Photolithography is the process of transferring a pattern from a mask to a substrate (GaAs in this case). The masks are glass or quartz outlines of the circuit design that is to be developed on the wafer. The idea of a mask is that holes or windows allow impurities to contact the desired portion of the wafer surface. After coating the wafer with a photoresist material, ultraviolet (UV) light is sent through the mask and onto the wafer. The pattern is then outlined in the photoresist using a stepper machine. Photoresist is then removed leaving the desired pattern. Processes of wet etching are then used to etch through the silicon-oxy-nitride and the GaAs layers. These basic steps are repeated with some variations until all the necessary layers are applied on the wafer. The entire process is done in a gold lit, dust free cleanroom to respectively keep from setting off the photoresist layer and to keep all dirt particles from resting on the masks or on the wafers. Photolithography can be compared to a very expensive photography process, and photolithographic machines are sometimes called printers or exposure towers. The difference between photography and photolithography is that the wavelengths in photolithography are much shorter. In photography, an image is created using the light of many different colors and intensities. However, in lithography, the range of exposing wavelengths is often extremely narrow. (Source: Murarka, S. P., and M. C. Peckerar, Chapter 8 in Electronic Materials : Science and Technology, San Diego: Academic Press, (1989).)
GaAsTEK receives semiinsulating gallium arsenide (GaAs) wafers 100mm in diameter and 625 microns thick from their supplier, M/ACOM. The very first step GaAsTEK takes in their wafer fabrication process is to check the received wafers for several acceptance criteria. The supplier has already checked the flatness and size specifications that GaAsTEK has required, therefore, their main concern is to test the wafers for electrical properties, such as conductivity. Conductivity is determined by measuring the resistivity (the inverse of conductivity) with a fourpoint probe, which does not touch the surface of the wafers. GaAsTEK implants the wafers with silicon for n-type doping and measures the sheet resistance with the contactless probes. After the wafers pass inspection the entire surface of each wafer is coated with an 820angstrom siliconoxynitride (SiON) layer. The purpose of this layer is to protect the wafer from damage caused by handling or etching during the photolithography process. The siliconoxynitride also protects against channeling during ion implantation. The wafers are now labeled with the product and lot numbers using a laser scribe. The next step in preparing the wafer for the photolithography process is to etch two alignment markers. The photolithography process itself forms the alignment markers. The flat edge of the wafer is used as the guide while putting down the more precise alignment marks. An adhesive, hexamethyldisilazane (HMDS), is applied to the wafer so that the photoresist layer sticks better to the wafer. The primer layer is baked on in a Vaporprime oven. The oven first depletes all the air from within the chamber, then backfills it with nitrogen and heats the wafer until it is dry. The HMDS causes the surface of the wafer to become hydrophobic. This layer prevents water from penetrating the surface of the wafer much like wax does for a car's paint. Click here to view a video that shows the photoresist process.
A small amount of photoresist is applied to the wafer as it sits still on the wafer
tracks. The name "resist" derives from its use to resist the action of the
etchant. Resists are applied to wafers as thinfilm coatings and then exposed selectively
to an energy pattern (ultra-violet light) that creates exposed areas. The film is then
developed so that it selectively removes either the exposed or the unexposed resist. This
is demonstrated in the following illustration.
The wafer is then rotated very rapidly so that a uniform layer is made from the original
drop of photoresist. The desired thickness is acquired by changing the speed of the
revolutions. The wafer is then placed onto a hot plate to remove all of the solvents
(which would later cause problems by turning into gases during critical photolithography
processes) from the wafer. The wafer is heated on a hot plate generally between 600șC and
1000șC for 20 to 60 seconds. It is then moved down the track to a station where it is
cooled. The cooled wafers are placed into a cassette, which holds the wafers during
transportation for the remaining processes. Click here
to view a video that shows the wafer track system.
The cassettes are inserted into a machine called a stepper. Below is a picture of a
commonly used stepper.
Today the two major photolithography techniques used are contact printing and optical
stepping. The main difference between the two techniques is the location of the mask to
the substrate. In contact printing, the mask is vacuum clamped directly on the wafer for
exposure. There are several obvious disadvantages to this process. The contacting surfaces
may not be totally flat and therefore, cause a leakage of photoresist in areas it is not
wanted. Also, the masks get damaged quickly from the constant contact with the substrate
and cleansing procedures used to clean off the photoresist. This method is still used
because it is much cheaper and works well in low volume applications. For higher volume
applications optical steppers are used because there is no contact between the mask and
the substrate keeping it from wearing down. The optical stepping technique is suited for
GaAsTEK because a more efficient and longer lasting technique is needed for better
production. Unfortunately, a drawback is that the optical stepper machine costs over a
million dollars. Optical steppers, the dominant machine for high yield photolithography,
are available in three different configurations. The pattern on the reticle (mask) can
either be the same size as that on the wafer, five times greater, or even ten times
greater. GaAsTEK uses a reticle which has a pattern that is five times the size as that on
the wafer, and therefore, the UV light that is projected through the mask is also sent
through a reduction lens to reduce the size of the image. (Source: Williams, Ralph, Chapter 2 in Modern
GaAs Processing Methods, Boston: Artech House, Inc., (1990).)
The stepper uses masks, which are glass or quartz plates covered with patterns of the
circuit design that is desired to be projected onto the wafer. Each pattern consists of
opaque and transparent areas that define the size and shape of all circuit and device
elements.
At GaAsTEK, chrome is used to produce the opaque surfaces. Although GaAsTEK does not make
the masks themselves, they are designed and drawn by them using a computeraideddrawing
(CAD) CAD system. The digital representations of the
masks are then sent to a mask fabricator, in this case, Photronix. A thin plastic
cover almost like plastic wrap is used to keep the mask clean of debris. The masks need to
be cleaned regularly in order to assure that no extraneous marks will be etched onto the
wafers.
