Search

Transition metal carbides, especially Mo2C, are praised to be efficient electrocatalysts to reduce CO2 to valuable hydrocarbons. However, on Mo2C in an aqueous electrolyte, exclusively the competing hydrogen evolution reaction takes place, and this discrepancy to theory was traced back to the formation of a thin oxide layer at the electrode surface. Here, we study the CO2 reduction activity at Mo2C in a non-aqueous electrolyte to avoid such passivation and to determine products and the CO2 reduction reaction pathway. We find a tendency of CO2 to reduce to carbon monoxide. This process is inevitably coupled with the decomposition of acetonitrile to a 3-aminocrotonitrile anion. Furthermore, a unique behavior of the non-aqueous acetonitrile electrolyte is found, where the electrolyte, instead of the electrocatalyst, governs the catalytic selectivity of the CO2 reduction. This is evidenced by in situ electrochemical infrared spectroscopy on different electrocatalysts as well as by density functional theory calculations.

Ternary metal (hydro)oxide amorphous catalysts are attractive oxygen evolution reaction (OER) catalysts due to their high performance and cost-effectiveness, but a fundamental understanding of their structure-property relationships remains elusive. Herein, we fabricate a highly active ternary metal (hydro)oxide (NiFeCo) OER catalyst, showing an overpotential of 146 mV at 10 mAcm-2 and ~300 hours durability in 1M KOH. Inspired by this finding, a dataset with first-principles adsorption energies of reaction intermediates at over 300 active-site structures for both oxides and hydroxides is computed and used to train a machine-learning model for screening the dominating factors and unveiling their intrinsic contributions. The computational work confirms that adding Fe and Co makes the original Ni (hydro)oxide reach ultra-low overpotentials below 200 mV through the modulation from hydroxide towards oxide and the formation of active-sites of ternary metallic components. A series of physical properties of the Fe, Co and Ni atoms in the active-sites are further included in the analysis, and it is found that the magnetic moment (mag) plays an important role in the OER activity. This work demonstrates the application of machine-learning methods in atomic-scale active-site engineering to understand the activity mechanism of ternary metal (hydro)oxide amorphous catalysts for water oxidation, and it has the potential to be extended to wider applications.

Local field effects at the electrical double layer change the energies of reaction intermediates in heterogeneous electrocatalysis. The resulting dependence on (absolute) electrode potential can be pivotal to a catalyst's performance in acid or alkaline media. And yet, such local field effects are very difficult to describe theoretically and are often ignored. In this work, we focus on O2 adsorption as the first step of the oxygen reduction reaction (ORR) on Au(111). Different physical effects of the local field are elucidated and compared by systematically improving the model of the double layer: from an applied saw-tooth potential in vacuum, to an implicit solvent model, and explicitly modeled water via ab initio molecular dynamics (AIMD). We find all models predict a dominant dipole-field type interaction that significantly strengthens O2 binding at increasingly reducing conditions. However, only an atomically-resolved solvent model such as provided by AIMD can properly capture the additional stabilization due to explicit H-bonding from the water network. This contribution comes with the formation of a peroxo-like surface species and a more dramatic field response around the ORR onset. Our results overall demonstrate the importance of including local electric field effects in models of the electrochemical interface, while assessing the level of detail that is required to be accounted for.

Many state-of-the art machine learning (ML) interatomic potentials are based on a local or semi-local (message-passing) representation of chemical environments. They therefore lack a description of long-range electrostatic interactions and non-local charge transfer. In this context, there has been much interest in developing ML-based charge equilibration models, which allow the rigorous calculation of long-range electrostatic interactions and the energetic response of molecules and materials to external fields. The recently reported kQEq method achieves this by predicting local atomic electronegativities using Kernel ML. This paper describes the q-pac Python package, which implements several algorithmic and methodological advances to kQEq and provides an extendable framework for the development of ML charge equilibration models.

Machine learning (ML) has been widely applied to chemical property prediction, most prominently for the energies and forces in molecules and materials. The strong interest in predicting energies in particular has led to a local energy-based paradigm for modern atomistic ML models, which ensures size-extensivity and a linear scaling of computational cost with system size. However, many electronic properties (such as excitation energies or ionization energies) do not necessarily scale linearly with system size and may even be spatially localized. Using size-extensive models in these cases can lead to large errors. In this work, we explore different strategies for learning intensive and localized properties, using HOMO energies in organic molecules as a representative test case. In particular, we analyze the pooling functions that atomistic neural networks use to predict molecular properties, and suggest an orbital weighted average (OWA) approach that enables the accurate prediction of orbital energies and locations.

