Efficient transition state (TS) search algorithms such as the partitioned-rational function optimization (P-RFO) typically require the hessian of the energy as input in addition to the TS guess structure. The cost of exact hessian calculation is prohibitive when electronic structure methods are used to study large molecules. Instead, an approximate hessian can be constructed using a unit matrix in delocalized internal coordinates, as well as curvature and reaction coordinate information generated by any interpolation algorithm that determines the TS guess. The resulting matrix has exactly one negative eigenvalue by design, which is highly desirable for a first-order saddle point search. An analysis of the performance of P-RFO using exact and approximate hessian inputs demonstrates that the cost reduction achieved in sidestepping the exact hessian calculation outweighs the increase in iterations associated with an approximate input. The approximate hessian construction method, therefore, renders the study of large systems more computationally viable.

Transition state search without exact hessian evaluation

245th National Meeting (2013)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.