. And we will go long way
because it is quite linear, there is a quite linear characteristics. And let us say this was all
plotted at Vds is equal to 1 volt. So, over here, we have a Vgs of 5 here, we have a Vgs of
10 and here the current is about 5 e minus 5,ok or let us say 0.5 e minus 4.
So, the slope that is your 𝛿ID by delta 𝛿VG, is essentially 1.5 minus 0.5 which is 1.0 e- 4
divided by this voltage difference which is 5 volts. So, this slope is about 2 e- 3, ok. And
that is supposed to be our ( 𝜇CoxW/L)Vds. So, let us say Vds is 1 volt. So, that is 1 volt,
let us say the aspect ratio is also 1 that means, the channel width is equal to the channel
length, ok.
Now, we know Cox is 200 Nano Farad per centimeter square, 200 e-9 per centimeter
square. And that is all and that into 𝜇 is equal to 2 e-3 which implies that, that is going to
be e 6 and that is going to be e 4, which if I have not made a mistake here say. So, the
mobility is quite high it is in fact, dramatically high and it is about 10000 centimeter
square per volt second, I am quite sure, I have made a mistake here, but, but nevertheless
that is the way you go about calculating this, ok.
So, let us just cross check this 1 e minus 4 by 5 volts, that is 10 e- 3 by 5 which is 2 e
minus 3. And the capacitance was 200 Nano Farad per centimeter square, ok. So, since
we have made an assumption on the aspect ratio, the aspect ratio let us say is 10. And let
us say the Vds is 1 volt .and let us say we do not know Mu and therefore, this slope must
be equal to this number, ok. So, which means that 6 and your mobility will be about
1000 centimeter square per volt second, ok.
737
So, yes numbers do not make complete physical sense because we are just arbitrarily
picking up Vds and aspect ratio. So, there is no typo, it is just a mobility of 1000
centimeter square per volt second is more acceptable considering the fact that we are
using silicon as our role model, ok. So, these curves are by no means actual data sets,
they are just plots prepared on an excel file using the parameters using a certain, certain
number of parameters and using these models, ok. But, it is just for the sake of this
exercise.
And then let us look at the log Id versus log Vds characteristics we have already looked
at this.
So, we saw that the sub threshold swing is in order to get this past 2 decades, you need
about 1 volt. And therefore, it is about 0.5 volts per decade. So, that is the swing and let
us look at the output resistance or the MOSFET. So, here we can see that the small signal
output resistance; you know the 𝛿VDS/IDS is you can calculate that quite easily. So, let us
just mark two points, let us say that we keep one point here at 0.0015 and the next point
here which is actually which (Refer Time:36:43) and the next point here which is close
to the edge which is say 0.0018 which means the difference is about 0.003,ok.
So, the difference in the currents is the order of 3 e minus 5. So, that is the N 𝛿I and that
happens for a voltage difference of about 3 volts, ok. So, that is quite a large output
resistance, but not large enough. So, it is about 100 kilo ohms, ok. So, this is the way you
738
go about extracting parameters, if I made a mistake in these calculations that I am doing
here, I do apologize. But you have the idea you know what has to be done. So, this is a
very powerful tool to extract parameters.
739
Lecture – 45
MOSFET characterization: Trapped charges, contact resistance
So, next let us come to using the CV characteristics to extract trapped charges and
mobile charges in the oxides. So, we looked at the charges in. So, you have your metal
insulator, you have your gate insulator semiconductor back contact in the MOS structure
and you have your source and drain electrodes there. And we saw that there is a
possibility of different kinds of trapped charges in the insulator.
So, we could have electrons holes which are your oxide trapped charges and then we
could have mobile ions, which are due to poor handling of the substrate etcetera etcetera.
So, how do we is there a possibility to identify the kind or the nature of the trapped
charges in the insulator. So, what we do here is we can use the CV characteristics tool as
a very powerful tool. And I strongly recommend this textbook, but Dieter Schroeder
which is called semi conductor materials and device characterization ok. It is there in the
reference list, but I just thought its a good textbook to you know sort of to redefine over
here, because it talks about different characterization techniques for semiconductor
characterization in general. And this includes the characterization of mobility you know
740
via different techniques, its the hall effect measurement, etcetera etcetera, and it goes all
the way to interfacial trap charges and techniques for MOSFET characterization.
So, it is a good textbook to look at. But once again you know, this is again a very
exhaustive topic in a just characterizing insulator track charges, and we do not have the
time and nor is it the goal of this course to focus on this aspect for so long.
But I will just give you an example as to why I tell you that CV is a very powerful tool
ok. So, if you would remember the impact of trap charges it was influencing the flat band
voltage ok.
741
The flat band voltages normally this, but then due to the presence of trap charges we had
the possibility of having; we have the possibility of having additional term which was
gamma times Q trapped by C ox, but gamma would take a value of 1 or 0 depending on
whether these trap charges were present close to the insulator semiconductor interface or
whether they were present at the gate insulator interface.
So, the trap charges were present at the gate insulator interface gamma was 0 by gamma
was this mean position are estimator ok. So, it was giving you just giving an indicators to
the position of these charges. And if when the charges were close to the gate were close
to the insulator semiconductor interface, gamma was equal to 1.
So, these were the two extreme values of gamma. So, when the charges are present here
they do not impact the flat band voltage and the charges are present here they impact the
flat band voltage very significantly. And flat band voltage in turn impacts the threshold
voltage of the MOSFET.
So, let us say we want to understand the mechanism of this trap charges in the insulator.
742
So, let us say we are seeing mobile charges in the insulator, but we do not know whether
its because of electrons and holes or whether its due do mobile ions. So, let us say that is
the problem. So, we have a semiconductor, you have an insulator and you have a gate
and the question is, I do see charges trapped there, but I do not know whether these are
oxide trap charges which is electrons and holes or whether it is mobile ions.
So, how do I go about identifying this? So that is that is an example. So, what we do first
is? We will perform a low voltage CV characteristic, which is your regular CV
characteristics. So, let us say we do a high frequency CV characteristics ok. And the CV
characteristics turned out to be this. Now what we do is, we take out the device and we
apply a very high voltage; very high voltage applied across the MOS capacitor. So, if
you apply a very high voltage if there are mobile ions.
743
All the mobile ions will drift towards the semi conductor insulator interface; that is they
will all take up a value of 𝛾 equal to 1 ok. And if they all take up a value of gamma equal
to 1 then the flat band voltage will be significantly influenced and what would happen to
the flat band voltage.
So, since these are all positive charges ok. It is like as though without the application of
gate voltage. We have already applied a gate voltage ok. So, we are talking about an N
MOS devices. So, it is a P type bulk. And these positive charges near the gate insulator
interface are playing the same role as adding positive charges on the gate. So, in some
sense they have already started inverting or already started creating negative charges
inside the semiconductor. And therefore, if we push the mobile ions to the gate insulator
interface and to the semi conductor insulator interface, the flat band voltage will reduce.
And now if mobile charges of the cause the flat band voltage would have reduced. And if
you now perform a CV characteristics at low voltages, you will find that the CV
characteristics are shifted a little bit to the left. And this is an indicator, it is a potential
indicator of mobile ions being present in your MOS in your insulator.
On the other hand if it is oxide trapped charges what would happen? So, let us say the
cause of tap charges is not mobile ions, but it is oxide trapped charges. Now once again
we initially do the low voltage CV measurements ok.
744
What do you mean by low voltage CV? Simply means that the voltages here are not very
high ok. So, let us say we just do your regular CV measurements and now when we
apply, we again apply a very high positive voltage.
But now since it is electrons and holes that are responsible, let us say it is not mobile
ions, but electrons and holes what would happen? If you have a very positive gate,
voltage electrons would be injected into the insulator; so the insulator is weak and
therefore, its housing electrons inside. And once electrons are injected into the insulator
we now perform another CV characteristics ok; we now remove the high voltage and
perform another CV characteristics. And now since it is electrons that have been
injected, the flat band voltage would have increased and we will see that the CV
characteristic shifts to the right. And why did the flat band voltage increase? It is because
even before we applied any positive charge, we already have some negative charge in the
insulator.
So therefore, any positive charge you apply on the gate we will first have to compensate
for all this negative charge on the insulator before achieving flat band and it is only after
that that any additional charge, positive charge you added on the gate will lead to
inversion. So, it is as though you have increased the threshold voltage of your MOSFET
and therefore, your CV characteristics will shift to the right because the flat band voltage
is now increased.
745
So, that is a very clear indicator as to how the CV characteristics can be used to identify
mobile ions and oxide trap charges? You can also do a cross check ok.
So, what we do is, let us say we apply a very large positive bias and now perform CV
characteristic, you then remove the large positive bias and perform CV characteristics.
So, you will find that if it is mobile ions. So, let us say this is the normal, this is the mid
characteristics and now we will apply a very large positive bias and perform the CV
characteristics and because if it is mobile ions they are all shifted to the left. And now we
apply a very large negative bias we take out the device and apply a very large negative
bias ok, which means these mobile ions will now start drifting towards the gate insulator
interface ok. If it is mobile ions they will all start drifting to the gate insulator interface if
we perform, if you apply very large negative bias and when you do the measurement
again the curve would have shifted back to where it was.
So, this is the mean position that is without insulator trapped charges ok. We do not
know this and therefore, we are oscillating around this point and we apply very large
positive bias the curve will go this way and we apply very large negative bias, the curve
will go the other way. On the other hand it will go the other way in the sense it will get
back to where it was the threshold voltage will not become hugely large, it will not
become heavily positive ok.
746
But the other hand if its oxide trapped charges when we apply very large positive bias
and then perform low voltage CV measurement. We would have injected electrons and
you will find that the curve has shifted to the right and then when you apply very large
negative bias you will inject holes into the device. And then when you perform CV
characteristic, the curve would have shifted to the left of the mid position and therefore,
if the curve CV characteristic curve swings widely on either sides, it is very likely that
you have oxide trap charges and not mobile ions ok.
So, these are just some examples to illustrate how the CV characteristics can be used to
make these measurements? Now there are other techniques if you want to look at mobile
ions you know you also perform IV measurements and you do this at very high
temperature. So, for example, if you sweep the gate voltage, you sweep the gate voltage
implying that you are changing the field here slowly increasing the field and different
mobile species will respond at different velocities to the increased field and why are we
sweeping the gate voltage. We not only want to create a field, but we also want to
measure the current through the capacitor ok.
So, we will have a steady displacement current and if there are mobile ions that drift in
the mobile ions will give some blips in the current and these blips will exist, because the
mobile ions start moving and then they reach an interface and then they stop moving and
therefore, the current dies out and it will give you the current will get back to the
747
depletion current. So, with this you can identify whether the ions are positively charged,
negatively charged you know what is there, what is that drift velocity etcetera?
So, it is a again another powerful tool to identify the presence of mobile ions. The other
important parameter with regards to trap charges in the oxide is the interfacial trap
charge density. So, how do you identify interfacial trap stains ok? So, there are many
techniques for this, because it is a very important measurement, there are many
techniques and they are very powerful transient techniques.
And I once again recommend the textbook I mentioned for studying these techniques,
but since we have been talking about CV curves, here is one simple technique ok, which
may or may not work all the time, because you are heavily dependent on the nature of
the traps, but the idea is to perform a low frequency CV measurement and then perform a
high frequency CV measurement. So, in a low frequency CV measurement at this point
you will see the influence of the series capacitance of Cox C depletion and the interfacial
trap capacitance, because the charge is trapped in the insulator can respond to the low
frequency AC signal. But at high frequency it is unlikely that the interfacial trap
capacitance will respond and we will only see the influence of Cox and C depletion. And
from the high frequency measurement we calculate, from the accumulation mode region
we calculate what Cox is, from the high frequency measurements, we calculate what Cs
748
is, because Cox is known and then we use Cox and Cs to calculate the intubation trap
capacitance from the low frequency measurement.
So, this is a technique that works for several devices particularly if it is a disordered
semiconductor, but and in disorder semiconductors you will not see a CV curve that
looks like that. So, there are other things that we need to do. We will see the impact of
frequencies creating different shapes and the CV characteristics.
749
So, it is a technique that can be used to estimate interfacial trap capacitance, but then
there are other powerful techniques also available first. And yet another option is to
measure the sub threshold slope, because the sub threshold slope has got this term of C
depletion+Cinterface by Cox built into the definition of the sub threshold swing.
So, this is yet another technique to measure the interfacial trap capacitance. So, that is all
with regards to the extraction of trapped charges in the dielectric. And the one last topic
that we will look at which is you know which is last, but its quite important is the contact
resistance in a MOSFET.
So, this is the very last topic with regards to the device physics of MOSFETs and it is
also the last topic with regards to the characterization techniques for MOSFETs. And the
point is got to do with regards to the presence and estimation, you know the detection of
the presence of any contact resistances in the MOSFET and what do we mean by contact
resistance?
So, you have your MOSFET structure, you have your drain and source and the idea of
the MOSFET is to have a current flowing between drain to source when it is on. Now it
is possible that this pathway from the metal contact through the semiconductor to the
channel has got some resistance and this resistance is, it is preferable that this resistance
is ohmic, but if you end up having significant barriers, this resistance will not be ohmic
and we have issues with regards to the contact. But it is preferable that this resistance be
750
ohmic and if it is ohmic how do we go about extracting that, how do I go about
measuring or characterizing that contact resistance?
So, it becomes a very useful extraction process to feed back into your process flow
because in most of the cases problems with regards to power MOSFET performance can
be attributed to poor contacts ok, in most cases therefore, it is a good characterization to
perform and to get information with regards to the development of your source and drain
electrons.
So, this is a very notorious problem. So, its any student whose trying to fabricate
transistors will experience problems with regards to contact resistance and it can end up
with the transistors showing very poor performance. Now how do we extract this ok. So,
first let us go to the theory of the extraction.
751
So, what we will do is we need to measure the I V characteristics of the MOSFETs. So,
we have the gate source and drain and the point is we are going to apply, we are going to
take this MOSFET to a probe station and we are going to apply a drain voltage or source
voltage and gate voltage to the MOSFET.
Now, after applying the source drain and gate voltage the MOSFET we are measuring
the current the drain to source current in the MOSFET, but if there is a contact resistance
present these are not the voltages the MOSFET actually sees. The MOSFET in fact sees
is a different voltage, it sees the voltage across this point and across that point. So, we
call that voltage is V D’ dash and V S’ or we know V D prime and V S fine. So, that is the
voltage we have applied and that is the voltage that we think that is the drain voltage we
think the MOSFET is seen. But in reality due to the presence of contact resistance the
MOSFET is seeing a lower drain voltage and it is seeing a larger source voltage.
Because if you have a current flowing through this device, you have a current flowing
through this device, it is going to result in a voltage drop here and therefore, V’D dash is
going to be V D-IDS r c that is what is given here. And VS’ dash is going to be at a higher
potential with respect to V S and it is going to be V S+I DS rc where, rc is an ohmic contact
resistance that is why we are using this relation to determine the voltage drop across the
resistor.
752
So, this is the situation, we think that the MOSFET is seeing this and we are going to
start modeling or MOSFET parameters based on this current voltage characteristics, but
in reality because of the contact resistance the MOSFET is seeing a different drain to
source voltage, it is seeing a smaller drain to source voltage.
753
So, how do we extract rc? How do we extract the drain to source voltage?
So, let us take the actual, let us write since we are developing a theory let us say that the
actual gate to source voltage seen by the MOSFET is VG-VS’. VG is there VG seen by the
MOSFET will be the VG you applied because the current is too small. There is ideally
there should not be any gate current. If you are seeing significant gate current then
contact resistance is not your fundamental problem. You need to sort out the insulator
first ok.
754
So, the V GS as seen by the MOSFET is VG-VS’ and the drain to source voltage seen by
the MOSFET is VD’-VS’. So, let us plug in these real parameters as seen by the
MOSFET into the MOSFET equation, this MOSFET equation should be correct. So, you
have mu Cox W/L, the MOSFET is in linear mode of operation its 𝜇CoxW/L+ and V GS-
VT is VG-V’S -V T and V DS-V T V DS is V’D -V’ S /2 and the whole thing into V’D -VS’.
So, this is the current voltage relation of the MOSFET, but now we will rewrite VS’ dash
in terms of VS which is VS+I DS rc and V’D as VD-IDS rc. So, we have 𝜇CoxW/L into V GS -
VT-I DS rc in minus and this is the VD’ dash minus V S dash. We have V DS-2I DS rc by 2
and the whole thing into V DS-2 IDS rc.
So, you find that this minus I DS rc term cancels with this plus I DS rc ok. So, these two
make it a plus I DS rc and the two divides and these two terms cancel off ok. So, what we
are left with is the actual ideas in the MOSFET is m 𝜇CoxW/L(VGS-VT-V DS/2).
So, this term actually remains unchanged, but the contact resistance influences this term
here, it is V DS-2 IDS rc and now if you simply divide everything by I DS and you know
you separate out this equation, just perform some algebra we want to rewrite that
equation, this equation here can be rewritten as this expression here which is V DS/I DS is
equal to this term, 𝐿/𝜇Cox WV GS-V T-V DS +2 rc ok.
755
So, we have made the rc, the contact resistance component come out as a separate term
and now what we do is we build MOSFETs. So, let us say we are developing a process
flow where and we want to get the contact resistance right. So, what we do is we develop
MOSFETs of different channel lengths ok.
So, we have a wafer with MOSFETs having different channel lengths. So, let me just
wrote like that ok. So, essentially what this means is this is the layout of the MOSFET.
So, you will have your source, you will have the drain you will have the gate and that is
the channel lengths.
So, we will have MOSFETs having different channel lengths and we will perform an IV
measurement on all these MOSFETs, in linear mode and then we will take the data and
we will plot V DS by I DS for different V GS as a function of channel length. So, we will
probably have say five different channel lengths and therefore, we will have these five
points, these will be your data points for the different V GS. And we when we plot these
data points we will end up with all these theoretically speaking they should all lie on a
straight line, and why should they lie on a straight line? Because we are plotting V DS
versus I DS versus channel length and you can see that it is linearly dependent and that
would be the slope.
So, the slope would be depending on V GS and the intercept the Y intercept will always
be constant and its 2 rc. So, if all your devices are well built and they all have the same
756
contact resistance then we will find that all these straight lines of this plot will all
intersect at the intercept and it will give you a value of two times rc. And therefore, it is
possible to extract the contact resistance which is your rc which is on the drain and
source side.
Now, it could so happen that you end up with a plot that looks like this ok.
You might have the three lines. Firstly, they might all be lines that is given, but the three
lines intersect over here, they do not intersect at L equal to 0. They in fact intersect at
757
some other value of L, what does that mean? It means that it is very likely that we are
having a mask error that is when you are doing a photolithography to fabricate your
devices it is possible that there is an error in the mask. And what we think as L is
actually L-∆L or you know minus or plus ∆ depending on whether this point is lying to
the right or left.
So, that is indicated above from mask error. So, but this is a general idea behind extract
the; behind the process of extracting contact resistance. So, this is the experiments that
you need to perform to extract your contact resistance in the device. So, with that we
conclude all our discussions on the MOSFETs and I hope you will find this very useful.
And from this point on we will change the flavor of the course and we will start we will
move from device physics and towards the talks on circuit design with the MOSFET as
the prime example.
758
Lecture - 46
MOSFET as a switch
We have now approached the last unit of this course, which is the unit on the circuits
aspects of all the devices that you have learned looked at so far, ok. So, if you just let us
say summarize our journey ok, it is we started off with some ideas on quantum
mechanics. And we looked at some interesting properties like the wave particle, the wave
function etcetera. And we use those ideas to develop you know concepts on how carriers
would behave in a solid, so electrons in a solid.
And we looked at that the formation of the band gap, what the valence band is, what the
conduction band is very briefly. And we use those ideas to then develop a solid structure
for device physics. And here we studied the basic properties of semiconductors. We
looked at the formation of junctions, and we look at some of the devices you know
particularly the metal semiconductor junction, the P-N junction; and in transistors, we
particularly paid attention to the BJT and the MOSFET.
And here we developed huge set of tools ok, such as the Parson’s equation to study the
behavior of all these devices. And throughout the study, they always tried to finish or
759
wrap up every device or you know every concept with the impact on circuits, right. We
looked at the small signal models of many of these devices.
And we now are ready to make use of all this knowledge to see the circuit implications
ok. And what we will be doing in this unit is essentially is focusing on the MOSFET, we
will not use any other device ok, because circuit design I mean circuit theory or analog
circuits is an entire course in itself, and we are not here to completely explore every
detailed of analog circuits. Although this is a course that I do teach at the graduate level,
but at the senior undergrad level, but this is not the place for a exhaustive course on
analog circuits.
But, instead the point here is to introduce all of you to circuits introduce all of you to you
know, so that all of you at least get a flavor of what circuit design is all about. And most
importantly try to see how different circuit parameters the circuit level circuit level
parameters, how the circuit level parameters are connected to device physics, which in
turn is connected to the behavior of electrons in a solid. So, every time we introduce a
parameter, we will try to hunt or way back to some elementary concepts.
So, we will focus on the MOSFET, and we will explore some very basic circuits,
particularly the use of the MOSFET as a switch, and small signal analysis with the
MOSFET, and using the MOSFET to build an amplifier. And we will only look at a
single stage amplifier, and particularly something that gives you gain, which is
something called as a common source amplifier ok. And of course, once you understand
the methods of using these techniques of small signal analysis for amplifiers, you can
you can extend those techniques to other circuits.
And finally, we will look at two important properties very quickly the frequency
response of MOSFETs, and secondly the noise in MOSFETs. And with that we will
complete this course. So, I hope all of you find this last unit to be useful, because it
completes the arc. It so this so, what we did is we briefly looked at many topics. So, this
is this course is more wide than deep. It looked at many topics from quantum mechanics
to solid state physics, through device physics and engineering. And finally it is going to
terminate with the understanding of circuits.
760
So, let us start off by looking at the MOSFET as a switch. Again, we are going to go
back to a traditional example that we used to introduce transistors, which is I have point
A and point B. And I want to send a message from point A to point B, and I need a
switch in place. So, when the switch is closed, the message here gets to point B; and
when the switch is open, nothing happens at point B. So, here is a more complete circuit.
Now, let us say we want to use the MOSFET as a switch a simple switch. And this kind
of a circuit is extremely useful for memories. For examples, where you have a MOSFET
is a switch and a capacitor, and you write information on the capacitor, you open the
switch, and the information is retained on the capacitor. So, this is a very it is a form of a
memory.
The other applications are your circuits and displays or any active matrix systems,
wherein you have a MOSFET, and you have a capacitor again, and you store and then
you have light emitting materials, for example say an LCD. And you store information
on this capacitor, and you open this MOSFET, and you have the voltage here driving the
LCD. So that you can see the color, the different levels of brightness on your display
screen.
So, this use of the MOSFET is a switch is extremely important, even when it comes to
analog circuit applications. And more than that the use of the MOSFET is a switch is the
761
essence of digital circuit applications ok. So, let us study some of the important aspects
of the MOSFET working as a switch.
So, for a transistor could replace the switch, because you need a control signal, and you
need these two terminals. So, you we have looked at BJTs, but we are not worried about
BJTs in this section of the course. And instead, we will focus on using the MOSFET all
right.
762
So, the first question ok. If I were to use the MOSFET as a switch and I want low I want
low small signal resistance across the switch, do I bias the MOSFET a linear mode of
operation or saturation mode of operation that is the first question. So, just to recollect
what is the linear, what is the saturation mode of operation?
So, you have your current your output characteristics of the MOSFET, you have I DS
versus V DS ok, and let us say it looks like that. And we know that for some particular V
G. We know that when V DS is greater than or equal to VGS-VT, the MOSFET is in
saturation mode of operation. And when V DS is less than VGS-VT, we say that the
MOSFET is in linear mode of operation.
Now, what is the property of an ideal switch, the ideal switch should have a low
resistance when it is turned on or when the switch is closed. So, this is the resistance of
the switch ok. So, this is R ds let us say, and that it is desirable that this resistance be
low.
And now, the question we are asking is do we biased a linear saturation mode operation,
when for having a low small signal resistance small signal resistance. So, what do you
mean by small signal resistance, it is dVDS/dIDS that is you have fixed the drain voltage at
some (Refer Time: 09:37) point, and we are having some current in response to the drain
voltage. Now, if we fluctuate the drain voltage around this point that is we have a small
signal, which we will represent with a small VDS. If we have a small fluctuation here,
how does the current fluctuate, how much does the current fluctuate. So that is what you
mean by, small signal analysis?
Now, where is the small signal resistance lower in saturation or a linear? So, if you were
to draw demarcate the saturation region, and then you draw the linear region here, what
is the small signal slope in the linear region. So, you may have set the V DS, and now we
are going to fluctuate a small signal VDS around this, and we are going to see a
fluctuation in the current in response to that. So, this slope is essentially the resistor. So,
as we move the current voltage up and down, the current moves up and down ok, so the
current changes by that level. And therefore, the resistance here is some values, which
we recall as r linear which is your dVDS/dIDS.
Now, what about in saturation mode, if the MOSFET has got ideal characteristics, and
this is perfect saturation that is there is no channel length modulation. A large fluctuation
763
in the voltage does not fluctuate the current too much that is even if dVDS is very large,
your dIDS is this is almost 0. And what is dVDS/dIDS in saturation, it is going to be
infinite. Therefore, the resistance the small signal resistance for a MOSFET in saturation
is infinite. And this resistance is something called as the output resistance of the
MOSFET ok, whether it is a linear or saturation. This resistance between source and
drain that we are talking about is something called as the output resistance of the model.
So, r o for a MOSFET in linear is finite it is small, and ro for a MOSFET in saturation is
infinite. Therefore, it is definitely preferable to bias the MOSFET in linear mode of
operation, if you are using it as a switch, so which means that if I am going to have to
send data from point A to point B, and I have a control. So, let us say point B is that our
reference, so let us say that is the source. And we have a control signal, which is the gate
to source voltage, so that DC is got a certain DC level here ok. And we have the data
voltage the gate to source voltage, and we have a drain to source voltage.
So, let us say that with time, the information being sent at point A has got the certain DC
value, so that is VDS. And on top of that, we have a small signal information that is being
sent to B. So, it is this information that we want appearing at point B.
So, what is needed is that this DC level ok. So, this fluctuation is very small, so that is
not going to be very large. So, this DC level is expected to be smaller than V GS minus
VT. So, in that case the MOSFET will be linear. But, if this DC level is increased, and if
this DC level is going to be larger than VGS-VT, then we have the MOSFET in saturation.
764
So, so here is the simulation ok, so the answer to this question is we want the MOSFET
in linear. And to prove the point here is the simulation, which is done using AIM spice,
so which is the software ok. And the simulator takes into account this circuit, so we are
simulating this circuit here shown here.
So, the MOSFET we are considering has got a channel width of 100 micron, channel
length of 10 micron, threshold voltage of 2 volts, the mobility that is very low. And we
will; so, basically this is a thin film transistor MOSFET, which has got disordered
semiconductors Cox which is 9.73 e-8 farad per centimeter square. And we have point B
here, which is got this terminal circuit. So, naturally if the MOSFET is open, point B is
pulled to ground, it is kept at ground because of charge draining off through this resistor.
765
Now, what happens, so let us say we bias the MOSFET in linear mode? How do we
know that? So, this diagram corresponds to linear mode. How do we know that this is
linear mode? This signal that you see here, the square response is the gate of this
MOSFET ok, so it is the gate voltage of this MOSFET. And the gate voltage is going
from 0, it climbs to 10 volts, and stays at 10 volts for some time, and comes back down
to 0, so that is the gate of the MOSFET. The source is generally at ground, because that
is my point B. So, this line you see here is point B.
766
And this is the voltage fluctuation at point A. So, you can see a small signal that is trying
to get across ok, it is a dying out sin wave. But, what is the average value of the signal,
what is the DC value of the signal, the DC value of the signal is around at this point, so
which is about 2 volts. And what does the DC value of V GS that is about 10 volts. And
the threshold voltage is 2 volts. So, definitely V DS is less than (VGS-VT), which means
that the MOSFET is in linear mode operation. So, when it is in linear mode operation let
us look at what happens at point B ok.
So, point B is seeing a signal like this. When the MOSFET is got 0 gate to source
voltage, the MOSFET the switch is open, so point B sees nothing, because you have A
that is fluctuating, and then the circuit is open, and B that is connected to ground with the
500 kilo ohm resistor ok. So, point B sees none of this fluctuation that is what is
expected. But, the moment the MOSFET turns on, you can see point B climbing up, and
almost following point A except for a small delay, it is almost following point A, and the
moment the switch closes again point B loses connection with point A, and it goes back
to 0. So, you can see that the signal from A is successfully transferred to B, if you keep
the MOSFET in linear mode of operation.
767
Now, what about keeping the MOSFET in saturation mode of operation ok? Now, this is
the case, where the MOSFET is in saturation mode. How do we know it is in saturation
mode? So, the gate voltage is still the same; when it is on, it is still 10 volts.
But, let us look at the voltage at point A ok. The voltage at point A is the this is 500 kilo
ohm that is point B, and this is point A, this is where the signal is coming in, so that is 10
volts that is V GS. What is V DS? So point A has got a signal that looks like this right.
And what is the DC value of that signal that somewhere here and that is about 9 volts.
768
So, V DS is about 9 volts, V GS is about 10 volts. And what is V GS minus V T, it is about
8 volts. Therefore, V DS is greater than (VGS-VT), which means the MOSFET is in
saturation.
So, now let us look at what happens to point B, is it able to capture the signal. So, this is
point B. You can see that when say MOSFET is turned on, its voltage increases, but it is
not able to so well, I mean so clearly track the message coming in from point A. You can
see there is some very light movements, but generally it is a constant, and then it dies out
ok. So, point B in this case is not able to track the signal from point A, because the
MOSFET is in saturation, and it has got a very high output impedance. So, this is a very
illustrative example, where you say that we need the MOSFET to be in linear mode
operation, and because it is going to have a low output impedance in linear mode
operation. So, this is the summary.
So, in linear mode operation, the MOSFET has the small signal output impedance of the
MOSFET is very low. And in saturation mode operation, the small signal output
impedance of the MOSFET is very high. And therefore, we want the MOSFET biased in
linear mode operation for use as a switch ok. So, we have decided it is linear mode
operation.
769
Now, what is the resistance of the switch ok? So, what do you mean by the resistor, what
is the resistance of the MOSFET switch? So, do forgive this plot having all these scripts,
because this is the image that is transferred from a Mac to a PC, and it generally happens
to have these kinds of errors.
So, what we are trying to show here is the output characteristics, it is V DS versus I DS
or the rest of the details are not important ok. So, the point of this is going to be in the
previous slide, we realize that we should bias the MOSFET in linear ok. Now, the
question is what is the value of the resistance, what is the output resistance of the
MOSFET. So, we are talking about the small signal resistance of the MOSFET.
770
We are not talking about resistance measured as V DS/IDS, no because if it is VDS/I DS,
even the saturation mode in the saturation mode, the MOSFET does have a finite
resistance ok. You have let us say a V DS of 5 volts. And let us say the V GS is whatever 3
volts, and therefore the MOSFET is in saturation.
And let us say there is a current of 10 micro amps, so the MOSFET has got the V DS by I
DS of 5 volts by 10 micro amps. So, you could say that effectively that is the resistance
of the MOSFET, no but that is not what we are talking about, because that is the large
signal behavior of the MOSFET, and that is the non-linear behavior. So, it is really not
appropriate to define a resistance there.
771
What we are interested in is the small signal resistance ok. So, I am reiterating this point
small signal resistance of the MOSFET, which is for a small fluctuation, after you have
set all the DC voltages, you set all the DC voltages. Sources at ground, the gate is at
some voltage, so that you know with respect to ground you have a certain V GS, and the
drain is at some voltage, and with respect to ground you have a certain V DS. And this V
DS is less than (VGS-VT). So, this is smaller than that.
And now, on top of that you have a little fluctuation. And therefore, you have a little
current that is DC, and then you have a fluctuation in the current. And this fluctuation is
defined by the small signal resistance of the MOSFET, so that is the resistance we are
after. So, what is the value of that?
772
So, the value of that, since we have biased the MOSFET in linear mode operation, what
is I DS, I DS is 𝜇CoxW/ L(VGS-VT- VT/2) V ds. And if my V ds is much much smaller than
(VGS-VT), we could say that we are going to ignore this term. And therefore, I DS is
approximately 𝜇CoxW/ L(VGS-VT- VT/2) V ds ok. So, this is something called as a deep
linear, I mean it is a very very linear operation. V ds is much much smaller than VGS-VT.
So, now what is the small signal output impedance, what is the output impedance of the
MOSFET, it is dVds/dIds, which is what 1/𝜇CoxW/ L(VGS-VT). This is nothing but the
slope of the i d v g characteristics in the linear region.
So, since it is deep linear, they are going to set the Vds to be very low, and we are looking
talking about a slope in this region here. So, this slope is given by that is inverse of that
slope if you wish, because that is an I DS versus V ds characteristics the inverse of the
slope, which is dVds/dIds is given by 1/𝜇CoxW/ L(VGS-VT). So, this is the quantitative
estimate of the small signal resistance of the MOSFET.
So, here there should not be a Vds/2 term existing, so that part is probably not very
accurate that is a typo on that slide (Refer Time: 25:06). So, let us now you know play
this game of connecting all these properties to our understanding of the device physics.
So, let us see whether we can use this is the refresher for your device physics.
773
Now, as a circuit designer, you have identified that this is the small signal output
impedance of the MOSFET, when it is in linear mode operation. So, first it does not
make sense. When the mobility goes up, should the resistance increase or decrease. And
the mobility goes up, the carriers move faster, the current increases. Therefore the
resistance should decrease that makes a lot of sense. If Cox goes up, the field effect
improves, the current increases, and the resistance decreases that also makes a lot of
sense.
If the MOSFET has got a large channel width over channel length large aspect ratio, the
resistance should definitely go down, so that makes sense. And if the V T is very small or
(VGS-VT) is very large, definitely you have larger inversion charge, and you have a lower
resistance, so that also makes a lot of sense. But, let us now go ahead and start defining
each one each one of these terms.
What is mobility? You can say we are talking about the output resistance of the
MOSFET. Now, what is the mobility, the mobility is connected to the mean collision
time, it is q t collision by the effective mass of the carrier. What is q; q is your 1.6 into 10
power minus 19 coulombs. What is mean collision time, mean collision time is the time
for the is the mean time between the scattering events of the carrier that is scattering
through the lattice and through the defects etcetera, thereby giving you a definition of
mu, which is your v d is equal to mu times electric field.
What is the effective mass, the effective mass is dependent on the shape of your E K
diagram, you have an energy, you have a momentum or the K space. And let us say you
have your E K diagram say that is let us say that is the conduction bandage, we saw that
this the valley of this conduction bandage can be approximated as a parabola ok.
And why can it be approximate as a parabola, because you are in a classical sense your
energy is P2/2m, and this P is nothing but the momentum, which is also h bar K, so is h
K2/2m. And this m is nothing but m*, where m* is the effective mass. And therefore, this
is a coefficient attached to the K2 dependence or a parabolic dependence, so it is an
approximation here.
774
And what is the effective mass, we identified it from the second derivative of E, so dE by
dK is h2/m into K, and d2E/dK square is h square h2 / m*. And therefore, your m star was
nothing but h bar square by d square E by dK square. So, for therefore, the shape of the E
K diagram is directly going to impact the output impedance of your MOSFET, so that is
your mu right.
What is Cox? Cox is nothing but the capacitance the gate capacitance or the insulator
capacitance per unit area of your capacitance voltage characteristics or your MOSFET
sorry, and that is nothing but epsilon 𝜀 ox /t ox, it depends on the thickness of the oxide.
Aspect ratio is the geometry with which you have designed the MOSFET. V gs is the
applied gate to source voltage. And what is V T, V T is the threshold voltage, it depends
on the flat band voltage, it depends on it depends on the depletion charge divided by
Cox, it depends on 2 ∅F, the surface potential reaching 2∅F.
And what is V fb, V fb is the flat band voltage, which depends upon the metal
semiconductor work function for the gate metal and the semiconductor, it depends on
any trapped charges in the insulator divided by Cox. And what does q depletion depend
on, q depletion depends on qNAxd, sorry it is a very crowded slide here. So, q depletion
depends on q let me just mark it somewhere else, it depends on qNAxd.
And what does what does NA, N A is nothing but your dopant ion concentration in your P
type or N type doped body of the MOSFET. And x d is the depletion width, which in turn
775
depends on the surface potential, and which should be 2∅F. at the time of turn on of the
MOSFET. And 2∅F. again depends on the dopant concentration. So, here is a complete
breakdown of the output impedance of the MOSFET a linear mode with respect to all the
device physics that we have learned so far.
So, this connection is the is the value or the message of all these lectures ok. So, it is the
value of talking about circuits, so that you understand not only how to design the circuit
or what the circuit the implication of the circuit design is in terms of circuit performance
ok, but also what to do and where these parameters came from, so that it helps you debug
at a very fine level. Of course, this is a very it is an introductory course in some sense,
but nevertheless it is got a lot of valuable messages in this in the in this content. So, now
let us move on. We have answered this question, which is what is the resistance of the
switch.
So, we know that we are going to use the MOSFET as a switch, and we know that it has
got a resistance of 1/𝜇CoxW/ L(VGS-VT). Now, here is the next question. So, suppose we
use this MOSFET as a switch in a switch capacitor circuit, which is the switch in series
with the capacitor, and let us say the MOSFET is on, so the now MOSFET is on. And the
small signal circuit or the equivalent small signal circuit can be written as replacing this
MOSFET with a small signal resistor, and here is the capacitor.
776
So, what is the time constant of the switch capacitor circuit, what is the RC time constant
of the switch capacitor circuit. So, what do you mean by RC time constant ok? (Refer
Time: 32:56) people are not I hope all of you are familiar with RC circuits, but just in
case you have forgotten what RC circuits are what an RC circuit is let us just go down
the track for a in a minute.
So, let us see you have a series resistor and a capacitor circuit. You have a capacitor, it is
connect and series the resistor. And let us say you have an input voltage here, which is
appearing as a step. So, you have a voltage versus time, and it is a step input voltage, it
goes from 0 volts to say some voltage Vo naught. And the question is what happens to the
voltage on the capacitor. So, at time t equal to 0, the capacitor had a voltage of 0. But, as
t greater than 0, what is the voltage on the capacitor, so that is the question we want to
answer ok, how do we solve this.
So, since the current through the resistor and the current through the capacitor should be
the same, we say that the current through the resistor is the movement this is turned on at
this point just after a turn on. So, let us say at t equal to 0 plus, the current through the
resistors V naught minus V c which is the voltage drop across the resistor divided by R,
and the current through the capacitor CdVc/dt ok, which is basically the dq/dt through the
capacitor. And solving this differential equation, you will see that your V c is and you
777
apply this boundary condition of t equal to 0, V c is 0, we will find that Vc is equal to V
naught into 1 minus e to the power minus t by RC, you will end up with this answer.
This RC has the dimensions of time right. And therefore, RC is called as the time
constant of the circuit. So, what is the time constant mean. So, technically when t is
equal to RC, the voltage in the capacitor is Vo *1-e-1 ok, but I mean that is only the
accurate quantitative definition.
But, essentially the time constant is a useful parameter that gives you an estimate; it is an
estimate or a measure of the charging time of the circuit you know. If you want to
compare two RC circuits you compare two RC circuits, and you say that this one has got
a smaller time constant than that one, which means to say that this circuit will charge
faster as compared to circuit 2 ok, which means to say that circuit 1 is faster compared to
circuit 2.
So, time constant is a good measure, it is not the total time taken to charge the capacitor
ok. The total time taken to the charge the capacitor is the time taken to bring Vc=Vo
naught, and that is technically infinite, it means infinite time to charge the capacitor that
is not so, RC the time constant is not the total time taken to measure charge the capacitor
is just a measure or an estimate of the delay or the time taken during charging. Is the
capacitor charging fast or slow compared to another capacitor that is the answer it gives
you.
778
So, we need to find the RC time constant. What is the time constant of the switch
capacitor circuit with the MOSFET, so that is the purpose of this slide. So, now after this
discussion, the answer is quite simple. It is simply the output impedance of the MOSFET
multiplied by the capacitance, and that gives you a good enough answer.
So, if you have a MOSFET operating in linear mode operation, and somebody I asks you
is to what is the small signal capacitance a small signal time constant of the MOSFET, it
is simply r o linear into the capacitance of the capacitor, it is an approximate model. And
what is the approximation, we have linearized the MOSFET.
Now, if somebody asks us to what is the exact charging dynamics ok, so do not linearize
it, how do we go about it. We simply equate the current through the MOSFET into the to
the current in the capacitor, which is we have taken into account the non-linearities of the
MOSFET. So, let us say that is the V C that is the input.
So, what is the current through the MOSFET, it is going to be 𝜇CoxW/ L(VGS-VT). This
must be equal to the current through the capacitor, so it is c dV c by dt. You solve this
equation, and you will end up with the exact charging dynamics, which also takes into
account the nonlinearities in the MOSFET.
So, if the voltage is concerned, if the input voltage is all very large, if the step input is
very large, this is the more appropriate equations (Refer Time: 38:38). But, you will find
779
that for most cases, this is not very far. The time constant, you will get in this case will
not be very far from the time constant you get by simply linearizing the MOSFET and to
using that value ok.
So, here is the simulation, and it shows you the charging of the MOSFET, so the gate
voltage goes up, the MOSFET is on. And then when the drain voltage takes up a step like
response, you can see that the capacitor starts charging. And this charging time constant
although it is the simulators using the exact model, you could easily model it using a
much simpler linearized model instead of worrying about solving this differential
equation, so that is the strength of having these kind of small signal models. Of course, it
is going to be very useful, when we study amplifiers etcetera. So, it talks about the
linearization of the non-linear device around a certain operating point ok.
Now, here is another useful circuit, but although this really does not connect us to our
device physics, nevertheless it is a useful circuit for all of you to know ok, which is
basically you have a capacitor that is charged with some charge Q 1. And you have
another capacitor say C 2, which does not have any charge in it. And you use a MOSFET
or any switch for the matter. So, although we are talking about MOSFETs as a switch,
you could use a MOSFET, but it really does not matter, you use a switch of any kind ok,
and you close the switch.
780
The question is what is the charge and the capacitor C 2 ok, what is Q C 2 at time t equal
to infinity. And what is Q C 1, when t is equal to infinite. We already know that Q one C
1, when t equal to 0 is Q 1 ok.
So, I think in this diagram, I have used let us just use the variables, I have used here,
which is at time t equal to 0 q I mean C 1 has a charge of Q initial 1. And at time t equal to
0, C 2 has a charge of Q 2 initial, which is 0. At time t equal to infinite, C 1 will have a
charge of Q 1 final. And at time t equal to infinite, C 2 will have a charge of Q 2 final. And
what has happened here is that the switch has been closed. The question is what is the
charge on the capacitors at time t equal to infinite; so how do we go about solving this?
So, we simply use charge conservation right.
781
So, if you use charge conservation what is the total charge, so the total charge is Q 1 initial
that was the entire charge in the system, because Q 2 initial was 0. So, this is simply, so this
is nothing but 0, and that must equal the final charge in the system.
So, what is charge conservation? So, suppose you have 10 electrons, and you had to take
3 electrons away, and keep them here. How many electrons do you have left here, the
answer is 7. And how did we get that number, because charge is conserved. So, you took
3 away, and you have 7 and you had and you had 10 in the beginning. And therefore you
have 7 remaining.
782
So, so according to charge conservation, you had up get one equation, which is basically
this. And after at t equal to infinite, after you have short circuited this at t equal to
infinite, these two points must have the same voltage ok, they cannot have different
voltages, they are short circuited. And at t equal to infinite, they must have the same
voltage.
So, therefore, the voltage on C 1, at t equal to infinite is Q 1 final by C 1, I must equal the
voltage on C 2 at t equal to infinite, which is Q 2 final by C 2. And using these two
relations, we can identify what Q 1 final and Q 2 final are so. We will leave that to us as an
exercise.
But, here is another interesting detour. So, let us say we said charge conservation is what
we need to use here. So, let us say we use energy conservation. So, what is that initial
energy in the system? So it is Q 1 initial 2/ 2C1. What is the final energy in the system? It is
Q 1 final whole square by (2C 1+Q 2 final )2/2C2.
Now, is it for me to equate the initial energy in the system to the final energy in the
system, can we say charge energy conservation energy is conserved. And therefore the
initial energy in the system is equal to the final energy in the system. If we define the
final energy in the system in this manner, the answer is this equation will not hold. And
why will it not hold, because what we have ignored to consider here is any loss due to
joule heating in the resistor that we used.
783
No matter what the switch and what the resistance of the switch, there will be some
amount of joule heating ok. And in fact, it turns out that that heating or the energy loss
becomes independent of the resistor used ok, which is an interesting offshoot from this.
So, we will look at that also in some assignment.
So, therefore this these two can be equated, the initial energy of the system and this can
be equated, only if we consider the loss due to joule heating in the resistor of the switch
resistance. So, this equation will hold true. And if you actually solve this out, you should
get the same answer as you get for charge conservation ok. So that is a message from this
slide.
784
Lecture – 47
MOSFET as a switch – Continued
So, we have looked at using the MOSFET as a switch. We are going to operate the
MOSFET in linear mode operation, we are we know that the on resistance of the
MOSFET small signal on resistance of the MOSFET is 1/𝜇CoxW/ L(VGS-VT). And we
know that a good estimate of the RC time constant in a MOSFET capacitor circuit is r on
into the capacitance of the capacitor. Now finally, what errors could creep? So, here is a
simulation or for MOSFET trying to write some data onto the capacitor.
So, take a look at this. So, this curve which I show you here in a red line is the voltage on
the capacitor, so that is the voltage on the capacitor. And what do we have here, so here
we have the gate voltage. So, the MOSFET turns on at this point. And you see that
without anything happening the data voltage is 0, so that is the data voltage ok; so that is
the data voltage line. So, the data voltage line is all 0, but the moment the gate voltage
turns on, you see this little blip on the capacitor, so why did this blip occur.
And then the moment the data voltage turns on, the capacitor starts acquiring this charge,
it starts acquiring charge; it gets starts charging up till the data voltage. You have already
785
seen the RC time constant for this. And then the gate voltage turns off here, the data
voltage is still the same. And the gate voltage turns off which means what is happening,
the switch was open initially; and when the switch closed, this was 0, and this still 0,
when the switch closed, we saw this little blip.
And then some data was provided, we provided 5 volts of data here and this capacitor
now started charging to 5 volts. And then suddenly the gate is now open, which means
the switch is open; and here you can see the sudden drop in the voltage that was being
stored, this is still 5 volts, but the moment the switch opens you see that there is a blip. In
fact, the voltage is no longer 5 volts on the capacitor. So, capacitor loses some charge.
So, why does the gate influence the voltage on the capacitor plates? The answer is it is
got to do with two different phenomena, which causes parasitic errors in the MOSFET,
in a MOS in the capacitor for MOSFET capacitor circuit.
So, the errors are because of two different phenomena which are called clock feed-
through and charge injection. And both of them have got to do with the operation of the
MOSFET, when the gate turns on and both these both these mechanisms, they create this
error and the capacitance during the operation of the MOSFET that is the gate turns on
and turns off.
786
So, in order understand clock feed-through, we need to understand this gate drain overlap
region. So, let me just exaggerate the MOSFET structure little. So, we have the gate, we
have the insulator and we and if you remember, the region near the source drain so this
was the let us say the drain and the source regions. The gate had a little overlap with the
drain and source electrodes.
And when we discuss the geometry of the MOSFET, we had marked this overlap region
with L overlap; and this overlap is important for the operation of the MOSFET, because
it provides complete get field control through the entire channel. And it is all right to
have an overlap equal to 0, but the moment you start having overlaps less than 0, then
you have regions in the MOSFET that are not controlled by the gate. Therefore, this
overlap is necessary, but it is also responsible for all this some of this behavior that you
see here.
So, let us look at this cross section here under this overlap. So, what do we have? We
have the gate metal, we have an insulator and we have a heavily doped drain electrode,
which is essentially a parallel plate capacitor. So, essentially you have a capacitor which
is we which we call as the overlap capacitor, and what is the capacitance of this overlap
capacitor; it is basically your Cox, so if you measure this in farads, Cox is the capacitance
per unit area, and area which we need is W*L overlap, so that is the area we are talking
about.
So, this overlap capacitor appears across the gate, it couples the gate signal to the source
and drain electrodes. So, if this were to be your circuit, if this is the switch capacitor
circuit that we are looking at. So, let us say that is the capacitor and we are observing all
the voltages on this capacitor, and we saw that there were some blips coming in and
going out.
So, what the circuit that should actually be drawn, is the circuit with having this overlap
capacitance between the gate to the drain. And now from this circuit you see that when
you have a gate voltage that is fluctuating. So, we saw the gate; when the gate voltage
went up, we saw this blip and the gate voltage came down, we saw another blip. So,
when the gate voltage goes ups up and down you have a transient or time varying voltage
across these series capacitors, which is C overlap and your load capacitance, so that is
your C L that is the load capacitance.
787
So, we have these two capacitors and series and you have a gate voltage that is going up
and coming down. And it is this voltage that is creating a fluctuation on your load
capacitance. So, when the voltage when the gate voltage climbs up you see that there is a
transient and the C overlap is coupling the gate charge to the C to the load capacitor. So,
you have this quick transient and the moment the voltage stabilizes, there is no more
current through the capacitance series, capacitor circuit and the current quickly dies out
to 0, so that is one mechanism. What is the other mechanism, the other mechanism is
charge injection. So, let us imagine there is a puddle of water.
And so let us have a little puddle of water. And let us say a child were to jump into this
puddle, you will automatically see that the water will also splash out and it will all head
to the sides. So, the charge injection mechanism is a bit like that.
788
When the gate goes high so we have a MOSFET when the gate voltage goes high,
inversely layer charge is formed, where did this inversion layer come from, where did
the charge in the inversion layer come from; the inversion layer charge came in from the
source. And the moment the gate voltage goes off, the inversion layer has to disappear.
So, you have a lot of charge sitting here you have the gate, insulator, semiconductor you
have a lot of inversion layer charge sitting here.
And the moment the gate switches off, this is the inversion layer charge has to leave this
place. Now, where does the inversion layer charge go back, it goes back to the source
and drain electrodes, there is nowhere else for this charge to go, so this charge here is
you like a little puddle of water. And the gate turning off is the child jumping in and
trying to get the puddle of water out of that place, and the puddle and this little puddle of
charge they just moves to the source and drain electrodes.
Now, if you were to gently turn off the gate, and give this charge enough time. So, you
turn off the gate very slowly, you turn off the gate very slowly. And you give all this
charge enough time to realize which is the low impedance path, which means with the
line the part which is got a larger capacitor then you will find that most of the charge
goes to that part and very little charge goes to the low impedance or the higher
impedance path.
789
The basically the charge distributes itself, you have a current that distributes itself
depending on the external impedances, but on the other hand if you were to turn off this
gate very rapidly, you do not give enough time for the charge and you will find that an
equal amount of charge exits the source and drain electrons.
So, you now have another mechanism which corrupts the charge on the charge store on
the capacitor. So, we had this capacitor load capacitor, we already saw that by turning off
the gate, whatever charge this gate transient caused a current through this path and it
created an error on the stored charge in the capacitor; but there is another error and that is
the charge in the channel will also exit the MOSFET, and will also jump into this
capacitor here, and it will also corrupt the data on the capacitor.
Now, which of this is a bigger problem? The bigger problem in some sense is the charge
injection, why is it bigger; the clock feed-through mechanism, the amount of error it
creates is fixed if you have fixed your gate voltage, or your select line voltage. If the V
GS is fixed, it does not matter how rapidly the gate transits, it does not matter what your
V DS is, if this is fixed, your error here is fixed.
And therefore, this error can be corrected off as an offset if you know what the gate
voltage levels are, and you know what your overlap capacitance is, and you know what
your load capacitance is, in some sense you can correct this error of as an as an offset, it
790
is an offset error; and you can just overcompensate for this offset, but on the other hand
the charge injection depends is a function of the drain to source voltage.
There is the data that you are applying, that is the data. And this data could vary from
one operation to the other; it could be 2 volts, now it could be 1 volt; the next second, it
could be 3 volts etcetera, etcetera. Whereas the turning on and turning off is generally
constant, you know the MOSFET will always turn on at a fixed voltage, let us say 10
volts to 0 volts, this turn on voltage the MOSFET does not change. You would not keep
the MOSFET in linear mode of operation and so that really does not change too much.
So, this charge injection error, the amount of charge stored in the channel depends on the
drain to source voltage. And the amount of charge exists exiting the channel, also
depends on the speed; the turning on and turning off speed of the gate signal. The gate
signal would turn on slowly, and turn off slowly; you will have a different error. And if it
turns on quickly, you will have half the channel charge entering the capacitor. So,
because of these two reasons its quite hard to predict the amount of charge in the amount
of charge injection, its quite hard to predict it.
So, therefore it is a problem in itself, but if you just look at the quantity so, here you have
the overlap capacitance that is the charge that is coupling through, but here you have the
entire channel region, entire channel charge its coupling through. So, for these reasons
both these, I mean the charge injection is a little bit slight slightly more complicated than
the clock feed-through phenomena, but nevertheless these two are very well studied.
791
And there are several techniques to compensate for both. In fact, for charge injection it is
very useful technique is to have a complementary device so, for example and you
provide this complementary device with the opposing signal. So, let us say the gate goes
on and off here, you know the clock drives the opposing gate. So, which means that
when electrons come out of this holes also, an equal number of holes also come out of
the complementary device which is basically your P-MOSFET, and that is your N-
MOSFET.
And you have electrons and holes and therefore, you do not have any charge injection.
So, there are techniques to overcome these, but nevertheless it is a problem that you
should be aware of.
792
So, with that we will conclude our discussion on the use of the MOSFET as a switch.
So, to summarize let us see what connections we can make with the circuit application.
The circuit requirements of the MOSFET as a switch, the reality of what can be
expected, and the device or solid state, physics or the property that governs this, that
governs this connection. So, here is the connection between circuit requirement and the
device physics.
793
So, the first thing what is the ideal property of a switch, when the switch is closed? So,
let us say what is the on resistance of the switch that is from drain to source; since, we
are using MOSFET we will just say drain to source. Now, what we would really like is
that this be 0 ohms. So, we are talking about small signal resistance, let us say small
signal resistance, we would like it to be 0 ohms; but in reality it is 1/𝜇CoxW/ L(VGS-VT).
And the connection to each of these terms you know, I already elaborated upon each of
these terms.
So, mobility comes from the E-K diagram, and it is dependent on your tcol and effective
mass, so that is that is the connection to the E-K diagram. And Cox is basically your
epsilon 𝜖ox/tox, and V T is basically your flat band voltage which is phi m s plus the
trapped charge in the insulator plus the depletion charge qNAxd/C ox plus phi F which is
connected to which depends on NA; and x d also depends on 2𝜙𝐹 on the surface potential
which is again connected to 2 phi F during turn on; so that was the connection there.
The second point is what is the off resistance of the switch? So, when the switch is
turned off, what resistance would be ideally like, we would like infinite ohms, but is that
the case it is not, because we have leakage mechanisms in the MOSFET. And we already
did an exhaustive study, not an exhaustive study at least we did a introductory study on
all the different leakage mechanisms.
We had leakage through the gate let us keep the gate aside for now. We had sub
threshold leakage, which we found was due to diffusion of carriers and had an
exponential dependence on the surface potential, and exponential dependence on V DS
etcetera. And we had other mechanisms like you know, we had mechanisms like drain
induced barrier lowering, and gate induced drain leakage, and leakage through the p-n
junctions at the source drain contacts, it is the contributing to the leakage currents. And
therefore, definitely the off resistance of the MOSFET is much less than infinite.
Third, what about the turn on speed we want that switch to go from on to off, and to on
very rapidly right, we want what is the turn on speed of the switch. We would like it to
be ideally we would like it to be 0 seconds, but is that truly the case it is not. So, if we
take the MOSFET alone. So, when the gate goes from you know below V T to above V T,
what is the turn on speed dependent on; it depends on the subthreshold swing of the
MOSFET.
794
And what is the subthreshold swing depend on, subthreshold swing is proportional to
1+C depletion plus C, interfacial trap by C ox. So, if your depletion capacitance is large, and
if and which can which can happen if your x t is small, which can happen due to high
doping. And if your interfacial trap capacitance is large, then your sub threshold swing
becomes very large ok; which means that you need a larger gate voltage to make the
current climb up by 1 decade. And then what else did we see? So, let me just find some
space here. So, let us since we have understood what V T is, let me just erase that.
So, what about the turn on speed in a circuit? So, you have a MOSFET RC circuit, the
turn on speed is typically your r on into the capacitance. We would ideally like it to be 0,
but it is not going to be 0; it is going to be r on into capacitance, and we have already
defined what r on is, so that it is the same answers that and is there anything else. So, we
would so these are the key parameters.
So, you would also like other things like for example, we would like know parasitics,
and we would like we would like that the MOSFET be able to take infinite amount of
current, when it is on; and have an infinite voltage drop, when it is off etcetera, but these
are all related to these basic device properties. And you see that the reality is quite
different from what one would desire as a circuit designer, but nevertheless these are
important.
795
And just one more thing though, one more important property which is, what is the input
impedance of the MOSFET so, let me call it impedance, input impedance of the
MOSFET. We would ideally like it to be infinite, so what do you mean by input
impedance we mean that the gate current is 0. We do not want the gate to be the control
signal to be influencing any message that is going in from source to drain. We would like
this current to ideally be 0.
But is it truly the case? It is not, because you have leakage mechanism such as tunneling,
you have space charge limited currents to the dielectric, you have you know other
mechanisms like hopping, and mobile ions, and trap charge and the dielectric etcetera,
etcetera. So, therefore, this two is not close to ideal. And that the key is that you
understand what the values are, and you understand the demise physics and the material
and these solid state properties that control these different aspects.
796
Lecture – 48
Amplifiers using MOSFET
So, now let us look at voltage amplifiers ok. So, we the lectures over the next few slides,
we will have a very strong flavor oriented towards circuit design and we will cover some
very fundamental aspects of how to analyze, and how to build some basic circuits. But,
nevertheless the point is not the point of the scores is not so much in perfecting the
methods of analysis of circuits, but rather trying to understand at least qualitatively as to
what happens and then connect those properties of the circuit with that device physics of
the device being used.
And in this case, we have completely focused on only using MOSFETs. So, it is
essentially going to be connecting the device properties of the MOSFETs with the
behavior of the circuit.
So, let us so in our discussion of an voltage amplifier, we should start off by looking at
what an ideal voltage amplifier is and firstly, what is an voltage amplifier and what is it
used for. So, let us say you have a sensor ok. Now for example, this could be a
microphone that is picking you know acoustic waves, and it is converting it all to
797
voltage. So, you have a voltage across this microphone that is varying in correspondence
to the voice that the microphone is picking up.
And you have an output resistance of the microphone, which is not really important at
this moment. And this voltage waveform from the microphone, it is too weak ok, it
cannot it cannot travel too far without being completely corrupted by ambient noise. And
therefore, it needs to be immediately amplified. And in order to do that we use something
called as a voltage amplifier.
Now, the voltage amplifier has got is, this circuit that you see inside this box ok, and it is
basically got an input impedance. So, you have this voltage appearing across an input
node of the amplifier. And then the input from the input impedance the input impedance
is the place, where the amplifier actually picks your signal ok, it picks the signal there.
And then the amplifier does its magic, and it provides a massive gain A. And then the
output voltage is sent to a through the output node of the amplifier. So, this is where the
amplifier sends it off to the load element or the external environment. So, here you
should have a amplified voltage signal of whatever came inside, so that is the purpose.
So, what are the properties of an ideal voltage amplifier? So, we will in the previous
section, we looked at an ideal switch, and we saw how far how far behind we were in
terms of achieving an ideal switch because of the limitations of the device. So, let us do
the same thing for a voltage amplifier. So, what are the properties of an ideal voltage
amplifier? So, I post these as a series of questions ok. So, what do you think the input
impedance of the amplifier should be ok. Do you think it should be low or should it be
high? So, let us ask ourselves that.
798
So, think of your sensor as a little voltage source ok. So, think of it does it, it is an
alternating voltage source. And it is a got some output impedance right no sensors
perfect, it is got some output impedance. So, essentially this is the circuit the sensor sees.
And we are trying to take the amplifier needs to get a large portion of this voltage source,
preferably it would like all the voltage source to all of this voltage to drop across the
amplifier input, because, this voltage is going to set up a current in this path.
And this is going to lead to resistance drops, so you will have a drop of Z s i and Z in i
across these two resistors. And therefore, there is a going to be some voltage drop here,
and some voltage drop here. You do not want all the voltage to be dropping across the
resistor, which is the output impedance of the sensor. You, in fact want as much of the
voltage being dropped across the input of the amplifier. You want the amplifier to sense
a strong enough input voltage, so that it can work very efficiently.
So, in order to do that we want to make Z in as high as possible right, theoretically we

