Multigrid solver for the reference interaction site model of molecular liquids theory

Area of Science:

• Computational chemistry
• Physical chemistry
• Numerical analysis

Background:

• The Reference Interactions Site Model (RISM) is crucial for calculating solvation properties.
• Existing RISM solvers face computational challenges, particularly in achieving high accuracy efficiently.
• Optimizing RISM parameterization is key to reliable hydration free energy predictions.

Purpose of the Study:

• To introduce a novel multigrid-based algorithm for solving RISM integral equations.
• To analyze the impact of algorithm parameters on the numerical accuracy of RISM hydration free energy calculations.
• To provide guidelines for parameter selection to minimize computational cost.

Main Methods:

• Development of a new multigrid-based algorithm tailored for RISM integral equations.
• Systematic analysis of the algorithm's performance using polar and nonpolar compounds.
• Comparative performance evaluation against one-grid Picard iteration and nested Picard iteration methods.

Main Results:

• The proposed multigrid method demonstrates significant speed improvements over traditional RISM solvers.
• Achieved over 30x speedup compared to the one-grid iteration method.
• In high-accuracy scenarios, the method is nearly 7x faster than nested Picard iteration.

Conclusions:

• The novel multigrid algorithm offers a computationally efficient and accurate approach for RISM calculations.
• The findings provide practical guidance for optimizing RISM parameter selection.
• This advancement has the potential to accelerate molecular simulations and solvation studies.