Nonadiabatic Derivative Couplings Calculated Using Information of Potential Energy Surfaces without Wavefunctions: Ab...

Area of Science:

• * Theoretical Chemistry
• * Computational Chemistry
• * Quantum Chemistry

Background:

• * Nonadiabatic coupling matrix elements (NACMEs) are crucial for understanding polyatomic system dynamics.
• * Accurate NACME calculation is computationally demanding, often requiring wavefunction-based methods.
• * Existing methods face challenges in scalability and applicability to diverse electronic structure approaches.

Purpose of the Study:

• * To implement and validate an approximate algorithm for calculating NACMEs.
• * To assess the accuracy of the PES-based algorithm against wavefunction-based methods.
• * To evaluate the efficiency and accuracy of combining the algorithm with machine learning models.

Main Methods:

• * Developed an approximate algorithm to compute NACMEs using only potential energy surface (PES) information (energies, gradients, Hessians).
• * Applied the algorithm to the CH2NH system for comparison with accurate wavefunction-based calculations.
• * Integrated a machine learning (ML) model with the approximate PES-based algorithm for enhanced efficiency.

Main Results:

• * The approximate PES-based algorithm yields highly accurate NACMEs, comparable to wavefunction-based methods.
• * Accuracy is maintained except at energetically degenerate points (conical intersections).
• * The ML-integrated approach provides similarly accurate NACMEs with significantly improved computational efficiency.

Conclusions:

• * The approximate PES-based algorithm offers a viable and accurate alternative for NACME calculations.
• * Its compatibility with various electronic structure methods and ML models broadens its applicability.
• * This work facilitates the study of nonadiabatic dynamics in larger, complex systems.