Q&A with Frances Hellman, Department of Physics & Calit2 Academic Participant
Frances Hellman is a professor of physics at UCSD and an academic participant in Calit². She organized and hosted the institute's first Worshop on Non-Volatile Memory September 21-22 in San Diego.
Q. Why is the workshop focusing on non-volatile memory?
A. The workshop takes a comprehensive view of non-volatile memory, which really means the future of magnetic recording and magnetism generally. UCSD is one of the top centers of magnetism in the world, and the breadth is enormous-from the very theoretical and basic science-driven, to things that are very close to current application. We are especially strong in the materials end of magnetism and how they get used. Calit² takes the far view of the applications end, as far as 20 years out, so the goal of this workshop is to pull people together and develop new collaborations and new directions that are related to Calit².
Q. Which technologies will be the focus of discussion?
A. Some participants will want to discuss extensions of current technology in magnetic recording. Companies are also working on rival technologies to make MRAM-magnetic random access memory-and I would call that a medium-term technology. Those companies include IBM and Motorola, and both will be at the workshop. MRAM is also part of an even longer term vision of the future of magnetic recording, which is spin electronics.
Q. What is spin electronics?
A. Spin electronics is the futuristic view of magnetic recording. If you look at your PC, there's the hard drive, which is entirely based on magnetic materials-iron, cobalt, etc-and the processor, which is entirely silicon-based. They're completely separate technologies, they're manufactured in different ways, and they're physically separate pieces of your computer. If the hard drive crashes, you replace it; if the CPU dies, you replace it. The futuristic idea behind spin electronics is to merge the two functions together, so that memory is not a distinct function from processing. In a sense, flash memory, which is semiconductor-based, is one way to combine the two functions together, but spin electronics would use magnetism for both purposes.
Q. Why is it called "spin" electronics?
A. In a very fundamental sense, memory-the way it is currently implemented-is based on the spin of the electron. Electrons spin up and down, and magnetic materials have two directions of magnetization and that constitutes their ability to store 0s and 1s. That's the spin. They also have a charge. If you take a broad view of what a processor and memory are about, the processor handles only the charge of the electron, and the memory only cares about the spin of the electron. So memory and the CPU are not using electrons in the same way. If you can marry the two, it will be faster, and there may be whole new ways to use them that we cannot even imagine today.
Q. How far are we from a true merger of the two functions?
A. There is some merging already in a certain class of functions of magnetic recording. In particular, the "read" head on most drives today is based on giant magneto resistance (GMR). GMR really does take advantage of both the spin and the charge. It came out about 15 years ago, and now it's built into most computers. That's a very rapid adoption of technology from first concept to widespread commercialization. But the "write" function is still completely separate, and processing has nothing to do with the spin.
Q. What is the ultimate goal of research into spin electronics?
A. The holy grail of spin electronics is a spin transistor which can be switched by a magnetic field rather than an electric field. It's one possible future, but there are a lot of technical and fundamental science problems before we get there.
Q. How does the work on magnetism and non-volatile memory fit into the mission of Calit²?
A. Let's envision a huge array of sensors, like the ones Calit² sees deployed throughout the physical world. If it's a trivial sensor and it's just going to transmit the data, you need wireless technology and some optical fiber to transmit the data on land. But let's assume it's not trivial and you're collecting a lot of sophisticated information locally. What you want is a local processing unit that can take a vast amount of local information, distill it down, and do some local processing before you send it. You want low power consumption, you want non-volatile memory so if the power fails you don't lose the data. You may also want to send something out to that sensor that directs it to do something slightly different-like point in a different direction, or collect a different set of data. The more compact and powerful the local processing units, the better off you are, and we can already foresee the day, six or seven years from now, when magnetic will overtake silicon. That's because magnetic recording density, which is inversely proportional to the size of the bits, is rising faster than silicon device density.
Q. What types of research could benefit from these magnetic-based sensors?
A. One small but intriguing example I read about recently in a magazine is with great white sharks. Little is apparently known about how they travel and where they go. Researchers are already implanting sensors that record how deep they swim, in which direction, how cold the water is, and so on. After six months, the sensor sends a current to a small wire that burns the attachment away, and this sensor floats to the surface and begins transmitting the information. Obviously the amount of information recorded in six months can be enormous. So what you really want is some way of processing that information, distilling it down into something that is meaningful and useful, and then transmitting it. So having the ability to have low power and high storage density has a lot of applications.
Q. What do you hope to achieve at the workshop on non-volatile memory?
A. First of all, if we-the UC San Diego and UC Irvine researchers in Calit²-hear what industry perceives as challenges and problems, we may have ideas for new approaches to solve them. Hopefully, out of this workshop we can develop some productive collaborations with our industrial partners. Also, for students, this is a good opportunity for them to meet some of the researchers in industry and vice versa.
Related Links
Summary of Proceedings
Presentations
Announcement