• Google
    This Blog Web

October 2011

Sun Mon Tue Wed Thu Fri Sat
2 3 4 5 6 7 8
9 10 11 12 13 14 15
16 17 18 19 20 21 22
23 24 25 26 27 28 29
30 31          

RSS Feed

Bookmark and Share

Email Feed

  • Powered by FeedBlitz

« Security Council | Main | Climate Change and Molecular Manufacturing »

November 01, 2004


Feed You can follow this conversation by subscribing to the comment feed for this post.


Of course, bottom-up "molecular manufacturing" is possible. I don't see the argument in that. The argument is whether that bottom-up "manufacturing" is going to be based on solution-phase reactions (like biology) or whether mechano-synthesis is possible.

Many people, like myself, are still not convinced of the possibility of mechano-synthesis for the reasons that are very eloquently expressed by Richard Jones in his book "Soft Machines".

Chris Phoenix, CRN

Kurt, you've been led astray by several red herrings at once. Let's try to sort them out.

Mechanosynthesis is a subset of molecular manufacturing; there are other ways to achieve most of the performance that Drexler claims. You can build quite strong chemicals underwater--including graphite.

Soft Machines doesn't address mechanosynthesis much; rather, it talks about mechanisms. And yes, it's eloquent; that's its main virtue, because it's wrong. Jones's basic argument is that if a machine doesn't work the way biology does, it's probably not as good as biology. And the basic reason for that is that he thinks biology, in its billions of years of evolution, has probably found the best way to do things.

Never mind the fact that biology has been completely unable to explore huge ranges of chemistry, including all of vacuum-phase chemistry. Never mind that molecular shape change is demonstrably unnecessary for low-friction sliding. And never mind that biology has had to solve completely different problems from nano-engineering--including metabolism.

Jones's specific arguments against semi-stiff machines boil down to Brownian motion and surface forces. He warns us that Brownian motion will make the machines too imprecise to use--as though this hadn't been taken into account over a decade ago, and in detail. He claims that surface forces will join surfaces "irreversibly"--and here he's obviously wrong, as VdW force between flat surfaces is about 1% of bond strength (so the surfaces can be pulled apart) and can be reduced even more by simple spacers.

This is a slightly more sophisticated version of Mark Ratner's blatantly false argument: "Nanomachines have no power source. Therefore they can't work." Step 1: assert, contrary to a decade of publication, that the technology has a nonsensical limitation. Step 2: assert that that limitation is fundamental, distracting people from the fact that the first claim is baseless. Eric Drexler's comment on Ratner's argument: "Oh, gee, Mark, you mean robots need power to work? I'll go right away and tell Marvin!" That would be Marvin Minsky, Eric's PhD thesis advisor at MIT. I can imagine what Eric might say to Richard Jones: "Oh, wow, things wiggle at the nanoscale? Gee, it's a good thing I just happened to design a "Stiff Manipulator" back in 1992 that can hold its position within a fraction of an angstrom even at room temperature."

To recap:
1) Jones doesn't argue against mechanosynthesis, he argues against semi-rigid machines.
2) Jones has a prejudice against semi-rigid machines, because biology doesn't use them, and Jones likes biology.
3) In warning us of Brownian motion and surface forces, he's giving us very old news.


Richard Jones

Chris, maybe you should read my book again, this time with an open mind.

We discussed Brownian motion at length elsewhere, and I don't propose to restart that. Suffice to say that when you say "a decade of publication", what you mean is that the effect of Brownian motion was worked out in one case for one simple component of a nanomachine and nothing more has been done since. You were left arguing the very weak position that Drexler's undoubted awareness of the problem is sufficient to render detailed calculations unnecessary.

I don't have to imagine what Drexler would say to me because I had an extensive correspondence with him over the summer. Everyone can read the public outcome of that correspondence. I did learn something from this - I was reminded that Drexler's published work makes it clear that he, too, likes biology, and that he views the most likely implementation path for radical nanotechnology as being through biomimetic routes.

You say about my book: "And yes, it's eloquent; that's its main virtue, because it's wrong." I take great exception to this. I'm waiting for you to produce a serious critique of the book. I was looking forward to such a critique, in fact, since it might move the argument forward. But all we've had so far from you are ex cathedra statements like this coupled to a refusal to engage with the real argument.

