VLSI Technology

Dr. Nandita Dasgupta

Department of Electrical Engineering
Indian Institute of Technology, Madras

Lecture - 3
MOSFET Fabrication for IC

In the two previous lectures, we have discussed the basic processing steps to realize a
bipolar junction transistor and a MOSFET. So, you have seen so many steps go into that -
oxidation, photolithography, diffusion, epitaxy, etc, etc. But, the starting point is always
the same. That is, in order to fabricate any device, you must first have a single crystal
silicon wafer, right? So, that is the first unit step. How does one realize a single crystal
silicon? Not only that, if you look at it more closely, for bipolar junction transistor we
mentioned that the single crystal wafer is 111  oriented. On the other hand, if you are
going for a MOSFET fabrication, universally the starting wafer will be 100  oriented.

Why is this difference? Is it just an arbitrary choice? No, let me assure that this is not an
arbitrary choice. The properties of the material are influenced by the crystal orientation
that is to say different crystal planes have, may have different properties and which will
influence the device performance. That is why, while bipolar junction transistors can be
fabricated on 111  substrate, for MOSFET fabrication you would prefer a 100  wafer.
So, in today’s class, let us take a look at the silicon crystal.

Now, the simplest crystal structure that you can think of is called a cubic lattice.
(Refer Slide Time: 3:38)

If you look at the card we have here, where the picture of a simple cubic lattice is there,
to the left of the picture you can see the simple cubic lattice. What is it? It is nothing but a
simple cube with atoms placed at each corner of the cube and if I think of a three
dimensional co-ordinate system, three axis x, y and z and if I consider that the
dimensions of this cube is unity that is, it is an unit cell, then the co-ordinates of each of
this corner atoms will be given like  0,0,0  ,  0,1,0  ,  0,0,1 etc, etc. Ok? So, this is
the simplest structure that one can think of - a simple cubic lattice.

Let us take it one step further in complexity. You know, a cube will have 6 faces - front
and back, top and bottom, left and right. Now, in addition to the corner atoms, 8 corner
atoms placed in the corners of the cube, if we think of atoms in the middle, in the center
of each of the faces, then what we get is a face centered cubic lattice and that is the
picture you can see towards the right of the card. In addition to the 8 corner atoms, you
also see 6 extra atoms placed in the center of each face and their co-ordinates are given

1 1  1 1   1 1  1 1 1 1 1 1
by  , ,0  ,  , ,1 ,  0, ,  , 1, ,  ,  ,0,  and  ,1,  . Right? Very
2 2  2 2   2 2  2 2 2 2 2 2
simple, is it not?
Silicon however is neither a cubic lattice nor a face centered cubic lattice. We have to
take one further step in complexity. You have seen the face centered cubic lattice, fcc
lattice. Now, think of two fcc lattice, two fcc sub-lattice rather, one pushed inside the
other along a major diagonal by a distance of one fourth. That is, we have two face
centered cubic sub-lattice; two interpenetrating f centered cubic sub lattice, one is pushed
inside the other along a major diagonal and displaced by one fourth in all the direction.
What we have is this model here.

(Refer Slide Time: 6:57)

This is called a zincblende structure, two interpenetrating fcc sub-lattice. Look carefully
at this model.
(Refer Slide Time: 7:37)

For the time being, forget about the blue balls and concentrate only on the yellow balls or
part of the balls. What do you see? This is nothing but an fcc lattice, right? I have atoms
at the corner of the cube and also at the center of the each of the face. In addition to this,
now see, that I have 4 blue balls, the blue balls, to distinguish between the first sub lattice
and the second sub lattice and what are the coordinates of these blue balls? Let us
consider that this point is the origin. This is the x- axis, this is the y-axis, this is the z-

1 1 1
axis. Then, this blue ball here, it has the co-ordinates of  , ,  . It is displaced from
4 4 4
the origin in all directions by one fourth.

