Actually, implants (especially for moats) are
usually done at a sufficiently high energy so that the dopant(phosphorus) is already pretty far into the substrate (often
several microns or so), even before the diffusion starts. Theanneal/diffusion moves the impurities into the wafer a bit more,
and as we shall see also makes the n-region grow larger.
"The n-region"! We have not said a thing about
how we make our moat in only certain areas of the wafer. Fromthe description we have so far, is seems we have simply built an
n-type layer over the whole surface of the wafer. This would bebad! We need to come up with some kind of "window" to
only permit the implanting impurities to enter the silicon waferwhere we want them and not elsewhere. We will do this by
constructing an implantation "barrier".
To do this, the first thing we do is grow a layer
of silicon dioxide over the entire surface of the wafer. Wetalked about oxide growth when we were discussing MOSFETs but
let's go into a little more detail. You can grow oxide in eithera dry oxygen atmosphere, or in a an atmosphere which contains
water vapor, or steam. In
, we show oxide
thickness as a function of time for growth with steam. Dry
does not behave too much differently, the rate is just somewhat
slower.
Oxide thickness as a function of time
On top of the oxide, we are now going to deposit yet another
material. This is silicon nitride,
or just plain "
nitride " as it is usually
called. Silicon nitride is deposited through a method calledchemical vapor deposition or "CVD". The usual technique is to
react dichlorosilane and ammonia in a hot walled low pressurechemical vapor deposition system (LPCVD). The reaction is:
Silicon nitride is a good barrier for impurities, oxygen and
other things which do not want to get into the wafer. Take alook at
and see what we have so far. A
word about scale and dimensions. The silicon wafer is about250m thick (about 0.01") since it has to be strong enough
not to break as it is being handled. The two deposited layersare each about 1m thick, so they should actually be drawn as
lines thinner than the other lines in the figure. This wouldobviously make the whole idea of a sketch ridiculous, so we will
leave things distorted as they are, keeping in mind that thedeposited and diffused layers are actually
much thinner than the rest of wafer, which
really does not do anything except support the active circuitsup on top. (There we go again, wasting silicon. Good thing it's
cheap and plentiful!)
Initial wafer configuration
Now what we want to do is remove
part of
the nitride, so we can make our n-well, but not put inphosphorous where do not want it. We do this with a processes
called
photolithography and
etching respectively. First thing we do is coat the wafer with yet
another layer of material. This is a liquid called
photoresist and it is applied through a process
called
spin-coating . The wafer is put on a vacuum
chuck, and a layer of liquid photoresist is sprayed uncap of thewafer. The chuck is then spun rapidly, getting to several
thousand RPM in a small fraction of a second. Centrifugal forcecauses the resist to spread out uniformly across the wafer
surface (most of it in fact flies off!). The solvent for thephotoresist is quite volatile and so the layer of photoresist
dries while the wafer is still spinning, resulting in a thin,uniform coating across the wafer
.
Photoresist is spun on
The name "photoresist" gives some clue as to what this stuff
is. Basically, photoresist is a polymer mixed with some kind oflight sensitizing compound. In
positive photoresist, wherever light strikes it, the polymer is weakened,
and it can be more easily removed with a solvent during the
development process. Conversely,
negative photoresist is strengthened when
it is illuminated with light, and is more resistant to thesolvent than is the unilluminated photoresist. Positive resist
is so-called because the image of the developed photoresist onthe wafer looks just like the mask that was used to createit. Negative photoresist makes an image which is the opposite of
what the mask looks like.
We have to come up with some way of selectively
illuminating certain portions of the photoresist. Anyone who hasever seen a projector know how we can do this. But, since we
want to make
small things, not big ones, we
will change around our projector so that it makes a smallerimage, instead of a bigger one. The instrument that projects the
light onto the photoresist on the wafer is called a
projection printer or a
stepper
.
A stepper configuration
As shown in
, the stepper consists of
several parts. There is a light source (usually a mercury vaporlamp, although ultra-violet excimer lasers are also starting to
come into use), a condenser lens to image the light source onthe
mask or
reticle . The mask contains
an image of the
pattern we are trying the place on
the wafer. The projection lens then makes a reduced (usually 5x)image of the mask on the wafer. Because it would be far too
costly, if not just plain impossible, to project onto the wholewafer all at once, only a small selected area is printed at one
time. Then the wafer is
scanned or
stepped into a new position, and the image is
printed again. If previous patterns have already been formed onthe wafer, TV cameras, with artificial intelligence algorithms
are used to
align the current image with the
previously formed features. The stepper moves the whole surfaceof the wafer under the lens, until the wafer is completely
covered with the desired pattern. A stepper is notcheap. Usually, TI or Intel will fork over several million
dollars for each one! It is one of the most important pieces ofequipment in the whole IC fab however, since it determines what
the minimum feature size on the circuit will be.
After exposure, the photoresist is placed in a
suitable solvent, and "developed". Suppose for our example thestructure shown in
is what results from
the photolithographic step.
After pr expose/develop
The pattern that was used in the photolithographic (PL) step
exposed half of our area to light, and so the photoresist (PR)in that region was removed upon development. The wafer is now
immersed in a hydrofluoric acid (HF) solution. HF acid etchessilicon nitride quite rapidly, but does not etch silicon dioxide
nearly as fast, so after the etch we have what we see in
.
After nitride etch
We
now take our wafer, put it in the ion
implanter and subject it to a "blast" of phosphorus ions
.
Implanting phospohrus
The ions go right through the oxide layer on the RHS, but stick
in the resist/nitride layer on the LHS of our structure.