Machine Learning Design of Perovskite Catalytic Properties

Abstract

Discovering new materials that efficiently catalyze the oxygen reduction and evolution reactions is critical for facilitating the widespread adoption of solid oxide fuel cell and electrolyzer (SOFC/SOEC) technologies. Here, machine learning (ML) models are developed to predict perovskite catalytic properties critical for SOFC/SOEC applications, including oxygen surface exchange, oxygen diffusivity, and area specific resistance (ASR). The models are based on trivial‐to‐calculate elemental features and are more accurate and dramatically faster than the best models based on ab initio‐derived features, potentially eliminating the need for ab initio calculations in descriptor‐based screening. The model of ASR enables temperature‐dependent predictions, has well calibrated uncertainty estimates and online accessibility. Use of temporal cross‐validation reveals the model to be effective at discovering new promising materials prior to their initial discovery, demonstrating the model can make meaningful predictions. Using the SHapley Additive ExPlanations (SHAP) approach, detailed discussion of different approaches of model featurization is provided for ML property prediction. Finally, the model is used to screen more than 19 million perovskites to develop a list of promising cheap, earth‐abundant, stable, and high performing materials, and find some top materials contain mixtures of less‐explored elements (e.g., K, Bi, Y, Ni, Cu) worth exploring in more detail.

Redefining the Stability of Water Oxidation Electrocatalysts: Insights from Materials Databases and Machine Learning

“…Thus, this approach should be viewed merely as a preliminary exercise for TM X-ides. While these ML approaches are robust and have been extensively utilized to derive significant insights into the catalytic performance of various materials, ,− researchers must exercise caution when making claims about the relevance and impact of specific descriptors. Future investigations must prioritize the use of more reliable data sets, as the quality of data fed into ML models critically determines their output’s accuracy. Simply put, the efficacy of ML outcomes is directly proportional to the quality of the input data.…”

Advancements and prospects of perovskite-based fuel electrodes in solid oxide cells for CO2 electrolysis to CO

“…100 More specifically for SOFCs/SOECs, appropriate descriptors such as ASR values and ionic Lewis acid descriptors can further enhance the performance of ML models. 101,105 Choosing a suitable ML algorithm is an essential task. The ideal model should aim to maximize computational accuracy while minimizing resource consumption.…”

“…To create the database, Jacobs et al gathered 749 data points from 313 studies that included 299 distinct perovskite compositions. 101 Considering that biases arising from various experimental conditions during data collection can directly impact ML outcomes, it is imperative to conduct a comprehensive analysis of the underlying statistics for each catalytic property in the database. This analysis aims to mitigate or eliminate these biases, thereby ensuring the reliability of ML outcomes.…”