Transition metal nitrides are a fascinating class of catalyst materials due to their superior catalytic activity, low electrical resistance, good corrosion resistance and earth abundance; however, their conventional synthesis relies on high-temperature nitridation processes in hazardous environments. Here, we report a direct synthesis of Ni3N/Ni enriched with N-vacancies using one-step magnetron sputtering. The surface state of Ni3N(001) with 75% N-vacancies is hydrogen-terminated and exhibits four inequivalent Ni3-hollow sites. This leads to stronger H* binding compared to Ni(111), and is affirmed as the most stable surface termination under the electrochemical working conditions (pH ≈ 13.8 and E = −0.1 V) from the Pourbaix diagram. The Ni3N/Ni catalyst shows low crystallinity and good wettability and exhibits a low overpotential of 89 mV vs. RHE at 10 mA cm−2 in 1.0 M KOH with excellent stability over 3 days. This performance closely matches that of the Pt catalyst synthesized under the same conditions and surpasses that of other reported earth-abundant catalysts on planar substrates. The application of Ni3N/Ni as a cocatalyst on Si photocathodes produces an excellent ABPE of 9.3% and over 50 h stability. Moreover, its feasibility for practical application was confirmed with excellent performance on porous substrates and robustness at high operating currents in zero-gap alkaline electrolysis cells. Our work demonstrates a general approach for the feasible synthesis of other transition metal nitride catalysts for electrochemical and photoelectrochemical energy conversion applications.

Tantalum cluster cations react with CO2 either via transfer of oxygen atoms to the clusters or the adsorption of an entire molecule. The released energy and vibrational heat capacities are assigned to determine the branching ratios of the pathways.

The molecular reorganization energy $\lambda$ strongly influences the charge carrier mobility of organic semiconductors and is therefore an important target for molecular design. Machine learning (ML) models generally have the potential to strongly accelerate this design process (e.g. in virtual screening studies) by providing fast and accurate estimates of molecular properties. While such models are well established for simple properties (e.g. the atomization energy), $\lambda$ poses a significant challenge in this context. In this paper, we address the questions of how ML models for $\lambda$ can be improved and what their benefit is in high-throughput virtual screening (HTVS) studies. We find that, while improved predictive accuracy can be obtained relative to a semiempirical baseline model, the improvement in molecular discovery is somewhat marginal. In particular, the ML enhanced screenings are more effective in identifying promising candidates but lead to a less diverse sample. We further use substructure analysis to derive a general design rule for organic molecules with low $\lambda$ from the HTVS results.

The computational discovery of novel materials has been one of the main motivations behind research in theoretical chemistry for several decades. Despite much effort, this is far from a solved problem, however. Among other reasons, this is due to the enormous space of possible structures and compositions that could potentially be of interest. In the case of inorganic materials, this is exacerbated by the combinatorics of the periodic table, since even a single crystal structure can in principle display millions of compositions. Consequently, there is a need for tools that enable a more guided exploration of the materials design space. Here, generative machine learning (ML) models have recently emerged as a promising technology. In this work, we assess the performance of a range of deep generative models based on Reinforcement Learning (RL), Variational Autoencoders (VAE) and Generative Adversarial Networks (GAN) for the prototypical case of designing Elpasolite compositions with low formation energies. By relying on the fully enumerated space of 2~million main group Elpasolites, the precision, coverage and diversity of the generated materials is rigorously assessed. Additionally, a hyperparameter selection scheme for generative models in chemical composition space is developed.

The adsorption energies of molecular adsorbates on catalyst surfaces are key descriptors in computational catalysis research. For the relatively large reaction intermediates frequently encountered, e.g., in syngas conversion, a multitude of possible binding motifs leads to complex potential energy surfaces (PES), however. This implies that finding the optimal structure is a difficult global optimization problem, which leads to significant uncertainty about the stability of many intermediates. To tackle this issue, we present a global optimization protocol for surface adsorbate geometries which trains a surrogate machine learning potential on-the-fly. The approach is applicable to arbitrary surface models and adsorbates and minimizes both human intervention and the number of required DFT calculations by iteratively updating the training set with configurations explored by the algorithm. We demonstrate the efficiency of this approach for a diverse set of adsorbates on the Rh(111) and (211) surfaces.

