At present, common resist models utilized in photolithography to correct for optical proximity effects(OPC) became complex and less physical in order to capture the maximum of optical and chemical effectsin shorter times. The question on the stability of such models, purely empirical, become topical. In thisthesis, we study the stability of existing OPC models by examining the origins of model instability towardsprocess parameters. Thus, we have developed a perturbative method in order to evaluate the stabilitycriterion. However, achieving stable and simple models needs a separation between optical and otherchemical effects. Besides, multiple approximations, widely utilized in the modeling of optical systemsoperating at high numerical aperture, lead to residual errors which can degrade OPC model accuracyand stability. Thus, we were interested to study the limits of validity of the Kirchhoff approximation,a method which, so far, is the most commonly used in mask near-field modeling. Other semi-rigorousmethods for mask topography effect modeling were also evaluated. These approximate methods canimprove the accuracy but degrades the run time. We then suggested different techniques to correct formask topography effects, while keeping the Kirchhoff approximation in the modeling of the optical part.Among them, we showed that errors due to mask topography effects can be partially captured by asecond empirical model. However, in order to ensure a good accuracy, it is necessary to increase themodel complexity by using more additional empirical terms. Finally, in order to achieve a numericalstability of the empirical model, we introduced a new hybrid fast and accurate method, the multi-levelmethod, which allows us to correct for mask topography effects through a multi-level decomposition ofthe thin mask and discussed its advantages and drawbacks.
