SkaSim

The oftentimes stark approximation of nature is the biggest shortcoming of conventional phenomenological methods, which in turn limits their predictive capability. Numerous practice-oriented problems in the engineering science require thermodynamic property data for real fluids which are either very expensive or even impossible to sample experimentally. This is a problem that can be circumvented by the use of force field based simulation methods. Highly scalable molecular simulations can realise thermodynamic property data computation of complex fluid mixtures with semi-interactive response times of minutes, even seconds if HPC architectures, efficiently parallelized programs and appropriate molecular models are at hand. It is within the scope of this project to investigate the two parallelization methods OpenACC and OpenMP in terms of performance for highly parallel molecular dynamics simulations and the development of mathematical optimization methods for molecular models. The focus lies on the prediction of pure fluid properties, non-ideal mixture behavior of condensed phases and the investigation of nanoscale processes. Furthermore, the development of novel methods in the realm of fluid phase boundaries and nucleation in chemically reacting systems will be fostered.