Minimum Mode Saddle Point Searches Using Gaussian Process Regression with Inverse-Distance Covariance Function.

Area of Science:

• Computational chemistry
• Materials science
• Chemical physics

Background:

• Finding saddle points on energy surfaces is crucial for understanding chemical reactions and material properties.
• Traditional methods can be computationally intensive, especially when accurate energy and gradient evaluations are costly.
• Exploiting previous computational data is key to enhancing performance in these calculations.

Purpose of the Study:

• To develop a more efficient method for locating saddle points on potential energy surfaces.
• To reduce the computational cost associated with finding transition states in chemical systems.
• To improve the convergence rate of saddle point searches in computationally demanding scenarios.

Main Methods:

• Utilizing a Gaussian process model based on inverse interatomic distances to approximate the energy surface.
• Performing accurate energy and gradient evaluations at the saddle point of the approximate surface.
• Iteratively correcting the Gaussian process model with new accurate data points.
• Employing the minimum mode following method guided by the lowest curvature mode.

Main Results:

• The proposed method significantly reduces the number of energy evaluations required for convergence to a saddle point.
• The Gaussian process-based approach effectively accelerates the search for transition states.
• The method's performance was validated across various systems, including H2 adsorption on Cu(110) and gas-phase reactions.

Conclusions:

• The integration of Gaussian process models with minimum mode following offers a computationally efficient strategy for saddle point identification.
• This approach is particularly beneficial for systems where energy and gradient calculations are expensive.
• The enhanced performance demonstrated can accelerate the exploration of reaction mechanisms and material phase transitions.