Haze and sun for mask symposium

12/01/2008

Ever since lithographers began exposing with 193nm DUV light, photomasks have accumulated photochemical haze that eventually congeals into crystalline defects. At this year’s SPIE Symposium on Photomask Technology (October 6-10 in Monterey, CA), maskmakers learned that clearing off the haze uncovered something new: a “sun effect” where images appeared brighter (and clear CDs bigger) in the center of the reticle, with dimmer regions at the edge. Perhaps symbolically, the days in Monterey began with thick haze, progressed to bright sun, and ended in darkness.

Damage to photomasks was always a major topic at the BACUS Photomask Symposium, but it has become more crucial since replacing high-end reticles has become more expensive. The pellicles that protected masks from particles never prevented haze, but now they are being damaged by aggressive 193nm radiation. In fact, according to Gregory Hughes of SEMATECH, who gave the annual Mask Industry Self-Assessment report, exposure-induced pellicle damage (which still darkens images, according to Jin-Ho Ryu of Hynix) was the leading cause of 193nm photomask returns. Pellicle replacement is a routine matter???but in a private interview, Franklin Kalk of Dupont warned that removing the adhesive that holds them on leaves sulfate ions on the surface, which then facilitates haze formation. Jaehyuck Choi of Samsung reported the existence of a mask cleaning procedure that left a mysterious protective layer, which slowed haze formation even when chemical contaminants were present.

Gavin Rider of Microtome Precision reported a damage mechanism in which chromium migrated into clear spaces, bridging gaps, changing CDs, and darkening areas of the image. He attributed the effect to electric-field induced migration of ions at electrostatic fields well below the threshold for conventional electrostatic damage (ESD). That chrome migration might be the origin of the “sun effect,” suggested Steven Labovitz of Pixer Technologies (now a subsidiary of Zeiss) which has developed a tool that measures the early onset of transmission loss due to haze. Cleaning the haze off (which often seems heaviest in the center) sometimes uncovers the “sun effect.” Kalk privately commented that, while electrostatic fields had something to do with chrome footing and CD changes, the amount of 193nm exposure and the atmosphere inside the stepper mattered more. He recommended replacing chrome with the new OMOG
MoSi material for critical applications.

That new material for binary masks, opaque molybdenum on glass (OMOG), developed by Shin Etsu MicroSi, was another featured topic. While very few such masks are being made today, it has major advantages, according to Greg McIntyre of IBM. In particular, the images it projects are just the same as those predicted in the simple thin mask approximation, facilitating modeling. The more complex EMF simulations needed for other materials complicate model-based OPC, too much even for supercomputer vendors. Thomas Faure described the successful process development for 32nm node photomasks with this material.

Sub-65nm geometries constitute a minuscule part of mask industry output, according to the industry self-assessment, even though they consume a disproportionate part of industry attention. In fact, 60% of reticle volume is still for geometries larger than 250nm, according to Hughes, and the increased volume of low-end reticles is one explanation for the decrease in mask industry revenue in the last year, even though volume remained stable at about 700,000 reticles.

Keynote speaker Aart de Geus of Synopsys, in his talk that emphasized the impact of economics on recent the technology direction, also touched on the difficulty of migrating to smaller dimensions. EDA customers hold on at older nodes because of lower cost and predictable schedules. Process development that cost companies $800M at 45nm will consume $1.1B at 32nm, predicted de Geus. He also warned that the variability of nanoscale physics was bubbling back into design, threatening 40 years of progress. He advocated greater discipline, with unified design flows, systematic verification, re-use of IP, and extensive modeling of the coupled design and manufacturing process. However, de Geus noted that computer modeling has failed to be predictive for the modern (but formally chaotic) economic system. He advocated developing even more powerful computers and models for the economists and designers. ???M.D.L.

Check out solid-state.com for Griff Resor’s report on a SPIE/BACUS expert panel that discussed the challenges maskmakers must overcome to reach the 32nm node, and what challenges lie beyond.