VLSI Technology

Dr. Nandita Dasgupta

Lecture - 20
Ion-implantation systems and damages during implantation
So, in our discussion about ion implantation, so far you have seen that ion implantation is
a more sophisticated, more controlled, more flexible way of doping a semiconductor and
the profile that you get by ion implantation is given by a Gaussian distribution. This
Gaussian distribution is strictly valid when the target is amorphous, but even for
crystalline targets the Gaussian profile can be assumed, provided we follow certain
precautions. Ok. Now, let us come to the actual ion implantation system. What are the
different parts in that ion implantation system, how exactly does it function and how do
we really control the dose by varying the parameters of the implantation system? So, first
of all, in an implantation system the thing that is needed that is an ion source. We must
have a source for the dopant and it must be ionized, right?

(Refer Slide Time: 2:31)

1
So, the first important aspect of an implantation system is an ion source. What do we
have in this ion source? We have actually a dopant source and an ionization scheme. First
of all, I must have a dopant source and then some means to ionize that. Dopant source
that we use is obviously a compound of the dopant material and it is usually a gas, a
gaseous compound of the dopant material and then, we have to have an ionization
scheme. Normally an electronic discharge is used in order to ionize this gas, so that you
get the ion beam of the dopant, ionized dopant species. Sometimes magnetic field is also
used in addition to the electronic discharge to improve the ionizing efficiency. So, at the
end of the ion source you have a small outlet, through which now the ionized dopant
species can pass. They now go to the next stage in the implantation system which is the
accelerator.

(Refer Slide Time: 4:19)

As the name suggests, what is an accelerator going to do? It is going to accelerate the ion
beam. It is going to impart energy to the ion beam and accelerate it down the line. Ok?
Next comes a very important concept, the concept of mass separation. Now, before
talking about what a mass separator really is, let me tell you once again, so far what has
happened. I have created the ionized dopant species, I have accelerated them, so that now
I have an ion beam of particular energy and this is going to impinge on the

2
semiconductor. But remember, we said that ion implantation is a process much less prone
to contamination. Ok? Now, you have used a compound of the dopant species and you
have ionized this. So, there are various ions in this ion stream; possibility of having
various type of ions in this ion stream. How then are you going to select a particular ion
species to come and impinge on the semiconductor? This is done by doing the mass
separation. In a mass separator, usually for ion implantation, what we use is a
homogenous field magnetic analyzer, homogenous field magnetic analyser.

(Refer Slide Time: 6:30)

What are your ion beams, actually? This is a combination of a stream of charged particles
with a particular mass. Ok? So in general, I can say a charged particle of mass m moving
with a velocity v. That small m and small v characterize the particular ion species; it is
going to be different for different species of ions, agreed? Now, you are subjecting them
to a uniform magnetic field B, perpendicular to its direction of motion, perpendicular to
its velocity, right? So, under these circumstances, this particle is going to experience a
force which is given by F, q is the electronic charge, small v is the velocity and B is the
uniform magnetic field that is applied, ok and because of this force, this will tend to move
the particle in a circular direction, will try to move it like this and that circular path will
have a radius say, small r and this particle will therefore experience a centrifugal force

3
 mv 2
given by , right? This centrifugal force is equal and opposite to the direction of this
r
force, right?

(Refer Slide Time: 8:56)

mv 2
So, I can say that this is equal and opposite to the centrifugal force ; their
r
magnitudes are the same, right? So in this, I can therefore get that v is equal to.., can I
not? At the same time, remember how this velocity was imparted to this ion beam? This
velocity was imparted in the accelerator section by using an electric field. Ok?

