Search

Machine learning models with uncertainty quantification have recently emerged as attractive tools to accelerate the navigation of catalyst design spaces in a data-efficient manner. Here, we combine active learning with a dropout graph convolutional network (dGCN) as a surrogate model to explore the complex materials space of high-entropy alloys (HEAs). We train the dGCN on the formation energies of disordered binary alloy structures in the Pd-Pt-Sn ternary alloy system and improve predictions on ternary structures by performing reduced optimization of the formation free energy, the target property that determines HEA stability, over ensembles of ternary structures constructed based on two coordinate systems: (a) a physics-informed ternary composition space, and (b) data-driven coordinates discovered by the Diffusion Maps manifold learning scheme. Both reduced optimization techniques improve predictions of the formation free energy in the ternary alloy space with a significantly reduced number of DFT calculations compared to a high-fidelity model. The physics-based scheme converges to the target property in a manner akin to a depth-first strategy, whereas the data-driven scheme appears more akin to a breadth-first approach. Both sampling schemes, coupled with our acquisition function, successfully exploit a database of DFT-calculated binary alloy structures and energies, augmented with a relatively small number of ternary alloy calculations, to identify stable ternary HEA compositions and structures. This generalized framework can be extended to incorporate more complex bulk and surface structural motifs, and the results demonstrate that significant dimensionality reduction is possible in thermodynamic sampling problems when suitable active learning schemes are employed. [ABSTRACT FROM AUTHOR]

Heterogeneous catalysts are widely used to promote chemical reactions. Although it is known that chemical reactions usually happen on catalyst surfaces, only specific surface sites have high catalytic activity. Thus, identifying active sites and maximizing their presence lies at the heart of catalysis research1–4, in which the classic model is to categorize active sites in terms of distinct surface motifs, such as terraces and steps1,5–10. However, such a simple categorization often leads to orders of magnitude errors in catalyst activity predictions and qualitative uncertainties of active sites7,8,11,12, thus limiting opportunities for catalyst design. Here, using stepped Pt(111) surfaces and the electrochemical oxygen reduction reaction (ORR) as examples, we demonstrate that the root cause of larger errors and uncertainties is a simplified categorization that overlooks atomic site-specific reactivity driven by surface stress release. Specifically, surface stress release at steps introduces inhomogeneous strain fields, with up to 5.5% compression, leading to distinct electronic structures and reactivity for terrace atoms with identical local coordination, and resulting in atomic site-specific enhancement of ORR activity. For the terrace atoms flanking both sides of the step edge, the enhancement is up to 50 times higher than that of the atoms in the middle of the terrace, which permits control of ORR reactivity by either varying terrace widths or controlling external stress. Thus, the discovery of the above synergy provides a new perspective for both fundamental understanding of catalytically active atomic sites and design principles of heterogeneous catalysts. Stress release at stepped platinum surfaces is shown to influence the strain experienced by atoms near the steps, resulting in effects on the catalytic activity of the whole surface. [ABSTRACT FROM AUTHOR]

Heterogeneous catalytic reactions are influenced by a subtle interplay of atomic-scale factors, ranging from the catalysts' local morphology to the presence of high adsorbate coverages. Describing such phenomena via computational models requires generation and analysis of a large space of atomic configurations. To address this challenge, we present Adsorbate Chemical Environment-based Graph Convolution Neural Network (ACE-GCN), a screening workflow that accounts for atomistic configurations comprising diverse adsorbates, binding locations, coordination environments, and substrate morphologies. Using this workflow, we develop catalyst surface models for two illustrative systems: (i) NO adsorbed on a Pt3Sn(111) alloy surface, of interest for nitrate electroreduction processes, where high adsorbate coverages combined with low symmetry of the alloy substrate produce a large configurational space, and (ii) OH* adsorbed on a stepped Pt(221) facet, of relevance to the Oxygen Reduction Reaction, where configurational complexity results from the presence of irregular crystal surfaces, high adsorbate coverages, and directionally-dependent adsorbate-adsorbate interactions. In both cases, the ACE-GCN model, trained on a fraction (~10%) of the total DFT-relaxed configurations, successfully describes trends in the relative stabilities of unrelaxed atomic configurations sampled from a large configurational space. This approach is expected to accelerate development of rigorous descriptions of catalyst surfaces under in-situ conditions. A combination of electronic structure calculations and machine learning strategies is developed to predict structures of complex heterogeneous catalysts in realistic environments, yielding new opportunities for optimization for energy applications. [ABSTRACT FROM AUTHOR]

