Search

While the vision of accelerating materials discovery using data driven methods is well-founded, practical realization has been throttled due to challenges in data generation, ingestion, and materials state-aware machine learning. High-throughput experiments and automated computational workflows are addressing the challenge of data generation, and capitalizing on these emerging data resources requires ingestion of data into an architecture that captures the complex provenance of experiments and simulations. In this manuscript, we describe an event-sourced architecture for materials provenance (ESAMP) that encodes the sequence and interrelationships among events occurring in a simulation or experiment. We use this architecture to ingest a large and varied dataset (MEAD) that contains raw data and metadata from millions of materials synthesis and characterization experiments performed using various modalities such as serial, parallel, multimodal experimentation. Our data architecture tracks the evolution of a material’s state, enabling a demonstration of how stateequivalency rules can be used to generate datasets that significantly enhance data-driven materials discovery. Specifically, using state-equivalency rules and parameters associated with statechanging processes in addition to the typically used composition data, we demonstrated marked reduction of uncertainty in prediction of overpotential for oxygen evolution reaction (OER) catalysts. Finally, we discuss the importance of ESAMP architecture in enabling several aspects of accelerated materials discovery such as dynamic workflow design, generation of knowledge graphs, and efficient integration of theory and experiment.

Rational solid-state synthesis of inorganic compounds is formulated as catalytic nucleation on crystalline reactants, where contributions of reaction and interfacial energies to the nucleation barriers are approximated from high-throughput thermochemical data, and structural and interfacial features of crystals, respectively. Favorable synthesis reactions are then identified by a Pareto analysis of relative nucleation barriers and phase-selectivities of reactions leading to the target. We demonstrate the application of this approach in reaction planning for solid-state synthesis of a range of compounds, including the widely-studied oxides LiCoO2, BaTiO3 and YBa2Cu3O7, as well as other metal oxide, oxyfluoride, phosphate and nitride targets. Pathways for enabling retrosynthesis of inorganics are also discussed.

We introduce a new agent-based framework for materials discovery that combines multi-fidelity modeling and sequential learning to lower the number of expensive data acquisitions while maximizing discovery. We demonstrate the framework's capability by simulating a materials discovery campaign using experimental and DFT band gap data. Using these simulations, we determine how different machine learning models and acquisition strategies influence the overall rate of discovery of materials per experiment. The framework demonstrates that including lower fidelity (DFT) data, whether as a-priori knowledge or using in-tandem acquisition, increases the discovery rate of materials suitable for solar photoabsorption. We also show that the performance of a given agent depends on data size, model selection, and acquisition strategy. As such, our framework provides a tool that enables materials scientists to test various acquisition and model hyperparameters to maximize the discovery rate of their own multi-fidelity sequential learning campaigns for materials discovery.

Machine learning has emerged as a powerful approach in materials discovery. Its major challenge is selecting features that create interpretable representations of materials, useful across multiple prediction tasks. We introduce an end-to-end machine learning model that automatically generates descriptors that capture a complex representation of a material's structure and chemistry. This approach builds on computational topology techniques (namely, persistent homology) and word embeddings from natural language processing. It automatically encapsulates geometric and chemical information directly from the material system. We demonstrate our approach on multiple nanoporous metal-organic framework datasets by predicting methane and carbon dioxide adsorption across different conditions. Our results show considerable improvement in both accuracy and transferability across targets compared to models constructed from the commonly-used, manually-curated features, consistently achieving an average 25-30% decrease in root-mean-squared-deviation and an average increase of 40-50% in R2 scores. A key advantage of our approach is interpretability: Our model identifies the pores that correlate best to adsorption at different pressures, which contributes to understanding atomic-level structure--property relationships for materials design., Comment: 14 pages main text, 8 figures

Battery evaluation and early prediction software package (BEEP) provides an open-source Python-based framework for the management and processing of high-throughput battery cycling data-streams. BEEPs features include file-system based organization of raw cycling data and metadata received from cell testing equipment, validation protocols that ensure the integrity of such data, parsing and structuring of data into Python-objects ready for analytics, featurization of structured cycling data to serve as input for machine-learning, and end-to-end examples that use processed data for anomaly detection and featurized data to train early-prediction models for cycle life. BEEP is developed in response to the software and expertise gap between cell-level battery testing and data-driven battery development.

A recent meeting focused on accelerated materials design and discovery examined user requirements for a general, collaborative, integrative, and on-demand materials research platform.

