Enhanced climbing image nudged elastic band method with Hessian eigenmode alignment.

Area of Science:

• Computational Chemistry
• Materials Science
• Chemical Physics

Background:

• Accurate transition state determination is crucial for understanding reaction kinetics.
• Existing methods like CI-NEB and MMF have limitations in computational cost, convergence, and relevance of identified saddle points.
• High-throughput automated chemical discovery requires efficient methods for mapping atomic rearrangements.

Purpose of the Study:

• To develop and benchmark an adaptive hybrid algorithm that integrates CI-NEB and MMF methods.
• To achieve faster and more reliable convergence to relevant saddle points for transition state calculations.
• To reduce the computational expense associated with identifying atomic rearrangements.

Main Methods:

• An adaptive hybrid algorithm switching between CI-NEB and MMF methods was developed.
• The algorithm was benchmarked on the Baker-Chan saddle point test set and the OptBench set (heptamer island on Pt (111)).
• A PET-MAD machine-learned potential was used for the calculations.

Main Results:

• The hybrid method demonstrated a median reduction of 57% in energy and force calculations for the BC set and 31% for the heptamer island transitions, compared to CI-NEB.
• A simple switching strategy (CI-NEB to MMF below 0.5 eV/Å) required 46% more force calculations than OCI-NEB.
• Bayesian analysis confirmed the efficiency gains of the adaptive hybrid approach.

Conclusions:

• The adaptive hybrid method offers a highly efficient alternative for transition state calculations.
• This approach can accelerate high-throughput automated chemical discovery of atomic rearrangements.
• The integration of CI-NEB and MMF provides a robust and computationally economical solution.