An interesting article on the Better Humans website, Unraveling the Big Debate over Small Machines, quotes me, and adds that my position on nanotechnology isn’t very different to Drexler’s. This is at first sight rather puzzling since my recent article in Physics World, The Future of Nanotechnology, and indeed my book Soft Machines, have been read by many people, including Drexler himself, as attacks on the Drexlerian position. Indeed, I would say myself that my views are actually pretty similar to those of MNT arch-sceptic George Whitesides, though I possibly express them a bit more politely, and with a little less self-confidence.
But on reflection, I find this rather a welcome perception. Perhaps it does mean that a space is growing on both sides of the debate for some rather more nuanced positions than we’ve seen in the past. The Better Humans article gives a lot of attention to the Drexler-Smalley debate. It seems to me that we need to move on from this. MNT sceptics need to recognise that Smalley did not deliver the knock-out punch that they were hoping for. This was brought home to me in Santa Barbara this week in a conversation with an old friend who teaches a sophomore class in nanotechnology at the University of Pennsylvania. She’d set her class the task of studying the debate and deciding which side they thought had prevailed; an overwhelming majority favoured Drexler. So a reasonable sample of educated and intelligent young people was not convinced by Smalley. On the other hand, I think that MNT devotees are wrong to think that this means there are now no rational grounds for scepticism about MNT. While the possibility of some kind of radical nanotechnology is proved by the existence of biological nanomachines, the question of what the best approach to making synthetic nanomachines is is by no means decided. My book Soft Machines argues that MNT has many more disadvantages and potential difficulties than some of its supporters admit, and it will be interesting to see whether its arguments prove more convincing than Smalley’s.
[…] horthand expression of the scientific case against MNT. On the other hand, as I discussed below in the context of the Betterhumans article, I don’t think that the now-famous exc […]
Richard,
I think that you are completely correct, politeness and a little humility should go a very long way to improving this debate and gaining a real understanding of how to design at the nano-level.
I think that your contribution to the debate is to point out the importance of the environment that the nano-machines operate in.
I think that you are correct, the Diamondiod design strategy will only work in a vacuum, but I think that you are wrong when you say it will only work at 3K. I think the right design should be able to work up to several hundred degrees K. One of the good points of Diamondiod designs is they are, computationally speaking, much easier to model compared to soft systems and that should allow us to throw out a whole lot of designs before we build them experimentally.
The soft, warm, and wet approach (man, that sounds sexy) to nano-machines does require a completely different approach. The really big advantage to this approach is you get to steal designs from nature.
* side note * following your lead, on focusing on the nano-environment, three phase nano-systems ( liquid, gas, solid ) might be a quite useful system. The gas phase may be able to provide a compressable / expandable surface that can provide some order to molecules at the gas / liquid interface.
Jim, the main issue here is the degree to which Brownian motion compromises the effective dimensional tolerances that we take for granted in macroscopic mechanical engineering. I think this will lead to difficulties at 300K. But you are right, this is easy to model and I would like to see such modelling done for a system of moderate complexity (a gear-box, say).
Richard,
Would a computer simulation of a 300 k diamondiod gear box that fits inside a cube 100 nm edge length be small enough to convince you that thermal noise and proportionally looser engineering tolerances are design problems that can be over come with the diamiondiod approach? (if it is possible to design a workable gear box at all)
Jim, that’s just the sort of thing I’d have in mind. Or how about asking, how small could you make a watch escapement?
Hello, Richard! Thanks again for your participation in my article. Your comments were an important contribution, I think. The point I was driving at in the article, comparing you with Drexler, is that you both seem to appreciate that there are many ways molecular manufacturing might be realized. Drexler’s mechanosynthesis and your soft-machines are just two of those possibilities. What’s important to the discussion, I think, is that you offer skepticism without a wholesale rejection of the realization of molecular manufacturing, as Drexler proposes. For, as you stated yourself, it’s not that Drexler’s proposal might not work, rather you believe it may not be the most efficient means to the end. This sort of attitude is healthy skepticism, for it lends itself open to new and better evidence. So, even if you don’t agree with Drexler on the details, what’s important is that you both share a “many means” approach (in terms of possibilities), even though you both champion very specific proposals.