Materials discovery is increasingly being impelled by machine learning methods that rely on pre-existing datasets. Where datasets are lacking, unbiased data generation can be achieved with genetic algorithms. Here a machine learning model is trained on-the-fly as a computationally inexpensive energy predictor before analyzing how to augment convergence in genetic algorithm-based approaches by using the model as a surrogate. This leads to a machine learning accelerated genetic algorithm combining robust qualities of the genetic algorithm with rapid machine learning. The approach is used to search for stable, compositionally variant, geometrically similar nanoparticle alloys to illustrate its capability for accelerated materials discovery, e.g., nanoalloy catalysts. The machine learning accelerated approach, in this case, yields a 50-fold reduction in the number of required energy calculations compared to a traditional “brute force” genetic algorithm. This makes searching through the space of all homotops and compositions of a binary alloy particle in a given structure feasible, using density functional theory calculations.

Assessing the synthesizability of inorganic materials is a grand challenge for accelerating their discovery using computations. Synthesis of a material is a complex process that depends not only on its thermodynamic stability with respect to others, but also on factors from kinetics, to advances in synthesis techniques, to the availability of precursors. This complexity makes the development of a general theory or first-principles approach to synthesizability currently impractical. Here we show how an alternative pathway to predicting synthesizability emerges from the dynamics of the materials stability network: a scale-free network constructed by combining the convex free-energy surface of inorganic materials computed by high-throughput density functional theory and their experimental discovery timelines extracted from citations. The time-evolution of the underlying network properties allows us to use machine-learning to predict the likelihood that hypothetical, computer-generated materials will be amenable to successful experimental synthesis., Comment: 20 pages, 5 figures

We discuss three concepts that have made it possible to develop a quantitative understanding of trends in transition-metal catalysis: scaling relations, activity maps, and the d-band model. Scaling relations are correlations between surface bond energies of different adsorbed species including transition states; they open the possibility of mapping the many parameters determining the rate of a full catalytic reaction onto a few descriptors. The resulting activity map can be viewed as a quantitative implementation of the classical Sabatier principle, which states that there is an optimum “bond strength” defining the best catalyst for a given reaction. In the modern version, the scaling relations determine the relevant “bond strengths” and the fact that these descriptors can be measured or calculated makes it a quantitative theory of catalysis that can be tested experimentally by making specific predictions of new catalysts. The quantitative aspect of the model therefore provides new possibilities in catalyst design. Finally, the d-band model provides an understanding of the scaling relations and variations in catalytic activity in terms of the electronic structure of the transition-metal surface.

Density functional theory calculations can be used to gain valuable insight into the fundamental reaction processes in metal–oxygen systems, e.g., metal–oxygen batteries. Here, the ability of a range of different exchange-correlation functionals to reproduce experimental enthalpies of formation for different types of alkali and alkaline earth metal oxide species has been examined. Most examined functionals result in significant overestimation of the stability of superoxide species compared to peroxides and monoxides, which can result in erroneous prediction of reaction pathways. We show that if metal chlorides are used as reference structures instead of metals, the systematic errors are significantly reduced and functional variations decreased. Using a metal chloride reference, where the metal atoms are in the same oxidation state as in the oxide species, will provide a computationally inexpensive and robust approach to significantly improve accuracy. The approach can potentially be applied to improve acc...

This paper demonstrates the utility of a new tool for mapping out the activities and selectivities of transition metals for catalytic reactions. A database of reaction and activation energies has recently been made available to the public via a web-based application called “CatApp”. Combined with microkinetic modeling, CatApp allows researchers to gain insights into the energetics and mechanisms of a given catalytic reaction. This paper illustrates this utility using the anhydrous dehydrogenation of methanol to formaldehyde as an example. There are presently no stable catalysts for this potentially important process. The model developed in this paper presents possible explanations for this observation and suggests certain properties that would be needed to design a good catalyst.

Low electronic conduction is expected to be a main limiting factor in the performance of reversible lithium–air, Li–O2, batteries. Here, we apply density functional theory and non-equilibrium Green's function calculations to determine the electronic transport through lithium peroxide, Li2O2, formed at the cathode during battery discharge. We find the transport to depend on the orientation and lattice matching of the insulator–metal interface in the presence of Au and Pt catalysts. Bulk lithium vacancies are found to be available and mobile under battery charging conditions, and found to pin the Fermi level at the top of the anti bonding peroxide π*(2px) and π*(2py) levels in the Li2O2 valence band. Under an applied bias, this can result in a reduced transmission, since the anti bonding σ*(2pz) level in the Li2O2 conduction band is found to couple strongly to the metal substrate and create localized interface states with poor coupling to the Li2O2 bulk states. These observations provide a possible explanation for the higher overpotential observed for charging than discharge.