would like it to be infinite. If it is infinite, there is no current in this path and every
voltage all in the complete voltage drop drops across the amplifier. But, on the other
hand if it is very small, then a large amount of voltage will drop across this output
impedance the sensor and the amplifier will get a very small amount of voltage.
799
So, therefore we want the answer to these questions, we want it to be very very high ok,
so that is the input impedance that we are design. What should be the output impedance
of the amplifier, which is the impedance here. So, the amplifier now received this
voltage, and it amplified it. And now it is going to act like a voltage source, it is going to
play the role the sensor played, but it is going to play that role on the output side ok.
So, now the job of the amplifier must be to take this amplified voltage, and make sure
that the load gets to see all of this ok. And therefore, using the same argument as before,
we want Z out or the output impedance of the amplifier to be very very small. We do not
want it to be large; we want we do not have control on Z l, because that is chosen by an
external circuit. But, we would like we would like that Z naught be so small that no
matter what Z l has chosen, it really does not matter, because the amplifier can easily
submit all the voltage to the lower.
So, we want to keep Z out as small as possible, and we want to keep Z in as high as
possible. What about the gain shall we have a gain less than one, if we have a gain less
than one, there is no point in having an amplifier unless it is used to modulate say Z out
or Z in ok, which is in some sense exactly what certain kinds of amplifiers do, but we
would like good gain at the end of it. We would like good gain that is the point of a
voltage amplifier ok; it is not a voltage attenuator.
800
So, what about the speed of the amplifiers? So, do forgive all if you find there lot of
circuit jargon here it is, because I have taken these slides from a circuit course that I
teach, but you can ignore that. Essentially, the next question is what about the speed of
the circuit, would you like it to be fast or slow what is there how should the circuit
respond as the frequency of the voltage changes.
So, let us say you have a sinusoidal input of that frequency, and you have another
sinusoidal input of this frequency, I get a higher frequency. Would you like the gain to
be it frequency dependent, I would you like it to be frequency independent, the answer is
we would like it to be frequency independent of course, because irrespective to the
frequency we want to get a good amount of gain ok.
So, if we want the amplify, if you were to plot if you what plot the amplifier gain versus
frequency, we would like it to be constant. And should the amplifiers gain vary with the
input voltage magnitude ok. So, what a what happens, if I have a frequency fixed, but the
input voltage varies it just doubles in size, would you like the gain to vary ok. And the
answer is no, because that will induce non-linearity ok. So, there are several aspects that
we will not talk about in these lectures, but there are some aspects that we will cover, but
nevertheless these are the properties of an amplifier.
So, in general you will see that all the MOSFET amplifiers, all the voltage amplifiers
have a tendency to work in a manner shown here ok. So, it is just an analogy, because we
801
have always use this analogy of water in a bucket to understand Fermi levels to bands
band and to Boltzmann’s coefficient etcetera. I will continued with the same analogy
here.
So, imagine to you know two pipes ok. So, one is a large pipe and other one is a much
narrower pipe. And both these pipes have got a certain amount of water, there is a water
it is a continuous volume of water that fills these pipes ok, this is all water. And the
amount of water is charged. So, the amount of water is essentially charged, and the
fluctuation in this level of water is the voltage, so this is the voltage. So, we design an
amplifier such that one pipe is very wide and the other pipe is very narrow. And we will
design it such that the input sensor submits the fluctuation here, which allows the water
to move up and down a little ok.
The input voltage is very very so this is the input voltage. The input voltage is very
small, but since the flux has to be the same or the current through these two pipes. The
flux of water has to be the same. A small fluctuation here, simply because the geometry
of this design becomes a very large fluctuation there ok, so that is exactly how a voltage
amplifier works that is the best intuition ok.
And this so these are basically going to be construct with the two devices, it could be a
resistor or a transistor acting as a thin capillary, which is called as the load. And a
MOSFET acting as a large capillary, which is called as the driver. So, when we say
802
driver, it is the device that is going to receive the input signal and which is going to
modulate the current or the flux through the circuit.
And the load is the one that is going to receive this change in flux, and it is going to drop
a voltage across it, which is the amplified version of the input voltage received by the
driver. So, this is the voltage amplifier. So, this is the very good intuition to the what the
voltage amplifier is going to do, and you are going to see this happen in all the different
circuits ok.
Now, before we begin so now, what we are going to do now is going to draw some
topologies of circuits. And we are going to perform an analysis because, without that
analysis you are not going to be able to feel the operation of the amplifier. And you are
not going to be able to relate the operation of the amplifier to the device properties that
are required.
So, therefore do bear with me. So, what we have defined here is the different symbols
that you will see. So, everything in capital is refers to DC voltages, these are not
fluctuating, these are constant values, these are DC levels. And everything in small
alphabets refer to ac or small signal responses.
And why do we use all these small signal, because we want to linearize the MOSFET we
want to linearize the device ok, so that is going to be the output voltage of the MOSFET
803
that is going to be the input signal received; that is the output signal that is the current
through the MOSFET and these are all small signal terms. And what is the gain of the
voltage amplifier, it is a small signal output voltage divided by the small signal input
voltage ok.
So, we will always be trying to calculate the gain. And what is the input impedance of
the amplifier. So, we are always talking about small signal parameters, it is the input
voltage the small signal input voltage fluctuation divided by the small signal input
current into the circuit. And what is the output impedance of the circuit, it is the output
voltage divided by the total current that is coming out at the output or in fact we will
look at certain techniques to look at all this ok.
So, the very first circuit we will study is a series combination of a resistor and a
MOSFET. And if you remember your capillaries the two capillaries with water, we said
that was the load and this is the driver. So, here we have the driver, and we have a load.
804
So, we have a load element, which is the resistor and we have a driver element, which is
a N-channel MOSFET ok. So, this is the way the circuit is connected.
So, you see a power supply here. And the power supply is in capital V D D that is
ground. And the power supply is designed, so that there is a DC voltage V out, which is
not shown here. There is a DC voltage V out, which is kept here such the and there is a
DC voltage V in, which is not shown here. And everything is done, so that the MOSFET
is in saturation.
805
So, we were building voltage amplifiers, we will always try to keep the MOSFET in
saturation, and I will tell you why ok. The MOSFET is going to be in saturation mode
operation, which means that the DC value of the input signal. So, the input signal has got
a certain DC value, and it is got a small fluctuation around and this is the AC signal.
And similarly, output you will find that there is a DC value to the output. And then the
amplified input is going to fluctuate around, so that is your v out small v out. And it is the
small v out by small v in which is going to be the amplifier gain. We are not interested in
capital V out by capital V in ok, but the capital V out and capital V IN are chosen by some
external circuits is already designed in such a manner that the MOSFET is in saturation.
And this is going to be the case everywhere, we are never going to show these DC
values, they will only show the small signal values, and the MOSFET is in saturation
everywhere.
So, V out the MOSFET has to be in saturation, your capital V out must be greater than
small VIN-V T, because this is V DS that is V GS and that is VT. So, we are only illustrating
this fluctuation. We have ignored we have ignored all this; we are only interested in
these small signal fluctuations of the MOSFET ok. So, we can draw two circuits, we can
draw two different circuits. So, we can say that there is AC and DC all happening
together it would be too confusing to analyze both of them together.
806
So, what we will say is from this point on whenever we do AC analysis, we will ground
all DC points. So, what we are saying is there is V DD plus small signal v d d is there a
small signal power no, this has to be 0 we would like the power to be constant ok. So, so
before I do this, let me just reiterate the point I made.
So, if there is a DC plus ac, AC is all shown in small signal ok, so everywhere you have
some DC nodes and ac nodes. We are going to say that this circuit is a combination of
two circuits. One of them is purely DC; we will only look at DC. So, we will only
consider capital DC everyone. And the other one is only ac, we will only consider capital
the small ac signals everywhere ok. So, we are only interested in analyzing this ac
circuit, because we are only interested in small signal analysis. And when needed, we
will do the large signal analysis ok.
807
So, now I probably should have made a separate slide here. So, do I apologize, if it is not
clear, but you can only send me a mail. So, now let us try to build our two different
circuits. We have capital V IN plus small v in going here. We have capital V OUT plus
small v out going here. We have ground, and there is no small there is no small signal
here, ground is DC ground, it is AC ground its ground ok.
And the power supply is constant ok, it is a DC voltage. And there is no small signal
fluctuations here. So, there are no small signal fluctuations in power supply and ground,
definitely these are not desired, you want the power supply and ground to be as steady as
possible. So, what is the AC circuit for this? If you draw only the AC circuit, so we will
only consider these portions ok. And if you draw the AC circuit, we find that this is
ground. The power supply can be thought of as AC ground, although it has got at large
DC voltage.
The output node will only have v out, and the input node will only have small signal v in.
And what is the DC circuit; the DC circuit will have your ground. Now, you cannot
ignore the power supply. There is a large DC voltage, there is some AC I mean DC
voltage at the output, and there is a DC voltage at the input ok. So, this understanding is
key ok, you need to understand what we are doing here, before we proceed.
And we have now biased this circuit in such a manner that it since saturation, you want
to keep this transistor in saturation. So, what do you mean by saturation, let us go to the
808
DC circuit. We want the current in the transistor to be beta ok. So, beta is nothing but my
mu 𝜇CoxW/L/2, so that is called as beta.
So, we want the current through this transistor is in saturation, so its beat it is
𝜇CoxW/L/2, so that is the current in saturation is mu 𝜇CoxW/L/2(VGS -VT)the whole
square, so which we are writing as 𝛽(VGS -VT) the whole square.
And we want that to be beta into (VIN -VT) the whole square, so that is your V g s. And
what is V out capital V out is V D D - IR drop here. So, VDD-I ds into R L is my V out. And
to keep this in saturation, we want V out to be greater than V in-V T, so that is the DC
condition required to keep this driver in saturation mode, and that is the DC circuit ok.
And when we do AC analysis, we will not worry about this circuit, we will only use this
circuit, where all the DC nodes like power supply and ground have all been grounded,
and we only look at small v out and small v in ok. So, I hope this is clear. I have
probably raced through this part a little quickly, but I have nevertheless try to make it as
clear as possible. Ok, so let us move ahead.
Now, when we do small signal analysis, we would like to define certain terms ok. So, we
are now interested in the small signal behavior of the MOSFET. And what do you mean
by a small signal behavior to reiterate, you have the transfer and output characteristics.
So, let us say that is the transfer characteristics of the MOSFET. So, the entire curve if
809
you take as such is an the MOSFET is definitely a non-linear device ok, but in a small
region ok.
So, if I fixed my DC voltage, and I am only looking at my AC signal there is going to be

