Predicting Catalytic Activity of High-Entropy Alloys for Electrocatalysis

Recently in Joule, Batchelor et al. developed a theoretical method for predicting catalytic activity of high-entropy alloys for oxygen reduction electrocatalysis. This versatile strategy could pave a new path toward designing metal-based catalysts for diverse energy applications. Recently in Joule, Batchelor et al. developed a theoretical method for predicting catalytic activity of high-entropy alloys for oxygen reduction electrocatalysis. This versatile strategy could pave a new path toward designing metal-based catalysts for diverse energy applications. Tailoring surface reactivity of nanomaterials is important for optimizing energy efficiencies while reducing environmental footprints of various industrial processes and renewable energy technologies. Recent developments of quantum-chemical methods, e.g., density functional theory (DFT), have shown promise for understanding kinetics of catalytic reactions and predicting active catalysts from first principles.1Nørskov J.K. Abild-Pedersen F. Studt F. Bligaard T. Density functional theory in surface chemistry and catalysis.Proc. Natl. Acad. Sci. USA. 2011; 108: 937-943Crossref PubMed Scopus (1242) Google Scholar Because of scaling relations between adsorption energies of reacting species, including transition states, the binding energies of simple adsorbates can often serve as descriptor(s) for catalytic activity. According to the well-established Sabatier principle in catalysis, the adsorption energies of key reaction intermediates should be neither too strong nor too weak, and optimal binding strengths will result in maximal activity, i.e., a volcano-shape dependence of catalytic activity with reactivity descriptor(s). With that as a guide, the reactivity descriptor(s) can then be optimized toward the top of the Sabatier volcano through tailoring the structure and composition of active sites. Historically, bimetallic and trimetallic alloys have been widely explored for maximizing catalytic performance of a given active metal. In this regard, high-entropy alloys (HEAs), which could be defined as multicomponent alloys with a well-ordered crystal structure and randomly distributed constituent elements,2Miracle D.B. Senkov O.N. A critical review of high entropy alloys and related concepts.Acta Mater. 2017; 122: 448-511Crossref Scopus (3896) Google Scholar show great potential for further tuning adsorption properties of metal surfaces. The reason that HEAs attract attention in catalysis is their inherent surface complexity, which can potentially provide a near-continuum distribution of adsorption energies at surface sites and maximize the fraction of optimal active sites. However, it is difficult to search for HEAs with desired properties through experimental trial-and-error and/or high-throughput computation. To accelerate the exploration of the immense catalyst space, there have been great efforts in developing predictive models relating the structure of a metal site to its local chemical reactivity. One well-known model is the d-band model developed by Hammer and Nørskov.3Hammer B. Nørskov J.K. Electronic factors determining the reactivity of metal surfaces.Surf. Sci. 1995; 343: 211-220Crossref Scopus (1821) Google Scholar Within this theoretical framework, electronic descriptors evaluated from self-consistent DFT calculations, e.g., the d-band center, can capture the complex interplay of geometric strain and metal-ligand effects at an active site and prove to be useful for catalyst design. Rooted on the d-band model, the purely geometry-based descriptors, e.g., the generalized and orbitalwise coordination numbers, were developed for facilitating catalyst screening.4Calle-Vallejo F. Martínez J.I. García-Lastra J.M. Sautet P. Loffreda D. Fast prediction of adsorption properties for platinum nanocatalysts with generalized coordination numbers.Angew. Chem. Int. Ed. 2014; 53: 8316-8319Crossref PubMed Scopus (301) Google Scholar, 5Ma X. Xin H. Orbitalwise coordination number for predicting adsorption properties of metal nanocatalysts.Phys. Rev. Lett. 2017; 118: 036101Crossref PubMed Scopus (105) Google Scholar Furthermore, recent initiative in promoting data science in catalysis6Tran K. Ulissi Z.W. Active learning across intermetallics to guide discovery of electrocatalysts for CO2 reduction and H2 evolution.Nat. Catalysis. 2018; 1: 696-703Crossref Scopus (355) Google Scholar, 7Ma X. Li Z. Achenie L.E.K. Xin H. Machine-learning-augmented chemisorption model for CO2 electroreduction catalyst screening.J. Phys. Chem. Lett. 2015; 6: 3528-3533Crossref PubMed Scopus (247) Google Scholar has demonstrated great success by relating geometric and electronic structure descriptors to reactivity properties with learning algorithms, such as artificial neural networks. But challenges exist for HEAs because of the intrinsic complexity and infinite potential combinations in this materials space. Recently in Joule, Batchelor et al. tackled this challenge by developing predictive reactivity models of high-entropy alloys by using local coordination and composition features of surface sites.8Batchelor T.A.A. Pedersen J.K. Winther S.H. Castelli I.E. Jacobsen K.W. Rossmeisl J. High-entropy alloys as a discovery platform for electrocatalysis.Joule. 