Molecules & ComputersModeling, Design and Visualization
Ralph Merkle, co-chair of this year's Foresight Molecular Nanotechnology Research Conference, talked to Foresight Update managing editor Jane Nikkel about the substance of the conference. He received his Ph.D. from Stanford University in 1979 and joined Xerox PARC in 1988 where he is pursuing research in molecular nanotechnology. Dr. Merkle has six patents and has published extensively, including "Computational Nanotechnology" in the journal Nanotechnology.
Nikkel: This year's conference focused on modeling molecular systems. What was your general impression of it?
Merkle: Overall the conference was excellent. We got feedback from a number of people who attended saying that they enjoyed it a great deal. The mix of topics was something that a lot of people found quite stimulatingthe idea of having a range of different subject areas being discussed at one conference was something people liked. Obviously there was a great deal of emphasis on computational methods and on modeling, reflected by the large number of speakers who were talking about various aspects of computational chemistry, molecular modeling, modeling of the diamond CVD, modeling of mechanochemical processes, modeling of the removal of individual silicon atoms from silicon surfaces. So there was quite a bit of emphasis on the computational aspects of this and I think understandably so.
As we move forward in our efforts to understand what molecular manufacturing is and how to do it, one of the critical things that we must do is accurately model the proposals and analyze a range of different alternatives. The most convenient method available of analyzing a range of alternatives without actually going to the expense of doing experiments is to analyze them computationally and using computational chemistry techniques. I expect that this will be very useful in the future as we review the different approaches and decide on which approach will actually be implemented experimentally or at least pursued experimentally.
Merkle: Charlie Musgrave's talk on modeling mechanochemical processes discussed the fundamental issues involved in modeling the site-specific chemical reactions that would take place on a diamondoid structure. In fact, his talk discussed techniques that are very similar to the modeling of diamond CVD, CVD being chemical vapor deposition. So much of his talk was similar to what would take place in discussions about the growth of diamond today. The difference, of course, is that the reactions that Charlie was proposing and analyzing are reactions in which one of the reactive species was positionally controlled. In other words, the reactive species would be at the tip of some positional device. Whether that positional device was a robotic arm or a Stewart platform or STM or whatever was not specified, but the orientation of his talk was, "given positional control how do you then go about analyzing the specific reactions that would be necessary in the synthesis of a diamondoid structure?" This was evident both in the hydrogen abstraction reaction and in the carbon deposition reaction he discussed.
It was also, I think, the background and rationale for his discussion of the ab initio quantum chemistry methods which are used in this and which are necessary to get sufficiently accurate results -- particularly in terms of the energy barriers of the transitions -- to allow us to say with some confidence that a specific reaction will in fact be useful, and could in fact be used in the mechanosynthesis of a diamondoid structure.
Nikkel: Isn't he part of Bill Goddard's group at Caltech?
Merkle: Yes, Charlie Musgrave is one of Bill Goddard's graduate students and currently is looking for a position, because he will be graduating with his Ph.D. fairly soon. To date he does have an offer from Japan to pursue his research in nanotechnology but does not yet have any comparable offer in the United States, which I think is unfortunate. It is also, I think, indicative of the state of affairs today with respect to research in the United States and research in Japan in this area.
This is an area which has a great deal of potential value; however, the value is something that will be most observable realized over many years, or a few decades. In the United States, most organizations have planning horizons that are relatively short compared to those in Japan. In Japan I think there is a broader recognition that long-range objectives are of value, and also I think in Japan there is greater interest in molecular manufacturing in a broad sense.
For example, in Japan there is a great deal of interest in manufacturing. There is great deal of interest in manufacturing very small things. So miniaturization is of importance in Japan and molecular manufacturing obviously will let you build the smallest possible devices. There is a great deal of interest in manufacturing things of the highest possible quality. Molecular manufacturing will let you build things where essentially every atom is in the correct place. This is the highest possible quality that can be achieved. There is also an interest in manufacturing things at low cost. Low-cost manufacturing is again essential to molecular manufacturing through the concept of self-replication.
Nikkel: Wasn't one of the conference speakers in the modeling area from a Japanese research project?
Merkle: Yes, Makoto Sawamura was from Japan and I was very pleased to hear his discussion. Obviously, they are very interested in silicon. The research work that he was reporting on and is participating in -- the activities of the Aono Atomcraft project -- were aimed at analyzing reactions on silicon. Silicon, of course, is very valuable in the semiconductor industry today and so I think that the Japanese research, although very basic, is aimed at improving our ability to manipulate and modify silicon surfaces with the expectation that the improved ability will have value, perhaps many years or even a few decades from now, but will have value in the long run.