4
(Refer Slide Time: 10:05)

1
So, the kinetic energy of these particles is given by mv 2 and that is imparted by the
2
potential that is applied to it in the accelerator and if that potential is V, then this is the
relationship; the potential energy gets converted into the kinetic energy. From this,

2qV
therefore I can say that v  ; this V is the potential that is applied, divided by m.
m
qxBxr
Now, if I substitute this value of v here, what do I get? I get this is equal to .
m

5
(Refer Slide Time: 11:20)

So, the radius of the circular motion of the particle that is r, r is therefore given by, r

1 2Vm
therefore can be given as , agreed? Now you see, in a particular system, I know
B q
what the value of B is, I know what the value of potential is and of course electronic

charge that is a constant. Therefore, r is simply proportional to m , where m is the mass

6
(Refer Slide Time: 13:02)

If I have three different ion species with mass, masses m1 , m 2 and m 3 , m1 being the

smallest, m 3 being the largest, m 2 being somewhat intermediate, then when they are
passing through this magnetic analyzer, they will move with three different radius of
curvature like this, agreed? So now, if I have a slit like this, if I place a screen here, with
just a slit, then this one gets stalled here, this one gets stalled here, but this one can pass
through the slit. So, this is basically the mass separation scheme. Because you know that
the radius of curvature is proportional to m, you can place a screen with a slit suitably, so
that only that particular ion species with this particular mass will pass through this slit
and therefore you have kept your semiconductor somewhere here, this is where your
target is, it will go and hit the semiconductor, right? If you want even better control on
this, then instead of using the simple homogenous field magnetic analyzer, you can use
an electric and magnetic field together in mutually perpendicular directions, which is
called the wind filter, you know the wind filter, the eXh filter that can also be used, ok
and we have a fourth part of the implantation system.

7
(Refer Slide Time: 15:59)

If you have a small target, then this is good enough; you have the mass separation and
this is coming and hitting on the surface. But, in actual practice you know you may want
to scan over a larger area. So, this beam has to be scanned up and down, and to and fro.
So, that is the final part of the implantation system, XY scanning. What you have now is
an ion beam of a particular ion species after the mass separation stage. Now, this ion
beam you can raster it all over the semiconductor wafer, right? So, you can use an XY
scanning to move it up and down, to and fro, in order to implant all over the
semiconductor, wherever necessary.

So, these are the four important parts of the implantation system. First of all, the
ionization - the ion source, the dopant material as well as the ionization scheme, where
you are creating the dopant species, the ionized species. Then, you have the accelerator
where you are imparting velocity to this ion beam. Then, you have the mass separator,
where you are selecting a particular ion species and finally the XY scanning in order to
implant selectively over the semiconductor material.

Now, the question is how do we control the dose? What are the parameters we play with
in order to control the dose? Controlling the dose is done in the first stage itself that is in

8
the ion source itself. What we do is the parameter that the system operator has is the
beam current. Now, what exactly is the beam current?

(Refer Slide Time: 18:24)

The beam current is I; that is the parameter, the system operator has in hand. Ok? Now, if
I have an ion species of mass, of charge q and say the charge state that is whether it is
I
singly or doubly ionized, ok if I call that charge state say, n for example, then this is
nq
actually the number of ions hitting the semiconductor per unit time, right?

9
(Refer Slide Time: 19:26)

It
So, if I carry out implantation for a time t, then is the total number of ions that have
nq
It
hit the semiconductor surface, right and if the target area is A, then is equal to the
nqA
total number of ions incident on the target per unit area. What is this? Total number of
ions incident on the target per unit area. That is equal to my total dose. So you see, you
can control the dose either by controlling I or by controlling t or by controlling them
together. So, for a practical system what would you want?

You would want a high beam current, right? If you have a high beam current, then you
can impart a large dose within a relatively short time, agreed? t can be made small; if I is
large, then t can be made small and still I can get a large enough dose. What is the
advantage? Your throughput is faster, you can use the same implantation for various
implantation run and the time for each implantation can be kept short. So, this is how the
dose is controlled and of course, how deep the dopant is going to go that will depend on
its energy and on its mass. If the dopant ion is light, then with a relatively less energy, it
will go much deeper. If the dopant ion is heavy, then it will need comparatively greater
energy, larger energy in order to penetrate to a comparable distance. Ok? So, you see
these two things you can control separately. This is controlled at the ion source itself and

10
the energy of course, you are going to impart in the accelerator. So, you have two
parameters to play with: the beam current and the energy. Ok?

