Navigating molecular space for reaction mechanisms: an efficient, automated procedure

Abstract

Mechanism is a core chemical concept that has vital implications for reaction rate, efficiency and selectivity. The discovery of mechanism is not easy due to the great diversity of possible chemical rearrangements in even relatively simple systems. For this reason, mechanisms involving bond breaking and forming are usually proposed via chemical intuition – which limits the scope of considered possibilities – and these hypotheses are then tested using simulation or experiment. This article discusses an automated simulation strategy for investigating multiple elementary step reaction mechanisms in chemical systems. The method starts from a single input structure and seeks out nearby intermediates, optimises the proposed structures and then determines the kinetic viability of each elementary step. The kinetically accessible intermediates are catalogued and new searches are performed on each unique structure. This process is repeated for an arbitrary number of steps without human intervention, and massively parallel computation enables fast searches in chemical space. Importantly, this strategy can be empirically shown to lead to a finite number of accessible structures, not a combinatorial explosion of intermediates. Therefore, the method should be able to predict multi-step reaction pathways in many interesting chemical systems. Demonstrations on organic reactions and a hydrogen storage material, ammonia borane, show that the herein proposed strategy can uncover complex reactivity without relying on existing chemical intuition.