npj Computational Materials最新文献

High-entropy carbide ceramics (HECCs) commonly exhibit non-stoichiometric compositions and short-range order (SRO) arising from diverse elemental mixing. In this study, taking (TiZrHfNb)C as a representative HECC, we explore the coupling effects of SRO and carbon non-stoichiometry based on density-functional theory (DFT) and machine learning (ML). DFT results indicate that carbon non-stoichiometry is favored in Ti and Nb environments due to enhanced local atomic relaxation and charge transfer, which contribute to improved d-d bonding interactions. DFT-based Monte Carlo (MC) simulations further reveal a clustering tendency of Ti and Nb elements that compete with carbon non-stoichiometry formation. These local features are effectively captured by ML models, enabling rapid assessment of the interplay among carbon deficiency, SRO, and their influences on the mechanical properties of HECCs. This work elucidates the microscopic local properties responsible for the macroscopic behavior, offering key insights for designing HECCs through careful element selection and local chemistry control.

Machine learning potentials (MLPs) have become an indispensable tool in large-scale atomistic simulations. However, most MLPs today are trained on data computed using relatively cheap density functional theory (DFT) methods such as the Perdew-Burke-Ernzerhof (PBE) generalized gradient approximation (GGA) functional. While meta-GGAs such as the strongly constrained and appropriately normed (SCAN) functional have been shown to yield significantly improved descriptions of atomic interactions for diversely bonded systems, their higher computational cost remains an impediment to their use in MLP development. In this work, we outline a data-efficient multi-fidelity approach to constructing Materials 3-body Graph Network (M3GNet) interatomic potentials that integrate different levels of theory within a single model. Using silicon and water as examples, we show that a multi-fidelity M3GNet model trained on a combined dataset of low-fidelity GGA calculations with 10% of high-fidelity SCAN calculations can achieve accuracies comparable to a single-fidelity M3GNet model trained on a dataset comprising 8 × the number of SCAN calculations. This work provides a pathway to the development of high-fidelity MLPs in a cost-effective manner by leveraging existing low-fidelity datasets.

The piezoelectric materials enable the mutual conversion between mechanical and electrical energy, which drive a multi-billion dollar industry through their applications as sensors, actuators, and energy harvesters. The third-rank piezoelectric tensor is the core matrices for piezoelectric materials and their devices. However, the high costs of obtaining full piezoelectric tensor data through either experimental or computational methods make a significant challenge. Here, we propose an equivariant attention tensor graph neural network (EATGNN) that can identify crystal symmetry and remain independent of the reference frame, ultimately enabling the accurate prediction of the complete third-rank piezoelectric tensor. Especially, we perform an irreducible decomposition of the piezoelectric tensor into four irreducible representations to efficiently reserve the symmetry under group transformation operations. Our results further demonstrate that this model performs well in both bulk and two-dimensional materials. Finally, combining EATGNN with first-principles calculations, we discovered several potential high-performance piezoelectric materials.

Molecular and lattice vibrations are able to couple to the spin of electrons and lead to their relaxation and decoherence. Ab initio simulations have played a fundamental role in shaping our understanding of this process but further progress is hindered by their high computational cost. Here we present an accelerated computational framework based on machine-learning models for the prediction of molecular vibrations and spin-phonon coupling coefficients. We apply this method to three open-shell coordination compounds exhibiting long relaxation times and show that this approach achieves semi-to-full quantitative agreement with ab initio methods reducing the computational cost by about 80%. Moreover, we show that this framework naturally extends to molecular dynamics simulations, paving the way to the study of spin relaxation in condensed matter beyond simple equilibrium harmonic thermal baths.

Cluster expansions are commonly employed as surrogate models to link the electronic structure of an alloy to its finite-temperature properties. Using cluster expansions to model materials with several alloying elements is challenging due to a rapid increase in the number of fitting parameters and training set size. We introduce the embedded cluster expansion (eCE) formalism that enables the parameterization of accurate on-lattice surrogate models for alloys containing several chemical species. The eCE model simultaneously learns a low dimensional embedding of site basis functions along with the weights of an energy model. A prototypical senary alloy comprised of elements in groups 5 and 6 of the periodic table is used to demonstrate that eCE models can accurately reproduce ordering energetics of complex alloys without a significant increase in model complexity. Further, eCE models can leverage similarities between chemical elements to efficiently extrapolate into compositional spaces that are not explicitly included in the training dataset. The eCE formalism presented in this study unlocks the possibility of employing cluster expansion models to study multicomponent alloys containing several alloying elements.

