Turning Data into Knowledge – Data-Led Catalyst Optimisation

Abstract

This thesis presents work focussed on screening ligands for Buchwald-Hartwig coupling
catalysis, analysis of trends observed in catalytic activity across ligand chemical space,
and seeks to make recommendations of optimal ligands for a given coupling.
After a review of the relevant literature and methodologies in Chapters 1, Chapters 2 and
3 describe the acquisition and analysis respectively of three sets of ligand screening data
for the coupling of aniline with 1-bromo-4-fluorobenzene. Data analysis with a variety of
regression techniques indicated that the underlying trend in ligand catalytic activity was
nonlinear in nature, and furthermore that there was likely a discontinuity in this trend,
hindering fitting with a single continuous function. Two non-parametric methods for
identification of optimal ligand properties were developed based on calculation of
weighted centroids and weighted dual Kernel Density Estimation (KDE). Application of
these methodologies allowed the recommendation of BippyPhos as a highly catalytically
active ligand for this coupling, and identification of the factors important for promoting
catalytic activity in this dataset.
Chapter 4 expands on the results of Chapters 2 and 3 by presenting eight ligand screening
sets gathered in collaboration with Bayer, and quantifying the manner in which optimal
ligand properties are affected by changes to reagent steric and electronic properties.
Alterations to the electronic properties of para-substituents on either coupling partner
were shown to have a substantial influence on the optimal ligand properties for catalysis,
but only when their steric properties were not also altered. Ligand recommendations
were also made for each of the eight pairs of coupling partners.
Chapter 5 presents a computational investigation of the catalytic cycle for the coupling
studied in Chapters 2 and 3, using seven of the ligands screened experimentally. By
comparison of energetic barriers for oxidative addition and reductive elimination in each
catalytic cycle, it was determined that the observed discontinuity in the trend guiding
ligand catalytic activity was likely due to a change in rate-determining step as a result of
changing ligand properties.
Chapter 6 details the optimisation of the synthesis of a phosphine/carbene ligand, and
demonstration of its varidentate binding ability. The ligand was also applied to the
Buchwald-Hartwig coupling studied in Chapters 2 and 3, though it was not active for this
catalysis.
As an additional note, Appendix E summarises the expansion, adjustment, and
correction of errors in the monodentate phosphine Ligand Knowledge Base (LKB-P). 75
new ligands were added to the LKB-P as part of this work, with a particular focus on
ligands relevant to coupling catalysis. The necessary exclusion of a subset of LKB-P
descriptors for the purposes of this research was discussed and justified.

Date of Award	17 Mar 2026
Original language	English
Awarding Institution	University of Bristol
Sponsors	UKRI EPSRC & Bayer AG
Supervisor	Natalie Fey (Supervisor), Paul G Pringle (Supervisor), Chris A Russell (Supervisor), Philipp Holstein (Supervisor) & Robert Webster (Supervisor)