Many mechanisms and structures in nanotechnology can be built from "large, polycyclic molecules with a family resemblance to bits of diamond"1. One threshold for nanotechnology is therefore the ability to build fairly arbitrary structures of this type. This paper describes a procedure for building such structures in a relatively mature nanotechnology.
One way to imagine constructing diamond-like structures is to consider extending our current techniques for producing diamond itself to small scales. We currently produce diamonds by two basic methods: crystallization from molten metal solvents at high pressure, and film growth from a number of highly reactive gases and plasmas. Unfortunately, both of these appear rather unattractive from a nanotechnological point of view. Molten metal solvents are, after all, high temperature systems. The same temperatures that allow crystallization under thermodynamic control to occur at reasonable rates also ensure that any nanomachinery used to build a diamond-like structure would be quickly destroyed2. The plasma alternative, on the other hand, contains many species which are so reactive that they are likely to be difficult to control. For instance, one species that is considered important in diamond film growth is atomic hydrogen3, one of the most reactive species in chemistry.
A second approach to building diamond-like structures is to attempt to scale down existing bulk technology. This leads to such things as attempting to construct structures by cutting materials, much as one does with lathes or other machine tools. This leads to one of the usual objections to nanotechnology: how can you manipulate atoms when the only tools that you have are other atoms, all of which have comparable radii of curvature and many of which form bonds with comparable strengths. Attempting to carve a diamond substrate with a diamond cutting edge held by a general purpose manipulator might be equally likely to modify the substrate and the cutting edge.
The simple approach to scaling bulk processes does not represent anyone's actual proposal for reaching nanotechnology. To a large extent, it is a reflection of my own early misconceptions on how nanotechnology would work. It is retained here, and later in this paper, as an aid in presenting the actual proposal.
This paper suggests an alternative for building structures that takes advantage of the principle of microscopic reversibility. Any path over the potential energy surface of a system that can be traversed from state A to state B can also be traversed, with the same barriers, from state B to state A4. If we know a reaction that destroys a structure, we can examine the reverse reaction (and related reactions) to see if we can create the original structure. Diamond can burn. More to the point, carbon can burn in fluorine, yielding mainly carbon tetrafluoride (CF4) at 500-700 degrees Celsius5. Obviously, the reverse reaction will not be thermodynamically favored. Consider, however, the decomposition of carbon tetraiodide:
n(CI4) --> Cn + (n/2)I2.
This reaction is thermodynamically favored6, and it ought to proceed in a very similar way to the reverse of
Cn + (n/2)F2 --> n(CF4),
since the molecular geometries and orbital symmetries are identical. There are, however, differences due to the greater size of the iodine atoms. These lead to possible steric problems which will be discussed below. A steric problem is a reaction that can't be done because it requires two atoms to occupy the same space at the same time.
The proposed procedure for physically constructing diamond-like structures is to:
Design the final structure, including all atom and bond positions.
Simulate the etching of the structure, noting each bond broken, noting each departing CF4, and noting all of the intermediate geometries.
Use the intermediate geometries from the simulated etching to construct a simulated reaction sequence which assembles the original structure, bond by bond, from CI4 molecules. Each CI4 molecule is brought in with the same orientation as the etched CF4 molecules departed. (During this stage all of the transition states for each of these reactions should be noted.)
For each transition state of each assembly reaction, design a catalyst that stabilizes the transition state by binding to the transition state.
Synthesize the catalysts.
Some of the reactions add CI4 molecules to the partially completed structure (as opposed to other reactions which, for instance, remove iodine atoms). Treat the catalysts for these addition reactions with CI4 before their use, in order to loosely bind Cl4 molecules to these catalysts, preloading these catalysts for use in the assembly sequence.
Build the structure by applying each catalyst molecule to the partially constructed structure at the proper place in the proper sequence.
Because of the heavy use of custom-designed catalysts in this procedure, I will call this procedure "the custom-catalyst procedure" in the rest of this paper.
The first phase of this procedure is the general problem of design in nanotechnology. A number of researchers are starting to design molecular systems with various desirable properties, typically with molecular mechanics software1,7,8. Because the design particulars will vary widely with the application, this phase will not be covered any further in this paper.
The second phase of this procedure requires simulating a reaction between a structure (and a series of partially etched structures) and an etchant. There are a number of factors working in our favor here. Since diamond-like structures are insulators (or at least semiconductors with large band gaps), any particular etching step is rather local. Only a few atoms in the structure change their bonds at one time and require quantum mechanical analysis. The rest merely shift position slightly and can be treated with molecular mechanics calculations9 or can be treated even more simply as linear springs (for atoms far from the reaction center). Molecular mechanics simulations of epitaxial growth on this scale are now being performed10.