Howard may well be right that discussion of molecular machine systems is moving back into the mainstream; I hope so, not least because one purpose of my book was to try and shift the debate back from a discussion of nanomaterials to these more interesting issues. But you, by your closed-minded adherence to one particular vision for nanotechnology, are missing the opportunity to move back towards the mainstream too.


Aside from brownian motion and surface forces, my other problem with mechanosythesis is that of complexity. Biological systems have several levels (a hierarchy) of organization, starting with DNA/RNA/ribosome to cellular to the whole organism. Each level has lots and lots of redundancy to cope with the inexact nature of biological processes. It is true that some reaction mechanisms are based on mechano properties (the passage of certain molecules through cell membranes) but other mechanisms are based on diffusion-driven chemistry and the bulk of it based on control of the reaction site by the complex folding of the proteins such that only the intended reaction site is exposed.

There is an incredible amount of redundancy to this and this redundancy exists on all the levels of complexity. The self-repair/regeneration nature of biological systems is one manifestation of this redundancy.

I am not a biologist or a chemist by training. However, I have read enough about these things to believe that building this kind of capability into a "nano-mechanical" system is going to be rather difficult (in system organization) and that, when done, the nano-mechanical system will not be much different from a biological system.

Also, solution-phase chemistry can make a huge variation of chemical compounds and structures, as biology can attest to. There are alot of chemical reactions that can occur in solution-phase chemistry, where the solvant acts not only as a solvant, but as catalyst as well. Nano-mechanical based reactions (presumably based in vacuum environment) may or may not be able to do this, but you would have to recreate all of the needed reactions to make things from scratch. Reverse-engineering biology seems to be easier to do, in comparison.

Biology may be needlessly complex due to the vagrancies of evolution, but much of the complexity may be due to the need for redundancy in order to compensate for all of the reactions that don't produce anything of value to the system.

Perhaps there will be several flavors of nanotech in 20-30 years, both wet and dry, with various advantages and limitations with respect to each other.

Michael Vassar

Yes Chris, you have been effective in promoting mainstream acceptance of MNT, but even more-so in promoting at least tenative acceptance of a terrifyingly fast prospective timelineby those who already accepted MNT's theoretical feasibility. That said, your timeline is not yet accepted by many MNT advocates, among them Robert Frietas. Serious examination of the basis for different time-line possibilities remains a high priority.
With regard to bio vs. vacuum, let's keep it civil. Remember the progress made in the Atkinson debate. Arguments regarding the relative feasibility of vacuum and biomimetics are very worth considering, even if you are confident (as am I) that diamondoid in vacuum is possible in principle, it may not be revolutionary if bio is sufficiently flexible. This means that a MNT revolution may occur without some key technologies we are used to assuming, which has major implications.

Chris Phoenix, CRN

Where did you-all get the idea that I'm arguing against biopolymers and squishy systems? I'm not. I'm not even arguing against wet chemistry. Molecular manufacturing, as I define it, could very well be done by wet chemistry before it's done by diamondoid mechanosynthesis. In fact, that might be the safest course, since it'll slow the transition. And I'm leaning heavily toward promoting the development of wet MM ASAP, as a recent blog article of mine indicated.

What I'm arguing against is the idea that this is the *highest performance* way to do molecular manufacturing. Let me be clear: Most of Richard's book is not wrong. What I take exception to is not his descriptions of soft machines, but his attacks on hard machines.

Why does it matter? Because there's reason to think that hard machines will have orders of magnitude higher performance than soft machines. If soft MM is only the beginning, then those who focus on that will still be missing many of the implications of MM.

Richard: Above, you say my position was weak: "the effect of Brownian motion was worked out in one case for one simple component of a nanomachine and nothing more has been done since." I disagree with that. Several components, including the planetary gear, have been simulated at the atomic level. And the effect of Brownian motion--that is, positional uncertainty--was considered (for example) in a paper of Ralph Merkle's on stiff manipulators.

In fact, in the Brownian discussion you referenced above, you didn't answer my last post: "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."

There are lots of ways to do things. Molecular transport can be done by diffusion or by conveyor belt. Only the latter works in vacuum--and in vacuum, it can be very efficient and high-throughput. Low friction can be achieved by squishy molecules or by superlubricity. Superlubricity requires great cleanliness, but can support a lot faster speeds.

My comment above, to the effect that surface forces are not "irreversible" as you claim, was meant to be a critique of your book. Let's talk about that.

