Original URL: https://www.theregister.co.uk/2009/06/20/euv/
Intel goes to ultraviolet extremes
The EUV prototype that (nearly) blows up
Research@Intel A long-awaited breakthrough in chip-making technology is inching closer towards commercial reality. But progress is slow and remains uncertain.
At Thursday's Research@Intel event in Mountain View, California, The Reg and other press folk sat down with Mike Mayberry, Intel's VP of technology and manufacturing, to discuss the company's future chip-fabrication goals. Among them was extreme ultraviolet lithography (EUV), which will help to cost-effectively shrink chip manufacturing below 32nm.
Intel is currently gearing up for its next-generation 32mn chips, but they will be built using a highly-refined version of tried-and-true immersion lithography technology - not EUV.
Unfortunately, to reach such a tiny process size, immersion lithography must use a complex and costly multiple-mask scheme. With prototype EUV technology, on the other hand, Intel Research has managed to use a more-straightforward masking system to print down to 24nm - but only in the lab.
Mayberry's labs have managed to "play tricks," as he put it, with immersion lithography to print features down to 15nm. "There are alternatives" to EUV, he said, "but they are not necessarily better alternatives than EUV working," adding that "You'd rather have a single-mask EUV approach."
But many EUV refinements remain to be made. According to Mayberry, "We've made considerable progress, but we're not yet to the point where we can safely predict that production will be on this date, on a Thursday, but we're continuing to work in that direction."
When EUV is introduced, Mayberry said, it will first be used in conjunction with immersion lithography. "We typically use a combination of wavelengths for every process," he said, "so we reuse some of the tools from the previous generation - you have to do that for cost-effective manufacturing."
The current energy source for EUV prototypes, said Mayberry, can produce around 10 wafers an hour once the process is finalized. "That's sufficient to do development," he said, "but not sufficient to do production."
That same source has been used in what he referred to as "burst mode" to produce five times that output, but Mayberry cautioned that "Burst mode means that you run it and - quick - you turn it off before it blows up."
Theoretically, he said, the system under development works, but researchers "have not yet shown commercial viability where you can run a tool for, let's say, a year in a row without anything more than regular maintenance."
Development EUV tools - the next step after the current experimental prototypes - should start to become available next year, said Mayberry, but their development cycle will most likely take two to three years. "That would put production in the 2012, 2013 time frame."
Though that date is not certain. As Mayberry put it, "If you went out and asked 20 people, you'd get probably 25 different opinions on when that would be." ®