Characterizing the reaction energies and barriers of reaction networks is central to catalyst development. However, heterogeneous catalytic surfaces pose several unique challenges to automatic reaction network characterization, including large sizes and open-ended reactant sets, that make ad hoc network construction the current state-of-the-art. Here, we show how automated network exploration algorithms can be adapted to the constraints of heterogeneous systems using ethylene oligomerization on silica-supported single-site Ga3+ as a model system. Using only graph-based rules for exploring the network and elementary constraints based on activation energy and size for identifying network terminations, a comprehensive reaction network is generated and validated against standard methods. The algorithm (re)discovers the Ga-alkyl-centered Cossee-Arlman mechanism that is hypothesized to drive major product formation while also predicting several new pathways for producing alkanes and coke precursors. These results demonstrate that automated reaction exploration algorithms are rapidly maturing towards general purpose capability for exploratory catalytic applications. This study demonstrates how reaction network characterization can be performed on heterogeneous catalytic surfaces predictively, rather than retrospectively, using automated exploration algorithms on an ethylene oligomerization exemplar reaction. [ABSTRACT FROM AUTHOR]

In heterogeneous catalysis, olefin oligomerization is typically performed on immobilized transition metal ions, such as Ni2+ and Cr3+. Here we report that silica-supported, single site catalysts containing immobilized, main group Zn2+ and Ga3+ ion sites catalyze ethylene and propylene oligomerization to an equilibrium distribution of linear olefins with rates similar to that of Ni2+. The molecular weight distribution of products formed on Zn2+ is similar to Ni2+, while Ga3+ forms higher molecular weight olefins. In situ spectroscopic and computational studies suggest that oligomerization unexpectedly occurs by the Cossee-Arlman mechanism via metal hydride and metal alkyl intermediates formed during olefin insertion and β-hydride elimination elementary steps. Initiation of the catalytic cycle is proposed to occur by heterolytic C-H dissociation of ethylene, which occurs at about 250 °C where oligomerization is catalytically relevant. This work illuminates new chemistry for main group metal catalysts with potential for development of new oligomerization processes. Silica-supported, single site, main group Zn(II) and Ga(III) ions catalyze ethylene and propylene oligomerization. Here, experimental and theoretical evidence suggests a Cossee-Arlman reaction mechanism similar to that for transition metal catalysts. [ABSTRACT FROM AUTHOR]

The impact of metal/oxide interfaces on the catalytic properties of oxide-supported metal nanoparticles is a topic of longstanding interest in the field of heterogeneous catalysis. The significance of the metal/oxide interaction has been shown to vary according to both the inherent reactivity of the metal nanoparticle and the properties of the oxide support, with effects such as the metal d-band center, the nanoparticle shape, and the reducibility of the oxide believed to contribute to the overall system reactivity. In recent years, the water gas shift (WGS) reaction, wherein carbon monoxide and water are converted to carbon dioxide and hydrogen, has emerged as a model chemistry to probe the molecular-level details of how catalysis can be promoted in such environments, and this reaction is the focus of the present contribution. Using a combination of periodic Density Functional Theory calculations and microkinetic modeling, we present a comprehensive analysis of the WGS mechanism at the interface between a quasi-one dimensional platinum nanowire and an irreducible MgO support. The nanowire is lattice matched to the MgO support to remove spurious strain at the metal/oxide interface, and reactions both on the nanowire and at the three-phase boundary itself are considered in the mechanistic analysis. Additionally, to elucidate the consequences of adsorbate–adsorbate interactions on the WGS chemistry, an ab-initio thermodynamic analysis of CO coverage is performed, and the impact of the higher coverage CO states on the reaction chemistry is explicitly evaluated. These results are combined with detailed calculations of adsorbate entropies and dual-site microkinetic modeling to determine the kinetically significant features of the WGS reaction network which are subsequently, validated through experimental measurements of apparent reaction orders and activation barrier. The analysis demonstrates the important role that the metal/oxide interface plays in the reaction, with the water dissociation step being facile at the interface compared to the pure metal or oxide surfaces. Further, explicit consideration of CO interactions with other adsorbates at the metal/oxide interface is found to be central to correctly determining reaction mechanisms, rate determining steps, reaction orders, and effective activation barriers. These results are captured in a closed-form Langmuir–Hinshelwood model, derived from a simplified version of the complete microkinetic analysis, which reveals, among other results, that the celebrated carboxyl mechanism of Mavrikakis and coworkers is the governing pathway when accounting for reaction-relevant CO coverages. [ABSTRACT FROM AUTHOR]

Heterogeneous catalysts constitute a crucial component of many industrial processes, and to gain an understanding of the atomic-scale features of such catalysts, ab initio density functional theory is widely employed. Recently, growing computational power has permitted the extension of such studies to complex reaction networks involving either high adsorbate coverages or multidentate adsorbates, which bind to the surface through multiple atoms. Describing all possible adsorbate configurations for such systems, however, is often not possible based on chemical intuition alone. To systematically treat such complexities, we present a generalized Python-based graph theory approach to convert atomic scale models into undirected graph representations. These representations, when combined with workflows such as evolutionary algorithms, can systematically generate high coverage adsorbate models and classify unique minimum energy multidentate adsorbate configurations for surfaces of low symmetry, including multi-elemental alloy surfaces, steps, and kinks. Two case studies are presented which demonstrate these capabilities; first, an analysis of a coverage-dependent phase diagram of absorbate NO on the Pt3Sn(111) terrace surface, and second, an investigation of adsorption energies, together with identifying unique minimum energy configurations, for the reaction intermediate propyne (CHCCH3*) adsorbed on a PdIn(021) step surface. The evolutionary algorithm approach reproduces high coverage configurations of NO on Pt3Sn(111) using only 15% of the number of simulations required for a brute force approach. Furthermore, the screening of potentially hundreds of multidentate adsorbates is shown to be possible without human intervention. The strategy presented is quite general and can be applied to a spectrum of complex atomic systems. [ABSTRACT FROM AUTHOR]

