When Ashraf Alam was a young student in Bangladesh following that country's 1971 war of independence from Pakistan, a class of 30 pupils would share perhaps five textbooks. Maybe that history helps explain why Alam, a Bell Labs researcher whose theories about breakdowns in ultrathin gate oxides are upsetting conventional wisdom, is so willing to share his ideas with others.
Alam, 38, heads the gate oxide research group at Agere Systems' Bell Laboratories in Allentown, Pa. His team, including Bonnie Weir and Jeff Bude, upended accepted theory five years ago when they correctly predicted that gate insulators made from tried-and-true silicon dioxide could be thinned to less than 15 angstroms.
That work was countered by a less-optimistic view from IBM Corp.'s gate oxide research team, and a spirited debate ensued. The models from Bell Labs and from IBM's Thomas J. Watson Research Center at Yorktown Heights, N.Y., took different views of a seemingly arcane issue: damage at the anode from atomic-level holes.
The gate insulator is key to performance scaling, and the more aggressive stance by the Bell Labs team emboldened companies, including reliability-conscious IBM, to thin gate oxides to today's 11 to 15 angstroms for the highest-performance devices.
Now, Alam and his colleagues are making another bold assertion: that soft breakdowns in the oxide can essentially be ignored. Soft breakdowns are slight dislocations in the oxide that may result in lower performance or current leakage. But the chip keeps operating. In hard breakdowns, by contrast, the oxide ruptures and the device shorts out.
Bell Labs has concluded that soft errors diminish in importance as voltages and device dimensions scale, so much so that "oxide reliability may no longer be a fundamental roadblock to scaling," Alam said in a presentation at the International Reliability Physics Symposium in Austin, Texas, last month. That assertion again is being challenged from Yorktown Heights.
IBM oxide researcher Jim Stathis argues that "the Bell Labs theory is elegant, but not the only way to interpret the data-and there are careful experiments [at IBM] which indicate that the theory is wrong."
At technical conferences, Alam often is seen sketching out the mathematical expressions behind his soft-breakdown theory, explaining his thinking to small groups of researchers from competing research labs. It is the kind of intellectual give-and-take-a modern-day Academy-that Alam excels at, and one that brings to mind a youngster sharing textbooks with a half-dozen other students.
While many engineers suffer from speaking styles that race and stumble, Alam's slightly accented English, deep voice and knowledge of his subject combine to make him a master communicator.
Steve Hillenius, director of device technology at Agere, said Alam has "one of those unique combinations of very deep knowledge of physics, with an acute ability to condense very complex issues. He has an incredible ability to communicate the essence of a very difficult problem."
Any theorist looks at experimental data and tries to make sense of it all. Alam said the educational environment in Bangladesh-where engineering students had relatively limited access to computers but excellent mathematics teachers-contributed to his theoretical powers. "In my country, we had teachers who had no money, very few computers, but who would spent a great deal of time with their students. They were very dedicated."
Alam came to the United States in the mid-1980s to get his master's degree. At graduate programs at Clarkson University in upstate New York and at Purdue University in Indiana, Alam found mentors in physics and mathematical theory who supplemented his electronicengineering studies. A widely celebrated engineering professor at Purdue, Mark Lundstrom, had the ability to understand "what is important in a problem, how to distill the problem to its essence," Alam recalled.Coming from such a poor country, Alam said he feels "extremely fortunate"-a phrase he says with such intensity that it is clear the words come from the deepest part of his soul.
He arrived at the gate oxide problem in a circuitous fashion. At Purdue and later at Bell Labs, Alam developed a simulator, one that is still in use today, for the design of optoelectronics components. Designers had used a trial-and-error method to combine electronic and optical functions, the result being lengthy design cycles and high costs. Alam's simulator, based on a theory of how the two domains interact, brought an intellectual order to the optoelectronic-design challenge.
Then, one day, Weir, a Bell Labs scientist, stopped by Alam's office with a stack of computer printouts. She asked him to work on developing a theoretical foundation for the experimental data, a request that marked the beginning of the 1990s scaling debate. Alam now heads the labs' 15-member gate oxide team.
A reporter gingerly approaches the subject of how Alam, as a Muslim living in the United States, feels about the war in Iraq. The subject proves fertile ground.
After Alam earned his doctorate in 1991, his father-an engineer at Bangladesh's state-run telecom provider-arranged for him to marry an 18-year-old woman, Salmina, now a degreed engineer and programmer with a career of her own. The couple have a young daughter.
'Shades of gray'
Though her family had its roots in Bangladesh, Salmina spent much of her youth in Baghdad and then in Kuwait, where her father taught civil engineering. "It is a very complicated part of the world. Iraq was not a country until 1918, and there is a very real possibility it will break up again," Alam said. "There are a lot of shades of gray to this conflict, and a nagging concern I have is whether the United States really knows what it takes to establish democracy there."