Automated Reaction Discovery Tools and their Application to Understanding Organometallic Reactions

Abstract

Using computational simulations to model reaction mechanisms has been a common way to garner kinetic and thermodynamic information of a chemical system that can help to understand reaction rates and selectivities. The most widely used tools available for reaction path modeling, however, lend themselves to working best when chemical intuition or experimental data for a chemical reaction is available to help build the simulation. In order to model reaction mechanisms of chemical systems that have not been well studied, or even tested experimentally yet, new tools for simulation are needed that do not rely on chemical intuition or experimental data. Herein, new methods developed for reaction discovery are laid out and applied to understanding transition metal catalyzed reactions on a deeper level. The first tool, ZStruct, allows for the systematic exploration of chemical space for unimolecular and bimolecular reactions for both organic and organometallic reactions. This method requires minimal user guidance, with the only prior knowledge required being that of which atoms on each reactant may be participating in bond forming/breaking processes. This method has been applied to explore the following reaction: (1) methane activation by cisplatin, where 10 previously unexplored reaction pathways were identified. (2) Ni(II)-catalyzed beta-hydride elimination, where off cycle Ni-THF or Ni-pyridine intermediates were found to be thermodynamically favorable. (3) Palladium-catalyzed piperidine arylation, where the full catalytic cycle, including the rate limiting C-H activation step, as well as roles of important Ar-I and Cs-salt additives, were elucidated through ZStruct’s ability to explore chemical space. The second tool, CGen, allows for sampling conformational changes that reactants of interest can undergo, and use them to generate metal-reactant complexes. Using the conformers that are generated, reaction discovery using ZStruct can be done to understand how conformational changes impact the mechanism by which a reaction occurs. Additionally, CGen can also be used to create catalyst-reagent complexes (reagents could be counterions, solvent molecules, or any other molecular additives used experimentally) by aligning a catalyst and reagent of interest in different orientations, thus allowing for better sampling of molecular interactions during a reaction. CGen was used to explore the chemical space of ethylene polymerization with a Ti-constrained geometry catalyst (Ti-CGC). With this system the impact of the polymer chain conformation on the mechanism to ethylene insertion was investigated. The favorable pathways to insertion that were found involve polymer chain conformations that maximize the distance between Ti and the polymer chain. The barriers to insertion found with the naked cation system were lower than experimental results, implying that inclusion of the experimentally required borane counterion is important for generating a model that reflects experimental observations. Therefore, the same reaction was remodeled once with Me-B(C6F5)3- and again with B(C6F5)4-, and in both cases CGen was used to generate Counterion-catalyst complexes. The sampling of various counterion alignments and polymer chain conformations demonstrated that alignment/conformer combinations can actually impede monomer insertion from occurring. Additionally, the models showed that the difference in nucleophilic strength between the two counterions impact how energetically challenging monomer uptake prior to insertion is, as well as how each counterion positions itself with respect to the catalyst during monomer insertion. The insights gained using reaction discovery methods in this work gave way to a deeper understanding of how chemical interactions in-situ may be hindering or helping a chemical reaction to occur.