What about this one? It is displaced in x and in z direction by three fourth and in y

3 1 3
direction by one fourth. That means its co-ordinate is  , ,  . What about this one?
4 4 4
This is displaced by three fourth in x and y direction and one fourth in z direction. So,

3 1 3 3 3 1 1 3 3
accordingly you have  , ,  ,  , ,  and  , ,  . So, the coordinates of this
4 4 4 4 4 4 4 4 4
4 blue balls are actually they belong to the second fcc sub lattice and it has been pushed
inside the first fcc sub lattice and it is, it has these following co-ordinates. That is, the co-
ordinates will be . Right?

(Refer Slide Time: 9:37)

Now, when I say it is a zincblende structure, what I mean is that the two sub lattices they
are made of different atoms. For example, gallium arsenide. Gallium arsenide is a
zincblende crystal.

(Refer Slide Time: 10:30)

That is in this model you can consider that this yellow balls refer to gallium atoms and
blue balls refer to arsenic atoms. Inside the gallium sub lattice you have pushed an
arsenic sub lattice or the other way round. You could also think that the yellow atoms are
arsenic atoms and the blue atoms are gallium atoms, makes no difference. All I am trying
to say is that each zincblende lattice is comprised of two interpenetrating fcc lattice; one
lattice, two interpenetrating fcc sub lattice, let us say. One sub lattice is comprised of
gallium atoms, the other sub lattice is comprised of arsenic atoms. Now silicon is a
degenerate form of zincblende structure. This is called a diamond lattice.

(Refer Slide Time: 11:32)

When I say it is a diamond lattice, what I mean is that there is no difference between the
two sub lattices. They are made of identical elements. In this particular case, both the sub
lattices are comprised of silicon. So, silicon is a degenerate form of zincblende structure
which is called the diamond crystal structure. Now that you know what the crystal
structure of silicon look like, let us look a bit closer to the zinc blende or the diamond
structure and see what information we get from here. First of all, let me tell you that in
the unit cell we have considered that the cube is of unit dimensions. Ok? Now, if the unit
of this dimension is called a, a is called the lattice constant of the particular material. That
is each side of the cube is a, it has a length a. So, a is called the lattice constant. Ok?
For a zincblende structure, what is the distance between two neighboring atoms? That is,
can you think what is going to be the distance between the two neighboring atoms in a
zincblende structure like this? Let us take the case of the atom which is placed at one

1 1 1
 , , .
4 4 4

(Refer Slide Time: 13:52)

1 1 1
That is here, the blue ball,  , ,  shifted in all directions from the origin. What are
4 4 4
its nearest neighbour? One nearest neighbour is of course the atom at the origin  0,0,0  .
What are the other nearest neighbours? One is this face centered atom, the other one is
this face centered atom and the other one is the bottom face centered atom. Is it not? So,
the distance between two nearest neighbors can be calculated by simply calculating the

1 1 1
distance of this  , ,  atom from the origin and what is that distance?
4 4 4
(Refer Slide Time: 14:52)

That distance is going to be and if I consider that this cell, it has a lattice constant of a,

3
then this distance will be a . Is it not? So, each atom of the second sub lattice, it
4
forms a tetrahedron with the atoms of the first sub lattice that is, with its four nearest
neighbour. If you look at this blue ball, it has these 4 nearest neighbours and it forms a
tetrahedral structure. What is a tetrahedral structure? It is like this.

(Refer Slide Time: 16:12)

You have this blue ball here and the other one going inside the board. Yes? Now, if I
consider that these atoms are nothing but hard spheres that is called the hard sphere
model; just as you are seeing here, hard sphere model, then what can be their maximum
radius? How close can they be spaced when they are touching each other? Is it not? Very
like in this model; you can see here that the blue ball is touching its nearest neighbours on

3
all four sides and you know that the distance between them is a.
4

So essentially I have a situation, this is my blue ball, this is one of the nearest neighbours.
How close can we place them when they are touching each other? Right? Maximum
closest possible positioning will be when they are touching each other. Ok? So, when
they are touching each other, let us forget about all the other neighboring atoms and
concentrate on only one.

