Debating Radical Nanotechnology

Philip Moriarty reports that the Nottingham Nanotechnology Debate can now be viewed on streaming video here. The debate, held last summer, featured two proponents of Drexler’s vision of molecular nanotechnology, Josh Storrs Hall and David Forrest, discussing the feasibility of these visions with a couple of more sceptical observers, myself and Saul Tendler, a bionanotechnologist from Nottingham University. The audience included many distinguished nanoscientists, and even with the video available, it’s worth reading the transcript of the debate, which can be downloaded from The Nottingham Nanoscience Group’s webpages, if only to identify the authors of the many perceptive questions.

The aftermath of the debate included these additional points from David Forrest, which attracted some discussion on Soft Machines here. For my part, I organised my thoughts on the problems which I think the MNT program needs to address and overcome in this post: Six Challenges for Molecular Nanotechnology.

So where does the debate go now? I can’t conceal my disappointment that the MNT community has reacted with complete indifference to this set of challenges, which I set out in as constructive and concrete way as possible. Nanodot, the blog of the Foresight Nanotech Institute, simply ignored it. The most vocal proponents of the MNT position are now to be found in the Centre for Responsible Nanotechnology, but rational discussion of MNT in that forum is hampered by the fact that its proprietors simply refuse to engage in debate with informed critics such as myself and Philip Moriarty, preferring simply to assert, in the absence of any evidence, that the MNT revolution comes ever nearer. The usual outcome of a refusal to engage with people outside one’s own circle of believers is, of course, complete marginalisation. I regret this situation, because even though I think many of the ideas underlying MNT are flawed, Drexler’s writings have been very valuable in highlighting the potential of radical nanotechnology, and the process of thinking through what might work and what won’t is likely to be a very productive way of establishing research directions.

17 thoughts on “Debating Radical Nanotechnology”

  1. I’m planning on learning enough about SPMs to learn how to make/assemble a critical SPM component (a PZT actuator or a novel tip geometry). I’d like to attain a small business grant and grow the business enough to start making other SPM components. Eventually I’ll apply for a government Innovation Grant and attempt mechanosynthesis. But I’d like to use my own crappy tools with an eye to MNT scale-up decades from now rather than trying to be the first to achieve the mechanosynthetic step.

    Richard, do you or anyone else have any thoughts about what tools would be required for diamond mechanosynthesis? I’m pretty sure there are or soon will be nanotech facilities whose time will be available for public bidding in Canada (at least that’s the way they do things with the synchrotron, I assume SPMs are the same). Does the central hurdle appear to be in loading an AFM tip with a single molecule moeity? In a sense, half of the basic mechanisynthetic step has already been achieved; hydrogen has been abstracted from a diamond surface by dragging an STM tip over top. Is there any chance these “exposed” surfaces can be reacted with a hydrocarbon dimer using presently available tools, or is metrology still too imprecise. If a whole surface “chunk” of diamond has been surface abstracted, I wouldn’t expect the reaction co-ordinates to migrate, so only the Z co-ordinate would present a difficulty. Are our SPMs still too crappy and should efforts be made by the diamond surface community to try to pursuade SPM engineers to design some custom tools?
    I’m still torn whether diamond surface chemistry or SPM engineering is more important. Definitely vacuum conditions and an unterminated diamond surface for the basic mechanosynthesis step, right?

  2. Ooops. There was a problem with the link in the comment above. I’ll try again.


    PZT actuators are available in very many “breeds” at (in most cases) rather modest cost: shear piezos, stacks of shear piezos, tubes, tripods etc… Indeed, over the past few years I’ve set and supervised a project in our third year undergraduate labs. which involves building a rudimentary STM from a $5 piezoelectric buzzer. (And this year it worked! – see David Alexander’s website on the ‘buzzer design’ ). There are many small (and not so small) companies out there which make SPM components including, for example, Nanosurf, RHK Technology, Attocube, and Nanograph Systems Ltd. (the latter is a spin-off from the Nottingham Nanoscience group).

