Search

In this work, we present ænet-PyTorch, a PyTorch-based implementation for training artificial neural network-based machine learning interatomic potentials. Developed as an extension of the atomic energy network (ænet), ænet-PyTorch provides access to all the tools included in ænet for the application and usage of the potentials. The package has been designed as an alternative to the internal training capabilities of ænet, leveraging the power of graphic processing units to facilitate direct training on forces in addition to energies. This leads to a substantial reduction of the training time by one to two orders of magnitude compared to the central processing unit implementation, enabling direct training on forces for systems beyond small molecules. Here, we demonstrate the main features of ænet-PyTorch and show its performance on open databases. Our results show that training on all the force information within a dataset is not necessary, and including between 10% and 20% of the force information is sufficient to achieve optimally accurate interatomic potentials with the least computational resources. [ABSTRACT FROM AUTHOR]

Machine-learning potentials (MLPs) trained on data from quantum-mechanics based first-principles methods can approach the accuracy of the reference method at a fraction of the computational cost. To facilitate efficient MLP-based molecular dynamics and Monte Carlo simulations, an integration of the MLPs with sampling software is needed. Here, we develop two interfaces that link the atomic energy network (ænet) MLP package with the popular sampling packages TINKER and LAMMPS. The three packages, ænet, TINKER, and LAMMPS, are free and open-source software that enable, in combination, accurate simulations of large and complex systems with low computational cost that scales linearly with the number of atoms. Scaling tests show that the parallel efficiency of the ænet–TINKER interface is nearly optimal but is limited to shared-memory systems. The ænet–LAMMPS interface achieves excellent parallel efficiency on highly parallel distributed-memory systems and benefits from the highly optimized neighbor list implemented in LAMMPS. We demonstrate the utility of the two MLP interfaces for two relevant example applications: the investigation of diffusion phenomena in liquid water and the equilibration of nanostructured amorphous battery materials. [ABSTRACT FROM AUTHOR]

X-ray absorption spectroscopy (XAS) is a premier technique for materials characterization, providing key information about the local chemical environment of the absorber atom. In this work, we develop a database of sulfur K-edge XAS spectra of crystalline and amorphous lithium thiophosphate materials based on the atomic structures reported in Chem. Mater., 34, 6702 (2022). The XAS database is based on simulations using the excited electron and core-hole pseudopotential approach implemented in the Vienna Ab initio Simulation Package. Our database contains 2681 S K-edge XAS spectra for 66 crystalline and glassy structure models, making it the largest collection of first-principles computational XAS spectra for glass/ceramic lithium thiophosphates to date. This database can be used to correlate S spectral features with distinct S species based on their local coordination and short-range ordering in sulfide-based solid electrolytes. The data is openly distributed via the Materials Cloud, allowing researchers to access it for free and use it for further analysis, such as spectral fingerprinting, matching with experiments, and developing machine learning models. [ABSTRACT FROM AUTHOR]

The atomistic modeling of amorphous materials requires structure sizes and sampling statistics that are challenging to achieve with first-principles methods. Here, we propose a methodology to speed up the sampling of amorphous and disordered materials using a combination of a genetic algorithm and a specialized machine-learning potential based on artificial neural networks (ANNs). We show for the example of the amorphous LiSi alloy that around 1000 first-principles calculations are sufficient for the ANN-potential assisted sampling of low-energy atomic configurations in the entire amorphous LixSi phase space. The obtained phase diagram is validated by comparison with the results from an extensive sampling of LixSi configurations using molecular dynamics simulations and a general ANN potential trained to ∼45 000 first-principles calculations. This demonstrates the utility of the approach for the first-principles modeling of amorphous materials. [ABSTRACT FROM AUTHOR]

Fluorine substitution is a critical enabler for improving the cycle life and energy density of disordered rocksalt (DRX) Li‐ion battery cathode materials which offer prospects for high energy density cathodes, without the reliance on limited mineral resources. Due to the strong Li–F interaction, fluorine also is expected to modify the short‐range cation order in these materials which is critical for Li‐ion transport. In this work, density functional theory and Monte Carlo simulations are combined to investigate the impact of Li–F short‐range ordering on the formation of Li percolation and diffusion in DRX materials. The modeling reveals that F substitution is always beneficial at sufficiently high concentrations and can, surprisingly, even facilitate percolation in compounds without Li excess, giving them the ability to incorporate more transition metal redox capacity and thereby higher energy density. It is found that for F levels below 15%, its effect can be beneficial or disadvantageous depending on the intrinsic short‐range order in the unfluorinated oxide, while for high fluorination levels the effects are always beneficial. Using extensive simulations, a map is also presented showing the trade‐off between transition‐metal capacity, Li‐transport, and synthetic accessibility, and two of the more extreme predictions are experimentally confirmed. [ABSTRACT FROM AUTHOR]

