Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
We know how photoresist works and how to get it on your material. Or my material rather; most of you aren’t going to be running laminators but no matter. How about transferring a pattern to the photoresist? That’s what we’re going over today. To change the photoresist you’ve got to hit it with an ultraviolet photon. Here, let me demonstrate:
I set that out on my cubicle desk several hours before snapping this photo. The small amount of ambient ultraviolet light in the white office light has been bombarding it for hours. Now let’s take all those funny-shaped tools off and see what’s underneath
You can see how the light has turned the resist from a light to a deep blue. Of course, we don’t manufacture circuits this way; it takes several hours to get that effect and I really don’t have the cubicle acreage to spare. To be economically viable we need to get the time down to the seconds range.
Enter the autoexposer. You run your web through the machine about a foot at a time. (“about a foot”? The web is subdivided into panels largely based on how much we can fit into an exposer at a time. Your standard panel is exactly 308 mm long. That number varies in practice by about 0.01 mm. One foot works out to 304.8 mm. A panel is about a foot long.) A modern wafer fab will also work on a step and repeat process. All the steps will be on one wafer; it’ll expose a small part of it (called a ‘die’) and move on to the next. The fact that you have to do these things one after another like that makes it one of the slower processes in any wafer fab.
To get ultraviolet photons in quantity you can use a mercury lamp. Mercury? Yeah. Recall what I said about electrons falling into holes a while back. When an electron drops from one orbital into another it loses energy, which it emits in the form of a photon. How much energy it loses is related to the strength of the positive charge in the nucleus. You can get a relatively high energy photon out of a mercury atom.
Visual light is (broadly speaking) in the 400- to 700-nanometer regime. The longer the wavelength the less energy the individual photon has. The shorter wavelengths are at higher energies. The stuff from about 400 to about 250 nm on the scale there is in the “suntan” region. Much shorter than that you get into ‘show the doctor your bones’ and ‘acquire superpowers’ ranges.
The graph shows you the intensity of light at various wavelengths you’d get out of a mercury bulb. Those spikes correspond with specific orbital transfers in the mercury atom. The other parts (note how the graph doesn’t go to zero everywhere else) come from other effects. Still, you get most of your energy out of the spikes. What kind of processing you’re doing is dependent on what kind of spikes you’re working with. Follow me? Not at all? Yeah, sort of guessed that. Let me tackle it from another angle.
How small are modern transistors? Pretty darn small. You see the 254 nm line on the left of the graph? At least ten times smaller than that. So how are you going to expose things that small? There are ways to generate even harder UV light (an excimer laser can get down further), but frankly, my textbooks are a little old, and a bit vague on these things. As a practical matter I’m working in the micron range; the smallest features we deal with are on the order of ten thousand nanometers wide. Your standard UV light is what I know, and what I can deal with.
A mercury lamp is an arc lamp in a mercury atmosphere. Okay, what’s an arc lamp? An arc lamp generates light by causing electricity to arc from an anode to a cathode. In this case, we’re using the electricity to strip electrons from the mercury. When the mercury reacquires electrons it’ll emit light at predictable wavelengths. These photons are what we use to expose the photoresist. Simple, right?
A mercury bulb has to stand high currents, high temperatures, and high pressures. They’re made out of silica or quartz, and the anodes and cathodes are tungsten. You might remember that as the filament material in your illicit incandescent light bulb. Sticking the quartz to the electrodes such that no atmosphere leaks in or out is an adventure in its own right.
In addition to the suntan, you’d get if you yourself were exposed to one of these lights there are other issues. The mercury plasma will tend to degrade the cathode over time. The tungsten will sputter onto the inside of the silica bulb. For one thing, this reduces your energy output. For another thing, the energy that isn’t output will heat up your bulb. It’s best to switch these things out after the manufacturer-recommended number of hours. They’ll explode if you leave them run too long.
Alright, you’ve got your bulb, it’s safely isolated from your face. How do you get your light from the bulb onto the web?
Around your bulb  you have a reflector  which, what else, reflects the light upwards. Exactly like the one in your flashlight. Above that, you’ve got a dichroic mirror . It passes your light through a columnator  which hopefully gets your light propagating in parallel lines. Just before your columnator, you have your shutters [not pictured]. These mercury bulbs take about 45 minutes from starting up to producing an even and predictable amount of energy out; you don’t want to do that every panel. You leave it running all the time and blink the shutters on and off when you need to expose something.
After the columnator, there’s another reflective mirror . It sends the light up into a parabolic mirror  and then straight down through your phototool and into the web. The phototool is held in place on a granite chuck .
A quick note of math; that parabolic mirror. Why a parabola? The parabola shape has an interesting property. Imagine, if you will, that you’re dropping a ball into a parabolic arc of concrete. The ball drops straight down, then bounces off the curved side. Inevitably it’s going to bounce sideways. If you dropped that ball at a different spot it’d bounce sideways again. If you kept doing this you’d notice that the balls always bounced through the same spot. Eventually. Hey, maybe you’re quicker on the uptake than you look.
That point is the focus of the parabola. Now let’s run it backward. Throw the ball at an angle into the parabola. Throw it such that its path passes through the focus. What does the ball do? It bounces straight up. No matter what angle you hit, as long as you threw it through that focus the ball always bounces straight up. Property of the parabola. You shine your light through the focus of the parabola so that it’s all traveling in a perfectly vertical direction as it goes down through the phototool and hits the web.
I keep using that word, phototool. What is it? It’s another name for the mask you use to print your pattern on the parts. At HTI we do something called contact exposing, which means your phototool is directly on top of the material. Technically speaking that’s 1970’s technology, but again we’re in the 10-micron range not the 10-nanometer range.
The phototool is made of clear quartz. On the side facing the material, we print a pattern in chrome (okay, we order it out. To print one of these you burn it with an electron microscope. Our electron microscope can’t handle plates of the size we need.) Why chrome? It ranks about 8 or 8.5 on the Mohs hardness scale. That means your stainless steel or your granite won’t scratch it. The only thing that will is the diamond on your wedding band. Seen a couple tools scratched that way.
You vacuum your phototool down to a granite chuck in part because it’s so big and massive. You don’t want any stray vibrations getting in and shaking up your exposing. There are oh so many ways that an exposer can screw up. It’s one of the more critical and delicate steps in the whole operation. We’ll go over a number of them next week in “Repeaters and Repeating Defects: How Repeating Repeaters Repeat.” or “No, really, They Said there Would be No Math.
This is part 12 of my ongoing series on How to Build a Computer, the “Panic at the Disco” way. You may find previous parts under the tag How to Build a Computer. This week’s post has been brought to you by Shorty’s Barber Shop. Shorty has been trimming hair since the Korean war, so show the man some respect. Shorty’s Barber Shop!