You asked me to "engage with the real argument." I thought I had done that. As I understand it, your argument is that surface forces and Brownian motion will prevent hard machines from working.

Your other argument is that soft machines work very well. We have no disagreement there--as long as "very well" is understood to be relative to today's engineering, not relative to the limits implied by scaling laws and material strength. By that criterion, soft machines don't look so good. And hard machines do.

(I should say, to avoid nitpickers, that I'm well aware the word "hard" is inappropriate even for diamond at the nanoscale. But "relatively stiff machines" takes too long to type.)


Chris Phoenix, CRN

Michael, the Atkinson debate did not start in a spirit of goodwill and civility--and it turned out well anyway. People are only human, and frustration will creep in now and then. Perhaps I should have been politer in saying that I thought Richard was demonstrably wrong in at least some of his core arguments.

As to timeline, I agree that it's a very important question. But I don't think my timeline arguments depend on the things I'm busy disagreeing with Richard about. Now, if Kurt is right about complexity, then my arguments are shot; I'll answer him in the next post. But the timeline is based on the ability to engineer at the nanoscale--not the ability to build hard machines. We still find it hard (no pun intended) to engineer proteins, but there are other molecules than proteins.

What does depend on the Richard/physics questions is the performance, and probably the cost, of the products. If we can only build with biopolymers and run the machines underwater and use floppy interacting chains to avoid friction (at slow speeds), then we don't have to worry nearly so much about weapons or manufacturing revolutions. So this issue is very important. But I don't think this issue calls my timeline into question.


Chris Phoenix, CRN

Kurt, you raise an interesting question, and I'd like to co-author a Wise-Nano page to explore it. Will you release that post under GFDL?

Meanwhile, I'll give a preliminary answer here.

Cells have to deal not only with internal errors, but with external varying conditions. They have to reconfigure themselves constantly, which means being able to metabolize their components. Also, anything larger than a cell has to grow from a much smaller seed/embryo--more reconfiguring and metabolizing. And the chemistry has to be able to evolve. And the first life must have started with very simple processes and very few chaperones. All this argues for self-assembly and loose imprecise molecules.

An engineered machine would not have any of these constraints. It would have to deal with errors. But if it were made by robotics rather than self-assembly, it could have quite a low error rate--low enough that it would not need to self-repair once it was built. (Error checking and correction during construction is a separate topic.) Its behavior would be quite limited compared to life; it would have a greater ability (and need) to ignore external conditions.

So I'm not sure that redundancy is needed at the level of molecular machines. It'll be needed at higher levels, say between 100 nm and 1 micron, but at those levels redundancy can be implemented engineering-wise.

You're right that solution-phase chemistry can make lots of reactions and structures. It's worth noting that Drexler says that mechanosynthesis can also be quite rich. (There are certainly vacuum-phase reactions that are not possible in solution phase--and probably vice-versa.) But I fall back on a simpler and very reliable argument: a few simple reactions, repeated in programmable positions, can build complicated shapes. I lose performance by this: the feature size may be 1 nm instead of 2 A. But that's OK; I'll give up that order of magnitude, in exchange for the multiple orders that are gained by carbon lattice machines in vacuum.

You say, "Perhaps there will be several flavors of nanotech in 20-30 years, both wet and dry, with various advantages and limitations with respect to each other." Sure, maybe. There's nothing wrong with wet nanotech. But there's nothing wrong with analog circuitry either. And the trend for the past few decades has been to digitize everything. And then to run it through CPUs. And then high-level programming languages--even interpreted languages! Why? Because it's easier to engineer. In this case, we know that we're giving up many orders of magnitude of performance. In the wet-vs-dry case, it looks like dry has *higher* performance.

My personal prediction is that wet nanotech after dry nanotech is developed will be in the position of analog computers after the transistor. Before long, even medical devices will use dry nanotech with wet interfaces. I could be wrong about this, and it wouldn't matter much. Because the basic point of my argument is that nanotech will rapidly approach the performance implied by scaling laws and bond strength. If wet nanotech can do that--well, then it's done.

What's dangerous is the following argument:
1) Wet nanotech may be better than dry nanotech.
2) Biology may have reached the limits of wet nanotech.
3) Therefore, biology is already as good as nanotech can be.

The logical flaw is obvious here; though it's sometimes hidden when people assert "is" rather than "maybe."