Large language models, commonly known as LLMs, are showing promise in tacking some of the most complex tasks in AI. In this perspective, we review the wider field of foundation models—of which LLMs are a component—and their application to the field of materials discovery. In addition to the current state of the art—including applications to property prediction, synthesis planning and molecular generation—we also take a look to the future, and posit how new methods of data capture, and indeed modalities of data, will influence the direction of this emerging field.

Virtual melting (VM) as alternative deformation and stress relaxation mechanisms under extreme load is directly validated by molecular dynamics (MD) simulations of the simple shear of single crystal Si I at a temperature 1383 K below the melting temperature. The shear band consisting of liquid Si is formed immediately after the shear instability while stresses drop to zero. This process is independent of the applied shear rate. A new thermodynamic approach is developed, and the thermodynamic criterion for VM, which depends on the ratio of the sample to shear band widths, is derived analytically and confirmed by MD simulations. Since stress-free melt is unstable at 300 K, with further shear, the VM immediately transforms to a mixture of low-density amorphous a-Si, stable Si I, and metastable Si IV. Cyclic transformations between a-Si ↔ Si I, a-Si ↔ Si IV, and Si I ↔ Si IV with volume fraction of all phases mostly between 0.2 and 0.4 and non-repeatable nanostructure evolution are reveled. Such cyclic transformations produce additional important carriers for plastic deformation through transformation strain and transformation-induced plasticity due to volume change, which may occur in shear bands in various material systems but missed in experiments and simulations. The release of shear stresses quenches the microstructure, and shows reasonable qualitative correspondence with existing experiments.

The quest for efficient and robust deep learning models for molecular systems representation is increasingly critical in scientific exploration. The advent of message passing neural networks has marked a transformative era in graph-based learning, particularly in the realm of predicting chemical properties and expediting molecular dynamics studies. We present the Moment Graph Neural Network (MGNN), a rotation-invariant message passing neural network architecture that capitalizes on the moment representation learning of 3D molecular graphs, is adept at capturing the nuanced spatial relationships inherent in three-dimensional molecular structures. From benchmark tests on public datasets, MGNN delivers multiple state-of-the-art results on QM9, revised MD17 and MD17-ethanol. Its generalizability and efficiency are also tested in additional systems including 3BPA and 25-element high-entropy alloys. The prowess of MGNN also extends to dynamic simulations, accurately predicting the structural and kinetic properties of complex systems such as amorphous electrolytes, with results that closely align with those from ab-initio simulations. The application of MGNN to the simulation of molecular spectra exemplifies its potential to offer a promising alternative to traditional electronic structure methods.

The design and high-throughput screening of materials using machine-learning assisted quantum-mechanical simulations typically requires the existence of a very large data set, often generated from simulations at a high level of theory or fidelity. A single simulation at high fidelity can take on the order of days for a complex molecule. Thus, although machine learning surrogate simulations seem promising at first glance, generation of the training data can defeat the original purpose. For this reason, the use of machine learning to screen or design materials remains elusive for many important applications. In this paper we introduce a new multi-fidelity approach based on a dual graph embedding to extract features that are placed inside a nonlinear multi-step autoregressive model. Experiments on five benchmark problems, with 14 different quantities and 27 different levels of theory, demonstrate the generalizability and high accuracy of the approach. It typically requires a few 10s to a few 1000’s of high-fidelity training points, which is several orders of magnitude lower than direct ML methods, and can be up to two orders of magnitude lower than other multi-fidelity methods. Furthermore, we develop a new benchmark data set for 860 benzoquinone molecules with up to 14 atoms, containing energy, HOMO, LUMO and dipole moment values at four levels of theory, up to coupled cluster with singles and doubles.

Despite extensive research on magnetic skyrmions and antiskyrmions, a significant challenge remains in crafting nontrivial high-order skyrmionic textures with varying, or even tailor-made, topologies. We address this challenge, by focusing on a construction pathway of skyrmionic metamaterials within a monolayer thin film and suggest several skyrmionic metamaterials that are surprisingly stable, i.e., long-lived, due to a self-stabilization mechanism. This makes these new textures promising for applications. Central to our approach is the concept of ’simulated controlled assembly’, in short, a protocol inspired by ’click chemistry’ that allows for positioning topological magnetic structures where one likes, and then allowing for energy minimization to elucidate the stability. Utilizing high-throughput atomistic-spin-dynamic simulations alongside state-of-the-art AI-driven tools, we have isolated skyrmions (topological charge Q = 1), antiskyrmions (Q = − 1), and skyrmionium (Q = 0). These entities serve as foundational ’skyrmionic building blocks’ to form the here-reported intricate textures. In this work, two key contributions are introduced to the field of skyrmionic systems. First, we present a novel combination of atomistic spin dynamics simulations and controlled assembly protocols for the stabilization and investigation of new topological magnets. Second, using the aforementioned methods we report on the discovery of skyrmionic metamaterials.