The Abbe’s method and Hopkin’s method are among the most popular microlithography aerial image simulation methods. In particular, the Hopkin’s method is generally more popular for the high speed aerial image simulation domain used in model based Optical Proximity Correction (OPC). This is due to a general perception that the Hopkin’s method can generate more compact sets of kernels than the Abbe method due to the application of a SVD (Singular Value Decomposition) process to the large Hopkin’s Transmission Cross Coefficient (TCC) matrix. On the other hand, the Abbe method is capable of easily decomposing the source field into independent point sources with a 2D partitions criteria, but this makes the kernels generated from Abbe’s method much larger than the Hopkin’s kernels.

In this paper, we demonstrate that by applying SVD to the original Abbe’s kernels, the essential kernels according to their singular values can be efficiently extracted. The experimental results our algorithm, the Abbe-SVD method, shows over 100X of both runtime and memory saving over traditional Hopkin’s SVD methods for kernel generation.

This methodology is fundamental research into improving ProGen runtime.

For more information on this work, please refer to the work presented at the 2008 International Conference on Electron, Ion, and Photo Beam Technology and Nanofabrication (www.eipbn.org/). The full paper is available at the Journal of Vacuum Science & Technology B website.