Richard, what do you think of the planetary gear? It has been simulated, and at reasonable speeds and room temperature it works just fine.
Would you care to comment on the analysis of diamondoid mechanical properties in Nanosystems, particularly Chapter 5: Positional Uncertainty? Does it seem to you to be an adequate basis for calculating the mechanical stiffness of e.g. Drexler’s robot-arm design (section 13.4) or Merkle’s analysis of strut-based positional devices? Do you think that this method of calculating the stiffness allows calculation of positional variance vs. temperature? And are you willing to accept that the plans for mechanosynthesis have taken this calculated variance into account?
Where’s the problem?
Chris
Chris, I think chapter 5 of Nanosystems is just fine as a basis for calculating the effect of Brownian motion on mechanical systems. My complaint is twofold – firstly, the calculations of the effects of Brownian motion have simply not been done for systems of adequate complexity. I’m having this conversation with Drexler by email at the moment as well, and his best counterexample is also the robot-arm design. I think this analysis as far as it goes is fine, it’s just a system that has no interlocking parts and is absurdly simple, compared say to a typical everyday macroscopic piece of mechanical engineering like an internal combustion engine. Which brings me to my second point; if the sums have not yet been done for nano-systems, then we have to appeal to our experience of macroscopic mechanical engineering, where typical tolerances are much higher than those anticipated in MNT at room temperature – the mechanical workshop in my department can produce pieces with tolerances of 1/10 of a thou without trying very hard, and significantly this level of precision was already attainable in the mid-19th century (and this was a major facilitating factor in the industrial revolution). Let’s be clear – the equivalent in MNT of machining a 1 inch piece to 1/10 of a thou accuracy is having a positional uncertainty in a 10 nm nano-component of 0.01 ?Ö. It’s a tall order!
As for the planetary gear, I note that on the Foresight website J. Storrs Hall’s new Stage 2 project is described with these words:
“In the ’90’s a group at Caltech did some MD simulations of Drexler and Merkle’s planetary gear‚Äîsurprisingly, the only simulations of these nanomechanisms done to date. However, because of the amount of CPU time available and high computational complexity of the algorithms they were using, they tried to see too much action with too little time. The result was to simulate the gear at something like 1000 times its designed speed, with predictable results.” That’s why Foresight seems to think it’s so important to return to this work. I think they’re right.
So, my answer to your question, “And are you willing to accept that the plans for mechanosynthesis have taken this calculated variance into account?” is no, and this answer seems to be shared by the Foresight Institute.
Figures 13.14 and 13.15 show cylindrical parts overlapping and held in place by circular ridges. What do you mean by “interlocking”?
Positional uncertainty of 0.01 A in machine components is not a tall order, it’s a red herring. With relatively soft material properties and no wear penalty for flexing, you can use lower tolerances. Make the inner part slightly larger than the outer part, so that the outer part stretches a bit. And if you need small bumps, you can use atoms.
You confuse “take into account” with “verify for a particular design.” Obviously, more things have been proposed than have been simulated. But that has little to do with whether the problem has been ignored or addressed. In Drexler’s and Merkle’s designs, preserving adequate stiffness in the face of thermal noise has always been a key guideline. No atomic-level simulation-verified designs exist for the strut of a Stewart platform. But we can calculate pretty accurately a safe diameter for that strut, and this has been done.
Chris
If you haven’t done the maths, you can’t claim to understand the impact of a potential problem on your designs. Without the maths, you’re relying on blind faith.
Your assignment for today is to calculate the interfacial stress generated for a given degree of misfit between the inner and outer part of a bearing, and quantitatively test your assertion that you are able to use poorer tolerances without penalties of additional friction or wear due to uncontrolled mechanochemistry.
Trick question, since you won’t accept any assertion I would want to make about friction.
