Efficient Two-Step Procedures for Locating Transition States of Surface Reactions

Area of Science:

• Computational Chemistry
• Surface Science
• Reaction Dynamics

Background:

• Accurate identification of transition states (TS) is essential for understanding reaction mechanisms.
• Locating TS for reactions involving multiple atoms on metal surfaces presents significant computational challenges.

Purpose of the Study:

• To develop and evaluate efficient two-step strategies for accurately locating transition states (TS) of reactions.
• To assess the performance of various path-finding and optimization methods for TS identification.

Main Methods:

• Employed a combination of four path-finding methods (nudged elastic band, standard string, climbing image string, searching string) for initial TS structure localization.
• Utilized conjugate gradient or modified steepest-descent optimization methods within two coordinate systems (Cartesian and mixed).
• Incorporated local refinement using two dimer methods, including a novel dimer-Lanczos variant.

Main Results:

• The dimer-Lanczos variant reduced gradient calls by approximately 20% compared to the standard dimer method.
• Using mixed coordinates during path finding decreased gradient calls by an additional 21% on average.
• Employing a modified steepest-descent method further improved efficiency by an average of 13%.

Conclusions:

• The proposed two-step strategies significantly enhance the efficiency of locating transition states for complex surface reactions.
• Specific method combinations (mixed coordinates, modified steepest-descent, dimer-Lanczos) offer substantial computational savings.
• Tailored strategies are recommended for addressing particularly challenging reaction cases.