Active Learning Training Strategy for Predicting O Adsorption Free Energy on Perovskite Catalysts using Inexpensive Catalyst Features. Issue 10 (4th August 2021)

• Record Type: Journal Article
• Authors: Shambhawi, Shambhawi; Csányi, Gábor; Lapkin, Alexei A.
• Is Part Of: Chemistry methods. Volume 1:Issue 10(2021)
• Page Start: 444
• Page End: 450
• Publication Date: 2021-08-04
• Subjects: active learning -- adsorption energy prediction -- electrochemistry -- heterogeneous catalysis -- perovskites; Chemistry, Analytic -- Periodicals; 543
• DOI: 10.1002/cmtd.202100035
• Languages: English
• ISSNs: 2628-9725

Abstract

Abstract: Machine learning (ML) based energy prediction models are among the most effective descriptor‐based catalyst screening tools for heterogeneous reaction systems. However, their implementations are limited due to expensive data labelling, ab initio feature evaluation and lack of universal catalyst features, that is, beyond d‐band theory. Herein, we propose an inexpensive geometric feature for application on systems beyond d‐band theory, for example perovskites comprising of s‐, p‐, d‐ and f‐block elements. We outline a workflow that inputs these features into an active learning algorithm that enables effective data labelling, whilst improving prediction accuracies of existing models. We then use batch sampling to define termination criteria and to implement time‐series error forecasting for further reducing the number of expensive data labelling for training. We implement this workflow to train ML models for predicting oxygen adsorption free energy on perovskites and achieve similar, if not better, prediction accuracies as obtained from ab initio features. Abstract : The study reports geometric features that can be generalized to catalyst compositions beyond d‐band metals. By employing these features into an active learning workflow, required prediction accuracies are achieved without expensive an‐initio features. The workflow itself enables effective data labelling, but feeding it into a forecasting model further limits the number of expensive data labelling.