Pd–Methyl Bond EnergyProperty Correlations, Noncorrelations, Machine Learning Models, and Application to Polymerization Catalysis

Metal–carbon bonds are a key intermediate in a
variety of
homogeneous organometallic transformations and often determine the
critical thermodynamics and kinetics of catalytic processes. Surprisingly,
the influence of different ligands on metal–carbon bond strengths
has been largely overlooked. Here we evaluated nearly 700 experimental
Pd–methyl complexes by calculating their bond dissociation
energies using density functional theory (DFT) and compared these
bond strengths to several fundamental molecular properties, and this
revealed several surprising correlations and noncorrelations. Most
surprising was that several fundamental properties, such as the bond
length, bond force constant, and bond electron density, have no correlation
with bond strength, despite these correlations often holding for main-group
compounds. We were indeed able to identify key ligand-dependent chemical
features/descriptors that provided a highly accurate machine learning
model and provided insight into the general factors that control the
Pd–carbon bond strength, such as radical delocalization and
nucleophilicity. Insights gained from the Pd–Me bond energy
analysis were then applied to CO migratory insertion steps that are
part of copolymerization reactions.