Transition metal complexes catalyse many reactions important in synthetic organic and organometallic chemistry. However, we rarely have a quantitative understanding of what will happen when we change the input variables (ligands, substrates, experimental conditions) in a catalytic reaction and new catalysts are usually discovered by chance or after extensive experimental screening and optimisation of the reaction conditions.

My work and that of members of my group has been focussed on computational studies of synthetically relevant organometallic catalysts, especially those used in hydroformylation and cross-coupling reactions.

To achieve this, we have combined different computational approaches to study the effects of changing the ligands on the mechanism of reaction. For many ligands used synthetically, this also means exploring the effects of conformational changes and multiple isomers on the reaction pathway, as well as probing multiple mechanistic possibilities. We extract structural and energetic parameters from these calculations and then explore whether such descriptors could be used to make predictions for other ligands, using multivariate regression models and related statistical approaches (all the way to what you might call machine learning). More recently, we have moved beyond ligands to also consider substrates and catalysts more broadly. We collaborate closely with synthetic chemists in academia and industry to test such predictions against experimental reality and many group members are co-supervised by experimentalists. Putting prediction before synthesis is a key long-term goal, but it is often the availability of “the right kind” of experimental data which hampers the application of machine-learning in our field.

Group webpage:

https://feygroupchem.wordpress.com/