Hydrogenated graphene (H-Gr) is an extensively studied system not only because of its capabilities as a simplified model system for hydrocarbon chemistry but also because hydrogenation is a compelling method for Gr functionalization. However, knowledge of how H-Gr interacts with molecules at higher pressures and ambient conditions is lacking. Here we present experimental and theoretical evidence that room temperature O2 exposure at millibar pressures leads to preferential removal of H dimers on H-functionalized graphene, leaving H clusters on the surface. Our density functional theory (DFT) analysis shows that the removal of H dimers is the result of water or hydrogen peroxide formation. For water formation, we show that the two H atoms in the dimer motif attack one end of the physisorbed O2 molecule. Moreover, by comparing the reaction pathways in a vacuum with the ones on free-standing graphene and on the graphene/Ir(111) system, we find that the main role of graphene is to arrange the H atoms in geometrical positions, which facilitates the activation of the O═O double bond.

ChemCatChem : the European Society Journal for Catalysis 14(19), e202200300 (2022). doi:10.1002/cctc.202200300 special issue: "Catalysts and Reactors under Dynamic Conditions for Energy Storage and Conversion", Published by Wiley-VCH, Weinheim

Detrimental to the performance of solid oxide cells (SOCs) is the interface between electrolyte and each electrode: for purely electron-conducting electrodes, it is part of the triple phase boundary where the reactions take place, and for mixed ionic and electronic conductor (MIEC) electrodes, the oxygen ions coming from or moving into the electrolyte are transported through this phase boundary. As such, the performance of SOFCs and SOECs is strongly dependent on the chemical nature of this interface. By combining state-of-the-art electron microscopy, synchrotron-based X-ray spectroscopy and theoretical calculations, we were able to obtain an atomically resolved picture of the interface between yttria-stabilized zirconia (YSZ, with 8 mol% Y2O3) and lanthanum strontium manganite (LSM, (La0.8Sr0.2)0.95MnO3-δ).[1] Energy-dispersive X-ray (EDX) spectroscopy measurements in the transmission electron microscope (TEM) reveal the presence of a significant amount of inter-diffusion between YSZ and LSM. Only strontium does not show any diffusion into YSZ, resulting in a shift of its intensity distribution profile of approximately 0.8 nm with respect to all other elements. All this is in agreement with Monte-Carlo-based simulations that were used to theoretically model the YSZ/LSM boundary including the ion- as well as site-specific swapping probabilities for the diffusion process.[1] According to these simulations, this diffusion region is slightly amorphous. By means of atomically resolved scanning TEM (STEM), this reduction of long-range order was observed experimentally for an approximately 1.5 nm wide slab on the YSZ side of the boundary, clearly distinguishing it from the bulk fluorite structure. In materials science, distinct ‘2D-like’ layers at grain boundaries and interfaces are often referred to as ’complexions’. While these can in general occur in different types, depending on thickness and retained order,[2] they are all characterized by a self-limited (finite) width and thermodynamic stability obtained only by confinement in between two bulk phases (i.e. they cannot exist on their own without neighboring phases).[2] These complexions have recently also been discovered in other energy-related systems such as battery materials.[3] The presence of such a stable complexion might be the reason why the YSZ/LSM interface is not as prone to the formation of lanthanum or strontium zirconates as other perovskites such as lanthanum strontium cobalt ferrite (LSCF).[4] Consequently, it might have significant influence on the chemical stability of such SOCs and a thorough understanding of the complexions between electrolyte and electrode may in the future allow fine tuning of SOC performance and stability. This requires the determination of electrochemical and thermodynamic properties of these interface complexions for which we will, in a next step, employ theoretical simulations and experimental techniques in order to find out how different transport phenomena, widths and stabilities of the layers behave with temperature, time and chemical environment. [1] H. Tuerk, F.-P. Schmidt, T. Götsch et al. in preparation [2] S. J. Dillon et al. Acta Mater. 2007, 55, 6208–6218 [3] J. Timmermann et al. Phys. Rev. Lett. 2020, 125, 206101 [4] W. Wang et al. J. Electrochem. Soc. 2006, 153, A2066