my small V g s. In this region, I can define the slope to be constant, it is an infinite. If it
is infinitesimally small section, and I can define my Vgs a fluctuating within this
segment. And therefore, the slope here is constant. And that slope determines the amount
of fluctuation of the AC drain to source voltage ok.
So, in other sense by keeping the slope here constant, we have linearized the behavior of
the MOSFET, we are saying that the MOSFET is like this, if I were to bias my DC
voltage at that point. So, here it is useful to define this term, which is your small signal
fluctuation of ids by small signal fluctuation of v g s is something called as the
transconductance of the device of the MOSFET; it is called g m or the transconductance
of the MOSFET. So, if somebody gave us the input node fluctuation at the gate of a
MOSFET, and asked us to identify the current the small signal current in the MOSFET.
The answer is simply g m times the input node fluctuation ok, so that is the useful term.
So, now we are all set to go ahead, and do the ac or small signal analysis small signal
analysis or ac analysis of the circuit.
810
So, let us say that you have this resistor, and now we have drawn only the AC circuit. So,
you can see that we have grounded the DC nodes, and we have only got the AC node.
So, we have small v out and small v in, we have a small v in going in, and we want to
know what is the gain of the circuit. What is v out by v in, which is the gain, we are
talking about small v out by small v in which is the small signal gain of the circuit. So,
we said that this is an amplifier, so what is the gain of this amplifier.
So, all you have to do is solve Kirchoff’s law current law at this point, just solve
Kirchoff’s current law at this point, which means what which means all the current going
in here is equal to the current coming out here ok. So, what is the current from here to
here through this resistor, it is 0 minus v out that is the voltage drop across the resistor
divided by R L that is the current in this direction. And that current, so you should keep
the direction very it is very important to keep the direction right ok.
Otherwise, you will end up with a sign mistake, and that current must be equal to this
current. Now, this MOSFET is in saturation, so if it is an ideal MOSFET in ideal
saturation, there is no channel length modulation would a fluctuation in v out or in the
drain to source voltage would a fluctuation here, cause any current variation the answer
is no. So, therefore this v out does not contribute to any current in the MOSFET.
But, on the other hand you have a gate voltage fluctuation v in ok. What is important to
see, whether there is a gate to source voltage fluctuation is there a fluctuation in either v g
811
or v s. Now, v s is not fluctuating, because it is a DC, but the gate is definitely fluctuating
ok. And the fluctuation is basically v in. And therefore, yes there will be a current
through the MOSFET, and the current as we looked at will be given as g m times v in,
where g m is the transconductance ok.
Therefore, we just write Kirchhoff’s law. We say that 0 minus v out by R L is the current
through the resistor, and that is equal to gm*vin which is the current through the
MOSFET. And therefore, what is v out /v in = -g m* R L, so that is the gain of the circuit
that is the gain of the circuit.
So, now what let us connect all this the device physics, this is quite easy. R L is simply
an external resistor; the only thing that you need to worry about is g m. So, what is g m?
G m is d I D S by d v d V g s when the MOSFET is in saturation ok, it is a small signal it is
its telling you what is how much the current varies, when the voltage vary. So, I the
small signal v current through the MOSFET divided by the small signal gate to source
voltage, which is basically a dID S/dV g s in the MOSFET.
So, in saturation mode your I D S is beta or which is your mu 𝜇Cox W/ 2L into V g s -V T

