A formally exact theory to construct nonreactive forcefields using linear regression to optimize bonded parameters.

Area of Science:

• Computational Chemistry
• Theoretical Chemistry
• Materials Science

Background:

• Developing accurate and efficient nonreactive forcefields is crucial for molecular simulations.
• Existing forcefields often face challenges in formal exactness, parameter optimization, and accurate representation of molecular dynamics.

Purpose of the Study:

• To establish the theoretical foundations of Force Field Functional Theory (FFFT).
• To introduce novel angle-bending and bond-stretch potentials for improved accuracy and differentiability.
• To demonstrate efficient parameterization and feature selection methods for forcefields.

Main Methods:

• Derivation of theoretical framework for FFFT.
• Development of continuously differentiable angle-bending and first-principles-derived bond-stretch potentials.
• Application of linear regression for parameter optimization via a new ansatz.
• Implementation of embedded feature selection for computational efficiency.

Main Results:

• New angle-bending potential shows good agreement with CCSD calculations for various molecules.
• New bond-stretch potential accurately reproduces vibrational energy levels for H2 and O2.
• The new ansatz significantly reduces sensitivity to nonbonded parameters in force constant optimization.
• Optimized flexibility models yield vibrational frequencies within a few percent of experimental values.

Conclusions:

• FFFT provides a robust theoretical framework for developing advanced nonreactive forcefields.
• The novel potentials and parameterization methods enhance accuracy, efficiency, and applicability of molecular simulations.
• This approach offers significant advantages for computational chemistry and materials science applications.