SPIE Microlithography '99: 157 nm Lithography Faces Challenges
By: Katherine Derbyshire, Semiconductor Online
In the last few months, 157 nm lithography has emerged as the most likely successor to 193 nm. According to William Arnold of ASM Lithography (ASML; Veldhoven, The Netherlands), speaking at last week's SPIE Microlithography meeting (SPIE; Santa Clara, CA), alternatives like extreme ultraviolet (EUV) and projection electron beam (SCALPEL), are unlikely to be ready in time for the 100 nm (0.10 µm) device generation.
Even as the technical obstacles facing post-optical lithography have become clear, 157 nm development has made rapid progress. The most effective argument in favor of 157 nm lithography may be psychological rather than technical, however. Lens systems and photomasks must use different materials at the shorter wavelength, but are otherwise quite similar to their 193 nm counterparts. The industry views 157 nm lithography as a further extension of optical lithography, rather than as a revolutionary step into an unknown future.
Building the infrastructure
Any exposure technology begins with a light source. Even early in the development stage, the likely characteristics of the light source constrain optical designs and set exposure parameters for photoresists. For 157 nm lithography, a fluorine (F2) excimer laser will provide exposure radiation. At SPIE, U. Stamm and coworkers from Lambda Physik (Göttingen, Germany) discussed F2 excimer laser feasibility. The group reported stable operation at a 1 kHz repetition rate. Output power was 20 W at a 500 Hz repetition rate. The laser power and repetition rate limit the amount of light the exposure system can deliver to the photoresist, and therefore the throughput of the exposure system.
A narrow bandwidth laser is also desirable because refractive optics transmit different wavelengths to different focal positions. Since the amount of error increases with the number of refractive elements, catadioptic designs can tolerate broader spectra. With an optical resonator, the Lambda Physik group reduced the laser's bandwidth to below 10 pm. According to Jim McClay, VP of Silicon Valley Group's (SVG; San Jose, CA) 157 nm program, this bandwidth would be tolerable for a catadioptic design, but not for a refractive design.
Completely reflective lens designs do not require line narrowing at all. Unfortunately, Arnold said, these designs probably cannot achieve a high enough numerical aperture (NA) for sub-wavelength 157 nm lithography. Silicon Valley Group already uses catadioptic designs in its 248 nm and 193 nm Micrascan systems, while other vendors rely on refractive designs.
Exposure tools for 193 nm lithography use some calcium fluoride (CaF2) elements to minimize damage due to absorption. Absorption is even more severe at 157 nm, requiring CaF2 elements throughout the lens. Existing applications for CaF2 lenses do not require the large size and high precision needed for semiconductor equipment. Lens blanks and polishing expertise are likely to be in short supply until materials and lens suppliers catch up with the new demand. Both SVG and ASML are establishing alliances with vendors.
According to T. M. Bloomstein and coworkers at the Massaschusetts Instititute of Technology's Lincoln Laboratory (Lexington, MA), laser damage due to color center formation is not likely to be a serious problem. Surface contamination of the lens can significantly affect transmittance, however. In situ lens cleaning, required due to photoresist outgassing, may affect optical performance.
Photomasks are the next important element of the lithography infrastructure. Since the 100 nm device generation is the projected entry point for 157 nm lithography, optical proximity correction and phase shift masks will be needed from the beginning. Silica glass, used for conventional masks, absorbs strongly at 157 nm. Promising results have been achieved with fluorinated SiO2, McClay told Semiconductor Online. Corning Inc. (Corning, NY) researchers Charlene Smith and Lisa Moore reported nearly 80% transmission in low-OH-content fused silica at 160 nm, compared to less than 15% transmission in conventional high-OH content material. Low-OH fused silica would not require significant changes to mask processing or substrate polishing. Transmission drops sharply below 156 nm, however.
Missing pieces
Photoresists, the final element of the lithography process, are nearly impossible to develop without exposure tools. SEMATECH has solicited proposals for a research-grade mini-scanner for photoresist development.
In summary, 157 nm development is still in its infancy. Researchers have identified critical questions, and only begun to suggest plausible answers. Years of development remain. Nonetheless, these preliminary results justify the technology's position as the post-193 nm front-runner.