Configuration-Space Sampling in Potential Energy Surface Fitting: A Space-Reduced Bond-Order Grid Approach

Area of Science:

• Computational chemistry
• Theoretical chemistry
• Quantum mechanics

Background:

• Accurate potential energy surfaces (PESs) are crucial for simulating reactive molecular systems.
• Traditional methods for generating PESs involve fitting or interpolating ab initio energies, requiring extensive sampling of nuclear configurations.

Purpose of the Study:

• To develop an automated procedure for optimal configuration-space sampling in generating ab initio energies for PES construction.
• To improve the efficiency and accuracy of PES fitting/interpolation for few-atom reactive systems.

Main Methods:

• Proposed an automated sampling scheme within the GEMS (grid-empowered molecular simulator) framework.
• Utilized a space-reduced formulation of bond-order variables for balanced representation of configuration space.
• Benchmarked performance using H2 and H3 systems with local-interpolation (modified Shepard) and global-fitting (Aguado-Paniagua) schemes.

Main Results:

• Uniform grids based on space-reduced bond-order variables demonstrated superior convergence compared to bond-length variables.
• The proposed method achieved better fitting/interpolation of PESs to ab initio data with increasing grid points.
• Validated the effectiveness of the automated sampling and bond-order variable approach on prototype systems.

Conclusions:

• The automated configuration-space sampling using space-reduced bond-order variables offers an efficient and accurate approach for constructing potential energy surfaces.
• This method enhances the reliability of dynamics calculations for reactive systems.
• The findings provide a valuable tool for computational chemists and physicists studying molecular reactions.