First you gotta find a dealer. Right, not that kind of doping. Today we’re going to discuss how to how you mix your dopant atoms into your silicon wafer so you can make transistors.
How Do I Dope My Wafer?
The easiest way is to do it while you’re still making it. Going back to wafer production, look at the Czorchalski (Cho RAL ski) method. Let’s say you mixed some phosphorus into the molten silicon. It’ll diffuse through your melt and get pulled up into the forming boule. Some small percentage of your crystal will consist of phosphorus atoms.
It’s harder to get good uniformity with the float-zone method. One thing you can do is include the dopant material in the core of the silicon before you start melting it together. The dopant will mix as you melt and reform your boule bit by bit. Another way, and this is by far my favorite bit about the entire subject is that you can shoot it with radiation.
Silicon comes in three natural isotopes: Si28, Si29, Si30. Shooting an atom with a neutron adds one to the atomic mass, but doesn’t change the atomic number. You shoot the two lighter isotopes and not much happens; it becomes a slightly heavier isotopes. If you shoot one of the heavies though, you’ll transmute it into Si31, which is radioactive. The newer, heavier silicon will eventually split off a beta particle. That increases the charge of the nucleus by one and the silicon becomes (stable) phosphorus. Beta radiation is relatively easy to block, and none of the rest of your silicon is going to stay radioactive. Here’s the nuclear equations, first the neutron capture then the radioactive decay:
Si30 + η — > Si31 (radioactive)
Si31 — > P31 + β (decay)
The half-life of Si 31 is 2.62 hours. That’s relatively short; you’re going to have to let your wafers ‘cool’ for a day or two before they stop producing a detectable amount of beta particles. Annoying, but it doesn’t require a mountain in Utah.
You hear that medieval alchemists? We’re transmuting base elements into, well, other base elements. Even so! We’re doing the kind of stuff that you could only manage in your fever dreams! Choke on that!
*Ahem*. Where what I? Right, doping. All those methods give you dopant for the bulk of your silicon wafer. There’s one more general case I’d like to discuss before we get to doping specific areas (which is necessary to build transistors).
Once you’ve got your shiny new wafer, and before you do anything else to it, you can grow an epitaxial layer. Because you’re starting with a perfect silicon crystal (perfect-ish; we covered that last week) you can grow a thin layer of also-perfect(ish) crystal on top. Importantly, the epitaxial layer doesn’t have to have the same dopant profile as the rest of the wafer. You can have an n-type layer atop a p-type wafer, or even two different concentrations of the same type of dopant.
Okay, you’ve got your wafer, it’s doped one way or another. To do something useful with that you’re going to also have to apply dopants to specific regions. You only get diodes and transistors once you stick an n-doped region up against a p-doped region. Okay, so how do I change the dopant profile of a specific part of my, er, part?
There’s two general methods. The older-fashioned one is called bulk diffusion. You put a whole bunch of another element on top of your silicon and some will diffuse in. That’s why we went over all that crystallography last week. You need to see that in the form of cookies? Fine.
Things diffuse easier if the temperature is high. The way this typically works you put your wafers in an oven, you pump in your dopant as some kind of gas (say, borane — B2H6 — for p-doping), and you heat up your wafer. The temperature as you go through the whole oven cycle will look something like this:
No preheating; we’re talking hundreds of degrees C. (Also hundreds of degrees Freedom, just more of them.) The exact temperatures depend on the dopant you’re using and the amount of diffusing you want it to do. In the first flat part of the profile you’re diffusing material into the silicon, and it tends to diffuse as interstitial defects. In the second, hotter portion you’re driving it in; you require the hotter temperature to stick your interstitial dopant into the crystal matrix.
People generally don’t use bulk diffusion anymore. It’s not accurate enough. The smaller you make your transistors (and that’s pretty darn small these days) the more accurate you need it to be. You know how you have to heat it up to get stuff to diffuse into the silicon? What happens if you have to do more than one operation like that? What if you’re doing something else that requires those thousand C temperatures? Any dopants you’ve already got in there are going to keep diffusing. To get more accurate dopant placement, these days people tend to shoot their wafers with something called an Ion Implanter. It’s also used to menace meddling superheros.
An ion is an electrically charged atom. To get that you make a plasma (“Oh sure,” I hear you saying “you just make one. What, do you rub two sticks together?” I’ll cover plasma-making in a future episode. Sorry to make you wait.) You make a plasma of free electrons in a high vacuum (Vacuums, both high and otherwise, will also be covered in a future episode). Can I finish my sentence now? Youmakeaplasmaoffreeelectronsinahighvaccuumwhichyouusetoionizeyourdopantspecies. Hah! Got you there. Let’s say you want to implant boron into your wafer. You float some borane gas into your electron plasma. The plasma is going to break up your borane molecules and strip electrons off of your atoms. That gives you your ions.
Now you’ve got a plasma of free electrons, BH3, BH2+, BH++, B2H5+, B2H4++, and more unmentionable things. You let your borane filter into the chamber from a slit on one side, and your ions float out through a slit on the other side. Now you accelerate them. Because it’s got a charge you’ve got a handle on it that you can pull. Your ion generator is going to have a positive charge on it’s outer shell (using it as the cathode for plasma creation). That’ll repel your (positive) ion species. You can use a negatively charged electrode to accelerate the ions in that direction, and another positively charged electrode to focus your ion beam. Good to go so far.
We only want one of those though; BH2+. How do we get that? Run your beam of ions through a magnetic field. Charged particles in a magnetic field will curve based on the mass of that particle. You can calibrate your magnetic field to only allow one species of ion to pass. BH++ is too light and too charged, it’ll curve too much and miss on one side. B2H5+ is too heavy; it won’t curve enough and hit on the other side. Here’s a diagram of an ion implanter:
No, wait, that’s the wrong thing. Let me try again:
You accelerate your atoms to a specific energy using a linear accelerator, like the first stage acceleration for a supercollider. Using that specific energy you shoot them into the wafer and you therefore precisely control the depth they sink to. Neat! Of course it’s more complicated than that; there are ways to tune your ion beam, and sometimes you raster it across your wafers like a old cathode ray tube TV, sometimes you move your wafers and let different parts of them hit the ion beam that way. But again I’m going long, so I’ll cover just one more point about this process.
When you blast your ions into a wafer this way it takes quite a while for them to slow down. It’s like breaking in pool; the implanted ion comes hurtling into the crystal with enough velocity to hurl those silicon atoms around all over the place. It’ll bounce off some, deflect of others, knock atoms out of their place in the lattice, and generally collide with something like two hundred other atoms before it stops. This leaves your crystal in a mess, and you’ve only got five minutes to clean it up before your mother gets here!
You clean it by running it through an annealing cycle. Heat the wafer up enough to reform the crystal. Heat it enough to drive in your dopants too; odds are your ions didn’t settle onto a lattice site. And do it all quickly so that your dopants don’t diffuse very far.
Got all that? Good. But what good is it? We can all draw our names in silicon now like a boy in snow, but what useful things can we do with it? For the answer to that question join us next week for “Why Ohio Diodes” or “Transistor? He’s just my Brother”.
This is part three of my ongoing series on building a computer, the A-Team way. You may find previous parts here: 1 (silicon) 2 (crystallography) or all of them under the tag How to Build a Computer. This week’s post has been brought to you by Mr. T. Mr. T pities the fool that pays full price for a computer that he can build himself. Don’t be a jive-sucka!