Gaussian Process Regression for Transition State Search.

Area of Science:

• Computational Chemistry
• Chemical Physics

Background:

• Transition state (TS) search is crucial for understanding chemical reaction mechanisms.
• Existing methods like the dimer method can be computationally expensive.

Purpose of the Study:

• To develop and validate a novel gradient-based algorithm for efficient transition state search.
• To provide a robust method for identifying starting points for TS optimization.

Main Methods:

• Implementation of a gradient-based optimization algorithm incorporating Gaussian process regression.
• Development of a strategy to determine initial guesses for optimization using reactant and product minima.
• Benchmarking against established methods (dimer, PRFO) using the DL-FIND library on 27 test systems.

Main Results:

• The proposed algorithm significantly reduces the number of energy and gradient evaluations required.
• Demonstrated efficiency across a diverse set of 27 chemical systems.
• Successful identification of starting points for optimization.

Conclusions:

• The Gaussian process regression-based gradient algorithm offers a more efficient approach to transition state searches.
• This method provides a practical solution for determining TS optimization starting points.
• The findings suggest potential for accelerating computational chemistry simulations.