This is actually quite interesting, but I think in Japan there is a concept of basic technology as opposed to the kind of basic research concept which we have in the United States. In basic technology, you are pursuing technological goals and technological objectives, but with long time horizons so you are no longer tied to a specific short range project. Nonetheless, what you are pursuing is research where there is an expectation that there will be practical benefits in the long run.
Bill Goddard, Charlie Musgrave's advisor and head of the Materials and Molecular Simulation Center at Caltech, also gave an excellent talk. Bill participated in the "Can today's computational chemistry model tomorrow's molecular machines?" panel discussion and provided an excellent background to the kind of things that can be done, to the kind of things that can be modeled, and to the broad capabilities of computational chemistry techniques -- and could have been a panel by himself. Bill is very well-known in the field and has a broad range of knowledge and understanding, and I think his contributions to the conference and the panel were quite substantial.
Nikkel: What was the answer to the question posed to the panel, i.e. "Can today's computational chemistry model tomorrow's molecular machines?
Merkle: I think broadly the answer is yes. One has to qualify that by saying that there will be some molecular machines in the future whose performance is a little bit difficult to assess now, because we can't quite tell whether it will work, but we can adopt a conservative position and when we are unable to reliably say that the device will work, we can say, okay, we will look at that problem again in five years or whenever our techniques improve so we can answer that question. We'll confine ourselves to systems where we can model them with sufficient accuracy to show the particular device in the particular context will perform the desired function.
I also thought that Ian Foster and Rick Stevens gave an excellent talk. They created a great deal of excitement with their concept of a "nanotechnology colaboratory" which looks like it is a very good idea. It is a method which would allow people across the United States, indeed across the world, to cooperate over the Internet on the advancing research front which is carrying us toward molecular manufacturing. I think the idea of having a collaboration among researchers which is fostered by email, by anonymous ftp of relevant files, and by using the computational and networking capabilities that are now so common, is a very good one indeed. They were clearly interested in furthering this. The informal discussion that evening after their talk was quite exciting.
Nikkel: They were from Argonne National Lab. Didn't quite a few national lab people attend this time around?
Merkle: There was quite a lot of interest on the part of the national labs. In fact, we had an informal lunch on Saturday where we talked about the interest from the national labs. There is an email mailing list which has been formed, specifically aimed at people in the national labs or national level research organizations including NRL. If you are in the national labs and want to be on that mailing list you should contact Jim Belak (email@example.com).
Nikkel: One major subject at the conference was molecular visualization. There were several speakers who talked on related topics.
Merkle: I think that visualization is another area that is very important. Certainly, University of North Carolina researcher Russell Taylor's description of the virtual reality system was something that fascinated everyone. The ability to actually reach out and "touch" an atom using a virtual reality interface is one which everyone found quite exciting. They are working with a goggle and glove system that will let you both see and also provide tactile feedback. So it provides a very visceral understanding of the system. [Editor's note:Web page of The nanoManipulator at the University of North Carolina - Chapel Hill Computer Science Department]
In fact, he commented that when taking research that was given to them and then presenting it in this mode, researchers that already knew the data but had not used this very vivid method of portraying the data were able to discover new insights -- were able to look at the same data and now see it in a way which let them understand what was going on better than they had understood without this mode of presentation.
Michael Pique of Scripps was also discussing a broad range of representations of visualizations of molecular structures. Both of those were very good talks and were quite central to understanding what is going on. Being able to see, to visualize, to reach out and "touch" the molecular structures is very important in understanding and developing an intuition about how these structures will behave.
That leads us into the design area. Geoff Leach gave a talk on Crystal Clear, a molecular CAD tool that he has written. Part of that work was done here at Xerox PARC, and he is continuing to pursue that work now that he is back at the Royal Melbourne Institute of Technology in Australia. His work is providing the first case of actual molecular CAD tools that are aimed at diamondoid structures.
There are a lot of research efforts aimed at proteins, aimed at amino acid or biologically-based molecules. Obviously, people are trying to understand biological systems and in many cases are now trying to design proteins. But Crystal Clear is the first design tool aimed at diamondoid structures. The expectation is that, in the future, we will be designing diamondoid structures within very different kind of molecular machines. These molecular machines will have stiff parts that interact very rigidly the way you would expect mechanical components and mechanical devices on the microscopic scale to interact.
