Molecular Rotors in Viscous Fluids - A Numerical Analysis Aid by GPU Computing.

Molecular rotors are microscopic structures composed of two coupled parts that can rotate mutually. They have promising applications in viscosity measurement and flow detection in micro-sized and complex geometries. We propose a model for studying molecular rotors as classical systems moving in viscous environments, subjected to internal interactions and stochastic forces. Our model is expressed as a set of non-linear stochastic Langevin equations that are solved numerically using a Brownian Dynamics procedure. Attention is focused on the calculation of the two-time correlation function of the internal angular variable. For small internal forces, this correlation turns out to have a very slow time decay and its correct estimation requires long numerical experiments. We propose a CUDA implementation for a computer cluster with TESLA GPUs that calculates angular correlation functions in parallel. The implementation reduces the used computational time considerably in comparison with the one consumed by a usual serial scheme. It allows for the simulation of massive molecular rotors ensembles from which reliable results can be obtained. It is discussed how the GPU implementation can be improved in modern GPUs architectures.