Semiconductor Optoelectronics

Prof. M. R. Shenoy
Department of Physics
Indian Institute of Technology, Delhi

Lecture - 16
Fabrication of Heterostructure Devices

This is the last lecture of part one, where we were recalling or reviewing the basic
semiconductor device physics. The objective of this talk is primarily to indicate to you what
are the various methods which are used in the fabrication of devices. It is impossible to
discuss the details of fabrication in one lecture, because there are several intricacies.
However, it is just to indicate what methods are used and those of you who are interested you
can look at the references.

(Refer Slide Time: 01:13)

For beginners the good references will be 1) Sze, Physics and technology of semiconductor
devices for beginners, 2) Semiconductor optoelectronics by Jaspreet Singh, and you can also
see 3) Semiconductor optoelectronics by Pallab Bhattacharya. These are some of the basic
references. However, there are plenty of specialist books and articles which are available on
these growth processes and growth technology.

We would simply introduce you to the various growth processes. So if I take a typical
semiconductor optoelectronic device, it has a substrate, a lower contact electrode which is a
metal, and an upper contact electrode which is also a metal. What I am showing is a
longitudinal cross section of a heterostructure device. If you see a typical heterostructure
device, then the substrate would occupy most of the device, because 60 to 100 micron is the
substrate thickness.

Over that you have layered structures, 1, 2, 3, and there may be many more layers, and of
course on top and bottom the contact electrodes. For example, in a double heterostructure
LED, I start with an n plus substrate, on which you deposit n aluminum gallium arsenide,
gallium arsenide, p aluminum gallium arsenide, p plus gallium arsenide. I’m just taking a
simplest structure. On top there is a metal contact.

So this is a simple double hetero structure LED or a laser diode. As you know now, why we
need n plus and metal, p plus and metal, and then we have contact between n and n plus, and
between p and p plus? Because these are all ohmic contacts. So, this would be a typical
structure. Therefore what do we have? We have a substrate which is a bulk crystal, and
followed by on top there are layers. These layers could be typically of thickness anywhere
from 0.1 to few microns, may be 2 3 4 microns.

So the thickness of these layers is quite small compared to the thickness of the substrate.
Therefore the substrate is grown by different techniques compared to the top layers here.
These are grown by bulk crystal growth techniques. And the top layers are grown by epitaxial
techniques, because the thickness of the layers which are grown here are very small
compared to the bulk crystal, and these epitaxial techniques are generally used when the
thickness is very small, anywhere from tens of angstroms to a few microns. Only then these
techniques would work, otherwise you have to go for bulk crystal growth techniques. So I
will briefly discuss both bulk crystal growth techniques and epitaxial growth techniques, but
just to give you an idea. As I indicated, for details you may have to see several texts, or
specialist paper depending on the detail that you are looking for.
(Refer Slide Time: 07:21)

If you take bulk crystal growth here, that is essentially from our point of view to grow the
substrate, there are two important techniques which are used. One is called the furnace
method or Bridgman method, there are several variations of this method; and other one is
very well known, Czochralski.

There are others and variations of these, but these two are basic processes to have bulk crystal
growth. To have epitaxial layers there are three important techniques, as I had indicated once
earlier, liquid phase epitaxy LPE, vapour phase epitaxy VPE, and molecular beam epitaxy
MBE. Each one has their advantages and features. We will briefly discuss all these three
techniques.
(Refer Slide Time: 09:47)

Let me start first with Czochralski method. In this method, you have a chamber, a lead nitride
crucible. There are different structures, different variations. Sometimes this process is done at
atmospheric pressure, or sometimes it could be done at high pressure, therefore in enclosed or
sealed volumes at controlled pressure. What is done is, one begins with a seed crystal, this is
a molten a melt. Melt means molten starting solution. Here, for example, if you are growing a
single crystal gallium arsenide, or let us say silicon, then you start with the molten silicon
here, and then there is a single crystal. The temperature of this melt is just above the melting
temperature.