Richard Jones

Chris, you need to learn the difference between molecular modelling and computer simulation. As to the importance of surface forces, remember that my consistent position is not that there is a physical law against MNT. It's just that it's going to be much more difficult than you think. Decades of experimental results in colloid science, and more recent experience in MEMS, shows that stuff at the nanoscale sticks. Your back of envelope calculations are characteristic of your general approach, in which you gaily trust your intuition and crude guesstimates over experimental experience. There's one fundamental law of experimental physics and engineering that you never take account of. This side of the Atlantic, we call it Sod's law.

This is highly relevant to the point that Michael Vassar raised. Even if I thought that MNT was entirely unproblematic, your estimates of the timescales needed to implement it would still be ludicrously unrealistic. You often make comments like, "it's just engineering". Well, I'm just a scientist, and I still know very well that the science is the easy bit. If you get something to work in a laboratory, you're still years away, not to mention most of the total expenditure away, from a manufacturable product. With MNT we have no laboratory prototype, we don't even have any idea how to start making one (and let me tell you if it was as obvious as you think it would have been done many times by now, Smalley or no Smalley). As Michael says, I don't even think anybody else even in the hardcore MNT community agrees with you on this.

Maybe it's this that underlies the fact that you view MNT more like a cargo cult than a scientific development. To you, nothing is going to happen until MNT drops from the sky. This attitude leaves you with an unhealthy passivity, well illustrated by your post about climate change. "The more important question is whether humans can do something about it. With molecular manufacturing, we can. Without it, probably not", you say. That statement is just risible.

There are many things that can be done to reduce our carbon emissions and our dependency on imported fossil fuel. Many of them depend on ordinary boring incremental nanotechnology, and in this Richard Smalley is exactly right. How can we get solar cells with respectable efficiency but an order of magnitude lower capital cost, capable of being manufactured in square mile quantities? There are such candidate technologies in the lab now, Grätzel's dye sensitised nano-titania cell being one example (an example, incidentally, which illustrates nicely how you can learn from biology without necessarily slavishly copying it). But no, you'd rather wait for the MNT enabled nirvana than press for these technologies to be developed now.

Ah, but MNT will have orders of magnitude higher performance, you say. All this illustrates is that you have a ludicrously one-dimensional view of what constitutes performance. If you are making a solar cell, what does it matter that diamond is stronger than silicon? It's issues of electronic structure and nanoscale morphology that are important. Different applications have all kinds of different figures of merit, and it's absurd to imagine that a single technology based on a single material will meet all human needs.

Chris Phoenix, CRN

I'm not quite sure how to answer your latest sally. You say "Decades of experimental results in colloid science, and more recent experience in MEMS, shows that stuff at the nanoscale sticks." I've never doubted that. What I doubt is that, once it sticks, it can't be pulled apart again. You have neither defended nor retracted your use of "irreversible." Please clarify your position.

You say I have no experience in engineering. But that's not quite true. I have six years experience at a company that designed high-end computer hardware. They did this an order of magnitude faster than the competition. As an embedded software engineer, I was involved with both the hardware and software side; I not only know what ground bounce is, I know what it looks like on a scope.

One of the things I saw was that hardware design usually takes an order of magnitude longer than software design. Why? Rapid prototyping. The exception that proves the rule is FPGAs: computer chips that can be reprogrammed in minutes. Even though it was part of a very complicated system, the FPGA designer took just a few days to get his chip working.

If molecular manufacturing were limited to complex and slowly-built systems, it would indeed take a long time to develop. The reason I think it'll be fast is because development will be programmable at every level. Once a level works, it can be used to build test-configurations at the next level in hours or days instead of months.

You said, "As Michael says, I don't even think anybody else even in the hardcore MNT community agrees with you on this." But Michael didn't say that. He said "your timeline is not yet accepted by many MNT advocates, among them Robert Frietas." Watch your accuracy, especially when attributing opinions to other people. By the way, there are other MNT people who agree with me, but not on the record AFAIK, which is why I haven't cited them.

I don't know where you get the idea that I'm passive. As I wrote in my previous response, "I'm leaning heavily toward promoting the development of wet MM ASAP." That means working with actual researchers to get them to work on wet MM.

There are indeed many things that we can do to reduce carbon emissions somewhat over the next few decades. But that won't be enough to mitigate the cumulative effect of climate change. That was what I said we'd probably need molecular manufacturing for.

