Google Custom Search

IMM Report Number 21

In conjunction with Foresight Update 43

Self-Replication and Pathways to Molecular Nanotechnology

by J. Storrs Hall, Research Fellow, Institute for Molecular Manufacturing

J. Storrs Hall, PhD
J. Storrs Hall, PhD

The extent of investor interest in nanotechnology has increased recently. Nanotechnology holds out a host of profitable opportunities, but the big bonanza remains a full molecular manufacturing capability as envisioned by Drexler in Nanosystems and elsewhere.

This capability would depend heavily on two aspects of the technology — molecular scale manipulation, specifically positional synthesis; and a self-replicating architecture.

The first part is the subject of most current nanoscience efforts. The second is examined less often, so let us do so now. A self-replicating technology does not necessarily mean a specific self-reproducing machine. It more generally means a set of manufacturing capital equipment which includes as a subset of its output everything that is necessary to make more of (each kind of) itself. Indeed no industrial base could survive without this property for the economy as a whole.

The smaller you can make the basic kernel of your technology that has this property, and the tighter the feedback loops, the less expensive its products will be. The capital doubling times for molecular manufacturing should be on the order of hours, compared to years for current technology. This would decrease the price of its products drastically.

What good does it do us to know this, if we propose to invest in nanotechnology? It means that there are more pathways than one might think. Instead of getting small and then becoming self-replicating, one might consider becoming self-replicating and then getting small. This is implicitly what Feynman described in his scheme of the succession of ever-smaller machine shops.

It’s worth pointing out that Zyvex Corporation appears to be taking an interesting hybrid approach — first going to an intermediate scale (MEMS), then going self-replicating, and then presumably proceeding to molecular scale (Some of this work was presented at the Eighth Foresight Conference — see coverage elsewhere in this issue). More power to them!

If only to delineate what the “other side of the envelope” might look like, though, let’s consider a pure Feynman-style pathway. Essentially we would build a self-replicating system in whatever macro-scale technology we could, and then set it to copying itself recursively, but just a little smaller each time.
This is not as simple as it sounds, or it would have been done long ago. One must have a design that is scale-invariant, or a way to change designs as scaling laws invalidate techniques. One of the advantages of this capability-first/size-second approach is that when scale-affected techniques do give out, you have a broadly capable system you can use to experiment with alternative techniques.

The system must be able to do sensing, handling, and assembly as well as fabrication, since those operations cannot be done by a human operator as at the micro scale.

This means that the system is going to be much more complex than a mere machine shop that can make all of its own parts; it is going to have to be able to make fairly sophisticated robotics. This is a substantial task, but far from impossible. Robots of the requisite capabilities are commercially available today — but they are enormously complex, which would require a hugely complex manufacturing system.

The trick, of course, is to design robots as simply as possible, to reduce the complexity of the system. While we’re at it, we need to pick the techniques we’re using in the knowledge that they have to scale down as far as possible. An example is to build robots that operate by feel instead of vision, even though that may be slower and less efficient at macroscale: at nanoscales, vision won’t work but feel does. Electrostatic motors work at nanoscales but electromagnetic ones don’t. We are expecting to use depositional techniques for fabrication at small scales; use them instead of subtractive ones to start with.

The problem with a machine that tries to build a copy of itself is that all the design tradeoffs that help you make precision parts tend to force you to build parts smaller than the manufacturing machine. One way around this, described in my paper [1], involves a two-phase system which has big fabricators that make small parts, and assembly robots (as in the picture) that put the parts together into fabricators and robots. My paper describes a number of techniques which should simplify design at the molecular scale, but the architecture itself is scale-independent.

proposed assembly robot

So how many steps would it take to get to molecular scale? The first thing to realize is that you have reached molecular scale not when your parts are the size of atoms, but when your tolerances are. A typical fine machining accuracy for most of the twentieth century was a ten-thousandth of an inch (2.5 microns). If we can halve the tolerance at each stage, we are only 12 or so steps from molecular accuracy. If it takes a year for each one… !


[1] Hall, J. S. (1999): “Architectural considerations for self-replicating manufacturing systems”, Nanotechnology 10(3) pp. 323-330

Dr. J. Storrs Hall is an IMM Research Fellow. He can be reached at