Future Current

Tihamer Toth-Fejel is a senior research engineer at General Dynamics Advanced Information Systems. He is a member of the advisory board of the Nanoethics Group and chair of the Society for Manufacturing Engineers Nanomanufacturing Technical Group. At various times in the past he has been a fellow for the NASA Institute for Advanced Concepts and on the scientific advisory board of Nanorex Inc. At the At the CRN conference on the Future of Nano and Bio he went through various nanotechnology applications that could potentially be used together to build a nanofactory for molecular manufacturing.

The following transcript of Tihamer Toth-Fejel’s CRN Conference presentation “Building a Nanofactory” has not been approved by the speaker.

Building a Nanofactory

I’m a research engineer. Therefore, by definition, I don’t know what I’m doing. If I knew what I was doing, it wouldn’t be research. You’ve heard a lot of presentations on DNA assembly, atomically precise manufacturing, and mechanosynthesis, and I will be touching on some of those issues in my talk. But how do you use these things to put it all together?

First, I should start, I first found out about nanotechnology when I was in grad school. I met this guy with long hair and he was talking about putting together all these nanorobots, putting them inside our bloodstream. This is 1978. I had just gotten back from Cuba and I was very disillusioned with the wrestling I had been doing because I had found out that it was not a game, just a war fought by other means. I was ready to go out and save the world, and that was how I met Eric Drexler.

At first I thought he was nuts. It took him five years to convince me that the nanotechnology he was talking about–in fact I don’t think he had coined the word yet–that this stuff was real. In the conversation I had with him, this was at Barbara Marx Hubbard‘s mansion after one of the meetings on space colonization, he was talking about this stuff within the next thirty years.

The thing that has plagued the whole nanotechnology idea for a long time is nomenclature. We have to define our terms. So I’m going to start with looking at the National Nanotech Initiative definition of nanotechnology as anything from 1 to 100 nanometers. They’re interested in unique properties. The difference in where Eric is coming from, especially being a mechanical engineer, is that he is interested in machines. How do you build finite state machines. This has been an interesting dichotomy that seems to have plagued the whole field for a long time.

Now, this past year I decided to talk to the chemical engineering department of the University of Michigan, which actually had a course in nanotechnology. I took their “Fundamentals of Nanotechnology,” and most of the classes I would be arguing with the professor about his concepts. He was a materials scientist, a really bright guy, and hopefully we will be collaborating on some projects. But his idea of nanotechnology is from a chemical point of view. The difference is that, the top-down approach is what chemists do. They basically throw stuff in a cement mixer and expect to get watches out. The bottom-up approach is what we’re trying to do. We build small tools that manipulate parts that are about the same size as the tools themselves. This is what the chemists just don’t quite understand yet.

When you talk about a nanofactory, you have to look at what factories actually are. Factories are not robots. They are actually much simpler. But because the four step process starting with the blacksmith, which Jim explained, we have things that are going to be similar between a factory and a nanofactory. You have concepts like mass production and interchangeable parts. But down below five nanometers, you are going to have electrical properties and leakage from tunneling effects of electrons, and you also have the exciton distance between the electrons and the holes for pure silicon which is going to be around five nanometers. When you have some interesting structure that has not been characterized, we don’t even know what the exciton distance is. That is going to make the electronic properties a little difficult. In situations like that it is important to do the science so that you can predict the behavior of the devices you are going to build. At this point, just about everything we know about for nanofactories are additive. There are no subtractive layers, unlike in today’s factories where much of the processes are either additive, in which case it is assembly, or it’s subtractive. It takes different types of cognitive tools to think about things in that direction.

One of the things I think NNI is doing right is that they are realizing that if you cannot measure it, you can’t build it. You have to measure it quickly, cheaply and repeatably. Otherwise, you can’t build a manufacturing process around it, and you can’t sell it. So there are a whole bunch of concepts that come along with when you actually talk about a nanofactory that people don’t have to worry about in the labs. When you are looking at measurement on the nanoscale, a lot of the concepts from metrology change, because at that level accuracy and precision become the same. Accuracy has to do with how close your measurement is to perfection, and precision has to do with how good your measuring tool is. Other concepts such as surface finish do not really make any sense at the nanoscale, while some things like statistical methods will still be the same, unless you are going to be able to measure 10^23 atoms. A lot of things will be different when you are trying to build nanofactories.