(Refer Slide Time: 17:58)

 3
So, then it is like this and the distance between them is this and this is a . Therefore,
4

3
each atom can have a radius maximum half of this distance that is, a . This is called
8
the tetrahedral radius.
(Refer Slide Time: 18:34)

This is, if we assume hard sphere models and each atom its sphere is touching the other,

3
then the tetrahedral radius is a . Ok?
8
(Refer Slide Time: 19:22)

For example, silicon has a lattice constant that is, a for silicon is 5.43 A 0 and therefore

3
its tetrahedral radius a , that is 1.18 A 0 .
8

(Refer Slide Time: 19:49)

Tetrahedral radius is, usually the symbol that is used is r0 . Ok? Now, let us look at the
packing density of this zincblende structure.
(Refer Slide Time: 20:13)

You can see in this model itself that not the entire space is taken up by the atoms. It
cannot be, right? How does one calculate the packing density of a particular crystal
structure? First of all, we must see how many atoms are there in a particular unit cell. So,
this is my zincblende unit cell. How many atoms are there? Let us look at the corner
atoms first. Ok? Take this corner for example. This corner atom does not belong
exclusively to this unit cell. Yes? It is shared by many others. How many others? That is,
I can think of placing another unit cell here, another unit cell here, another unit cell here,
that makes it 4 and then I can place another set of 4 such unit cells on top of this. See, this
is the top corner. I can think of another set of 4 unit cell for which this atom will be the,
at the bottom corner. So, each corner atom is shared by 8 unit cells. So, each of their
contribution to one unit cell is actually one eighth.

(Refer Slide Time: 22:13)

One eighth for each corner atom and how many corner atoms I have? 8. Therefore, the

1
total contribution is x8  1 . Right? Next, let us consider the face centered atoms. So,
8

1
the contribution of each face centered atom to 1 unit cell is . It is shared by two, one
2
here and the other on top of this. Ok? So, contribution of face centered atoms and how

1
many such face centered atoms I have? 6, because there are 6 faces. So, x 6  3 and
2
now, for the zincblende structure, I have 4 atoms of the second sub lattice and they are
not shared by anybody, they are present there in full.

(Refer Slide Time: 23:53)

Therefore, the contribution of the second sub lattice atom is actually 1x 4 . So, I have a
total of 8 atoms in the unit cell. Agreed? I have a total of 8 atoms that is, 8 full spheres.
Considering hard sphere model, I have 8 full spheres inside the unit cell and each atom

3
can have a maximum radius of a . Ok? So, what is the total volume occupied by
8
these atoms inside the sub lattice? What is the volume of a sphere?

(Refer Slide Time: 25:01)

 4
You know, it is r 3 . r is actually 3 a and there are eight such spheres.
3 8

(Refer Slide Time: 25:21)

3 3
So, it works out to be a .
16
3
4  3  3 3
8x  a   a
3  8  16

This is the volume occupied by the atoms inside the unit cell and what is the volume of
the unit cell? a 3 .

(Refer Slide Time: 26:27)

So, the area, the volume that is occupied by the atoms inside the crystal with respect to

3 3
the total crystal volume is simply , which is about 34%; works out to be
16 16
about 34%. Let us for a moment think about a simple fcc lattice, face centered cubic
lattice. We have seen that for a zinc blende structure, the packing density is 34%. That is
34% of the total unit cell volume is occupied by the atoms. Now, if you compare it with a
face centered cubic lattice, is this packing density more or less? Think about a face
centered cubic lattice. You can have it here, just ignore the blue balls.

(Refer Slide Time: 27:39)

We will have a face centered lattice. So, the nearest neighbour is this with this. Ok?

(Refer Slide Time: 27:56)

So, the distance between the nearest neighbours can be calculated to be; it is actually
2 2
1  1  2 1
  0    0  0 
 2   2  2
Right? So, this works out to be; and if I consider the crystal lattice to be a, then this is

a
actually and in the fcc lattice, you do not have this interpenetrating second sub
2
lattice atoms. Therefore, the total number of atoms present inside is only 4. Ok? So, the
volume that is occupied will be.