It's true that solar cells can be built without MM. But what about power conditioning and storage? What about the installation hardware and skills? Without MM, the best solar cell you can imagine would take decades to make even a limited difference. So your argument appears as "one-dimensional" as your characterization of my argument.

Now that we've shown we can each cut down each other's ideas, let's quit that game and get constructive. I've just created a category over on Wise-Nano: Energy products. I added a first-draft page showing what carbon-lattice molecular manufacturing could do to free us from fossil fuels. Please fill in the comparable page showing what nanoscale technology can do.

Your closing statement: "it's absurd to imagine that a single technology based on a single material will meet all human needs." I'm not sure it's more absurd than imagining that a single technology could meet all information processing needs. Of course computers aren't a single technology. But neither is carbon lattice. There are lots of materials that can be built out of just diamond and fullerene. Once we have the reactions to do that, the rest is blueprints. There are some products that will take a lot of R&D, but there are many simple but useful ones that should be easier to design with this material and manufacturing process than with today's technology base.




Go ahead and use my post.

Richard Jones

For irreversible, get a colloid science textbook and look up the difference between flocculation (reversible attraction, stuck in the secondary minimum of a DLVO potential) and aggregation (irreversible aggregation in the primary minimum). Or for a more practical demonstration, just run your car without oil for a few weeks and see how reversible that is. Or to be more formal, a surface has an associated positive free energy. Two surfaces placed together can reduce that free energy by erasing the surface. As this is a process that increases the total entropy of the universe, it is irreversible.

Your comments about software versus hardware are irrelevant. What has not yet been designed is the hardware.

You haven't in the least shown that you can cut down my ideas. If you're going to talk about being constructive, describing my book with the words "And yes, it's eloquent; that's its main virtue, because it's wrong" is not a great way to start.

jim moore

As someone who spends his days de-aggregating and stabilizing nano-scale particles Richard is not completely correct. Adding energy to your system in the right way can easily break up agglomerated particles. The traditional way is to use collisions and shear forces to break up the particles. The key to stabilizing the particles is to modify their surface so that the particles cannot get too close to one another. Typically you use steric hindrance or electrostatic repulsion to keep the particles separated.

Now Richard’s very general point about surface forces dominating at the nano-scale is from my experiences completely correct. One of the big areas of R&D for nano-science / engineering is to develop controllable (programmable?) ways for nano-scale surfaces to interact.

Chris Phoenix, CRN

1) I'm well aware that *some* surfaces will stick irreversibly--like metals. That's not the point. Graphite, for example, does not stick irreversibly.

You wrote in your book, "The continual shaking ... and the large forces that will stick any surfaces that come into contact irreversibly together are inescapable obstacles to the kind of precision nano-engineering that would be needed to miniaturise systems of gears and cogs to the nanoscale."

I disagree that every surface that comes into contact will be stuck irreversibly. If your statement was too strong, then there is still hope for the gears-and-cogs approach.

(I can give you a good argument on the shaking/tolerance question too, but let's deal with one at a time.)

2) Hardware can be like software, if it's computer-controllable. That's the point of the FPGA example.

3) I already said I should probably have been politer. Note I asked for a _shift to_ constructiveness.


Michael Vassar

Richard: I would really appreciate a more nuanced position on what I said. Chris is definitely not alone in his position. He may not even be in the minority among the tiny set of people who have actually worked on and contributed to the theory of MNT, e.g. the people who understand it best. To me, one of the most striking features of MNT is that the more we learn the easier it looks, precisely the opposite of most technologies.
For all that, other forms of nanotech are almost certain to have significant impact first. I advocate that not just you but everyone try to put together a relatively complete breakdown of major expected nanotech applications together with estimated development costs, both absolute and as a function of assembler cost of development.
That said, Chris focuses on replicating technologies because while sustainable energy, cures for cancer, etc may stem from ordinary nanotechnology, the truly revolutionary developments come either from the ability to program matter digitally, or from self-replicating manufacturing, two intimately related technologies. There is definitely reason to advocate nanotechnology, but only molecular nanotechnology promises a change in human affairs that makes extrordinary demands on human Foresight and Wisdom. (this isn't quite true, I can think of non-mnt nanotech doomsday weapons or singularities, but they seem relatively unlikely).

