Mask Synthesis in the Era of GPU Computing and Deep Learning
TimeThursday, December 9th10:30am - 11:00am PST
Special Session (Research Track)
Hosted in Virtual Platform
DescriptionIn semiconductor high-volume manufacturing (HVM), lithography and etch modeling of the patterning process from photomask to silicon plays a fundamental role. In the past few years, Deep Learning (DL) has been successfully applied to capture the residue error of traditional optical and etch models, with greatly improved results. This talk will highlight recent progress using DL for process modeling. Inverse Lithography Technology (ILT), a scientifically rigorous and satisfying approach, was originally proposed, advocated, and demonstrated in 2003 by Luminescent Technologies, a startup company founded by Dr. Peng. At the time, wide adoption of ILT was hindered by two issues in HVM: long computation times and long mask write times using the existing VSB mask writer. Tools to address both issues rose in the past few years: GPUs and Multiple-Beam mask writers. In this talk, Dr. Peng will share recent work on supporting GPUs in lithography and etch modeling and ILT.