NiFe and CoFe (MFe) layered double hydroxides (LDHs) are among the most active electrocatalysts for the alkaline oxygen evolution reaction (OER). Herein, we combine electrochemical measurements, operando X-ray scattering and absorption spectroscopy, and density functional theory (DFT) calculations to elucidate the catalytically active phase, reaction center and the OER mechanism. We provide the first direct atomic-scale evidence that, under applied anodic potentials, MFe LDHs oxidize from as-prepared α-phases to activated γ-phases. The OER-active γ-phases are characterized by about 8% contraction of the lattice spacing and switching of the intercalated ions. DFT calculations reveal that the OER proceeds via a Mars van Krevelen mechanism. The flexible electronic structure of the surface Fe sites, and their synergy with nearest-neighbor M sites through formation of O-bridged Fe-M reaction centers, stabilize OER intermediates that are unfavorable on pure M-M centers and single Fe sites, fundamentally accounting for the high catalytic activity of MFe LDHs. NiFe and CoFe layered double hydroxides are among the most active electrocatalysts for the alkaline oxygen evolution reaction. Here, by combining operando experiments and rigorous DFT calculations, the authors unravel their active phase, the reaction center and the catalytic mechanism. [ABSTRACT FROM AUTHOR]

Glycerol decomposition via a combination of dehydrogenation, C-C bond scission, and C-O bond scission reactions is examined on Pt(111) with periodic Density Functional Theory (DFT) calculations. Building upon a previous study focused on C-C bond scission in glycerol, the current work presents a first analysis of the competition between C-O and C-C bond cleavage in this reaction network. The thermochemistry of various species produced from C-O bond breaking in glycerol dehydrogenation intermediates is estimated using an extension of a previously introduced empirical correlation scheme, with parameters fit to DFT calculations. Brønsted-Evans-Polanyi (BEP) relationships are then used to estimate the kinetics of C-O bond breaking. When combined with the previous results, the thermochemical and kinetic analyses imply that, while C-O bond scission may be competitive with C-C bond scission during the early stages of glycerol dehydrogenation, the overall rates are likely to be very low. Later in the dehydrogenation process, where rates will be much higher, transition states for C-C bond scission involving decarbonylation are much lower in energy than are the corresponding transition states for C-O bond breaking, implying that the selectivity for C-C scission will be high for glycerol decomposition on smooth platinum surfaces. It is anticipated that the correlation schemes described in this work will provide an efficient strategy for estimating thermochemical and kinetic energetics for a variety of elementary bond breaking processes on Pt(111) and may ultimately facilitate computational catalyst design for these and related catalytic processes. [ABSTRACT FROM AUTHOR]

Small clusters are known to possess reactivity not observed in their bulk analogues, which can make them attractive for catalysis. Their distinct catalytic properties are often hypothesized to result from the large fraction of under-coordinated surface atoms. Here, we show that size-preselected Pt8−10 clusters stabilized on high-surface-area supports are 40–100 times more active for the oxidative dehydrogenation of propane than previously studied platinum and vanadia catalysts, while at the same time maintaining high selectivity towards formation of propylene over by-products. Quantum chemical calculations indicate that under-coordination of the Pt atoms in the clusters is responsible for the surprisingly high reactivity compared with extended surfaces. We anticipate that these results will form the basis for development of a new class of catalysts by providing a route to bond-specific chemistry, ranging from energy-efficient and environmentally friendly synthesis strategies to the replacement of petrochemical feedstocks by abundant small alkanes. [ABSTRACT FROM AUTHOR]

The discovery of more efficient, economical, and selective catalysts for oxidative dehydrogenation is of immense economic importance. However, the temperatures required for this reaction are typically high, often exceeding 400 °C. Herein, we report the discovery of subnanometer sized cobalt oxide clusters for oxidative dehydrogenation of cyclohexane that are active at lower temperatures than reported catalysts, while they can also eliminate the combustion channel. These results found for the two cluster sizes suggest other subnanometer size (CoO)x clusters will also be active at low temperatures. The high activity of the cobalt clusters can be understood on the basis of density functional studies that reveal highly active under-coordinated cobalt atoms in the clusters and show that the oxidized nature of the clusters substantially decreases the binding energy of the cyclohexene species which desorb from the cluster at low temperature. Current oxidative dehydrogenation processes are based on petroleum cracking that is indirect, environmentally unfriendly, and energy intensive. Here, the authors discover that subnanometer sized cobalt oxide clusters are active for oxidative dehydrogenation of cyclohexane at lower temperatures than reported catalysts. [ABSTRACT FROM AUTHOR]