A combined method for determining reaction paths, minima, and transition state geometries

Mapping out a reaction mechanism involves optimizing the reactants and products, finding the transition state and following the reaction path connecting them. Transition states can be difficult to locate and reaction paths can be expensive to follow. We describe an efficient algorithm for determining the transition state, minima and reaction path in a single procedure. Starting with an approximate path represented by N points, the path is iteratively relaxed until one of the N points reached the transition state, the end points optimize to minima and the remaining points converged to a second order approximation of the steepest descent path. The method appears to be more reliable than conventional transition state optimization algorithms, and requires only energies and gradients, but not second derivative calculations. The procedure is illustrated by application to a number of model reactions. In most cases, the reaction mechanism can be described well using 5 to 7 points to represent the transition state, the minima and the path. The computational cost of relaxing the path is less than or comparable to the cost of standard techniques for finding the transition state and the minima, determining the transition vector and following the reaction path on both sides of the transition state.