An additional degree of freedom in this phase of the procedure is the choice of where the simulated etchant molecules "land." Unlike simulating true natural etching, where the etchant molecules land randomly, the sites of etching can be selected to maximize control over the simulated assembly reactions. In particular, etching sites can be selected so that the structure is etched away fairly uniformly, avoiding the creation of inaccessible areas or cavities during subsequent growth simulations.
The third phase of this procedure (simulating assembly reactions) is similar to the second in that it requires simulation of a series of reactions with localized reaction centers embedded in a larger structure. All of the comments on the second phase apply. This phase includes careful calculations of the potential energy surface for each elementary reaction (roughly speaking, an elementary reaction is the creation or destruction of a single bond), and determining the precise geometry of the transition state of the reaction.
The fourth phase of this procedure requires that we design a catalyst for each elementary reaction. A wide variety of reactions are catalyzed by proteins in natural systems. Most elementary reactions that would arise from this procedure could probably be catalyzed by a protein designed to bind to the reactions' transition states11,12. One would select atoms that surround the transition state atoms at the proper distance to give maximum stabilization, then design the rest of the protein to create this geometry. Given a mature nanotechnology there will be a variety of other possible catalyst designs, but one wants to avoid designs requiring too many steps to synthesize. An infinite regress of catalysts to build catalysts is not helpful.
It may be helpful to slightly modify the third and fourth phases of this methodology to make catalyst design easier. While the n(CI4) --> Cn + (n/2)I2 reaction is thermodynamically favored, some of the elementary steps in the reaction may not be. If, for instance, the reaction takes place by a two-step free radical chain reaction such as:
(1) I· + R3CI --> I2 + R3C·
H + 15 kcal/mole (unfavored)
(2) R3C· + R3C --> R3C-CR3 + I·
H - 32 kcal/mole (favored)
then the first step is thermodynamically unfavorable. However, the presence of the unfavorable step is not aufficient to rule out this approach. There are many biological enzymes which perform unfavorable reactions by coupling them to favorable ones (notably to ATP hydrolysis), and it may be desirable (and possible) to design enzymes that couple steps 1 and 2 together in an analogous way.
Another possible avenue for improving the procedure is to depart slightly from the reverse of the etching process by adding another reagent to make the elementary steps more favorable. In particular, if one adds acetylene as an iodine sink, one can use the following three-step chain reaction:
(3) I· + C2H2 --> IC2H2·
H + 3 kcal/mole (slightly unfavored)
(4) IC2H2· + R3CI --> C2H2I2 + R3C·
(5) R3C· + R3CI --> R3C-CR3 + I·
H - 32 kcal/mole (favored)
This makes the first two reactions substantially easier to drive than reaction 1 in the previous example. (All reaction enthalpies are estimated from bond energies in reference #6.)
There are some possible problems with these reaction sequences due to the bulk of the iodine atoms, as compared with fluorine atoms. Reactions 1 and 4 have no steric problem because the incoming radical is removing an iodine atom. The radical can approach the iodine from the opposite side of whatever that iodine is bound to. Reaction 3 is unlikely to have a problem because acetylene is linear, allowing the iodine atom to approach without intersecting any atoms other than the carbon to which it will bind.
Reaction 2(=5) can cause problems. Here the carbon radical is attacking a tetrahedral carbon. In the worst case this is a -CI3 group on the surface of the workpiece. A likely transition state is a trigonal bipyramid with the iodines in the three equitorial positions and with the existing bond to another carbon and the bond to the attacking carbon in axial positions. This structure is sterically congested, requiring nonbonded I/I distances of 3.67Å (versus a normal nonbonded distance of 4.32Å) and nonbonded C/I distances of 2.64Å (versus a normal nonbonded distance of 2.93Å). The nonbonded I/I interactions are actually less severe than in the initial -CI3 group, so they favor the transition state. The nonbonded C/I interactions will raise the activation energy for the reaction. This does weaken the analogy with the corresponding fluorine reaction. Fortunately this step is the most exothermic reaction of the cycle, so this may be less serious than at other steps. A full quantum mechanical calculation of the barrier for this step should be done. Note13 that CI4 does decompose at 170 C, so there must be some sequence of reactions that remove iodine from CI4 with reasonably low activation energies.
The fifth step in the custom catalyst procedure is synthesis of the catalysts themselves. This can be done with standard peptide chemistry, using the same feedstocks as are used today, albeit on a smaller scale. The sixth step, binding CI4 to the appropriate catalysts, is a standard chemical operation.
The seventh and last step in the custom catalyst procedure, application of the catalysts to the workpiece, requires a programmable positioner. In a mature nanotechnology such a positioner may be built out of diamond-like components itself. During exploratory nanotechnology an AFM-derived positioner may serve14.
