AUG. 24-31, 1998
|SPECIAL REPORT CONTENTS|
Q: What makes buckytubes compelling to industry?
A: Take a look at the preface and introductory sections of Sematech's National Technology Road Map for the Semiconductor Industry, 1997. (see http://notes.sematech.org/97pelec.htm) It's a fascinating document. In a sober, knowledgeable way, it looks at the next 15 years of the semiconductor industry. By the year 2006, if we stay with Moore's law, circuits on semiconductors will shrink to just 100 nanometers across [one hundred billionths of a meter]. The report notes that "there are many areas of technology for which no potential solutions are known," and calls for innovative research to get across the 100-nanometer barrier. I've had meetings with folks at [chip consortium] Sematech. The notion that they will eventually have to leave silicon was discussed in depth. They see so many problems on the horizon that they can't get around. So now they are ready to think about things like carbon.
Q: That's a big departure.
A: Yes. And this gets back to the old dreams of "molecular electronics." The idea is that little transistors, diodes, and other elements on a device will all be individual molecules. This is being discussed very seriously. When you realize how wedded the semiconductor, computer, and telecom industries are to Moore's law, you see that inevitably, within about 20 years, everything will change. There is a huge electronics industry, well in excess of $200 billion a year, with a great desire to maintain Moore's law for another 50 years. It's likely that tens of billions of dollars will be spent on breaking the 100-nanometer barrier. And the only thing on the other side of that barrier is molecular electronics. It's "nano-land," by definition.
Q: What are the main hurdles?
A: At this scale, quantum effects start biting at your heels in a serious way. The only solution is to make quantum effects your friends. That means you have to make entities sufficiently small that the quantum effects dominate over random thermal jiggling. To simplify it in extreme: If you take particles, usually electrons, and you confine them in boxes only a couple of nanometers on a side, the particles can't just sit there. They have to occupy some distinct level of jiggling, called quantum states. If you make the box small enough, the energy between one quantum state and another is much more than random thermal jiggling. You know what the electron is doing.
Q: The nanotube would serve as the box?
A: That's right. This is all to explain why we may need molecular electronics. Once you're below 100 nanometers, a device has to go way down in size to remain stable. It can't be a little under 100 nanometers. It'll have to be more like one nanometer, or about 3-4 atoms across. At the same time, this device has to live in the real world, with air and water around it. So how can we be sure that this tiny entity we so carefully crafted will stay just that, without adding or subtracting one or more atoms? The answer is, it will be a molecule. That means all the atoms stay together, as we made them. They're happy that way, and don't want to change. It's now an entity that maintains its identity as it joggles around with other atoms, just the way water molecules in a glass bump around, but maintain their identity.
Q: How does the molecule become a useful device?
A: If we are ever to cross the 100-nano barrier in electronics, we need to develop nano structures that let electrons move through, as they do through wires and semiconductors. And these structures must survive in the real world of air, water, boiling temperatures. Until recently, we had no clue how to do that. Molecules that were happy at high temperatures were terrible conductors. Wires were a problem. Even with gold, which is inert, if you make a wire just one nanometer in diameter, it beads up into little balls, and gold atoms on the surface of the ball interact with air and moisture. Silicon is no better at one nanometer in diameter. It reacts with air, and makes silicon dioxide or glass.
Q: Buckytubes solve that problem?
A: Yes. In the 1970s, there was much discussion of molecular electronics, but nothing came of it, mostly because people didn't have good molecular metallic wires. But now it looks like we do, and the name is "buckytube." They are fullerenes -- carbon nanotubes, in which every carbon atom is surrounded by three others in an hexagonal lattice. The way we have to make these things is nowhere near perfect. And because of all the imperfections, many of them are cruddy conductors -- not usable in molecular electronics. But since 1996, we've been able to make perfect ones, sealed off on both ends, with no defects. If you launch an electron, it comes out the other end without anything slowing it down. Depending on how its cut, it can be a semiconductor, or an excellent wire. There's hardly any electron loss.
Q: Are such tiny transistors useful?
A: The question is, how would we make a computer memory that has a thousand billion such transistors on a square-centimeter chip? Could we do this using techniques more reminiscent of chemistry than physics? Well, we have good days and bad days. On good days, we figure of course we'll do it. On bad days, we can imagine that silicon will always be the answer, and that there is no way across the 100-nanometer barrier. But a decade ago, we couldn't even have an argument about it, because we couldn't make the necessary wires or circuits. The buckytubes we have now are as rigid as anything we know of in the universe, and they are only about 1.5 nanometers in diameter. In wire form, as long as they are straight, electrons can move down them with little loss. The longest one we've seen is one-tenth of millimeter, or 100 microns. But much of what we are making is probably longer than that.
Q: Is it possible to build a full-blown computing device?
A: We aren't close to doing that. In fact, I think it has to be done through chemistry. We have to put buckytubes in an environment where they assemble themselves into a device because that's what they want to do. It's possible. Chemists find tricky ways of making molecules jiggle around and do what we want, because we've put them in just the right environment. This is called self-assembly, and it will no doubt take years to develop. It is the high science and technology of molecular electronics. The little bits and pieces, the wires, junctions, etc., will all be molecular events. One day, we hope, there will be a fabrication facility that churns out nanochips, where one just appears every few minutes. It'll have 10 to the 15th bits in it, with such a high degree of reliability that every chip can be sold.
Q: Will these chips be cheaper than today's chips?
A: Yes, in our dreams. Moore's second law says the cost of fab facilities rises exponentially. That means maybe $30 billion to $100 billion per fab in 2020. Then you get the problem that eventually, the fab costs more than the total output of the industry. Right now every last detail in a Pentium chip is crafted out of a hunk of stuff. Like Michelangelo creating a sculpture. Every last bit of chip removed by lithography. Well, the chemistry in every cell of the body doesn't do that. The more we understand what happens in living cells, the more incredibly powerful you realize things can be when they work from the bottom up, by interaction of one molecule and another.
There may be some reason why it's impossible to build a Pentium by randomly jiggling molecules. Maybe the universe wasn't designed to let us build computers at a molecular scale. But right now, we see no reason why it should not be possible. I think it's inevitable that tens to hundreds of billions of dollars will be spent over next two decades trying to make such things happen. Then, the world will be dramatically different. Fabrication will be a lot faster and cheaper. And the entities we make could be incorporated in your clothes. You'll have tremendous power in tiny places, and all of it made very cheaply.
Q: Can you say a little more about self-assembly?
A: For things to be made through self-assembly, they must be able to jiggle, rub against each other, and stick where you tell them to stick. Big old things like Pentium chips don't jiggle around. But if you can put the atoms where you want, you can engineer behavior to achieve amazing, exotic results that wouldn't be possible without control over those atoms.
With buckytubes, you have incredible strength, perhaps 100 times that of steel, and conductivity better than copper. That's inherent in the carbon atoms you stick together in just this way. If nature had to build a living cell, with macroscopic globs of glue, it would never work. Nature -- how, we don't know -- has technology that works in every living cell, and that depends on every atom being precisely in the right spot. Enzymes are precise down to the last atom. They're molecules. You put the last atom in, and it's done. Nature does things with moleculer perfection.
Could we train bacteria to make buckytubes? We'd dearly like to. But there may be some fundamental reason why no organism that lives in water will ever be able to make a buckytube. These buckytubes are creatures of the dry world -- the world that nature has not managed to access so far by random evolution. Maybe we are nature's way of making buckytubes.
Updated Aug. 13, 1998 by bwwebmaster
Copyright 1998, Bloomberg L.P.