Electrostatic Actuation

Now how do you pull them along? That's not very hardI'll give you a design for pulling. [At the blackboard. Feynman draws a rectangular block with a set of alternating electrodes creating a path for the block.] If you had, for example, any object like a dielectric that could only move in a slot, and you wanted to move the object, then if you had electrodes arranged along the slot, and if you made one of them plus, and another one minus, the field that's generated pulls the dielectric along. When this piece gets to a new location, you change the voltages so that you're always pulling, and these dielectrics go like those wonderful things that they have in the department store. You stick something in the tube, and it goes whshhhht! to where it has to go.

There is another way, perhaps, of building the silicon circuits using these sliding devices. I have decided this new way is no good, but I'll describe it anyway. You have a supply of parts, and a sliding device goes over, picks up a part, carries it to the right place, and puts it inthe sliding devices assemble everything. These devices are all moving, of course, under the electrical control of computer stuff below them, under their surfaces. But this method is not very good compared to the present evaporation technique, because there's one very serious problem. That is, after you put a piece in, you want to make electrical contacts with the other pieces, but it's very difficult to make good contacts. You can't just put them next to each otherthere's no contact. You've got to electrodeposit something or use some such method, but once you start talking about electrochemically depositing something to seal the contact, you might as well make the whole thing the other way by evaporation.

Another question is whether you should use AC or DC to do the pulling: you could work it either way. You could also do the same thing to generate rotations of parts by arranging electrostatic systems for pulling things around a central point. The forces that will move these parts are not big enough to bend anything very much; things are very stiff at this dimensional scale.

If you talk about rotating something, the problem of viscosity becomes fairly important; you'll be somewhat disappointed to discover that if you left the air at normal air pressure in a small hole ten microns big, and then tried to turn something, you'd be able to do it in milliseconds, but not faster. That would be okay for a kit of applications, but it's only milliseconds. The time would be in microseconds, if it weren't for viscous losses.

I enjoy thinking about these things, and you can't stop, no matter how ridiculous things get, so you keep on going. At first, the devices weren't movingthey were in place. Now they can slide back and forth on the surface. Next come the tiny, free-swimming machines.