Compare the custom catalyst procedure with a simple approach to scaled bulk processes in the following figure. Imagine building a structure with a general purpose manipulator, with a chemically inert end effector, using stable starting materials, and driving arbitrary reactions using mechanical force.
In the custom catalyst procedure the manipulator arm can be sloppier than the scaled bulk approach can. The catalyst will tend to snap into place due to the forces that bind it to the workpiece. This "self-alignment" happens automatically since the catalyst has been designed to bind to the workpiece/reagent-complex in order to speed their reaction. Another result is that the partly constructed workpiece will often act as a natural jig to hold the reactant molecule in place rather than as an obstacle to moving it to the right place.
The custom catalyst procedure can use a thicker, coarser manipulator arm than the scaled bulk approach can. An arm manipulating a catalyst may be too large to fit into the reaction site at all, yet still function. The "sharp end" of the arm is really the reactant molecule itself, which only needs to be held in place by the catalyst, not actively and precisely positioned by a general purpose rod. A side benefit is that a catalyst that binds well to the transition state is likely to bind reasonably well to the reactants and products, providing a way for the positioner to manipulate these molecules. Using the catalyst as a handle for incoming and outgoing small molecules such as CI4 and I2 avoids having them rattle around at random, and possibly jamming a mechanism we are using or building. An intermediate case is one where the arm binds specifically to the reagent14 but not to the entire transition state.
A nice feature of the custom catalyst procedure is that none of the reactions involved will run into steric problems due to interference from the workpiece (the solid angle problem). As is visible in the following figure, when a solid structure is built from small subunits, the angle of approach for each subunit is very restricted by the subunits already in place. This is a pervasive problem for building solid structures with nanotechnology. A solid is built by adding layers. Each of these layers is built by adding lines of subunits to the steps at the edges of the layers. The edges, in turn, are uneven, with kinks where a line of subunits is being added. The individual subunits attach to the solid at these kinks.
If we tried to use arbitrary reactions to build a structure, some of them, such as SN2 reactions, would require that reactants or products enter or leave the reaction site from the inside of the workpiece. Reverse etching reactions cannot have this sort of problem, since there is a clear path for etching products to leave the solid, and a corresponding path for the reagents for reverse etching reactions to approach the growing solid. The custom catalyst procedure reduces solid angle problems with the reagents, since it uses reverse etching reactions, and with the manipulator arms, since the arms need not fit into the reaction site.
A final advantage of the custom catalyst procedure is due to the choice of fairly unstable reactant molecules. Since the reactions used are thermodynamically favored, the positioner need not exert substantial forces in order to drive the reaction. It is sufficient for a positioning arm to be able to peel a catalyst off the workpiece once a reaction is complete. In contrast, many of the analogs to bulk assembly techniques (such as force-fitting or inertial welding) require large forces at the end effector. These forces may create problems, such as warping a manipulator arm, particularly if the arm is slim in order to avoid solid angle problems.
All the advantages of the custom-catalyst procedure help allow us to use low-tensile strength materials such as protein catalysts to shape diamond structures. The catalysts needed to create or destroy stiff, diamondlike structures need not be as strong themselves. You can carve a block of steel with a wooden toothpick--provided you repeatedly dip that toothpick in acid.
I owe many thanks to B. C. Crandall and K. Eric Drexler for many helpful comments on this paper.
9. M.J. Field, P.A. Bash, M. Karplus, "A combined quantum mechanical and molecular mechanical potential for molecular dynamics simulations," J. Comput. Chem (USA), Vol.11, No.6, July 1990, pp. 700-33.
10. D. Srivastava and B.J. Garrison, "Modeling the growth of semiconductor epitaxial films via nanosecond time scale molecular dynamics simulations," Langmuir, Vol.7, No.4, 1991, pp. 683-92.
11. A. Tramontano, A.A. Ammann, R.A. Lerner, "Antibody catalysis approaching the activity of enzymes," J.Am.Chem.Soc., Vol.110, No.7,30 Mar 1988, pp. 2282-6.
12. D.E. Hansen and R.T. Raines, "Binding energy and enzymatic catalysis," J. Chemical Education, Vol.67, No.6, June 1990, pp. 483-9.
13. CRC: Handbook of Chemistry and Physics 52nd Edition, Chemical Rubber Publishing Company, 1971.
14. K.E. Drexler, "Molecular tip arrays for AFM imaging and nanofabrication," Fifth International Conference on Scanning Tunneling Microscopy/Spectroscopy and First International Conference on Nanometer Scale Science and Technology; Baltimore, 23-27 July 1990.
Jeffrey Soreff is a researcher at IBM with an interest in nanotechnology.