It is basically a molecular CAD tool -- it does not approach the synthesis problem. It does not ask how do you build a particular molecular building block? It asks instead, what does the molecular structure look like? What does the structure of this diamondoid mechanical component look like? How would you design a particular component? So it provides a convenient set of primitives that allow you to introduce systematic dislocations into a bearing, for example, so the bearing could be bent around into a tubular shape without an excess of strain in the diamondoid material. It also allows convenient specification of a broad range of other complex structures as well as termination of surfaces using a variety of reconstructions with patterns. It is a very broadly based function which lets you edit the crystal structure of the diamond in the same sense that you might edit the structure of text with a text editor. I think it really is quite a unique effort at this point in time. I don't think there are any other software tools that allow you to do this with diamondoid structures.
Of the other issues involved in the synthesis of the structures, the issues of the specific chemical reactions are addressed at least partially in Charlie Musgrave's talk on mechanochemical processes. He was discussing what specific chemical reactions might be used. The issue of the kind of tools you would need to analyze a series of reactions to build a complex structure is just beginning to be addressed.
Carol Shaw has been working on MAP (molecular assembly planning) software. That software is designed to help you map out the synthesis by going through the actual sequence of steps that would be involved in the synthesis of a particular diamondoid structure. The actual structure of the program is very similar to retrosynthetic analysis which is quite common in chemistry. In retrosynthetic analysis you start with the finished molecule and then you consider the reactions that would be involved in taking the molecule apart and work backwards from the final product to the initial components. In the same way, in the MAP software the idea is to start with the final structure and work backwards. The difference between the MAP software and retrosynthetic analysis is primarily in the assumption of positional control which is assumed to be available in the MAP software and strongly influences the kind of reactions that you will consider.
There is a difference also in the particular reactions involved. The reactions involved in synthesizing diamondoid are very likely going to be the kind of chemical reactions we see today in diamond CVD growth. Those reactions typically occur in vacuum, and they involve highly reactive species that -- if they were in solution -- would react fairly readily with the solvent, and so would not be able to react with the surface. In addition, because the reactions involve highly reactive species, the reactive species themselves cannot be allowed to touch the wrong thing. So you have to actually hold or position the reactive compounds away from other surfaces to prevent them from reacting where they should not react.
Institute for Molecular Manufacturing's Marcus Krummenacker was pursuing the design of molecular building blocks and was considering a very different approach. In essence, if we are to develop molecular manufacturing systems, there seem to be two broad approaches: one is to advance the state of the art in scanning probe microscopes and improve them to the point where we can actually start building molecular structures using them, the other is to expand our capabilities in self-assembly and use self-assembly to build up molecular structures. A critical requirement for self-assembly is the ability to design molecules which are relatively rigid, at least in the context of the chemicals systems that you want to be relatively rigid and which have surfaces which are complementary to other building blocks.
If you can design and synthesize building blocks which have complementary surfaces -- and where you can not only make two building blocks which have a complementary surface but also insure that the other surfaces do not have any particular attraction for each other so you do not accidentally get binding of the wrong surfaces -- then you can put the building blocks together in a test tube and they will self-assemble into some structure that you might desire to build. This is a critical requirement for self-assembly, and if we can achieve a combination of the design technology and a synthesis technology which gives you the ability to control the complementary surfaces which are required for self-assembly, then we should be able to design and in fact build rather complex structures.
Indeed using self-assembly it should be possible to design and build simple assemblers, thus achieving what is a basic initial goal for molecular manufacturing. In principle, you could do it without using scanning probe microscopy at all. You would only need positional control at the very end when you built the actual assembly device itself. In practice, I would expect that there is a great deal of opportunity for synergy between the probe microscopy and self-assembly building block based approach that Markus was talking about. The two approaches could be used together, so you might, for example, have a system where you have building blocks that are synthesized by the synthetic technology but which now are positioned using a less accurate scanning probe technology. The building blocks might be relatively large and so lack the requirements for positional accuracy in the positional device, which would be quite useful.
The requirements for accuracy in diamondoid synthesis are quite high. You have to be within roughly the atomic diameter of the correct site which is a matter of an angstrom or two. So thermal vibration in the diamondoid synthesis process which cause errors in the tip position of more than one or two angstroms would be unacceptable. On the other hand, if you were using building blocks that were ten nanometers on a side, then you could tolerate much larger positional errors; and the building blocks, because they would presumably have complementary surfaces, would compensate for those errors by latching on to the appropriate other building block when it was brought to a position that was relatively close to the right position.
Nikkel: So you see these two approaches -- the scanning probe and self-assembly as complementary?