So the melt here is just above melting point, and this is a crucible, usually lead nitrate
crucible, PbN, and this is heated. What I am showing now from the sides is the heating coils,
and there is a crystal, which is a seed crystal. As we draw all the parts, the setup will become
clear. The seed crystal is slowly pulled up. The melt has a temperature just above the melting
point. As you slowly pull up, the temperature here is much less than the melting point.
Therefore, the melt crystallizes on the seed, and as you pull up, the crystal keeps on growing
on the seed.

The melt crystallizes on the seed with exactly same lattice structure, with the same directions,
and therefore you have a single crystal. If you control the pulling speed then this diameter can
be varied, so the size of the single crystal could be varied. It is simple, pulling through this,
but the environment is controlled, it is not in open. Sometimes this chamber may have
atmosphere anywhere from 1 to 20 atmosphere, because there are several issues involved in
terms of the vapour pressure of the starting material.

This is called Czochralski method to grow single crystals from a melt. You can grow for
example, silicon crystals here with diameter as large as 30 centimeters. This is one of the
techniques to grow boules or ingots. Like a huge cylinder, and then you cut these into slices
to make wafers out of this. Normally it is called dicing, to make the substrates. So the silicon
wafers that you see are obtained by dicing these boules which are formed by the Czochrolski
technique.

Let me now discuss the Bridgman technique, and then you see the advantage that using
Czochrolski technique you can grow very thick, very large boules, or large single crystals,
whereas the Bridgman technique or furnace technique is used to grow smaller crystals.

(Refer Slide Time: 17:22)

As the name indicates, it has a furnace. So let me now discuss the Bridgman technique here.
In this technique, this is the furnace, usually these are graphite boats, and this is quartz
cylinder, and there are heating element, or heating coil. What is done is, at one end of the boat
a single crystal is placed, and it is filled with the melt. The melt comprises of, for example, if
it is silicon, it is polycrystalline silicon. How do we get polycrystalline silicon?
Polycrystalline silicon is obtained by chemical methods.
(Refer Slide Time: 19:38)

Let me show you an equation here. How to get polycrystalline silicon? You start with
chemical reactions in a furnace, silicon carbide and SiO 2 are the starting point, and in a
furnace at high temperatures. CO is the gas which goes out, and what you get is metallurgic
grade of silicon. SiO2, silica is plenty in sand. So this is the first step of the crude process to
get metallurgic grade of silicon. Metallurgic grade is approximately 98 percent purity.

The metallurgic grade silicon then interacts with the HCl here and you get trichlorosilane,
SiHCl3, and hydrogen gas. This is reduced further in a hydrogen environment, and reduction
in hydrogen environment gives you silicon plus 3HCl, the HCl goes as a gas. This is the
chemical reaction of heating and reduction. So what you get is high purity silicon which is
electronic grade silicon. High purity means impurities are 1 part per billion or less. So 10 -9 or
10-10 is the impurity concentration. This is electronic grade silicon.

So you can get polycrystalline silicon by the above technique. Both here and Czochrolski
technique, the melt comprises of chemically obtained silicon and then you have to grow
single crystal. Growing single crystal means what? Single crystal does not mean growing 1
cubic lattice or 1 cell. Single crystal means that all over that material you have a periodic
structure of silicon. So single crystal means everywhere there is same crystalline structure.
Polycrystalline means you have crystalline structure, but there are domains.

(Refer Slide Time: 22:32)

So, this is polycrystalline because crystalline structure is not over the entire structure. Single
crystal means over the entire material or entire structure or entire piece, you have 1 crystal
setup, that is the lattice structure is the same everywhere. So that is obtained by one, the
Czochrolski technique which I discussed, and the second one is the furnace method. In the
furnace method that I started discussing, this boat is pulled. I’m only giving you an idea, the
techniques are quite involved. The boat is pulled through a temperature gradient towards
cooling. So this end is cooled end and this here is hot end.

