A Graphics Processing Unit Implementation of Coulomb Interaction in Molecular Dynamics.

Area of Science:

• Computational chemistry
• Materials science
• Physical chemistry

Background:

• Long-range electrostatic interactions are crucial in molecular dynamics simulations of electrolytes.
• Traditional Ewald sum methods on CPUs can be computationally intensive.
• Efficient simulation methods are needed to study electrolyte behavior.

Purpose of the Study:

• To report a GPU implementation of the orientation-averaged Ewald sum in HOOMD Blue.
• To compare the performance of the GPU implementation against a CPU-based Ewald sum in LAMMPS.
• To validate the accuracy of the GPU implementation for electrolyte simulations.

Main Methods:

• GPU implementation of the orientation-averaged Ewald sum scheme in HOOMD Blue.
• Performance comparison with an optimized CPU version of the Ewald sum in LAMMPS.
• Molecular dynamics simulations of monovalent and divalent hydrated salts at various concentrations.

Main Results:

• The GPU implementation demonstrates significantly higher performance than the CPU version for systems up to ~10^5 particles.
• Thermodynamic properties (electrostatic energy, heat capacity) and structural properties (radial distribution functions, integrated charge) were calculated.
• Excellent agreement was observed between the GPU and CPU methods for all calculated properties.

Conclusions:

• The GPU-accelerated Ewald sum method provides a computationally efficient and accurate approach for simulating electrolytes.
• This method enables the study of bulk electrolyte properties across a wide range of ionic concentrations.
• The HOOMD Blue GPU implementation offers a valuable tool for advancing research in electrolyte systems.