Summary The FE 2 method is a renown computational multiscale simulation technique for solid materials with fine‐scale microstructure. It allows the accurate prediction of mechanical behavior structures made heterogeneous nonlinear material behavior. However, leads to excessive CPU time and storage requirements, even academic two‐dimensional problems. In order allow realistic three‐dimensional two‐scale simulations, significant reduction memory usage required. For this purpose, authors have...

We present a protocol for automated fitting of magnetic moment tensor potential explicitly including moments in its functional form. For the this we use energies, forces, stresses, and forces (negative derivatives energies with respect to moments) configurations selected an active learning algorithm. These are computed using constrained density theory, which enables calculating their both equilibrium nonequilibrium (excited) states. test our on system B1-CrN demonstrate that automatically...

We propose a machine-learning interatomic potential for multi-component magnetic materials. In this we consider moments as degrees of freedom (features) along with atomic positions, types, and lattice vectors. create training set constrained DFT (cDFT) that allows us to calculate energies configurations non-equilibrium (excited) and, thus, it is possible construct the in wide configuration space great variety moments, Such makes fit reliable potentials will allow predict properties excited...

Segregation of solutes to grain boundaries (GBs) is an important process having a large impact on mechanical properties metallic alloys. In this work, we show how accurate density functional theory (DFT) calculations can be combined with machine learning methods obtain reliable GB segregation energies significantly lower computational efforts compared full ab initio approach. First compare various descriptor sets respect their efficiency in predicting for arbitrary types. Second, demonstrate...

While first-principles methods have been successfully applied to characterize individual properties of multi-principal element alloys (MPEA), their use in searching for optimal trade-offs between competing is hampered by high computational demands. In this work, we present a framework explore Pareto-optimal compositions integrating advanced ab initio-based techniques into Bayesian multi-objective optimization workflow, complemented simple analytical model providing straightforward analysis...

A well-known drawback of state-of-the-art machine-learning interatomic potentials is their poor ability to extrapolate beyond the training domain. For small-scale problems with tens hundreds atoms this can be solved by using active learning which able select atomic configurations on a potential attempts extrapolation and add them ab initio-computed set. In sense an algorithm viewed as on-the-fly interpolation initio model. large-scale problems, possibly involving thousands atoms, not...

Abstract Machine-learning interatomic potentials (MLIPs) have made a significant contribution to the recent progress in fields of computational materials and chemistry due MLIPs’ ability accurately approximating energy landscapes quantum-mechanical models while being orders magnitude more computationally efficient. However, cost number parameters many state-of-the-art MLIPs increases exponentially with atomic features. Tensor (non-neural) networks, based on low-rank representations...

We present an automated procedure for computing stacking fault energies in random alloys from large-scale simulations using moment tensor potentials (MTPs) with the accuracy of density functional theory (DFT). To that end, we develop algorithm training MTPs on alloys. In first step, our constructs a set $\ensuremath{\sim}10$ 000 or more candidate configurations 50--100 atoms are representative atomic neighborhoods occurring simulation. second use active learning to reduce this...

In a previous work [M. Hodapp and A. Shapeev, Mach. Learn.: Sci. Technol. 1, 045005 (2020)], we proposed an algorithm that fully automatically trains machine-learning interatomic potentials (MLIPs) during large-scale simulations, successfully applied it to simulate screw dislocation motion in body-centered-cubic tungsten. The identifies local subregions of the simulation region where potential extrapolates, then constructs periodic configurations 100--200 atoms out these nonperiodic can be...

We developed a method for fitting machine-learning interatomic potentials with magnetic degrees of freedom, namely, Moment Tensor Potentials (mMTP).The main feature our consists in mMTP to forces (negative derivatives energies respect moments) as obtained spin-polarized density functional theory calculations.We test on the bcc Fe-Al system different compositions.Specifically, we calculate formation energies, equilibrium lattice parameter, and total cell magnetization.Our findings demonstrate...

The "flexible boundary condition" method, introduced by Sinclair and coworkers in the 1970s, remains among most popular methods for simulating isolated two-dimensional crystalline defects, embedded an effectively infinite atomistic domain. In essence, method can be characterized as a domain decomposition which iterates between local anharmonic global harmonic problem, where latter is solved means of lattice Green function ideal crystal. This local/global splitting gives rise to tremendously...

Machine-learning interatomic potentials (MLIPs) have made a significant contribution to the recent progress in fields of computational materials and chemistry due MLIPs' ability accurately approximating energy landscapes quantum-mechanical models while being orders magnitude more computationally efficient. However, cost number parameters many state-of-the-art MLIPs increases exponentially with atomic features. Tensor (non-neural) networks, based on low-rank representations high-dimensional...

We developed a method for fitting machine-learning interatomic potentials with magnetic degrees of freedom, namely, Moment Tensor Potentials (mMTP). The main feature our consists in mMTP to forces (negative derivatives energies respect moments) as obtained spin-polarized density functional theory calculations. test on the bcc Fe-Al system different compositions. Specifically, we calculate formation energies, equilibrium lattice parameter, and total cell magnetization. Our findings...

Understanding the physical origin of deformation mechanisms in random alloys requires an understanding their average behavior and, equally important, role local fluctuations around average. Material properties can be computed using direct simulations on configurations but some are very difficult to compute, for others it is not even fully understood how compute them sampling, particular, interaction energies between multiple defects. To that end, we develop atomistic model does averaging...

Understanding the physical origin of mechanisms in random alloys that lead to formation microstructures requires an understanding their average behavior and, equally important, role local fluctuations around average.Material properties can be computed using direct simulations on configurations.However, some are very difficult compute, for others it is not even fully understood how compute them sampling, particular, interaction energies between multiple defects.To end, we develop atomistic...

We present a protocol for automated fitting of magnetic Moment Tensor Potential explicitly including moments in its functional form. For the this potential we use energies, forces, stresses, and forces (negative derivatives energies with respect to moments) configurations selected an active learning algorithm. These are computed using constrained density theory, which enables calculating their both equilibrium non-equilibrium (excited) states. test our on system B1-CrN demonstrate that...

In (M Hodapp and A Shapeev 2020 Mach. Learn.: Sci. Technol. 1 045005), we have proposed an algorithm that fully automatically trains machine-learning interatomic potentials (MLIPs) during large-scale simulations, successfully applied it to simulate screw dislocation motion in body-centered cubic tungsten. The identifies local subregions of the simulation region where potential extrapolates, then constructs periodic configurations 100--200 atoms out these non-periodic can be efficiently...