the whole square. So, what is dID S/dVg s? dIDS/dVgs is mu 𝜇CoxW/LV gs -V T, so that is
the transconductance gm. So, we already know what mu depends on, we know what C ox
depends on, and we know what V T depends on ok. It is got a very it is quite coincidental
812
that this g m value of a MOSFET in saturation is the same as the inverse of the r on value
for a MOSFET in linear mode ok.
So, you can it is basically the inverse of the is 1/r on of the MOSFET, but when the
MOSFET is in linear bias and deep linear bias ok. So, the second thing that is that is the
that is so we are already we already understand the device physics, so there is nothing
new to connect here. So, the other thing to note is this negative sign ok, why is the
negative sign occurring ok. So, it is showing that its showing that when v in goes up, v
out is actually coming down ok, its phase shifted. And why is that actually quite easy to
understand.
Ok, so if you were to look at the DC circuit and AC circuit together, let us put the DC
and AC together. Essentially this voltage goes up, the current in the MOSFET increases.
And if the current in the MOSFET so this is the gate voltage of them, say V IN goes up,
the current in the MOSFET increases. If the current in the MOSFET increases, the
voltage drop across the resistor increases ok, so that is V out.
And what is Vout? V out is V DD -I R L. So, if the current increases V out comes down.
Therefore, if VIN increases V out comes down; and if V IN decreases, V out goes up. And
therefore, there is a negative sign here showing the phase shift.
813
And finally, look at this pattern ok. So, you start observing this pattern, you have g m
into R L. Since, this an amplifiers like a lever, you have a driving strength into the load
output impedance. So, you will always find no matter how complicated your voltage
circuit voltage amplifier circuit. You always find that there is an effective
transconductance of the driving stage multiplied by the effective output impedance of the
circuit will be the gain of your circuit ok, but it is all right. If you do not observe the
pattern for several cases, it will just it is just a matter of practice ok. So, this is how you
calculate the gain.
Now, let us calculate I will show you the technique to calculate the output impedance of
the circuit of the circuit. So, we want to know, so this is the circuit we analyzed right. So,
you want to know what happens, what is the impedance. So, suppose there is another
stage here ok, the circuit is going to drive some other stage.
Now, when this stage looks inside the circuit, we want to know what is the impedance it
sees or in other words what is the output impedance of the circuit. So, if you go back to
our definition of the voltage amplifier here ok. So, we have understood what the gain is
ok, but we still did not understand the input impedance and the output impedance ok.
And we are now looking at the output impedance, which is if this circuit here, were to
look back into this amplifier what resistance would it see here, what it is the value of that
resistance ok, we ideally want that value to be small. So, it is good to understand, what
814
constitutes that value. So, you want the gain to be large and we already saw that the gain
is dependent on g m* R L.
So, now let us see what constitutes the output impedance of the circuit. So, in order to
calculate the output impedance, what we do is we just get rid of all are input signals, you
know let us just take the circuit alone separately ok. Let us say there is no input being
applied. So, if there is no input being small signal input being applied, the input node is
also grounded. And this is power line is grounded the gate ground is ground, because we
are doing AC analysis.
And we will apply an artificial test voltage here ok. So, what we will do is we will apply
a test voltage, it is an artifact. We just doing an experiment a thought experiment. And
we will measure we will try to calculate the current that will go inside this inside the
circuit ok. We will apply an artificial test voltage, and we will try to calculate the current
that will go inside the circuit. And this small signal test voltage divided by the small
signal current will basically be our output impedance of the circuit.
Because, if I were to put this entire circuit in this black box, and say it has got some
output impedance. What we are doing is we cannot see inside this black box, but we are
saying that we are going to drive, we are going to apply a test voltage here, we are going
to send a test say current here. And this ratio is basically going to be my R out ok.
815
So, let us see what the currents are now if you were to do this ok, what are the cons, we
will we just solve Kirchoff’s law here, so where will the currents go. So, this i t will go it
sees two parts, it can go down this way, and it can go down that way ok.
Now, what is the current through the MOSFET, the MOSFET is in saturation. And any
fluctuation that the drain to source node is not going to create any current. So, ask
yourself if the MOSFET is in saturation, the fluctuation in v d s really does not matter,
there is going to be no current ok. If it is an ideal MOSFET, because we are not
considering channel length modulation if you consider channel length modulation, yes
there will be a current. And is there a gate voltage fluctuation a gate; gate or source
voltage fluctuation gate minus source right. So, gate or source voltage fluctuation will
create a current, but since there is no gate voltage fluctuation the current is again 0.
So, therefore the current through this path due to this test signal that you apply is 0, there
is no current through this path. All the current only goes through this path. And what is
the current in this path, it is the V t that you have applied, so it is going to be V t minus 0
minus this node, which is 0 divided by R L. So, therefore my output impedance is equal
to V t /i t. And what is i t? The i t is 0 through the MOSFET plus V t/R L, which means
that my output impedance is simply R L. So, the output impedance determined by the
load, it is just a method to show you how to calculate the output impedance.
816
Lecture – 49
Amplifiers using MOSFET- continued
So far we have been assuming that if the MOSFET is in pure saturation say saturation,
and it is if it is an ideal MOSFET, then any fluctuation in the drain to source voltage does
not create a current. There is no current. But if there is channel length modulation ok, so
ideally the MOSFET so far we have been looking at MOSFETs that look like this, but if
there is channel length modulation the MOSFET characteristics would look like that.
And technically speaking a small fluctuation in Vds, now will cause a fluctuation in I DS
here ok. And the way we want to take into account this channel length modulation effect
is we will consider two circuits in parallel.
From now on whenever we say channel length modulation is important, we will say that
there is a MOSFET that is where there is a no channel length modulation and a finite
output resistance in parallel to this MOSFET, which takes into account this channel
length modulation term. This ro is essentially the slope it is a 𝛿Vds/𝛿IDS when the
MOSFET is in saturation ok, so that is the output impedance that we are going to add on
to the MOSFET in this parallel fashion.
817
So, now when we say that there is a Vds fluctuation there is going to be no current in this
MOSFET, but there will be a current here. And this current is basically the channel
length current of the non ideal MOSFET. So, the non ideal MOSFET is a represented as
a combination it is a parallel combination of an ideal an a ideal MOSFET an a resistor in
parallel and that resistor corresponds to the channel length modulation. So, now this is
the physic, this is the connection to the device physics, because you all understand what
channel length modulation is.
But, now let us see what impact it has on circuits. So, now let us redo the understanding
of the gain ok. What is the small signal gain now, now we will consider non ideal
MOSFET ok. So, you see this there is now a channel length modulation resistor. So, then
we have a non ideal MOSFET. So, what is the gain? So, we solve Kirchoff’s law here
current law.
818
So, the current coming down this node is the current through this, and the current
through this ok. So, let us call that i 1, i 2, i 3. So, i 1 is nothing but 0 minus V out/R L, i 2 is
0, because this is an ideal MOSFET. In parallel to this ideal MOSFET is the channel
length modulation parameter. And therefore i 3 is Vout/ro. And these two together give you
the non ideal MOSFET ok. So, therefore, we have i 1 is equal to i 2 plus i 3 and therefore,
you have this equation which is I am sorry your.
819
So, let us now consider using the non ideal MOSFET in your amplify it is a exact same
circuit as we discussed before, but the only difference is we now have a non ideal
MOSFET ok, which is essentially the combination of this ideal MOSFET and this
channel length modulation resistor. Now, what impact does channel length modulation
have on the voltage gain of your circuit ok. So, once again we will solve Kirchoff’s law
at this point.
We will have we will call that current i 1 the current through the MOSFET is i 2 this is
the ideal MOSFET, and this will be the channel length modulation parameter, which in
combination with i 2 gives you the non ideality effect. So, now what is i 1? So, we need i
1 equal to i 2 plus i 3. What is the i 1, i 1 is 0 minus Vout/R L. What is i 2, i 2 is going to
have there is going to be no impact of V out on i 2, because this MOSFET is now ideal
ok.
And all the channel length effect has been take channel length modulation effect has
been taken into account in r o. So, this current due to V out is going to be 0, but there is a
gate to source fluctuation, which is V in. So, i 2 is going to be g m times V in. And i 3,
which is a channel length modulation effect is going to be V out by r o.
So, these two together represent the non ideal MOSFET, which is ideal MOSFET with
no channel length modulation, and this is the channel length modulation effect. So, this
will result in a gain being minus -g mRL parallel with ro. So, earlier with no channel
820
length modulation if recollect the gain or simply -gmRL, but now the channel length
modulation resistance will appear in parallel with R L ok. So, it reduces the gain a little
channel length modulation reduces the gain a little.
So, now what is the output impedance of the non ideal MOSFET. So, we will do our
same experiment, we will apply a test voltage, and we will measure the test current that
is going in. And we will forget about any input voltages etcetera we will ground all those
nodes. So, what is the current through this branch now through these two devices. This
unit will not take any current, because the fluctuation in v t does not matter, but there
will be a current due to channel length modulation effect.
And therefore, this node will carry a current of v t/ r o. And what about the current
through that resistor it is going to be v t / R L. So, therefore, the total current is going to
be v t/RL +v t/r o, which is basically saying that my Z output is v t by i t is 1 by 1/R L+1/r o,
which is R L parallel with r o so that is the output impedance. So, again you can see the
pattern here. You can see the output impedance into the effective trans conductance is
the gain ok, but that is how you calculate the output impedance. So, you are now familiar
with using small signal analysis to calculate different circuit aspects.
821
So, now let us look at the next question. So, we know what the gain of the circuit is. So,
how do I increase gain ok. I want to make the gain higher, how do I increase gain ok.
The first since this is going to be the gain, how do we increase, the first could be try and
get rid of channel length modulation, but that is not really a very practical solution right.
You cannot get rid of channel length modulation. So, r o is not going to change that is
there to stay. What about changing g m, now that is possible, because what is g m, we
said that g m= 𝜇Cox W/ L( VGS-VT).
So, how do I change gm, I can increase the mobility, but not that is not very practical,
because mobility is defined by a semiconductor, and it is the foundry that decides the
kind and the quality of semiconductor they use. Can I increase C ox, it is possible, if you
ask the foundry to use a higher epsilon ox and a lower t ox ok, but that is again in the
hands of the foundry or the person who makes the devices for you.
Can we increase the aspect ratio, most definitely, because that is an under your control.
You do the layout, you do the design, you can decide the channel width in the channel
length as long as they fall under the design rules of the foundry. And therefore this is
something you could definitely use to improve g m ok. What about the gate to source
voltage, yes we can increase that, but only to a certain limit right. We cannot make it
greater than the maximum supply voltage used for the technology.
822
And what about VT, VT cannot we cannot play around with VT too much, because it is
the foundry that decides the dopant concentration the flat band voltage etcetera, etcetera.
So, this is not something that we can clear round with. So, these are not things that we
can play around with, but the aspect ratio, and v g s is definitely something we can
adjust.
So, let us say we have done our best, and picked the best possible g m. We have played
around with this. We cannot do anything with the channel length modulation. What
about the load resistor, can we increase the load resistor, technically yes you can increase
the load resistance to give you high gain, but there is a certain limit right. You cannot
make the load impedance very very large.
Say for example, let us say I keep I say I am going to continue to I am going to continue
increasing the load resistance, and I am simply going to have an open circuit that. So,
this is going to be my amplifier circuit. So, I have my V in here, I have my V out there
does it mean there I am going to have very high gain.
Can I use this equation, and say that R L is infinite, therefore infinite and to the let us say
r o limits everything infinite, and parallel with r o is basically, r o and therefore my g m is
very very large. Can we say that? The answer is no, because clearly you can see that the
V out here is not going to remain V out it is going to go down all the way till 0.
823
R is going to become very very small. And there is going to be no current in the circuit
ok. So, there are it does not seem like a solution that that is going to be readily that is
going to solve your problem.
So, what is the what is the catch here, so what is it that you observed, so, as you increase
R R L you will find that the D c value of the output node starts dropping ok, because if I
want to set a constant current let us say the current is kept constant. And we start
increasing R L ok.
824
The voltage drop here becomes IRL. If the voltage drop starts increasing, and therefore
V out the Dc voltage the V out starts dropping. And if it drops too much, if you keep
increasing R L, and if the voltage drops too much, then this transistor will move out of
saturation, and it will enter linear region of operation it will sort operating in a linear
mode.
And what is the problem in linear operation, the problem in linear operation, when you
use it as an amplifier circuit is that any fluctuation in V out will now start pushing a lot
of current through this device, because, now this device has got the resistance of r o
linear ok, which is very very small. It is much smaller than r o. And because if you if you
recollect your output characteristics of the MOSFET, this is the linear resistance that is r
o as you have shown it here, because that is r o, and saturation, and this is the r o and
linear. And this these two are pairing in parallel will simply be dominated by r o linear.
And the effort of all your large R L in order to achieve this will essentially be limited,
because this parallel combination is also going to be there may be a linear.
So, in you cannot increase R L forever so that is the message of the story plus there are
many other factor such as your swing of your output goes down etcetera, but
nevertheless you cannot increase R L forever. You can you can increase it a little, but not
forever, but there is a limit to it. This you cannot do anything and this you can definitely
825
increase, but and it you can increase it by improving your aspect ratio, but now let us say
I am not happy with the situation.
And I want to definitely improve R L what options do I have. So, in our attempt to try
increase R L let us try to have a circuit which has 2 n-MOSFETs ok. We have an n-
MOSFET driver, which is biased and saturation and an n-MOSFET load which is biased
in saturation. And in order to make sure that these two are biased and saturation, you can
do a Dc analysis, which I will leave as a homework problem wherein you just equate the
currents in these two devices.
So, you see that this is the current in one saturation device. And this is the current in the
other saturated MOSFET and by equating these two currents we know we get obtain a
relation for V out and all you have to do is ensure that V out is greater than v g 2- V T, and
that V D D is greater than and equal to V g 1- V out -V T for both these MOSFETs to be in
saturation.
So, we have a situation, where the DC conditions are set such that the driver MOSFET is
in saturation, the load MOSFET is in saturation, and both are NMOS devices, both of
them are NMOS devices. And since this is the driver we are going to replace V g 2 with a
small signal input from the sensor. And V g1 is a DC voltage which is used to just bias
keep things under right bias conditions. So, there is no AC signal there ok.
826
And therefore, this will end up being coming AC ground when we do the AC analysis.
So, let us take this case and why I have we chosen an NMOS device, the hope is that the
NMOS device can be engineered through aspect ratio, through applying v g 1 in a suitable
manner etcetera.
To make sure that V out is always kept finite. And it is always kept keeping you know
MOSFET the driver MOSFET in saturation unlike the case of using a resistor, so that is
the hope. And we hope that this is going to give us large gain, because now we can use
many more parameters to tune the resistance or the MOSFET, but let us see whether that
works out.
So, the small signal analysis is like this. So, you have that was the load MOSFET that is
the load NMOS transistor, which is sitting in saturation. This is the channel length
modulation resistor of that load. And this is like an ideal NMOS transistor, and in
combination they give you the non ideality effect. And the gate is grounded, because this
was a DC bias signal that we had applied, so that is grounded it is ac ground.
827
And now we have a small signal a c v in into the driver of the circuit the driver
MOSFET, which is also an NMOS, which is in saturation; so, now, again writing
Kirchhoff’s law here. What are the currents in these four nodes ok, so the current through
this transistor here. This current is not influenced by V out, because this is representing
the ideal moss.
And therefore, this current is simply g m of the driver into v in that is this. What is the
current here, the current here is V out by r o ok, so that current is V out divided by r d o.
What is the current through this channel length modulation parameter, so that current is
minus V out by r l o. And this current here is there a current here firstly so this current
here is not going to be present, because of a v d s difference, because this is in saturation,
but let us look at the v g s, if there is a v gs fluctuation, there will be a current of g m l
into v g s.
And what is v g s it is a v g minus v s the v g is grounded, but what about the v s the v s is
basically v out. So, therefore, there is a current and that current is minus gm l into v out.
And this is the key to this forms the bottleneck of using the circuit in order to address
gain. So, let us use Kirchhoff 's law. So, the total current through this branch here is this.
And the total current through that node there is that and by equating these two you end
up with a gain that looks like this.
828
So, the gain has got g m d which is the driver multiplied by a parallel combination of
these three resistors. You have RLO, which is high, because it is the output impedance. It
is the output impedance of the MOSFET in saturation is very large in parallel with r d o,
which is also the output impedance of the driver MOSFET in saturation it is large, but
the parallel combination with g m l is going to reduce this entire term.
It is going to make this entire term small, because g ml 1/g ml is a very small number,
why is that 1/g ml is nothing but 1/𝜇CoxW/L (VG2-VT) ok. And this if you think of it is
numerically, it has got the same numerical value as the output resistance the MOSFET in
linear mode. When we looked at the analysis of a switch, and we already know that this
is going to be designed to keep it very low.
829
So, 1/ g m l is very small, this is a small term. And therefore this entire parallel
combination is going to become small, and therefore the gain is going to become small.
If this if only if this g m l, if this term are not present, the gain would be very large. And
what is that g ml due to, it is due to the gate to source voltage of this MOSFET of this
load fluctuating, because it was an n-MOSFET.
So, how do I get my high gain it looks like adding another n-MOSFET does not work.
So, the answer to high gain is to find a device that will give us a current source, so that
830
no matter what the fluctuation across this device, the current is always a constant, and
which means the small signal output impedance is close to infinite so that is the device
we want.
It will give you a finite voltage drop finite voltage drop, but infinite output impedance, so
that is the magical device where looking for.
So, from our knowledge of device physics is there a device of that kind, the answer is
yes. And the device is a complementary NMOS device. So, you have an you use an
831
NMOS driver, which is biased and saturation. And we use a PMOS load, which is also
biased and saturation ok. And this structure is supposed to be an amplifier that has got
very high gain, why is that in the previous case in the case of an NMOS, if we when we
looked at the NMOS load, we found that the gain was being killed, because of the 1 by g
m l term that is the g m term of the n-MOSFET.
The n-MOSFET was permitting a current through this path in response to the fluctuation
in V out. So, we want a device that irrespective of the fluctuation V out the current will
the will have no fluctuation or no response, and current through it we want such a device.
And we are saying that the PMOS is such a device why is that what is the gate to source
voltage of the PMOS, the gate to source voltage of the P MOS is this, the gate to source
voltage to the NMOS is this. So, it is only if either this or that nodes fluctuate, well an
ideal P MOS have any current through it. So, a fluctuation in V out will not cause a
current through the P MOS device ok.
So, in an case of an NMOS device, if either this node or this node where to fluctuate, you
will have a current through the device, but in the case of PMOS device it is only when
these two nodes fluctuate that you will have a current through it. And therefore,
fluctuation in the output node will not cause a current.
832
And therefore, it is going to be a device that is going to give you the resistors resistance
that you desire. So, once again now, let us analyze a final circuit, which is the amplifier
with P MOS load. So, we have a PMOS load and an NMOS driver. It could be the other
way around as well.
So, it is basically CMOS technology or complementary MOS technology ok, which is

basically the cornerstone of all our digital analog circuits now. So, if we look at this, we
have the currents through both the devices being equal. And therefore firstly you will
identify that there is no definition for the output voltage. And therefore it becomes quite
difficult to bias the circuit, but that being put aside.
833
Let us take let us take let us consider the small signal analysis of your new amplifier,
which has got a P MOS load and a NMOS driver. So, the P MOS load has got a bias
voltage a dc bias voltage no ac fluctuation there, so we just ground it. So, you do not
worry about that the power supply is grounded the ground is ground. This is v in that is v
out, that is the channel length modulation of the P MOS that the channel length
modulation of the NMOS.
And now we see the primary difference, the current through this is the same. It is g md
Vin, the current through this is V out by r d o, the current through this is V out minus V out
by r l o, and the current through this is now 0, because neither is this node fluctuating, nor
is that node fluctuating. So, the current is now so 0, and that is the fundamental
difference between having an NMOS and a PMOS.
And therefore, using Kirchoff’s law this current here is equal to gm, this current in this
node that is g m d into Vin +VdVout by r d o. And it turns out the gain is this it is minus g m
d into r d o parallel r l o. This is very high and that is also very high, because these are the
saturation mode output resistances of the MOSFET. And this is existing only, because of
channel length modulation So, if you say that there is no channel length modulation, then
this becomes infinite and that becomes infinite, and the gain is theoretically infinite.
834
So, CMOS technology can give you very high gain ok, so that is why it is desirable to
have CMOS technology. It can give you high gain ok, but does it follow the property of
low output impedance the answer is no. The high gain comes with the cost of high output
impedance. And therefore, you need to have this circuit give you high gain, and you need
to have another circuit to give you low output impedance, which is not going to be
covered in these set of lectures.
But if you are interested, then you can look up any book and analog circuits, and you can
look up common drain amplifiers, and you know other amplifier topology such as
common gate etcetera, etcetera. So, this gives you a brief understanding of amplifier
design ok. And to summarize everything you have learned and connect everything to the
device physics.
835
What is the key message? You all know what gm is and how gm is connected, so gm is
important. High gm generally implies high gain. The next message is the channel length
modulation r o in due to channel length modulation brings down gain a little ok. So,
channel and modulation resistance is still large it is large, but it brings down it
contributes to lowering the gain a little small.
And CMOS technology can give you very high gain. And the gain for most voltage
amplifiers is a combination of the effective trans conductance of the driver, and the
effective output impedance of the circuit ok. So, this is the take home message from
everything that you have learned. So, CMOS the combination of PMOS and NMOS
transistors is a very powerful tool.
836
So, here is the summary of you know using the CMOS technology in an inverter
configuration. So, if you are all familiar digital circuits, you will find that this is a circuit,
which is called as an inverter. And from the point of view of analog amplifiers what we
are doing here is we are using the NMOS as a driver, and the PMOS as a load and, also
the PMOS as a driver, and the NMOS as a load. So, it is a very powerful very strong
amplifier. And you will find that the output and input characteristics look like this. And
you can see a region of very high gain. Any small fluctuation in the input will lead to a
very large fluctuation in the output.
837
Lecture – 50
Circuits: Frequency Response, Noise
So, let us now come to the topic of the frequency response of circuits and may not going
to do this in a very exhaustive manner, but rather look at just a couple of important
points, which are related to the engineering of the devices that we use. And these two
points that we want to worry about are something called as a Miller effect and something
called as a transit frequency.
And to understand the Miller effect we need to get back to using a circuit and we will use
a very are very early circuit of a common source amplifier with an NMOS driver which
is on saturation and the and a resistor load. And a transit frequency is a more device level
study and therefore we will try to do that first ok.
So, what do you mean by transit frequency? Now, so far we have implicitly assumed that
we are talking about input signals that are time varying. We have always said it is a time
varying input signal is a small easy waveform, but the frequency has been very low. And
why do we say that we have implicitly made that assumption, because we never bothered
838
about any current going into the gate. So, if you were to take a capacitor and say you
have an input Vin, which is say some Vino sin 𝜔t. And say the capacitance is called a the
capacitors got a capacitance C, and let us say this end is grounded.
So, what is the current through the capacitor? The current through the capacitor is C
dV/dt. So, in this case it will be dVin/dt, which will effectively be C into V in o into
omega into cos of omega t, where this is basically a phase shift term. So, the magnitude
of the current ok, so if you look at the magnitude of the current through the capacitor, it
is C V in o into omega so, as omega goes up, the current increases.
And since we never bothered to calculate the current through this gate ever during
analysis we were talking about very low frequency signals. So, now we approach a point
where the frequencies so large that we need to take these currents seriously and that is
where the concept of a transit frequency comes in.
So, let us take a MOS transistor and we apply this input as your the we apply that as your
input to the gate of t MOSFET. So, what is the current between the source and drain of
the MOSFET? It is i d s, which is gmV in. And what is the magnitude of this current, i d s
magnitude is, so this is nothing but g m into V ino sin 𝜔t.
And the magnitude of this or which is not capturing the phase information, but just
looking at the magnitude, it is gm into V in o, so that is the magnitude of the current
through the drain to source, now because this frequency is considerably high we also
need to take into account the gate current. So, let us call that as ig igs ok so, what is the
gate current ok? So, if I have to use this data we used here, the magnitude of the gate
current igs is going to be C, which is the capacitance of the MOSFET. So, let me just
rewrite it more neatly right here.
839
So, the magnitude of iGS is going to be the capacitance of the gate capacitance which is
nothing but Cox W/L, so that is going to replace the term C here. Cox into WL let me just
remove this, because we do not need that into V ino 𝜔𝑡 so that is your total gate
capacitance. This is the capacitance per unit area into the area, which is W* L into Vin
into 𝜔, so that is the total gate current.
So, you now at high frequencies, you now have a gate current going in and you also have
i d s. Now, the transit frequency is the frequency at which igs equals i d s. And why is
that frequency critical, because at that frequency the leakage or the input resistance of the
gate has started matching the effective resistance of the source to drain, which is the
basic which basically defines the maximum limit or the frequency to which you can push
your device to operate.
So, in some sense the technology determines the transit frequency. And it tells you that it
is a frequency, when the frequency when the gate to source current, matches the drain to
source current in magnitude, so which means that this is when C ox WL into V in o into
omega is equal to g m into V IN o. And what is the g m? Gm is mu C ox W over L into v
g s or the V IN D c the D c value of V IN, which is not shown here.
So, there is a D c value of V in, and on top of that you have this very high frequency v
small v n going here. So, the V IN D c-VT into small v in o and we will give a special
symbol to this omega we will call it omega T, which is to represent the transit frequency.
840
So, C ox cancels on both sides, W cancels on both sides, V in o cancels on both sides and
what we are left with is this definition for the transit frequency, which is omega T is mu
by L square into V IN dc -VT.
And if you write it in terms of the frequency in hertz, if you say f T, then you divide this
by 2𝜋. So, this is a very key parameter and any technology ok. So, it is telling you that at
a certain frequency that the gate input gate current will match the drain to source current
and that frequency is given by f T, which is mu by 𝜇/L2 ( V IN Dc- VT).
So, how do I increase that frequency, we increase that frequency by lowering the channel
length. And therefore, transistor scaling plays a great role in improving the f T. And the
other thing to do is to increase the mobility. And what does mobility depend on; we are
already quite comfortable with that it depends on the purity of the semiconductor. In the
sense what is the mean collision time, what is the effective mass etcetera ok? And then
we have of course, have the threshold voltage. So, this is something called as a transit
frequency which I felt was an important concept.
Now, the second concept that we are going to look at is something called as the Miller
effect and that is got to do with the circuit that you are using. And in particular it is going
to do with circuits that give you high gain voltage amplifiers that give you high gain.
And this circuit that we have used here is a good example of that. So, what is the low
frequency gain of the circuit that we have used, so, let us just clear some space here.
841
So, what is the low frequency gain of the circuit, we have already calculated that we said
that, so in here we let us ignore channel length modulation you all know how to we use
channel length modulation. So, the current through this is g m into v in, and the current
through that is minus V out/ RL. And therefore, by equating these two currents at this node
we have minus V out by R L is equal to V in into g m and therefore, V out by V in is
equal to the gain and V is equal to g minus g m into R L. And this gain is the low
frequency small signal voltage gain of the amplifier.
Now, what happens when the frequency starts getting larger. Now, a key concern is to do
with the pole, again what do you mean by a pole for the people, who are not familiar
with control systems theory or if you are not taking a control systems course. Let us just
take a small detour here. Let us go back to the example of an R C circuit. So, we have an
R C circuit.
842
Let us try to define these things in a manner, where we do not require the use of you
know Laplace transforms etcetera. So, in an R C circuit, so let us so we already some of
the step response of an R C circuit. We said that you know there is going to be the
voltage here is going to climb as V naught into 1 minus e to the power minus t by R c,
where t was the time constant of the circuit.
Now, imagine in the R C circuit you, now have a pulse, a periodic pulse that beings that
is being supplied. And you can control the frequency of this pulse so, you can basically
control this delay. So, let us draw the waveform let us say that that is the input pulse and
that is the voltage waveform on the capacitor.
How do you think the voltage waveform would be, so let us say this is a starting point.
So, the moment the pulse came up let us just start draw these three nodes. Let us say the
pulse went up that is the input, the capacitor is going to start charging. It is going to
charge to a certain point, it is going to charge to a certain point and then the pulse
dropped off. And now the capacitor is going to start discharging. You give it enough
time it is going to discharge and then the pulse is started again, it is going to charge and
then it is going to discharge and your behavior would look like this. And all this is your
exponential functions.
843
And now what we do is let us start reducing the width of the pulse ok, so that was what
that pulse did, but now what let me just keep that there and now let us start reducing the
width of the pulse. So, let us say we now have a new pulse, which is going to to be much
narrower ok, we not going to give the capacitor so much of time to charge up. So, now,
what is going to happen, so now you have a pulse that it is got a higher frequency or
increase the frequency input frequency of the pulse.
So, now the capacitor tries to take it is time. It is got a time constant of R c, R c has not
changed you only change the time period of this pulse. So, it is going to try to climb, but
you are not given it enough time to charge fully. So, it is going to climb till here and then
it is started to a discharge and then it is going to climb till then and started to discharge.
And therefore, earlier the capacitor was climbing a lot more it is getting charged the lot
more, but now we are allowing it to charge only a little. And therefore, if you look at the
magnitude of the voltage here and if you take the gain of the circuit the say the V c by V
IN and if you take that gain, you will find that the gain has gone down, because V c is
not climbing up to what it was climbing up to ok, because the frequency has increased.
And therefore, if R or C increase, then the time constant increases, and the circuit cannot
respond tR high frequency input pulse.
So, in other sense the gain of the circuit will start going down ok, so that is enough
introduction. So, now, let us come back to this amplifier here so, you can actually break
844
down any circuit into R s and C s series combinations of Rs and Cs and by just calculating
the time constants we will know as to what how whether the circuit can respond quickly
or not.
Now, R interest here in this study on Miller effect has got to do with the input sensor. So,
let us say there is a input sensor, which is sending in all your V in signals and those
signals are all coming here, but the sensors has got some resistance it is got some output
resistance, which is normally the case it could also be a previous amplifier.
845
So, let us say there was another amplifier here and we know that we know how to
calculate Zout. So, it has got some output impedance and that amplifiers sending out some
signals and it is being sent to this amplifier. So, it could be any situation.
But let us just keep it simple we will say that there is a sensor and the sensor is throwing
out you know these high frequency signals, but sitting at a large DC voltage to keep all
this in saturation. And the sensors got some output impedance and it is sending these
signals here. Now, when these signals were very low frequency, we really did not care
about it, because we knew that the MOSFET could easily respond to these low
frequencies, but now we need to be very careful.
846
And in particular we need to identify what is the input R c, so that is the sensors
resistance. And I am now going to be very interested in knowing what the capacitance is
at this node, because each time this signal goes up this R c node has to charge and it is
only then that the gate voltage sees that voltage rise and it is only then that the current
here will change. So, the transistor will not respond to the voltage and the gate increases.
And the voltage on the gate will have to depend on the charging of this R C circuit here
ok. And there is a small problem if you use a amplifier what is the problem?
847
So, let us say normally what constitutes this capacitance, what does this capacitances
comprise of? The way you define this capacitance is to see what are the elements the
transistors driving? So, on the face of this capacitance is basically your gate to source
capacitance of the MOSFET, it is the it is there is a channel here and there is a gate to
source capacitance the in the MOSFET. So that is not the problem that capacitance is
there in any case ok. Now, the problem arises, because of the gate drain overlap.
So, if you remember the geometry of the MOSFET, so you had your gate, you had an
insulator, you had a semiconductor and let us say that was the gate. And we had a source
drain region and there was a overlap, so let us say that is the drain, so there was an
overlap between the gate and the drain. And we saw this having a impact when we
looked at clock feed through in switches. So, it is a same overlap capacitance that we
need to worry about now ok.
So, consider that overlap capacitance there ok. So, now, look at the current sent by the
sensor you have an input signal that is going let us say you have an input signal that is
going like a blip I am going to exaggerate the size of the signal. And now because the
amplifier is giving you again and again or minus g m R L with a phase shift.
The output signal is going to go with a very large negative blip ok. So, what is the
current driven by the sensor so, at this point there is it is all AC signals. So, all the
capacitors have going to sink current in it is all CdV/dt so at this point you have a current
through this. And you also have a current through this overlap capacitor, but now since
the since the circuit is giving you again your V out is minus A times V in where A is
nothing but your g mRL ok. So, we will just say plus A times V in where A is minus g m
R L A is a negative number.
And since A is designed to be much greater than 1 your V out is going to be an amplified
negative going voltage is compared to V in. And therefore, the MOSFET is you can
imagine this capacitor the C overlap capacitor has got a voltage which is A times V in,
where A is minus g m R L and this end having a voltage V in. And therefore, it is going to
sink in a current.
848
And since and you can represent you can if you want to represent the capacitor as an
effective resistor you can do, so by simply replacing it with it is an impedance of 1 by
omega C overlap so that is basically your resistance. So, you have a current of V in the
voltage drop across this is Vin-AVin 𝜔C overlap is basically the current through the
circuit or in other words you have a current of Vin(1-A𝜔C) overlap through the overlap
capacitor.
So, from the point of view of the sensor, so if this is the sensor from the point of view of
the sensor, the sensor is driving is putting out V in and that V in is driving a capacitor C g s,
which is connected to ground, but it is also driving this current through the overlap
capacitor. So, from the point of view of the sensor it is like as though you have another
capacitor to ground, which is got a capacitance of 1 minus A into C overlap. So, it is V in
into this capacitance, now the V in omega into that capacitance now that is constituting
the current.
So, it is like as though you can imagine that overlap capacitor to be another capacitor
connected to ground it is got a capacitors a 1- A overlap, but what is A? A is minus -g m R
L. Now, therefore, this capacitor has got a capacitor 1+ g m RL into C overlap. And what
is g m R L? G m R L is designed to be very large and therefore, this effective capacitor
appears to become very large.
849
So, it is as though your sensor is now driving two capacitors, which are connected to
ground it is got some resistor here R and there is the C g s, but there is also this 1+g m R L
into C overlap. And therefore, your total capacitance is now very large and therefore,
your R c time constant is very small and therefore, the amplifier circuit will respond a
little slower at high frequency so that is another problem. So, what is the connection to
device (Refer Time: 21:16) in the transit frequency we already saw it is the mu by L
square term is very important, but here what is the key concern? The key concern is the
overlap capacitance.
We cannot compromise the gain, but we must definitely do something about the overlap
capacitance so that is why it is important to have processes that are called self aligned
processes, which makes sure the way the devices are fabricated ensure that the overlap
capacitances minimized it is reduced. So, there are many ways you can make use the
Miller effect also, you can artificially create a pole, so as to say to stabilize response of
amplifiers etcetera, but nevertheless this is a problem that you must be aware of. And it
is not only true for you know MOSFET circuits or MOS capacitor.
So, here I have given you a slide where you know any amplifier with any feedback
impedance across it well experience this equivalent of a Miller effect. So, if you have the
time do go through these arguments, although I will not discuss it right now.
850
So, the very last topic we will look at before we close circuits and therefore close the
course is related to the noise and circuits. And in particular we will look at the noise and
MOSFETs. So, what do we mean by noise in circuits? So, we have several electronic
devices and because of the way electrons respond to fields in these devices, the currents
through these devices and the voltages across these devices are all. I have got a very
predictable component, but there is also a lot of randomness to all this and that is what
we mean by electronic noise.
So, we cannot completely predict the currents or voltages to the infinite accuracy,
because there is a lot of randomness. So, for example, if we were to measure the current,
you will find that there is a lot of fluctuation and randomness in the current a small
amount of randomness hopefully. And this is what is meant by noise, but very quickly
you know just to you know just to see how to characterize noise. So, let us say you walk
into a room and there are many people talking how do we characterize noise?
851
The 1st is of course, the amplitude of the noise. Say an what do you mean by amplitude?
Since noise is a random signal you cannot end up taking the average, the average of this
kind of a random signal would actually be. So, let us say you have some signal and that
average will be basically 0, instead you must go ahead and take you must take you must
first square the signal. And probably calculate the average of the square of that signal or
maybe take the average of the root of the square that is the root mean square of the signal
in order to get your estimate of the amplitude. The 2nd estimate has got to do with the
frequency content in the noise.
So, let us say you walk into a room and you have many children who are talking as
compared to adults. You will find that there is noise and the noise has got more high
frequency content as compared to if say adults were talking, where the noise would have
more low frequency content. And the frequency content of the noise is very nicely
described by something called as the power spectrum of the noise.
So, you look at the power of the noise although we says power it is all measured in
whole square per hertz and you look at the distribution of this in frequency. So, this is a
very quick summary of noise. And the 3rd has got to do with whether the noise is
correlated ok so, if you will work into a classroom and there are lots of kids and let us
say you bring in a nice surprise then all the kids will shout out in unison.
852
So, there all the noise sources will be correlated. And then there are other methods to
characterize noise you could say whether the distribution from, which these random
points are picked up, whether the distribution is, whether that probability distribution is
varying in time or whether it is not. So, whether it is the noise stationary or not
stationary? So etcetera.
But, we will just very quickly look at some typical noise in circuits. The very first and
most important in some sense noise source is noise in the resistor and this noise is simply
because of the random movements of electrons in a conductor ok. And this typically
happens when you have a conductor and when you have a large number of electrons that
are moving through you are not talking about a sparse population. So, since many of you
might be familiar with some certain crowded cities I have put a picture of traffic in a
very crowded city.
See, if you look at these roads, there completely jam packed with the vehicles ok. And if
you were to measure the flux of this traffic, you will find that there is a lot of scattering
with the you know with the road itself. There is a scattering between the vehicles and
therefore you will see there is some flux in the traffic. And it is quite possible that this
will have a thermal noise like behavior although I have I am not very sure, but
nevertheless it happens when you have a large number of electrons creating a good
amount of flux in the device.
853
And the thermal noise is all the noise is represented as a mean voltage square or a mean
current square. And the thermal noise is given by 4KT, which is the Boltzmann
coefficient in a temperature into the resistance of the device. And the thermal noise is
quite frequency independent, which means that the power of the noise remains constant
at all frequencies. And you could also represent the thermal noise in terms of the current,
which is to say it is the voltage square divided by the Rs2, which means that this becomes
the thermal noise of the resistor.
854
Now, in a MOSFET in conduction, you do have thermal noise when you have an
inversion channel formation, but once the inversion layer is formed you have a large
number of electrons between the source and drain. And therefore the MOSFET has
thermal noise through it. So, when we talk about MOSFETs and circuits we must take
care of introducing the thermal noise, if you are worried about the noise behavior.
And for a MOSFET in operating and saturation mode above threshold operation, it is got
a resistance, which is approximately or you know in this case you have your thermal
noise current, which is supposed to be 4 KT/R right so that was the thermal noise current
and that term here is approximately 4 KT into 2/ 3gm, where g qw is the trans conductance
of the MOSFET. And gm therefore determines the amount of noise so, if you want to
have smaller amount of current noise, you want to reduce the gm of the MOSFET. So,
this is the output noise of the MOSFET.
And then you have something called as short noise. So, I am just going through this just
to give you a glance through the devices ok. So, here you can again compare it to traffic,
but traffic which is very sparse ok. So, let us say you have a electrons moving in a
conductor a moving in a in a device and the electrons are very sparse.
And this is a classic case for example, say in a diode that is a reverse biased or say in a
for photodiode applications for measuring to capture the intensity of light. So, here you
have electrons moving at some velocity and let us say we put a marker line here. And we
855
the way we are going to count or the current as we are going to count the number of
electrons that cross this marker over a period of time.
So, let us say we say in 100 seconds we counted the number of electrons and we say
thousand electrons went past that marker. And then we are going to restart the elec
experiment you are going to say that ok. Let me start the count for another 100 seconds
and we saw that there were about 900 electrons that went through. And then there were
980 electrons that went through the next time. And therefore, there is some intrinsic
fluctuation in this.
So, it was 1900 980 and so on so forth and that is what contributes to noise and that noise
is something called a short noise. And it is given by this particular term; here your
current noise current is dependent on the DC current through the device. And therefore,
this is not seen in resistors it is also not seen in MOSFETs that are you know in strong
inversion, but it is very much seen and say devices such as photo diodes or photo
sensors, which are you know diodes which are in reverse bias ok.
And then finally we will talk about the last noise source which is something called as a
flicker noise. And this is the very interesting noise source it is quite ubiquitous and I
strongly recommend reading this particular paper that I have linked here. And flicker
noise is quite interesting, because it is got a 1/f responds ok. So, if you plot the intensity
856
of flicker noise, versus frequency is got it reduces as frequency increases. And this is the
flicker noise in the MOSFET ok.
And what is flicker noise due to? Ficker noise is due to typical variations in the number
of electrons available for conduction and typical variations in the mobility of these
electrons and these two can be linked. So, there are two possible models that are very
strongly connected. And they both explain the presence of flicker noise and MOSFETs.
And essentially what does flicker noise saying, it is saying that rare events occur big
events occur very rarely.
So, at very low frequency the power of the noise is very large. So, large fluctuations
occur very rarely and small fluctuations occur very frequently. So, if you think of
earthquakes large earthquakes occur very rarely and small tremors occur quite frequently
so that is essentially the message of flicker noise. So, this is a flicker noise per hertz and
since it is got a frequency dependence you can integrate that to get the total noise in a
certain bandwidth.
And finally, the very last topic is something it is got to do something called as a corner
frequency with regards to noise and the MOSFET. And what is corner frequency? We
know that thermal noise. So, in a MOSFET there are two key components of noise, we
have the thermal noise and we have the flicker noise. And thermal noise is rather
frequency independent it is almost constant all frequencies. And therefore, it is called
857
white, because it is called all frequencies in it, but flicker noise has got a 1 over f
dependence. So, it is very strong at low frequencies and very small at high frequencies.
So, the frequency at which the flicker noise power matches the thermal noise power is
got a given a special symbol is called as corner frequency. So, if you are designing
circuits where the operation is going to be here, then you need to worry about flicker
noise. If you design circuits which the operation frequency is very large, then you only
worry about thermal noise. So, it is useful to know where the corner noise frequencies.
And just to conclude you know I will put up the slide, but we will not go through it. If
you have an RC circuit, then the noise on the capacitor of this R C circuit is actually
independent of the resistor. Provided the resistor is the only source of noise, which is
your thermal noise and provided you are looking at the noise over an infinite bandwidth.
You will find that the total noise voltage noise in the capacitor ends up being KT/ C and
I have given the derivation here, it requires and understanding of certain formulas.
And therefore I will not quiz you or examine you on this point, but nevertheless I felt it is
useful enough to put down on a slide and leave it here ok. And it and this noise is
typically just K T/C, it is independent of the resistor. And therefore if you are using a
MOSFET as a switch, this is some something that that you can be aware of. So, with this
we will complete noise in the circuits, we complete at discussions on circuits itself and
858
we if at complete our discussion on the crystalline semiconductor physics part of the
course.
859
Lecture – 51
Introduction: Amorphous Semiconductors
So, so far the course is focused on Crystalline Semiconductors. Wherein, we had a

