Welcome. This is Fundamentals of Micro and Nanofabrication.

I am Sushobahan
Avasthi from IISC, Bangalore. This starts the module on chemical vapor
deposition, but before we get into the nitty-gritty of the technique itself, I
like to take a little detour and explain some of the definitions and
terminology that we will use in the rest of this module. So, there are two
types of depositions that you can theoretically do, one uses chemical
vapors and one uses physical vapors. In chemical vapor, the idea is that you do
not have the material that you are depositing completely fabricated yet. So,
what you have is just the precursor. So, supposing you are depositing silicon
dioxide, then you would flow a precursor of silicon and you would flow a
precursor of oxygen, and actually do the reaction at the wafer to form silicon
dioxide. In physical vapor deposition, you already have the silicon dioxide pre-
made either in a solid-state form or in a pellet form or in a powder form, and
then you figure out a way to transport this already formed silicon dioxide from
your source to your substrate. In general chemical vapor deposition gives
slightly better quality which is why it is used for semiconductor deposition; by
physical vapor deposition tends to be cheaper simpler and is often used for
metal and dielectric depositions, a non-critical deposition if I may say so.
So, this is the module that we are going to first take which is chemical vapor
deposition, and after CVD we will then discuss physical vapor deposition. So,
before that let us get some other terminology right. So, there is a concept of
uniformity. So, you may have a certain notion of what uniformity means well let
us standardize the definition a little bit more carefully. So, in deposition
uniformity refers to the fact that how does your thickness change over your
wafer. This is important because after deposition you often have an etching
step. So, here is an example. So, you have taken silicon and you
deposited silicon dioxide was some thickness. But because of some problem,
your silicon dioxide was not uniform, so it was thinner on the left-hand side,
thicker on the right-hand side. And after that, you did some etching, so that
etching recipe was probably optimized by somebody assuming that the silicon
dioxide wafer, silicon dioxide thickness is uniform. So, they gave you a recipe
that would work just right, if the silicon dioxide thickness was also just right.
However, in some places the thickness is lower in those cases you would over
etch the sample. So, the opening would be larger than expected. And in some
cases, you would under etch the sample because the in which case you would
do not have the opening at all because the silicon dioxide was thicker. So, these
sorts of problems can only be avoided by having the silicon dioxide uniformity
to be very precise. So, you should aim to have a uniform deposition
everywhere. The second concept is of conformality. So, conformality means that
is you follow the underlying topology, so best explained by a figure again. So,
here is silicon supposing you have silicon dioxide on top and that silicon dioxide
is patterned, and now you are going to deposit metal on top. So, now, that
metal can have a conformal coverage. And conformal coverage it basically
maintains the thickness no matter what the underlying topology is. So, the
thickness is consistent across this whole topology. Practically you may not get
this. So, let me grab my laser pointer, practically you may not get this. You may
see that over sharp corners that a position is a little more thin, and where the
position is a little more thin, then the thickness is not exactly the same all the
way through it is thinner here than it is there. So, this is what is called a good or
a perfect conformal coverage; this should be called a suboptimal or poor
conformal coverage. Conformality can sometimes be important, but it really
depends upon your application, sometimes it is not, but it is a good idea to
always know in advance whether the process that you are using will lead to
conformal deposition or not. A related concept is one of step coverage. So, often
in papers and processings, you will hear this word step coverage. What it
basically means is whether the film deposition is conformal or not. So, again
here is your wafer surface, and you have a step because you have deposited
something and then patterned it. And then the question is that this black
layer that you are depositing on top is it following this step exactly or not? So, if
it is following this step exactly, then you have a good step coverage; if it is not
exactly following this step and you have a thinner deposition here, then it is
poor or bad step coverage. A little bit of introduction into gas or fluid flow would
also be important to understand chemical vapor deposition a little. So,
whenever a fluid is traveling in a constrained medium say a tube, a pipe or a
reactor, it tends to follow some it tends to follow the laws of fluid dynamics.