Compound materials, such as transition-metal (TM) carbides, are anticipated to be effective electrocatalysts for the carbon dioxide reduction reaction (CO2RR) to useful chemicals. This expectation is nurtured by density functional theory (DFT) predictions of a break of key adsorption energy scaling relations that limit CO2RR at parent TMs. Here, we evaluate these prospects for hexagonal Mo2C in aqueous electrolytes in a multimethod experiment and theory approach. We find that surface oxide formation completely suppresses the CO2 activation. The oxides are stable down to potentials as low as −1.9 V versus the standard hydrogen electrode, and solely the hydrogen evolution reaction (HER) is found to be active. This generally points to the absolute imperative of recognizing the true interface establishing under operando conditions in computational screening of catalyst materials. When protected from ambient air and used in nonaqueous electrolyte, Mo2C indeed shows CO2RR activity.

Cyclic voltammograms (CVs) are a central experimental tool for assessing the structure and activity of electrochemical interfaces. Based on a mean-field ansatz for the interface energetics under applied potential conditions, we here derive an ab initio thermodynamics approach to efficiently simulate thermodynamic CVs. All unknown parameters are determined from density functional theory (DFT) calculations coupled to an implicit solvent model. For the showcased CVs of Ag(111) electrodes in halide-anion-containing solutions, these simulations demonstrate the relevance of double-layer contributions to explain experimentally observed differences in peak shapes over the halide series. Only the appropriate account of interfacial charging allows us to capture the differences in equilibrium coverage and total electronic surface charge that cause the varying peak shapes. As a case in point, this analysis demonstrates that prominent features in CVs do not only derive from changes in adsorbate structure or coverage but can also be related to variations of the electrosorption valency. Such double-layer effects are proportional to adsorbate-induced changes in the work function and/or interfacial capacitance. They are thus especially pronounced for electronegative halides and other adsorbates that affect these interface properties. In addition, the analysis allows us to draw conclusions on how the possible inaccuracy of implicit solvation models can indirectly affect the accuracy of other predicted quantities such as CVs.

The combination of modern machine learning (ML) approaches with high-quality data from quantum mechanical (QM) calculations can yield models with an unrivalled accuracy/cost ratio. However, such methods are ultimately limited by the computational effort required to produce the reference data. In particular, reference calculations for periodic systems with many atoms can become prohibitively expensive for higher levels of theory. This trade-off is critical in the context of organic crystal structure prediction (CSP). Here, a data-efficient ML approach would be highly desirable, since screening a huge space of possible polymorphs in a narrow energy range requires the assessment of a large number of trial structures with high accuracy. In this contribution, we present tailored Δ-ML models that allow screening a wide range of crystal candidates while adequately describing the subtle interplay between intermolecular interactions such as H-bonding and many-body dispersion effects. This is achieved by enhancing a physics-based description of long-range interactions at the density functional tight binding (DFTB) level—for which an efficient implementation is available—with a short-range ML model trained on high-quality first-principles reference data. The presented workflow is broadly applicable to different molecular materials, without the need for a single periodic calculation at the reference level of theory. We show that this even allows the use of wavefunction methods in CSP., Using a cluster-based training scheme and a physical baseline, data efficient machine-learning models for crystal structure prediction are developed, enabling accurate structural relaxations of molecular crystals with unprecedented efficiency.

Potential dependencies of adsorption energies at metal electrodes are often interpreted in terms of a dipole-field interaction picture. Classical electrostatics would then suggest particularly strong dependencies at protruding defect sites like steps and kinks due to a local field enhancement. Here, we use first-principles density functional theory calculations that capture the capacitive charging of the electric double layer to analyze these dependencies for prototypical adsorbates at vicinal Pt(111) surfaces in an implicit aqueous electrolyte. This analysis confirms effective dipole-field-type dependencies but rationalizes why they can neither be estimated on the basis of accessible molecular dipoles nor the mere atomic-scale geometry of the adsorption site. The observed magnitude questions mechanistic analyses of surface catalytic reactions made on the basis of the prevalent computational hydrogen electrode approach that is agnostic to these potential-induced adsorption energy variations.

Computational screening of metal oxide catalysts is challenging due to their more localized and intricate electronic structure as compared to metal catalysts and the resulting lack of suitable acti...

Computational screening based on first-principles microkinetic modeling has evolved into a widespread tool for catalyst discovery. Efficiently exploiting various scaling relations, this approach dr...

The surfaces of compound catalyst materials generally exhibit a wide range of geometric and compositional motives that could act as active sites. We address this inherent complexity by extending co...

