Fast fragments: the development of a parallel effective fragment potential method

Area of Science:

• Computational chemistry
• Quantum mechanics
• Molecular modeling

Background:

• The Effective Fragment Potential (EFP) method reduces computational cost for solvation studies by partitioning systems.
• Current EFP implementations require optimization for larger systems and dynamic simulations.
• Efficient calculation of fragment-fragment and QM-EFP interactions is crucial for broader EFP applicability.

Purpose of the Study:

• To enhance the computational efficiency of the Effective Fragment Potential (EFP) method.
• To implement parallelization strategies for EFP calculations within the GAMESS software.
• To improve the scalability of EFP methods for larger chemical systems.

Main Methods:

• Parallelization of EFP fragment-fragment and QM-EFP interaction energy and gradient computations.
• Development of a new algorithm for the iteratively self-consistent polarization term using nonblocking communication.
• Integration of these parallelization techniques into the GAMESS computational chemistry suite.

Main Results:

• Achieved reasonable speedup in computational time for EFP calculations.
• Demonstrated improved scalability with increasing numbers of processors.
• Successfully applied parallelized EFP methods to water clusters of various sizes.

Conclusions:

• Parallelization of EFP computations offers significant performance improvements.
• The new polarization algorithm enhances scalability for QM-EFP interactions.
• Optimized EFP methods are essential for studying larger, complex chemical systems and dynamics.