Smooth Things Come in Threes: A Diabatic Surrogate Model for Conical Intersection Optimization.

Area of Science:

• Computational Chemistry
• Theoretical Chemistry
• Quantum Chemistry

Background:

• Optimizing conical intersection structures is challenging due to non-differentiable adiabatic potential energy surfaces.
• Accurate characterization of conical intersections is crucial for understanding photochemical reactions.

Purpose of the Study:

• To develop a computationally efficient method for optimizing conical intersection structures.
• To create a smooth and differentiable surrogate model for non-differentiable adiabatic surfaces.

Main Methods:

• Gaussian process regression was employed to build a pseudodiabatic surrogate model.
• The surrogate model consists of three smooth, differentiable surfaces approximating the adiabatic surfaces.
• Restricted variance optimization was used in conjunction with the surrogate model.

Main Results:

• The pseudodiabatic surrogate model accurately reproduces the adiabatic potential energy surfaces.
• A notable decrease in computational effort was achieved for locating minimum energy crossing points.
• The combination of the surrogate model and optimization method proved effective.

Conclusions:

• The pseudodiabatic surrogate model offers a viable solution for optimizing conical intersection structures.
• This method significantly enhances computational efficiency in theoretical chemistry.
• The approach facilitates the study of systems involving conical intersections.