In fact everywhere this is maintained at a high temperature but on one side it’s a lower
temperature. Cool does not mean it directly goes into atmospheric temperature. It is just at a
lower temperature. So the boat is pulled through this. What is the idea? Idea is as this passes
to temperature below the melting point, it starts crystallizing. Because there is a single crystal
seed present, the solidification takes crystalline form. Therefore if you pull this through, the
entire thing forms a single crystal, and this is called the furnace method or Bridgman method.

As I said, for details please see references and specialist articles on this. So from these single
crystals, we get the substrates. Substrates are usually diced wafers of around 100 micron
thick. It depends on the material also; for some materials, thicker substrates are used. In some
materials the substrate is slightly thinner, so typically 60 to 100 micron.

(Refer Slide Time: 25:50)

So the above two methods that I’ve described is to obtain the substrate. If you have to have
doped substrate, in the melt before the single crystal is formed, required dopants are included
in the starting melt. For example I wrote n plus substrate. How do we get n plus substrate? If
you add arsenic in silicon melt, then you will have n plus substrate. So the dopants have to be
added in the melt before the single crystal is formed if you want a doped substrate.

Now we want to have these epitaxial methods to grow layers on top. Epitaxy is apparently a
combination of epi plus taxis. epi means upon or on top, this is a Greek word. And taxis is
ordered arrangement. So epitaxial growth means ordered growth on top, i.e., ordered
arrangement of atoms to forms layers. From epitaxy, comes epitaxial techniques. I have
indicated the three techniques: 1) liquid phase epitaxy, 2) vapor phase epitaxy, also called
CVD, chemical vapor deposition, and if the chemicals involve organometallic compounds,
then MOCVD. In fact most of the gallium arsenide, aluminum gallium arsenide are grown by
MOCVD. MOCVD stands for is metal organic chemical vapor deposition.

What do I mean by Organometallic? I will give you an example. For example,

In the remaining part of this talk, let me discuss briefly the three techniques which are widely
used for epitaxial layers. LPE this is the first technique, liquid phase epitaxy. As the name
indicates, the starting material is a liquid. This is done in a chamber you have. Let me draw
this. You will see nice figures in books and you will also see in internet there are several
articles and which show videos. I am drawing it because as you see you will understand what
it is.

It’s a graphite container through which there is slot. Also on top, these are cylindrical boxes,
like pill box. Means you can pour some liquid into this. These cylinders go down into this,
and on a slider. There is a slider on which you place the substrate, let’s say gallium arsenide.

I did not describe how to get gallium arsenide single crystals. I talked about silicon because
that was the easiest. Gallium arsenide is more complicated, it actually has to be done in a
sealed closed furnace because of the vapor pressure considerations, but anyhow let me not go
into that right now. In the box you put the melts, gallium arsenide and aluminum gallium
arsenide. Now what is this gallium arsenide melt? Gallium arsenide melt is super saturated
solution of arsenide in gallium.

Gallium is the solvent, and arsenide is dissolved in gallium. Why super saturated? You can
imagine it as a basic crystallographic process: you remember even in the schools we have
making sugar crystals from sugar solution? You have super saturated sugar solution and you
cool it, and when you cool, sugar crystals are formed because it was super saturated at a
slightly elevated temperature. When you cool it then it cannot hold anymore of the dissolved
sugar and so the sugar crystallizes.

It is the same process which is use in this, so you have super saturated melt and this is the
gallium arsenide substrate. You slide it so that when it comes exactly under this pill box here,
it just sets there. All of these are at some temperature, let us say approximately around 800
degree centigrade. The melting point of gallium arsenide is actually very high. It is I think
1237 or 1238 degree centigrade. However, when it is in a solution, it is in the liquid form,
because a solution has a lower melting point compared to the pure solvent.

