Editorial Note: This document is retained mainly for historical reasons. Trying to answer the raised questions led to the paper "Steps Towards Molecular Manufacturing" which discusses design rules for molecular building blocks in some detail.

Assuming that a positional synthesis method of some sort, based very likely on an AFM (Atomic Force Microscope) with a MTA (Molecular Tip Array) [Drexler 91], could be used to construct a first protoassembler and other large structures in an aqueous solution-phase, a detailed analysis of how universally employable building-blocks should look like and to what kinds of constraints their design is subjected, will prove very valuable. The questions that will have to be addressed will cover aspects of modularity, ease of designing with these building-blocks, assembly strategies, and feasibility of the chemistry used to bond together the building-blocks. It seems like some tough and non-trivial problems are in need of an elegant and sophisticated solution. The main problem will be to find a method to knit together the building-blocks in a three-dimensional and covalently bonded structure in order to avoid the complexities of designing cascades of conformational changes to achieve three-dimensionality by means of non-covalent adhesion, otherwise known in biochemistry as the "protein folding problem". The linking of building-blocks into a 3D structure will pose severe steric problems that will have to be examined in close detail so that the success rates of chemical reactions will be high enough to definitely obtain the desired product after up to 100,000 consecutive steps.

presented at the

Second Foresight Conference on Molecular Nanotechnology
Palo Alto, 7-9 November 1991

Author:

Markus Krummenacker
620 Barker Hall (Neilands lab)
Dept. of Molecular and Cell Biology
University of California
Berkeley CA 94720

Tel.: (510) 642-7452
FAX: (510) 642-7846

(Please note that the above address is obsolete. It was current in 1991.)

Molecular Building-Blocks

Regardless of the choice that will be made concerning the specific construction method that will be used to build a first protoassembler and other, more advanced molecular machinery, a detailed analysis of how universally employable molecular building-blocks should look like and to what kinds of constraints their design is subjected, will prove very valuable. Envisioned is a system of modular components which can be combined flexibly in any desired fashion, something not unsimilar to a molecular version of the ingenious toy "LEGO(TM)". In the following, an array of questions is suggested that should be kept in mind while creating and evaluating designs of such building-blocks. The different aspects can be divided roughly into three categories:

1. Modularity/Flexibility/Usefulness - issues: are concerned with the easiness of designing large structures and the limits of the building-blocks regarding universality (what desirable structures could not be built ?).

2. Chemistry/Feasibility - issues: discuss yields and error rates and possibly energetics and waste production (economy in utilization of resources).

3. Grey Zone, where both former issues are so intricately connected to each other in trade-off issues, so that they can't be separated.

1. Modularity/Flexibility/Usefulness

- 1.1. Granularity

How large should building-blocks be ? If they are larger, fewer blocks and fewer synthetic steps would be needed to build large scaffolding structures but a diversity of different angular geometries and finely detailed architectures would be more difficult to cover.

- 1.2. Amount

How many building-blocks of a given granularity would be needed to build an assembler with the desired minimal capabilities ? 10,000 ? 100,000 ?

- 1.3. Rasters/Symmetries

Is it reasonable to specify an overall (3D) raster-geometry over the machine to be designed ? How easily would building-blocks be able to fit into this structure ? With zero strain ? When would one have to employ specialized units in order to break out from the raster to realize special ("unorthodox") device geometries ? How could movable parts be incorporated into a largely rasterized structure ? How much would a raster restrict one's flexibility to switch to building-blocks of a different class of rasters ? Interface-units between differing rasters ?

- 1.4. Exchangeability/Expansibility

Can certain building-blocks be designed in such a fashion that they could be easily replaced by a close cousin without having to change anything else in whole structure ? Such replacers could supply exotic functional groups that could provide additional catalytic capabilities once it is proven that the overall design works as anticipated.

- 1.5. Quantifiable ?

Is the difficulty of a design process using building-blocks of a given class quantifiable ? Could one conceive a measure of how easy it is to cover a diversity of desirable geometries using certain building-blocks ?

- 1.6. Rigidity/Mechanical Properties

How stable towards applied forces and torque are the specific building-blocks and especially the large structures built from them ? How much would the rigidity of well-designed, nearly strainless cagelike skeletons in the building-blocks become compromised because of strain introduced by knitting them together into larger structures without having perfectly fitting geometries ? Would such "higher-order" strain effects add up to a dangerous amount ?