Now, here the fluid can be the liquid or gas, of course, in chemical vapor
deposition, we are primarily looking at gas. The actual physics of fluid flow is
fairly complicated. So, in order to simplify things, we typically look at three
classes of flow. So, one is called the molecular flow this is what happens when
the particles or molecules are so rare that they do not miss collide with each
other. So, they are not they do not know of each other s presence. In that case,
the only thing that the molecule or particle interacts with is the physical
consider the physical containment container itself, so if that can be a pipe, a
wall or a chambers etcetera. In these cases, the flow is actually not proportional
to the pressure drop across this pipe. So, simply because we have a pressure
gradient does not necessarily mean the flow will scale with that pressure. This is
the case that happens in vacuum chambers so right. So, your vacuum chamber
and very low pressure, then you are in a molecular flow regime. The more
familiar to you may be the laminar flow regime which is often discussed in
introductory physics courses, where you must have seen Bernoulli s principle,
you must have heard of Navier-Stokes equations etcetera. So, they all tend to
define the laminar flow. So, in this case, the particles collide with each other,
but the flow is streamlined. So, the particles tend to maintain these streams
right, so that is what the laminar flow means. So, this particle will flow sort of
parallel to other particles and along the shape of the tube. This occurs at lower
Reynolds numbers and the flow is proportional to the pressure. So, depending
on the amount, the amount of mass transfer from left to right will depend upon
how much pressure you apply from left versus right. And, the third and the most
complicated case is actually the turbulent flow where the flow velocity becomes
so high that the particles collide with each other so much that they are not
streamlined anymore. The physics of this is fairly complicated it typically occurs
at high Reynolds numbers. These two regimes are what are called viscous flow
regimes, and they are typically present in a CVD reactor. By and large in
practical CVD reactors tend to maintain laminar flow, why, it will be obvious in a
minute. So, which regime we are in would depend upon the pressure of the tube
at the amount of there the pressure which tells you the density of the gas, it
depends upon the tube diameter, the reactor diameter, the pressure you have
across it, the flow rate, viscosity, density. Cumulatively all of this is captured in
two numbers one is called the Knudsen number and one is called the Reynolds
number. And, if depending upon what those numbers are, you are either in the
viscous flow regime which is laminar or turbulent, or you are in the molecular
flow regime or you are somewhere in the middle. So, for CVD you tend to be in
this region above this red curve in the viscous flow regime. In order to maintain
laminar flow, the Reynolds number must be 2000 which leads to the
question what is the Reynolds number. We will not query on this, but just for
your information, Reynolds number is simply a unitless quantity which is given
here which basically tells you how fast the fluid is flowing through your tube
right. And if it is flowing too fast, then flow becomes turbulent which is bad.
Typically, you get a laminar flow for Reynolds numbers less than a 1000; and
you get turbulence for Reynolds number above 2000 or so. So, here are some
numbers for hydrogen at various temperatures, one thing to notice is that for a
very large range of temperatures, the viscosity, and the density do not change
much. So, that is good news, that means, is that you design a CVD reactor that
is a laminar 700 degree C, there is a good chance that it is also a laminar
around 1200 degree C. So, the variation of temperature does not change the
fluid dynamics of too much. Now, the reason the CVD regime tries to remain in
the laminar flow is highlighted in the slide. So, if you are in the molecular flow
regime, typically your density is very low, your pressure drop and flow rates are
very small which means your growth rates are very small, which is impractical
and you do not want to do that. However, if you are at the turbulent flow
regime, even though you have a lot of pressure drop, you have a lot of flow
rate, you have a lot of material going through, the growth is very unpredictable.
