Machine Learning Interatomic Potentials Research Articles

Building a DFT+U Machine Learning Interatomic Potential for Uranium Dioxide

Despite uranium dioxide (UO2) being a widely used nuclear fuel, fuel performance models rely extensively on empirical correlations of material behavior, leveraging UO2 historical operating experience. Mechanistic models that consider an atomistic understanding of the processes governing fuel performance (such as fission gas release and creep) will enable a better description of fuel behavior under non-prototypical conditions such as in new reactor concepts or for modified UO2 fuel compositions. To this end, molecular dynamics simulation is a powerful tool for rapidly predicting physical properties of proposed fuel candidates. However, the reliability of these simulations depends largely on the accuracy of the atomic forces. Traditionally, these forces are computed using either a classical force field (FF) or density functional theory (DFT). While DFT is relatively accurate, the computational cost is burdensome, especially for f-electron elements, such as actinides. By contrast, classical FFs are computationally efficient but are less accurate. For these reasons, we report a new accurate machine learning interatomic potential (MLIP) for UO2 that provides high-fidelity reproduction of DFT forces at a similar low cost to classical FFs. We employ an active learning approach that autonomously augments the DFT training data set to iteratively refine the MLIP. To further improve the quality of our predictions, we utilize transfer learning to retrain our MLIP to higher-accuracy DFT+U data. We validate our MLIPs by comparing predicted physical properties (e.g., thermal expansion and elastic properties) with those from existing classical FFs and DFT/DFT+U calculations, as well as with experimental data when available.

Thermal transport in disordered wurtzite ScAlN alloys using machine learning interatomic potentials

Disordered wurtzite ScAlN alloy is an emerging material with a strong piezoelectric coefficient that promotes its application in 5G communication devices. To extend the service lifetime of the device, it is an urgent task to understand the thermal transport properties of this alloy for thermal management. Yet a thorough understanding of the thermal transport in the disordered ScAlN alloy is hindered by the inaccuracy of empirical interatomic potentials adopted for atomistic simulations. Herein, we developed an accurate and efficient neuro-evolution potential and combined it with Allen and Feldman theory and molecular dynamics to investigate thermal transport properties of disordered ScAlN alloys with varying ratios. This potential is shown to have the same computational accuracy as first principles by comparing the phonon spectrum and radial distribution function. Based on the homogeneous non-equilibrium molecular dynamics simulations, the lattice thermal conductivity of the disordered alloys reaches its maximum value (less than 12W/m·K) at room temperature. Further investigation of the temperature effects on the vibrational modes shows that it affects the mode lifetimes for the propagons and the mode heat capacities for the diffusions. In addition, it was found that both mode lifetime and diffusivity decrease with increasing Sc concentration. This work aims to train a universal machine learning interatomic potential of disordered ScAlN alloys, seeking to accelerate the thermal management design of related devices.

Experiment-Driven Atomistic Materials Modeling: A Case Study Combining X-Ray Photoelectron Spectroscopy and Machine Learning Potentials to Infer the Structure of Oxygen-Rich Amorphous Carbon.

An important yet challenging aspect of atomistic materials modeling is reconciling experimental and computational results. Conventional approaches involve generating numerous configurations through molecular dynamics or Monte Carlo structure optimization and selecting the one with the closest match to experiment. However, this inefficient process is not guaranteed to succeed. We introduce a general method to combine atomistic machine learning (ML) with experimental observables that produces atomistic structures compatible with experiment by design. We use this approach in combination with grand-canonical Monte Carlo within a modified Hamiltonian formalism, to generate configurations that agree with experimental data and are chemically sound (low in energy). We apply our approach to understand the atomistic structure of oxygenated amorphous carbon (a-COx), an intriguing carbon-based material, to answer the question of how much oxygen can be added to carbon before it fully decomposes into CO and CO2. Utilizing an ML-based X-ray photoelectron spectroscopy (XPS) model trained from GW and density functional theory (DFT) data, in conjunction with an ML interatomic potential, we identify a-COx structures compliant with experimental XPS predictions that are also energetically favorable with respect to DFT. Employing a network analysis, we accurately deconvolve the XPS spectrum into motif contributions, both revealing the inaccuracies inherent to experimental XPS interpretation and granting us atomistic insight into the structure of a-COx. This method generalizes to multiple experimental observables and allows for the elucidation of the atomistic structure of materials directly from experimental data, thereby enabling experiment-driven materials modeling with a degree of realism previously out of reach.

Interatomic Interaction Models for Magnetic Materials: Recent Advances

Interatomic Interaction Models for Magnetic Materials: Recent Advances

Two-Step Growth Mechanism of the Solid Electrolyte Interphase in Argyrodyte/Li-Metal Contacts.