2. Chemistry/Feasibility

- 2.1. Types of Chemistry

What chemical reactions could be used to link building-blocks ? Probably only well known and preferably extensively documented reactions. But how much does that restrict the array of possible choices ? Will these reactions work in an aqueous medium ? Most organic chemical reactions seem to have been investigated mainly in pure organic solvent non-water media. Could these reactions still be used by supplying some kind of reaction-enforcing tools that largely emulate an organic environment much like nature herself seems to employ this as a principle in her enzymatic catalysts ?

- 2.2. Reaction Geometry

How tolerant are the reactions to small amounts of misalignment of the attacking components ? Can a reaction geometry close to the ideal case simply be achieved by using a given set of rasterized building-blocks or will one have to implement "unorthodox" device geometries ?

- 2.3. Success of Reactions

Is it possible to calculate the mean yield for specific synthetic steps and to grade them according to efficiency ? Can the yield for individual steps be made so high that it becomes possible to link the necessary 10,000 to 100,000 consecutive chemical reactions to obtain with certainty a functioning product ?

- 2.4. Error Rates

Do certain reactions pose the danger of easily slipping into a side reaction ? If this happens, would it be impossible to reverse this in order to free the misplaced building-block again ? Here arises a question that is closely connected with the issue 2.1., the type of chemistry used to form the links: What is the trade-off of benefits in reversible versus irreversible chemistries in the light of error correctibility ? If it would turn out to be extremely difficult to exert very precise control over the placement and alignment of reactants, especially when trying to establish two bonds simultaneously, a reversible type of chemistry would allow the new building-block to settle in the designated place in a minimum-energy fashion and if it happens that the wrong bonds accidentally get formed, it probably would be possible to pull out the misplaced building-block again and give it another try. On the other hand, the added unit has to stay in place securely until other stuff is built over before it could get a chance to hydrolyze off on it's own. With irreversible chemistry this problem will not occur but if a wrong bond gets formed it probably will not be possible to correct this and so the whole structure built so far has to be essentially discarded as garbage.

- 2.5. Cycle Time

How long does it take to weld a bond? Would one usually have to go through several trials, several unreactive collisions until finally the chemistry snaps things into place ? Is the cycle time directly related to the tolerance in reaction geometry and the energetics of the reaction or is there a more complicated dependence ? What is the time needed for the actual reaction process compared to the overhead costs, like the time required to select or bind a next building-block and to reposition the construction center to a new location on the substrate ?

- 2.6. Product Characterization

How would one check if a synthetic step has occurred in the anticipated way and how long would that product characterization procedure take compared to actual synthetic steps ?

3. Grey Zone

- 3.1. Functional Groups

Should the functional groups that are attached to the carbon skeleton of building-blocks be strictly classified in two categories:

- link-forming groups that are needed to knit building-blocks into the larger structures,

and

- additional decoration, e. g. to render the large structures solvable in water or that participate in interactions in a receptor site but are not of structural importance ?

The other possibility would be to employ generally useful functional groups that could serve all these purposes, and only the choice to use an individual group for link-formation would give this chosen functional group structural importance.

- 3.2. Recognition

What is the minimum number of different building-block types that is necessary to cover the diversity in geometries and functional groups required to assemble a functioning large structure ? Would it be reasonable to provide each building-block with a separate domain reserved only for recognition purposes by the receptor site of a tool ? Even though that this would be a very modular, flexible and adaptable approach, it possibly could almost double the size of the individual building-blocks and so half the mass of the final product structure would essentially be functionally useless, a ballast of former handles which were needed to put the individual components in their place. Or one could use the decorating functional groups on the building-blocks for the identification but then the freedom of group placement is much more severely constrained. In any case one will have to watch out that the groups intended to be used for link-formation always remain exposed and are clearly accessible.

A different approach would be to not recognize the building-block itself that is to be incorporated in the growing structure, but a supporting handle, onto which the building-block is fastened. This could be coupled with at the same time activating a link-forming functional group of the block. So then the growing structure would be attacking the functional group on the building-block, automatically displacing the handle including the whole tool plus constructor arm as a "leaving group". After this reaction one just would have to make sure that the left-over handle, which has been now converted into garbage, gets removed from the tool within a reasonable time, so that it does not block progress and the binding of a new unit, which otherwise would resemble the problem that traditional enzyme researchers would call "product inhibition".

- 3.3. Linking Strategies

Assembling three-dimensional, covalently bonded structures in a precisely prescribed fashion will create difficult problems without precedent. If only one bond is being established per added building-block, the process would resemble something more like a linear polymerization as is seen in the conventional cases of nucleotides leading to nucleic acids and of amino acids leading to peptides. A more three-dimensional way of weaving the net has to be found. Could merely non-covalent interactions possibly hold together large parts of the structure ? Correctly self-folding strings are very difficult to design because of the combinatorial explosion of possibilities. Covalent bonding strategies which are designable in advance seem to be more promising. Is it then possible to forge two or more bonds at the same time, in the same reaction cycle ? Otherwise one would have to reach down into a sterically very crowded environment to tie the other links between the structure and a newly added building-block. In which case one would have to calculate with about two or three synthetic steps per unit added. With how many links should a block be tied into the structure anyway ? Probably not very often less than four, which is at first sight the minimum number required to obtain continuous 3D structures.

