The performance of functional materials is dictated by chemical and structural properties of individual atomic sites. In catalysts, for example, the thermodynamic stability of constituting atomic sites is a key descriptor from which more complex properties, such as molecular adsorption energies and reaction rates, can be derived. In this study, we present a widely applicable machine learning (ML) approach to instantaneously compute the stability of individual atomic sites in structurally and electronically complex nano-materials. Conventionally, we determine such site stabilities using computationally intensive first-principles calculations. With our approach, we predict the stability of atomic sites in sub-nanometer metal clusters of 3-55 atoms with mean absolute errors in the range of 0.11-0.14 eV. To extract physical insights from the ML model, we introduce a genetic algorithm (GA) for feature selection. This algorithm distills the key structural and chemical properties governing the stability of atomic sites in size-selected nanoparticles, allowing for physical interpretability of the models and revealing structure-property relationships. The results of the GA are generally model and materials specific. In the limit of large nanoparticles, the GA identifies features consistent with physics-based models for metal-metal interactions. By combining the ML model with the physics-based model, we predict atomic site stabilities in real time for structures ranging from sub-nanometer metal clusters (3-55 atom) to larger nanoparticles (147 to 309 atoms) to extended surfaces using a physically interpretable framework. Finally, we present a proof of principle showcasing how our approach can determine stable and active nanocatalysts across a generic materials space of structure and composition.

Accurate theoretical simulation of electrochemical activation barriers is key to understanding electrocatalysis and guides the design of more efficient catalysts. Providing a detailed picture of proton transfer processes encounters several challenges: the constant potential requirement during charge transfer, the different time scales involved in the processes, and the thermal fluctuation of the solvent. Hence, it is prohibitively expensive computationally to apply density functional theory (DFT) calculations in modeling the potential-dependent activation barrier at the electrode-solvent interface, and the results are dubious. To address these challenges, we have developed an analytical approach based on charge conservation and decoupled potential energy surfaces to compute charge transfer barriers. The method makes it possible to simulate an electrochemical process at different potentials and explicitly include thermal fluctuations of the solvent at the electrode-solvent interface. We use the Pt-catalyzed alkaline hydrogen evolution reaction (HER) as our benchmark reaction, and we model the microkinetics of HER with consideration of the spatial fluctuations between the metal surface and the first solvent layer at room temperature. The distribution of water-metal distances has a large effect on the barriers of the charge transfer processes, and an accurate account of the statistical fluctuation in the reaction network leads to a several orders of magnitude increase in HER current as compared to transfer from a static solvent. The trends of the different reaction mechanisms in HER were successfully simulated with our model, and the theoretical I-V curves obtained are in good qualitative agreement with experimental results.

Understanding the mechanism behind the superior catalytic power of single- or few-atom heterogeneous catalysts has become an important topic in surface chemistry. This is particularly the case for gold, with TiO2 being an efficient support. Here we use scanning tunneling microscopy/spectroscopy with theoretical calculations to investigate the adsorption geometry and local electronic structure of several-atom Au clusters on rutile TiO2(110), with the clusters fabricated by controlled manipulation of single atoms. Our study confirms that Au1 and Au2 clusters prefer adsorption at surface O vacancies. Au3 clusters adsorb at O vacancies in a linear-chain configuration parallel to the surface; in the absence of O vacancies they adsorb at Ti5c sites with a structure of a vertically pointing upright triangle. We find that both the electronic structure and cluster-substrate charge transfer depend critically on the cluster size, bonding configuration, and local environment. This suggests the possibility of engineering cluster selectivity for specific catalytic reactions.

