nagle@lll-crg.llnl.gov (John Nagle) (08/26/89)
Most of the nanomachines discussed here, or, for that matter, in
Engines of Creation, are far beyond what we can hope to construct for
some time. I suggest that some thought be devoted to the design of
simpler nanomachines, far simpler than an assembler.
One such is a smart etchant. This is a tool for photolithography.
A film of fluid, with etching nanomachines, is laid down on a surface.
A mask is projected on the surface, but instead of being simply projected
while the etchant works, it is used to control which machines receive
messages expressed as patterns of light flashes. The machines respond to
low-level commands, such as "remove a 1-atom layer", "remove a 10 atom
layer", "remove all non-silicon atoms", "deposit metallic atom layer",
and such.
The result is a means of producing ICs with a simple, one-pass
process.
John Nagledoom@portia.stanford.edu (Joseph Brenner) (09/01/89)
>I suggest that some thought be devoted to the design of >simpler nanomachines, far simpler than an assembler. Good idea. >One such is a smart etchant. [...] >The result is a means of producing ICs with a simple, one-pass >process. Not so good, I think. As I remember it, the atomic density of Silicon is on the order of 10^22 atoms per square centimeter. Let's say we're going to make a chip that's about one square centimeter, and assume we need to do at least one operation on every atomic site in this area. How fast can we perform an operation? Does 10,000 times a second sound like a reasonable upper bound? Then, even if you allow 100,000 seconds to make the chip (that's over a day), we'd need 10^13 nanomanipulators to do the job. Now all you need to do is to figure out how we're going to manufacture over a trillion nanomanipulators without something like Drexler's self-replicating assemblers. It has been suggested that a Scanning Tunneling Microscope could be used to perform repairs on photolitho IC masks: maybe you could use some form of your dumb nanomanipulators to write very fine traces on some kind of template, which is then duplicated by some other means. -- Joe Brenner (J.JBRENNER@MACBETH.STANFORD.EDU Materials Science Dept/Stanford, CA 94305)
Joseph.Bates@a.gp.cs.cmu.edu (09/01/89)
I agree with John Nagle that it is very interesting to look at the design and construction of simple nanomachines. It would be great to see sample solutions of some of the interesting problems, such as power distribution and heat dissipation. I wonder, though, if the particular example of a smart etchant is really much simpler than an assembler. Are there "cute mechanical tricks" that let us build these special purpose robots, or do the functions required of the etchant mean that it is best built with general purpose processors and general purpose atomic manipulators? The etchant seems like a rather flexible automaton. How about just a reversible glue that can be turned on or off by some sort of signal applied at its periphery? It seems to me that this would involve a small memory, communications paths, and grasping/ungrasping ability at each nanomachine. If it were to be able to cycle for long periods then power might be a problem, too. Generalizations could include synthetic muscle, where the global control pattern of grasp/pull/ungrasp becomes more complex. I wonder what level of design automation is needed to attack this problem? Are we there yet? Do we need much better tools? Joe Bates