(Refer Slide Time: 29:19)

a
The distance between the nearest neighbors is . Therefore, the tetrahedral radius can
2

a
be and I have only 4 spheres, 4 such spheres. Therefore, the total volume that is
2 2

3
4  a 
occupied is only this much; 4 x   .
3 2 2
3
4  a  
4x    a3
3 2 2 3 2

(Refer Slide Time: 30:05)

So, this works out to be or the percentage of volume that is occupied that is going to be


simply ;
3 2

(Refer Slide Time: 30:53)

which actually works out to be more than 70%. So, you see, compared to a zincblende
structure, an fcc lattice is much more densely packed. The packing density is much higher
for a face centered cubic lattice compared to a zincblende structure. Now, when you try
to dope silicon that is, we would like to introduce impurity atoms inside the silicon
crystal and these impurity atoms, you know, should replace the silicon atoms. Right? So,
one of this positions, one or many of this positions should be taken off by the impurity
atoms. It should push the silicon atoms out and take that position instead, right?
Incidentally, this is called a substitutional kind of introduction, right? That one impurity
atom it goes and substitutes one silicon atom.

Now, silicon has a particular tetrahedral radius. Just now we have said that it has a lattice
constant of 5.43 and therefore, its tetrahedral radius works out to be 1.18 A 0 . The
impurity atom may or may not have the same tetrahedral radius. Yes? So, what will
happen? If it does not have the same tetrahedral radius, then you can introduce strain in
the lattice. The lattice will be strained. This is a very important consideration. If the
tetrahedral radius of the impurity atom differs considerably from that of silicon, then the
lattice will be under a lot of strain and this will eventually dictate the amount of
electronic activity.

Now, what do I mean by electronic activity? Suppose you are introducing impurity atoms
inside silicon. They will be electronically active only when they have occupied the proper
positions in the lattice that is, when they have substituted for silicon. Otherwise most
likely they will not be electronically active. They might be precipitated, they might stay
in the interstitial points. So, they will not be electronically active. So, you can keep on
putting a lot of impurity. You can try to put a lot of impurity inside the silicon, but finally
the carrier concentration that will not be at par with the amount of impurity that you have
introduced. This is what we mean by electronic activity. All the impurity atoms you have
put inside the silicon may not be electronically active and the difference in the tetrahedral
radius will dictate to a large extent as to how much electronically active a dopant material
is going to be. I have the tetrahedral radii of a few common dopant impurities in silicon.

(Refer Slide Time: 35:17)

So, silicon, phosphorus, arsenic, antimony and boron. For silicon it is 1.18. For arsenic
also this is 1.18. For phosphorus this is 1.1. For antimony this is 1.36 and for boron this is
only 0.88. In this context, we also define a term called the misfit factor, which is denoted
usually by the symbol  . We define the radius, the tetrahedral radius of the impurity
atom as r0 1    .

(Refer Slide Time: 36:44)

r0 is the tetrahedral radius of silicon,  is the misfit factor and rimpurity is the

tetrahedral radius of the impurity atom. So, we co-relate the two tetrahedral radii by the
misfit factor and obviously from this, you can see that the misfit factor is going to be very
large for boron, because the tetrahedral radius of boron differs maximum from that of
silicon.

(Refer Slide Time: 37:42)