periodic arrangement of atoms and we developed models such as the chronic penney
model, wherein we used blocks theorem to identify a density of states profile. And in
particular we looked at the density of states for 3 d materials, 2 d materials, etcetera.
And, we identified a energy a density of state distribution where in we had states above
the conduction band edge and below the valence band edge, but then no states inside a
band gap.
And, now we reach the last bit of the course, wherein we are going to generalize all these
ideas that we develop for a crystalline semiconductors, and in particular talk about non
crystalline semiconductor. So, what happens, if the semiconductor is amorphous for
example, what happens, if it has got a large number of defects; and, what is the physics
that governs this and how do we build devices out of this and how to be model in study
the transistors based on such materials.
So, we now look at the device which sort of defines non crystalline semiconductors. And
in particular we look at something called as a thin film transistor.
860
So, what are the applications of non-crystalline semiconductors?
The strongest application is in the development something called as large area

electronics. So, let us consider a display screen ok.
Now, if I have to build transistors etcetera on a substrate that is as large as a display

screen. For example, a display screen is several inches to give in several feet in length,
and if we need to build all the semiconductors and electronic devices, that are present on
861
the screen using crystalline silicon. One would require either a large tile of these
crystalline materials or 1 would require a very very large wafer.
And, both these are going to be quite expensive and it is a probably not the best way to
go about building such a large area system. So, what is normally done is you know in
order to develop these kinds of largely electronic systems. One starts with a substrate ok.
And, this substrate is not a semiconductor wafer, when it could be any material. For
example, glass or even plastics ok. And, on this substrate the different layers of your
semiconductor device or the different, the different materials used to build your
electronics are deposited as thin films ok. So, for example, if you want needs to build a
MOSFET on this large area substrate. This is one possible method that it could be done,
we first develop a thin we first deposit a very thin metallic film.
That will constitute the gate of the MOSFET we will then deposit an insulator a thin film
that constitutes the insulator. And, then a thin film that constitutes the semiconductor.
And, this would constitute the metal insulator, semiconductor stack. And, then we have a
source strain electrodes, which if required would need the doping of the semiconductor
or some kind of treatment in order to ensure low contact resistance and a good injection
of carriers.
862
So, by depositing is various thin films using processes such as you know thermal
evaporation for metals, or you know dc sputtering or RF sputtering for semiconductors
or insulators etcetera. We develop these thin filaments of with each layer being about 50
nanometers; thin films of these materials to develop electronic devices.
So, by doing these over large areas and at low temperatures one develops larger
electronic systems, for these various applications. Now, the problem of using these kinds
of low temperature processes or you know is that we often end up with semiconductors
that are not crystalline. So for example, if we take a material and you give all the atoms
of the semiconductor.
Enough time and enough energy to move about, and arrange themselves in there least
energy positions or the most favorable positions one would end up with the, one would
end up with the semiconductor lattice that is almost perfectly crystalline. So, for
example, if you look at your silicon wafers; the silicon atoms are allowed to arrange
themselves into a periodic lattice, and they are grown from a melt and they given and the
growth is done very very slowly.
863
So, this you are given the atoms enough time and enough energy, but on the other hand if
you would do not give enough time or do not give enough energy to the atoms. For
example, if you take the melt and you suddenly quench it, you will end up with the
material that is quite disordered. The atoms did not have enough time and energy to end
up in there least energy positions. And they instead formed a network of randomly
arranged atoms. And, these atoms could have different bond lengths it could have they
could have different bond angles etcetera.
And, they form a solid, but it is not a crystalline solid. So, you end up with materials that
are amorphous in nature. So, generally and not I mean there are there are examples to
counter this, generally speaking the kinds of the nature of the semiconductor that is used
in the development of these kind of MOSFETs is amorphous ok.
864
So, you end up with the amorphous semiconductor or a non-crystalline semiconductor

that forms that becomes a part of these devices. And, since these devices are based on
thin films, these are also called as thin film transistors. So, they are field effect
transistors, but they are just built in a different way on different substrates and called as
thin film transistors or TFTs. So, you might have heard of TFT display and that is what
this refers to.
Now, the general architecture of a displays like this ok, as shown here. You have gate
drivers and you have data drivers, but you have the display has got many tiny pixels ok.
And, the size of these pixels determine the resolution of this display. And, inside this
pixel inside each pixel, you basically have a switch a capacitor and say a light
modulating material ok. Now, the light modulating material could be an led it could be
an LCD where in the LCD crystal is oriented in accordance to that brightness that we
need etcetera, but the switch is most often a thin film transistor.
So, this switch capacitor circuit is almost a fundamental building block of any active
matrix display system ok. And therefore, if one looks at a display screen, the number of
pixels present there is equal to the number of transistors present in the display. For a
simple for a simple system of this kind of course, you could have you could have larger
thin film transistor circuits in each pixel.
865
And, in general this larger electronics has you know has led to a lot of interesting
development in the area of displays, in the area of energy storage systems, in the area of
variable electronics. And, since all these transistors can be grown on flexible material,
mechanically flexible materials such as plastics, you could also have electronics that you
can bend and flex.
So, the amount of applications or the creative use of these kinds of systems is quite large.
The application space is quite large.
So now, that we have seen you know how thin film transistor has grown, what does the,
what is the problem? What are the materials used? What are the problems? And, what
are the advantages? The advantages clearly are that all deposition etcetera can be done at
relatively low temperatures, as compared to crystalline silicon. On the other hand it is
come to lot of problems ok.
Now, before we understand the problems, let us look at the typical materials that one
could use. For in terms of inorganic some of the common semiconductor materials that
are used are amorphous silicon. And, hydrogen is added to amorphous silicon to
passivate, you know all some of the disorder particularly dangling bonds.
You could also have poly silicon, which is much much better in terms of performance,
but still less compared to crystalline silicon and you also have a amorphous metal oxides.
866
So, for example, zinc oxide or gallium indium zinc oxide etcetera. And, the
semiconductor materials booklet is not limited to just inorganic materials, you also have
organic materials.
Say for example, small molecule organic materials such as Anthracene or Rubrene. And,
you also have long chain polymer like organic materials. So, it is a it is a quite a large set
of materials that one could choose a semiconductor from. Now, nevertheless in all these
cases we have some very fundamental limitations.
And, the biggest limitation comes in the form of the mobility of carriers. Now, since we
are talking about a disordered material, we find that the effective mobility of your
electrons in the material is quite low; so just to give you a comparison.
Here is a comparison of crystalline silicon with amorphous hydrogenated silicon. So, you
have a field effect mobility for your transistors ok. The electron mobility in the case of
crystalline silicon is about is greater than 1000 centimetre square per volt second ok.
Whereas, in the case amorphous silicon is less than one centimeter square per volt
second, which means that the speeds are about a 1000 times slower on the other hand the
on currents.
And therefore, the on currents in the case of crystalline silicon are about 1000 times
larger than amorphous silicon. So, you have currents of the order of milliamps in the case
of a transistor that is on whereas, in the case of amorphous silicon and the currents are of
the order of micro amps for a transistor that is on. The operating voltages or the voltages
required to create the field to make these electrons move is again, you need much smaller
voltages in the case of crystalline silicon, whereas, in the case of amorphous materials
the voltages do vary from say anywhere between 5 volts to 20 volts. And, this very much
depends upon your process, the device structure, and you know the nature of the
semiconductor, and particularly the semiconductor insulator interface on the contacts.
And, in crystalline silicon we have both N MOS and C MOS transistors, which means
you have a comp complementary metal oxide semiconductor process.
Whereas and we saw that this has a lot of advantages, when it came to circuit design
particularly in the development of amplifiers.
867
Whereas, in the case of amorphous semiconductors most of the time. Or in the case of
most of the materials practically speaking we only have N type transistors or we only
have P type transistors. So, most of the organics gave you p type transistors and say
amorphous silicon for example, gives you N type behaviour.
And not and we do not have complementary MOS transistors in most materials. But,
having said that, there are also several organic materials that allow you to have both n
and p type transistors, which means you do have a C MOS process in these materials.
868
So, let us now look at the consequence of the structure of crystalline and amorphous
materials. And, you know what the consequences with regards to the density of states.
Now, the case of crystalline materials we had a periodic lattice ok, you had a nice
arrangement of atoms, and you we did have defects, but that is those we have a quite rare
ok. And, because of this nice arrangement of atoms one could we saw that, we could use
the chronics penny model, and we could use the blocks theorem. And all that led us to
the fact that, we have several energy levels and we have clean energy gaps between these
energy levels.
So, we had a forbidden gap and that was very very clean.
869
And, in terms of the density of states picture for a crystalline for 3 D crystalline
materials; we saw that, we have the valence band edge, that was the conduction band
edge, and we had a density of states. Density of states that varied as the square root of
energy, square root of E minus E c and the square root of E minus E v minus E, in that
conduction bands and valence bands respectively.
That is behind states in these regions and then we had a band gap with say band gap of
energy E g and in the band gap there were absolutely no states ok. So, this was the
density of states diagram for a crystalline material.
And, what this meant is meant was that, if you want to count the number of carriers, say
the number of electrons, inside the band gap. The number of electrons is dependent ok.
And, any energy level E is dependent on the Fermi the value of the Fermi function, at
that energy level into the density of states of that energy level d E. And, so, this is the
number of electrons per unit volume. And, although the Fermi function had a value that
was you know between 0 and 1.
The density of states had a value of 0 inside the band gap. And therefore, this product
was equal to 0 ok. So, in other words there are no electrons to be found in the density of
states.
870
However, if say a defect, due to a defect if a few states appear in the band gap, it is
possible that that is possible for us to find electrons at these energy levels, depending on
the value of the Fermi function and the number of states that appear. So, this was the
story in the case of crystalline silicon or crystalline semiconductors.
Now, the story in case of amorphous semiconductors is quite different.
So, here we could draw. So, this example this example is shown for amorphous
hydrogenated silicon ok, but there is something equivalent for any amorphous
semiconductor. So, we could draw something called something equivalent to our
conduction band edge. And we could draw something equivalent or valence band edge.
And, this is the plot of the density of states versus energy for an amorphous material. So,
the first thing that one observes is that there is no clean band gap.
So, here is your conduction band edge here is the valence band edge in the case of
crystalline materials, this entire region was void of any states, but in the case of
amorphous materials you find that you have a large number of states. So, you have states
throughout the band gap or throughout the energy distribution ok.
So, it becomes quite hard to state as to which is the conduction band edge and which is
the valence band edge. So, you have states throughout the distribution ok; that is the first
point to notice.
871
And, the only definition for what constitutes the conduction band edge and what
constitutes the valence band edge comes from the definition of the mobility of carriers
and we will come to that in a minute. So, it is generally seen that carriers sitting above E
c move a little faster as compared to carriers that are below E c.
And therefore, all these states that sitting above the conduction bandage, and below the
valence band edge are also called as extended states ok.
872
And, we will talk about this in a minute. So, now let us focus on the states inside the
band gap. Now, this dotted line here indicates the Fermi level position. Now, all these
states that are present in the band gap, and that lie above the Fermi level are called as
acceptor like states. And, these are called acceptor like states. Because, they are neutral
when they are empty, and the moment an electron occupies these states these states
become negatively charged ok. And, the states that lie below the Fermi level are called
donor like states.
So, they are neutral when they are filled with an electron. And, the moment an electron
leaves them they become positively charged. There are 2 further classification of the
states in the band gap. The states that are located near the band edges.
So, that is these regions here and these regions here are called as tail states; or band tail
states ok. And, the states that are located in the middle of the band gap, that is closer to
mid gap are something called as deep defect states or deep states ok.
So, therefore, one can have acceptor like band tail states that are basically the states here;
one can have acceptor alike deep states which are basically the states here close to the
mid gap, but above the Fermi level. One could have donor like deep states, which are the
states located here, and one could have donor like tail states, which are essentially the
states located here.
873
So, these are definitions these are the definitions associated with that density of states
diagram for amorphous semiconductors.
So, a key idea with regards to understanding the behaviour of carriers in amorphous
materials is defined by something called as Anderson’s theory of localization. So, in the
case of crystalline materials we found that the crystalline material had a very periodic
potential; well, which we used in our model, which was called as a chronic Penney
model.
And, we use along with the blocks theorem we found that the electrons, the electron
wave function could propagate through the lattice ok. Although, there were very clean
gaps, which the electron could not possess, the electron wave function extended through
the lattice. And, we had and it permitted electron conduction through the solid. But, in
the case of amorphous materials we find that the potential well has got varying levels of
disorders.
So, you have you have the floor of the potential well that is quite disordered. In means
some sense the material is not periodic or the potential well is not periodic. So, it is a
consequence of that, we find that in Schrodinger equation, your potential energy term is
no longer periodic. And, as a consequence of that you find that there are 2 possible states
in which the electrons could exist.
874
The electrons that stay above your conduction and valence band, you know which we
defined as the extended states. Have a wave function that actually propagates through the
solid ok. Whereas, the electrons that are sitting inside the gap states inside the band gap,
they have something called as a localization of the wave function.
And what that means, is that, if you look at the amplitude of the wave function, you find
that the amplitude or the carriers the amplitude of the wave function is quite large at
certain points and is negligible at other points. Which implies so, the probability of
finding the particle at these locations is very large and therefore, whereas, the probability
of finding the particles in the other locations is very small. Therefore, the electrons could
be thought off as just particles that sit there that remained trapped inside these particular
states. And, they are not available for conduction in the material.
Now, all this is a consequence of the disorderness or the disorderness of this potential.
And, this happens this is a direct consequence when this disorderness crosses a certain of
threshold. So, that is what Anderson’s theory tells us.
So, to summarize the different conduction mechanisms in an amorphous material, all

