HP research results satisfy Williams
03/01/2007
Stan Williams and his colleague Greg Snider at HP Labs in Palo Alto, Calif., have completed research that could lead to making field-programmable gate arrays (FPGAs) up to 8x denser-while using less energy for a given computation-than those currently being produced. The work was featured in the January 24 issue of Nanotechnology, published by the British Institute of Physics. It uses an idea developed by Dmitri Strukov and Konstantin Likharev of Stony Brook University in New York for connecting a nanowire crossbar to CMOS.
Williams, a chemical physicist, and his research group are also working on ways to grow nano-sized switches and wires via chemical reactions and have them assemble themselves into electronic circuits. He discusses the challenges and delights of directing research in this field with Small Times’ Jo McIntyre.
 Stan Williams |
Q: Is this recently announced FPGA breakthrough the most important nano project you are working on now?
In our lab there are at least 30 distinct projects going on. I love all my children. I’m not going to say any one is better than any other. It’s a very broad palette covering mechanics, electronics, photonics, and metamaterials. In a metamaterial, you are creating an ordered structure. With this you can now make new types of materials that have never existed in nature that have very specific technical applications. Our electronics efforts are on wires and switches and manufacturing and memories and logic devices. This FPGA is one example. I am not bored.
Q: When did you get involved in this FPGA research?
This is a long story. It is at least 10 years of evolution. It goes all the way back to work we did trying to understand how to build switches at the molecular scale. In about the mid-1990s we had been looking at means other than transistors to do information management tasks (e.g., to perform logic operations). We started with some fairly interesting ideas as to how we would have something in place that would be needed when Moore’s law reaches the end of the line.
We were inspired, or informed, by work going on at Stony Brook. They have been reading our papers and came up with a clever idea for connecting these nano-wires to underlying CMOS. But because we had already been thinking about this for several years, we immediately recognized the importance. This is the way scientific research is supposed to work.
Q: Did that idea lead to the paper published in Nanotechnology?
Yes, this recent paper that came out is the result of a continuous phase of research over 10 years. The paper is a realistic, near-term road map for introducing a new technique into silicon CMOS that should not be much more difficult than many of the other changes in materials that have been done.
The most important thing we’re learning is how to keep Moore’s law going as long as possible without hiccups or interruptions. We have some new fabrication techniques, so we asked how we could make hybrid circuits, nanowires plus larger transistors, with the idea of extending the technology.
Q: What were your goals for this research?
It looks like a tremendous opportunity to improve one type of circuit FPGA dramatically without having to scale down transistors, but also, with an ulterior motive, to get people to incorporate it into an existing circuit.
Before any industry is going to start using our switches and wires, it has to be in the manufacturing infrastructure. The issue is to broaden the technology portfolio of the entire industry that was making integrated circuits.
We viewed the FPGA as the lowest-hanging fruit. It was the one we felt could be done. The problem four or five years ago was trying to figure out how to put switches into the interconnect of the surface and make it better.
Q: How did you solve that switch interconnect problem?
Through an interesting series of circumstances, we finally figured out how to do it. We had some ideas we had been batting around. The group at Stony Brook had an idea for a crossbar structure, or network of nanowires and switches to CMOS. A light bulb went off for us. Ideas have continued to flow. We’re now getting more ideas for other circuit types.
Q: That problem of having tiny nanowires connect to comparatively large transistors has existed for years. How did you manage to do it?
It depends on the size of circuits. It’s not such a stretch anymore. And we have developed a new lithography process-imprint lithography-and can fabricate chip sizes as small as 15 nanometers that have never been used in a fabrication environment before. What we’ve put together is not so totally crazy. We can put this in the fab process.
Q: Is this work defined as research or development?
My group is defined as research, but we have now spanned the range from more than 10 years out to looking at tomorrow.
The fact that we have been pushing to do this has elevated our standing within the eyes of our management. We could have just sat back and done wonderful research all that time, but we volunteered to step up and help our research find its way into the industry so HP can benefit from it.
Q: What has been the reaction to the announcement of the breakthrough?
Very favorable. I’m hearing from people all over the world about it. It was a demonstration of technical prowess on one hand, showing we can do significant things here at HP, but beyond that, it really is something useful. Reading about this in The Wall Street Journal increased internal enthusiasm for the project.
Q: Is anybody else close to doing this?
This takes time. It’s not a sprint; it’s a marathon. My group has existed for 12 years now. Anyone who is going to get into this is going to have to go on a significant journey.
There are a lot of other somewhat similar types of things going on at IBM and Infineon Technologies, and at several Japanese companies. Lots of people would like to use carbon nanotubes, but they’re not thinking of using active switches.
Q: How realistic is the goal of having a laboratory prototype completed within the year?
I wouldn’t have said it if I didn’t think it would happen. Would I guarantee it will happen? No. There are a lot of things that can go wrong.
The fundamental research is over, and now the hard work begins. Lots of unanticipated things can happen. That’s why we wanted to get things into a fab. If everything works out as we truly hope it will, in a year or so we should have a nice prototype. But it could take much longer.
Q: How involved are you going to be in this as it gets handed off to the fab people in Corvallis, Ore.?
The handoff never works very well. This is a very complex thing we are trying to do. It’s easy for us to make an assumption that something is easily doable, but when it meets the test in the fab environment, there’s a big distance between the lab and the fab. That will be a major accomplishment. If we can do it, then we start to have a platform, a hybrid-type chip. Now you can think of having it go off in a dozen different pathways.
We respect our colleagues in Corvallis. We have a strong collaborative effort now. There are 10 people there who specifically work in areas close to ours. There is a lot of e-mailing, phone calls, and physical visits back and forth, which assist in technology transfer. We will be on-call and willing to help out whenever they need some assistance.
Q: What do you consider the most promising direction nanotechnology research is taking these days-electronics, medicine, optics, cosmetics ...
There is an explosion in all of these areas. It’s becoming real a lot faster than I was anticipating. I was one of the people who was cautioning against over-hyping the area, especially in the materials area. Look at all these incredible composite materials that have been made. There’s a huge, multi-billion dollar market for nanomaterials. That’s taking off like a rocket. More and more people are admitting they are doing it.
Q: Admitting?
There’s the whole issue of dealing with activist-type fallout. I’ve actually given talks at various conferences and had people stand up and start screaming and yelling at me. Fringe groups can be very vocal and you don’t know whether people will get physical. You kind of make a decision about whether to stand and be counted. Early on, HP established a policy that we weren’t going to hide anything.
Q: What accomplishments at HP Labs are you most proud of?
I’m mostly proud of the fact that we have brought together a multi-disciplinary team of really brilliant people to discover and invent. The most difficult part was inventing a new language or dialect so we could all understand each other. People were using the same words for entirely different concepts. Finally, things started to click. Now we have nearly 60 people working together amazingly well-people who [otherwise] would have had no reason to talk to each other.
They had the same boss who kept telling them to “just talk to each other.” It was difficult to get some buy-in to this. It was not preordained that this could work. Finally having all of us being able to work with each other so well was an interesting journey in and of itself.
The Williams File
Stan Williams, 55, joined HP Labs in 1995 as principal lab scientist. He is now an HP senior fellow and director of a group he founded as Quantum Structures Research Initiative, now called Quantum Science Research. The group’s purpose is to explore nanometer-scale electronics.
Williams received the Feynman Prize for Nanotechnology and the Julius Springer Award for Applied Physics. He co-authored and edited Nanotechnology Research Directions, which proposed the National Nanotechnology Initiative that Congress created in 2000 with $485 million in initial funding.