Now, so far you see, we have discussed only the positive aspects of implantation. That is
we have discussed how good implantation is, how much better control, how much
flexibility it offers, compared to diffusion for example. But, do not let us forget that
implantation actually means, physically you are hitting the semiconductor surface with a
stream of high energy particles and in the process it is going to create damage. You
already know that once these ionized particles penetrate the semiconductor surface, it is
brought to rest by two stopping mechanisms. One is nuclear stopping, the other is
electronic stopping. Of these two, the nuclear stopping is the primary cause for damage,
because it is colliding with the lattice atoms and in the process, imparting energy to the
lattice atom. Sometimes this energy may be sufficient so as to cause displacement of the
lattice atom.

Now, let us investigate this a bit more closely.

(Refer Slide Time: 24:39)

11
Suppose the incident ion energy is E 0 and suppose E d is the energy required to cause
displacement of a target atom.

(Refer Slide Time: 24:57)

To dislodge a target atom from the lattice site we need an energy E d , energy required to

dislodge a host atom. Then, if E 0  E d , no problem, no displacement; the incident ion

energy is not sufficient to cause any displacement. If however E 0  E d , but it is less than

2E d , then we will have a single displacement. It is greater than E d , it can cause a

displacement, but it is less than 2E d . So, we cannot have multiple displacement, right?

But, suppose if it is greater than 2E d , then we can have multiple displacement or a

So, on an average you can say that number of displaced atoms per incident ion will be
E0 E0
given by ; on an average, you can say that it will be . Now for silicon, this
2E d 2E d

value of E d is 15 keV and for implantation, we frequently use energies in excess of 100

12
keVs. So, you can understand how much damage there is going to be created, how many
displaced lattice atoms there will be. So, obviously the atoms are, host atoms, are
displaced from their original lattice site. There can be, they can create, they will create a
vacancy as well an interstitial, so that means a Frenkel pair. So, a lot of Frenkel pairs can
be generated, lots of damages will be there.

How to estimate and rectify this damage caused by implantation that is the major
challenge. How to estimate the extent of damage and then how to rectify it? So, damages
caused by implantation is the major flip side or the major negative aspect of implantation
- damages caused by it. Unless and until we can rectify these damages, the semiconductor
will be of no use. So, first of all how do you, how do you make out whether there is
damage or not? A very crude way of making out whether there are damages is by visually
inspecting. You know that a normal silicon surface is going to be mirror polished, right?
The top surface on which you are going to fabricate, you are going to dope, you are going
to create your devices close to the top surface, that top surface is going to be mirror
polished. If there are damages, then the mirror polish will be the first casuality. The
mirror polish will be lost; you will have a sort of cloudy or milky surface.

(Refer Slide Time: 29:47)

13
So, damages can be inspected first visually. This is of course a very crude way of looking
at things. Immediately after implantation, we will find that the surface has become cloudy
or milky. So, you know that the surface is damaged; that is all. You cannot estimate any
better than that, it is very crude way of looking at it. If you really want to estimate the
extent of damage, then directly, then you can try electron microscopy or you can try
Rutherford backscattering, which will tell you about the exact crystalline imperfections
present in the damaged surface. So, you can try electron microscopy. But we, the
technologists, the VLSI engineers, do not forget we are primarily electrical engineers, so
we are concerned about the electronic activity of the surface, right? We are not interested
in the crude visual inspection or in the crystal imperfections, the direct investigation of
the crystal imperfections. We are interested in figuring out how the device performance is
going to be affected; what are the parameters, what are the factors which have got
affected, that is of concern to us. What are these factors?

First of all for an electronic engineer, the most important idea is the conductivity, right?
In a damaged surface, the conductivity is going to be extremely low. You are going to
have a highly resistive surface, because of the damages; the electronic activity will be
extremely low. So, change in conductivity and then, in a damaged surface, the mobility is
also going to be extremely low. So, mobility is the second casuality and finally the life
time that is the third casuality. So, we are most interested about these three factors. How
rho has changed - how the conductivity has changed, how mu has changed - how the
mobility has changed and what has happened to the life time?

