Machine learning approach for screening alloy surfaces for stability in catalytic reaction conditions

Abstract

A catalytic surface should be stable under reaction conditions to be effective. However, it takes significant effort to calculate the stability of a surface, as this requires intensive quantum chemical calculations. To more efficiently estimate stability, we provide a general and data-efficient machine learning (ML) approach to accurately and efficiently predict the surface energies of metal alloy surfaces. Our ML approach introduces an element-centered fingerprint (ECFP) which was used as a vector representation for fitting models for predicting surface formation energies. The ECFP is significantly more accurate than several existing feature sets when applied to dilute alloy surfaces and is competitive with existing feature sets when applied to bulk alloy surfaces or gas-phase molecules. Models using the ECFP as input can be quite general, as we created models with good accuracy over a broad set of bimetallic surfaces including most d-block metals, even with relatively small datasets. For example, using the ECFP, we developed a kernel ridge regression machine-learning model which is able to predict the surface energies of alloys of diverse metal combinations with a mean absolute error of 0.017 eV/atom. Combining this model with an existing model for predicting adsorption energies, we estimated segregation trends of 596 single-atom alloys with and without CO adsorbed on these surfaces. As a simple test of the approach, we identify specific cases where CO does not induce segregation in these single-atom alloys.

Bayesian optimization of single-atom alloys and other bimetallics: efficient screening for alkane transformations, CO2 reduction, and hydrogen evolution

“…14 Other studies show ML being used to predict SAA properties such as stability and segregation energies. 15–18 Overall, these studies demonstrate the utility of ML in the rapid screening of SAAs for different applications. However, for SAAs (or other SACs), it would be desirable to have a ML method that efficiently screens smaller design spaces, in contrast to most ML methods that require a relatively large dataset.…”

Theoretical Calculation Assisted by Machine Learning Accelerate Optimal Electrocatalyst Finding for Hydrogen Evolution Reaction

“…Montemore and co‐workers systematically designed TM atoms doping on a series of metal surfaces to uncover the stability of the doped systems. By implementing the KRR model, the obtained mean square errors were lower than 0.02 eV for all the systems, further ensuring the high accuracy of the KRR in the catalytic field [53] . Apart from those, the supported Vector machine (SVM) is a classical machine learning model for data classification and is popularly used in the electrocatalytic field [54] .…”

“…By implementing the KRR model, the obtained mean square errors were lower than 0.02 eV for all the systems, further ensuring the high accuracy of the KRR in the catalytic field. [53] Apart from those, the supported Vector machine (SVM) is a classical machine learning model for data classification and is popularly used in the electrocatalytic field. [54] SVM is a binary classification algorithm that could spontaneously find a hyperplane to classify the training data and make new judgments to split the new data based on the trained data, thus ensuring low errors between the distinct data.…”

Single Atom Alloys Segregation in the Presence of Ligands

“…Han et al developed a model that predicted E seg in the presence of H through compressed-sensing data-analytics approach (SISSO), utilizing multiple DFT inputs. 16 More recently, Sulley et al applied machine learning techniques to determine the stability of single atom alloys in the absence and presence of CO. 24 Although these models were able to screen through different SAAs in the presence of H and CO, an understanding of how different adsorbates, specifically ligands, affect E seg has yet to be unraveled.…”