COMPASS: Double-ended saddle point search as a constrained optimization problem

Area of Science:

• Computational Chemistry
• Chemical Physics
• Materials Science

Background:

• First-order saddle points on potential energy surfaces (PES) are crucial for understanding chemical reactions and molecular transitions.
• Accurate identification of these transition states is essential for predicting reaction rates and mechanisms.

Purpose of the Study:

• To develop and present a novel algorithm for efficiently locating first-order saddle points on the PES.
• To provide a robust and flexible method for exploring reaction pathways and transition states in complex systems.

Main Methods:

• The algorithm formulates saddle point finding as a constrained optimization problem using two sets of atomic coordinates (images).
• It employs a time-varying distance constraint and an energy difference constraint, pulling images from different valleys towards each other.
• The optimization combines quasi-Newton methods with linear constraints, recursively handling path splitting if images cross the barrier.

Main Results:

• The method was successfully tested on Lennard-Jones-38 cluster transitions and 121 molecular reactions using density functional theory (DFT).
• Demonstrated superior efficiency in energy and force evaluations compared to competing methods, particularly when not switching to single-ended approaches.
• Showcased robustness and flexibility through continuous search path construction and the ability to focus on arbitrary path subsegments.

Conclusions:

• The presented algorithm offers an efficient and reliable approach for identifying first-order saddle points on potential energy surfaces.
• Its ability to handle complex pathways and its computational efficiency make it a valuable tool for computational chemistry and materials science research.
• The method enhances the study of molecular reactions and cluster dynamics by providing accurate transition state information.