Now, this is approximately around 800. When this gallium arsenide is sitting under the pill
box, please see this whole thing is in a oven. When the temperature is reduced, what happens
is crystalline gallium arsenide deposits on this. It is the same process of crystallization of
sugar, but now there is a substrate gallium arsenide on which there are gallium arsenide layer
which gets deposited. The time for which the temperature has been lowered determines the
thickness of the layer that is grown on the gallium arsenide under this.

Next if I want to grow aluminum gallium arsenide, the slider is pushed further to the next slot
under this pill box. There is aluminum gallium arsenide melt and you again lower the
temperature, and aluminum gallium arsenide will get deposited. So you have gallium arsenide
on which aluminum gallium arsenide is deposited. If you need again gallium arsenide to be
deposited or vice-versa, to make double heterostructures, you can have more chambers here,
so there can be more chambers to deposit layers after layers. What is the simplicity that you
see here? The arrangement is very, very simple. Only a simple graphite container with a slider
and pill boxes. You put the melts, and whole thing is in a furnace.

I will next describe you the other two, and you will see the complications in these. This LPE
is simplest, and even today they are used. When you need to deposit layers of thickness
greater than or of the order of 1000 or 100 nanometer, LPE can be used. Why greater than
this? Because the control on the thickness is not very good, because you have to slide it to the
next point. As the control is not very good, you cannot grow or have control of 5 or 10
Angstroms. So you have the other two techniques. So if you do not need abrupt junctions
with precise control, then LPE is the best technique. It is simplest, most cost effective and
whenever you need to grow thicker layers LPE is the best technique. Every device does not
require such abrupt precise junctions to make quantum well structures. You cannot make
quantum well structures with LPE because you do not have so much of control, but for many
applications you do not need that, you need to use thicker layers, and LPE is still a
commercial technique which is used to grow optoelectronic devices.

Let me go over to vapor phase epitaxy. I keep repeating please refer to literature for more
details. I have given you an idea what the technique is, and for details you can always go
through other material. So let me describe vapor phase epitaxy; this is an interesting
technique.

(Refer Slide Time: 39:49)

Let me give you some equations that describe deposition of gallium arsenide, aluminum
gallium arsenide layers using MOCVD reaction.
(Refer Slide Time: 40:03)

You have a reaction chamber in which there are substrate holders. Substrate holders means on
top of these there are substrates sitting, 1 2. These are substrates, like I showed you in the
LPE, so say a gallium arsenide substrate will sit first. There is an exhaust here, a mixture of
required gases enters. What are these gases I will show you in a minute.

So a mixture of gas enters from here and reaction takes place on the surface of the substrate
inside the reaction chamber. This is chemical reaction which takes place on the surface of the
substrate, and layers get deposited. The remaining gas is exhausted. Now, there are different
techniques used, like there could be lamp heating, so these are heating lamps, or there could
be RF heating. It is not coil now, it is lamp array. So there are different heating techniques.

Now we have to see what is this mixture? This is actually hydrogen gas H 2 which is bubbling
and through it the vapours are carried. Everywhere there are mass flow controllers which are
shown as taps, say T1 T2 T3 T4. So we have hydrogen gas from here and arsine from here. It’s
a highly poisonous gas. This is trimethylgallium TMG, trimethylaluminum TMA, and arsine
here. Hydrogen is the carrier gas. This entire mixture is entering the vapor chamber, or the
reaction chamber. So see the reaction. Trimethylgallium plus arsine here gives you gallium
arsenide plus methane gas. Trimethylgallium plus trimethylaluminum plus arsine gives you
aluminum gallium arsenide plus methane.

Chemical reaction is taking place inside this chamber, so the CVD process is taking place
here. It is a vapor phase epitaxy because the trimethylgallium and trimethylaluminum are
carried by hydrogen gas in the form of vapors; hydrogen is bubbling through this. The mass
flow controllers control the rate at which the vapors have to go. Here is arsine coming and if
there is any dopant if you want to add, you can add the dopant gas. The reaction is taking
place on the surface of the substrate, and the gallium arsenide which is formed here is directly
deposited on the surface.

