Barely two years ago, everyone from computer chip manufacturers and that industry's support associations to stepper/scanner manufacturers thought the global semiconductor market would pass over F2 excimer lasers. They said the 33-nm advantage afforded by F2 laser sources only represents a half-step improvement in critical dimension (CD) capabilities and respective increases in microprocessor clock speeds. Such minor improvements weren't worth the cost of retooling microchip fabs around the world, not to mention the extreme costs and labor associated with beating several technological hurdles.
Since then, however, several industry factors have conspired to elevate F2 lasers to a viable node on the semiconductor roadmap. Next-generation lithography (NGL) techniques such as extreme UV (EUV) and ion- and electron-beam lithography will take longer to bring to production. And while existing technologies such as KrF (248 nm) and ArF (193 nm) are extending their usefulness, they are unlikely to be enough to close the gap between excimer and NGL. These factors, in combination with lessons learned during the development of ArF-based lithography in CaF2 lenses, deep UV (DUV) photoresist, masks, and pellicle design will give F2 a shot at carrying the microchip industry through to the NGL level. Plotting a course
The debate on the future of semiconductor design, testing, and manufacturing has no better forum than that provided by Sematech (Austin, TX), a support association for the semiconductor industry. Since its formation, Sematech has been the guardian of the International Technology Roadmap for Semiconductors, a forward-looking document that encompasses the consensus of semiconductor manufacturers. The roadmap details the anticipated technologies required for subsequent generations of faster microchips.
Figure 1. Pellicles are the protective coverings attached to photomasks, which represent a large portion of the cost of ownership for microlithography operations. Micro Lithography Inc. has introduced several new techniques for DUV lithography, including liquid coating on the sides of the frame and multilayered fluoropolymer antireflection coatings. Some experts question whether progressing farther into the DUV (157 nm) will require new approaches to pellicle design because of the absorption and refraction caused by nonuniformity. (Illustrations courtesy of MLI).
According to Gerhard Gross, director of lithography for Sematech, a common consent began to build behind 157-nm lithography last December when EUV lithography was also named the forerunner for NGL followed by electron projection lithography (EPL). "By the time of our meeting last year in December, some opinions were starting to form," Gross said. "Based on the survey results of that workshop, the International SEMATECH Executive Steering Council decided we would fund NGL projects only for EUV and EPL technologies. There was some support for 157-nm optical lithography, so we began a feasibility projectBy June, the enthusiasm for 157-nm technology had grown and it was the preferred technology at the 70-nm node, followed by EUV and EPL."
The first betas for 193-nm ArF-based lithography machines entered testing last year and the first production machines are beginning to emerge, said Tom Kroehle, director of lithography for Lambda-Physik. Lambda-Physik, Cymer, and Komatsu are the leaders in developing optical lithography light sources.
The systems are similar in operation to photographic printing machines. Laser light is directed through a series of focusing lenses to a mask containing a pattern of the microchip. The pattern on the mask, which is covered by a protective pellicle (mask cover), is then reduced through another series of lenses until a demagnified light pattern falls on the photoresist on the wafer. The chemical composition of the photoresist layer on the wafer changes based on the presence or absence of the patterned light. Then the wafer is etched to remove excess material around the pattern, the pattern is moved to a new location on the wafer (hence the names "stepper" and "scanner"), and the process repeats until a complete microchip is created.
Excimer lasers are considered excellent sources for microlithographic applications because of their short wavelengths and low-coherence light. However, as these laser sources move deeper into the UV, traditional lenses, pellicles, masks, and photoresists become a problem. Many of the materials used to create these devices such as fused silica lenses and polymer-based pellicles absorb DUV light. Other elements of the system, such as pellicles, also require tighter manufacturing controls because small variations in thickness can result in refraction that blurs the light pattern, ruining the microlithographic process. F2 lasers will be ready
Only a few years ago, the F2 laser itself was a part of the technological problem plaguing 157-nm lithography. Very few of the lasers existed, the power outputs were too low, and control mechanisms were not sufficient for the exacting science of semiconductor production. The introduction of solid state excitation mechanisms went a long way toward solving those control problems, while increasing the repetition rates and boosting overall power. Kroehle said F2 lasers will be ready in time for the first beta scanner releases scheduled for 2002-03.
Lambda-Physik already has a standard broadband 600-Hz F2 laser that delivers 10 W. This laser has been tested at the MIT evaluation center, besting the 2.5-billion-pulse marker on the present tube without replacement. Internal optics are still under test, recently achieving 800 million pulses without degradation. In the early part of next year, the company will release a 1-Khz, 10-W F2 laser with line selection with a total bandwidth greater than 2 pm. By the end of third quarter 2000, a 2-KHz, 20-W model is expected, Kroehle said.
Three of the four main stepper/scanner manufacturers, including Nikon, Canon, and SVGL, have ordered F2 lasers for test and research and to determine the crucial cost of ownership. The final player, Ultratech, has ordered an F2 laser from Cymer. Although duty cycles and lifetimes continue to improve, 157-nm lasers may not meet every performance need expressed by the big four. At least one manufacturer has requested an F2 laser with full line narrowing down to 0.25 pm. Lambda-Physik officials say that is theoretically possible, but perhaps not within the time frame required by the semiconductor roadmap or in an acceptable cost-effective manner.
"Most vendors go with catadioptric (a combination of reflective and refractive) lenses, which can use normal bandwidths," Kroehle said. "Those who want standard refractive lenses have requested the extremely narrow line width narrowing, but I don't see that happening."
Currently, ArF lasers (193 nm) with 1-Khz and 2-Khz repetition rates and 10- and 20-W outputs are available for beta microlithographic scanning systems. By the time ArF-based systems hit full production, Kroehle expects performance to be increased to 4 Khz and 40 W. He expects F2 lasers will also reach 4 Khz and 40 W by the time they enter full production-assuming several side issues related to DUV lithography can be resolved. Remaining challenges
During a recent symposium held at Lambda-Physik on the future of 157-nm lithography, industry officials representing all elements of the semiconductor manufacturing process gave presentations about the state of the many elements that will go into DUV microlithography. According to industry experts, several technological hurdles for 157-nm lithography have been identified, including photoresists, pellicles, and optics.
Dave Collier, president of Alpine Research Optics and symposium participant, said, "Based on what I saw, some significant issues remainbut there don't appear to be any show stoppers at this point. The pellicles are the weaker -- if not the weakest -- link (at 193 nm) and the same will apply at 157 nm."
Micro Lithography Inc. (MLI), one of the world's largest suppliers of pellicles for microlithography, has met the 193-nm challenge by using fluoro-polymer anti-reflective coatings for the pellicles and carefully controlling the uniformity (0.1 percent) and film thickness (2.85 µm plus/minus 0.2 µm). Many experts expect pellicle manufacturers will need to find new materials to meet the transmission requirements of an F2 lithographic system. According to C.B. Wang, director of research and development for MLI, the company has tried several new materials. However, he said a clear winner has yet to arise.
Optics may be less of an issue than first considered. Early reports on DUV lithography suggested that the market's ability to manufacture sufficient CaF2 optics was far behind required levels. However, Collier said that longer cycles between nodes and the entrance of Corning and Schott along with market leaders Optovac and others will deliver the large numeric aperture optics on time and with sufficient purity for the application.
SPIE's 25th Annual Symposium on Microlithography will be held 27 February -- 3 March in Santa Clara, CA.
R. Winn Hardin
R. Winn Hardin is a science and technology writer based in Jacksonville, FL.