The proceedings from the first joint SPIE Photomask and EUVL conference (2017) are now posted. A few remarks are in order:
The ASML tool progresses, with throughput now around 100 wafers per hour. The resolution, line edge roughness, sensitivity (RLS) trade-off still has not been overcome. Several speakers observed that the EUV scanner is dose limited, and that LER suffers as a result. Additionally, EUV photomask TAT exceeds that for 193i. Samsung reported costs of 8X greater, while the Ebeam Initiative reported EUV mask yields around 64%, compared to 90% or greater for AttPSM (193i). Mask corrections specific for the EUV scanner lead to data volumes 3X greater than for 193i.
Much work is being done to develop more sensitive EUV resists. I think that much of this effort is misplaced; until the LER is greatly improved, printing faster simply produces more failing devices. Far more promising is the fundamental research into resist exposure mechanisms. Noteworthy advances were reported by IMEC and by LBNL.
LBNL reminded us that, even with 100% absorption, 30 mJ/cm2 is about 20 EUV photons per nm2 . Therefore, the expected 1σ noise is around 4.5 photons per nm2 ! This is the root cause of the LER problem, and why the dose must greatly increase.
[For those unfamiliar with the challenges posed by EUVL, a look at Wikipedia will be enlightening.]