Mesoscopic simulation of photoresist processing in optical lithography

The reduction of semiconductor device dimensions necessitates, amongst other things, a reduction in linewidth fluctuation of the individual device components. The achievement of specified tolerances for future technology is an as yet unsolved problem for mass manufacture. The impact of many process and material parameters on resulting linewidth fluctuation is known experimentally, but no models yet exist that allow a suffciently accurate prediction of linewidth fluctuation. In this thesis, new models for the mesoscopic (i.e., discrete and stochastic) simulation of photoresist patterning in optical lithography have been developed and implemented. It has been proven that modeling of the Poisson distributed number of photons (so called "shot noise") is unnecessary. This implies, contrary to common believe in literature, that the average number of photons absorbed during photoresist exposure has no direct impact on linewidth fluctuation. The new photoresist post-exposure bake simulation algorithm reduces the required computing time and memory resources when compared with the standard approach for mesoscopic simulation of reaction and diffusion processes. The new algorithm for the subsequent photoresist development simulation combines an overlap-free description of the photoresist polymers with calibrated development rates. This enables quantitative match of the average profile linewidth predicted by mesoscopic models with established macroscopic models and experimental data for the first time. Profile data obtained from mesoscopic simulations requires additional post-processing in order to carry out an automated evaluation. An efficient algorithm has been developed for the unambiguous determination of the surface position of the developed photoresist. The new models have been used to analyze the impact of process and photoresist material properties on linewidth fluctuation. Comparisons with experimental data from literature show a very good agreement.