Obviously, as a result of the implantation damage, all these three factors are going to get
degraded and how much they have got degraded will give me an estimation of the
damage. This is the only estimation I am interested in, because it is going to reflect
directly on the device performance and I will say that I have rectified the damage only
when I can restore these three factors to their expected value, right; because these are
directly concerned with the device performance. I have to get back the electronic activity
that is I have to get the active carrier concentration. You see, I may have put so many
dopants inside the system, inside the semiconductor by ion implantation. But, they are not

14
electronically active. That is why the resistivity is very high,  is very high. I must get the
active carrier concentration. Ok? Secondly, the mobility of that damaged region will be
extremely low. I must get back the expected high mobility and then I must get back the
expected carrier life time.

Before we try to get back these three factors to their expected value, let us look at the
nature of the damage as created when the implantation is due to a, I mean, when we have
doped the semiconductor with a heavy ion such as phosphorus, compared to when we
have doped the material with a light ion such as boron. Will there be any basic difference
in the nature of the damage? Let us try and investigate that first. You know what happens
when the ion beam is incident on the semiconductor. It is going to lose its energy and
primarily, nuclear stopping is the mechanism that is responsible for the damages. Ok?

(Refer Slide Time: 35:37)

So, when a light ion is incident on the semiconductor, ok let us talk about light ion; it has
got a small mass, it hits the substrate, undergoes collision with the lattice atoms. So,
during each collision, it is transferring a relatively small amount of energy, ok and
because this incident ion in itself is a light ion, at each collision it is getting displaced
through a large scattering angle, right? The lattice atom is much heavier compared to the

15
incident ion. So, the incident ion is getting deflected by a large scattering angle, but it is
transferring only relatively less amount of energy. So, if this is the semiconductor surface
and if this is the ion beam coming, then you see its path is like this. It hits here, ok in the
process it gets deflected by a large angle; it is absolutely random, it can be this way or
this way. Then again, it hits a lattice atom. Again it gets deflected by a large angle and
the process continues and at each collision, since it is transferring a relatively smaller
amount of energy, it is moving through a considerable distance before coming to rest. So,
what you see is actually a long narrow branch of damage.

The damage will be relatively less in this case, because you see at each collision it is
transferring relatively small amount of energy to the lattice atom. So remember, that I
need minimum of E d 15 keV for silicon; E d is 15 keV in order to dislodge the host atom.
Because the light ion is transferring a smaller amount of energy, it may not be sufficient
to cause multiple dislocations. So, at every point you see, I have just a small damage.
But, the incident ion is moving much deeper. Ok? So, I have relatively less damage, but a
long narrow branch. This is what is going to happen in case of a light ion incident; ok, an
extended region with small branching dislocation trap here and here, may be a small this
thing; if any multiple dislocation is taking place, it is very small at these points.

(Refer Slide Time: 38:57)

16
Now, let us take the case of a heavy ion. As before, this is the surface. Now, what is
going to happen? This is a heavy ion. At every collision, it transfers a large amount of
energy to the lattice atom and it gets deflected only by a small scattering angle, right?.
So, because it transfers a larger amount of energy to the lattice atoms, those lattice atoms
in their turn will again cause dislocations and I will have multiple dislocations. At the
same time since the incident ion is suffering, it is getting deflected only by a small
scattering angle, I will have the maximum damage within a small region, agreed? So,
what I will have is actually a horribly damaged region, but within a small volume. At
every point, there is a collision and that collision is sufficient; the lattice atom has
obtained sufficient energy, so as to cause further damage. So, it is full of damages, that
region is full of damages, but within a smaller volume. So, you see essentially, the nature
of the damages for a light ion and a heavy ion is basically they are very different. For the
light ion I have an extended region with relatively less damage, for a heavy ion I have a
lot of damage confined within a small volume. Ok? So, before I try to rectify the damage
I must know whether the incident ion was light or heavy, then accordingly we can plan
our actions. Ok?

