Transition State Search Using a Guided Direct Inversion in the Iterative Subspace Method.

Area of Science:

• Computational Chemistry
• Chemical Physics
• Theoretical Chemistry

Background:

• Transition state (TS) optimization is crucial for understanding chemical reaction mechanisms.
• Traditional methods like quasi-Newton rational function optimization (RFO) can struggle with complex potential energy surfaces.
• Efficient algorithms are needed to accurately locate saddle points on these surfaces.

Purpose of the Study:

• Introduce and evaluate a novel transition state optimization method, guided energy-represented direct inversion in the iterative subspace (gEn-DIIS).
• Compare the performance of gEn-DIIS against the established RFO method.
• Develop and assess a hybrid approach combining RFO and gEn-DIIS for improved optimization efficiency.

Main Methods:

• Implementation of the gEn-DIIS algorithm for transition state optimization.
• Development of a hybrid optimization strategy integrating RFO and gEn-DIIS.
• Comparative analysis using a suite of test molecules to evaluate optimization speed and robustness.

Main Results:

• The gEn-DIIS method demonstrates significantly faster optimization convergence compared to RFO.
• The hybrid RFO-gEn-DIIS approach shows practical advantages across various convergence stages.
• gEn-DIIS proves particularly effective for challenging optimizations involving flat reaction paths.

Conclusions:

• The gEn-DIIS algorithm represents a computationally efficient advancement for transition state optimization.
• The hybrid method offers a versatile and robust tool for exploring chemical reaction pathways.
• This work provides a valuable alternative for computational chemists dealing with difficult saddle point searches.