Cornell University researchers have reported significant progress in making a new generation of transistors based on gallium nitride, a material that promises to deliver up to a hundred times as much power at microwave frequencies as the semiconductors now used in cellular telephones, military radar and satellite transmitters.
Lester F. Eastman, the John L. Given Foundation Professor of Engineering, and James R. Shealy, professor of electrical engineering, say they have tested gallium nitride transistors with output power of up to 2.2 watts per millimeter at a frequency of 4 gigahertz (GHz) and expect to see power figures five times higher as soon as improved test equipment is installed.
The researchers reported on their work, largely supported by a grant from the Office of Naval Research, August 5 at the Fifth Wide Bandgap Nitride Semiconductor Workshop in St. Louis, Mo.
"Others have gotten higher power," Eastman said in a recent interview, "but not at as high a frequency."
As demand for space in the electromagnetic spectrum increases, Eastman said, devices that operate at higher frequencies will be needed.
The current generation of chips used in microwave devices use arrays of gallium arsenide (a material Eastman also helped develop). Typically, Eastman said, a single gallium arsenide transistor can handle a power of about 0.7 watts per millimeter, and one can be made on a single chip that will put out about 1 watt to an antenna at 10 GHz.
Among the most important uses for gallium nitride transistors, Eastman said, will be a new generation of communications satellites serving portable telephone users. Current cellular telephones communicate through transmitting towers spotted around a city. New systems in development will use hundreds of low-orbit satellites to allow users to make phone calls anywhere on Earth, not just in places where cellular towers are located. With greater transmitter power, Eastman said, higher-orbit, and thus fewer, satellites will be needed to cover the globe, resulting in great cost savings.
The output of transistors is measured by the amount of power produced per millimeter of length the transistor occupies on a chip. (Actually each transistor is broken up into a number of tiny strips called "fingers.") Typically one transistor on a chip will be about 0.3 microns wide by 250 microns long. Currently, Eastman said, his laboratory is testing chips with arrays of gallium nitride transistors that add up to a total length of about half a millimeter. "We are using much smaller devices at first and checking to make sure the heat can get out," Eastman said.
"We believe we can make transistors with an output power of 12.5 watts per millimeter," he said. "We plan to combine four devices, each 2 millimeters long, on a monolithic integrated circuit to make a chip with an output power of 100 watts at a frequency of 10 GHz."
Current gallium arsenide devices operate at frequencies well above 12 GHz, but there is a trade-off between power output and frequency. As frequency increases, the power output of a given device drops rapidly.
The crystal from which the chips are made is grown on a heat sink made of either silicon carbide or sapphire. Silicon carbide, Eastman said, conducts heat about 10 times as well as sapphire and makes the high power possible, but it is still a very expensive material. Northrup Grumman and Cree Research, two companies also working to develop gallium nitride transistors, supply wafers of silicon carbide to the Cornell researchers as a contribution to science education.
Shealy is responsible for making the crystals, using an innovative technique. Ordinarily transistors are made by "doping" a semiconductor with a few atoms of some other material. The extra atoms break up the pattern of the crystalline material in such a way as to create either free electrons or "holes" where electrons are missing. This makes the semiconductor into a conductor and one that can be switched on and off by the application of a small voltage.
Instead of doping, the Cornell team has chosen to make crystals in which a very thin layer of gallium aluminum nitride is laid on top of a base of gallium nitride. The bond between the two layers places a strain on the upper layer that allows free electrons to flow into the gallium nitride layer, a phenomenon known as a piezoelectric effect.
"With the piezoeletric effect we get enormous charge densities, resulting in a material that has very low resistance," Shealy said, "and then we can also put very high voltages across it. We believe we are presently the only ones in the country pursuing an all-piezoelectric design. This also makes the devices valuable for making power supplies for electronic equipment."
After Shealy and his students grow the crystals, Eastman's team builds transistors and integrated circuits on them, using the equipment of the Cornell Nanofabrication Facility. The crystals still have many defects, Shealy said.
"We've reached the point where we can start to see problems in the devices that relate back to problems in the crystal fabrication. If we close that loop and go around a few times, we ought to be able to perfect the materials and devices," he said.
The Cornell research is principally supported by a three-year, $1 million per year grant from the Multi-Disciplinary University Research Initiatives program of the Office of Naval Research (ONR), with an additional $1 million per year from the Defense Advanced Research Projects Agency and other federal agencies. Dr. John Zolper of ONR monitors the program. Cornell itself funded the construction of a special clean room for crystal growth.
Others at Cornell participating in the research are Arthur Ruoff, the Class of 1912 Professor in Engineering; Joseph Silcox, the D.E. Burr Professor of Engineering; senior research associate W.J. Schaff; and graduate students Joseph Smart, Kenneth Chu, Bruce Green, Eduardo Chumbes, Tyler Eustis, Brian Foutz, Jun Ting Liu, Michael Murphy and Nils Weimann.