2019; https://doi.org/10.1016/j.joule.2018.12.015Scopus (268) Google Scholar They used the oxygen reduction reaction (ORR) as a case study because of its sluggish electrochemical kinetics in limiting the performance of low-temperature polymer electrolyte membrane (PEM) fuel cells. The state-of-the-art elemental metal electrocatalyst, consisting of 2∼5 nm platinum (Pt) nanoparticles, sacrifices ∼400 mV for appreciable current densities.9Marković N.M. Schmidt T.J. Stamenković V. Ross P.N. Oxygen reduction reaction on Pt and Pt bimetallic surfaces: a selective review.Fuel Cells (Weinh.). 2001; 1: 105-116Crossref Google Scholar Therefore, there has been a great deal of interest in searching for more efficient oxygen reduction electrocatalysts, preferably with a reduced amount of precious metals. It has been demonstrated that the adsorption energies of oxygen-containing species, e.g., *OH, at an active site are predictive ORR reactivity descriptors.10Norskov J.K. Rossmeisl J. Logadottir A. Lindqvist L. Kitchin J.R. Bligaard T. Jonsson H. Origin of the overpotential for oxygen reduction at a fuel-cell cathode.J. Phys. Chem. B. 2004; 108: 17886-17892Crossref Scopus (6594) Google Scholar The activity volcano plot suggests that Pt sites with optimal catalytic performance should bind the *OH intermediate slightly more weakly (∼0.1 eV) than Pt(111). This has been used as a guiding principle for finding more efficient electrocatalysts for oxygen reduction since its discovery. Building on this catalyst design framework, the goal of the current study is to develop predictive models for estimating the binding strength of *OH on HEAs relative to Pt(111). The basic idea of site representation is to divide the local chemical environment of a surface atom or its ensembles within a certain cutoff radius into different neighboring shells, including the site itself (Figure 1A ). The number of atoms for each particular element and neighboring shell is used as a factor (i.e., feature) contributing to the local chemical reactivity of an active site. The length of the feature vector is determined by the number of elements used in alloys and the number of neighboring shells considered within the cutoff. With this representation of an active site, Batchelor et al. developed a linear regression model to link the feature vector to the adsorption energy of *OH at an atop site with the ordinary-least-squares algorithm. They used DFT calculations of *OH adsorption energies on surface sites of randomly picked IrPdPtRhRu HEAs to train and test the model. Developed with only ∼800 data points, the model showed a root-mean-square error of ∼0.06 eV for *OH adsorption energies on surface sites not used in the training stage (Figure 1B), demonstrating the robustness of the model.8Batchelor T.A.A. Pedersen J.K. Winther S.H. Castelli I.E. Jacobsen K.W. Rossmeisl J. High-entropy alloys as a discovery platform for electrocatalysis.Joule. 2019; https://doi.org/10.1016/j.joule.2018.12.015Scopus (268) Google Scholar Aiming to find candidate materials with maximized catalytic activity for oxygen reduction, the authors then applied the regression model to find the composition of HEAs with the sequential least-squares programming optimization method. They used the Sabatier volcano plot to estimate the catalytic activity of a surface site for a given binding energy of *OH relative to Pt(111). Because Pt(111) was used as the reference, the co-adsorbates and water layers present under working conditions had little influence on the activity enhancement. Random initial conditions were applied for the algorithm to find the atomic fractions that maximize the activity considering the large number of possible arrangements of surface sites. Optimization of the HEA composition led to the catalytic system Ir10.2Pd32.0Pt9.30Rh19.6Ru28.9 (Figure 1C) with a lower overpotential (∼40 mV) than that of Pt(111). This composition will form a stable HEA according to the stability rules involving the ratio of formation enthalpy to entropy and the difference in atomic radii of constituent metal elements. Interestingly, if the constraint on the number of elements in optimization is relaxed, i.e., allows for the exploration of conventional alloys, the algorithm predicts a Pt-Ir binary alloy with a globally optimized activity and experimentally validated stability (Figure 1D). Thus, the HEAs can be used as an unbiased discovery platform for electrocatalysis. In summary, Batchelor et al. have developed a theoretical approach for predicting catalytic activity of HEAs by learning from data and used the model to optimize IrPdPtRhRu HEAs for catalyzing oxygen reduction. The HEAs and binary IrPt alloy identified in this study show significant activity enhancement over Pt(111), although the cost is still high. Incorporating materials cost and stability into such a design approach might prove to be fruitful for developing cost-effective, robust, and highly active high-entropy alloys for various electrocatalytic applications. High-Entropy Alloys as a Discovery Platform for ElectrocatalysisBatchelor et al.JouleJanuary 18, 2019In BriefA theoretical method for finding active alloy electrocatalysts is proposed, and the method is applied to the electrochemical half-cell reaction of reducing oxygen to water, which is vital for improving the efficiency of, for example, hydrogen fuel cells. Our method predicts adsorption energies between reaction intermediates and the alloy surface to discover which sites on the surface are the most active. Starting from the multicomponent alloy IrPdPtRhRu, the alloy composition with best predicted catalytic activity is found. Full-Text PDF Open Archive