From: Robert J. B. <bra...@ae...> - 2003-10-20 12:58:05
|
On Mon, 20 Oct 2003, E.L. Willighagen wrote: > Not sure what the problem is. You move around a big blob of atoms towards a > second blob, and then you want to move it to that position where the first > bonds are created? To some extent. The bonds created in most nanoparts (other than most biological nanoparts) are usually covalent bonds. Chemists have lots of tricks for setting up reactions to get them to go the way they want them to. Displaying two nanopart subcomponents and determining what reactions might join them is part of the problem. But let me cite a concrete example. The M&D Fine Motion Controller is intended to be a flexible nanocomponent that might serve as a tip on a complete nanoassembler arm (see Nanosystems pgs 401 & 477). It consists of a number of sub-components (1 tip, 6 extension arms (2 types), 6 control rings (2 types), 1 shaft and 2 2 shaft end-caps). The tip is connected to the arms which are connected to the rings by only a few bonds. The rest of the atoms in the structure have a high covalent bond density with the exception that the rings are not bonded to the shaft and are free to rotate. The entire nanopart is ~2600 atoms -- about an order of magnitude beyond what chemists would currently try to synthesize. However, the subcomponent pieces are small enough that you could hand them to a retrosynthetic analysis program and come up with a plausible method for synthesizing them. One would then want to design the reactions that would for example join the arms to the tip, then join the other end of the arms to the rings. The tricky part will be figuring out how to slide the rings onto the shaft. (Or perhaps one might want to figure out how to bind all of the rings and shaft together during the synthesis process I'm not sure.) It is *not* an easy problem. But it is one we can currently think about. It is also an easier part to consider the synthesis of because it doesn't have the circular structures that buckytubes and the neon pump and differential gear have. Those have relatively high bond stress and I think will be much trickier to synthesize. (IMO, I think we are going to have to design enzymes or jigs to facilitate the synthesis.) Now, back to the question of the FMC and synthesis analysis paths. If one had not stared at the FMC for several hours it would not be obvious where the points are that have the lowest bond density -- i.e. the points where the chemist would normally prefer to setup the reactions to join the subcomponents. That is why I would like some aspect of Jmol or CDK to have a feature that helps to determine this. That way if someone designs a Nano@Home part and it gets loaded back into the database, then a clever chemical synthesist can pull it out of the database and perform a quick analysis to see where he should focus his expertise. Miguel is most probably correct that a normal graph search approach is probably NP complete -- but computers keep getting faster and so for some parts up to a certain size that approach will work. But years ago there was a method for solving differential equations called the Simplex Method. It involved searching through the phase space of solutions in such a way that one found an optimal solution. However it suffered from the problem that in some circumstances it might locate only a local optimal solution and not a global optimal solution. Then a clever scientist at Bell Labs came up with a geometric algorithm that allowed him to search the entire phase space in a single pass and find the global optima. I think the bond density problem may be similar -- you want to search the volume of the molecule with a much smaller volume -- one that might contain only 2 or 3 bonds instead of 4 or more -- and identify those volumes in the molecular space which are bond deficient. The only thing I don't know is how one might deal with the various orientations your small volume might need (for example how many orientations a cube might need) to perform an effective search. Or perhaps the shape of the volume and its orientations might be different depending on the atom type or bonding characteristics. The basic goal is to help break a nanopart down into subcomponents where retrosynthetic analysis becomes feasible. Robert |