The Zen of Device Making
When you watch someone make a device you quickly find out that there are many steps involved and many ways to do each step. If you ask “what’s the difference?” there is often no definite answer. If the difference between two methods is unknown my instinct is to try them both. I often do this when I’m cooking or fixing my bicycle. Usually both ways work, but one method is better. This helps me learn about the system and helps me find the most efficient methods. However, applying the try-every-way philosophy to device making has caused me some grief. It is easy to spend too much time trying to understand the mysteries of device making. After a couple of months you’d rather be advancing mankind’s understanding of physics (or climbing). I hope that these stories help put device making in the right context so you can get on with the physics as soon as possible.
The following are some everyday steps that can be done in multiple ways: storing your wafer, dicing your wafer, keeping track of chips, doing the initial clean, growing tubes, developing e-beam resist, lifting off resist, cleaning off resist residues, wirebonding. When you do any of these steps START WITH SOMEONE ELSE’S METHOD. You have to notice everything when you observe a new method. Did he use glass, teflon or plastic beakers? Did he use pointy or rounded tweezers, stainless steel or plastic? What settings did he use on the wirebonder? Answers might be: glass beakers are used because isopropanol dissolves plastic beakers and the dissolved plastic leaves a residue on the oxide. Rounded tweezers are used because chips often flip out of pointy tweezers. Plastic tweezers were used because metal reacts with hydrofluoric acid. We use this setting because the other settings have broken the oxide layer.
If you note down everything you’ll have a Good Recipe. But you have to do the entire process more than once to be sure you’ve written down a Good Recipe. I thought I had figured out how to scatter catalyst particles across a chip: 1) Clean the chip in Piranha solution 2) Oxygen plasma the chip 3) Sonicate the methanol-catalyst suspension for 1 minute 4) Cover the chip with a droplet of methanol-catalyst suspension 5) Spin at 3000 rpm until methanol evaporates. But the process was flaky, the recipe only worked sometimes. Often catalyst would clump together. Sometimes there was strange crap deposited. It was not a Good Recipe. To make it a Good Recipe I’ve added some extra instructions: -Cleaning and deposition must be done on the same day (perhaps dust accumulates on the surface and nucleates methanol beading). -Catalyst must be stored in upright vials (perhaps crap came from the vial’s plastic seal inside the cap). I haven’t spent time trying to understand these extra instructions. They are plausible and the recipe works. I can get on with making a device.
A very frustrating aspect of device making is that the last step of a process may go to hell if you change Step 1. Here’s an example. My ‘initial cleaning’ procedure is 1) Dip the chip in Piranha solution. 2) Put the chip in an oxygen plasma. A week ago I stopped doing the oxygen plasma because it seemed superfluous, the chip was already clean after the Piranha solution. Everything went fine until growing cvd tubes. No tubes grew. Perhaps it was nothing to do with skipping the plasma. Or maybe oxygen plasma effects the oxide surface in a way that helps tubes grow. With enough effort such mysteries can be figured out. But don’t bother! If a recipe works you don’t need to figure out why something else doesn’t work. Make a device instead.
I have wondered why device making is so different from cooking or fixing a bicycle. Here’s how I see it: 1) The process takes days to complete so cause and effect is hard to determine. 2) It’s hard to figure out what is going on at the oxide surface. 3) “Seeing” a device takes time. There’s only so many times I want to focus an electron microscope, wait for an AFM scan or set up the probe station.
Baking scones, and then testing them, is much easier.
//Ethan, Nov 2000//
