19 July 2011—Extreme ultraviolet (EUV) light, the sort that’s needed to create the transistors of tomorrow, isn’t so easy to make. For years, the companies building EUV sources have struggled to deliver enough photons to match the output of a few incandescent lightbulbs.
But the days of paltry power could be coming to a close. Suppliers of EUV light say they will soon deliver a big jump in brightness, putting EUV lithography on track to produce ever-smaller transistors and pick up Moore’s Law where existing lithography machines leave off.
The revamped light sources are meant to supply 13.5-nanometer light to new lithography tools built by the Dutch company ASML Holding. These EUV machines have been designed for "preproduction," as proof-of-principle demonstrations that future machines will be able to expose wafers effectively and at a high enough rate to be economical.
By the end of 2011, six of the ASML machines, dubbed NXE:3100s, will have shipped to research facilities at six different companies, including Samsung and TSMC. Some of the machines, including one installed in February at an Imec research facility in Leuven, Belgium, are already in use, printing patterns onto chip wafers.
While the machines are working as designed, their EUV light sources are dimmer than what’s needed, so each wafer needs a longer exposure time. "All the parts move at the rate we need them to move, but we don’t have enough light yet," ASML’s chief scientist, Bill Arnold, told IEEE Spectrum in April.
At the moment, the ASML machine at Imec is producing wafers at a rate that’s roughly "an order of magnitude" lower than the target of 60 wafers per hour, says Imec CEO Luc Van den hove. According to Arnold, that’s typical for the EUV preproduction tools.
But the machines could be in for a speed boost. At Semicon West, an industry meeting held last week in San Francisco, representatives from ASML and the two suppliers of light sources for the NXE:3100s all said that they have worked out a road map that should allow them to deliver nearly 100 watts of power into the machine’s optics by roughly the end of the year—enough to process 60 wafers an hour. If they succeed, it will bring EUV a big step closer to wafer production.
San Diego–based Cymer’s light sources, which will be used in five of ASML’s six machines, are currently being upgraded to produce 11 W. But Cymer’s David Brandt says the firm expects to be able to boost the power to 80 W by the end of the year, with additional adjustments pushing it even further.
Cymer’s approach, dubbed "laser-produced plasma" (LPP), uses a powerful carbon dioxide laser to vaporize speeding droplets of molten tin. As the resulting plasma cools, excited tin electrons relax back to lower-energy orbits, emitting EUV light that is steered through the machine and eventually used to cast patterns on a wafer.
Brandt says Cymer can boost the power by shooting another laser at the tin droplet before the main laser hits it. This "prepulse" will heat the tin, causing it to expand and allowing the second laser pulse to penetrate further into the droplet, creating more plasma.
ASML’s other light source provider, Xtreme Technologies GmbH, of Aachen, Germany, is also aiming for big gains in power. Xtreme, now owned by the Tokyo-based lighting firm Ushio, uses laser light to vaporize tin atoms off the surface of two wheels that rotate at high speed through baths of molten tin. The resulting cloud of tin atoms hangs between these wheels and is zapped with a large current, creating an EUV-producing plasma. The technique, which was previously called "discharge-produced plasma" and has now been renamed "laser-assisted discharge plasma" (LDP) to distinguish it from an older technique, has long been the underdog to LPP. But the approach has been gaining attention. Xtreme provided the light source for Imec’s NXE:3100 machine. Although it is not yet bright enough, "we have a road map to have 100 W by the end of the year," says Xtreme’s president, Marc Corthout.
Corthout says the LDP source will get two upgrades in the coming months that will help boost the brightness of the machine. One upgrade aims to make the machine more transparent to EUV light. A second set of changes are designed to boost the power of the source itself by increasing the rate that tin clouds are created and by running a higher current through each one. The team expects to triple the plasma cloud production rate from 6 000 to about 18 000 clouds per second by increasing the spin rate of the wheels.
If all goes well, the improvements should allow the NXE:3100s to reach 60 wafers per hour. But estimates vary regarding how productive an EUV machine would have to be to prove economical. ASML aims to begin shipping its next EUV tool, which is intended for production, in 2012. It is designed to produce 125 wafers an hour using 250 W of EUV light.
However, predictions of progress in EUV lithography have hardly been reliable in the past. EUV was once expected to enter production as early as 2005, notes analyst G. Dan Hutcheson of VLSI Research. So far, he says, the light suppliers "haven’t met any of their schedules." But even if ASML and the light-source makers don’t reach the latest target on time, Hutcheson says he’s optimistic EUV will eventually make its way to chip fabs: "I have faith they can do it, but I don’t have proof they can."
Rachel Courtland, an unabashed astronomy aficionado, is a former senior associate editor at Spectrum. She now works in the editorial department at Nature. At Spectrum, she wrote about a variety of engineering efforts, including the quest for energy-producing fusion at the National Ignition Facility and the hunt for dark matter using an ultraquiet radio receiver. In 2014, she received a Neal Award for her feature on shrinking transistors and how the semiconductor industry talks about the challenge.