So,  for boron is actually 0.25, for arsenic on the other hand it is zero. They have the
identical tetrahedral radii. For phosphorus also it is pretty small and it is about 0.068, for
antimony it is about 0.15. This brings us to a very interesting point. Boron is a Group III
element. So, we need boron in order to dope the silicon p-type. You see, unfortunately
the tetrahedral radius of boron differs considerably from that of silicon, so that the misfit
factor is actually the largest among all. On the other hand the Group V elements,
particularly arsenic and phosphorus, they have zero or very small misfit factors. Think of
what I said about electronic activity. The misfit factor is going to determine the electronic
activity. The larger the misfit factor, the more difficult it is going to be to make the
dopant electronically active. Yes?
So, you see, it is more difficult to heavily dope silicon p-type, because boron has such a
large misfit factor. It is more difficult to have a heavily doped p-type region in silicon
compared to an n  region in silicon. Corollary - it is easier to fabricate an npn transistor
than a pnp transistor. Why? Because in a pnp transistor, the emitter needs to be p-type.
The emitter is the most heavily doped region, should be the most heavily doped region in
the transistor and since we use boron for doping it p-type, it is going to be more difficult
to have, to achieve very high doping concentration, very high carrier concentration, not
just doping concentration, electronically active dopant that is high carrier concentration,
if you are going to have a pnp transistor. This is one of the various reasons why, for
integrated circuit we prefer to use npn transistor wherever possible rather than pnp
transistor. There are various other reasons why you prefer npn transistor, but one of them
is related to the lattice constant and misfit factor of silicon and boron. So, now you know
about the crystal structure of silicon, about the packing density of the particular
zincblende structure and about the misfit factors.

Next, let us concentrate on the crystal planes that is, the crystal orientation. We
mentioned that for certain particular applications, we prefer certain planes. Why is it so?
Let us first look at some of the common crystal planes.

(Refer Slide Time: 41:41)

This is a crude sketch of the unit cell. I have just drawn the cubic structure, I have not
drawn the atoms, ok? Now, a plane is actually, you know, defined by the intercepts on the
three axis x, y and z axis. So, you have considered that this is the x-axis, this is the y-axis
and this is the z-axis.

(Refer Slide Time: 42:42)

So, a plane is actually defined by this equation that is where a, b, c are the intercepts
made by the plane on x, y and z axis respectively. I can also define this plane as
hx  ky  lz  1 , where h, k, l are the reciprocal of the intercepts.
(Refer Slide Time: 43:34)
These h, k, l are referred to as the miller indices and the planes are commonly referred to
by this miller indices; h, k, l values that actually defines the plane.

(Refer Slide Time: 44:03)

Coming back to this diagram, let us consider this plane, this plane or this plane, you
could say. On the board you can see this is the plane. This makes an intercept on the z-
axis and it is parallel to both x and y axis. That is its intercepts on the x and y axis will be
infinite. What is this plane called? This plane is therefore called, its z- axis intercept is 1;
the reciprocal of 1 is also 1. Its x- axis intercept is infinite, reciprocal of that is zero and
its y-axis intercept is also infinite, the reciprocal of this is zero.

(Refer Slide Time: 45:15)

Let us take this plane. This makes an intercept on the y-axis alone. So, this plane will be
called  0,1,0  . Its y-axis intercept is 1, its x and z-axis intercepts are both infinite,
therefore, this plane will be called  0,1,0  . Similarly you can take the top plane and you
can see that this is actually 1,0,0  .

(Refer Slide Time: 46:21)

The entire family of such planes which makes intercept only on one axis and are parallel
to the other two axis will be given as, notice the difference. When I put them inside the
first bracket, I am being specific. I am specifically telling what is this particular plane.
When I am putting the miller indices inside the second bracket, I am referring not to one
particular plane, but to a class of planes which makes collectively all those, make
intercepts only on one axis and are parallel to the other two axis. They are referred to as
1,0, 0 planes.

(Refer Slide Time: 47:07)

Similarly, I can think of a plane going like this and this plane will be one in the family of
1,1, 0 planes, 1,1, 0 planes. So, this plane will be belonging to the family of
1,1, 0 planes.

(Refer Slide Time: 48:13)

Similarly I can even think of another plane like this, a triangular shaped plane and this
plane you can see makes intercepts all the three axis and therefore this belongs to the
family of 1,1,1 planes. So, they are the three basic families of planes we need to

discuss when discussing the crystal orientation of silicon. They are 1,0, 0 , 1,1, 0 ,
1,1,1 . Depending on the particular plane, the properties of the material will differ and
in the next class we will see how the properties of the material is going to depend on the
particular crystal plane.