Configuration-Sampling-Based Surrogate Models for Rapid Parameterization of Non-Bonded Interactions.

Area of Science:

• Computational Chemistry
• Molecular Modeling
• Chemical Engineering

Background:

• Accurate thermophysical properties, particularly vapor-liquid equilibria (VLE), are crucial for molecular simulations.
• These properties heavily rely on the precise parameterization of non-bonded interactions in classical force fields.
• Traditional parameterization methods are computationally intensive, requiring extensive direct molecular simulations.

Purpose of the Study:

• To develop and compare rapid approaches for force field parameterization and uncertainty quantification.
• To evaluate the efficacy of surrogate models in approximating direct molecular simulation results for non-bonded interactions.
• To assess the performance of Multistate Bennett Acceptance Ratio (MBAR) and Pair Correlation Function Rescaling (PCFR) methods.

Main Methods:

• Implementation of two configuration-sampling-based surrogate models: MBAR and PCFR.
• Coupling of surrogate models with the Isothermal Isochoric (ITIC) thermodynamic integration method.
• Estimation of vapor-liquid saturation properties using the combined methodologies.

Main Results:

• Demonstrated the utility of surrogate models for efficient high-dimensional parameterization and uncertainty quantification.
• Identified complementary roles for MBAR and PCFR: PCFR excels in exploring distant parameter space regions.
• MBAR proved more effective for refining parameters within local regions of the parameter space.

Conclusions:

• MBAR and PCFR serve as valuable, complementary tools for accelerating force field development.
• The proposed approach significantly reduces the computational cost associated with parameterizing non-bonded interactions.
• This methodology enhances the accuracy and reliability of molecular simulations for predicting thermophysical properties.