The stepper uses laser beams to align the alignment marks on the wafer with the
corresponding marks on the mask. This prevents any error due to the slightest
misalignment. The UV light that is sent through the mask and reduction lens is generated
from an ILine mercuryarc lamp. The wavelength of the light is 365 nanometers (nm),
which is an industry standard. Many fabrication plants today are trying to reduce this
wavelength to approximately 250 nm. The reason for the push into using smaller wavelengths
is because smaller wavelengths allow for greater resolutions. A simple formula to
illustrate this point is the following.
R = sin(a) / l
where:
The image of the mask is developed numerous times onto the wafer stepping from one
location to the next (hence, the name of the machine). For each step the process takes
approximately thirty seconds to align the wafer and one minute for the exposure. Below is
a drawing (Fig. A) showing the setup for aligning and exposing the wafer. Also, a second
illustration (Fig. B) shows the relavent areas of photoresist being developed.
The decision for the time length of exposure is dependent on how thick (usually between
0.3 and 2.5 microns) the photoresist layer is. This "stepping" procedure is
repeated until all the required fields are filled with the appropriate pattern. The
stepper can be programmed to change mask sets easily for many different applications. For
GaAsTEK, this means very little down time when changing from one design of integrated
circuit to another design.
A process called etching removes the photoresist. Etching removes the unwanted material from a wafer by chemical, electrolytic or plasma means. GaAsTEK uses wet chemical techniques because it best suites the application of GaAs wafers. The siliconoxynitride is removed with hydrofluoric acid (HF) using a wet etch and the GaAs is also removed in a wet etch process with a mixture of ammonia, peroxide, and water.
The same carbon based liquid resist is used for both positive and negative photoresist.
This feat is successfully done by using two simple procedures. If negative photoresist is
desired then the coated substrate is exposed and then baked. If positive photoresist is
desired then the coated substrate is exposed only. The main difference between the two
photoresists is that the negative exposed area is not removed, while the positive exposed
area is removed. Below is an image that illustrates this process.
The negative photoresist is used to take off excess metal during metallization; for
instance, titungstennitride (TiWN). There are several advantages and disadvantages to
using both positive and negative photoresist. Below is a comparison of characteristics of
negative and positive photoresists:
| Characteristic | Negative Resist | Positive Resist |
| Exposure | Rely upon cross-linking for image formation | No chemical change takes place in resist that forms the image |
| Molecular Weight | High molecular weight products formed during exposure | No molecular weight changes chemical reaction in non-image areas |
| Oxygen sensitivity | Have oxygen sensitivity, causing exposure problems | No oxygen sensitivity |
| Removal | Are difficult to remove, due to high molecular weight | Easy removal, has no high molecular weight products present |
| Developing | Solvent developing results in image swelling. Also, disposal is more difficult | The image is unaffected by the aqueous developer. Disposal is relatively simple |
| Coating thickness and resolution | Coating thickness must be 1/3 the minimum image size | Coating thickness can be equal to or greater than minimum image size |
| Step Coverage | Marginal due to thin coating limitations | Excellent, since thick coatings (2-3 um) can be used |
Photoresist is not stable at high temperatures in a vacuum system therefore; the substrate
is baked with a machine that has a photostabilization system. This stabilizer uses deep
UV rays with short wavelengths to make a crustapproximately 250șC in order to cook the
photoresist thoroughly. Now that the beginnings of the circuit design have been deposited
onto the wafer, the photoresist will be stripped off using the wet etching process. The
wafer then goes to the ion implantation process and then back to the photolithography
steps. The wafers are again primed, coated, aligned and exposed. However, this time the
photoresist is stripped off using an oxygen plasma. Metallization is the next
step for the GaAs wafers. These steps are repeated until the desired circuit is completed
on the substrate.
Once the metallization is complete, a metal "liftoff" procedure is
performed. For the "liftoff" procedure, the resist is applied to the wafer,
exposed, and developed to the desired pattern. Then metal is applied to the wafer by
evaporation methods. A solvent is then used to dissolve the resist, leaving the metal that
is in contact with the wafer. Unlike the etching, the "liftoff" process is
highly sensitive to the edge profile of the patterned resist, as shown below.
Since we are doing this "liftoff" process, we need to use a negative
photoresist to cover the wafer. When the photoresist is dissolved off of the wafer with a
solution, the metal that it is stuck to it is also removed. The negative photoresist gives
an undercut profile, which is better for taking off metal. The profile gives a nice clean
break between the metal on the photoresist and the metal that goes through the openings
where we want it to deposit. The wafer goes into the stepper again. This time, the mask
pattern is different. The openings on the chrome are now where we want to keep the
photoresist instead of where we want to remove it. Now we are left with a negative image
of what was on the chrome mask. This is an illustration on how negative photoresist is
carried out. (Source: Williams, Ralph,
Chapter 2 in Modern GaAs Processing Methods, Boston: Artech House, Inc., (1990).)
After the entire circuit is done, the front side of the wafer is glued onto a heavy sapphire plate with epoxy and the back is ground down. The reason it must be attached to the sapphire is because GaAs is so brittle that in order to grind it, we need to stabilize the wafer so it does not shatter. The wafer is ground down to approximately 100 microns, about 5/6 of the original thickness. The purpose of making the wafer thinner is to enhance the thermal properties of GaAs. The thermal problems arise when the wafer is attached to another material acting as a heat sink, since GaAs is a poor thermal conductor. Reducing the size of the wafer in turn reduces the thermal resistance so that the heat dissipates quicker. All of the excess GaAs material ground off the back of the wafer is disposed of with care by an inhouse filtration system.