High-throughput screening and literature data-driven machine learning-assisted investigation of multi-component La2O3-based catalysts for the oxidative coupling of methane

Abstract

Herein, multi-component La₂O₃-based catalysts for the oxidative coupling of methane (OCM) were designed based on high-throughput screening (HTS) and literature datasets with multi-output machine learning (ML) approaches including random forest regression (RFR), support vector regression (SVR), Gaussian process regression (Bayesian), and itemset mining (LCM). The combined use of HTS data and SVR successively assisted the search for 11 types of multi-component La₂O₃-based OCM catalysts in 20 validations with C₂ yields appearing at 450 °C based on indirect ML assistance. The appropriate multi-component predicted from ML contributed to the determination of a characteristic feature of the lower onset temperature for an La₂O₃-based OCM catalyst. The LCM application on the SVR extended HTS data area supported the observation of the effective elements in the HTS area. However, a challenging subject remains, i.e., 2 types of multi-component La₂O₃-based catalysts afforded an effective C₂ yield (>5.0%) at 450 °C, as inferred from the 20 selected types of catalyst validation. Thus, to predict unique multi-component La₂O₃-based OCM catalysts further, a combination of HTS and literature data was applied for four ML approaches. This was helpful to discover 17 additional combinations of multi-component La₂O₃-based catalysts affording effective C₂ yields (>5.0%) at 450 °C in the 38 selected types of predictions. In total, 30 new types of multi-component La₂O₃-based catalysts with a C₂ yield greater than 5.0% at 450 °C in CH₄/O₂ = 2.0 condition were found based on the indirect ML assistance driven by HTS and literature data.