One of the interesting parallels that I think people have talked about today is that you have a 3D printer. The idea is that a 3D printer is a very promising technology and there are two things that need to happen. One of them is that it needs to have higher resolution, it needs to have more kinds of input materials, and there are people who are working on these problems. The other problem is with orientation. When you are printing something with the inkjet printers we use for our documents, the orientation of the particles is not important. But when you are talking about a nanofactory, the orientation is important.

There are a whole bunch of options in terms of how we get to a nanofactory from where we are right now with 3D printers, nanolithography, and other technologies. In engineering you put together modules. With the nanofactory using convergent assembly that Ralph came up with so many years ago, you have modules that you put together. It turns out that additive layers are just as fast, but the important thing is that you do have modules.

With mechanosynthesis, your modules are very small. When I worked with Chris on the NASA project where we were trying to design a nanofactory we focused on trying to make the blocks bigger. We don’t have mechanosynthesis right now. What we have is ordinary chemical engineering. How big of a block can we build that is relatively stiff? If we use DNA, we can build something fairly large, but it’s floppy.

Other people have been looking at what you could build if you had some sort of large molecule. Sharon Glotzer at the University of Michigan assumes that you can build nanoparticles of a precise size and orientation in terms of electric fields or just in terms of shapes, but usually in terms of polarity. She was able to simulate, using basically string and block constructs, the behavior of conventional diblock copolymers. Depending on what size and density you mix these “diblock copolymers” you can get different types of large scale behavior. That is the self-assembly path. This is the way that chemical engineers and chemists think. The problem is that this is not really what we need. This is useful for some things, but if you are building a computer, medical devices, or Robert Freitas‘s reciprocytes, this will not work.

What we have here are simple modular robots that pick up things and put them together. Each module does not assemble itself, but the system can make copies of itself. This type of self-replication is very different from self-assembly, but this assumes that you have got a finite state machine. Self-assembly, all it means is that you can do some orientation with respect to each other. It depends on weak forces, and there is no such thing as air detection, self-tests or anything like that.

If you are trying to build yourself a module, you have to look at what requirements we have today. You are going to get something between one and ten nanometers in terms of building blocks. You want something that is orthogonal so that you can build 3D structures. This is a very useful paradigm for building regular 3D structures. You have to start with a building block. Here I am using a silsesquioxane nanocube. Basically you have got silicon on the corners with oxygen connecting them. This was first invented in 1930. In the last ten years there has been a lot of development by the defense department, because one of the things you can do is add organic groups to the outside and do this under fairly controlled conditions. What you actually have here is a silica core with organic corners that can combine the properties of both.

One of the things that happens is that generally the temperature at which this degrades goes up quite a bit. This is stable at around 500 degrees C. This is a nice nanocube, but what if you could build a perfect nanocube? A perfect nanocube is something where not only do you control what goes on in each one of these corners, but you can control what happens to these spaces. If you could do step-wise hierarchical construction, then you could throw blocks like this into a cement mixer and get watches out, as long as you do it one step at a time.

Let me see how you might be able to build that. The first thing you do is start out with your ideal silsesquioxane nanocube. Instead of putting organic groups on the outside, you put another silsesquioxane. You have to have links in between the nanocubes, and then you can start growing them just like you do dendromers.

Now, there are a lot of other requirements. You need to be able to control the lengths of the links so that all your different types of nanocubes are the same size. Then we use something that we borrow from DNA synthesis. You basically rip off this DMT molecule, and at the same time rip off this DIP molecule, then you get them to connect. The important thing about this reaction is that it is step-wise. In other words, you only connect one of these things. Once you have these perfect building blocks, then you could build complex structures. The problem is, we can’t build them. Right now, trying to build six different independent chemical reactions is very difficult to do and you would also have to control exactly where they are. The chemists are simply not up to the task.