We use XPS and isotope labeling coupled with differential electrochemical mass spectrometry (DEMS) to show that small amounts of carbonates formed during discharge and charge of Li-O2 cells in ether electrolytes originate from reaction of Li2O2 (or LiO2) both with the electrolyte and with the C cathode. Reaction with the cathode forms approximately a monolayer of Li2CO3 at the C-Li2O2 interface, while reaction with the electrolyte forms approximately a monolayer of carbonate at the Li2O2-electrolyte interface during charge. A simple electrochemical model suggests that the carbonate at the electrolyte-Li2O2 interface is responsible for the large potential increase during charging (and hence indirectly for the poor rechargeability). A theoretical charge-transport model suggests that the carbonate layer at the C-Li2O2 interface causes a 10-100 fold decrease in the exchange current density. These twin "interfacial carbonate problems" are likely general and will ultimately have to be overcome to produce a highly rechargeable Li-air battery.

The formation and oxidation of the main discharge product in nonaqueous secondary Li−O2 batteries, that is, Li2O2, has been studied intensively, but less attention has been given to the formation of cathode−electrolyte interfaces, which can significantly influence the performance of the Li−O2 battery. Here we apply density functional theory with the Hubbard U correction (DFT+U) and nonequilibrium Green’s function (NEGF) methods to investigate the role of Li2O2@Li2CO3 interface layers on the ionic and electronic transport properties at the oxygen electrode. We show that, for example, lithium vacancies accumulate at the peroxide part of the interface during charge, reducing the coherent electron transport by two to three orders of magnitude compared with pristine Li2O2. During discharge, Li2O2@Li2CO3 interfaces may, however, provide an alternative in-plane channel for fast electron polaron hopping that could improve the electronic conductivity and ultimately increase the practical capacity in nonaqueous Li−O2 batteries.

The use of methanol as a fuel and chemical feedstock could become very important in the development of a more sustainable society if methanol could be efficiently obtained from the direct reduction of CO2 using solar-generated hydrogen. If hydrogen production is to be decentralized, small-scale CO2 reduction devices are required that operate at low pressures. Here, we report the discovery of a Ni-Ga catalyst that reduces CO2 to methanol at ambient pressure. The catalyst was identified through a descriptor-based analysis of the process and the use of computational methods to identify Ni-Ga intermetallic compounds as stable candidates with good activity. We synthesized and tested a series of catalysts and found that Ni5Ga3 is particularly active and selective. Comparison with conventional Cu/ZnO/Al2O3 catalysts revealed the same or better methanol synthesis activity, as well as considerably lower production of CO. We suggest that this is a first step towards the development of small-scale low-pressure devices for CO2 reduction to methanol.

