Density functional study of methane activation by frustrated Lewis pairs with group 13 trihalides and group 15 pentahalides and a machine learning analysis of their barrier heights

Frustrated Lewis pairs (FLP) are an important advancement in metal-free catalysis. FLPs activate a variety of small molecules, notably dihydrogen. Methane activation, however, has not been reported despite it being an abundant chemical feedstock. DFT calculations were utilized to elucidate the reaction mechanism of methane activation by triel trihalide and pnictogen pentahalide-ammonia Lewis pairs. Two reaction mechanisms were modelled for methane activation: proton abstraction and hydride abstraction. In all cases, deprotonation was thermodynamically and kinetically favored versus hydride abstraction. The use of heavier pnictogens and bigger triels were calculated to be more favorable for the activation of methane. To discern factors affecting the activation energies, different descriptors were correlated – ground state thermodynamics, orbital energies, transition state strain energies, etc. - but no consistent patterns were identified. Thus, machine learning methods were used to correlate ground state parameters to barrier heights. A neural network was used to correlate ground state descriptors (global electrophilicity index, bond dissociation energies, reaction energies) to activation free energies (R² = 0.90).