Richard Jones

A good thought experiment to get a grip on the sticking issue is to imagine your two surfaces in intimate contact, then ask how much free energy is gained when the interface is erased. For graphite and some layer minerals like talc, the answer is none, because even at equilibrium sheets are held together by van der Waals or electrostatic forces. For everything else, the lowering of free energy is substantial. This tells you about equilibrium; now you have to ask how easy it is to get to equilibrium, and whether in practise you can fend off the inevitable consequence of the second law for long enough to do something useful. For clean metals, getting to equilibrium is easy because metallic bonding is not strongly directional so energy barriers are very small. For covalently bonded materials the energy barriers are higher and depend strongly on issues like whether the surface is effectively passivated, on the one hand, or whether the surface structure is strained or reconstructed on the other. As Jim rightly says, the aim of colloid technology is to keep the surfaces apart, away from the primary minimum, by steric or charge stabilisation, and this is the way you can get redispersible colloids.

Michael, I'm sorry if you think I took your name in vain. The reason what you said resonated with me so much was that in my correspondence with Drexler, he said two things that surprised me a little. Firstly, he explicitly dissociated himself from people who did predict a very fast timeline for MNT, and secondly he said that he himself was sceptical of claims that there would be a direct route via mechanosynthesis to MNT, rather than a route via an initial biomimetic stage.

I think your other point is really interesting even though I disagree with you. I think that MNT supporters consistently underestimate the potential transformational importance of non-MNT nanotechnologies, and in this they actually end up on the same side of the argument as the nanobusiness types, with their emphasis on short-term and rather trivial applications. I think sustainable energy, ambient computing and ubiquitous intelligent artefacts, and developments in nanomedicine may well be revolutionary even in the absence of MNT. On the other hand, I think that MNT supporters overstate the importance of manufacturing and materials. As I've said before, only half flippantly - if you want material goods that cost, to a first approximation, nothing, you've got them already. Just go to IKEA.

Michael Vassar

I think that there is simply a difference in what we mean by revolutionary. Non MNT nanotechnology including biotechnology, information technology, etc, can probably deliver a rate of change comparable to that of electrification. This is rapid enough to make most long term plans made today irrelevant, and goes way beyond the level of change that the mainstream population sees as "realisitic". It can provide energy independence, environmental sustainability, waste management, and probably even affordable space access. It can create overwhelming military supremacy while greatly reducing risk to both sides in armed conflict, and can probably be used to make post-nuclear superweapons. Aubrey De Gray and others even see indefinite human life extension as probable without MNT. Here is an attempt I made to present some of its practical implications in a minimal future-shock form. http://www.futurist.com/portal/future_trends/SuburbsFuture.htm My article's timeframe of 50 years is highly unambitious, to say the least.
However, it is not clear to me why non MNT nanotechnology would need a center for Responsible nanotechnology. Advocacy, sure. Funding, lots. But specific efforts to be responsible? Not as far as I can tell. Surveillence could be a problem. Neurological interfaces might raise ethical issues. But existing balances of power would be generally preserved, and existing political institutions could deal with surveillence issues etc as they came up. CRN is necessary if and only if MNT is developed rather abruptly during the next 30 years. During that time-frame, the abrupt nature of the changes enabled by MNT will demand prior analysis if they are to be handled safely. This concern remains valid even if one abandon's Chris's extreme timeline, and even if one projects an extremely optimistic rate of technological development in nanotechnology as a whole, so long as MNT develops abruptly relative to historical technologies and prior to incremental technology producing outcomes on par with my third scenario listed above.

Michael Vassar

"Well, I'm just a scientist, and I still know very well that the science is the easy bit. If you get something to work in a laboratory, you're still years away, not to mention most of the total expenditure away, from a manufacturable product." is actually a very good summary of the argument for thinking that MNT offers a much better return on the amount of necessary effort than other nanotech. Simply put, with nanotech solar cells, the prototype>product gap is huge, but with assemblers, the prototype assembler gives you products as soon as it reproduces itself a few dozen times. This is at the heart of the software for matter analogy. For someone who's writing the first word-processing program, the gap from prototype to marketable product is very small. Likewise for someone making the first MNT devices in a large set of product classes.

Mike Treder, CRN

Richard Jones wrote:

...in my correspondence with Drexler, he said two things that surprised me a little. Firstly, he explicitly dissociated himself from people who did predict a very fast timeline for MNT, and secondly he said that he himself was sceptical of claims that there would be a direct route via mechanosynthesis to MNT, rather than a route via an initial biomimetic stage.

