Through the combined efforts of researchers and manufacturers all over the world, the capabilities of microprocessors continue to grow exponentially as predicted by Gordon Moore in 1965. Progress in numerous fields is on track to continue to uphold Moore's Law. But a new obstacle looms: powering very high-current, very low-voltage processors. The difficulties in delivering this power with sufficient stability and fast enough response time have earned the industry nickname the Power Wall. This project will develop innovations in microfabricated magnetic components that will allow microprocessor power delivery with performance well beyond the capabilities of other techniques now being studied, and will lay the foundations for advanced magnetics in a wide range of applications.
Future microprocessors will typically require supply currents on the order of 100 A. The power requirement is mitigated by scaling supply voltages to lower levels, but the resulting low impedance makes stable power delivery more difficult. A small-size, fast-response power converter will need to be located immediately adjacent to the processor. Portable, battery powered systems impose even more stringent efficiency and size constraints. Improved inductors are critical for meeting these requirements. State-of-the-art inductors remain the largest and most expensive components in power converters, and they limit the efficiency and response time. A new technology-such as microfabrication-must be applied. But existing microfabricated magnetics exhibit poor efficiency, poor power density, or both.
This project will introduce several innovative approaches that will dramatically boost performance. New magnetic materials will achieve low hysteresis and eddy-current losses, yet allow operation with high flux density in the 5-20 MHz range. These properties will be obtained by using composite materials comprising nanoscale particles of magnetic metal in a ceramic matrix, deposited by vacuum evaporation. In these materials, the ceramic will insulate the particles to prevent eddy currents, while the ultrafine particles will reduce coercivity and hysteresis loss. In addition, a newly proposed fabrication process will use anisotropic silicon etching and other microfabrication techniques to form inductors in a configuration optimized for low-impedance high-current applications. The result will be higher power density and efficiency with a streamlined process flow that will simplify magnetic material deposition. Inductors using the new material and process will be fabricated and tested, and applications will be developed in cooperation with industrial partners including Intel Corp. and Volterra, who will collaborate on implementation of circuits and provide financial support.
The new magnetics technology will have a broad range of applications including power conversion in other systems and inductors for RF communications circuits; the new magnetic materials are expected to have good properties up to frequencies on the order of 1 GHz.
The project is designed to capture synergy between research and education. Students including first-year undergraduates, upper-level undergraduates, and Ph.D. candidates will be actively involved in the research program. To meet the significant challenges associated with providing undergraduates with a meaningful research experience, successful, proven programs in undergraduate research will be combined with refinements in student teaming. The research program will be structured to provide opportunities for useful and satisfying work at appropriate skill levels for undergraduates, by taking advantage of the unique opportunities afforded by work in microfabrication.
Outreach and curriculum innovation will help develop the next generation of engineers. Outreach activities involving high-school and junior-high-school students will attract them to engineering; courses that let undergraduates sample the fun of real-world problem solving will inspire them to continue; and a streamlined class in power electronics and electromechanical energy conversion based on modem applications will encourage them to study this critical area. ***