them all the electrons or all the carriers that sit above the conduction band edge are said
to be sitting in the extended states. And, these electrons are available for conduction.
And, you can define a certain mobility for these carriers, and you can define that
875
transport in a manner that is similar to what was defined in the case of crystalline
materials.
But, all the electrons that are sitting in the trap states, which is basically the band tails
states or the deep defect states. These electrons are more or less trapped ok. And, they
are not available for conduction in a manner as seen in the case of electrons in extended
states. And, the only way these carriers can move about is via tunnelling or via
mechanism called as hopping.
And, a general nature and a general method or a model that describes the transport of
carriers in trap states is as shown here. So, you can imagine 2 trap states ok. And, these
electrons have got a certain wave function, and this wave function is heavily localized.
So, it is got a very large amplitude that sits, you know that very large amplitude at
around these locations. And, the probability that an electron moves from state 1 to state 2
or vice versa is given by P 1 2. And, this probability depends upon the distance between
these states. As well as the energy differences between these 2 states.
And, it depends as the exponential of the negative of the distance and exponentially the
negative of the energy, which implies that if these states are very far apart or if they are
located they have very large energy gaps between them. There is a much smaller
probability of the electron actually getting from state 1 to state 2.
876
Now, with regards to the electron movement in the extended states, the movement can be
thought off as shown in this picture here. So, let us say that is the conduction band edge
and all these states are the extended states. And, below the conduction band edge and
very close to it, there are the trap states that that would try to trap the electrons and hold
it and localize them.
So, the movement of the electron in the extended states is described as something called
as a multiple trap and release model. And, what happens here is that the electrons move
quite quickly in the extended states and they have some mobility called as mobility of
mu naught ok. Now, once in a while the electrons get trapped in these trap states.
And, they spend some time in these trap states, but if the electrons are trapped in trap
states that are quite close to the extended state edge, which is close to E c. A small
amount of energy is can release these electrons ok.
So, let us say a small amount of thermal energy allows these electrons to get free from
the trap state and get back into the extended state. So, the movement of the electron is a
periodic is a sequence of a extended state transport, and then some amount of trapping,
wherein the electron spends some time in the trap states.
And, then gets free again and then moves through the extended states and then gets
trapped again etcetera. And, therefore, one can define if one defines the extended state
877
mobility of the carriers ok. Which is which we call the free carrier mobility as mu
naught, then the effective mobility that on sees for carriers moving in a disordered
material is say mu d which we will call a drift mobility. And, that mu d is given as Mu
naught into the time taken time spent by the carriers in the free states divide by the total
time spent by the carriers in the semiconductor, which is basically a summation of the
total time spent per the carrier in the free states plus the time spent per that carriers in the
trapped states.
So, this is the definition of the effective or the drift velocity or the drift mobility of
carriers. And, this is the mobility definition, that we would use while trying to define the
drift velocity of the carriers.
So, just to summarize ok: so everything we have seen so, far. So, we see that we want a
amorphous semiconductors, amorphous semiconductors for applications and say may a
larger electronic systems or variable electronic systems etcetera, and the consequence of
amorphous semiconductors or the consequence of the disorderness of the lattice.
Is that, you have firstly; you have states in the band gap ok. So, the band gap is not clean
you have states in the band gap. And therefore, you have carriers trapped in the band
gap. And, if one if one is interested in calculating the carrier concentration inside the
878
gap, then we just need to define, you know the density of states in the gap and multiplied
with the Fermi function ok.
The third consequence is that the carriers in the gap ok. So, the carriers can either be in
extended states where they have a mobility say mu naught or they could be trapped in the
gap which are basically a localized states, where they are trapped. And, these carriers do
not in any practical sense contribute to conduction in your semiconductor devices.
And, the carriers that are present in extended states can move with an effective velocity
given by the drift mobility, defined by the drift mobility, which is nothing but your
extended state mobility into this ratio, which is the ratio of the time spent per the carrier
in the extended states divided by the total time spent by the carrier in the material which
is your Tfree+ Ttrapped. And, this is defined by something called as the multiple trap release
mechanism.
So, this is the summary of everything we have looked at so far.
879
Lecture - 52
Thin Film Transistors
So now let us try to quantify the number of trap charges ok. So, it is not only the
qualitative understanding of the disordered semiconductor that is important, but also the
methods and the means to quantify these estimates. And from this point on we will study
a lot of the methods and tools used to quantify and calculate carrier concentrations and
therefore, the currents and disordered semiconductor based devices.
But since the mathematics can get quite involved please keep in mind that the focus is
not to be on the mathematics, but on the general approach to solving these problems. So,
to start with the primary difference between crystalline materials and disordered
materials is the states inside the gap and therefore, the trapped carrier concentration
inside the gap. Now inside the gap the trapped carrier concentration is simply the density
of states into the fermi function f (E).
In the case of crystalline material density of states was 0 and therefore, the trapped
carrier concentration was 0, but in the case of disordered materials this density of states
has to be defined and a key towards understanding the density of states the key towards
880
understanding the carrier concentration and disordered semiconductors is the definition
of the density of states. So, on this page let us just look at you know how one would go
about in quantify the density of states.
So, this particular picture is taken from this reference and it identifies the density of
states picture for amorphous hydrogenated silicon. And such a picture can be developed
for any disordered semiconductor. So, this is just being used as an example. So, in order
to the experimentally estimate the density of states, a one could perform different
experiments such as the deep level transient inspector scopy. And one could get an
experimental; one could get experimental confirmation has to what the density of state
distribution is with energy. Now what is shown here is the acceptor like density of states
in the case of this particular material, where this could be considered to be the fermi level
and then you have the acceptor like deep states and you have the acceptor like tail states
and that is your conduction band edge. Now how does one quantify how does one model
the density of states? A very common model is typically if you have a density state
distribution that looks like this.
Wherein the deep states the logarithm of the density of states versus energy for the deep
states is like a straight line, and the same holds true for the tail state distribution with the
exception being that the slopes are different ok. So, how does one go about modelling
this? A good way to approach this is by defining the density of states to be exponentially
dependent on the energy. So, we say that the density of states is some pre factor into the
exponential of E- EC divided by a fitting parameter which is called as the characteristic
energy.
So, here EC k is nothing but the Boltzmanns coefficient and TC is a parameter which is
called as the characteristic temperature or one could also use the equivalent which is the
characteristic voltage which is k T C/ q. So, this can also be called as the characteristic
voltage. Now using this fitting parameter TC one could define a characteristic energy that
fits this definition that fits the experimental data using this kind of a model. Now this is a
general expression and therefore, you could have one characteristic temperature for the
deep states and one characteristic temperature for the tail state. So, one could say that the
density of states. So, this is a very general expression. So, this is not is only to define
there approached modelling.
881
One could say that the density the acceptor like density acceptor like deep states have a
relation which is says some god which is some pre factor that needs to be estimated from
the experimental data into the exponential of E minus EC by a characteristic temperature
for the deep states. And the acceptor like tail states is given by say g AT naught into
exponential of E-EC/ kT t which is the characteristic temperature for the tail states.
And similarly one could a define the donor like deep states and the donor like tail states
based on the experimental data found. So, the donor like states are not shown in this
picture here. So, this is the first step towards any modelling which is to define the density
of states and we can define this using an exponential definition and with something
called as the a fitting parameter called as the characteristic temperature.
So, the next step is to calculate the total number of localized carriers. So, what do you
mean by localized carriers? These are carriers which are trapped and are sitting in the
gap states. And how do we define the carrier concentration of localized carriers? It is
simply all carriers that set from the fermi level to the conduction band edge. So, here
again we are looking at only the carrier sitting in the acceptor like states. So, from EFO to
EC. And that is nothing but the product of the fermi function. So, here you can see the
fermi function into the density of states we had defined a density of states which is got a
characteristic temperature and exponential dependence on energy. So, that is the density
of states into the fermi function and this integral is performed the respected d from fermi
882
level to the conduction band edge. So, which implies that we are integrating and
calculating all the carriers that are present in this region here, which are our trapped
carriers in the acceptor like states.
Now this integral can the mathematics here is quite involved, but do not focus on the
mathematics depending on the what you say the characteristic temperature.
So, here I should define this particular parameter alpha. So, this particular parameter
alpha is nothing but the ratio of the characteristic temperature to the ambient
temperature. And depending on what alpha is or what the value of alpha is, this integral
can be the answer to this integral can be different. Now it is so, happens that in this
particular case for this particular example, if the alpha is greater than one which means if
the characteristic temperature is greater than your ambient temperature, which so,
happens in the case of deep states, then the number of localized carriers in the deep states
turns out to be this particular answer here.
883
And on the other hand if the alpha is less than 1 which so, which happens to be the case
in the case of tail states.
So, the characteristic temperature for the tail states is less than the ambient temperature
and therefore, this integral the same integral turns out to be a very different gives you a
very different answer which looks like this.
So, as I mentioned in the mathematics is a bit involved, but I am trying to keep the focus
on the approach. So now, we can have a complete definition of the carrier
884
concentrations. So, here are the localized carriers. So, there are 2 kinds of carriers and
disordered materials.
You have trapped carriers or localized carriers. So, these are nothing but the trapped
carriers which are trapped in the states in the gap. And you have the free carriers which
are sitting in the extended states. So, these are the carrier sitting in the extended states.
So, these are the ones that can that can provide that posses a large mobility and which
can contribute to the current.
So, the number of free carriers per unit volume is given by an expression that you are all
very familiar with, which is the density of effective density of states into E to the power
EC- EF-EC- Ef/ kT. So, this is this expression is very similar from the k compared to the
case of crystalline silica. But this is something new which does not exist in the case of
crystalline silicon which is the number of carriers that are trapped per unit volume is
given by the total number of trapped carriers in the tail states plus the total number of
trapped carriers in the deep states.
And those answers are given here for this particular example. So, here is a nice picture.
So, here is the charge per unit volume versus energy. So, you can see that the free carrier
concentration goes like this. And the trapped carrier concentration is shown here right
this is all due to the deep states and that is all due to the tail states. So, it is quite
important to note that at thermal equilibrium it could so, happen that the trapped carrier
concentration is much larger than the free carrier concentration.
885
So, having given that introduction to disordered semiconductors let us now look at thin
film transistors. And we will see we will look at the techniques used to model thin film
transistors, which are basically the field effect transistors using disordered
semiconductors and using a thin film process flow. So, asI mentioned before the thin
film transistor has got thin film metallic gate, a thin film insulator, a thin film of
semiconductor and your source and drain contacts which are tailored such that you have
good carrier injection and therefore, low contact resistance.
Now a key point to note is if you are building transistors out of many organic inorganic
materials or many organic materials the density of states largely determines the kind of
thin film transistor. Again what I mean by that is regards to whether that transistors n
type p type or whether the process allows a complimentary mos device and that is got to
do with the density of states picture. And we will have a look at the working of the thin
film transistor to understand this.
886
Now how does a thin film transistor work? Now as we so, what we have to look at here
is the density of states picture. So, let us say we have a we have applied a gate voltage
we built a thin film transistor and we start throwing in positive charge on the gate.
Now in the case of crystalline silicon, what was the consequence of us putting positive
charge on the gate? In the case of crystalline silicon, if we looked at the, the band gap
was very free and the Fermi level was free to move about or in other words the band
bending was very, very free. So, in crystalline silicon we had a device structure wherein
we had a doped semiconductor. And we had a insulator and the gate and the consequence
of us adding positive charge was firstly, be achieved flat band and then if you use aP-
type silicon body, we saw that the we first achieved depletion. Here we had our intrinsic
fermi level bend towards the fermi level or in other words the conduction band and
valence band began to bend towards the fermi level; we first achieved a depletion where
in all the positive charge here was compensated by the negatively charged ions acceptor
like ions in the bulk.
And then by adding more and more positive charge, we could achieve a significant
amount of band bending; the band bending became very significant and the intrinsic
fermi level bent below the fermi level near the interface. And therefore, the moment Ef
became greater than Efi Ei we found that electrons began to appear at the interface and
we had a region of operation called as the inversion region of operation. And these
887
electrons where the ones that enabled carrier transport between source and drain so, that
was the working of the crystalline silicon semiconductor.
Now in the case of disordered semiconductor the fermi level is not free to move. There is
band bending, but it strongly depends upon the density of states. Now as we add positive
charge in the gate of a non crystalline semiconductor MOSFET or in other words the thin
film transistor is that, the fermi level tries to start moving towards the conduction band
edge ok. In other words if you think off the band bending picture the conduction band
attempts to bend towards the fermi level.
But since there are so, many states sitting in the band gap, the movement of the fermi
level towards the conduction band implies that these states must now be filled with
electrons before the fermi level can make progress and move towards the conduction
band edge. Or in other words all the positive charge that we are on the gate is being
compensated by the trapped charge that is appearing in the gap states as the fermi level
makes it is move towards the conduction band edge.
So, let us say this is the let us say that figure 1 is the position where your Vgs is equal to
flat band. So, all bands are flat at this point.Now as we add more and more positive
charge on the gate, we find that the fermi level starts moving towards the conduction
band edge or in other words the conduction band tries to start bending towards the fermi
level, but since the fermi level is now climbed up by this distance as compared to it is flat
band point all these states are now filled with carriers there all filled with electrons and
that occupancy depends upon the product of the fermi function and the density of states.
Now it is these electrons that are going to compensate for the positively positive charge
that we have applied on the gate. In the case of crystalline silicon we found that it was
the band bending. And it was the exposed dopant ions that was contributing or
compensating for the positive charge on the gate, but here it is the trapped charge or
these electrons that are occupying acceptor like states that are going to contribute to the
charge balance. So now, the meaning of acceptor like states becomes very clear right.
So, as the fermi level moves up these states were filled with electrons. And therefore,
they all became negatively charged. So, they are playing the same role as the doped
acceptor like ions in the case of an n channel MOSFET. So, as the fermi level moves up
the acceptor like states get filled with electrons and this charge compensates for the
888
charge on the gate. So, as more and more positive charge is applied on the gate, the fermi
level starts heading closer and closer to the conduction band edge.
And at some point there will be enough free carriers sitting in the conduction band edge
above the conduction band edge in the extended states to permit significant current. And
this point is defined as your threshold voltage ok. So, look at the clear difference
between the operation of a MOSFET and that of a TFT.
So, let us spend some time looking at this difference in the case of crystalline MOSFETs
what defined are threshold voltage what was the threshold voltage in.
The case of crystalline silicon MOSFETs, we had a gate and we had an insulator and
when we do the band bending diagram then since the band gap was very clean the
conduction band could move to move towards the fermi level very freely. So, at first we
had to achieve flat band. That was the first milestone and then. So, that was the let us say
that is the intrinsic fermi level and that was the fermi level with the p type doped
semiconductor.
So, first we achieved flat band, but then as we started adding more and more positive
charge, the fermi level moved closer to the conduction band or in other words there was
some band bending. There was some band bending. And all the positive charge in the
gate was compensated for by the depletion charge in the semiconductor. So, the next
889
milestone toward achieving flat threshold voltage what is that we need to deplete the
semiconductor enough in order to get the bands in order to get the bands to bend
significantly. And finally, we had to achieve inversion. And how was the inversion point
defined? It was that point at which the band bending was.
So, significant that the electron concentration of the interface managed to match the hole
concentration in the bulk. So, we said that this was equal to ∅F and therefore, the surface
potential had to be 2∅F in order to achieve threshold voltage. So, this had to become 2∅F
and this point we said that there is a enough inversion charge that we considered to be
the and we considered the MOSFET to be working above threshold voltage. So, that was
our definition of threshold voltage. What about crystalline non crystalline semiconductor
a amorphous semiconductors? In this case there is no, there is nothing like a depletion
layer.
The semiconductor is intrinsic there is nothing like a depletion layer. There is nothing
like achieving these 2∅F. So, all this there is no there is no inversion point where in the
surface potential achieves to ∅F. So, these definitions do not exist in a similar manner as
in the case of crystalline semiconductors. Instead what we have is we have a conduction
band we have states inside the band gap. And let us say this is the fermi level at flat
bands. So, yes you do have flat band. So, we have the first milestone to crosses the flat
band, but threshold voltage is defined as that voltage, wherein the fermi level gets close
enough to the conduction band edge.
So, that all the states above the conduction band edge the extended states are populated
with some amount of free carriers that in enable a measurable level of conduction. So,
that is the definition of threshold voltages which is a very vague definition. Because it
depends upon your density of states profile. So, the density of state distribution is very
large. You need to trap a lot more carriers before the fermi level can climb till that point.
And therefore, your threshold voltage is going to be much larger ok and therefore, the
threshold voltage the proper definition of threshold voltage. So, let us say that this is the
location of the fermi level when the carriers and there is significant carrier population in
the extended states. So, the definition of threshold voltage is that, the fermi level needs
to, needs to climb up and passed.
890
These gaps or passed all these states which implies that all these states first need to be
filled with electrons and therefore, all this negative charge will balanced all the positive
charge applied on the gate. And therefore, the definition of threshold voltage is the flat
band voltage plus this total trapped charge in all the states divided by COx. So, how
much of charge do we need to trap in all these states so, as to get the fermi level close
enough to the conduction band edge.
So, as to get enough carriers in the extend with states. So, that is the definition of
threshold voltage. And this depends upon the density of states picture. So, we can see
how the density of states is controlling the operation of the MOSFET. Now there is
another point, let us look at the mobility of carriers. So, in the conduction band states ok,
let us look at the transients; let us look at the transients. Now as the fermi level moves
through the states.
We have something called as a sub threshold operation. So, here was the flat here was
the flat band voltage. So, there was the flat band voltage and between this point to that
point we have sub threshold operation. So, in sub threshold mode of operation the fermi
level sits mostly in the deep states. It is only when you are close to threshold voltages of
the fermi level as climbed far enough that has gone into that it has gone into tail states.
So, in sub threshold operation the fermi level sits in the deep states.
But then as the fermi level makes it is climbed the speed at which it is make it makes it is
climb towards the conduction band depends upon the density of states here. So, let us say
that that there are a large number of states to be filled. The fermi level will climb very,
very slowly. Because it needs to fill all these states with the electrons before it can move
upward. But then if there are very few states to be filled then the fermi level can climb a
little faster.
So, depending. So, let us say that the fermi level let us say that the density of states looks
like this. The fermi level is climbing through very slowly because there are large number
of states. And then it increases it is speed and then it goes up very, very quickly through
this region because there are very few states. And then it becomes slow again because
there are a large number of states again.
So, this implies that the threshold slope or the sub threshold swing would reflect this
behaviour you will find that the current is growing very, very slowly because the fermi
891
level is moving very slowly through the gap states and then in this region as the fermi
level begins to move quickly you will find that the current also increases a rapidly. And
once the fermi level enters these states you will find the current increases slowly and
then the MOSFET get enters above threshold operation. So, in some sense the sub
threshold slope and the sub threshold swing are all dependent on the nature of the density
of the deep states. So, that is the second difference.
So, therefore, in the case of crystalline silicon, we found that the sub threshold slope if
you look at the sub threshold slope it depended on something called as 1 plus C depletion
by COx. And because we had interfacial trapped charge, you have see interfacial. But
here one can define the sub threshold slope as 1+C Cinterfacial+C gap states by COx. The
third point the third point to note the third point to note has go to do with the mobility of
carriers. So, why did we say that why did we say that the amorphous silicon
semiconductor has more n type behaviour as compared to p type behaviour. And the
answer to that also lies in the density of states (Refer Time: 25:57).
So, let us say we have 2 options we have the fermi level sitting here at flat band and we
would by applying positive voltage like to move the fermi level closer to the conduction
band edge, as compared to moving the and by applying negative voltage we would like
to move the fermi level closer to the valence band edge. Now let us see which of these is
more feasible.
Now it so, happens in amorphous silicon, that the density of states in the valence band or
that valence band or the donor like tail states has got a much larger density as compared
to the acceptor like tail states. So, you can see that this slope here is very, very sharp.
Whereas, this slope here is very, very large. Now because this asymmetry you have a
very different different is got a very direct consequence on the practicality of having hole
type transport or electron like transport.
So, in order to create holes, in order to allow holes to transport, be the transport be the
carriers of current transport, we need the fermi level to move towards the valence band
edge. But this because of the large density of states this can happen only at very, very
large negative voltages. On the other hand it is much easier to get the fermi level closer
to the conduction band edge because of the smaller density of states and the tail states.
892
So, the presence of gap states determines. Firstly, the threshold voltage secondly, the sub
threshold slope and thirdly the type of transport is at n type or a p type transport. And
this is the significant difference between disordered semiconductor based MOSFET
which are called as thin film transistors and the crystalline semiconductor based
MOSFETs.
So, let us now look at, let us use all these information to try and see you know the
methods is to develop analytical models for thin film transistors. So, again here the
mathematics might be very involved, but the focus should be on the approach as
compared to the details of the derivations. So, the first step is let us look at the MIS
electrostatics. Which is equivalent to your MOS capacitor the electrostatics of there
MOS structure. So, the first step is to write down Poissons equation. In the case of you
know your crystalline silicon MOSFETs.
We found that Poissons equation was enough you write it in terms of the electric field, it
was given by d by d axis qNA/𝜀 s where NA was the acceptor ion concentration in the P
type bulk which we had used to develop our MOSFET. And in terms of the potential, it
was d2 ∅/dx2 is qNA/𝜖𝑠
Now in the case of amorphous semiconductors, what determines the Poisson equation?
So, what determines the charge carriers. So, this charge carrier concentration is given by
the trapped charge. So, what constitutes all this charge carriers is the trapped charge, plus
893
the free charge. So, if you go back to this general picture of which you know we looked
at the movement of the fermi level through the gap states, we saw that all these trapped
charges were the ones that were responding to the applied electric field or the applied
gate voltage. And we also had some amount of free carriers depending upon the location
of the fermi level. So, any electrostatics with regards to the band bending or to identify
the electric fields inside the semiconductor must consider the trapped charge
concentration and this free carrier concentration.
So, therefore, we now have a Poissons equation which is got a free carrier concentration,
which is got a trapped carrier concentration. And this is the trapped carrier in the tail
states and the trapped carriers in the deep states. So, you have 3 terms, which is the free
carrier sitting above the conduction band edge, the trapped carriers in the tail states and
the trapped carriers in the deep states. Now each one of these terms depends upon the
potential that is applied.
So, therefore, this is a Poissons equation which will also have the potential term on the
right hand side. And all of you are familiar as to the means to solve this particular
equation. But the free carrier concentration will have an e to the power phi by phi th term
where this is the thermal voltage, the trapped carrier concentration in the tail states will
have an E to the power phi by phi t term, where this is the characteristic voltage of the
tail state trapped charge. And the deep trapped charge carrier concentration will have an
E to the power e5/∅d where ∅d is the deep the deep state characteristic voltage.
So, using these 3 terms one can write out poissons equation and solve it in all it is detail,
but for the purpose of these slides what we have done here is we have merged all 3
exponentials and defined an effective carrier concentration. And an effective
characteristic voltage and written the sum of these 3 terms as simply this particular
exponential ok. So, this need not be this is definitely not the most accurate approach, but
it is being done only for the purpose of this example here. So, what is this mean? It says
that if the trapped carrier concentration is the largest if say one of these 2 is the largest
than any o this term will d will contain of mostly trapped carriers. And this characteristic
voltage will tend towards one of these characteristic voltages.
On the other hand if this term here has got mostly free carriers then any o will tend
towards this particular concentration that the concentration of nf o. And this term here
894
will tend towards ∅th. So, that is the implication of having this exponential. Now all that
is left for us to solve this particular exponential.
And here is the details of all the solution. So, we have this Poissons equation to be solved
and there are there are 2 boundary conditions. The first boundary condition says that at.
So, when you have band bending in the semiconductor the point at which phi is 0 that is
when we achieve flat band, we have defined phi the reference potential as on the
potential in the bulk. So, when phi is 0 you have a 0 electric field. That is when the bands
become flat the electric field is 0. The second boundary condition has got to do with the
interface.
And what we say there is that the electric displacement which is the permittivity into the
electric field at x equal to 0 is given by Ci.(Vgs-Vfb) minus the surface potential. So, that
is the second boundary condition. So, by applying these 2 boundary conditions and going
through all this mathematics, we will find that this is the expression for the electric field
and you know going through all this mathematics by applying both the boundary
conditions we find that the surface potential is defined in terms of a Lambert W function.
895
Which can be approximated as this particular expression here. And the surface potential
therefore, is and the potential through the semiconductor not the surface potential, the
potential through the semiconductor is given by this particular expression. The surface
potential in particular is the potential at x equal to 0.
Now, the next step towards modelling is now, we have the potential function and we
have this surface potential. Now the surface potential determines the free carrier
concentration per unit volume. Because of free carrier concentration per unit volume is
896
defined by this particular expression here which is your nfo into e to the power phi if e to
the power phi by phi th, but the surface potential defines the free carrier concentration
per unit volume near the surface. But however, for the definition of your current
densities it is useful to define the free carrier concentration per unit area.
And the way this is done is by integrating nf in thickness and you integrate it from 0
through the thickness of the semiconductor. So, the point is thin film transistors are made
up of thin film semiconductors. And this thin and what we are saying here is that nf is the
free carrier concentration through the entire thickness of the semiconductor. And by
performing this integral we identify the free carrier concentration per unit area.
So, that that integral is solved out here in full detail. And we find that the free carrier
concentration per unit area is given by this particular expression here. And you see 2
parameters that did not appear in the case of crystalline silicon MOSFETs. In the case of
crystalline silicon MOSFETs the free carrier concentration per unit area was simply Cox
Cox(Vgs-VT- Vc). But here you have 2 other parameters you have an exponent which is
called 𝛼 −1 and you have a pre factor which is called as 𝛾.
So, what are alpha and gamma? 𝛼 is your twice your effective is 2 times the effective
voltage divided by your ∅th. In other words your 𝛼 is nothing but the term that we
defined earlier which is your twice your characteristic temperature divided by the
ambient temperature.
897
And in the case of crystalline semiconductor materials, this characteristic temperature
was the same as the ambient temperature. Or in other words phi e was known to be the
same as ∅th. And therefore, 𝛼 was the same as 2. And therefore, this term here became 2
minus 1 which was 1. So, therefore, crystalline silicon MOSFETs did not have any factor
there. But this is the more general version of the semiconductor and you find that 𝛼 is
twice TC/ T.
So, that is your parameter 𝛼. That depends upon the characteristic temperature of the gap
states. And what is gamma? 𝛾 is defined by this particular term. The key point here is
that gamma defines the number of carriers in the free states divided by the total effective
number of carriers.
Which in some sense includes the carriers in the trap states plus the carriers in the free
states. Therefore, gamma as an indicator of this ratio of the free state carrier
concentration by the total carrier concentration. And using this particular relation, which
is number of the number of carriers per unit area in order to define your current voltage
characteristics we find that the current voltage characteristics is given by this expression
which can then be integrated to define the current voltage characteristics of your TFT.
Ok. So, in the case of MOSFETs alpha was simply 2 and we ended up with this
particular relation, but even in the case of TFTs if these particular assumptions are made
then you will find that there character is that current voltage characteristics of a TFT is
898
quite similar to the current voltage characteristics of a MOSFET. So, in some sense this
is a summary of how one goes about with regards to deriving there current voltage
characteristics of a TFT. Now as we saw this term threshold voltage in the TFT is given
by the total trapped charge concentration by Cox plus the flat band voltage.
And this plot here shows that typical transfer and output characteristics for a TFT. Now
since the threshold voltage is nothing but the trap charge by Cox+ Vfb this is a trapped
charge per unit area divided by Cox+ Vfb the threshold voltage ends up being a very
dynamic quantity.
899
And once C is a very unique phenomena in disordered semiconductors which is

