Molecular dynamics simulations are widely used in fields such as biology, thermodynamics, and physics to study molecular systems at the atomic scale. Each molecule is modeled as a particle consisting of one or more interaction sites. While single-site models simplify simulations, they may reduce accuracy. Multi-site models, in which each site represents an atom or interaction point, provide a more detailed representation by computing interactions between all sites of two particles. These simulations often involve large numbers of particles and require very small time steps in the femtosecond range, which makes them computationally expensive. To obtain meaningful results within a reasonable time frame, high-performance computing and efficient algorithms are essential. This thesis explores the use of the multiple time-stepping algorithm r-RESPA to accelerate multi-site simulations with the autotuning software AutoPas on a high-performance computing cluster. The r-RESPA algorithm allows forces to be integrated at different time step sizes, improving performance when some forces can be computed less frequently than others. To apply this idea, the force calculation is divided by interaction distance. For close particle pairs, all interactions between molecular sites are computed using the full multi-site potential. For more distant particle pairs, two strategies are investigated. The first continues to use the full multi-site representation at larger time intervals. The second replaces the detailed model with a coarse-grained, single-site potential that is derived using the Iterative Boltzmann Inversion. The evaluation methodology is presented in detail, results from both approaches are analyzed and compared, and possibilities for future improvements are discussed.     «

Molecular dynamics simulations are widely used in fields such as biology, thermodynamics, and physics to study molecular systems at the atomic scale. Each molecule is modeled as a particle consisting of one or more interaction sites. While single-site models simplify simulations, they may reduce accuracy. Multi-site models, in which each site represents an atom or interaction point, provide a more detailed representation by computing interactions between all sites of two particles. These simula...     »