Say in the next stage if you do not want gallium arsenide to be deposited, you simply close
the corresponding tap and then you may be only bubbling the aluminum gallium arsenide.
Similarly you can do vice-versa. So you can control by this technique the layers very
precisely, the control is very good because the mass flow controllers can adjust, and the
deposition rate is determined by the inflow of the reactant mixture. Therefore the control is
very precise both in MOVCD and the normal VPE. The reaction for CVD is also shown here.
It is called CVD because now there is no metal organic compound. Silicon tetrachloride plus
hydrogen gas gives you silicon plus HCl. Basically they are all VPE. MOCVD is only when
the reactants have metal organic chemical vapors. I go to the last technique that is MBE.

(Refer Slide Time: 49:20)

MBE, molecular beam epitaxy, is widely used to make quantum well structures. The setup is
quite involved but the control is extremely precise; you can deposit monolayers of gallium
arsenide or aluminum gallium arsenide. There are atomic beams which are coming to form
molecules directly on top of the substrate. You have effusion chambers, or effusion furnaces,
through atomic beams come out. There are shutters for each chamber. There is a main shutter
and in front of this here is the substrate holder. There are quite involved setups with gear
wheels and all those things. I am not showing all those arrangements, but I have a chamber,
and there is a RHEED gun here. Have you heard of RHEED? Reflection high energy electron
diffraction gun, RHEED gun. So, high energy electrons or electron beam is incident on this
here and it is diffracted here. There is a fluorescent screen, and there is a rotating mechanism.
This is the substrate, and this is loading chamber, the double barrier ultra-high vacuum lock.
Now let me explain and then everything will become clear.

These are sources, say gallium, aluminum, arsenic and some dopant, if you want to put some
dopant. These are atomic sources, it is atom by atom which come, so atomic beam is incident
on the substrate. The substrate is kept heated at a certain temperature. So atoms are arriving
on the surface, and if this is gallium arsenide, the atoms arrange themselves so that it fits to
the gallium arsenide lattice. So at a time it is growing one layer, and that is why you call it
monolayer. You are allowing atomic gallium to come from here, and arsenic. If you want
aluminum too, you can open its shutter. There is a main shutter. If you do not want any
reaction to take place or you are doing some setting, then you close the main shutter.

This is at ultra high vacuum, of 10 -11 torr. It is ultra high vacuum chamber in which we have a
substrate that is mounted. This is the substrate holder which you can rotate, so the substrate
keeps on rotating, and atomic beams come from here, like gallium, aluminum, arsenic all
these. In the effusion chamber, you have a crucible, usually boron nitride crucibles in which
you have aluminum. You have placed pure metallic aluminum and it is heated and the atomic
beam is coming out of these effusion guns or chambers or effusion furnace.

So independent beams are coming, and the atoms rearrange themselves in the lattice on the
surface of the substrate. So you can imagine the control that you have. You are allowing the
rate at which atoms are incident; the rate at which atoms are incident here is controlled by the
temperature. So in a controlled rate of gallium, aluminum, arsenide, you can change the ratio
in Ga1–xAlxAs, say if you are growing the ternary compound aluminum gallium arsenide.

You can control this x so that required composition of aluminum gallium arsenide can be
obtained, layer by layer. So the control that you have is 1 atomic monolayer which means the
control is of the order of 3 to 5 Angstrom. The layer thicknesses can be controlled correct to 1
monolayer which means you can make abrupt junctions. If you do not want gallium
aluminum arsenide anymore you just block aluminum, and the next layer will be pure gallium
arsenide.

One doesn’t call MBE as atomic beam epitaxy because the layers which are grown are
molecules. It is a very interesting technique and most of the quantum well structures are
grown by MBE technique. I will stop here. I hope I have given you an idea about the different
techniques which are used; plenty of details are required to understand each one of them
further. So we will stop here and go over to part two.