Structure-Activity Correlations for the Oxidation of CO over Polycrystalline RuO2 Powder Derived from Steady-State and Transient Kinetic Experiments

Accelerating the prediction of inorganic surfaces with machine learning interatomic potentials

The surface properties of solid-state materials often dictate their functionality, especially for applications where nanoscale effects become important. The relevant surface(s) and their properties are determined, in large part, by the material's synthesis or operating conditions. These conditions dictate thermodynamic driving forces and kinetic rates responsible for yielding the observed surface structure and morphology. Computational surface science methods have long been applied to connect thermochemical conditions to surface phase stability, particularly in the heterogeneous catalysis and thin film growth communities. This review provides a brief introduction to first-principles approaches to compute surface phase diagrams before introducing emerging data-driven approaches. The remainder of the review focuses on the application of machine learning, predominantly in the form of learned interatomic potentials, to study complex surfaces. As machine learning algorithms and large datasets on which to train them become more commonplace in materials science, computational methods are poised to become even more predictive and powerful for modeling the complexities of inorganic surfaces at the nanoscale.

SQERT-T: alleviating kinetic Monte Carlo (KMC)-stiffness in transient KMC simulations

Lattice based kinetic Monte Carlo (KMC) is often used for simulating the dynamics of systems at a supramolecular scale, based on molecular scale transitions. A common challenge in KMC simulations is rapid 'back-and-forth' reactions, which dominate the events executed during simulations and inhibit the ability for simulations to reach longer time scales. Such processes are fast frivolous processes (FFPs) and are one manifestation of a phenomenon referred to as KMC-stiffness. Here, an algorithm for staggered quasi-equilibrium rank-based throttling geared towards transient kinetics (SQERT-T) is presented. Within the SQERT-T methodology, a pace-restrictor reaction and an FFP floor are utilized along with throttling of the process transition rate constants to accelerate the KMC simulations while still retaining sufficient time resolution for sampling of the data. KMC simulations were performed for CO oxidation over RuO₂(1 1 0) and over RuO₂(1 1 1), and the results were compared to experimental data obtained using RuO₂ powders. The experiments and simulations were for transient conditions: the system was subjected to a temperature program which included temperatures in the range of 363 to 453 K. The timescales that were achieved during the KMC simulations in this study would not have been accessible without KMC acceleration, and were enabled by the use of SQERT-T.

Oscillatory Behavior in the CO-oxidation over Bulk Ruthenium Dioxide — the Effect of the CO/O2 Ratio

CO oxidation over polycrystalline ruthenium dioxide was monitored in an in-situ XRD setup. The evolution of the bulk state of the catalyst was followed by in-situ XRD during reaction, while the surface morphology and chemical state before and after reaction were investigated by HRSEM and EDX. The commercial RuO2 powder was calcined prior reaction to ensure the formation of completely oxidized RuO2. This pre-calcined RuO2 is initially inactive in CO oxidation regardless of the CO/O2 feed ratio and requires an induction period, the length of which strongly depends whether the catalyst is diluted with boron nitride or not. After this induction period oscillations in the CO2 yield occur under O2-rich conditions only. These oscillations exhibit two time constants for the diluted catalyst, while the low frequency oscillations were not observed in the case of undiluted RuO2. Furthermore, the state of the catalyst after activation in O2-rich feed conditions differs dramatically from the state after activation in CO-rich feed conditions. Firstly, the catalyst activated in an O2-rich atmosphere remains inactive under CO-rich conditions in contrast to the catalyst activated in CO-rich conditions which is afterwards active under all feed ratios examined. Secondly, the surface morphology of the catalyst is quite different. While the apical surfaces of the RuO2 crystals become roughened upon activation in the CO-rich feed, they become facetted under O2 rich activation conditions. Therefore, we conclude that at least two different active surface states on the bulk RuO2 catalyst exist.