An accurate determination of the potential energy is the crucial step in computer simulations of chemical processes, but using electronic structure methods on-the-fly in molecular dynamics (MD) is computationally too demanding for many systems. Constructing more efficient interatomic potentials becomes intricate with increasing dimensionality of the potential-energy surface (PES), and for numerous systems the accuracy that can be achieved is still not satisfying and far from the reliability of first-principles calculations. Feed-forward neural networks (NNs) have a very flexible functional form, and in recent years they have been shown to be an accurate tool to construct efficient PESs. High-dimensional NN potentials based on environment-dependent atomic energy contributions have been presented for a number of materials. Still, these potentials may be improved by a more detailed structural description, e.g., in form of atom pairs, which directly reflect the atomic interactions and take the chemical environment into account. We present an implementation of an NN method based on atom pairs, and its accuracy and performance are compared to the atom-based NN approach using two very different systems, the methanol molecule and metallic copper. We find that both types of NN potentials provide an excellent description of both PESs, with the pair-based method yielding a slightly higher accuracy making it a competitive alternative for addressing complex systems in MD simulations. [ABSTRACT FROM AUTHOR]

Machine learning interpolation of atomic potential energy surfaces enables the nearly automatic construction of highly accurate atomic interaction potentials. Here we discuss the Behler–Parrinello approach that is based on artificial neural networks (ANNs) and detail the implementation of the method in the free and open-source a tomic e nergy net work (æ net ) package. The construction and application of ANN potentials using æ net is demonstrated at the example of titanium dioxide (TiO 2 ), an industrially relevant and well-studied material. We show that the accuracy of lattice parameters, energies, and bulk moduli predicted by the resulting TiO 2 ANN potential is excellent for the reference phases that were used in its construction (rutile, anatase, and brookite) and examine the potential’s capabilities for the prediction of the high-pressure phases columbite (α-PbO 2 structure) and baddeleyite (ZrO 2 structure). [ABSTRACT FROM AUTHOR]

Nanoalloys are a promising class of (electro-) catalysts for applications in sustainable energy technologies, such as carbon dioxide conversion and proton exchange membrane fuel cells. In this article, we establish a methodology for the prediction of the composition and atomic ordering of alloy nanoparticles in thermal equilibrium. The approach is based on a combination of site-based Monte-Carlo simulations to sample the composition space and simultaneous molecular dynamics simulations to sample the structure space at set temperature and chemical potentials. To provide an accurate description of the atomic interactions that allows the required extensive sampling of the configurational space, we employ artificial neural network potentials to interpolate density-functional theory reference calculations. For the example of Cu–Au nanoalloys with 1.0 nm and 1.5 nm diameters, the equilibrium compositions and structures at 300 K and 500 K are compared to the ground state configurations. Consistent with previous reports, we find that the most stable structures exhibit Cu(core)–Au(shell) configurations. However, we observe a temperature dependence of the alloy composition that gives rise to some Au concentration within the particle core at greater temperatures. Therefore, we argue that equilibrium nanoalloy configurations should be used as starting point for the computational investigation of the catalytic activity of alloy nanoparticles. [ABSTRACT FROM AUTHOR]

The design of catalysts for CO2 reduction is challenging because of the fundamental relationships between the binding energies of the reaction intermediates. Metal carbides have shown promise for transcending these relationships and enabling low-cost alternatives. Herein, we show that directional bonding arising from the mixed covalent/metallic character plays a critical role in governing the surface chemistry. This behavior can be described by consideration of individual d-band components. We use this model to predict efficient catalysts based on tungsten carbide with a sub-monolayer of iron adatoms. Our approach can be used to predict site-preference and binding-energy trends for complex catalyst surfaces. [ABSTRACT FROM AUTHOR]

The shape, size, and compositionof catalyst nanoparticles canhave a significant influence on catalytic activity. Understandingsuch structure–reactivity relationships is crucial for theoptimization of industrial catalysts and the design of novel catalystswith enhanced properties. In this letter, we employ a combinationof first-principles computations and large-scale Monte-Carlo simulationswith highly accurate neural network potentials to study the equilibriumsurface structure and composition of bimetallic Au/Cu nanoparticles(NPs), which have recently been of interest as stable and efficientCO2reduction catalysts. We demonstrate that the inclusionof explicit water molecules at a first-principles level of accuracyis necessary to predict experimentally observed trends in Au/Cu NPsurface composition; in particular, we find that Au-coated core–shellNPs are thermodynamically favored in vacuum, independent of Au/Cuchemical potential and NP size, while NPs with mixed Au–Cusurfaces are preferred in aqueous solution. Furthermore, we show thatboth CO and O2adsorption energies differ significantlyfor NPs with the equilibrium surface composition found in water andthose with the equilibrium surface composition found in vacuum, suggestinglarge changes in CO2reduction activity. Our results emphasizethe importance of understanding and being able to predict the effectsof catalytic environment on catalyst structure and activity. In addition,they demonstrate that first-principles-based neural network potentialsprovide a promising approach for accurately investigating the relationshipsbetween solvent, surface composition and morphology, surface electronicstructure, and catalytic activity in systems composed of thousandsof atoms. [ABSTRACT FROM AUTHOR]

