Small MachinesHow to Use Them

What use would such things be? Now it gets embarrassing. I tried very hard to think of a use that sounded sensibleor semisensibleyou'll have to judge. If you had a closed area and a half wheel that you turned underneath, you could open and shut a hole to let light through or shut it out. And so you have light valves. But because these tiny valves could be placed all over an area, you could make a gate that would let through patterns of light. You could quickly change these patterns by means of electrical voltages, so that you could make a series of pictures. Or, you could use the valves to control an intense source of light and project pictures that vary rapidlytelevision pictures. I don't think projecting television pictures has any use, though, except to sell more television pictures or something like that. I don't consider that a useadvertising toilet paper.

At first I couldn't think of much more than that, but there are a number of possibilities. For example, if you had little rollers on a surface, you could clean off dirt whenever it fell, and could keep the surface clean all the time.

Then you might think of using these devicesif they had needles sticking outas a drill, for grinding a surface. That's a very bad idea, as far as I can tell, for several reasons. First, it turns out that materials are too hard when they are dimensioned at this small scale. You find that everything is very stiff, and the grinder has a heck of a job trying to grind anything. There's an awful lot of force, and the grinder would probably grind down its own face before it ground anything else. Also, this particular idea doesn't use the individualization that is possible with small machinesyou can individually localize which one is turning which way. If I make all the small machines do grinding, I've done nothing I can't do with a big grinding wheel. What's nice about these machinesif they're worth anythingis that you can wire them to move different parts differently at different times.

One application, although I don't know how to use it, would be to test the circuits in a computer that is being manufactured. It would be nice if we could go in and make contacts at different places inside the circuit. The right way to do that is to design ahead of time places where you could make contacts and bring them out. But if you forgot to design ahead, it would be convenient to have a face with prongs that you could bring up. The small machines would move their little prongs out to touch and make contact in different places.

What about using these things for tools? After all, you could drill holes. But drilling holes has the same problemthe materials are hard, so you'll have to drill holes in soft material.

Well, maybe we can use these tools for constructing those silicon devices. We have a nifty way of doing it now, by evaporating layers, and you might say, "Don't bother me." You're probably right, but I'd like to suggest something that may or may not be a good idea.

Suppose we use the small machines as adjustable masks for controlling the evaporation process. If I could open and close these masks mechanically, and if I had a source of some sort of atoms behind, then I could evaporate those atoms through the holes. Then I could change the hole by changing the voltagesin order to change the mask and put a new one on for the next layer.

At the present time, it is a painstaking job to draw all the masks for all the different layersvery, very carefullyand then to line the masks up to be projected. When you're finished with one layer you take that layer off and put it in a bath with etch in it; then you put the next layer on, adjust it, go crazy, evaporate, and so on. And that way, we can make four to five layers. If we try to make four hundred layers, too many errors accumulate; it's very, very difficult, and it takes entirely too long.

Is it possible that we could make the surfaces quickly? The key is to put the mask next to the device, not to project it by light. Then we don't have the limitations of light. So you put this machine right up against the silicon, open and close holes, and let stuff come through. Right away you see the problem. The back end of this machine is going to accumulate goop that's evaporating against it, and everything is going to get stuck.

Well then, you haven't thought it through. You should have a thicker machine with tubes and pipes that brings in chemicals. Tubes with controllable valvesall very tiny. What I want is to build in three dimensions by squirting the various substances from different holes that are electrically controlled, and by rapidly working my way back and doing layer after layer, I make a three-dimensional pattern.

Notice that the silicon devices are all two-dimensional. We've gone very far in the development of computing devices, in building these two-dimensional things. They're essentially flat; they have at most three or four layers. Everyone who works with computing machinery has learned to appreciate Rent's law, which says how many wires you need to make how many connections to how many devices. The number of wires goes up as the 2.5 power of the number of devices. If you think a while, you'll find that's a little bit too big for a surfaceyou can put so many devices on a surface, but you can't get the wires out. In other words, after a while this two-dimensional circuit becomes all wires and no devices, practically.

If you've ever tried to trace lines in two dimensions to make a circuit, you can see that if you're only allowed one or two levels of crossover, the circuit's going to be a mess to design. But if you have three-dimensional space available, so that you can have connections up and down to the transistors, in depth as well as horizontally, then the entire design problem of the wires and everything else becomes very easy. In fact, there's more than enough space. There's no doubt in my mind that the ultimate development of computing machines will end up with the development of a technologyI don't mean my technology, with my crazy machinesbut some technology for building up three-dimensional circuits, instead of just two-dimensional circuits. That is to say, thick layers, with many, many layershundreds and hundreds of them.

So we have to go to three dimensions somehow, maybe with tubes and valves controlled at small scale by machines. Of course, if this did turn out to be useful, then we'd have to make the machines, and they would have to be three-dimensional, too. So we'd have to use the machines to make more machines.

The particular machines I have described so far were just loose pieces that were moving in placedrills, valves, and so forth that only operate in place. Another interesting idea might be to move something over a surface or from one place to another. For example, you could build the same idea that we talked about before, but the thingsthe little bars or somethingare in slots, and they can slide or move all over the surface. Maybe there's some kind of T-shaped slot they come to, and then they can go up and down. Instead of trying to leave the parts in one place, maybe we can move them around on rollers, or simply have them slide.