On the CO-Oxidation over Oxygenated Ruthenium

The oxidation of carbon monoxide over polycrystalline ruthenium dioxide (RuO2) powder was studied in a packed-bed reactor and by bulk and surface analytical methods. Activity data were correlated with bulk phases in an in-situ X-ray diffraction (XRD) setup at atmospheric pressure. Ruthenium dioxide was pre-calcined in pure oxygen at 1073 K. At this stage RuO2 is completely inactive in the oxidation of CO. After a long induction period in the feed at 503 K RuO2 becomes active with 100% conversion, while in-situ XRD reveals no changes in the RuO2 diffraction pattern. At this stage selective roughening of apical RuO2 facets was observed by scanning electron microscopy (SEM). Seldom also single lateral facets are roughened. EDX indicated higher oxygen content in the following order: flat lateral facets > rough lateral facets > rough apical facets. Further, experiments in the packed bed reactor indicated oscillations in the CO2 formation rate. At even higher temperatures in reducing feed (533–543 K) the sample reduces to ruthenium metal according to XRD. The reduced particles exhibiting lower ignition temperature are very rough with cracks and deep star-shaped holes. An Arrhenius plot of the CO2 formation rate below the ignition temperature reveals the reduced samples to be significantly more active based on mass unit and shows lower apparent activation energy than the activated oxidized sample. Micro-spot X-ray photoelectron spectroscopy (XPS) and XPS microscopy experiments were carried out on a Ru(0001) single crystal exposed to oxygen at different temperature. Although low energy electron diffraction (LEED) images show a strong 1×1 pattern, the XPS data indicated a wide lateral inhomogeneity with different degree of oxygen dissolved in the subsurface layers. All these and the literature data are discussed in the context of different active states and transport issues, and the metastable nature of a phase mixture under conditions of high catalytic activity.

Does phenomenological kinetics provide an adequate description of heterogeneous catalytic reactions?

Phenomenological kinetics (PK) is widely used in the study of the reaction rates in heterogeneous catalysis, and it is an important aid in reactor design. PK makes simplifying assumptions: It neglects the role of fluctuations, assumes that there is no correlation between the locations of the reactants on the surface, and considers the reacting mixture to be an ideal solution. In this article we test to what extent these assumptions damage the theory. In practice the PK rate equations are used by adjusting the rate constants to fit the results of the experiments. However, there are numerous examples where a mechanism fitted the data and was shown later to be erroneous or where two mutually exclusive mechanisms fitted well the same set of data. Because of this, we compare the PK equations to "computer experiments" that use kinetic Monte Carlo (kMC) simulations. Unlike in real experiments, in kMC the structure of the surface, the reaction mechanism, and the rate constants are known. Therefore, any discrepancy between PK and kMC must be attributed to an intrinsic failure of PK. We find that the results obtained by solving the PK equations and those obtained from kMC, while using the same rate constants and the same reactions, do not agree. Moreover, when we vary the rate constants in the PK model to fit the turnover frequencies produced by kMC, we find that the fit is not adequate and that the rate constants that give the best fit are very different from the rate constants used in kMC. The discrepancy between PK and kMC for the model of CO oxidation used here is surprising since the kMC model contains no lateral interactions that would make the coverage of the reactants spatially inhomogeneous. Nevertheless, such inhomogeneities are created by the interplay between the rate of adsorption, of desorption, and of vacancy creation by the chemical reactions.

Monitoring in situ catalytically active states of Ru catalysts for different methanol oxidation pathways

One of the prerequisites for the detailed understanding of heterogeneous catalysis is the identification of the dynamic response of the catalyst surface under variable reaction conditions. The present study of methanol oxidation on different model Ru pre-catalysts, performed approaching the realistic catalytic reaction conditions, provides direct evidence of the significant effect of reactants' chemical potentials and temperature on the catalyst surface composition and the corresponding catalytic activity and selectivity. The experiments were carried out for three regimes of oxygen potentials in the 10(-1) mbar pressure range, combining in situ analysis of the catalyst surface by synchrotron-based photoelectron core level spectroscopy with simultaneous monitoring of the products released in the gas phase by mass spectroscopy. Metallic Ru with adsorbed oxygen and transient 'surface oxide', RuO(x), with varying x have been identified as the catalytically active states under specific reaction conditions, favouring partial or full oxidation pathways. It has been shown that the composition of catalytically active steady states, exhibiting different activity and selectivity, evolves under the reaction conditions, independent of the crystallographic orientation and the initial pre-catalyst chemical state, metallic Ru or RuO(2).