Home > Nanotechnology Columns > Center for Responsible Nanotechnology > From Basic Nanotechnology to Molecular Manufacturing
Robots building robots sounds convenient--but how do you build the first one, especially when they're too small to see?
June 26th, 2007
From Basic Nanotechnology to Molecular Manufacturing
Manufacturing at the nanoscale--imposing structure on nanoscale objects--can be done in several ways. One is to build all the information needed into the ingredients, so that they form the desired structure without external manipulation. This is called self-assembly, and it's a promising approach, but difficult because it's indirect. Another way to make nanoscale structures is to use relatively large and very delicate machinery, such as focused ion beams and scanning probe microscopes. This tends to be expensive, slow, and imprecise. The third way to manufacture at the nanoscale, in theory, is to use very small machinery--but that raises the question of where the machinery comes from.
Molecular manufacturing involves the use of tiny molecular machines which are capable of building more molecular machines. Such things exist in nature; in fact, that's what the molecular machinery of cells does--ribosomes build protein using instructions from DNA. However, the natural molecular machines are difficult to engineer: it is hard to make a cell build what you want, especially if you want something other than protein. So the goal of molecular manufacturing is to use artificial engineered machines to build more machines as well as other products. And for highest performance, these machines should be "dry"--not immersed in water--which makes them rather different from biological designs.
Few molecular machines have been demonstrated so far, and most of those are based on biology. For example, molecular motors from cells have been immobilized on a surface and tiny pieces of metal have been attached to them. When the right chemicals are added, the motors turn the "propeller." This is great for research, but it's a far cry from the robot-building-robot goal of mature molecular manufacturing. So how can we get from here to there?
The broader field of nanotechnology has advanced sufficiently that researchers, even when they are not working deliberately toward molecular manufacturing, frequently do work that is relevant to it. For example, Alexander Zettl has built a motor which is not based on biomaterials at all, is not immersed in water, and is not powered chemically. It is constructed by building a bearing out of nested carbon nanotubes, attaching a piece of metal to it, and then using electric fields to move the metal. It is a very mechanical structure, susceptible to engineering and analysis and improvement. But it was hard to build, because it required large machines working at their limits of precision. See
So the ultimate goal of molecular manufacturing is to build small "dry" mechanical-molecular machines that are intricate and well-engineered enough to be used as a complete set of molecular construction tools. And the starting point, today's level of technology, is at the level of individual motors and levers and bearings. Getting from "here" to "there" is not a trivial exercise. Fortunately, there are several possible pathways. Furthermore, a small set of molecular construction tools can build a wide range of molecules--this is one of the things that makes molecular manufacturing workable.
At the level of atoms and molecules, things come in discrete packages. Water and hydrogen peroxide differ by one atom, and there's no molecule that's halfway between the two. Making a molecule requires that each atom be in the right place--but once a molecule is made, it's completely finished--there's no painting, sanding, or polishing required. So the goal of molecular manufacturing is to build molecules directly, making exactly what is desired by transferring molecular fragments onto the object under construction. This kind of direct atomic manipulation has been done with scanning probe microscopes, but only in a few cases, and three-dimensional parts have not yet been built. Biology, by the way, does something very similar: it builds one-dimensional molecules by adding a few atoms at a time onto the end, and the resulting polymer chains then fold up (self-assemble) into three-dimensional structures.
So the question of how to achieve molecular manufacturing boils down to the question of how to make nanoscale devices that can manipulate and transfer molecular fragments well enough to be worthwhile. There are three increasingly stringent criteria of "worthwhile." First, it has to be better than large devices such as scanning probe microscopes. Second, it has to be good enough to build more nanoscale fabrication devices. Third, it has to be good enough to build useful products such as computers, medical devices, or even structural materials. To meet these criteria requires reliable, programmable, general-purpose mechanical/chemical systems.
One approach to building such systems is to design and build a nanoscale device which is capable of doing molecular fabrication operations, but is simple enough that it can be built atom-by-atom by a large, slow, awkward machine like a scanning probe microscope. This will require a simplified device design as well as significant advances in scanning probe chemistry. The Nanofactory Collaboration is currently working toward this goal.
Another approach is to build the first device by self-assembly, for example by making three-dimensional functional shapes out of folded DNA strands. Here again, this requires that the first device be relatively simple. Self-assembled structures tend to be relatively weak, since if they stick too strongly they would tend to assemble incorrectly in the first place and be unable to correct the mistake. And self-assembled structures with large numbers of components tend to have a high probability of error. But it may well be easier to build the required self-assembling ingredients than to place large numbers of atoms one-by-one using a scanning probe microscope.
A third approach is to build medium-small, imprecise devices and then use them to build smaller and more precise devices. For example, scanning probe microscopes have already been built using MEMS (micro-electro-mechanical systems) technology. Smaller systems have some advantages; they may operate more quickly, and smaller volumes may be easier to keep uncontaminated. But it is not clear whether, in general, small systems can overcome the advantages of large systems: large systems are easier for humans to build and maintain, and may be supported by decades of technological development.
It is clear that each of these approaches has problems and limitations associated with it. However, the approaches can be combined. For example, rather than placing individual atoms, which would be slow and finicky, scanning-probe systems might be used to join relatively large self-assembled molecules. This would be a lot faster and require less precision. Likewise, relatively simple MEMS devices might be useful to template self-assembly or to control the working environment for scanning probe systems.
So far, designs for a molecular manufacturing system have not been developed. Without such designs, it is difficult to say (for example) whether it would be most efficient to build a DNA machine to make silica, then use that to make diamond, or whether a more straightforward approach such as direct scanning probe diamond building would work better despite requiring more advanced tools. Computers and software are just about at the point where component designs are within reach of hobbyists; for an example, see the "Machine Phase" blog.
Ten years ago, this kind of simulation would have required expensive supercomputers, and would have been less reliable; it would have required cutting-edge lab work to try to reality-check the simulations. That's not to say no progress could have been made, if sufficient money and effort had been poured into the problem. And it's not to say that the simulations are totally trustworthy even today. But simulations are getting rapidly less expensive and more reliable. It won't be too long before hobbyists have developed some kick-the-tires models that can be used to assess the functionality of various mechanical designs. It will be possible to meaningfully discuss questions such as the precision and capabilities of a particular design, and whether the effort to debug and implement that design will be worth it in terms of contributing to molecular manufacturing capabilities.
As concrete designs are proposed and evaluated, molecular manufacturing discussion will move from guesstimates of when the end result will be achieved, to discussing the comparative merits of specific development pathways in detail. At that point, it will become easier to make schedules, not just roadmaps. And with each increase in predictability, the field should become easier to fund. Once funding rises significantly above the level that has been accessible to hobbyists, we may see far more rapid progress, perhaps leading to a bandwagon effect: what is barely accessible to a hobbyist may be low-hanging fruit to a large corporation, once they realize that it exists. And advances in computers, software, and laboratory equipment are making the fruit hang lower every year.