For clarification, I asked Eric about this, and he said:

Regarding very fast timelines: For some value of "very fast", timelines surely become implausible (I am, of course, willing to associate with people who expect faster timelines than I do, even if I disagree with some of their views).

Regarding direct mechanosynthetic routes to MNT: Direct routes are surely possible, but I expect that more-or-less biomimetic approaches (based on polymers and self-assembly) will yield results faster. This has been my view for some time.

Chris Phoenix, CRN

Richard, studying the energetics of end products will tell us almost nothing about the safe handling of dynamite. I'm not sure why you even brought it up in a discussion of dynamic processes gated by energy barriers. It appears to be a distraction.

I'm not going to be distracted from the issue of whether stiff machines can work. You wrote that all surfaces, pulled together by surface forces, would stick irreversibly. You claimed this irreversible sticking as a reason why stiff machines could not work.

Despite several requests, you have not directly defended nor retracted your statement. But you've come pretty close to a retraction in writing, "For covalently bonded materials the energy barriers [to sticking] are higher and depend strongly on issues like whether the surface is effectively passivated, on the one hand, or whether the surface structure is strained or reconstructed on the other."

So can we take it that well-designed surfaces can be prevented from sticking irreversibly? Can we take it that whether a stiff machine will stick depends on how well it's engineered, and that there is no known fundamental barrier to making stiff kinematic machines with separable surfaces?


Richard Jones

Understanding that dynamite is thermodynamically unstable is rather important when you decide how to handle it! It should make you rather cautious about anything that lowers the energy barrier that's keeping it kinetically stable. Exactly the same is true about systems with lots of surface; they have lots of energy stored in them, and you've got to work hard to keep them kinetically stable. (There was an entertaining and only slightly off the wall Russian theory of ball lightning going around a few years ago which held that the phenomenon was the result of the rapid aggregation of some ultrafine particulate which released all the surface energy in an explosive fashion).

Your formulation "no known fundamental barrier" looks pretty much like my formulation that MNT isn't forbidden by the laws of physics. To quote directly from my book's conclusions;
"I do not think that this approach (i.e. MNT) fundamentally contradicts any physical laws, though I think that some of its proponents underestimate the problems that features of the nanoworld, like Brownian motion and strong surface forces, will pose for it."

What we disagree on is simply a judgement call about how easy the good engineering you are going to need to do is going to be. I think it's going to be harder than you think. I may be wrong, you may be wrong, it's an issue that will be resolved one way or another by experiment. It's not worth having a culture war about (particularly since there seem to be enough of those going on already).

So I'll respond to your statement, "I disagree that every surface that comes into contact will be stuck irreversibly. If your statement was too strong, then there is still hope for the gears-and-cogs approach" by agreeing that "every" is too strong, and there is indeed still hope.

In this conciliatory mood, I would though urge you to read my book again and look for the positive message. This is that the different physics of the nanoscale offers new opportunities and new design philosophies for making nano-machines and devices. My message isn't that biology is the best possible nanotechnology you can do. It's that biology offers us unfamiliar design paradigms that have been developed and optimised for the nanoscale world, and that we need to think through why biology does things the way it does, to see if there are tricks that we would otherwise miss due to the fact that our engineering intuition is developed in a different domain. My major argument against MNT isn't so much that it won't work because it doesn't take account of surface forces, Brownian motion, etc; it's that it may not be the best way of doing things because it tries to engineer round these problems rather than exploiting them.

Hal Finney

Richard, I pre-ordered your book several weeks ago and just got word yesterday that Amazon had shipped it. So I am not familiar with the details of your arguments. Are there any specific designs or mechanisms proposed by Drexler in Nanosystems which you think won't work due to either surface forces or thermal motion (I don't think Brownian motion is the right word, because that refers to jostling due to molecular impacts in a fluid)?

Chris Phoenix, CRN

Richard, thanks for "agreeing that "every" is too strong, and there is indeed still hope." Please remember, as you defend your book, that I'm not the only one who reads it as saying Drexler's approach can't work. Remember Kurt's statement that started this discussion: "Many people, like myself, are still not convinced of the possibility of mechano-synthesis for the reasons that are very eloquently expressed by Richard Jones in his book "Soft Machines"."

