Molecular-dynamics simulations of nucleation from vapor to liquid were performed and different clustering techniques from miscellaneous domains of computer science and thermodynamics were used to automatically detect the emerging droplets. A new algorithm to detect these droplets is proposed which needs no additional information.

Additionally tracking of the droplets beyond various timesteps was implemented and a parallel version with both cluster detection and tracking was developed. The results of the sequential as well as the parallel algorithm were verified visually, checked for physical correctness and compared with results from literature.