    What you’re suggesting is, I think, to move beyond current PZT technology so as to manufacture a piezo-driven microfabricated device capable of mechanosynthesis. (Is this correct?). The central difficulty is, as you suggest, in precisely controlling the tip structure: to control, one must initially characterise. A useful characterisation tool which has the potential to provide significant insight into molecular conformation at the end of a tip is inelastic tunnelling spectroscopy (IETS, as pioneered by Wilson Ho and co-workers). In the vast majority of case this necessitates, however, temperatures of ~ 4K. If IETS were combined with a structural probe such as field ion microscopy then in principle both chemical and “geometric” information become available. This is a rather complicated experiment, however, requiring the combination of low temp. STM/AFM + field ion microscopy (and the associated tip transer protocols between the instruments). (One could also envisage near field optics (e.g. apertureless SNOM) playing a role in elucidating the atomic-level structure of the tip).

    Hydrogen has, as you point out, been abstracted from diamond in STM mode. But this was with an uncharacterised tip – i.e. was the tip purely W in composition or could it have had C “contamination” on the end due to accidental tip crashes? – via an E field or tunnel current (vibrational heating)-driven process. “Half of the basic mechanosynthetic step”, as you put it, has not been carried out – no mechanosynthesis has been achieved because an E-field or a tunnel current drove the abstraction process. Drexler’s entire mechanosynthesis concept is based around the application of precise molecular tools using mechanical force to drive reactions.

    I accept, however, that single H atom abstraction using a scanning probe has been achieved. (This was nevertheless via a much more subtle “non-contact” process than is suggested by your dragging an STM tip over top statement). The key question is not that our SPMs are “too crappy” but that our tip technology is not sufficiently well-advanced. This to my mind is the primary stumbling block. In my debate with Chris Phoenix (now almost 18 months ago), I suggested a simple 7 step process to demonstrate one step in a rudimentary mechanosynthesis process. This required the application of tips with different chemical functionalities and, moreover, a process to regenerate the tip following the deposition of an atom/dimer. It is precisely this process on which I plan to focus in the coming years. (The issue of our SPMs being “too crappy” will impact when we want to increase the bandwidth of the assembly process… )

    Definitely vacuum conditions and an unterminated diamond surface for the basic mechanosynthesis step, right? .

    Why an unterminated surface? The elegant concept at the heart of Drexler’s thesis is that dangling bonds are tied up so that radical/ dangling bond centres can be introduced one at a time. With an unterminated surface (unless it’s a surface which already has a low free energy and does not have a propensity to reconstruct, such as III-V(110) planes), the process is made much more difficult. In my opinion, I’d rather try to crawl before tearing off at breakneck speed into completely uncharted waters… (Yes, a horrible mixed metaphor but I’m existing solely on a mixture of adrenaline and caffeine (!) at the moment at the MAX-lab synchrotron in Lund, Sweden). If you have a tip carrying a reactive “payload” the last thing you want to scan is a reactive surface where ‘extraneous’ reactions may occur well away from the site of interest.

    [On a separate, but somewhat related matter: Mechanosynthesis driven by scanning probes is an innovative way to structure matter. Let’s not underplay the difficulty of the technique, however. I get immensely frustrated by claims by ‘hardcore’ MNT supporters that molecular manufacturing requires no new science. Ho et al. Eigler et al., and Meyer et al. (Berlin/ IBM Zurich) produce stunning work on single atom/molecule manipulation which is published in the highest quality scientific journals. Are MNT proponents really dismissing this research as free of scientific advance?]