something called as the threshold voltage shift. So, what is this mean? So, in the case of
crystalline silicon MOSFETs the threshold voltage was more or less a fixed quantity. I
mean we once you know the threshold voltage at the MOSFET you used it for modelling
your circuit behavior. But in the case of disordered semiconductors when a MOS when
the thin film transistor is an operation.
So, let us say you apply a certain gate to source voltage and you apply a certain drain to
source voltage. Then when this thin film transistor is an operation you find that the
threshold voltage in a thin film transistor becomes a function of time. In other words as
the MOSFET current as the MOSFET is driving current, the current in the MOSFET
begins to decrease in time. Why does this happen? So, this is a very key phenomena
which is very peculiar to disordered semiconductors.
Now there are 2 mechanisms which result in threshold voltage shift. The first mechanism
can be said to be both these mechanisms are related to charge trapping, but the first
mechanism is charged trapping in the dielectric and in interfacial states. So, although it is
not written here, you can say it is dielectric and interfacial states, that is at the
semiconductor insulator interface. So, these 2 are the first parameters of mechanisms that
contribute to charged trapping.
900
There is another mechanism which is which contributes to charged trapping and that is
related to something called as defect state creation. So, what is defect state creation? So,
we saw that there was a certain density of states in the semiconductor. So, that was the
energy and this was that density of states. Now its so happens that as the MOSFET is in
operation. So, you have your fermi level here, the MOSFET is an operation or the TFT is
an operation.
So what happens is that due to the disorder in the semiconductor and due to electrons
interacting with these weak and dangling bonds, stronger defects are created. In other
words new states are created deep inside the band gap ok. So, the presence of these new
states results in all the free electrons now getting trapped in these states or in other words
it appears like as though the fermi level is slipping down in energies even though we
have a large gate voltage.
And this movement of the fermi level down in energy simply because of the creation of
defect states results in it appearing like as though the threshold voltage of the transistors
increasing with time. So, it is a very dynamic phenomena and one finds the threshold
voltage the transistor increases with time and is given by a relation that looks like this.
Ok something called as a stretched exponential function and that experiments show this
which is if you apply a certain gate voltage and if the there is a that the gate voltage is
larger the rapidity with which the speed at which the threshold voltage increases will also
be larger. So, this is a phenomena that is quite unique to disordered semiconductors and
it is of quite some importance.
901
Lecture – 53
Tutorials Session - 1
Hello everyone, welcome to the course Semiconductor Devices and Circuits. We have
headed towards the end of the course.
So, we thought we will take a tutorial session for the remaining two or three lectures.
Hope you have enjoyed the course.
902
We have designed the Tutorial Session in such a way that you will get many problems
from the assignments and we will introduce some new concepts also so that overall it
will give you a recap of the entire course. So, let us first jump into the first problem; this
from the quantum mechanics.
The problem states that an electron is confined to an infinite 1D potential well of width L
extended from x equal to 0 to x is equal to L. So, first is you have to find out the
probability of finding the electron between x equal to 0 and x equal to L by 3 at the
ground state. And second one is the electron if the electron jumps from the n equal to 3
energy level to the ground state energy level and in doing so, it emits a photon of
wavelength 20.9 nanometer then what will be the width of the well?
So, there are few things you have to note down here. First is the infinite 1D potential
well which has been covered in the course, in the first topic. That first topic was that
quantum mechanics infinite 1D potential well. Second thing is basically the well is
extended from x equal to 0 to x=L. Now, the first problem states that you have to find the
probability of finding the electron between x equal to 0 and x=L/ 3 at the ground state.
So, just recall the wave function of the 1D potential, infinite 1D potential well. If the
particle is at nth state then the wave function is given by psi 𝜓nx equal to root over √2/L
sin integral multiple of 𝜋/L into x. Where basically, n is any integer and the
corresponding energy of the n th sate is given by En is equal to n2𝜋2h2 /2mL2.
903
So now, as the; so first thing is the; what will be the ground state? Ground state will
correspond to n equal to 1. Just remember n is equal to 0 is not the ground state because
if we put n equal to 0 in this equation, here this wave function will vanish because sin 0
is equal to 0. So, and also the energy will become 0. So, n equal to 0 is not any valid
state. So, n equal to 1 is your first state and that is why you call it a ground state. Now,
probability of finding the electron per unit length at x is given by modulus of 𝜓1x whole
square and this nothing but your pdf, probability density function.
So, the probability of finding the particle or the electron here between x and x plus dx
will be given by |𝜓|1 x whole square into dx. That is pdf into dx because you want to
find the probability between x and x+ dx. So, this is your dx length. So, what will be the
probability of finding the electron between this dx length that is given by this
expression? So, the probability of finding the electron between 0 and L by 3, you have to
basically integrate this expression from 0 to L by 3.
So, just go through the steps few steps here. So, integrate from 0 to L by 3. So,
magnitude of (𝜓1x)2 whole square will be given by 2/ L then, sin2 𝜋x/ L. So, you have to
convert this sin square to cos 2x in order to do the integration. So, that will come as a
standard integral. So, 2 by L integration over 0 to L by 3, you can take this 2 inside. So,
it will become 2 sin square. So, basically 2 sin2𝜃 that is equal to 1- cos2𝜃. So, you have
to plug in here 1- cos I will just erase this one this mid space
904
cos 2 pi x by L, there will be one dx. So, if you do this integration it will come as 1 by L.
Here it will be x minus. So, you have to do the integration of this. It will be L by 2 pi sin,
integral of cos is sin 2𝜋 x/L.
So, your limit you have to put the limit from 0 to L by 3. So, now just look into this
expression, if you put 0 both will be 0; this two term. So, you have to put this L/3. So,
only L/ 3, the upper limit will that will give you some value, L/ 3- L/2𝜋 sin 2 𝜋/3. So,
you just cancel out L here. So, it will come as one-third minus 1/ 2 𝜋 sin 2 𝜋/3.
So, if you just calculate this, this probability will come as 0.2; just check this with
calculator. So, what is the next problem? It was if the electron jumps from n equal to 3
energy level to the ground state energy level, so, again from n is equal to 3 to n is equal
to 1 your ground state corresponds to n=1.
905
So, basically, En that is given by this expression that is the energy of n-th state. So, if it
jumps from n equal to 3 to n equal to 1, it emits an energy which corresponds to del E is
equal to pi square h quartz square by twice 2mL2 square into n equal to 3 square minus n
equal to 1 square. So, it will give you some expression here.
Now, this energy is basically as it emits a photon, so this energy should be equal to the
energy of the released photon. So, photon energy then is given by E ph photon energy,
this photon energy. This is equal to hc by lambda and that should be equal to the energy
of that released energy. So, hc by lambda should be equal to 4 pi square h quarts square
by mL square. So, let us just do some simple algebra here; hc by lambda. This is equal to
4 pi square by mL square.
Let us convert this h quart into h/ 2𝜋; h quart this is just a reduced Planck’s constant. So,
this is equal to h/2 𝜋 pi. Let us plug in this expression here, h2 / 4𝜋2. So, this will be
cancelled out, one h we can cancel here, L will become equal to h𝜆/mc ok. So now, you
just plug in the numbers given in the problem. So, your wavelength is given as 20.9
nanometers. So, lambda is equal to 20.9 nanometer; h is the Planck’s constant whose
value is given by 6.626 E- 3; 34, m will be the mass of the electron; c will be the velocity
of light in free space.
So, if you plug in those numbers that L will come as 0.225 nano meter. So, that is the
width of your infinite 1D potential well.
906
So, now let us come into the next problem. This is from the semiconductor basics and
this basically deals with the continuity equation. This also was given in as an assignment
problem. So, I thought we could discuss in details that problem here. The problem states
that assume a p-type silicon sample with the following parameters at room temperature
300 Kelvin. So, 300 Kelvin means, the thermal voltage here will be equal to 0.026 volt.
Some parameters are given for example, the doping density that is NA is equal to 10 to
the power 17 per cc. Mobility of the electron is equal to 300 centimeter square per volt
second and tau n equal to 1 microsecond. Just remember here electron is your minority
carrier.
Because it is a p-type silicon and the sample is uniformly eliminated with light from the
top as shown in this figure. So here, light is coming and basically absorbing those
photons, electron hole pairs are generating. In this generation rate is also given optical
generation rate is GL 10 to the power 24 per centimeter cube per second. So, 1 constraint
here is given as the incoming photons are absorbed in a thin layer of 10 nanaometer of
the surface. So, if we assume here a 10 nanometer thickness on the surface, all the
photons are absorbed over this layer.
So, basically what does it mean that there the electron hole pairs are getting generated
over this thin layer only. There is no generation inside the bulk. So, if we consider this is
the x axis then just denote this mark as the x equal to 0. So, inside the bulk means for x
907
greater than 0, there is no generation of electron hole pairs. So, basically we can say that
this generation rate becomes 0 inside the bulk here. Now, the problem is you have to find
out the steady state excess electron concentration at a certain distance 1 micrometer in
the 2 following cases.
One is if the sample extends to infinity along the x axis. So, if this dimension a goes to
infinite in the first case and the second case states that if the sample is only 5 micrometer
long along the x axis. So, if this sample has a finite dimension, so, basically in the second
case it is given that this dimension of the device is 5 micrometer and also in the second
case it is given that on the other end, after this 5 micrometer length, there is an ideal
ohmic contact at x equal to L equal to 5 micrometer that always enforces the equilibrium
condition.
We will just discuss this in while solving the problem. So, there 2 assumptions you can
consider here. One is the there is no electric field present inside the bulk and you can
consider the low level injection. So, this will basically help you reduce the minority
carrier diffusion equation.
So, let us try to solve this problem. In the first case where this goes to infinity and in that
case, we can write down the minority carrier diffusion equation for x greater than equal
to 0 as del n t is equal to Dn 2 Δn Δ x2 – Δn/𝜏𝑛 +GL square minus del Δn by n+GL. So,
this is the first equation. So, here Δn is the excess minority carrier concentration, that is
908
the electron concentration and D n is the diffusion coefficient, tau n is the electron life
time and GL is the generation rate.
Now, so, basically what does it state that the electron this electron and holes are
generated here and this electrons will diffuse to the bulk according to this following this
relation. So, why we have considered this; because our electric field equal to 0 and if you
recall the continuity equation, we can reduce in case of electric field equal to 0, we can
reduce the equation to only diffusion term. So, that is why this minority carrier diffusion
equation looks like this.
So, assumption is electric field equal to 0 and another assumption is low level injection.
Under these two assumption, the minority carrier diffusion equation can be reduced to
this. Now, the problem ask about the steady state. So, now, when we are talking about
the steady state there will not be any change of a carrier concentration with respect to
time. So, del n del t will become 0, just one second. So, del n del t will become 0 in case
of steady state and other terms will remain.
So, this is the first step where we have reduced for the minority carrier diffusion
equation. The next reduction will come when considering the fact that there is no
generation of electron hole pairs inside the bulk. So, beyond that 10 nanometer thin layer
that generation rate will become 0. So, this will again become 0 for x greater than equal
to 0. So, for x greater than equal to 0, the minority carrier diffusion equation can be
reduced to this final term d 2 d del n dx square minus del n n square is equal to 0. So, just
probably you have noticed one thing that this dou has become now D.
So, basically the partial derivative we have converted to the direct derivative because this
del n, here in this first equation it is a function of x and t, but as we are talking about now
steady state, so, there is no variation of time. Hence, we are considering now del n as a
function of space only. So, there is only a single independent variable and hence we can
convert that partial derivative to the direct derivative d 2 dx square.
So, another thing we have just noticed this Ln. This is the diffusion length that is equal to
root over Dn tau n; this is the diffusion length of the minority carrier that is the electron
here. Now, so basically we have to solve this differential equation under a particular
boundary condition. So, what will be the general solution of this equation?
909
This general solution will be given as A ex to the power x by Ln plus B e to the power
minus x by Ln. So, you just recall your highest you just recall your undergrad maths
where this we consider a trial solution e to the power m x to solve this kind of
differential equation and L end up with this general solution.
So, now the neat thing is we have to solve this, we have to find out these 2 coefficients A
and B imposing a particular boundary condition. Before that just find out the what are
that these Ln and Dn values from the given parameters, Dn that will be that can be found
using the Einstein’s equation.
Dn is equal to thermal voltage into the mobility and that is given by 0.026 into 300
centimetre square per second is equal to 7.8 centimetre square per second. And what will
be the diffusion length? Diffusion length can be found using this equation and that is
given if you plug in those numbers, this Ln is coming as 27.9 micrometer. So, we will use
this while deriving the final answer ok. So, in the next step we will we will incorporate
the different boundary conditions.
The first boundary condition will come from the fact that at infinity, so as the sample that
is extended to infinity, so if we approach to infinity the minority carrier concentration,
excess minority carrier concentration will go down to 0 because all these excess minority
carriers that are coming from the thin interface layer to the bulk they will recombine with
the oppositely charged holes and hence the excess minority carrier concentration will go
910
down to 0. The second boundary condition will coming from the fact that will come from
this surface boundaries.
So, it tells that this minus Dn dΔ n dx at x=0 this is the rate at which the electrons they
diffuse away from the thin interface thin surface layer to the bulk and this rate should be
equal to the rate at which the electrons are getting generated at the thin interface layer.
GS is basically the generation rate, but this is surface generation rate not the volume
generation rate. So, basically you have to multiply the volume generation rate with the
thickness of that thin layer in order to find what is the number of charge carriers, that is
getting generated per unit area inside that thin semiconductor layer.
So, that is you just plug in those numbers and GS will come as this. Now, you have to use
these two boundary conditions to solve the to find out that these 2 coefficients A and B
in equation 4; del n so that is the general solution del nx equal to A e to the power x by
Ln plus B e to the power minus x/ Ln.
So, first plug in the first boundary condition, del n x tends to infinity that is equal to A e
to the power infinity plus B into e to the power minus infinity and this is equal to 0
according to the boundary condition. Now this term goes down to 0. So, basically it
comes as A e to the power infinity is equal to 0. So, your A must be equal to 0. So, only
the second term will that will exist, the first term will go down to 0 from the general
solution.
So, now the general solution under the first boundary condition is reduced to this term B
exp(-x/Ln). Now, you will use the second boundary condition; second boundary
conditions states that minus DnΔn dx at x equal to 0 is equal to GS. So, just solve the
second boundary condition. So, now, your del nx this is the reduced general solution B to
the power minus x=Ln.
911
So, d del n x dx this is equal to B- 1/ Ln e-x/ Ln. Therefore, dnxdx at x equal to 0 will be
equal to minus B/ Ln.
So, let us plug in this expression into the second boundary condition minus Dn into
minus B by Ln that is equal to GS minus will cancel out. Therefore, B will be GS by Dn/
Ln. So, that is what B comes under the second boundary condition. So, now, we get the
excess carrier concentration profile that is a spatial distribution of the excess electron
concentration that is given by this expression. So, now the you have to just plug in the
numbers.
GS/ Dn Ln you can simply directly you can directly find from the previously calculated
values of Dn, Ln and GS. So, you plug in those numbers here; x is given as 1 micrometer.
Basically, you have to find the excess electron concentration at 1 micrometer. So, you
plug in that number here. You will find that what is the excess electron concentration at x
equal to 1 micrometer.
912
Next part of this problem, where it states that you have to find the excess electron
concentration again at x equal to 1 micrometer, but here the boundary condition gets
changed because of the finite dimension of the device.
Now, the device dimension becomes L equal to 5 micro meter and this is x equal to 0.
So, instead of infinitely extended device, now the device has a finite dimension and that
is given by L=5 micrometer.
So, this is your starting point that is a minority carrier diffusion equation at steady state
and under the two assumptions that there is no electric field inside the bulk and there is a
low level injection. So, this is the differential equation which dictates the transport of the
minority carriers inside the bulk that is for x greater than equal to 0. Now, here as
calculated previously that the diffusion length of the electron is 27.9 micrometer and the
device dimension is given as 5 micrometers. So, if you compare these 2 values, Ln we
can consider as much larger much greater than L.
So, what does it signify that Ln is basically the diffusion length that means, on an average
after injecting an electron, on an average that electron goes 27.9 micrometer before it
recombines with a hole, but we have imposed a boundary at L equal to 5 micrometer. So,
we can consider that all the electrons that are coming from the thin interface layer to the
bulk they will go to the opposite to the electrode which is placed at l equal to 5
micrometer without recombining with the oppositely charged hole.
913
So, basically we can ignore any kind of recombination. So, we can reduce this minority
carrier diffusion equation further by considering Δn is equal to 0; that means, the
recombination term there is no recombination. So, this recombination term will go down
to 0. So, the minority carrier diffusion equation that gets reduced to this d 2 del n dx
square is equal to 0. So, the general solution of this differential equation is given by Δnx
equal to Ax+ B. That is the linear equation.
Now the boundary conditions, boundary conditions will be one will be the excess
electron concentration at x equal to L will be equal to 0. This is coming from the fact that
at x equal to well, we have considered a perfectly ohmic contact. Now, this ensures that
whatever excess electrons are coming at this point, they will be immediately taken that
will be immediately removed by that electrode. So, that there a thermal equilibrium
condition can exist.
So, your n at x=L becomes 0 because of that perfectly ohmic contact. The second
boundary condition that is coming from the surface, so, that is similar to the previous
case because again at steady state the generation rate of the electrons inside the thin
semiconductor layer should be equal to the rate at which that is moving away from the
interface to the bulk.
So, this equation gives that boundary condition. So, basically using these two boundary
conditions, now we have to solve this basically you have to find out these 2 coefficients
914
A and B. So, now the general solution is del n equal to Ax +B. Plug in the first boundary
condition. It states that del n at x equal to L that is equal to AL plus B that is equal to 0.
Therefore, B = - AL and if you plug in this into this expression basically you can reduce
further the general solution as del n equal to A into x minus L.
Now you plug in the second boundary condition. So, second boundary condition will
come from basically you have to differentiate this expression first; A into x minus L. Let
us differentiate this expression first that is equal to A. So, let us plug in this expression
into the second boundary condition minus Dn. Now, this is a constant. So, it does not
matter if whether we are taking at x =0 or at some other x value.
So, you can just plug in directly this number that is equal to GS. So, this will give you
the value of A that is equal to -GS/Dn. So, now the excess electron concentration will
come as del nx equal to GS/ Dn into L- x. So, now you have to just simply calculate the
numbers. GS/ Dn that is given by GS, we have calculated in the previous part. That is 1e
18 Dn that is again we have calculated previously as 7.8. So, GS/Dn is coming with this
number. So, now, therefore, del n is given as this GS/Dn.
So plug in that number into L-x. L is here 5 micrometer, so, 5 minus x. We basically
have constructed this expression such a way that x is in micrometer. So, you have to
convert this into centimetre in order to get the whole expression in per centimetre cube
value. Normally, we express the any concentration as per centimetre cube value. So, I
have converted that micrometer into centimetre 1e minus by multiplying with 1e minus
4.
So, this is the final expression where x is in micrometer. So, now, you just plug in at x
equal to 1 micrometer and you will get the answer for what is the excess electron
concentration at x equal to 1 micrometer.
915
Lecture – 54
Tutorials Session – 2
So, now let us move into the next problem. So, the problem states that there is a
chromium and p-type silicon metal semiconductor junction. The semiconductor is doped
with NA= 1017 per centimeter cube. So, first you have to calculate the Schottky barrier
height and the built-in potential at T is equal to 300 Kelvin. Second you have to calculate
the potential drop across the semiconductor in volt. The depletion layer width in
micrometer, the magnitude of electric field at the metal semiconductor interface in volt
per centimeter which is nothing, but the peak electric field and the junction capacitance
per unit area in nano farad per centimeter square all under the case when the junction is
subjected to a 5 volt reverse bias. And, some useful parameters are given.
916
So, let us try to solve this of the thermal equilibrium case first. Here this is a energy band
diagram when the junction is a thermal equilibrium. So, here we have a metal, this side is
the metal and this side is a p-type silicon this metal is nothing, but chromium and this p-
type silicon. So, at thermal equilibrium the Fermi level will be aligned, this is the Fermi
level which is aligned throughout the device and there is a band bending in order to
ensure that the there is a continuity of the energy vacuum level. This ensures that the
electron affinity of the semiconductor and the metal work function that remains constant
throughout the structure.
So, by looking into the energy band diagram you can see that the Schottky barrier this is
the Schottky barrier that is given by q∅B, this is the height of the Schottky barrier. So,
this is the height of the potential that is basically that will be filled by the electron when
it is trying to move from the metal to semiconductor. On the other hand, built in potential
is basically the barrier which is failed by the electron when it is trying to move from the
semiconductor to the metal. So, this is your built in potential.
Now, how will you find the Schottky barrier height? So, basically if you look into this
diagram so, this Schottky barrier height will be given by your, this entire height that is
the electron affinity plus the band gap, this total height, minus the metal work function
this side. So, that will give you this barrier height. So, that is what here it is written that
q∅B is equal to the electron affinity of the semiconductor this side plus the energy band
917
gap this side minus the metal work function this side. So, that is coming as 0.67 electron
volt by plugging in all the numbers given in the problem.
Now, when it is poured is what is the built in potential. So, basically I have to find out
the built in potential now. So, the built in potential you can find from the Schottky
barrier height. So, basically now we have to find out what is the value of this height and
that can be found from the Schottky barrier height minus the minus this much which is
written as q∅F which is nothing, but the separation of the Fermi level from the valence
bandage eV.
So, basically we know the Schottky barrier height we have to find out now the value of
phi F and value of ∅F can be easily found by using this expression phi F equal to kT/q ln
N V/ N A and that is given by 0.026 into ln plug in all the numbers 0.026 because it is a
300 Kelvin given in the problem. So, your ∅F is coming has 0.14 volt just subtract these
two you will find the built in potential that is given by 0.53 volt.
So, let us jump into the next part where it is given that now the junction is reverse bias.
So, basically you have this metal semiconductor junction which is a p-type silicon the
semiconductor. Now, the metal in case of reverse bias, the metal will be at higher bias
and semiconductor will be at lower potential. So, this is how the junction or the diode is
biased 5 volt.
918
So, in usual notation we can write this applied bias is equal to minus 5 volt. So, V a if it
is positive it signifies that it is forward biased when Va is negative it signifies that it is a
reverse bias phase. So, in this particular case Va is negative. So, V a is equal to minus 5
volt. This semiconductor is at a lower potential this Fermi level will basically rise. So,
there will be a separation of the Fermi levels between the metal Fermi level and the
semiconductor Fermi level. So, this is a separation now between these two energy levels.
And, this separation is given by the amount of the applied bias. So, q into magnitude of
Va is the separation between these two Fermi levels. So, the result is the more band
bending more amount of band bending because Schottky barrier height cannot change.
What will change effectively the height of the potential that is seen by the electron when
it is trying to move from the semiconductor to the metal, that is the built in potential was
that barrier in case of thermal equilibrium. Now, that built in potential will increase by
built in potential will not change, but effectively the electron will see phi b i plus
magnitude of V a; that is the applied bias. This is the resultant voltage that is observed by
the electron from when it is trying to move from the semiconductor to the metal.
So, the first part of this question was what is the potential drop across the semiconductor
and that is given by ∅bi - Va, because now this that this entire amount of band bending is
basically that corresponds to the total amount of voltage that drops across the
semiconductor. So, that is equal to phi b i plus magnitude of V a; because here V a is
negative. So, if you plug in those numbers you can find that potential drop across the
semiconductor is now 5.53 volt.
Now, the second is that what is the depletion width this again straight forward you just
plug in those numbers. The expression of the depletion width is x d equal to root over 2
epsilon s built in potential minus Va/qNA. So, you just plug in all the numbers given in
the problem, we will find the depletion width as 0.27 micrometer. Now, what is the
electric field at the middle semiconductor interface this is nothing, but your peak electric
field whose expression is given by qNAxd/𝜖 s plug in all the numbers again you will find
the electric field at the metal semiconductor interface.
Now, the fourth point was what was the junction what is a junction capacitance per unit
area in this under this reverse bias situation? The junction capacitance basically if you
recall the junction capacitance that constitutes of two capacitances; one is the depletion
919
capacitance, another one is the diffusion capacitance. Now, in case of reverse bias this
diffusion capacitance becomes negligible because of the absence of injected charge
carriers. So, only the depletion capacitance remains in case of the reverse bias. So,
basically I have to find out only the depletion capacitance. So, your junction capacitance
is nothing, but under the reverse bias case is nothing, but your depletion capacitance
which is given by 𝜖 s/xd this is the capacitance per unit area.
So, again plug in all the numbers depletion width already we have found here plug in all
the numbers you will get the junction capacitance at 39 nano farad per centimeter square.
Now, next come to the next problem which basically builds with a PN junction. Now, in
that course basically you have studied what is what is the, you have studied in details the
electrostatics, transport etcetera for an abrupt PN junction.
Now, here this problem discusses about the linearly graded PN junction. Now, what is
the linearly graded PN junction? In case of abrupt PN junction what you have consider
that the doping concentration on p-side and inside they remain constant uniform
throughout the structure. But, in case of linearly graded PN junction the doping profile
linearly varies with the distance from the interface and this expression of the doping
profile is given as ND- NA equal to ax, where a is nothing, but your proportionality
constant.
920
So, basically the here the doping spatial distribution of doping profile is a linear function
of the space is a linear function of the space. Now, here you have to find out the peak
electric field, built-in potential and depletion width at thermal equilibrium. So, basically
you have to do that entire analysis of electrostatics at thermal equilibrium in this spatial
kind of doping profile. So, some parameters are given like the intrinsic carrier
concentration is equal to n i the dielectric permittivity of the semiconductor epsilon s.
And some approximations you can use in your analysis, first is the fully depletion
approximation and the second one is the, you can consider that all the dopant atoms
inside the depletion width in inside the depletion region they are completely ionize. So,
you can consider that ND plus is nearly equal to N D and NA minus is nearly equal to NA.
Let us try to first solve this problem. So, the doping profile is given by N D minus N A is
equal to ax. So, the doping profile picture will look like this is the doping profile. This is
the linear doping profile so, with a. So, basically the slope of this line is equal to a and.
So, first we have to solve the Poisson’s equation.
So, between minus -xd/2 to xd/ 2 we are considering here that x d is the depletion width
and we are considering in symmetry that minus xd/is on the p side xd /2 on the inside and
the depletion layer is extended between these two regions between these two lines. So,
let us try to first solve the Poisson’s equation between these two points minus xd /2 and
xd /2. So, here Poisson’s equation is given by d electric field dx. So, let us write that d
921
electric field dx. This is equal to minus d 2 phi ∅dx2 square which is equal to your charge
per unit volume on the semiconductor by semiconductor permittivity.
Now, this rho x can be written as qND plus minus N A minus plus p- n, right. Now, here
the first assumption was the, you can consider it is a fully depleted region this region is a
fully depleted region. So, we can consider that p and n they can be consider as 0. The
next assumption you can consider is a fully ionization of the dopant atoms. So, we can
consider now ND plus is equal to ND and NA minus equal to N A. So, this can be further
reduced to q/∈s N D- N A.
Now, this is given by this doping profile N D- N A equal to ax. So, we can write it as q ax
by epsilon s. So, this is what given here the Poisson’s equation. So, basically now we
have to solve this Poisson’s equation. If you solve this Poisson’s equation so, first
integrate this one time, so, we will get the electric field expression this is a
straightforward integration this electric field is coming as a quadratic function of x along
with the integration constant.
Now, this integration constant can be found from the boundary condition the boundary
condition is that electric field that goes down to 0 beyond this the depletion width. So,
we can consider that x equal to plus or minus x d by 2, the electric field goes down to 0.
If you plug in this boundary condition here in this expression you will find the
proportionality constant C1 as minus qa xd2/8∈s epsilon s.
So, here just observe that here you get basically two boundary conditions, one is at plus
another one is at minus both will give you the same result because of the symmetry
because this is a this an even function because of that symmetry will get the same value
of C1. Now, you plug in this expression of C 1 into equation 2 you will find the total
electric field expression that is equal to minus d phi dx equal to this expression. So, this
is the electric field for a minus I mean this is a electric field inside the region minus x d
by 2 to xd /2 beyond that electric field will become 0.
So, from this electric field expression we can find out the maximum electric field that
maximum electric field will be at the interface or at the junction exactly at the junction.
So, maximum electric field will be when x equal to 0 just plug in that x equal to 0 here.
So, this term will be canceled. So, this expression will be minus qaxd2/8∈s. So, it is a
negative this is a negative field because the field that will be directed from n region to p
922
region. So, electric field here is a negative and the peak electric field hence also is a
towards the negative axis. So, you have to performed you can find the magnitude of the
peak electric field as this.
So, this is the diagram which basically shows the electric field profile this you can see a
quadratic profile these quadratic profile and this is the peak electric field and its
magnitude is epsilon m. So, minus that maximum electric field is this value, ok.
So, next you have to integrate that electric field in order to find the potential. So, what
this is this was our electric field. Electric field was that can be written as minus d phi dx,
let us write the equation first qa/28∈s by twice epsilon s x square - xd2/4, sorry. So, now,
you integrate this expression to find out the potential or potential ∅. So, if you integrate
this again a straightforward integration you will end up with qa by twice epsilon s. Here
this term will give x is q by 3 and this term will give - xd2/4into x. So, this is what comes
here along with an integration constant C 2.
So, now we have to find out this integration constant C 2 under the boundary condition
that the potential is 0 at x equal to - xd/2 that is a reference level. So, just look into the
diagram of the potential distribution along with the doping profile if we match the
potential distribution profile. So, it will look like this these are a non-linear expression
cubical expression you have. So, this is a potential profile. We can consider that this is
923
your reference level with respect to the reference level the potential at this point is a
constant potential here is consider the that is termed as a built in potential phi d i.
So, we can consider this as a reference level that is why we have kept phi at x equal to -
xd/2 equal to 0, that is the first boundary condition. That is the only boundary condition
we need to solve this expression because there is only a single integration constant. So, if
you plug in that number here you will find C2 is equal to qa xd3/24∈s. Now, if you now
just plug in that expression of C2 into this equation you will find the entire expression of
the potential.
So, again this is the potential profile valid within this range because beyond the depletion
range again this will for example, beyond x=-xd/2 it will be 0, that is the reference level
and beyond x equal to xd/2 it will be 5 Vbi that is a built in potential. So, within this range
this expression is valid. So, how we will you find built in potential you have to plug in
that x equal to xd/2. So, if you find the potential at x equal to xd/2 that will give you the
built in potential. So, just plug in that x equal xd/2 here, here you will end up with this
expression this expression built in potential equal to qa xd3/12∈s.
So, just observe that here we have find maximum electric field in terms of the depletion
width, here we have found built in potential in terms of the depletion width again. So,
now, the only thing is remaining to find the depletion width at this x equal to minus x d /2
and at x equal to xd/2 by magnitude this doping concentration will be equal.
924
And, that doping concentration is given by a into xd/2 because ND- N A is equal to ax.
So, you just plug in if x equal to ∓xd/2 x d by 2. So, the magnitude of the doping density
at x equal to xd/2or -xd/2that is at the ages of the depletion layer that value is given by
this a into x d by 2. So, at these two ages the doping densities are equal.
Now, we will use an approximate formula what we have used earlier in case of abrupt
junction that the built in potential is equal to kT/q in case of abrupt junction this formula
is you we have used this formula kT/ln N A N D by n i square. So, in this particular case
also we are using an approximate formula this not exact formula, but this an approximate
relation we are using here, built-in potential equal to kT by q ln instead of any N D now
we will use the doping density what we obtained at the two ages of the depletion width
and that is given by a xd/ 2 into (a x d /2/ ni )2.
So, this will be equal to (kT/q ln axd /2ni )2. So, just take this 2 out. So, now you are
built-in potential will be nearly equal to 2 kT/q ln axd /2ni. So, this is another relation.
So, now we have two relations between xd and built-in potential; one is this relation
which is an approximate relation and another one this relation this relation. So, basically
if you look into these two equations, this two are transcendental equation. So, analytical
it is not possible to solve these two equations. So, you have to sort for some numerical
analysis either by simulator or by some numerical technique you have to solve this built
in potential and depletion which separate together.
So, once you solve for the depletion width you plug in the depletion width number here
in order to find the maximum electric field. So, this is how this linearly graded a PN
junction electrostatics can be taken care.
925
Lecture – 55
Tutorials Session -3
Hello, I am Oppili Prasad serving as teaching assistant along with mister Prasenjith for
the Semiconductor Devices and Circuits course. So, in the earlier tutorial session, you
would have solved many problems from various topics from this course. So, in the
session, we will continue with a few more problems relating to topics like MOS CV
characteristics, MOSFET and also some other problem from circuits.
Let us move on to the first problem mode. So, this is a problem from the MOSFET topic,
the problem reads like this an n channel enhancement mode MOSFET has the following
device parameters which is basically oxide thickness is given tox and doping
concentration, the flat band voltage is minus 1.2 volts the aspect ratio is 10. And, the
channel length is given and you are asked to assume a constant mobility of 400
centimeter square per volt second and for this MOSFET device, you are asked to plot the
I D versus V DS characteristics which is nothing, but the output characteristics of the
MOSFET over the range of V DS values from 0 to 5 volts. And, you have given different
926
gate over dry voltage, which is nothing, but this V GS-VT values to be a 1 volt 2 volt and
3 volt and your ask to plot this ID versus V DS curve.
So, with regards to a problems in MOSFETs, there can be 2 interesting types of

