Molecular dynamics simulations are essential for studying atomic and molecular behavior,
utilizing potential functions to model interactions between sets of two or more particles.
While pairwise potentials, potentials computing interaction between two particles, are widely
popular, they can lack accuracy, necessitating the inclusion of interactions between three
particles. However, these interactions, referred to as three-body interactions, possess a
higher computational complexity, which poses a significant burden for their real-world
applicability. To mitigate this issue, this work implements zonal methods, a class of node-
level parallelization methods, to parallelize molecular dynamics simulations considering
three-body interactions. For this purpose, we have derived theoretical requirements for
applying zonal methods with three-body interactions, and created an efficient framework on
top of these insights. In this framework, the fullshell, halfshell and midpoint methods were
implemented, which, evaluated on a distributed system, demonstrated the high scalability of
zonal methods, thus enhancing the feasibility of high-fidelity molecular dynamics simulations.
«
Molecular dynamics simulations are essential for studying atomic and molecular behavior,
utilizing potential functions to model interactions between sets of two or more particles.
While pairwise potentials, potentials computing interaction between two particles, are widely
popular, they can lack accuracy, necessitating the inclusion of interactions between three
particles. However, these interactions, referred to as three-body interactions, possess a
higher computational complexity, which p...
»