Merkle: Yes. It was one of the things very evident at this year's conference -- more and more, we're able to see how different technologies will come together to make molecular nanotechnology happen.
Jane Nikkel is this newsletter's managing editor and assisted with the planning and running of this year's conference.
Special thanks go to the financial supporters who made the first Feynman Prize in Nanotechnology possible: Marc Arnold and Ted Kaehler. This new prize, in the amount of $5000 in 1993, will stimulate many researchers to take on new projects leading to nanotechnology. Thanks to Carl Feynman for arranging permission to name the prize after physicist Richard Feynman.
Thanks to Prof. Mike Kelly of Stanford's Dept. of Materials Science and Engineering, and to Art Olson and Mike Pique of the Molecular Graphics Society, for bringing their organizations on board as co-sponsors. Thanks also to individuals at the companies who sponsored the meeting: Rick Le Faivre of Apple Computer, Richard Nesbit of Beckman Instruments, Ralph Merkle of Xerox PARC, Jerome Wiedmann of Digital Instruments, Elizabeth Enayati of Fenwick & West, Rich Walker at Biosym, Gayle Pergamit at Nanoscale Progress, Pat Ryan at ARCO, and Michael Gray at JEOL.
More special thanks go to the workers and volunteers who made the Third Foresight Conference on Nanotechnology our best meeting yet. Preparation and on-site efforts by Jane Nikkel, Judy Hill, and conference planner Kathleen Shatter resulted in a well-attended, smooth-running meeting. Co-chairmen Ralph Merkle and Eric Drexler recruited speakers and unified the meeting around the theme "Computer-Aided Design of Molecular Systems."
Critical assistance with the press was donated to the meeting by Niehaus Ryan Haller Public Relations of South San Francisco; Ron Pernick was of major help here.
Able assistance was rendered by our many volunteers, including Robert Armas, Emmanuel Barros, Stephanie Corchnoy, Ed Goldberg, Chip Morningstar, John Oh, Eric Tilenius, Dean Tribble, and Eric Williams. Additional thanks to those who volunteered to demo software: Martin Edelstein (Tektronix's CAChe), Markus Krummenacker (Cambridge Scientific's Chem3D Plus), and Carol Shaw (her MAP software).
Apart from the conference, thanks go to the Foresight members who assist us in keeping up to date by forwarding articles we need to see; these include Fred Andree, Dave Forrest, Bruce Gaber, Jones Hamilton, Bill MacIntosh, Anthony Napier, Jim Palmer, Mark Reiners, Roy Russell, and Alvin Steinberg.
Foresight director Jim Bennett points out that a survey of nanotechnology work is incomplete without Japan.
Wired magazine's current (December 1993) issue carries an article by Charles Platt on nanotechnology entitled "Nanotech Engines of Hyperbole?". This was a brief piece giving a basic description of nanotechnology, and asking, as its title indicates, whether the predicted marvels of nanotechnology are reality or hype. Platt pursues this question by calling a variety of persons around the US, including staff at government laboratories, universities, and individuals and questioning them on nanotechnology. A variety of opinions and results are reported.
When I saw Platt's name on a nanotechnology article in Wired, my hopes had been elevated, as I had enjoyed the previous work of both magazine and writer. Although this was in many ways a positive article, I was disappointed. The article failed in two fashions. The first is that the juxtaposition of positive and negative opinions, although a staple of journalism, would tend to leave the reader with the conclusion that the opinion among informed experts regarding the feasibility of nanotechnology tended to divide evenly into pro and con camps; in fact, the opinion of those who have bothered to study the available information has been generally positive.
The second is that Wired in particular has made a point of looking beyond the borders of the United States and the boundaries of large corporate and governmental institutions to seek a more accurate and representative picture of the world as it is. Platt seems to feel that a survey taken entirely within the boundaries of the US could accurately represent the state of nanotechnology research. However, the problem is that almost all serious funded research on nanotechnology today is being done in Japan. Had Platt dropped an email to Makoto Sawamura at the Aono Atomcraft Project or any of the other researchers in Japan's $200 million nanotechnology initiative to assess their opinion of its feasibility, instead of querying American techno-bureaucrats with institutional incentives to dismiss nanotechnology, he might have obtained a different picture. In a less imaginative author or publication, these faults might have passed unremarked. For Platt and Wired this level of journalism represents a disappointment.
Jim Bennett is president of the Center for Constitutional Issues in Technology and a director of the Foresight Institute and the Institute for Molecular Manufacturing.