problems. Firstly, where you are given device parameters like this and you are asked to
you know kind of plot the I D versus V DS curve and the transfer characteristics by you
know substituting the values in the current equation to get these characteristics. And, the
other kind of problem can be vary a given the transfer or output characteristics and there
asked to extract various parameters of MOSFET like mobility threshold voltage and
etcetera. So, before starting to you know exactly solve this problem that is refreshing you
know concepts for with regards to MOSFET characteristics.
So, you might know that there is something called transfer characteristics and there is
something called output characteristics for the MOSFET. So, this transfer characteristics
looks at the variation in the drain current with respect to the gate source voltage for
different VDS value, which is nothing, but the drain to source voltage.
So, MOSFET is off some values and you know this it is on after the threshold voltage
and this can be VDS1 and VDS2 and VDS3; so, VGS less than VTS, where MOSFET is off
and VGS greater than VT, which is a threshold voltage. So, a MOSFET is on and the
output characteristics looks something like this, where it is nothing, but the plot of drain
927
to which is the drain current versus drain to source voltage for different gate to source
voltage values for different VGS values.
So, there is something called this linear region, where the current looks like this and the
there is a point, where the current saturates and this is the ID sat value and there is a point
which is called the you know VDS sat, which is nothing, but the pinch of point after
which the MOSFET enters saturation and till that the MOSFET remains in the linear
region and this is the saturation region. So, you can you know kind of define the curves
like this, which is VDS when it is less than VGS-VT, the gate over dry voltage, the
MOSFET is said to be in linear region. And, VDS exactly equal to VGS minus VT is
nothing, but the pinch of point and when VDS greater than VGS-VT you say that the
MOSFET is in the saturation region.
And you can also recollect from your lecture slides that in the linear region the drain
current equation for a MOSFET looks something like this is nothing, but mu n Cox by 2
W over L into you have 2 term here VGS-VT into (VDS- VDS )2. So, for very low drain to
source values, you can see that the plot is almost a straight line and in the saturation
region, you can say that the current equation looks something like this I D equals mu n
Cox by 2 W over L into VGS-VTH the whole square.
So, this is for the linear region and this is for the saturation region and you know you can
also see that in the saturation region the drain current value is remains independent of the
drain to source voltage it is almost constant. So, this is for you know a MOSFETs ideal
MOSFETs, where there are no channel length modulation effects. So, when there is
channel length modulation effect present then, you can see if there is a slope present in
the output characteristics of the MOSFET.
So, in that case the saturation drain current value is also a function of the VDS value. So,
this is a fix summary of what we have learnt in the lecture slides regarding MOSFET
output characteristics and transfer characteristics. So now, just looking into the particular
problem, where you are asked to calculate the ID versus VDS characteristics for the given
device parameters. So, looking at the so, you need for this particular problem, you need
to find the you know current equations for various gate dry or the gate over dry voltages
in the linear region and also in the saturation region and you also have to find out what is
the pinch of point for each case.
928
So, use you can you can see that you are given the mobility value and you have given
you know the different values like W over L. So, what you need to kind of first do is find
out calculate 2 values, which are unknown to us which is basically the Cox and this VT
So, we set off with calculating what is the value for Cox which is nothing, but the unit
area oxide capacitance of the MOSFET device. So, you know that Cox is nothing, but 𝜖 ox
over tox. So, plugging in all the known values, you can find that Cox comes out to be is
8.63 into 10 power minus 8 farad per centimeter square.
Secondly, we you know calculate the threshold voltage for the MOSFET device from the
given parameters and you can recollect that this threshold voltage has you know a
different components like this, which is basically this flat band voltage 2 phi fp term and
this depletion charge term q depletion by Cox term. So, to calculate threshold voltage first
we try to find out what is this ∅fp which is nothing, but VT lan NAni. And, you are given
the thermal voltage to be 25.9 millivolts and you are given the intrinsic carrier
concentration from which you can find that ∅fp comes out to be 0.389 volts and in the
question you are also given that the flat band voltage is minus 1.2 volts.
And to calculate the threshold voltage the other value, you need as the input is this
depletion charge, which is nothing, but qNA XdT. So, this you are calculating at a point,
where the you know maximum space charge width is achieved. So, at maximum space
charge width, you have this relation relating to ∅fp and qNA. So, substituting values you
929
get X dT the maximum spare charge width to be a 0.142 microns and substituting values
further you get Q d the depletion charge value per centimeter coulombs per centimeter
square.
And plugging in all the known values you find the threshold voltage for the device to be
0.9 volts. So, we have calculated 2 unknown parameters, now which is basically the Cox
unit area oxide capacitance on the threshold voltage for the device.
So now, next we set off to you know kind of plot the main requirement for the question
which is nothing, but the output characteristics I D versus VDS in volts this kind of in
probably in milliamps. So, first we you know extract the needed equations for the a gate
overdrive voltage for VGS- VT equal to 1 volt.
930
So, for VGS- V

equal to 1 volt, you see that in linear region, you know you take the known linear region
equation by substituting, all the known values. Now, we have got Cox and VT and all the
known values plugging in back you get that the drain current takes saturation for VGS
minus VT equal to 1 volt which is nothing, but 0.173 into 2 VDS -VDS VDS2 in milliamps.
So, we can try to plot this curve here in the linear region. So, this curve is represented by
the equation 1, which you have just obtained and the pinch off point is nothing, but we
know that it is nothing, but VGS - VT that is when the MOSFET enters saturation and that
is given to be VDS sat and this is 1 volt for this curve. And, then we you know kind of
calculate the saturation current equation, the saturation current value by using the
saturation current equation and you know you find that the saturation current value turns
out to be 0.173 milliamps.
So, we plot that here so, from then on the MOSFET into the saturation. So, this is for
gate overdrive voltage given in question a. So, a subdivision a of the question and for
subdivision b of the problem again you take the other gate overdrive voltages and find
the linear MOSFET equation in the linear region and you know also find the pinch off
point and also the saturation current value. So, let us plot for the subdivision b here, sorry
if I am not drawing it to scale. So, here the pinch of pinch of point comes out to be 2
931
volts and this equation is given by linear region equations given by equation 4 and the
saturation current value is 0.692 milliamps.
And for gate over dry voltage being 3 volts the linear region equation takes a form like
this and here the pinch off point is VGS - VT which is nothing, but 3 volts and the
saturation current value is nothing, but 1.557 milliamps. So, this is the required output
characteristics for the MOSFET given in the question ok.
So, let us move on to the next problem. So, this is a problem from MOS cap a MOS
capacitor topic and the CV characteristics of a MOS capacitor and I guess this problem
was given in one of your assignments and we thought that we will just replicate this
problem here and you know discuss the solutions ok. So, that a question drains like this.
Consider a MOS capacitor with p type silicon substrate at t equals to 300 Kelvin dope to
doping density of 1016 per centimeter cube given, the oxide thickness to be 55 nanometer
and the area of the device, which is the total area A is also given in a 200 10 power
minus 3 10 centimeter square and you are asked to assume some values for thermal
voltage and the intrinsic carrier concentration and I asked to find out calculate 3 values
for this question, which is nothing the nothing, but the total oxide capacitance of the
device in picofarads, which is nothing, but Cox total the total not the unit area
capacitance, the total capacitance of the device.
932
And you are asked to calculate the maximum depletion width it can be denoted as a X dT
and the effective unit area capacitance at a point, where the maximum depletion width
achieved in the mass capacitor ok.
So, the solution goes like this. So, it is very straightforward that Cox is epsilon ox over
tox and this is the unit area capacitance, the total capacitance is very straightforward and
it is that nothing, but the product of unit area capacitance and the total area of the device.
So, when you plug in values, you can see that it is 126 pico farads and next we you know
a move on to calculate the maximum depletion width for the device. So, you know that
the depletion width for the mass MOS capacitor takes a relation like this which is
nothing, but root over 2 ∈s 𝜙s/ qNA this 𝜙s nothing, but the surface potential for the
device.
So, at a point when maximum depletion width achieved, you can you can reclaim from
your lecture slides that this 𝜙s value takes value of 2 phi𝜙fp. So, your maximum
depletion width takes some form like this four root of 4 epsilon is phi fp by qNA and we
know that phi fp is nothing, but VT lan NA by ni. So, plugging in known values, you can
substitute the values to find that the maximum depletion width is 0.3 microns.
933
And in the last part of the question you are asked to calculate the effective capacitance c
effective corresponding to when maximum depletion width is achieved. So, when you
recollect your MOS capacitor, you can see that it is nothing, but the serious combination
of 2 capacitances, which is nothing, but Cox here and C depletion. So, these 2 capacitors
come to gather in Cs ess in a MOS capacitor. So, already we know the unit area
capacitance Cox been calculated in the earlier section of this problem and we need to
calculate, what is this C depletion? C depletion is nothing, but the epsilon s corresponding
to when maximum depletion width is achieved epsilon s over XdT.
So, when you plug in values you can find that this is 34.53 into 10 power minus 9 farads
per centimeter square and when you have 2 capacitances in C ess the effective
capacitance comes over to be Cox into C depletion/ Cox+ C depletion. So, and you are asked to
find the effective capacitance in terms of unit area capacitance. So, you arrive at this
number for this problem.
934
So now, we will move on to the next problem, which is the aim of this problem is to give
insight as to what kind of impact the overlap capacitances in a MOSFET device has
when it is being used in an amplifier circuit. So, you are given a given that the common
source amplifier has an n type driver biased with gm value and a resistive load and your
given a various device parameters like channel width channel length and the unit area
oxide capacitance and also you given that for this amplifier, there is an input resistance
measuring R in equal to 1 mega ohms.
And you are asked to find out, what is a input cutoff frequency for this amplifier which is
the that is where the input pole lies, which is nothing, but this f L 1 value. And you are
also given that another common source amplifier has the same configurations as above,
but except that the MOSFET device in that amplifier has got an additional Lov, which is
nothing, but this L overlap measuring to be 2 microns.
935
So, you are asked to find out what is the cutoff frequency for this amplifier which is f L
2, this amplifier 2. And you are also calculate: what is the ratio of both the input cutoff
frequencies. Let us represent this problem in terms of a circuit diagram to understand
what we really need to calculate here?
So, this is nothing, but a common source topology with resistive load, this being the
source terminal drain terminal in the gate terminal, you are given that the input resistance
here is 1 mega Ohms and this is resistive load of 10 kilo Ohms, you are just shown a part
of the circuits, it is shown that the a trans conductance turns out to be 10 power minus 3
amperes per volt and they are given the aspect ratio and for this amplifier 1, you do not
have any L overlap Lov, which is nothing, but the overlap length in the device.
And you have another amplifier circuit with the same configurations with, but with an
additional overlap length of 2 microns. So, I would like to refresh this concept from the
circuits week of the lecture metal and you know, we have this oxide here and you have
semiconductor and you would have seen that in the course that there is a drained region
here and this is nothing, but the overlap region, which is nothing, but the gate here metal
gate here. Since nothing, but the gate to drain overlap in the device and you are given
that the overlap length comes out to be 2 microns.
So, this adds to additional capacitance of C overlap which is nothing, but a capacitance
coming directly between the gate terminal and the drain terminal of the device in this is
936
the source terminal. So, this directly this overlap length of 2 microns leads to additional
parasitic capacitance between directly between the gate and the drain terminal of the
device and you know that for this common source, a configuration this is nothing, but the
input terminal and you take the output here.
So, this overlap capacitance acts nothing, but like a miller component directly because
this is a component directly connected between the input and the output terminals. So,
you would like to know: what is the miller capacitance at the input terminal, due to this
Cov.
So, first let us analyze the first case, where there is no Lov present. So, in for this
amplifier this is let us call, this to be amplifier 1. Here you can represent the effective
network at the input terminal assume that this is the gate terminal ok. So, this is the gate.
terminal.
So, what is the effective network here? And here we have R in to be 1 mega ohms and
between the gate and ground terminal, you have Cg let us call that the effective
capacitance between the gate and the ground terminal as C effective and for this device
this is nothing, but the Cgs value. So, the C effective this is the effective capacitance of
the gate input terminal is nothing, but Cgs for this configuration and you have given R in
value and for a common source topology, you know that the gain is nothing, but minus
gm RL.
937
So, from the given trans conductance value and resistive load we can find that the gain
comes out to be minus 10 x n a inverting stage. So, the gain is negative now we move on
to calculate: what is the cutoff frequency for this input, cutoff frequency for this
amplifier. So, which is basically determined by this RC network at the input so, you also
know that the gate to source capacitance can be calculated like this is which is nothing,
but the total a Cox is a unit area capacitance in the total Cgs is nothing, but WL into Cox.
So, when you plug in values, you can find that this is almost 10 fem to farads at the input
terminal. So, the input cutoff frequency is nothing, but a 1/ 2𝜋RC at the input. So, where
R is the effective input impedance input resistance and C is nothing, but the Cgs value.
So, plugging in values you can see that for this amplifier with you know no overlap, you
find that the cutoff frequency comes around to be 15.9 megahertz.
And now let us next move on to analyze this case where Lov of 2 microns is present in
the device. So, again we you know kind of draw the equivalent RC network at the input
which is the gate terminal here, R in and it is 1 mega Ohms and here again we calculate,
what is C effective. But here you see that it is not just Cgs, but in addition you have this gate
to C overlap coming directly between the gate and drain terminals, which is a miller
component and for a miller component, you know that the can be replaced at the input
effectively by a by a capacitance like this and here this is the miller component, where
938
Cov into 1- C is the additional capacitance, which gets reflected at the input, which
means that the Cov here gets amplified as Cov into 1 minus A at the input.
So, the effective capacitance of the input is nothing, but Cgs+ Cov into 1 minus A with
where this is the miller component. So, you know: what is the gain which is again ten for
this and you know that and you have to find: what is Cov which is nothing, but only the
overlap length to be considered for calculating the Cov capacitance. So, it is Wo into Lov
into Cox, which is nothing, but 4 fem to farad and again Cgs is 10 femto farads here.
939
And so, what is the effective capacitance of the input? So, I have I think I have again. So,
we have representative with the terminology called C in 2 in for the previous thing it is
nothing, but C in 1. So, sorry that I have put it C effective here ok so, C in 2 is nothing,
but Cgs plus Cov into 1 minus A. So, when you plug in values, you can find that the total
capacitance at the input is nothing, but 54 femto farads. So, you see that due to this
miller effect a Cov of 4 femto farads gets amplified and reflects as this much of
capacitance at the input terminals.
So, the effective a total capacitances 54 femto farads for this amplifier configuration so,
the cutoff frequency is nothing, but again 1 by 2𝜋 RC, but we use R in to be 1 mega
ohms and C in to be the effective value calculated in this problem. So, we can find that
the frequency comes down and it is just 2.95 megahertz compared to 15.9 megahertz in
the previous problem. So, you see that due to this presence of overlap capacitance, you
can see almost a 5 times increase in the capacitance at the input terminal and. So, it
affects the operating frequency such that the operating frequency goes down by almost 5
times than the overlap capacitances percent.
Now, we will just move on to the next problem in this tutorial session, which is a simple
problem in the circuits topic of the course hopefully in course lectures, you would have
analyze different configurations like common source amplifier with resistive load and
common source amplifier with active loads etcetera. And, you would have studied, what
940
is the DC biasing and you know DC analysis of the circuit and also a small signal
analysis of the amplifier circuits, where you try to calculate the voltage gain of the circuit
and the input impedance output impedance of the circuit.
And here is another one such amplifier circuit, there your small signal input is given at
the source terminal and the small signal output is taken at the drain terminal of the a
device and you are given at the gate terminal constant DC voltage is given and you have
a resistive load RL connected to VDD. So, this is called as a common gate topology. So, in
this question you are asked to find out, what is derive the small signal voltage gain AV
which is nothing, but VO by V in for the amplifier and you are given that V bias and VDD
are the DC voltages aiding the basing of the MOSFET and RL is the load resistance and
gm is the trans conductance of the device.
And let r ds be the inherent output resistance for the device itself and you are asked to
find out what is the small signal gain of the amplifier. So, here what you would like do is
like we do something called as the small signal analysis, where you first ground all the
DC voltages, you make them 0. So, you see that your input is applied here and you see V
out here and for the small signal analysis, you replace the MOSFET device with it is
equivalent small signal model, which looks something like this for a MOSFET device.
So, you have the gate terminal here and you have the brain terminal here and you have
the source terminal here. So, this VGS is at the input side and you have a current source
941
measuring gm VGS between drain to source and the rds, which is the inherent output
resistance of the MOSFET device coming between drain to source.
So, this is the you know simple first order small signal equivalent model for a MOSFET
device and for this particular a common gate topology, you we try to do the small signal
analysis to find out what is the voltage gain? Which is nothing, but VO by V in. So, here
we represent all the voltage values for here, you set this to ground and VDD and you have
set RL, which comes between drain and ground. So, which is RL here and at the source
you apply a input small signal input here and you see that the small signal output is being
taken at the drain terminal.
So, this is the equivalent circuits small signal equivalent circuit for the goal amplifier
circuit. So, here what you would like to do is to derive for the gain expression, you do a
Kirchhoff’s current law at the V out mode, which is nothing, but the drain mode. So, you
see that there are 3 branches here. So, we write the KCL equation for this which is
nothing, but gm VGS plus v out minus V in V out -V out-V in by rds +V out and RL between V
out and ground.
So, V out minus 0 by RL so, the summation is 0 and you also see that source here and you
know vn is between source and ground and here its ground here and VGS comes like this.
So, what it effectively means is VGS is nothing, but minus V in. So, substituting values
and you know rearranging, you can you substitute that VGS is minus V in and you know
rearrange all the terms, you can find that V in into gm plus 1 by rds is V out into 1/ RL+
1/rds ok.
So now, AV this is nothing, but V out by V in is gm plus 1 by rds by 1 by RL plus 1 by

rds. So, this is the you know expression for small signal gain for this topology and you
can see that when rds is very high, then you can see that this all gain is almost gmRL and
you see that there is a a positive term in this expression. So, it means that this common
gate stage is a non inverting stage unlike common source amplifier, where you know it is
acts as an inverting amplifier.
942
And at this point, we thought we will just give you an exercise problem from the moss
capacitor topic, which you can get off look at, you can take it up as your, you know at
your leisure time to solve this. So, the aim of this problem is very simple it is nothing,
but you intend to calculate the doping concentration for a MOS structure, where you are
decide the threshold voltage is given. So, unlike in the pre and unlike in your assignment
questions, where you are being given the doping concentration, all the equations here
you given that you need a specific threshold voltage and you know you have to back
calculate, what is the required doping concentration to achieve that specific threshold
voltage.
And you are given other parameters relating to device such as t ox and what is the and
you are also given that there is some, you know interface trap charges and the density of
that is given and you are also given the 𝜙ms value. And, you are also given constants like
intrinsic carrier concentration and thermal voltages, you know that this threshold voltage,
it has got this main 3 components 2 𝜙f and the depletion charge component and flood
band voltage component.
And you can see then in each and every expression this doping concentration comes in.
So, which is for example, here V 𝜙fn is nothing, but VT lan Nd by ni. So, since this is a n
type substrate, you see that the doping concentration is Nd and the maximum space
charge which is nothing, but 2 𝜙s into 2 𝜙fn by q Nd and depletion charge is given by qN
943
d into XdT. So, we can see that it will almost all the expressions, this you know Nd term
appears here in there.
So, there are 2 approaches to kind of solve this problem, the first approach is to kind of
resort to trial and error method, where you assume some value for the doping
concentration. You assume that you take some value as this is the doping concentration
in per CC value and you plugging those values two kind of see that if you reach to the
specific threshold voltage. And if you do not, you refine this doping assumption and you
know kind of it towards the unit threshold voltage.
And the other you know way is to use a computer generated solution to solve for these
equations to directly find out what is the Nd value. So, you can kind of take this problem
as an exercise and when you solve this, you can find that the answer comes to be 2.5 into
10 power 14 into centimeter cube. So, I guess with this we have almost come to the end
of the tutorial session and also to end of all the video lectures for the semiconductor
devices and circuit course. All these your you know welcome to ask any questions and
post all your doubts in the forum and with regards to these tutorials. If you have any
specific questions or you need any further explanations for the tutorial problems, you can
always post it in the forum.
Thank you.
944
THIS BOOK
IS NOT FOR
SALE
NOR COMMERCIAL USE
(044) 2257 5905/08

nptel.ac.in
swayam.gov.in

108108112

Uploaded by

Document Information

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

108108112

Uploaded by

Copyright:

Available Formats

INDEX

S. No Topic Page No.

(Refer Slide Time: 00:14)

(Refer Slide Time: 03:40)

(Refer Slide Time: 08:22)

(Refer Slide Time: 09:09)

(Refer Slide Time: 10:41)

(Refer Slide Time: 12:26)

(Refer Slide Time: 17:07)

(Refer Slide Time: 24:17)

(Refer Slide Time: 27:05)

(Refer Slide Time: 33:16)

The Schrodinger’s equation is nothing but energy balance equation or an energy

(Refer Slide Time: 45:07)

(Refer Slide Time: 00:37)

(Refer Slide Time: 09:05)

the condition that the particle must exist somewhere, right.

At n equal to 2 I have sine of 2 pi by L into x, and at x equal to 0 at x equal to L the

(Refer Slide Time: 29:36)

(Refer Slide Time: 00:20)

(Refer Slide Time: 06:39)

(Refer Slide Time: 09:19)

Now, how many terms do we need to identify, we need identify A 1, B 1, A 2, B 2, A 3

(Refer Slide Time: 12:16)

And therefore, at x equal to L by 2, phi 3 takes a value of L by 2 e to the power minus

(Refer Slide Time: 20:54)

(Refer Slide Time: 23:02)

(Refer Slide Time: 25:15)

(Refer Slide Time: 29:57)

(Refer Slide Time: 36:41)

(Refer Slide Time: 00:18)

(Refer Slide Time: 08:49)

(Refer Slide Time: 13:29)

(Refer Slide Time: 18:03)

(Refer Slide Time: 20:24)

(Refer Slide Time: 25:26)

(Refer Slide Time: 00:14)

(Refer Slide Time: 10:16)

(Refer Slide Time: 11:55)

(Refer Slide Time: 14:05)

(Refer Slide Time: 20:12)

(Refer Slide Time: 24:52)

(Refer Slide Time: 27:16)

(Refer Slide Time: 29:55)

(Refer Slide Time: 30:43)

(Refer Slide Time: 00:14)

Now, the first definition is to do with something called as an intrinsic semiconductor. An

(Refer Slide Time: 13:26)

Therefore, in a crystal in a semiconductor crystal; if say this is the semiconductor crystal,

(Refer Slide Time: 00:14)

(Refer Slide Time: 25:02)

(Refer Slide Time: 27:37)

(Refer Slide Time: 00:14)

(Refer Slide Time: 12:30)

(Refer Slide Time: 17:44)

(Refer Slide Time: 20:19)

Now, is something called as a packing or the effectiveness of packing in a crystal. And in

(Refer Slide Time: 25:51)

(Refer Slide Time: 28:32)

Now, what is the frequency it is essentially 1 by the spatial period. So, if it is a 1 by a

(Refer Slide Time: 33:33)

We now proceed to the next important topic with regards to Semiconductor

(Refer Slide Time: 00:26)

So, if you are looking at 3D semiconductors ok.