29 July 2009—The light sources powering the next-generation chipmaking technique, extreme ultraviolet lithography (EUVL), are too dim, according to industry experts. And that’s one of the main reasons why semiconductor manufacturers worry that the technology will not be ready to produce advanced chips a few years from now. But Almantas Galvanauskas, a professor of electrical engineering at the University of Michigan at Ann Arbor, believes he has a way to boost their power by using a type of industrial laser called a fiber laser. He presented his research two weeks ago at the International Workshop on Extreme Ultraviolet Lithography, in Honolulu.
Galvanauskas’s innovation is to replace the bulky, power-hungry chemical lasers used today with more efficient lasers based on special optical fibers. If Galvanauskas can make his method work—and it looks promising, according to Vivek Bakshi, formerly with semiconductor manufacturing research consortium Sematech, and Sam Sivakumar, a fellow at Intel—one of the three remaining problems preventing EUVL’s move into manufacturing will have been solved.
Galvanauskas’s fiber laser would replace carbon dioxide lasers in EUVL systems. To understand why using a solid-state fiber laser is attractive, you have to understand how 13.5-nanometer photons are generated. Inside an enormously complex machine that looks like a Jules Verne–style diving contraption, a laser beam zaps 20- to 40-micrometer falling droplets of tin, heating them at 1013 kelvins per second. When the zapped droplets reach 700 K, the superheated tin pulverizes into a plasma, which emits EUV-wavelength photons.
These EUV photons careen toward a light collector, which gathers the light and sends it bouncing off eight mirrors that variously focus, narrow, and steer the beam until the final manicured stream of EUV photons bounces off the lines and spaces of the photomask, thereby projecting a pattern onto a chemical-coated silicon wafer.
However, EUV photons are absorbed by everything, including air and optics, so a great many of them never make it to the wafer. After the absorption is accounted for, only 2 to 3 percent of laser photons are “converted” into EUV photons. Additional losses due to limited collection efficiency and reflection losses from each bounce from optics result in further reduction in EUV power by about factor of two. To produce sufficient EUV photons to create a pattern, researchers agree that the power of the light in the 13.5-nm band (the so-called in-band power) must be 115 to 180 watts at the intermediate focus. To attain 115 to 180 W, you need to start with 20 to 25 kilowatts of laser power.
And that’s where commercial EUVL is stalling. Existing EUV sources include synchrotron radiation, laser-produced plasmas, and so-called discharge-produced plasmas, where EUV-emitting plasma is produced without lasers. Carbon dioxide chemical lasers, the de facto industry standard, have been unable to produce that amount of source power. In use since the 1960s, they produce light via an electric discharge in a mixture of nitrogen, carbon dioxide, and other elements. But carbon dioxide lasers haven’t come close to getting an EUVL system to produce 180 W. Cymer, a manufacturer of industrial chipmaking light sources, announced last week that it had shipped a light source capable of 75 W. However, there is no independent confirmation yet for that number.
Galvanauskas thinks that industrial-scale fiber lasers are a better solution. Widely used for welding and other industrial purposes because they are cheap, long-lived, robust, and have no moving parts, fiber lasers are also smaller than flowing-gas lasers. Most important, fiber laser electrical-to-optical efficiency is approximately 4 to 5 times better than that of flowing-gas carbon dioxide lasers. Most industrial fiber lasers take light from a semiconductor laser and amplify it into a laser inside a special kind of optical fiber. The fiber’s core contains ytterbium-doped light-emitting atoms and is surrounded by another layer, which guides the light along the fiber. Unlike the paltry 8 percent efficiency of the carbon dioxide lasers, these can convert up to 30 percent of the input power to output. Their higher efficiency also means it’s easier to keep them cool.
But fiber lasers aren’t perfect. There are 50-kW industrial fiber lasers available as commercial off-the-shelf components (IPG Photonics makes one for welding); however, says lithography guru Chris Mack, they’re not appropriate for EUVL because they are continuous-wave lasers, not pulsed. "If you used continuous-wave lasers,"says Mack, "you’d be wasting an enormous amount of energy." And a single-fiber laser isn’t capable of reaching 25 kW, because pulses of high energy damage the fiber.
So instead of trying to build a single laser that would be capable of 25 kW, Galvanauskas ganged together modules of lower-power fibers. This pulsed system is seeded with a 1064-nm diode laser that is driven by electrical pulses to synchronize it with the falling tin drops. In 2005, his team’s laser demonstrated multi-kilowatt average power operation (more than 1.5 kW for a single-fiber laser and more than 10 kW when a few were stacked together in so-called multimode output beams).
Recently, Galvanauskas’s group at Michigan demonstrated a new technique of combining multiple pulsed laser beams into one, which opens a practical way of reaching very high powers. The goal is to reach 25 kW. The engineering solution is to create smaller modules (lower power, 200W to 400W apiece) to string together that will combine to form one 25-kW beam. "We are developing a spectrally combined system," says Galvanauskas. Together, the beams of the modules will combine to achieve the necessary power. But these modules can also be used separately for other functions, including fixing defects on blank wafers.
At the EUV Litho workshop, Galvanauskas reported that his source is ready to go from benchtop to commercial and that he has procured funding, although he wasn’t comfortable talking about the investor yet. "A few years ago, people said there was no way we could produce EUV in a fiber laser,"he says. Six months later, he had completed a benchtop demonstration, and the team is now working on ways to mass-produce them.