Computational catalyst discovery: Active classification through myopic multiscale sampling

Abstract

The recent boom in computational chemistry has enabled several projects aimed at discovering useful materials or catalysts. We acknowledge and address two recurring issues in the field of computational catalyst discovery. First, calculating macro-scale catalyst properties is not straightforward when using ensembles of atomic-scale calculations [e.g., density functional theory (DFT)]. We attempt to address this issue by creating a multi-scale model that estimates bulk catalyst activity using adsorption energy predictions from both DFT and machine learning models. The second issue is that many catalyst discovery efforts seek to optimize catalyst properties, but optimization is an inherently exploitative objective that is in tension with the explorative nature of early-stage discovery projects. In other words, why invest so much time finding a “best” catalyst when it is likely to fail for some other, unforeseen problem? We address this issue by relaxing the catalyst discovery goal into a classification problem: “What is the set of catalysts that is worth testing experimentally?” Here, we present a catalyst discovery method called myopic multiscale sampling, which combines multiscale modeling with automated selection of DFT calculations. It is an active classification strategy that seeks to classify catalysts as “worth investigating” or “not worth investigating” experimentally. Our results show an ∼7–16 times speedup in catalyst classification relative to random sampling. These results were based on offline simulations of our algorithm on two different datasets: a larger, synthesized dataset and a smaller, real dataset.

Insights and Activation Energy Surface of the Dehydrogenation of C2HxO Species in Ethanol Oxidation Reaction on Ir(100)

“…Therefore, the current studies on Ir(100) can be served as benchmark for comparison in the future studies on different facets, hydrogen bonding [88,89] the effect of the van der Waals interactions, [91] and for investigating alloy effects, which are important in developing better catalysts, [92–95] and comparison with other metals [96] . Finally, the current work will provide data set for kinetics modeling [68,81,97] and computational catalyst discovery [98,99] …”

“…[96] Finally, the current work will provide data set for kinetics modeling [68,81,97] and computational catalyst discovery. [98,99]…”

Data-driven design of electrocatalysts: principle, progress, and perspective

“…For instance, Tran et al developed an ML-based classification method combined with multiscale modeling of DFT computation. 89 By classifying the corresponding computational results, the catalyst candidates were divided into “worth investigating” or “not worth investigating” groups. This step can enable subsequent research to focus on more potential candidate materials, thus improving the catalyst development speed by 7–16 times compared to random experiments.…”

Generalizing Performance Equations in Heterogeneous Catalysis from Hybrid Data and Statistical Learning

“…Since large experimental and computational datasets are being made available through high-throughput techniques, the systematically generated data is amenable for statistical learning (SL) treatments. ,− Generally, the introduction of data approaches has been steered from the DFT community. Thus, approaches combining DFT and SL have been recently introduced in (i) the derivation of approaches to simplify the DFT computational burden of calculating the energies of many configurations; − (ii) applying these energies to traditionally MK­(DFT) derived volcanoes, providing candidates for electrochemical processes; and (iii) searching for optimized descriptors employed in theoretical studies. , In parallel, attempts to link experimental catalytic performance can be found in recent literature (for example, in the work of Foppa et al, ∼40 tabulated experimental observables were used to predict the annotated consistent conversion and selectivity of nine vanadium-based catalysts using a symbolic regression protocol (SISSO) to derive a nontrivial ensemble of models).…”