One other problem you have is how to connect the nanocubes to each other. The other thing you need for a nanofactory is a molecular actuator. There are a lot of them, it turns out. A lot of people are working on what kind of molecular actuators there are. I personally like the rotaxane dimers because they just look the coolest. The problem is that the yield is quite slow. The other thing that is really nice about them is that they can make these things contract and expand by using different wavelengths of light. If you could do that, you could build a modular nanocube motor.

The important thing about this for a nanofactory is that it is modular and you have interchangeable parts. If you want to build a nanofactory, you need interchangeable parts. Another way to build interchangeable parts if they are molecularly precise is the stuff that Jim spoke about earlier. The problem, as he pointed out, is that DARPA wants us to do a lot of them at once. DARPA is finally getting interested in this stuff. They would actually like to see people using arrays of probes to actually build things. It’s called tip-based fabrication.

They are asking for six-probe tips, a really low number, which is interesting because Mirkin at Northwestern has built 55,000 tips. It only took him thirty minutes to make 55,000 copies of that nickel over there. Each one of these features is 80 nanometers across, which is actually fairly large. Compared to diamondoid mechanosynthesis, which needs a resolution of two nanometers, this is actually still quite large. On the other hand, it’s got 55,000 tips. He’s not prying individual atoms off or adding them. What he is doing is using ink, so the name of the company is Nanoink. They have been able to get down to fifteen nanometers and it’s quite a fast procedure.

In addition to using probes and tips, another possibility is to use pores. In fact, some people have talked about using holes with electrodes on either side and measuring DNA as it gets forced through. This is a different way of using some sort of mechanism to add silsesquioxane nanocubes and have them placed in exactly the right place, in almost a reverse manner to the probe tips that I showed earlier.

I’m going to talk about Paul Rothemund‘s work in a little more detail than anyone else here has. Here is a guy at Caltech that took an M13 virus and did this. He says at one point in his paper that this is so easy a high school student could do it. Sure enough, Mark Sims took his daughter over to Caltech and she actually did this work for Nanorex.

How the actual process works, if you have not read the paper yet, you’ve got an M13 viral genome made of single stranded DNA, and then you have little pieces of DNA that hook up in particular places. That is the reason that you use DNA: it has such good molecular recognition properties. Instead of just using a regular helper strand like this, you can put a self-complementary region to it so you get a bump. Once you have a bump, then you can write. Once you can write, you have a template. It turns out, if you look at photolithography, that is really just writing.

This resolution is about three nanometers. The thing that is really interesting is that it seems like a huge step back, because Eigler at IBM was able to move around individual atoms, and this was back in 1989. The thing is that that was done with a whole team of scientists, took millions of dollars worth of equipment, and took them all day just to do one. Paul Rothemund is one guy in one lab and made billions of them.

The hard part is coming up with the helper strands, which are individually synthesized and you have to know exactly what the sequence is. This software was written in JAVA at General Dynamics and it took less than two weeks. Once someone paves the way, following is easy.

This is nice. But how do you make money to keep your employers happy? What we you really want to do is have one of those perfect nanocubes that I mentioned before and you connect it to the DNA. You can build interconnects, you can build transistors, and actually start building solid state devices by using the nanocubes and the DNA. As someone mentioned earlier today, the electronic properties of DNA are really bad. Silica is not too good either but there are quite a few polymers that are conductive or semi-conductive.

Because we are currently limited to the M13 virus, you are stuck with 7000 base pairs. This means you can only build these things so large. Here we have the NAND gates put together as a series of self-assembling puzzle pieces. Now, the question is, why are they self-assembling? It turns out the way DNA behaves, and Paul Rothemund saw this in his dozens of patterns, when you bend DNA you are exposing the inside amino acids, and they like to line up. You can take advantage of that by actually encouraging this behavior to try to get the DNA to line up underneath it and whatever device you would like to build connect up over it.

Now, there are some questions. Do you keep the DNA or do you throw it away afterwards? It probably depends. DNA is not a very high-temperature material, especially when you consider that the Pentiums today are running almost as hot as a toaster. They are putting out a lot of heat. We may have to find a way after the thing is assembled to hold it down in some other solid state fashion so that we don’t use the DNA. Another way to do hierarchical assembly is to use polyominoes. You have to use the DNA to form these shapes. Right now the technology is such that you can only have 7000 amino acids, but once you have these shapes you can build them either hierarchically like this or directly into whatever shape you would like. Interestingly enough, General Dynamics has been very supportive of what I am trying to do, but with this stuff which is more than five years in the future, they say they aren’t going to bother patenting it, which is why I am able to tell you about it.

