This page will show the various steps to building up our wafers in ECE 2224. We will spend five or six days in the clean room, and these are pictures from these days. I will update this page as I do each step. You will notice all of the background is yellow. This is because we have placed UV filters on all of the overhead flourescent lights so we do not prematurely expose the photoresist that we will place on the wafers later. The color of the filters is yellow.
What is a silicon wafer?
Our VERY first step is to get in a clean room suit. This consists of a suit with no pockets, a face mask, a hair net, goggles, gloves, and shoe coverings. Everything we take into the clean room must be wiped down with alcohol to knock any loose particles off. The paper and pens we use for notes are "special" clean room safe pens and paper. Apparently, the paper is the most expensive part of our suits. The suits are not so much to protect us from the environment inside as to protect the environment inside from us.
Now that we have been suited up, it's time to actually start production. First, we clean the wafer with acetone. To do this, we turn on a vacuum pump on the spinner, place the wafer inside a spinner. This holds the wafer in place. To get an even amount of a liquid on the wafer, we spin the wafer at a few thousand RPM and spray liquid in the top.
Now that we have cleaned the old SiO2 and all the particulate off the wafer surface, we must find some of its electrical properties, such as resistivity. We do this with a four point probe. We see from the resistivity that Ohm's law isn't 100% accurate, as we do not have a perfectly straight IV curve. The second picture is the wafer specifications that came from Motorola.
We now want to regrow some SiO2 on the surface of the wafer. We will later etch some of this away as we need, but for now, we want it all over the surface of the wafer. To speed up the growing process, we put the wafers in a furnace and heat them to around 1200oC and pump various pure gasses into the furnace. We start with the furnace at around half that value and ramp it up over time because otherwise the thermal gradient across the width of the wafer may vary too sharply, causing it to warp or even break. This process takes a few hours. The clear parts of the furnace are not glass, but rather quartz. The reason behind this is that quartz can be heated very hot, and then dumped in ice water without shattering. The wafer boat, the furnace door, tube, and tools to push the wafer boat in are all quartz.
After a few hours in a furnace, the oxidation shoud be thick enough to work with. In our case, we came back two weeks later (the TA removed the wafer after the appropriate amount of time). The first thing we did was check to see how thick the oxide really was. We did this with an ellipsometer, which emits a beam of light and sees how it is defracted. Our wafers measured around 5500 Angstroms.
The next step is to apply photoresist. To apply it evenly, we place the wafer on the spinner and drop the photoresist in, hoping it will be slung into an even layer. Unfortunately, the nice spinner was down, so we had to use a less-than great spinner. In fact, the vacuum pump for this one was too strong for our wafers, and bowed them slightly in the middle. This is part of the reason our wafers are multi-colored (red-green) instead of just an even color.
The photoresist is mostly dry now, but just to dry it a little more, we stick it on a hotplate at around 100oC for around 90 seconds.
The photoresist is completely dry now, so we want to put our first mask on the wafer. This process is known as photolithography. We are using positive photoresist, so we basically cover portions of the SiO2 we want to "keep," then expose the wafer to UV light. We exposed ours for 4 minutes. In the industry, we'd use a glass mask with chrome covering. We used a transparency with black inkjet ink. For our purposes and the size devices we are making, it should do fine.
Now for another slow part that was cool for the first 30 seconds. We place the wafer in a container of developer solution. This removes the exposed photoresist. We begin to see some of our devices...
We spin the wafer dry and put it in hyrdofloric acid to etch away at the remaining silicon dioxide. The acid part was pretty boring to watch, so I don't have any pictures of that. We then rinse the wafer with deionized water. As the second photo shows, not all of our devices turned out properly. This is probably due to thicker photoresist and the mediocre spinner. Good thing we aren't graded on whether or not our wafer works...
Here are a few close-up shots of the diffusion etch mask we did in the last lab.
Another few shots of our wafer.
The next step in our wafer development is to make the contacts out of aluminum. Rather than try to dab little bits of aluminum all over the wafer, we will cover the entire wafer in aluminum and then etch away what we do not want. We will do this by using an aluminum evaporator, where we lower the pressure and evaporate Renyolds Wrap onto the wafers (and everything else in the chamber, including the viewing door). Basically, aluminum foil is placed on a tungsten coil, the pressure is dropped via a roughing pump, high current is passed through the coil to get I2R heating. The coil glows white hot and the aluminum evaporates in less than a second. The pressure is raised and the wafers retrieved. As you can see, the layer of aluminum is uniform (since it is propgating more or less isotropically from a point on the coil) and produces a very reflective surface.
Now, the wafer is totally covered in aluminum. We run through the drill of spinning photoresist on the wafer, exposing it, developing it, and then placing it in aluminum etch. Then we have a finished wafer ready for testing.
