Two-state reaction path search using a quantum Monte Carlo-inspired approach

We present an algorithm for finding chemical reaction pathways using a Monte Carlo transition state search (MCTSS) scheme. Our strategy is a bidirectional two-state approach that simultaneously drives two Monte Carlo trajectories from reactants to products, and vice versa, until the trajectories meet. The trajectories are driven in a Metropolis-like procedure with transition probabilities based on the real-space diffusion Monte Carlo algorithm. A computationally inexpensive structure preselection procedure is used to guide the two trajectories toward each other. We performed a proof-of-principle demonstration of the MCTSS algorithm for the model two-dimensional double-well potential and for the halogen anion S$_N$2-substitution in halogenated methane. The MCTSS approach presented here is expected to be particularly useful when employing electronic structure methods that do not provide analytic gradients.