Making Precise Things from Imprecise Tools

These machines have a general problem, and that's the refinement of precision. If you built a machine of a certain size, and you said, "Well, next year I want to build one of a smaller size," then you would have a problem: you've only got a certain accuracy in dimensions. The next question is, "How do you make the smaller one when you've only got that much accuracy?" It gets worse. You might say, "I'll use this machine to make the smaller one," but if this machine has wobbly bearings and sloppy pins, how does it make an accurate, beautiful, smaller machine?

As soon as you ask that question, you realize it's a very interesting question. Human beings came onto the earth, and at the beginning of our history, we found sticks and stonesbent sticks and roundish funny stones, nothing very accurate. And here we are today, with beautifully accurate machinesyou can cut and measure some very accurate distances.

How do you get started? How do you get something accurate from nothing? Well, all machinists know what you do. In the case of large machinery, you take the stones, or whatever, and rub them against each other in every which way, until one grinds against the other. If you did that with one pair of stones, they'd get to a position at which, no matter where you put them, they would fit. They would have perfectly matched concave and convex spherical surfaces.

But I don't want spherical surfacesI want flat surfaces. So then you take three stones and grind them in pairs, so that everybody fits with everybody else. It's painstaking and it takes time, but after a while, sure enough, you've got nice flat surfaces. Someday, when you're on a camping trip, and everything gets boring, pick up some stones. Not too hardsomething that can grind away a little bit, such as consolidated or weak sandstones. I used to do this all the time when I was a kid in Boston.

I'd go to work at MIT and on the way pick up two lumps of snow, hard snow that was pushed up by the snowplow and refrozen. I'd grind the snow all the way till I got to MIT, then I could see my beautiful spherical surfaces.

Or, for example, let's say you were making screws to make a lathe. If the screw has irregularities, you could use a nut that's breakable; you would take the nut apart and turn it backwards. If you ran the screw back and forth through the nut, both reversed and straight, soon you would have a perfect screw and a perfect nut, more accurate than the pieces you started with. So it's possible.

I don't think any of these things would work very well with the small machines. Turning things over and reversing and grinding them is so much work, and is so difficult with the hard materials, that I'm not really quite sure how to get increased precision at the very small level.

One way, which isn't very satisfactory, would be to use the electrostatic dielectric push-pull mechanism. If this device were fairly crude in shape, and contained some kind of a point or tooth that was used for a grinder or a marker, you could control the position of the tooth by changing the voltage rather smoothly. You could move it a small fraction of its own irregularity, although you wouldn't really know exactly what that fraction was. I don't know that we're getting much precision this way, but I do think it's possible to make things finer out of things that are cruder.

If you go down far enough in scale, the problem is gone. If I can make something one-half of a percent correct, and the size of the thing is only one hundred atoms wide, then I've got one hundred and not one hundred and one atoms in it, and every part becomes identical. With the finite number of atoms in a small object, at a certain stage, objects can only differ by one atom. That's a finite percentage, and so if you can get reasonably close to the right dimensions, the small objects will be exactly the same.

I thought about casting, which is a good process. You ought to be able to manufacture things at this scale by casting. We don't know of any limitationexcept atomic limitationsto casting accurate figures by making molds for figures that match the originals. We know that already, because we can make replicas of all kinds of biological things by using silicone or acetate castings. The electron microscope pictures that you see are often not of the actual object, but of the casting that you've made. The casting can be done down to any reasonable dimension.

One always looks at biology as a kind of a guide, even though it never invents the wheel, and even though we don't make flapping wings for airplanes because we thought of a better way. That is, biology is a guide, but not a perfect guide. If you are having trouble making smooth-looking movable things out of rather hard materials, you might make sacs of liquid that have electric fields in them and can change their shapes. Of course, you would then be imitating cells we already know about. There are probably some materials that can change their shape under electric fields. Let's say that the viscosity depends on the electric field, and so by applying pressure, and then weakening the material in different places with electric fields, the material would move and bend in various ways. I think it's possible to get motion that way.