15 July 2009—Researchers at this year’s International Workshop on EUV Lithography in Honolulu, are discussing new approaches to many of the intransigent problems that plague this long-anticipated yet still-not-ready chip-printing method. These include new ways of generating the extreme ultraviolet (EUV) light, figuring out a better way to inspect nanometer-scale parts of the system, mitigating contamination generated by EUV light sources, and producing sharper nanometer-scale patterns on chips.
The EUV lithography workshop is organized by Vivek Bakshi, a researcher formerly with global chipmaking consortium Sematech, who went on to found EUV Litho, based in Austin, Texas. He says his workshop is more academic and focused on R&D and less commercial than Semicon West, a much larger chip-manufacturing gathering under way this week in San Francisco. Bakshi solicited solutions to problems that have kept EUV lithography from being adopted as the industry standard, even as its only alternative, double-patterning lithography, is strained further to its limits.
EUV lithography has a history of delays and the potential for further setbacks if its remaining issues are not dealt with quickly. Chipmakers boost processor performance by cramming ever more transistors onto the chips without making them bigger or raising their energy budgets. The only way to do that is by making those transistors smaller; today’s chips contain transistors whose features are as small as 32 nanometers. To print these features, chipmakers use massive multimillion-dollar tools that project light patterns through masks.
Today’s tools project the chip patterns with 193 nm radiation, deep ultraviolet photons with wavelengths that are smaller than those of visible light but still much larger than the patterns themselves. Toolmakers use myriad tricks to coax the light onto the chip in those dimensions, such as printing in water and using exotic lenses. However, even these tricks are running out of steam as features on microchips shrink further. ”The tricks are running out,” says Bakshi, ”and they’re getting prohibitively expensive.”
EUV lithography was supposed to step in and save the day. EUV radiation (which is, in reality, on the edge of the X-ray band of the electromagnetic spectrum) consists of photons of 13.5-nm wavelengths, a massive step down from 193 nm. ”There is no limit to it,” says Bakshi. ”It can take you all the way to the end of Moore’s Law.” Veldhoven, Netherlands–based toolmaker ASML Holding, one of three major tool competitors, along with Nikon and Canon, has installed two EUV lithography development machines, one at IMEC in Belgium and another at the College of Nanoscale Science and Engineering laboratories at the University at Albany, in New York. But there have been many problems preventing EUV’s adoption by the industry.
For one thing, the EUV photons are extraordinarily difficult to produce; researchers have had to dream up exotic contrivances, like blasting falling tin drops with a carbon dioxide laser, to coax these short-wavelength photons into existence. Then there is the problem of absorption. Because their wavelengths are so short, EUV photons are absorbed by absolutely everything, including air and glass. That precludes sending them through lenses to focus them. Thus EUV lithography printing systems operate in a vacuum and use mirrors instead of lenses.
Bakshi says scientists will discuss solutions to three fundamental challenges this week. ”The biggest one is high-power light sources,” he says. ”You need 180 watts of power as your baseline.” But a leading San Diego–based laser firm, Cymer, just shipped a source that puts out only 75 W, and even that, according to experts who were not comfortable speaking on the record, is probably an exaggeration.
One of the proposed alternatives to the messy and profoundly energy-inefficient CO2 EUV photon generator is being floated by Almantas Galvanauskas, a professor of engineering at the University of Michigan. He will discuss the possibility of replacing the CO2 laser with a fiber laser. Fiber lasers use an optical fiber doped with rare-earth elements, which can be kilometers long, as the active medium. They are currently popular for high-power needs such as industrial welding and directed energy weapons. Galvanauskas says that a fiber laser is more energy efficient by orders of magnitude than the chemical flowing-gas laser that currently produces the beam. It also produces less chemical waste and reduces the machine’s footprint. ”The overall system should be much smaller, efficient, and power-scalable than the CO2 lasers,” he says. The team has demonstrated power generation of half a kilowatt per fiber module. Galvanauskas is in talks with one of the major industry light-source manufacturers, which he would not name, but the fiber-laser idea has yet to be tested outside the lab.
In addition to energy efficiency, another major problem that needs to be overcome is contamination. The exploding tin droplets create a great deal of debris in the machine that is not easily cleaned out. Researchers from University College Dublin, University of Illinois at Urbana-Champaign, Tokyo Institute of Technology, and ETH Zurich are all working on the contamination issue and will report their results this week.
The masks, the mirrors on which the pattern of the chip to be produced is written, are also a difficulty. Creating defect-free masks won’t be possible without proper measurement tools, but as yet, no existing commercial tool is available to reliably examine defects on the mask at the needed scale of a few nanometers. Researchers from the University of Hawaii and Ritsumeikan University, in Japan, among others, are working on a high-brightness EUV light source for such metrology tools.
So-called line-edge roughness is another problem that will be addressed by researchers from Lawrence Livermore National Laboratory, in California, and Hanyang University, in South Korea. ”As features have shrunk, the need to control the line-edge roughness has increased,” says noted lithography expert Chris A. Mack, who will also present research at the conference. ”Line-edge roughness in a transistor means leakage current.” And more leakage means the chip will use more power.
The fundamental cause of line-edge roughness is an interaction between the optics and the photoresist, the material coating the silicon wafer that hardens on exposure to EUV light, helping to transfer the mask’s pattern to the silicon. In a lithography system, there is a trade-off among resolution, the brightness of the light source, and the sensitivity of the photoresist to EUV light. ”I can get any two of those, but I can’t get all three,” says Mack. Smaller features require higher resolution, and higher resolution requires a better light source. But a stronger light source means greater sensitivity, which means more errors.
”The bottom line is, there is no commercial EUV lithography today,” says Mack. He says that scientists at the gathering in Hawaii will be able to prioritize the list of EUV lithography’s many technical problems and provide goals for things like light source power that suppliers can work toward.