Solid catalysts form the backbone of the chemical industry and the hydrocarbon-based energy sector. Most catalysts and processes today are highly optimized, but there is still considerable room for improvements in reactivity and selectivity in order to lower energy consumption and waste production. In addition, the development of sustainable energy solutions is a tremendous challenge to catalysis science and engineering. The ability to store solar energy as a fuel calls for new catalysts, as does the development of a sustainable chemical industry that is based on biomass and other non-fossil building blocks. The development of new catalysts could be accelerated significantly if we had access to systematic data for the activation energies of elementary surface reactions. Once the key parameters that determine the activity or selectivity of a certain process have been established through experiments or calculations, such a database would enable searches for new catalyst leads. Ideally, data would come from detailed, systematic experiments, but it is generally not possible to find such data. Electronic structure calculations provide a powerful alternative. The accuracy is not such that detailed predictions of absolute rates of elementary reaction steps can be made, but for classes of interesting catalysts (such as transition metals) it is possible to create systematic data with sufficient accuracy to predict trends in reactivity. Herein, we introduce such a set of calculated reaction energies and activation energies for a large number of elementary surface reactions on a series of metal single-crystal surfaces, including surfaces with defects such as steps. We also introduce a simple web application (CatApp) for accessing these data. The data will be part of a larger database of surface reaction data that are being developed under the Quantum Materials Informatics Project. The database includes reaction energies for all surface reactions that involve C C, C H, C O, O O, O H, N N, C N, O N, N H splitting for molecules with up to three C, N, or O atoms on close-packed face-centered cubic fcc(111), hexagonal close-packed hcp(0001), and body-centered cubic bcc(110) surfaces, as well as stepped fcc and hcp surfaces. The metals included in the database are Ag, Au, Co, Cu, Fe, Ir, Mo, Ni, Pd, Pt, Re, Rh, Ru, Sc, V. The data have been compiled from previous reports, where details of the calculations can be found. The key point is that the values have all been calculated with the same code (DACAPO), the same exchange-correlation energy functional (GGARPBE), and similar calculational parameters. Therefore one adsorption energy or reaction barrier can be compared to another with some confidence. Gas-phase CO2 and O2 , for which the RPBE functional performs poorly, were corrected as described in Refs. [25] and [26], respectively In cases where there are no calculated data for a given reaction, we use the recently developed scaling relations to provide an estimate. The scaling relations link the adsorption energies of different molecules that contain varying amounts of hydrogen. In a similar fashion, we exploit the fact that transition-state energies are quite generally found to scale with reaction energies. We have developed a simple visual query tool for accessing the data presented above. On our homepage, we maintain a list of hyperlinks to the available versions of the tool, together with a list of references to the scientific data it employs. The tool is a web application implemented in JavaScript, SVG, and HTML, and runs in modern web browsers without any plug-ins. The application can be easily used on computers and portable devices with a touch interface. By running the application, one can choose a surface and an elementary reaction and be presented with a reaction path that reports the reaction and activation energy. If a DFT value is not found for a given reaction and a scaling estimate is used, the value is shown in italic type. In Figure 1 we have shown an example of the use of the application. The energy barrier needed to break the N2 bond on two different surface orientations of ruthenium, Ru(0001) and stepped Ru(0001), are extracted. The plots immediately show the structure dependence of this important step in the Haber–Bosch process (N2+ 3H2!2NH3). Our web application will allow anyone to download data such as that shown in Figure 1, and to quickly explore whether there may be other metals or structures where the N2 bond is broken more readily. All code and data are downloaded when the application is accessed for the first time and is kept in the local storage of the browser. This feature allows the application to be used even when the user has no internet connection. More importantly, it guarantees the user complete privacy, since all queries are performed locally in the browser and not by connecting to our server. The only information that is delivered from the user to our server is an anonymous [*] Dr. J. S. Hummelshoj, Dr. F. Abild-Pedersen, Dr. F. Studt, Dr. T. Bligaard, Prof. J. K. Norskov SUNCAT Center for Interface Science and Catalysis SLAC National Accelerator Laboratory 2575 Sand Hill Road, Menlo Park, CA 94025 (USA)

We present the first comprehensive comparison between free energies, based on a phonon dispersion calculation within density functional theory, of theoretically predicted structures and the experimentally proposed α (P6(1)) and β (Fddd) phases of the promising hydrogen storage material Mg(BH(4))(2). The recently proposed low-density [Formula: see text] ground state is found to be thermodynamically unstable, with soft acoustic phonon modes at the Brillouin zone boundary. We show that such acoustic instabilities can be detected by a macroscopic distortion of the unit cell. Following the atomic displacements of the unstable modes, we have obtained a new F 222 structure, which has a lower energy than all previously experimentally and theoretically proposed phases of Mg(BH(4))(2) and is free of imaginary eigenmodes. A new meta-stable high-density I4(1)/amd structure is also derived from the [Formula: see text] phase. Temperatures for the decomposition are found to be in the range of 400-470 K and largely independent of the structural complexity, as long as the primary cation coordination polyhedra are properly represented. This opens a possibility of using simple model structures for screening and prediction of finite temperature stability and decomposition temperatures of novel borohydride systems.

Structural transition properties of the LiBH4+xLiI (x=0–1.00) pseudobinary system were examined by powder x-ray diffraction and differential scanning calorimetry combined with periodic density functional theory calculations. We experimentally and computationally confirmed the stabilization of the high-temperature [hexagonal, lithium super(fast-)ionic conduction] phase of LiBH4 with x=0.33 and 1.00, and the results also imply the existence of intermediate phases with x=0.07–0.20. The studies are of importance for further development of LiBH4 and the derived hydrides as solid-state electrolytes. ©2009 American Institute of Physics

It is shown that nanopores are formed during desorption of NH3 from Mg(NH3)6Cl2, which has been proposed as a hydrogen storage material. The system of nanopores facilitates the transport of desorbed ammonia away from the interior of large volumes of compacted storage material. DFT calculations show that there exists a continuous path from the initial Mg(NH3)6Cl2 material to MgCl2 that does not involve large-scale material transport. Accordingly, ammonia desorption from this system is facile.