    I’m more than happy to discuss diamond mechanosynthesis with you at any point either via this blog or by e-mail ( I have a large backlog of data to write up over the summer (amongst other much less exciting adminstrative tasks), however, so please forgive me if my responses are not quite as timely as I’d like…

    Best wishes,


  3. Richard,

    It is interesting to read Jim Moore’s comment under your Six Challenges for Molecular Nanotechnology post. (For some reason I missed the lengthy discussion that the piece provoked at the time – apologies.) Jim, whose comments are of their usual high and perceptive standard, states:

    “This is the type of constructive criticism that supporters of MNT should rejoice in. You have taken the time and effort to understand the Pro-MNT position and have started to come up with research strategies that addresses some of the potentially problematic obstacles that may prevent MNT from being realized (in its current form). I am not sure why the Foresight Institute’s nor CRN’s blogs have not linked to it yet. “

    I find it remarkable that your MNT-specific comments have been ignored by CRN et al. while recent advances in nanoscience (such as the fascinating ‘nanocar’ synthesized by Kelly, Tour and co-workers) entirely unrelated to the molecular nanotech concept are touted as a key step towards the realisation of MNT. Kelly and co-workers have used elegant and sophisticated chemistry to synthesise the nanocar which then diffuses on a substrate in a highly intriguing fashion. But… where is the link to MNT? The nanocar has not been constructed in a deterministic fashion from the bottom up by computer-controlled assemblers. Nothing about the structure or diffusion of this molecule has got anything to do with the MNT concept.

    “All these things sound very familiar. It’s only a matter of time now — probably not very much time — until we have all the tools and skills to build a nanofactory. And then the nanotech revolution [PPT] will begin.” From US Scientists Doing MM

    Who has the hype? , indeed!


  4. I’m not defending specifically Drexlerian ideas; an exponentially scaling manufacturing technology is what interests me. If E-fields can actually be integrated into industrial scale-up and if Nanosystems would then need to be rewritten to accomodate E-fields, fine with me. I guess the diamond surface would need to then be “mechanosynthetically doped”, and the industrial thoroughput of the system might ultimately be limited by variables such as power requirements.

    “I accept, however, that single H atom abstraction using a scanning probe has been achieved. (This was nevertheless via a much more subtle “non-contact” process than is suggested by your dragging an STM tip over top statement). ”
    Okay, dragging an STM tip one nanometer over top. I’m sure there were a whole bunch of STM process operating conditions that had to be controlled to accomplish this.

    I don’t know if unterminated diamond surface is the correct terminology (carbon terminated?). What I mean is the sp2 carbon dimer surface that occurs on a diamond surface when there is an absence of hydrogen. The surface itself is very reactive, it tears apart any CNT probe tip contacted to it. It surely is less suitable for the MNTed creation of actually diamondoid products than is any hydrogen terminated diamond surface geometry. But there are two targets being discussed here: full Drexlerian scale-up and a proof-of-principle single moeity deposition. The Drexlerian scale-up on a hydrogen terminated diamond surface requires bumping off (specific) hydrogen atoms on the diamond surface, finding the newly created reactive surface site, and affixing a carbon moeity of some sort. 3 steps. For a proof-of-principle single moeity deposition demo, the “carbon terminated” diamond surface has no hydrogen to be removed. It only requires placing a carbon dimer (somehow) on the sp2 diamond surface and newly created sp3 bonds should immediately form just beneath the placement site. As you say, it will be difficult to ensure and confirm only the desired moeity is making contact with the diamond surface; the rest of the probe tip will want to interfere if such a reactive surface is used as the deposition site. I’m not sure which specific diamond plane would be best for this at the moment. But the difficult steps of abstracting hydrogen and then finding the reaction site for dimer depisition, would be eliminated.

    My comment about piezo actuators wasn’t related to MNT in general, just my own future plans. I see that all piezo actuators create imaging defects and need to be software corrected. I thought it might be possible to have an actuator manufactured that specifically corrects for imaging hysteris as a function of its geometry. Maybe there is even a better SPM actuator than a PZT tube, waiting to be discovered. I get a sweet business grant if I can come up with a viable business plan, and I want a small business that can be used to facilitate MNT. You seem to be saying focusing upon manufacture of novel SPM probe tips would aid MNT more than trying to cobble together a better piezo-actuator imaging platform?