The development of reliable interatomic potentials for large-scale molecular dynamics (MD) simulations of chemical processes at surfaces and interfaces is a formidable challenge because a wide range of atomic environments and very different types of bonding can be present. In recent years interatomic potentials based on artificial neural networks (NNs) have emerged offering an unbiased approach to the construction of potential energy surfaces (PESs) for systems that are difficult to describe by conventional potentials. Here, we review the basic properties of NN potentials and describe their construction for materials like metals and oxides. The accuracy and efficiency are demonstrated using copper and zinc oxide as benchmark systems. First results for a potential of the combined ternary CuZnO system aiming at the description of oxide-supported copper clusters are reported. [ABSTRACT FROM AUTHOR]

Artificial neural networks represent an accurate and efficient tool to construct high-dimensional potential-energy surfaces based on first-principles data. However, so far the main drawback of this method has been the limitation to a single atomic species. We present a generalization to compounds of arbitrary chemical composition, which now enables simulations of a wide range of systems containing large numbers of atoms. The required incorporation of long-range interactions is achieved by combining the numerical accuracy of neural networks with an electrostatic term based on environment-dependent charges. Using zinc oxide as a benchmark system we show that the neural network potential-energy surface is in excellent agreement with density-functional theory reference calculations, while the evaluation is many orders of magnitude faster. [ABSTRACT FROM AUTHOR]

Solid‐state proton conductors with high conductivity at intermediate temperatures (100–300 °C) have recently gained attention due to their potential for a wide range of new electrochemical applications. The proton conductivity of cesium hydroxide hydrate (CsOH.H2O) is reported on, a superprotonic conductor whose ionic conductivity is hitherto unreported. Via a thermal treatment, the hexagonal and tetragonal CsOH.H2O phases are prepared and the effect of the degree of hydration (H2O content) on the structure and proton conductivity is investigated. It is shown that the temperature in the thermal treatment has an enormous influence on the structure and water content, and consequently, the proton transport. The conductivity of the hexagonal and tetragonal phases exceeds 3 × 103 S cm−1 at 30 °C and 10−2 S cm−1 at 120 °C, making them potentially suitable for both low‐ and intermediate‐temperature electrochemical applications. 1H NMR spin‐lattice relaxation rate measurements reveal fast dynamic processes with rates reaching the GHz regime, resulting in diffusion coefficients as high as 1.3 × 10−11 m2 s−1 at 60 °C, in good agreement with the proton conductivity results derived from impedance spectroscopy. These findings will inspire the development of novel inorganic solid proton conductors based on hydroxide hydrates. [ABSTRACT FROM AUTHOR]

Machine-learning potentials (MLPs) for atomistic simulations are a promising alternative to conventional classical potentials. Current approaches rely on descriptors of the local atomic environment with dimensions that increase quadratically with the number of chemical species. In this paper, we demonstrate that such a scaling can be avoided in practice. We show that a mathematically simple and computationally efficient descriptor with constant complexity is sufficient to represent transition-metal oxide compositions and biomolecules containing 11 chemical species with a precision of around 3 meV/atom. This insight removes a perceived bound on the utility of MLPs and paves the way to investigate the physics of previously inaccessible materials with more than ten chemical species. [ABSTRACT FROM AUTHOR]

The development of reliable interatomic potentials for large‐scale molecular dynamics simulations of chemical processes at surfaces and interfaces is a formidable challenge because a wide range of atomic environments and very different types of bonding can be present. In recent years interatomic potentials based on artificial neural networks (NN) have emerged offering an unbiased approach to the construction of potential energy surfaces for systems that are difficult to describe by conventional potentials. Artrith, Hiller and Behler (pp. 1191–1203) review the basic properties of NN potentials and describe their construction for materials like metals and oxides. The accuracy and efficiency are demonstrated using copper and zinc oxide as benchmark systems. First results for a potential of the combined ternary CuZnO system aiming at the description of oxide‐supported copper clusters are reported. [ABSTRACT FROM AUTHOR]

Cation disorder is an important design criterion for technologically relevant transition-metal (TM) oxides, such as radiation-tolerant ceramics and Li-ion battery electrodes. In this Letter, we use a combination of first-principles calculations, normal mode analysis, and band-structure arguments to pinpoint a specific electronic-structure effect that influences the stability of disordered phases. We find that the electronic configuration of a TM ion determines to what extent the structural energy is affected by site distortions. This mechanism explains the stability of disordered phases with large ionic radius differences and provides a concrete guideline for the discovery of novel disordered compositions. [ABSTRACT FROM AUTHOR]