Methodology for modeling and simulating line-end shortening effects in deep-UV resist

This paper present a methodology for modeling and simulating line-end shortening (LES) effects in deep-UV photoresist and calibrating the simulation with experimental data. A reaction/diffusion mode is first calibrated using large area dose-to-clear versus bake time data and thickness loss from a top-to-top bake experiment. SPLAT and STORM are linked to simulate the exposure and post exposure bake processes of chemically amplified resist. A threshold mode is then applied to determine the line-end shortening effects. Verification experiments were conducted on two resists, APEX-E and UVIIHS, for several types of geometric features, and the resist model parameters were fine tuned using the Method of Feasible Direction. The measurement of LES agrees quite well when the simulation using an exponential diffusion model of post exposure bake is used. The fine tuning reduced the RMS error to below the noise level in the experimental data and improve the accuracy in predicting LES to 10 percent of feature size.