    Philip, thx for suggesting field ion microscopy, IETS and apertureless SNOM as potential tools for radical nanotechnology. Beyond AFMs and STMs, there are almost a hundred different nanotech tools out there. The other ones I did look into were far too coarse for the purposes of MNT.

  5. I had a discussion with a friend of mine who knew Eric Drexler from his L-5 days (late 70’s – early 80’s). Eric himself saw nanotechnology from the soft/biological approach when he wrote his first paper on it, back in 1981. He did not start up with the “hard” MNT stuff until he had heard about the STM. It was the STM that created all of the hard MNT discussion (and hand waving) that has never gone away. So, the original idea of “nanotech” was based on synthetic biology or something like it.

    Yes, I noticed as well that the CRN and Foresight people have yet to address the issues brought up in “Six Challenges for Molecular Nano technology”.

    I consider the hard MNT concept to be dead. So, I do not pay much attention to Foresight and CRN.

  6. It is irrelevant whether CRN or foresight address the six-challenges, what matters is that people doing actual work address them. As was pointed out earlier on this blog the latest paper by Freitas and Merkle et al. addresses some of these points. Considering the small number of people working on these ideas, I don’t know what work that is not going on that could be going on. Overall, there is Catch-22 situation, where you must convince people that MNT is possible to get people to work on projects that could either prove or disprove MNT.

    On the subject of Dr. Tour’s work there is an interesting audio interview available here. The gist is that Dr. Tour’s long-term vision for nanotechnology involves the general principle of manufacturing macroscale objects from the atom up. He addresses the objections that such manufacturing will be inefficient by pointing to the tremendous speed that many kinds of grass grow. The specifics of his ideas defy classification into “hard” or “soft”. He talks of masses of “nanotrucks” moving from place to place transporting atoms. Absolute precision does not concern him, because there are so many trucks it does not matter if many do not reach their destinations. So, you could say that this is more a stochastic (or soft,wet) manufacturing paradigm compared to MNT. On the other hand, many critics of MNT have said essentially that NO macroscale machines will work at the nanoscale; the existence of a “nanocar” however assembled puts doubt on that particular objection.

  7. Nanoenthusiast,

    Richard put forward the six challenges as a means to foster debate. This debate would in turn be useful to those attempting to develop ‘prototypical’ MNT processes. Importantly, you don’t need to be a proponent of molecular nanotechnology (in the Drexler sense) to work on problems which underpin MNT and molecular manufacturing. That is, perhaps somewhat paradoxically, I don’t agree with your statement that you must convince people that MNT is possible to get people to work on projects that could either prove or disprove MNT . Don Eigler – a very strong opponent of the molecular manufacturing concept – has carried out a series of experiments which have immense significance for the development of MNT.

    “So you could say that this is more a stochastic (or soft,wet) manufacturing paradigm compared to MNT”.

    I agree entirely re. the stochastic nature of the process. But my point is that MNT is deterministic to its core. Diffusion-driven processes where the trajectories of the atoms are not controlled by an external (or internal), pre-programmed device cannot be described as MNT. The first step on the road to MNT (working from a ‘hard, dry, diamondoid’ perspective) as I see it will be to demonstrate computer-controlled epitaxy in 3D. This is the challenge that is going to occupy me for quite some time! (By the way, it’s also perhaps worthwhile to note that the Kelly, Tour et al. experiment was carried out in UHV so “soft, wet” is not really an appropriate description in this case).



    What an interesting comment re. Drexler’s development of the ‘hard, dry’ approach to MNT – thanks for this. Interesting that you consider the hard MNT concept to be dead!



    On the points raised in your first paragraph, I briefly discussed the issue of mechanosynthetic doping with Chris Phoenix (CRN) some time ago. This is a very interesting topic and, again, is something that I’m keen to pursue from a fundamental solid state physics point of view. It requires, however, the development of protocols for transfer of a dopant atom into a deterministically “grown” nanocrystal. Not an easy problem. (This is particularly true in 3D – in 2D one could think of laterally manipulating a dopant atom across a passivated surface into a dangling bond site created via H extraction).

