Nudged Elastic Band Calculations Accelerated with Gaussian Process Regression Based on Inverse Interatomic Distances

Area of Science:

• Computational chemistry
• Materials science
• Surface science

Background:

• Calculating minimum energy paths for atomic rearrangements is crucial in materials science.
• The nudged elastic band (NEB) method is commonly used but computationally expensive.
• Gaussian process regression (GPR) can accelerate NEB calculations by reducing energy evaluations.

Purpose of the Study:

• To improve the efficiency and accuracy of GPR for accelerating NEB calculations.
• To address limitations of existing GPR approaches when dealing with high-force configurations.
• To develop a more robust method for studying atomic rearrangements on surfaces.

Main Methods:

• Modified GPR covariance function using inverted interatomic distances.
• Introduction of an early stopping criterion for path relaxation.
• Application to dissociative adsorption of H2 on Cu(110) and H2O diffusion on ice Ih(0001).

Main Results:

• The revised GPR approach significantly improves performance in challenging cases.
• Reduced number of energy and force evaluations required for convergence.
• Successful application to dissociative adsorption and diffusion processes, outperforming original GPR.

Conclusions:

• The enhanced GPR method offers a more efficient and reliable way to compute minimum energy paths.
• This advancement is valuable for studying complex surface chemistry and materials dynamics.
• The improved method accelerates discovery in surface science and catalysis research.