Electrochemical reduction of carbon-dioxide/carbon-monoxide (CO(2)R) to fuels and chemicals presents an attractive approach for sustainable chemical synthesis, but also poses a serious challenge in catalysis. Understanding the key aspects that guide CO(2)R towards value-added multicarbon (C2+) products is imperative in designing an efficient catalyst. Herein, we identify the critical steps toward C2 products on copper through a combination of energetics from density functional theory and micro-kinetic modeling. We elucidate the importance of atomic carbon in directing C2+ selectivity and how it introduces surface structural sensitivity on copper catalysts. This insight enables us to propose two simple thermodynamic descriptors that effectively describe C2+ selectivity on metal catalysts beyond copper and hence it identifies an intelligible protocol to screen for materials that selectively catalyze CO(2) to C2+ products.

This corrects the article DOI: 10.1103/PhysRevLett.120.226101.

Electrochemical reduction of carbon dioxide (CO2) over transition metals follows a complex reaction network. Even for products with a single carbon atom (C1 products), two bifurcated pathways exist: initially between carboxyl (COOH*) and formate (HCOO*) intermediates and the COOH* intermediate is further bifurcated by pathways involving either formyl (CHO*) or COH*. In this study, we combine evidence from the experimental literature with a theoretical analysis of energetics to rationalize that not all steps in the reduction of CO2 are electrochemical. This insight enables us to create a selectivity map for two-electron products (carbon monoxide (CO) and formate) on elemental metal surfaces using only the CO and OH binding energies as descriptors. In the further reduction of CO*, we find that CHO* is formed through a chemical step only whereas COH* follows from an electrochemical step. Notably on Cu(100), the COH pathway becomes dominant at an applied potential lower than −0.5V vs. RHE. For the elemental metals selective towards CO formation, the variation of the CO binding energy is sufficient to further subdivide the map into domains that predominantly form H2, CO, and ultimately more reduced products. We find Cu to be the only elemental metal capable of reducing CO2 to products beyond 2e− via the proposed COH pathway and we identify atomic carbon as the key component leading to the production of methane. Our analysis also rationalizes experimentally observed differences in products between thermal and electrochemical reduction of CO2 on Cu.

Single atomic sites often determine the functionality and performance of materials, such as catalysts, semi-conductors or enzymes. Computing and understanding the properties of such sites is therefore a crucial component of the rational materials design process. Because of complex electronic effects at the atomic level, atomic site properties are conventionally derived from computationally expensive first-principle calculations, as this level of theory is required to achieve relevant accuracy. In this study, we present a widely applicable machine learning (ML) approach to compute atomic site properties with high accuracy in real time. The approach works well for complex non-crystalline atomic structures and therefore opens up the possibility for high-throughput screenings of nano-materials, amorphous systems and materials interfaces. Our approach includes a robust featurization scheme to transform atomic structures into features which can be used by common machine learning models. Performing a genetic algorithm (GA) based feature selection, we show how to establish an intuitive physical interpretation of the structure-property relations implied by the ML models. With this approach, we compute atomic site stabilities of metal nanoparticles ranging from 3-55 atoms with mean absolute errors in the range of 0.11-0.14 eV in real time. We also establish the chemical identity of the site as most important factor in determining atomic site stabilities, followed by structural features like bond distances and angles. Both, the featurization and GA feature selection functionality are published in open-source python modules. With this method, we enable the efficient rational design of highly specialized real-world nano-catalysts through data-driven materials screening.

Nanoparticles have an immense importance in various fields, such as medicine, catalysis, and various technological applications. Nanoparticles exhibit a significant depression in melting point as their size goes below ≈10 nm. However, nanoparticles are frequently used in high temperature applications such as catalysis where temperatures often exceed several 100 degrees which makes it interesting to study not only the melting temperature depression, but also how the melting progresses through the particle. Using high‐resolution transmission electron microscopy, the melting process of gold nanoparticles in the size range of 2–20 nm Au nanoparticles combined with molecular dynamics studies is investigated. A linear dependence of the melting temperature on the inverse particle size is confirmed; electron microscopy imaging reveals that the particles start melting at the surface and the liquid shell formed then rapidly expands to the particle core.