The development of low-temperature fuel cells for clean energy production is an appealing alternative to fossil-fuel technologies. CO is a key intermediate in the electro-oxidation of energy carrying fuels and, due to its strong interaction with state-of-the-art Pt electrodes, it is known to act as a poison. Here we demonstrate the ability of Earth-abundant Cu to electro-oxidize CO efficiently in alkaline media, reaching high current densities of ≥0.35 mA cm−2 on single-crystal Cu(111) model catalysts. Strong and continuous surface structural changes are observed under reaction conditions. Supported by first-principles microkinetic modelling, we show that the concomitant presence of high-energy undercoordinated Cu structures at the surface is a prerequisite for the high activity. Similar CO-induced self-activation has been reported for gas–surface reactions at coinage metals, demonstrating the strong parallels between heterogeneous thermal catalysis and heterogeneous electrocatalysis. CO is a key intermediate in the electro-oxidation of energy carrying fuels which typically acts as a poison. Here, the authors demonstrate that Cu is an efficient CO electro-oxidation catalyst in alkaline electrolyte due to the continuous formation of undercoordinated active Cu adatom sites in the presence of CO and OH.

The Density-Functional Tight Binding (DFTB) method is a popular semiempirical approximation to Density Functional Theory (DFT). In many cases, DFTB can provide comparable accuracy to DFT at a fraction of the cost, enabling simulations on length and time scales that are unfeasible with first-principles DFT. At the same time (and in contrast to empirical interatomic potentials and force fields), DFTB still offers direct access to electronic properties such as the band structure. These advantages come at the cost of introducing empirical parameters to the method, leading to a reduced transferability compared to true first-principle approaches. Consequently, it would be very useful if the parameter sets could be routinely adjusted for a given project. While fairly robust and transferable parametrization workflows exist for the electronic structure part of DFTB, the so-called repulsive potential Vrep poses a major challenge. In this paper, we propose a machine-learning (ML) approach to fitting Vrep, using Gaussian Process Regression (GPR) to reconstruct Vrep with DFT-DFTB force residues as training data. The use of GPR circumvents the need for nonlinear or global parameter optimization, while at the same time offering arbitrary flexibility in terms of the functional form. We also show that the proposed method can be applied to multiple elements at once, by fitting repulsive potentials for organic molecules containing carbon, hydrogen, and oxygen. Overall, the new approach removes focus from the choice of functional form and parametrization procedure, in favor of a data-driven philosophy.

Using first-principles density-functional theory calculations combined with ab initio thermodynamics, we introduce a design protocol for RuO2-based core-shell catalysts which exhibit enhanced stability and activity under oxygen evolution reaction (OER) operating conditions. Due to their high activity and favorable stability in acidic electrolytes, Ir and Ru oxides are primary catalysts for the oxygen evolution reaction (OER) in proton-exchange membrane (PEM) electrolyzers. For a future large-scale application, core-shell nanoparticles are an appealing route to minimize the demand for these precious oxides. Here, we employ first-principles density-functional theory (DFT) and ab initio thermodynamics to assess the feasibility of encapsulating a cheap rutile-structured TiO2 core with coherent, monolayer-thin IrO2 or RuO2 films. Resulting from a strong directional dependence of adhesion and strain, a wetting tendency is only obtained for some low-index facets under typical gas-phase synthesis conditions. Thermodynamic stability in particular of lattice-matched RuO2 films is instead indicated for more oxidizing conditions. Intriguingly, the calculations also predict an enhanced activity and stability of such epitaxial RuO2/TiO2 core-shell particles under OER operation.

The discovery of new catalytically active materials is one of the holy grails of computational chemistry as it has the potential to accelerate the adoption of renewable energy sources and reduce the energy consumption of chemical industry. Indeed, heterogeneous catalysis is essential for the production of synthetic fuels and many commodity chemicals. Consequently, novel solid catalysts with higher activity and selectivity, increased sustainability and longevity, or improved prospects for rejuvenation and cyclability are needed for a diverse range of processes. Unfortunately, computational catalyst discovery is a daunting task, among other reasons because it is often unclear whether a proposed material is stable or synthesizable. This perspective proposes a new approach to this challenge, namely the use of generative grammars. We outline how grammars can guide the search for stable catalysts in a large chemical space and sketch out several research directions that would make this technology applicable to real materials.