 | |
| Vol. 18, No. 6 | Oct. 8, 1998 |
| |
| Vol. 18, No. 6 | Oct. 8, 1998 |
When Nobel Prize winner Leo Esaki discovered the tunnel diode in 1957, the super-fast, current-switching device was touted as a kind of Holy Grail for computer chip makers, but technological obstacles have so far hindered its widespread use in conventional, silicon-based circuits.
In the Oct. 12, 1998, issue of Applied Physics Letters, currently online, University of Delaware researchers-with scientists at the Naval Research Laboratory and Raytheon Systems Co.-will describe promising, new tunnel diodes that may help chip makers boost silicon's speed while further shrinking chips.
This UD research also was featured on the front page of the Oct. 1 Wall Street Journal.
The technology, described in a patent disclosure, should someday benefit soldiers in the battlefield-or business people on the go, said lead researcher Paul R. Berger, electrical and computer engineering.
If UD's silicon-based, tunnel diodes can be combined with semiconductor circuits to boost transistor switching speeds, the breakthrough could enhance silicon chips, which aren't expected to get any faster or smaller after about the year 2007, said Berger, who received a 1996 National Science Foundation CAREER award for promising, young investigators.
Alan C. Seabaugh of Raytheon, agreed. "This is the first tunnel diode that is compatible with a silicon integrated circuit process," he said. "Adding the diode to silicon circuits will increase circuit performance and could extend the life of existing production lines-a considerable savings, as a new fabrication facility can cost over $1 billion."
The UD tunnel diodes "could provide the same functions with half the components, requiring fewer interconnections, so that circuit speed increases and power losses are reduced," Seabaugh said. "Fewer components also means more chips could be made on a single, silicon wafer, lowering costs."
Nick Holonyak Jr., coauthor of a landmark 1959 article on tunneling events in silicon and germanium materials and maker of the first practical, light-emitting diode, said the UD research is "an exciting revival" of some of his work.
"I'm happy to see this work," the professor of the University of Illinois at Urbana-Champaign added. "It's very interesting to me because it all started with silicon and germanium!"
Newer fabrication techniques- specifically, low-temperature molecular beam epitaxy (MBE) processes pioneered by Phillip E. Thompson and Karl D. Hobart of the Naval Research Laboratory-allowed Berger's team to investigate questions first raised by Holonyak.
"MBE wasn't around when Holonyak was conducting those studies," said Thompson. "It took us a while to make it work at lower temperature, which freezes atoms in place as you add them to the surface of materials."
With Berger and Seabaugh, authors of the Applied Physics Letters article were Thompson and Hobart (both UD graduates); Roger Lake, Gerhard Klimeck and Daniel K. Blanks, Raytheon (Dallas); James Kolodzey, a UD professor whose work set the stage for Berger's current study; UD graduate students Sean L. Rommel, Mike W. Dashiell and Hao Feng; and UD undergraduate Thomas E. Dillon. Authors of the patent disclosure are Berger, Rommel, Thompson, Hobart and Lake.
Faster circuits using less power could revolutionize many military technologies, especially devices for converting analog data into digital signals, said Lt. Col. Gernot Pomrenke, USAF, manager of the Ultra-Electronics Program at DARPA, sponsor of the UD research with the National Science Foundation.
"This technology is absolutely promising," Pomrenke said. "It could increase the functionality of high-speed circuits for radar receivers, missile seekers, satellite communication systems and advanced communications networks capable of handling extremely high data-transfer rates." High-speed, logic circuits should support big advances in image processing and pattern recognition, too, he added.
In the 21st century, soldiers in the field and business people traveling with portable computers might use computer chips equipped with the new, tunnel diodes. Super-fast, highly efficient memory devices would require less power and fewer stops to recharge batteries.
A tunnel diode is a "versatile, high-speed, semiconductor switch," Seabaugh explained. "You can switch the electrical current back and forth, or you can store information with it." That's because electrons zipping through a tunnel diode don't follow the rules of classical physics and remain within designated pathways. Instead, they tunnel through barrier regions. As more voltage is applied to the diode, he said, more electrons flow, then less, then more again at a critical voltage level, a consequence of the quirky, quantum mechanical behavior of tunneling electrons known as "negative differential resistance"-an up-and-down pattern of current flow.
Existing tunnel diodes, fabricated individually from metal alloys, are limited to "niche applications," and U.S. companies sell only about 10,000 of the devices each year, Berger said. So far, he added, researchers haven't been able to mass produce silicon-based, tunnel diodes offering high "peak-to-valley" ratios-a measure of tunneling efficiency.
A key problem has been the amount of "dopant" substances or impurities that must be packed into tunnel-diode materials to closely confine electrons, so that they demonstrate quantum operational speeds. "Creating tunnel diodes using this traditional, alloy process is a black art," according to Berger. "You put a blob of something on a substrate and heat it up and it melts and diffuses just like grilled cheese. You can't make millions of these to go with your Pentium chips!"
Instead, Berger and his colleagues "grew" highly doped (delta-doped) silicon monolayers at the Naval Research Laboratory, using a technique known as molecular beam epitaxy (MBE), at a relatively low temperature-370 degrees Celsius (698 degrees Fahrenheit). The materials were then cured or annealed at 700 to 800 degrees C. The resulting diode resembles a sandwich, with a 4-nanometer layer of pure silicon-germanium at the center, encased by delta-doped slices of boron and antimony on either side. The boron and antimony layers are then encased by silicon with opposing electrical charges, all of which is heaped atop a "plate"-a silicon substrate with positive charge carriers.
"We've been able to mass-produce many of these tunnel diodes across a wafer," Berger said. "We believe this is the first viable technology for integrating silicon-based, tunnel diodes with conventional semiconductors."
To be useful in circuits, this type of tunnel diode-known as a resonant interband tunneling diode (RITD)- should offer a peak-to-valley ratio (PVR) of about 4 at room temperature. "We're more than halfway there," Seabaugh said.
Paul van der Wagt, a designer of tunnel diode-based circuits, said Berger's research "is very promising," perhaps also for low-power, memory applications if currents can be scaled down. "They achieved a negative differential resistance in a silicon system, which has rarely been obtained before," said van der Wagt, of the Rockwell Science Center in Thousand Oaks, Calif. "They also have obtained quite a reasonable peak current density level. It's significant work."
-Ginger Pinholster