Now that we've cleared up the sticky-surface question, I want to address the thermal noise question. In your book, you point out that thermal noise will cause rather large positional deviations in nanomachinery. I agree. You equate positional deviations with manufacturing tolerance. I disagree. So let's talk about what problems, if any, positional deviation will cause for nanoscale machinery.

First point: The average shape of two stiff molecules with identical chemical structures will be identical.

Consequence: An error in tolerance in a macro-scale machine can cause a bigger error in tolerance in the machined product. But at the nanoscale, unless the positional variance is large enough to cause chemical errors, the product will be just as precise as the fabricator.

Second point: Thermal noise can't be amplified by passive mechanisms to break bonds that otherwise wouldn't break. If thermal noise could "push uphill" like that, we could build a ratchet and recover it.

Consequence: The thermal noise-induced positional variation won't break bonds that would otherwise be stable. This appears to be true regardless of the complexity of the machinery--in other words, this variation can't break off gear teeth, etc. Compare this to macro-scale machines, where a tiny slop in a gear train can cause impacts that quickly destroy the teeth. But the thermal-noise-caused slop in a nanoscale gear train can't possibly convey more energy than the thermal noise.

So if it doesn't break stuff, and it doesn't cause manufactured products to be less accurate than their manufacturing system, the only remaining consequences I can see for thermal noise are 1) undesired kinematic-state transitions; 2) noisy sensors.

Undesired kinematic-state transistions means things like gear teeth slipping past each other. In general, this would require multiple large bond deformations--far too much energy for thermal noise to supply. Preventing this kind of problem can be achieved by basic engineering.

Sensor noise is in one sense a fundamental barrier--you can't do better than a certain bandwidth--but in another sense merely an engineering issue. In today's technology already some sensors and amplifiers are operating close to the informational limits. And it's certainly possible to get useful information with 10^-15 error using nanoscale hardware.

So what problems will thermal noise cause for nanomachinery?


Chris Phoenix, CRN

Richard: "My major argument against MNT isn't so much that it won't work because it doesn't take account of surface forces, Brownian motion, etc; it's that it may not be the best way of doing things because it tries to engineer round these problems rather than exploiting them."

The point is not which technology will be more efficient in the end. The point is that vacuum-phase engineering with stiff machines provides a lower bound of performance. Saying that bio techniques will be more efficient is very different from saying that there's anything wrong with the performance projections already in existence for stiff vacuum machines.

I agree with the suggestion that biology has techiques we can borrow for use by engineering. That says nothing against the MNT approach.

I am suspicious of the suggestion that we should try to use all phenomena in every device. That seems to require harnessing complexity, which would mean we couldn't engineer. Maybe as technology matures this will become more useful, but as long as technology is progressing rapidly, I think fast R&D will be preferable to efficient products.

The whole point of engineering is to ignore most factors while using a few predictable factors for a predictable effect. One doesn't engineer complex systems--almost by definition. If you want to use or build a complex system, don't do it by engineering. If you have a simple problem, then use engineering to bypass environmental complexity and design a solution quickly.

I also think a general limitation of bio machines is that they work underwater. Maybe some ideas can be translatd to a vacuum environment, and we're certainly open to suggestion. But once we gain the ability to build vacuum-phase machinery, I think anything that operates in water will have a very hard time keeping up.

But I am willing to be proved wrong about bio keeping up with vacuum-phase. Just keep in mind the distinction: The important question is not whether bio is better. The important question is whether there's any problem with the projections of stiff-machine performance.


Richard Jones

Kurt is quite right to read the book in the way in the way he does - as I remember from his posts he's a practical person who has to compare technologies and sell them to people. To the engineer or the venture capitalist "there's a glimmer of hope" or "it's not fundamentally impossible" are not compelling value propositions.

Here are two bad consequences of bad tolerances from Brownian motion - the first is allowing small molecules to get to places where they shouldn't be, the second is causing moving parts that shouldn't meet to contact, leading to highly localised energy inputs. This is a recipe for uncontrolled mechanochemistry and irreversible surface damage.

To answer Hal very briefly, two designs from Nanosystems that I think will be problematic are the molecular sorter and rod logic. As for why I call thermal fluctuations Brownian motion, this is common technical usage in the field I work in, and I think conceptually it's a helpful and sensible broadening of the original sense.

The comments to this entry are closed.