New Strategies for Direct Methane-to-Methanol Conversion from Active Learning Exploration of 16 Million Catalysts

Abstract

Despite decades of effort, no earth-abundant homogeneous catalysts have been discovered that can selectively oxidize methane to methanol. We exploit active learning to simultaneously optimize methane activation and methanol release calculated with machine learning-accelerated density functional theory in a space of 16 M candidate catalysts including novel macrocycles. By constructing macrocycles from fragments inspired by synthesized compounds, we ensure synthetic realism in our computational search. Our large-scale search reveals that low-spin Fe(II) compounds paired with strong-field (e.g., P or S-coordinating) ligands have among the best energetic tradeoffs between hydrogen atom transfer (HAT) and methanol release. This observation contrasts with prior efforts that have focused on high-spin Fe(II) with weak-field ligands. By decoupling equatorial and axial ligand effects, we determine that negatively charged axial ligands are critical for more rapid release of methanol and that higher-valency metals [i.e., M(III) vs M(II)] are likely to be rate-limited by slow methanol release. With full characterization of barrier heights, we confirm that optimizing for HAT does not lead to large oxo formation barriers. Energetic span analysis reveals designs for an intermediate-spin Mn(II) catalyst and a low-spin Fe(II) catalyst that are predicted to have good turnover frequencies. Our active learning approach to optimize two distinct reaction energies with efficient global optimization is expected to be beneficial for the search of large catalyst spaces where no prior designs have been identified and where linear scaling relationships between reaction energies or barriers may be limited or unknown.

Using Computational Chemistry To Reveal Nature’s Blueprints for Single-Site Catalysis of C–H Activation

“…In contrast, Mn­(IV)-oxo compounds, which should have more stable metal-oxo moieties according to the “oxo wall” theory, show comparable C–H bond reactivity to Fe­(IV)-oxo compounds, despite differences in spin states and electron configurations. − DFT and CASSCF studies found that HAT barriers are comparable between Mn and Fe compounds, emphasizing that Mn compounds are underexplored for C–H bond reactivity . Nandy et al recently explored a space of 16 million TMCs for C–H activation without assuming an N- or O-containing primary coordination sphere and found that low-spin Fe­(IV)-oxo compounds with strong-field (e.g., P- or S-coordinating) ligands have the best tradeoff between HAT energetics and methanol release (Figure ). Yadav et al have recently designed Fe TMCs to obtain product halogenation selectivity only seen in enzymes .…”

“…A Pareto front, as computed by DFT for catalysts with the optimal tradeoff between the two reaction energies, is annotated in gray. [Reproduced with permission from ref . Copyright 2022, American Chemical Society, Washington, DC.…”

Catalytic properties of the ferryl ion in the solid state: a computational review

“…For example, a database composed of 16-million Mn and Fe catalysts has recently been developed with novel tetradentate macrocycles and coordinating axial ligands. 234 Neural network have also been applied to the classification of nonperiodic atomic configurations relaxed using DFT methods (B3LYP, 113,115,116 LANL2DZ effective core potential 235 and 6-31G* basis set 236 ) to determine the reaction energies for the radical CH 3 ˙rebound mechanism for the conversion of methane to methanol, using N 2 O to regenerate the catalytic sites. In addition, deep neural networks have been employed very recently to carry out an extensive reactivity screening for all possible coupling intermediates, including ferryl species, to predict the degradation of organic micropollutants by iron oxychloride (FeOCl).…”

Computational evolution of new catalysts for the Morita–Baylis–Hillman reaction

“…In some cases (e.g. [12]) the process is repeated, i.e. the new QM data is used to update the ML model and the library is re-screened.…”

“…A typical state-of-the-art computational organometallic catalyst discovery study (e.g. [11,12]) involves libraries of 10 4 -10 6 catalyst candidates constructed using predefined metals and ligands. The activity is then predicted for a small subset [O( 103 )] using QM based on a linear free energy relation between energies of key intermediates in the catalytic mechanism and the reaction rate.…”