A long-standing question in electrocatalysis is how the electrolyte pH affects the catalytic activity of proton electron transfer reactions. A prime example for this is the hydrogen evolution/oxidation reaction (HER/HOR) over metal catalysts. While it has long been established that alkaline conditions result in a more sluggish reaction kinetics than in acidic conditions, the underlying reason for this trend remains contentious. We apply density functional calculations to evaluate prevailing hypotheses for the origin of this effect: shifts in hydrogen binding, proton donor, and water reorganization energy. We present a microkinetic model, based on ab initio reaction energetics of all possible elementary steps. Our model shows a good agreement with experimental trends. We find that with increasing pH, the proton donor changes from hydronium to water. Our model suggests that the intrinsically larger barriers for the splitting of water with respect to hydronium are the cause of HER kinetics being slower in alkaline than in acidic media.

Big data and artificial intelligence has revolutionized science in almost every field - from economics to physics. In the area of materials science and computational heterogeneous catalysis, this revolution has led to the development of scientific data repositories, as well as data mining and machine learning tools to investigate the vast materials space. The goal of using these tools is to establish a deeper understanding of the relations between materials properties and activity, selectivity and stability - the important figures of merit in catalysis. Based on these insights, catalyst design principles can be established, which hopefully lead us to discover highly efficient catalysts to solve pressing issues for a sustainable future and the synthesis of highly functional materials, chemicals and pharmaceuticals. The inherent complexity of catalytic reactions quests for machine learning methods to efficiently navigate through the high-dimensional hyper-surfaces in structure optimization problems to determine relevant chemical structures and transition states. In this review, we show how cutting edge data infrastructures and machine learning methods are being used to address problems in computational heterogeneous catalysis.

The detailed structure of silica and germania films supported on Ru(0001) metal substrates are compared to each other. Surface science techniques together with density functional theory calculations have been used to gain insights into the atomic arrangement of these prominent glass-forming materials. The monolayer films of these materials both show predominantly crystalline hexagonal lattices with characteristic domain boundary structures. For the germania monolayer films a large variety of ring elements within domain boundaries have been observed. Density functional calculations predict stronger interaction with the metal substrate for bilayer germania as compared to bilayer silica films. Scanning tunneling microscopy images with atomically resolved structural features have given access to silica and germania bilayer film structures. Both bilayer films form characteristic amorphous ring structures. However, the germania bilayer films appear to be more corrugated, pointing to a stronger interaction with the metal support thus giving rise to slightly different connectivity rules.

A new two-dimensional (2D) germanium dioxide film has been prepared. The film consists of interconnected germania tetrahedral units forming a bilayer structure, weakly coupled to the supporting Pt(111) metal-substrate. Density functional theory calculations predict a stable structure of 558-membered rings for germania films, while for silica films 6-membered rings are preferred. By varying the preparation conditions the degree of order in the germania films is tuned. Crystalline, intermediate ordered and purely amorphous film structures are resolved by analysing scanning tunnelling microscopy images.

First-row layered transition metal (oxy)(hydro)oxides (LTMOs) form an important class of earth-abundant materials. They are well-known as active alkaline oxygen evolution reaction (OER) catalysts,([1,5]) and are also often used as metal-ion battery anodes([6]) or as metal-air bifunctional electrodes.([7]) However, their electrochemical activities, particularly for the oxygen reduction reaction (ORR), across the whole 3d-element series remain largely unexplored. In this work, we perform a systematic screening of these catalysts for both OER and ORR using a surface edge-site model with exposed active sites for metal double hydroxides M(OH)(2), oxyhydroxides MOOH and oxides MO2. We establish OER and ORR activities and scaling relations of the whole series across +2, +3 and +4 oxidation states, and successfully reproduce the experimental activities of a few pure layered (oxy)(hydro)oxides. We predict CoOOH/CoO2 and NiOOH/NiO2 as active and stable OER catalysts. We also predict Fe(OH)(2)/FeOOH, Mn(OH)(2)/MnOOH and Co(OH)(2) as active and stable ORR catalysts. This makes Co-(oxy)(hydro)oxides only bifunctional catalyst in this series. Using linear regression, our results indicate that trends across the 3d-series can be obtained from only a few bulk, surface and atomic type descriptors. Particularly, we identify that the number of outer d-electrons at the surface-active site as the most important descriptor of activity.

