Impact of parallelism on data volumes for a multibeam mask writer

Abstract

Single beam mask writer architectures have satisfied mask patterning requirements for decades, but there is now great interest in multibeam mask writers to handle the throughput and resolution demands arising from the needs of sub-10 nm technology nodes. Future mask writers must transmit terabits of information per second and handle petabytes of data. For electron-beam direct write (EBDW) lithography systems parallelism and lossless layout image compression are techniques which have been considered together to approach the data transfer problem. For the multibeam mask architectures proposed by IMS Nanofabrication and NuFlare, the data being transmitted are significantly processed and modified from the initial proximity-corrected layout images, so the data compression problem the authors consider for this application appears to be for a new type of data. Just as the throughput requirements for EBDW lithography systems necessitate the lossless data compression decoders to swiftly reproduce the layout images from their encodings it is likewise interesting to study simple and effective lossless data compression algorithms for multibeam mask writers. The authors will examine how parallelism affects the total beam compressed data for a family of idealized multibeam architectures inspired by the IMS Nanofabrication eMET series.