    It only requires placing a carbon dimer (somehow) on the sp2 diamond surface … . That “somehow” qualification is all-important! I must admit that I’m with Drexler on this particular point. In my opinion, by far the most controlled method to carry out computer-controlled atom-by-atom or molecule-by-molecule assembly is to start with a low free energy surface, ‘expose’ radical centres, and then use the resulting dangling bond as a reactive site to which a reactive molecule can be attached. I am, of course, seeing this process from the point of view of a scanning probe microscopist where, to use the Nanosystems ‘parlance’, we’re in the rudimentary ‘boot-strapping’ phase of development. My problem with working with an unpassivated surface is that the level of precision, as I see it, must be much less (as compared to a depassivation – attachment process) because the probability of “dropping” a molecule where you don’t want it to go is much higher.

    The difficulty of relocating the depassivated site is very much alleviated if one uses a local ‘reservoir’ (held at temperatures low enough to quench diffusion) of “tool tip” molecules close to the work area to control the functionality of the tip.

    At this stage, I am focussed on tip structure and functionality. Non-linearities in piezoelectric actuators are much less of an issue for the demonstration of rudimentary mechanosynthesis steps. Working at LHe temperatures, one can quote the drift rate of some commercial UHV STMs in terms of atoms per day!

    Best wishes,


  8. …sigh…

    There’s a misplaced or missing “” in the post above which leads to some extraneous italicisation. Apologies. I think that I’ll leave out the HTML tags in future!

    (Richard, is there any possibility of adding a “Preview Comment” feature to your blog for those of us who are somewhat sloppy with HTML-ing?)


  9. Philip, I think I’ve fixed your html. Maybe its possible to have a “preview feature” but I’m not going to get round to doing it anytime soon! More comments from me later, but I’ve got a talk to finish.

  10. Phillip H., I’m glad Philip M. has replied to you in detail as he knows massively more about that side of things than me. I’ll just add that I think it’s commendable that you want to get on and actually do something practical.

    Philip M., I also was very surprised about the enthusiasm shown for the Tour nanocar by CRN. Sometime ago CRN published a “scorecard” by which you could evaluate the relevance of a piece of research for MNT. The nanocar scores very low even by CRNs own criteria of relevance. I can only conclude that enthusiasm and will-to-believe has overcome critical thinking.

    Kurt, you’re quite right – reading Drexler’s earlier (i.e. pre Nanosystems) work makes it clear just how biologically inspired it was. It’s important to remember just how much progress has been made in biology since the early 80s, particularly in the discovery of the structure of biological machines and in the biophysics underlying the understanding of how they actually work. In the early 80s it was entirely understandable to look at biological machines and assume that their mode of operation is fundamentally mechanical. This view is now, however, out of date, made obsolete by recent discoveries. This has not yet been appreciated by most MNT enthusiasts.

    Nanoenthusiast, the whole point is that the work is not getting done. Freitas is doing interesting theory, and Damian Allis is a very talented young scientist, but their output (necessarily) is a drop in the ocean compared to the scale of effort that would be needed to move this project along. CRN is an advocacy organisation, and as such part of its mission is convincing people that MNT is worth working on. (There’s no need actually to prove it is possible to demonstrate that it is worth working on, of course, as many interesting things often arise from projects that ultimately fail). Currently, Kurt’s perception that the hard MNT project is “dead” is very widely held. CRN will not, however, be able to persuade anyone if it refuses to talk to anyone apart from people who are believers anyway.