Journal Article In situ Probing of Nanostructure Surfaces Get access Thomas W Hansen, Thomas W Hansen DTU Nanolab, Technical University of Denmark, Kgs. Lyngby, Denmark Corresponding author: thwh@dtu.dk Search for other works by this author on: Oxford Academic Google Scholar Monia R Nielsen, Monia R Nielsen DTU Nanolab, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Pei Liu, Pei Liu DTU Nanolab, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Philomena Schlexer, Philomena Schlexer Department of Chemical Engineering, Stanford University, Stanford, CA, USA Search for other works by this author on: Oxford Academic Google Scholar Jacob Madsen, Jacob Madsen DTU Physics, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Anton B Andersen, Anton B Andersen DTU Nanolab, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Jakob Schiøtz, Jakob Schiøtz DTU Physics, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Bela Sebok, Bela Sebok DTU Physics, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Ib Chorkendorff, Ib Chorkendorff DTU Physics, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Jakob B Wagner Jakob B Wagner DTU Nanolab, Technical University of Denmark, Kgs. Lyngby, Denmark Search for other works by this author on: Oxford Academic Google Scholar Microscopy and Microanalysis, Volume 25, Issue S2, 1 August 2019, Pages 2080–2081, https://doi.org/10.1017/S1431927619011139 Published: 01 August 2019

We present a microkinetic model for CO (2) reduction (CO (2) R) on Cu(211) towards C 2 products, based on energetics estimated from an explicit solvent model. We show that the differences in both Tafel slopes and pH dependence for C 1 vs C 2 activity arise from differences in their multi-step mechanisms. We find the depletion in C 2 products observed at high overpotential and high pH to arise from the 2 nd order dependence of C-C coupling on CO coverage, which decreases due to competition from the C 1 pathway. We further demonstrate that CO (2) reduction at a fixed pH yield similar activities, due to the facile kinetics for CO 2 reduction to CO on Cu, which suggests C 2 products to be favored for CO 2 R under alkaline conditions. The mechanistic insights of this work elucidate how reaction conditions can lead to significant enhancements in selectivity and activity towards higher value C 2 products.

B. Holst, C. Büchner, S. D. Eder, T. Nesse, D. Kuhness, P. Schlexer, G. Pacchioni, J. R. Manson, M. Heyde, H.-J.Freund Department of Physics and Technology, University of Bergen, Allégaten 55, 5007 Bergen Norway Fritz-Haber-Institut der Max-Planck-gesellschaft, Faradayweg 4-6, 14195 Berlin, Germany Department of Physics, NTNU, NO-7491, Trondheim, Norway Department of Material Science, Università de Milano-Biocca, Via R. Cozzi 55, Milan, Italy Department of Physics and Astronomy, Clemson University, South Carolina 29634, USA bodil.holst@uib.no

Silver (Ag) clusters confined in matrices possess remarkable luminescence properties, but little is known about their structural and electronic properties. We characterized the bright green luminescence of Ag clusters confined in partially exchanged Ag-Linde Type A (LTA) zeolites by means of a combination of x-ray excited optical luminescence-extended x-ray absorption fine structure, time-dependent-density functional theory calculations, and time-resolved spectroscopy. A mixture of tetrahedral Ag4(H2O) x2+ (x = 2 and x = 4) clusters occupies the center of a fraction of the sodalite cages. Their optical properties originate from a confined two-electron superatom quantum system with hybridized Ag and water O orbitals delocalized over the cluster. Upon excitation, one electron of the s-type highest occupied molecular orbital is promoted to the p-type lowest unoccupied molecular orbitals and relaxes through enhanced intersystem crossing into long-lived triplet states.