Quantitative lithographic performance of proximity correction for electron beam lithography

Abstract

We have developed a method for measuring the improvement in lithography produced using proximity correction. We have used this method to characterize the performance of the Cambridge Instruments proximity correction software. The goal of this work is to explore the parameter space of the correction. We have used our measurement technique to show quantitatively the effect on lithographic performance of: scattering fit model (double Gaussian, triple Gaussian, and double Gaussian plus exponential), variation in scattering coefficients, and variation of computation parameters (integration grid size). For our study, we have used a common e-beam process: 0.5 μm thick PMMA resist on silicon, with a 20 KeV, 75 nm Gaussian beam. We have studied several geometries with dimensions as small as 0.2 μm. We have found that feature sizes of the proximity corrected exposures are quite insensitive to large variations in input parameters. Variation of the double gaussian parameters alpha and eta of ±30% from their best-fit value produced <1% change in feature size, while feature size changed by ≊2% for each 10% change in the beta parameter. Over the range of feature sizes measured (0.2–3.0 μm) there was less than 10 nm difference between patterns corrected using the double Gaussian, triple Gaussian, and the double Gaussian plus exponential models.