Scalable approach to high coverages on oxides via iterative training of a machine‐learning algorithm

Abstract

Understanding the interaction of multiple types of adsorbate molecules on solid surfaces is crucial to establishing the stability of catalysts under various chemical environments. Computational studies on the high coverage and mixed coverages of reaction intermediates are still challenging, especially for transition‐metal compounds. In this work, we present a framework to predict differential adsorption energies and identify low‐energy structures under high‐ and mixed‐adsorbate coverages on oxide materials. The approach uses Gaussian process machine‐learning models with quantified uncertainty in conjunction with an iterative training algorithm to actively identify the training set. The framework is demonstrated for the mixed adsorption of CHx, NHx and OHx species on the oxygen vacancy and pristine rutile TiO2(110) surface sites. The results indicate that the proposed algorithm is highly efficient at identifying the most valuable training data, and is able to predict differential adsorption energies with a mean absolute error of ∼0.3 eV based on <25 % of the total DFT data. The algorithm is also used to identify 76 % of the low‐energy structures based on <30 % of the total DFT data, enabling construction of surface phase diagrams that account for high and mixed coverage as a function of the chemical potential of C, H, O, and N. Furthermore, the computational scaling indicates the algorithm scales nearly linearly (N1.12) as the number of adsorbates increases. This framework can be directly extended to metals, metal oxides, and other materials, providing a practical route toward the investigation of the behavior of catalysts under high‐coverage conditions.

Machine Learning for Electrocatalyst and Photocatalyst Design and Discovery

“…For example, electronic and atomic descriptors are not able to differentiate between allotropes, isomers, conformers, and polymorphs that have very similar descriptors for each form. Techniques such as radial distribution functions (density distributions as a function of distance) and Voronoi tessellations (partition the crystal lattice into subregions close to each of a given set of objects) are useful in encoding the crystal lattice. , The Coulomb matrix is a global descriptor mimicking the electrostatic interaction between the nuclei and is widely used in structure featurization, especially in adsorbate research . Structures of molecules can be described by the simplified molecular-input line-entry system (SMILES), a string of characters composed of letters and symbols that can be converted into numbers using the ASCII values of the characters. , More refined implementations have been developed based on SMILES.…”

“…As mentioned, because PCA retains all original descriptors, models trained on principal components lose performance when many of the original descriptors are uninformative (noise). PCA is commonly employed in modeling of materials properties where a large number of features have been generated, such as heterogeneous catalysts, photovoltaics, and supramolecular materials. ,,, Given that PCA only builds linear projections but structure–property relationships are often nonlinear, additional learning algorithms have been developed to perform nonlinear dimension reduction. Kernel PCA uses standard PCA to perform nonlinear dimensional reduction by using a nonlinear transformation to project the data points into a higher dimension feature space where standard PCA can be used .…”

Electrocatalytic Hydrogenation of Biomass-Derived Organics: A Review

“…We would be remiss not to briefly discuss the emerging field of informatics for catalysis, which is growing rapidly because of its potential to revolutionize the discovery and design of catalysts. Recently, ChemCatChem published a series of papers as a special collection highlighting the use of data science in catalysis. − Although computational chemists have long used informatics approaches, , the confluence of high-throughput synthesis, characterization, and testing of catalysts, and the proliferation of computational capabilities have driven the development of machine learning and data science tools . These tools are enabling the analysis of both experimental and computed data in an effort to discover hidden relationships and connect chemical properties with catalytic activity .…”

A Deep Dive into Machine Learning Density Functional Theory for Materials Science and Chemistry

“…In the context of computational materials science and chemistry this usually translates to identifying catalysts for reactions of interest (for instance CO 2 reduction) and their respective activity and selectivity. The computational demand of such studies can be drastically decreased by employing ML, and a wide range of such investigations has been performed [155], [156], [157], [158], [159], [160], [161], [162], [163], [164], [165], [166], [167], [168], [169], [170], [171], [172], [173], [174], [175], [176], [177], [178], [179], [180], [181], [182], [183], [184], [185], [186], [187], [188], [189], [190], [191], [192], [193], [194], [195], [196], [197], [198], [199], [200], …”