- 3.4. Tools

What are tools supposed to do ? Probably receptors are going to be needed to specifically bind and hold on to building-blocks or their handles. How many different types of receptors will be needed ? As many as there are different building-blocks ? Or could certain classes of building-blocks share a receptor through sequential exchange of the liquid reaction medium ? But should the receptors' only purpose be merely presenting the right units ? Or could or should they rather be designed as reaction-enforcing, trying to shield the reaction from interfering water molecules, or even supply functional groups that would catalyse the reaction ?

Epilog

These are the kinds of questions that one will have to consider closely in order to come up with a design of building-blocks which is sufficiently useful for building a first protoassembler and other large structures.

Additional Illustrations for some of the Issues

- 1.3. Rasters/Symmetries

and

- 3.3. Linking Strategies

It appears to be feasible and reasonable to specify rasters, overall lattice structures, according to which all constructional activities are to be outlined and oriented. The need for rasterization arises out of the problem to link building-blocks with as little strain as possible, in order to a) not deliberately relinquish structural rigidity, and to b) be able to form the links at all, because chemistry needs precise alignment of the components, which becomes very difficult if the individual parts do not fit into the framework, so trying to "bridge the gaps" and attempting "final ring closure" operations in strained systems would cause a lot of headaches. By exploiting symmetries in the building-blocks and in the construction process, it should be possible to mutually cancel irregular angles introduced by particular chemistries in the links.

A very important question concerns the exact number of links that have to be established to integrate a building-block into a growing structure and in what sequence these bonds should be made. The first thing that comes into mind is that this number should be as low as possible. The lower the number, the fewer are the necessary chemical reactions and the higher is the yield of the final product. In addition, the difficulties of steric hindrance are relaxed a little bit if fewer bonds have to be established after the first one connects the newly added building-block with the structure.

Quite a nice solution is being offered with four bonds which are arranged in a tetrahedrical geometry. Nature shows that it is possible to link such units into infinite and regular 3D-lattices in the substance diamond, amongst others.

But probably one could even use units with only three links. In a very nice analysis of 3D-networks [Wells 77] it is being argued on p. 26-27, that in order to form a regular 3D-lattice, a repeat-unit of this lattice has to have six free links for hooking up with neighbors, arranged in three pairs of two diametrically opposed links, the link pairs being parallel respectively to three noncoplanar axes. But the repeat-units themselves can be built up from smaller units, for example by using two building-blocks capable of forming four links (the diamond lattice again in the case of tetrahedra being employed), or by using four building-blocks capable of forming three links, leading to a repeat-unit as illustrated symbolically in the figure.

The chemistry of the links indicated should also just be taken symbolically.

A macromolecular structure would be assembled not by adding these rather large and elongated, unconveniently shaped repeat-units, but by adding the smaller constituent building-blocks one at a time. On the average, one thus would have to form 1.5 links per block added, so one would alternate between forming one and two links per block. In this way, one has to go into the trouble of making two bonds at the same time only with half of all the building-blocks used.

Here an important trade-off issues arises: the obviously beneficial desire to use building-blocks with as few links as possible leads to more open and loose structures which are less densely interknitted which also facilitates the access to individual bonds. But those structures tend to show less mechanical rigidity and one will have to make sure that building-blocks at the construction site, which are tied into the growing structure by so far only one completed bond, do not obtain too many degrees of rotational freedom so that they still can be easily aligned for subsequent reactions. This trade-off issue has to be analyzed in closer detail before the optimal rasterization procedure can be conceived.

Literature

MTAs (Molecular Tip Arrays):

[Drexler 91] K. E. Drexler: "Molecular tip arrays for imaging and nanofabrication"
J. Vac. Sci. Technol. B9(2), (1991) p. 1394-1397

AFM-based Positional Synthesis:

[DreFos 90] K. E. Drexler, J. S. Foster: "Synthetic tips"
Nature 343, (1990) p. 600

Regular Lattices and Rasters:

[Wells 77] A. F. Wells: "Three-dimensional nets and polyhedra"
John Wiley & Sons, New York, (1977)

[Seeman 82] N. C. Seeman: "Nucleic Acid Junctions and Lattices"
J. theor. Biol. 99, (1982) p. 237-247