Machine Learning Accelerates the Discovery of Design Rules and Exceptions in Stable Metal–Oxo Intermediate Formation

Abstract

Metal−oxo moieties are important catalytic intermediates in the selective partial oxidation of hydrocarbons and in water splitting. Stable metal−oxo species have reactive properties that vary depending on the spin state of the metal, complicating the development of structure−property relationships. To overcome these challenges, we train machine-learning (ML) models capable of predicting metal−oxo formation energies across a range of first-row metals, oxidation states, and spin states. Using connectivity-only features tailored for inorganic chemistry as inputs to kernel ridge regression or artificial neural network (ANN) ML models, we achieve good mean absolute errors (4−5 kcal/mol) on set-aside test data across a range of ligand orientations. Analysis of feature importance for oxo formation energy prediction reveals the dominance of nonlocal, electronic ligand properties in contrast to other transition metal complex properties (e.g., spin-state or ionization potential). We enumerate the theoretical catalyst space with an ANN, revealing expected trends in oxo formation energetics, such as destabilization of the metal−oxo species with increasing d-filling, as well as exceptions, such as weak correlations with indicators of oxidative stability of the metal in the resting state or unexpected spin-state dependence in reactivity. We carry out uncertainty-aware evolutionary optimization using the ANN to explore a >37 000 candidate catalyst space. New metal and oxidation state combinations are uncovered and validated with density functional theory (DFT), including counterintuitive oxo formation energies for oxidatively stable complexes. This approach doubles the density of confirmed DFT leads in originally sparsely populated regions of property space, highlighting the potential of ML-model-driven discovery to uncover catalyst design rules and exceptions.

Seeing is Believing: Experimental Spin States from Machine Learning Model Structure Predictions

“…The emerging area of machine learning (ML)-accelerated high-throughput computational screening 20,21,63−75 has led to exploration of much larger chemical spaces 76,77 over which no one-size-fits-all exchange correlation functional can be expected to be predictive. Our group has developed representations 70,71 for training ML (e.g., artificial neural networks or ANNs) models to predict spin-state ordering to within sub-kilocalorie per mole accuracy of the DFT training data 70,71,78 and demonstrated the use of these models in the design of a range of spin-state-dependent catalytic 75 and materials 63,70−73,78−80 properties. In evaluating such ML model predictions, we treated DFT as the ground truth, despite its limitations in predicting ground-state spin.…”

Progress in Computational and Machine‐Learning Methods for Heterogeneous Small‐Molecule Activation

“…[87] In order to develop designing principles for the methane activation on the metal-organic framework catalysts, Rosen et al performed DFT calculations to understand the structure-activity relationship. [151] The study reveals that the activation energy of the CH scission and the metal-oxi site formation energy have an inverse relationship. Xu et al found that the classical BEP relationship is not sufficient to be used for metal oxide chemistry in the interest of methane activation, thus they have implemented LASSO to identify key descriptors for predicting.…”

“…In this section, we discuss the applications of fundamental theory and methodology for the conversion of H 2 , O 2 , H 2 O, N 2 , CO 2 , and CH 4 activation. We refer to other refs for in-depth reviews (H 2 , [149][150][151] O 2 , [152,153] H 2 O, [154][155][156] N 2 , [157][158][159][160] CO 2 , [161][162][163] and CH 4 [4,5] ). Here, we focus on the demonstrations of the catalyst design methods such as the Sabatier principle, d-band theory, high-throughput screening, and machine learning for the activation of small molecules discussed above.…”

Reaction mechanisms at the homogeneous–heterogeneous frontier: insights from first-principles studies on ligand-decorated metal nanoparticles

“…197,198 Chemical descriptors, which typically describe catalytic activity and selectivity, are usually employed as input for algorithms. On the homogeneous side, machine learning protocols have been recently employed to uncover metal-oxo intermediates 199 as well as to screen transition metal catalysts for cross-coupling reactions. 200 On the heterogeneous side, García-Muelas and López demonstrated how principal component analyses and regressions provide adsorption energies on metal surfaces from covalent (d-band center) and ionic (reduction potential) descriptors.…”