    As for Tour’s nanocar, I have no idea how you could integrate such a molecule in an overall atom-by-atom manufacturing system, and I don’t think he does either. As for demonstrating the possibility of a nanoscale machine, that’s not really very exciting in itself because we already know of many other nanoscale machines, both natural and synthetic. The point is whether mechanical engineering design paradigms are appropriate at the nanoscale. Since the mode of motion of the nanocar is diffusive they are clearly not in this case. It isn’t even obvious whether it’s a good idea to put wheels on it, as it were – see the post I made a few weeks ago about Jackie Krim’s work comparing rolling and sliding friction for fullerenes.

  11. To Phillip Huggan:

    I think the most important tool in molecular nanotechnology creation path would be modelling toolkit. In fact, that’s our approach: we can model several hypotheses in the time that would be necessary to carry the experiment and with the fraction of the experiment cost. And only then, basing of the understanding of the underlying system, we plan the key experiments that allow us to make final selection of model.

  12. Andrey –

    As I understand it, Nanorex is attempting this to some degree. Their software is supposedly going to hit the university campus this July for a special college course, and be available sometime thereafter. (Details both of capabilities and delivery dates are disappointingly scarce.) See for all I know on the subject.


  13. JB,

    I’m watching Nanorex efforts with NanoEngineer (actually, I’m in for alpha testing, so I’ll have more information later).

    But, judging on their presentation of Foresight website (not from my experience with NanoEngineer yet!) – its forcefield is pure molecular mechanics and, thus, fails to model _any_ one-electron properties like any chemical reaction, electric and magnetic fields, etc.

  14. Philip and Phillip: This idea of dehydrogenating a patch on the diamond surface and then gently placing two carbon atoms on the patch using some kind of tool tip has me a bit confused. Once you’ve created your dehydrogenated sites wouldn’t it be easier to pump in some gas (maybe ethene or ethyne) and have it absorb on the desired sites? It seems to me that with a few steps of dehydrogenating patches and pumping in gas it should maybe be possible to grow a single layer of diamond, then you could start all over again. What’s the advantage of using a tool tip to place your moiety mechanically rather than just absoring your moiety as a gas?

    Regarding nanorex, I’ve had a bit of a look at their website and I don’t think that the accuracy of the methods they’re using is sufficient for the kind of claims they’re making. They have some cool movies of nanoscale gears and bearings, but I’m not convinced that they’d actually work if you tried simulating them in a method with any sort of realistic treatment of chemical bonding. If I had some more time I’d like to try running a proper DFT-based molecular dynamics simulation of one of their bearings, to see whether it would really work as advertised. From what I’ve seen so far the nanoengineer program seems to be more suitable for hobbyists and fanboys than for scientists.

  15. Hugo,

    Exactly the model levels puzzled me most when the NanoEngineer was first mentioned on Foresight blog. I’ve tried to discuss this (e.g. chemical instability and reactivity of ther bearings) but the talk finaly declined to statement that later they will calculate most interesting _parts_ of structure via DFT (QM/MM, I suppose).

    So far, just got the email that they’ve released Alpha 8 version – will try to check it over the weekend.

  16. (I’m taking a break from public discussions for a while but it is rude to abandon a conversation…)

    Andrey, I agree modelling should precede experimentation and is cheaper. But a whole bunch of modelling must be accomplished before a nanofac parts library can be theoretically “proven”. A single (diamond surface) mechanosynthetic operation has not been achieved. In this case I think experimentally achieving this single target would be worth its weight in supercomputer time. Also, supercomputers are always getting faster/cheaper so it always seems to make sense to do the bulk of the modelling tomorrow. SPM-based tools do not experience the same deflationary forces. I suppose there are more subtle public relations issues regarding funding though, that ultimately determine whether a given $ is better spent renting supercomputer time vs. renting nanotech tools.

    Hugo it sounds like you are describing the conditions for a CVD furnace. The composition of the surface reaction site, composition of the feedstock gases, geometry of the furnace, type of furnace, composition of the gases, pressure; all have to be meticulously controlled before diamond will form as opposed to graphite or just plain nothing. It is expensive to run a CVD furnace and it doesn’t provide the atom-by-atom control needed for some diamondoid products.

Comments are closed.