Accelerating molecular dynamics simulations using fast Ewald summation with prolates.

Area of Science:

• Computational physics and chemistry
• Materials science
• Biophysics

Background:

• Long-range Coulomb interactions are computationally expensive in molecular dynamics (MD) simulations.
• Existing methods like Particle Mesh Ewald (PME) and Particle-Particle-Particle-Mesh (PPPM) offer near-linear scaling but still incur significant costs.
• Efficient calculation of electrostatic interactions is crucial for accurate and timely MD simulations.

Purpose of the Study:

• To introduce a novel ESP-Ewald summation method utilizing prolate spheroidal wave functions (PSWFs).
• To enhance the efficiency of Fourier-based electrostatics calculations in MD.
• To reduce computational overhead, communication, and particle-grid operations without compromising accuracy.

Main Methods:

• Developed ESP-Ewald summation employing PSWFs for improved Fourier representation.
• Integrated the ESP method into LAMMPS and GROMACS open-source MD packages.
• Compared ESP performance against PME/PPPM baselines across various error tolerances and core counts.

Main Results:

• ESP-Ewald achieved a 3-fold acceleration of electrostatic interactions and a 2.5-fold MD speed-up at error tolerances of 10^-3 to 10^-4 using ~10^3 cores.
• At higher accuracy (10^-5), ESP provided a 10-fold acceleration for electrostatics and a 5-fold MD speed-up.
• The accelerated codes demonstrated improved strong scaling with increasing core counts and were validated in biological and material simulations.

Conclusions:

• ESP-Ewald summation offers a significant acceleration for MD simulations by optimizing electrostatic calculations.
• This method provides a practical, drop-in solution to reduce simulation time and energy consumption.
• ESP-Ewald is a valuable advancement for large-scale molecular dynamics workflows in various scientific domains.