Calibration of resist model parameters becomes more and more important in lithography simulation. The general goal of such a calibration procedure is to find parameters and model options which minimize the difference between experimentally measured and simulated data. In this paper a multidimensional downhill simplex method and a genetic algorithm are introduced. We investigate the performance of different modeling options such as the diffusivity of the photogenerated acid and of the quencher base, and different development models. Furthermore, new objective functions are proposed and evaluated: The overlap of process windows between simulated and experimental data and the comparison of linearity curves. The calibration procedures are performed for a 248 nm and for a 193 nm chemically amplified resist, respectively
This paper focuses on a novel methodology for a fast and efficient resist model calibration. One of the most crucial parts when calibrating a resist model is the fitting of experimental data where up to 20 resist model parameters are varied. Although general optimization approaches such as simplex algorithms or genetic algorithms have proven suitable for many applications, they do not exploit specific properties of resist models. Therefore, we have developed a new strategy based on Design of Experiment methods which makes use of these specific characteristics. This algorithm will be outlined and then be demonstrated by applying it to the calibration of a Solid-C resist model for one ArF line/space resist. As characterizing dataset we chose: a) a Focus Exposure Matrix (FEM) for the dense array, b) linearity, c) OPE (optical proximity) curve and e) the MEEF (mask error enhancement factor) for a dense array. It turned out that a simultaneous fit of the complete data set wa s not possible by varying resist parameters only. Considering the optical parameters appeared to be crucial as well. Therefore the influence of the optical settings (illumination, projection, 3D mask effects) on the lithography process will be discussed at this point. Finally we obtained an excellent matching of model predictions and experimental results
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.