Analysis of Fitness Functions for Electron-Beam Lithography Simulation and Evolutionary Optimization

Abstract

Abstract-In a previous paper, an electron-beam lithography simulation and optimization tool based on a genetic algorithm was presented. The fitness function was defined as the inverse value of the Euclidian distance between a resampled computed resist profile and its targeted counterpart. In this letter, this previously proposed fitness function is analyzed and its limitations are presented. It is shown that due to out-of-sync effects, the fitness function is intrinsically flawed and is not well-defined for small indentations in the resist profile and/or a large number of nodes along the resampled chain. An alternative fitness function based on the computation of the enclosed area between the calculated resist profile and the targeted one is proposed. It is shown-both theoretically and experimentally-that this function is able to remove the small indentations that previously appeared in the optimized resist profile. This improvement has a direct impact on the quality of the fabricated structures.