Batteries, as complex materials systems, pose unique challenges for the application of machine learning. Although a shift to data-driven, machine learning-based battery research has started, new ...initiatives in academia and industry are needed to fully exploit its potential.
Bulk metallic glasses (BMGs) are a unique class of materials that are gaining traction in a wide variety of applications due to their attractive physical properties. One limitation to the wide-scale ...use of these materials is the lack of predictable tools for understanding the relationships between alloy composition and ideal properties. To address this issue, we developed a framework for designing metallic glasses using machine learning (ML) models that predict three key properties of candidate BMG compositions: ability to exist in an amorphous state, critical casting diameter (Dmax), and supercooled liquid range (ΔTx). Our models take only the composition of the alloy as input, and were created from a database of more than 8000 metallic glass experiments assembled from several dozen papers and handbooks. We employed these ML models to optimize the properties of existing commercial alloys and found, experimentally, several of our ML-predicted compositions can form glasses and exceed existing alloys in one of our two design variables, ΔTx.
Display omitted
The 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 the 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 the retrosynthesis of inorganics are also discussed.
The role of van der Waals (vdW) interactions in density functional theory (DFT) + U calculations of the layered lithium-ion battery cathode Li x CoO2 (x = 0–1) is investigated using (i) dispersion ...corrections in the Perdew–Burke–Ernzerhof (PBE) generalized gradient approximation functional, (ii) vdW density functionals, and (iii) the Bayesian error estimation functional with vdW correlation. We find that combining vdW corrections or functionals with DFT+U can yield lithiation voltages, relative stabilities, and structural properties that are in much better agreement with experiments for the phases O1-CoO2, O3-CoO2, layered-Li0.5CoO2, spinel-Li0.5CoO2, and LiCoO2 than using DFT+U or vdW-inclusive methods alone or using the hybrid Heyd–Scuseria–Ernzerhof functional. Contributions of vdW interactions to the lithiation voltages are estimated to have a similar magnitude with that of applying a typical U in the range 2–4 eV for cobalt, each accounting for 5–10% of calculated voltages relative to PBE. Relative stabilities of O1 and O3-CoO2 as well as layered- and spinel-Li0.5CoO2 are correctly predicted with vdW-inclusive methods combined with the +U correction.
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.
Metal oxide cathode coatings are capable of scavenging the hydrofluoric acid (HF) (present in LiPF6‐based electrolytes) and improving the electrochemical performance of Li‐ion batteries. Here, a ...first‐principles thermodynamic framework is introduced for designing cathode coatings that consists of four elements: i) HF‐scavenging enthalpies, ii) volumetric and iii) gravimetric HF‐scavenging capacities of the oxides, and iv) cyclable Li loss into coating components. 81 HF‐scavenging reactions involving binary s‐, p‐ and d‐block metal oxides and fluorides are enumerated and these materials are screened to find promising coatings based on attributes (i‐iv). The screen successfully produces known effective coating materials (e.g., Al2O3 and MgO), providing a validation of our framework. Using this design strategy, promising coating materials, such as trivalent oxides of d‐block transition metals Sc, Ti, V, Cr, Mn and Y, are predicted. Finally, a new protection mechanism that successful coating materials could provide by scavenging the wide bandgap and low Li ion conductivity LiF precipitates from the cathode surfaces is suggested.
Functional electrode coatings are critical for the enhancement of electrochemical performance of Li‐ion batteries. A materials screening approach to find novel HF‐scavenging lithium‐ion battery cathode coatings based on first‐principles thermodynamics is presented. This coating design approach can aid experimental efforts by allowing materials screening prior to actual battery production and by predicting effective coatings.
Realizing the growing number of possible or hypothesized metastable crystalline materials is extremely challenging. There is no rigorous metric to identify which compounds can or cannot be ...synthesized. We present a thermodynamic upper limit on the energy scale, above which the laboratory synthesis of a polymorph is highly unlikely. The limit is defined on the basis of the amorphous state, and we validate its utility by effectively classifying more than 700 polymorphs in 41 common inorganic material systems in the Materials Project for synthesizability. The amorphous limit is highly chemistry-dependent and is found to be in complete agreement with our knowledge of existing polymorphs in these 41 systems, whether made by the nature or in a laboratory. Quantifying the limits of metastability for realizable compounds, the approach is expected to find major applications in materials discovery.
We introduce atomate, an open-source Python framework for computational materials science simulation, analysis, and design with an emphasis on automation and extensibility. Built on top of open ...source Python packages already in use by the materials community such as pymatgen, FireWorks, and custodian, atomate provides well-tested workflow templates to compute various materials properties such as electronic bandstructure, elastic properties, and piezoelectric, dielectric, and ferroelectric properties. Atomate also enables the computational characterization of materials by providing workflows that calculate X-ray absorption (XAS), Electron energy loss (EELS) and Raman spectra. One of the major features of atomate is that it provides both fully functional workflows as well as reusable components that enable one to compose complex materials science workflows that use a diverse set of computational tools. Additionally, atomate creates output databases that organize the results from individual calculations and contains a builder framework that creates summary reports for each computed material based on multiple simulations.
Sequential learning (SL) strategies,
i.e.
iteratively updating a machine learning model to guide experiments, have been proposed to significantly accelerate materials discovery and research. ...Applications on computational datasets and a handful of optimization experiments have demonstrated the promise of SL, motivating a quantitative evaluation of its ability to accelerate materials discovery, specifically in the case of physical experiments. The benchmarking effort in the present work quantifies the performance of SL algorithms with respect to a breadth of research goals: discovery of any "good" material, discovery of all "good" materials, and discovery of a model that accurately predicts the performance of new materials. To benchmark the effectiveness of different machine learning models against these goals, we use datasets in which the performance of all materials in the search space is known from high-throughput synthesis and electrochemistry experiments. Each dataset contains all pseudo-quaternary metal oxide combinations from a set of six elements (chemical space), the performance metric chosen is the electrocatalytic activity (overpotential) for the oxygen evolution reaction (OER). A diverse set of SL schemes is tested on four chemical spaces, each containing 2121 catalysts. The presented work suggests that research can be accelerated by up to a factor of 20 compared to random acquisition in specific scenarios. The results also show that certain choices of SL models are ill-suited for a given research goal resulting in substantial deceleration compared to random acquisition methods. The results provide quantitative guidance on how to tune an SL strategy for a given research goal and demonstrate the need for a new generation of materials-aware SL algorithms to further accelerate materials discovery.
Benchmarking metrics for materials discovery
via
sequential learning are presented, to assess the efficacy of existing algorithms and to be scientific in our assessment of accelerated science.