I note from Fig. 3.9 of Nanosystems that changing r by 0.1 A changes force by about a factor of 2. So it would appear that it’s important in systems that hold sliding surfaces together stiffly to get the spacing right. But… from Figure 10.12 of Nanosystems, it seems that the gap can vary by 0.6 A without changing the rotational barrier height much. That is, the barrier height changes by orders of magnitude, but for most bearing parameters it’s far less than a zeptojoule. So Eric did look at this a decade ago.
If I recall correctly, the frictional contribution we most disagree on is whether atoms in stiff surfaces will go “sproing” sideways as they pass each other. It looks like the Nanosystems calculations assumed that the atoms would be held in place, maximizing adverse force; this kind of analysis wouldn’t notice atoms going “sproing” in a real molecule. From the 9 and 19 peaks in 10.12, it looks like there’s a barrier height of about 50 zJ per atom even at the smallest separation; from the caption of 10.10, this is over a circumferential distance of 1.5 A. That’s a sideways stiffness of ~2 N/m (assuming a linear force curve and a spherical cow). From Fig. 3.4, it appears that single bonds, even Si-Si, have a stiffness of hundreds of N/m. A surface atom will be bonded trusswise and will have a good fraction of that stiffness. So without going out and buying molecular modeling software, I can still be pretty sure that these atoms won’t be going sproing. So, no added friction.
For bond breaking, I don’t have to think too hard. Stretching a bond by 1% won’t even come close to breaking it.
Chris
I realize my comment on bond breaking was a bit hasty; I also have to compute whether pressing atoms together will make them bond to each other too easily. The short answer is no; a 1% change in separation distance will change the system’s energy by just a few zJ per atom, not nearly enough to overcome the energy barrier to breaking multiple bonds simultaneously, even if the rearrangement would be favorable.
Chris
Chris, good work getting your assignment in on time but I’m going to have to deduct marks as there’s evidence it isn’t all your own work, and as you didn’t reach the final answer.
Figure 3.9 is a good place to start, and already you’ve found one important feature, which is that the force is an approximately exponential function of separation, so as soon as you go into that “tight fit” regime the forces can rapidly get quite large, into the range of tenths of nanoNewtons per atom. If we go back to the Harrison simulations we find that this is exactly the kind of range of normal forces for which we move into the regime of substantial (i.e. mu>0.1) friction coefficients. As for mechanochemistry, that is a trick question because it’s going to be complicated and dependent on the particular surface and reconstruction involved. Good job you corrected yourself, obviously the force curves are very asymmetric and the situation in compression is very different to that in tension. Strain energies are in the ball park of tens of kJ/mole, compared to typical bond energies of hundreds of kJ/mole; since activation energies for energetically favourable rearrangements are only going to be a fraction of the bond energy (you don’t have to completely break one bond before you make another one) the margins aren’t that comfortable. The point is that more work is needed with proper computer simulations (not simply molecular modelling); the answers aren’t all in Nanosystems.
Yes, obviously it’s complicated. Many configurations will work, and many will not. Are you suggesting that all configurations will not, or we will be unable to find the ones that will? We only need a few different bearings, and so far, it looks like there’s a huge range of possibilities.
Note that the outer shell can be expanded rather than contracted, reducing stiffness (but not too much) and reducing both friction and reconstruction.
It’s been fun, but I have to get back to work now. Email me if you come up with an argument that shows we can’t make any bearings less than 10 nm across. Especially since Drexler already found and tested three of them by 1992, both strained-shell and special-shape.
Chris
The URL linked to my name in this comment actually goes to a sci.nanotech post I made in 1993 analyzing static friction in a Nanosystems ring bearing. Depending on how stiff the atoms are, the static friction could be as much as 12 orders of magnitude greater than that estimated by Drexler based on a completely stiff ring. Now, in practice the ring is probably pretty stiff compared to the forces involved, so my guess is that the actual increase in friction is relatively modest. But clearly the “sproing” problem is very relevant to the frictional situation.
Hal, the link didn’t work for me, but I’d be interested to see your calculation.
These google groups urls… I wish they had a “link to this page” link. When you find something by searching the URL doesn’t always work later. Try this one. Analysis of static friction in Nanosystems bearings.