It is very it is too unrepeatable. And if it is too unrepeatable, it is not a good
microfabrication process. So, by and large, most CVD reactors try to stick to the
laminar flow regime where the growth is both predictable and reasonably
high. Now, before we go further let us take a little take a look at how laminar
flow happens inside a pipe. So, if you have some liquid flowing through a pipe,
you must have some pressure difference. So, you must have a region of high
pressure and a region of low pressure, and this is creating the gradient that is
pushing the fluid from left to right. Now, as this fluid goes through this pipe, and
if this is a viscous liquid, then it would have a certain velocity profile. Under no-
slip conditions, you assume that the velocity at the interface is essentially 0,
and then that velocity profile increases. And, then it again falls as it reaches the
other interface. So, this is a imagine you looking at the cross-section of a tube
with a diameter of a radius R right. So, the velocity is lower at the interface with
the tube, and the velocity is highest somewhere in the middle. And, the reason
for this velocity profile is because this liquid is viscous. So, it has some friction,
it interacts with the other particles or other molecules, and hence maintains this
profile. Now, the viscous flow is given by this equation, the force due to
pressure is cross-section by this equation. And, under equilibrium, they are both
equal which gives you the profile of this velocity as a function of its radial
position, as a function of the radial the polar coordinate. Now, Poiseuille law
basically says that the amount of flow that you are getting is nothing but
velocity multiplied by this area. Now, since this velocity is non-uniform, you
must do some calculus to actually calculate the integral of this flow as a
function of radius and pressure. And, if you do all this calculation which is
simple calculus, you actually see that the total flow in this pipe is equal to the
pressure drop multiplied by a constant. And this constant is nothing but it
consists of geometric factors and viscous force. This equation is very similar to
the equation that you see of current flowing in a resistor which is that the flow
current I must be equal to conductance which is 1 over R, where R is resistance
into pressure drop here the pressure drop is just analogous to voltage drop
right. So, in some sense this is nothing very brilliant this is simply an ohms law,
but for gas flow. So, now, what you are looking at is instead of current, you are
looking at mass flow; instead of voltage, you are looking at pressure drop, and
this value represents conductance which is not electrical conductance, but
conductance of fluid flow. So, this is sort of an important insight because what
this allows you to do is to understand that when you are doing design of a CVD
reactor analyzing, the working of a CVD reactor, it can be thought of as a simple
electrical network problem where every component has a certain conductance,
has a certain has a certain conductance, and you can think of mass flow
controllers as current sources, you can think of pumps as grounds etcetera,
etcetera. So, the design of any vacuum system not just CVD equipment, but any
vacuum system can often be just thought of a series of conductances, either a
parallel or sorry a network of conductances either in series or parallel. And the
trick to understanding how much gas would flow would do, it would be to know
what is the pressure drop across the system and what is the effective
conductance. And, the effective conductance follows the same rules that you
know from series and parallel combinations of electrical resistances. So, in
series, the conductances are inversely added; in parallel, the conductances are
simply added. Now, in a real system, you do not necessarily build a system that
is conductance limited. Let me explain by giving it a practical example. So,
supposing you took if you if you had take your house right, and your house you
have a fan. Now, in order to control the amount of current that flows through
that fan, you do not calculate effective resistance and then use a wire of
precisely that resistance right that is not what you do. And the reason you do
not do that is because it is too unpredictable, it takes too much work, much
simpler is to connect the fan through a regulator. And this regulator has some
knobs and depending upon what is the setting of this knob you actually control
the current. So, you do not give a precise value of conductance of this wire, you
just put a wire with a very large conductance, and in series add a regulator
which can tune the conductance. In practical vacuum systems, you do the
same thing. So, you do not plan the conductance of your whole system, you just
apply very thick pipes which can conduct a lot of fluid, but somewhere in the
middle, you put a regulator. Now, that regulator for gas flow can either be a
pressure regulator, it can be a mass flow controller. And this is what actually
controls the flow of the gas through this system. And, the other hand you have
a pump, and pump can be thought of as nothing but something that
basically creates zero pressure which is equivalent to a ground right. So, in this
manner, you can actually analyze any complicated reactor to actually
understand how the gas flow works. Of course, the reality is a little more
complicated, CVD design is a fairly challenging problem, and there are a lot of
second-order effects to consider, a lot of computational fluid dynamics used. We
will not go into any of those details because that is far too advanced for this
course. I will just leave you with this generic idea that I introduced which is that
when you are analyzing vacuum system, when you are analyzing fluid flow, it is
often useful to think of pumps as essentially power supplies things that create
voltage gradient, things that create voltage differences. So, tanks and pumps
cylinders sorry gas cylinders and pumps are nothing but something that creates
high voltage and something that creates low voltage. The conductance of wires
can often be thought of as, the conductance of the tubes can often to be
thought of as conductance of wires in an electrical circuit. Mass flow
controllers, pressure regulators can often be thought of as a simple ways of
regulating the current in the circuit. And just like there is Ohms law for electrical
circuits, there is similar law for gas flow in a fluid medium when it is under
laminar flow regime. So, that is enough for this lecture in the next lecture we
get into more details of how CVD works.