Now, the other factor that is going to contribute to the damage is the dose. That is very
easy to understand. If I have a large enough dose that means here I have shown you just
for one single incident ion, if I have large enough dose, so many such ions, the whole
thing is getting multiplied. So, if I have a large enough dose the damage is more; if I have
a small enough dose, the damage is less. If now I have a large dose of a heavy ion,
phosphorus for example, a large dose of phosphorus, then the damage will be so heavy,
so that there is a possibility of an amorphous layer formation in silicon. So much damage
that its crystalline structure is completely lost and the material has become amorphous in
nature.

Obviously, for the light ions, for example boron, the dose required to amorphize the
silicon will be much larger compared to phosphorous, because I have to create the same
amount of damage in order to amorphize the region. Since boron is a much lighter ion,
its, the creation of damage is much less. Therefore, in order to amorphize silicon, the

17
dose of boron must be much, much heavier, right? Ok. But do not think, do not go away
with the impression that this is something very bad; that amorphization of silicon is
something very bad. It is not bad; in fact sometimes it can be used to our advantage.
What are these advantages? First major advantage is of course in the control of the
doping profile.

(Refer Slide Time: 43:55)

If you have an amorphous surface, then of course all the channeling and all that will be
much less and I will have a primarily Gaussian profile. If the dose is more, you will have
a more Gaussian profile than if the dose was less. That is why, sometimes prior to
implantation, silicon is pre-amorphized by self-implantation or by an inert implantation.
You can either implant argon or neon or you can self-implant. That is you can implant
silicon on silicon in order simply to amorphize the region, so that in that amorphous
region when you next carry out the implantation, the profile will resemble a Gaussian
more closely. So, now you have a damaged region depending on whether the incident ion
is light or heavy, the dose is small or big, you have extent of damages and you know how
to estimate the extent of damage also. For us the most important factors to estimate this
damage are the changes in conductivity, mobility and the lifetime.

18
Now, we have to rectify this problem which is called the annealing of damages. Damages
must be annealed out. Annealing essentially means a high temperature treatment; subject
the material, the damaged material to a thermal treatment, when the damages will get
annealed out. Now ideally speaking, ideally what should we have? We should have
complete recovery, complete recovery of the lifetime, complete recovery of mobility and
100 % carrier activation; that is on one hand, this is what we want. But in practice, we
may have to fall short of this ideal requirement. What are the reasons? Why do we have
to, may have to fall short of these things? One is in the material itself. Silicon can be
subjected to 12000C with no problem. But, if silicon already has a metal pattern on that, if
it already has an aluminium pattern on that, then you cannot subject it to a temperature
higher than 5000C. So, the material tolerance may be a problem and the second problem
is of course, the longer and the longer is the duration and the higher is the temperature of
annealing, remember there is going to be considerable change in the doping profile. It
may not resemble the doping profile that was dictated by the ion implantation, because in
this case, ion implantation is going to be followed by drive-in.

Effectively what you are doing is after ion implantation, you are subjecting it to a drive-
in. So, the doping profile may get considerably changed, if Dt is considerably large. Dt
may even become larger than your projected range and in which case, the ion
implantation profile may get completely changed. So, these are the two major factors
which will dictate your annealing time and temperature. On one hand, you would like to
have the maximum carrier activation – 100% carrier activation, full 100% mobility
recovery and recovery of life time. On the other hand, you may be constricted to use a
less than perfect temperature and time for the annealing.

Now, in this case, let me just tell you some general guidelines. First of all, the lower the
implantation dose is, the easier it is to anneal it out; lower dose lesser damage, lesser
damage, the easier to anneal it out. That is one factor. Second factor is even here you see,
the crystal orientation matters. It is found that 100 anneals faster than 111 , ok and
now when we discuss about annealing, there are essentially two cases. One case is where
the silicon was amorphized prior to or during implantation and the other is where the

19
silicon retains its crystalline nature. You will be surprised that the silicon which was pre-
amorphized or amorphized during implantation is relatively easy to anneal than silicon
which was not amorphized. That is why I said amorphizing may not always be a bad idea.
In fact, it does offer certain advantages. This we will discuss in the next class.

20