Nanorex just came out with a new version of NanoEngineer. This one can handle DNA, which is really cool. For the next version, you will be able to put in structures and get the files for your helper strands. Mark Sims, his plan is to go around the country putting on workshops where you go in, you design your structure, the file gets sent to a DNA ligamer synthesizer which manufactures the DNA right there. You mix it up, heat it to 90 degrees C, let it cool over two hours and then you take a look at it through your atomic force microscope. From an educational and publicity standpoint I think this is going to have a great impact.

In terms of applications, NAND gates can make a supercomputer, and if you can take care of the heating problems, you can have a supercomputer small enough to fit on the end of your finger. If you can just make pores, then you can make water filters. This is one way you can get clean water. Fuel cells are another possibility.

An extreme broadband reconfigurable fragmented aperture phase array is a mouthful. It basically means that you have an antenna, except you have a lot of them, and you can reconfigure them on the fly. Let’s say you could cover your outer skin with this sort of antennae. Your clothes are made out of this. All of a sudden, you have an aperture in the optical range that can see at a hundred miles what you can see at twenty feet. It will increase your visual acuity, for your camera or whatever you have, by three or four magnitudes. This is what the optical antennae can give you. You don’t need to build something that is very heterogeneous. These arrays are going to be fairly easy to build with primitive nanotechnology.

The ultimate goal of putting together nanocubes is to build a desktop nanofactory. One of the things that is kind of interesting though is that people have talked about how grey goo is going to be the danger. Having a nanorobot get loose and start eating everything in sight is about as easy as to build a car that runs wild and starts living on tree sap. Theoretically someday we may be able to build something like that, but it’s very easy to stop. On the other hand, a desktop nanofactory appliance that sits on your desktop sounds pretty harmless. We all have desktop printers, right?

Now, what if you could print not everything, but a copy of itself, computers, and maybe a few other things. As Ralph mentioned, one of the first things we want is diamond. When you have diamond, what happens is you can build skyscrapers a hundred miles up. I’m going to steal Josh‘s idea here. You don’t just want a building that’s a hundred miles up, you want to do something practical with it. What you want to do is build a string of them, you build a long structure that is 300 kilometers long and a hundred miles up. It’s an accelerator, and that is how you get to lower earth orbit. That takes a lot of diamond, and, of course, everyone is going to want one. Human greed is infinite and there is a limited amount of carbon.

Think how successful Google would have been if they charged for every single search. If it had cost a tenth of a penny, no one would have used them. Because it was free, everyone used it. The same thing is going to happen with carbon dioxide in the air as what happened with the Tragedy of the Commons.

Keith Henson realized this fairly early on and said what is going to have to happen is the Sierra Club is going to go dig up Wyoming and burn all the coal in it just to save the rainforest. It’s funny, but I have not found a good, logical reason why it is not going to happen. In order for the government to get involved and enforce a ban on using CO2 for your desktop appliance, they could not go around looking for pollution sources. You would have to look for who is sucking CO2 out of the air. That sort of enforcement requires nanotechnology, which is scary. Another possibility is the air is owned. You may have a right to life, but you don’t have a right to breathe. I told some of these things to the president of the Sierra Club once. He was very open to nanotechnology, and he held his head in his hands and said, “How do you people sleep at night?”

We have all sorts of issues that we are going to have to deal with in terms of simple things like a molecular printer that can pull CO2 out of the air. This can have a huge impact because it can build a copy of itself. If it costs only a dollar for a molecular printer and it can print a copy of itself it could be a pretty scary world. So I think nanofactories are a very important tipping point. There are many approaches to nanofactories. One of them is going to break through first and they will be available in your neighborhood. Thank you very much.

This entry was posted on Thursday, January 24th, 2008 at 1:06 am and is filed under CRN Bio & Nano, Nanotechnology. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.