Parallel unconstrained minimization of potential energy in LAMMPS
Sandia Labs., Livermore, CA (United States), T Plantenga · 1997
This report describes a new minimization capability added to LAMMPS V4.0. Minimization of potential energy is used to find molecular conformations that are close to structures found in nature. The new minimization algorithm uses LAMMPS subroutines for calculating energy and force vectors, and follows the LAMMPS partitioning scheme for distributing large data objects on multiprocessor machines. Since gradient-based algorithms cannot tolerate nonsmoothness, a new Coulomb style that smoothly cuts off to zero at a finite distance is provided. This report explains the minimization algorithm and its parallel implementation within LAMMPS. Guidelines are given for invoking the algorithm and interpreting results.