We develop a density functional theory model for the electrochemical growth and dissolution of Li(2)O(2) on various facets, terminations, and sites (terrace, steps, and kinks) of a Li(2)O(2) surface. We argue that this is a reasonable model to describe discharge and charge of Li-O(2) batteries over most of the discharge-charge cycle. Because non-stoichiometric surfaces are potential dependent and since the potential varies during discharge and charge, we study the thermodynamic stability of facets, terminations, and steps as a function of potential. This suggests that different facets, terminations, and sites may dominate in charge relative to those for discharge. We find very low thermodynamic overpotentials (0.2 V) for both discharge and charge at many sites on the facets studied. These low thermodynamic overpotentials for both discharge and charge are in very good agreement with the low kinetic overpotentials observed in recent experiments. However, there are other predicted paths for discharge/charge that have higher overpotentials, so the phase space available for the electrochemistry opens up with overpotential.

With surging interest in high energy density batteries, much attention has recently been devoted to metal-air batteries. The zinc-air battery has been known for more than a hundred years and is commercially available as a primary battery, but recharging has remained elusive, in part because the fundamental mechanisms still remain to be fully understood. Here, we present a density functional theory investigation of the zinc dissolution (oxidation) on the anode side in the zinc-air battery. Two models are envisaged, the most stable (0001) surface and a kink surface. The kink model proves to be more accurate as it brings about some important features of bulk dissolution and yields results in good agreement with experiments. From the adsorption energies of hydroxyl species and experimental values, we construct a free energy diagram and confirm that there is a small overpotential associated with the reaction. The applied methodology provides new insight into computational modelling and design of secondary metal-air batteries.

Energy storage in the form of ammonia bound in metal salts, so-called metal ammines, combines high energy density with the possibility of fast and reversible NH3 ab- and desorption kinetics. The mechanisms and processes involved in the NH3 kinetics are investigated by density functional theory (DFT) and quasielastic neutron scattering (QENS). The crystal structures of Mg(NH3)nCl2 with n = 6, 2, 1, which contains up to 9.19 wt % hydrogen and 0.115 kg hydrogen L−1, are first analyzed using an algorithm based on simulated annealing (SA), finding all the experimentally known structures and predicting the C2/m structure for the uncharacterized low temperature phase of Mg(NH3)6Cl2. It is found from DFT that the rotation of ammonia in the hexammine complex (n = 6) requires an activation energy of 0.09 eV in the low temperature phase of Mg(NH3)6Cl2 and 0.002–0.12 eV in the high temperature phases; effectively having free rotors as observed experimentally. The findings are supported by the QENS data, which identify C3 rotations of NH3 in the low temperature phase with an activation energy of 0.09 eV. The calculated diffusion rates were found to be 106–107 Hz at the desorption temperatures for all n = 6, 2, 1 systems. DFT calculations involving bulk diffusion of NH3 correctly reproduces the trends observed in the experimental desorption enthalpies. In particular, for n = 6, 2, 1, there is a good agreement between activation barriers and experimental enthalpies. These results indicate that the desorption of NH3 is likely to be diffusion limited.

The indirect hydrogen storage capabilities of Mg(NH 3) 6Cl 2, Ca(NH 3) 8Cl 2, Mn(NH 3) 6Cl 2, and Ni(NH 3) 6Cl 2 are investigated. All four metal ammine chlorides can be compacted to solid tablets with densities of at least 95% of the crystal density. This gives very high indirect hydrogen densities both gravimetrically and volumetrically. Upon heating, NH 3 is released from the salts, and by employing an appropriate catalyst, H 2 can be released corresponding to up to 9.78 wt % H and 0.116 kg H/L for the Ca(NH 3) 8Cl 2 salt. The NH 3 release from all four salts is investigated using temperature-programmed desorption employing different heating rates. The desorption is found mainly to be limited by heat transfer, indicating that the desorption kinetics are extremely fast for all steps. During desorption from solid tablets of Mg(NH 3) 6Cl 2, Mn(NH 3) 6Cl 2, and Ni(NH 3) 6Cl 2, nanoporous structures develop, which facilitates desorption from the interior of large, compact tablets. Density functional theory calculations reproduce trends in desorption enthalpies for the systems studied, and a mechanism in which individual chains of the ammines are released from the surface of the crystal is proposed to explain the fast absorption/desorption processes.

The possibilities for designing new materials based on quantum physics calculations are rapidly growing, but these design efforts lead to a significant increase in the amount of computational data created. The Computational Materials Repository (CMR) addresses this data challenge and provides a software infrastructure that supports the collection, storage, retrieval, analysis, and sharing of data produced by many electronic-structure simulators.