Yesterday we looked at CRN's recommended study of sensing, manipulation, and fabrication tools for working at the nanoscale. There are at least three plausible ways to begin building from the bottom up: nucleic acid, biological self-assembly, and diamondoid. (Others have been suggested as well.) Each of these areas should be studied separately, both for manufacturing potential and for development requirements.
Here is an outline for an exploration of the various steps needed to develop a complete manufacturing system based on diamondoid vacuum mechanosynthesis. This is CRN's proposed study #8: "What will be required to develop diamondoid machine-phase chemical manufacturing and products?"
Subquestion A: How much computer time and human creativity would it take to invent, then simulate and verify a set of diamondoid-building (and/or graphene-building) reactions?
Preliminary answer: Robert Freitas has proposed a $5 million, five-year project to do just that; the project would also simulate the construction of nanodevices using these reactions.
Subquestion B: What will be involved in developing a non-diamondoid manipulation system that can carry out the required manipulations to build the first system?
Preliminary answer: Unknown, but it should be noted that we can now lithographically fabricate features that are smaller than the molecules we can engineer. In other words, we can build pretty much any shape at any size scale.
Subquestion C: How reliably can the operation of diamondoid machine parts be simulated? What would be the cost and development time of a CAD/simulation system capable of extracting mechanical characterization from molecular dynamics simulation of such parts?
Preliminary answer: Unknown, but this is a much easier problem than characterizing proteins: the parts involved are much stiffer, and energetic computations can afford to be much less accurate. Hydrocarbon MM packages have been around for years (e.g. Brenner) and are now appearing in open source software (e.g. NanoHive).
Subquestion D: How many parts and surfaces would be needed to constitute a complete set of low-level structural and functional components? How much human effort would be required to develop them?
Preliminary answer: Unknown. Low-level components include rotational, helical, and flat bearings; conductive and insulating components; molecular interfaces between different surfaces and crystal orientations. Note that Freitas expects to design at least some working components as part of his $5 million proposal.
Subquestion E: What would be the cost and development time of a CAD/simulation/tracking system that could support the design of machines and systems from low-level components?
Preliminary answer: Unknown. Probably comparable to high-end software design tools, or semiconductor design tools circa 1990. It wouldn't have to handle a lot of different parts or physics, at least in early versions where performance can be sacrificed to reduce undesired interactions between parts.
Subquestion F: What would be the cost of developing a design for an integrated, hierarchical manufacturing system to build large products?
Preliminary answer: An architecture for such a design has been worked out. The molecular fabrication in that design is based on a simple robotic-chemistry design by Ralph Merkle. Many fabricators make parts in parallel, and the parts are then combined via convergent assembly. Merkle's design requires perhaps 100 moving parts and half a billion atoms (most of which don't have to be individually specified). Convergent assembly appears to require only simple robotics at several scales. Assembly and fabrication appear to require only simple control software. Much of the engineering, even at nanometer scales, will be more or less familiar to mechanical engineers. Overall engineering difficulty might be comparable to an aerospace project.
Subquestion G: How many of these steps could be accomplished concurrently in a crash program?
Preliminary answer: All of these steps could be started concurrently, with successive refinement. This may not happen due to caution on the part of the funders. However, a funding organization that was willing to fund a crash program could probably do all these steps in parallel.
Subquestion H: How precisely can costs and schedules be estimated?
Preliminary answer: Due to lack of study, very little information is available. For the sub-projects that we can estimate, the cost is consistently under $1 billion, and several appear to cost just a few million. Also, all of them (with the exception of software engineering, which should not be a major fraction of the total cost) appear to be getting easier rapidly. We can't rule out the possibility that the whole thing might cost less than $1 billion; in fact, that appears likely to us, though we don't say it loudly because it sounds too implausible. A project starting five or ten years from now very likely would find the cost greatly reduced. (However, other studies indicate that this is not a sufficient reason to delay; it's simply evidence that if we do delay, a rapidly increasing set of organizations will be able to do it.)
About schedules, again, very little information is available. The argument parallels the cost discussion. The project can be divided cleanly into sub-projects. In the areas where we can make estimates for the sub-projects, the estimates are surprisingly short. We don't see any sub-project that needs to take more than five years. Doing all sub-projects in parallel would require excellent management, visionary funding, and good communication to ensure smooth integration. But this appears feasible, and implies that the whole thing might be done in five years with sufficient effort and skill. (But government bureaucracy is not well suited to do this.)
Provisional conclusion: At a guess, the difficulty and schedule of developing a tabletop kg-scale manufacturing system producing kg-scale nano-featured products may be comparable to the Apollo Program. Or it may be quite a bit easier; we can't know without more engineering investigation. At this point, we can't rule out the possibility that it could be done in five years for less than $1 billion. Note also that work on this may have already started somewhere, and may be quite close to completion.
CRN's initial basic findings (preliminary answers and provisional conclusions) for all thirty studies should be verified as rapidly as possible. Because our understanding points to a crisis, a parallel process of conducting the studies is strongly preferred.
We are actively looking for researchers who have an interest in performing or assisting with this work. Please contact CRN Research Director Chris Phoenix if you would like more information or if you have comments on the proposed studies.