The structure and growth of the solid electrolyte interphase (SEI) region between an electrolyte and an electrode is one of the most fundamental yet less well-understood phenomena in solid-state batteries. We present an atomistic simulation of the SEI growth for one of the currently promising solid electrolytes (Li6PS5Cl), based on ab initio-trained machine learning interatomic potentials, for over 30,000 atoms during 10 ns, well beyond the capabilities of conventional molecular dynamics. This unveils a two-step growth mechanism: a Li-argyrodite chemical reaction leading to the formation of an amorphous phase, followed by a kinetically slower crystallization of the reaction products into a 5Li2S·Li3P·LiCl solid solution. The simulation results support the recent, experimentally founded hypothesis of an indirect pathway of electrolyte reduction. These findings shed light on the intricate processes governing SEI evolution, providing a valuable foundation for the design and optimization of next-generation solid-state batteries.

Machine-Learning-Driven High-Throughput Screening for High-Energy Density and Stable NASICON Cathodes.

The Na super ionic conductor (NASICON), which has outstanding structural stability and a high operating voltage, is an appealing material for overcoming the limits of low specific energy and larger volume distortion of sodium-ion batteries. In this study, to discover ideal NASICON cathode materials, a screening platform based on density functional theory (DFT) calculations and machine learning (ML) is developed. A training database was generated utilizing the previous 124 545 electrode databases, and a test set of 3126 potential NASICON structures [NaxMyM'1-y(PO4)3] with 27 dopants at the metal site and 6 dopants at the polyanion central site was constructed. The developed ML surrogate model identifies 796 materials that satisfy the following criteria: formation energy of <0.0 eV/atom, energy above hull of ≤0.025 eV/atom, volume change of ≤4%, and theoretical capacity of ≥50 mAh/g. The thermodynamically stable configurations of doped NASICON structures were then selected using machine learning interatomic potential (MLIP), enabling rapid consideration of various dopant site configurations. DFT calculations are followed on 796 screened materials to obtain energy density, average voltage, and volume change. Finally, 50 candidates with an average voltage of ≥3.5 V are identified. The suggested platform accelerates the exploration for optimal NASICON materials by narrowing the focus on materials with desired properties, saving considerable resources.

Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic potentials

Efficiently creating a concise but comprehensive data set for training machine-learned interatomic potentials (MLIPs) is an under-explored problem. Active learning, which uses biased or unbiased molecular dynamics (MD) to generate candidate pools, aims to address this objective. Existing biased and unbiased MD-simulation methods, however, are prone to miss either rare events or extrapolative regions—areas of the configurational space where unreliable predictions are made. This work demonstrates that MD, when biased by the MLIP’s energy uncertainty, simultaneously captures extrapolative regions and rare events, which is crucial for developing uniformly accurate MLIPs. Furthermore, exploiting automatic differentiation, we enhance bias-forces-driven MD with the concept of bias stress. We employ calibrated gradient-based uncertainties to yield MLIPs with similar or, sometimes, better accuracy than ensemble-based methods at a lower computational cost. Finally, we apply uncertainty-biased MD to alanine dipeptide and MIL-53(Al), generating MLIPs that represent both configurational spaces more accurately than models trained with conventional MD.

Plutonium oxide melt structure and covalency.

Advances in nuclear power reactors include the use of mixed oxide fuel, containing uranium and plutonium oxides. The high-temperature behaviour and structure of PuO2-x above 1,800 K remain largely unexplored, and these conditions must be considered for reactor design and planning for the mitigation of severe accidents. Here, we measure the atomic structure of PuO2-x through the melting transition up to 3,000 ± 50 K using X-ray scattering of aerodynamically levitated and laser-beam-heated samples, with O/Pu ranging from 1.57 to 1.76. Liquid structural models consistent with the X-ray data are developed using machine-learned interatomic potentials and density functional theory. Molten PuO1.76 contains some degree of covalent Pu-O bonding, signalled by the degeneracy of Pu 5f and O 2p orbitals. The liquid is isomorphous with molten CeO1.75, demonstrating the latter as a non-radioactive, non-toxic, structural surrogate when differences in the oxidation potentials of Pu and Ce are accounted for. These characterizations provide essential constraints for modelling pertinent to reactor safety design.

Global machine learning potentials for molecular crystals.

Molecular crystals are difficult to model with accurate first-principles methods due to large unit cells. On the other hand, accurate modeling is required as polymorphs often differ by only 1kJ/mol. Machine learning interatomic potentials promise to provide accuracy of the baseline first-principles methods with a cost lower by orders of magnitude. Using the existing databases of the density functional theory calculations for molecular crystals and molecules, we train global machine learning interatomic potentials, usable for any molecular crystal. We test the performance of the potentials on experimental benchmarks and show that they perform better than classical force fields and, in some cases, are comparable to the density functional theory calculations.

Do we really need machine learning interatomic potentials for modeling amorphous metal oxides? Case study on amorphous alumina by recycling an existing ab initio database

In this study, we critically evaluate the performance of various interatomic potentials/force fields against a benchmark ab initio database for bulk amorphous alumina. The interatomic potentials tested in this work include all major fixed charge and variable charge models developed to date for alumina. Additionally, we introduce a novel machine learning interatomic potential constructed using the NequIP framework based on graph neural networks. Our findings reveal that the fixed-charge potential developed by Matsui and coworkers offers the most optimal balance between computational efficiency and agreement with ab initio data for stoichiometric alumina. Such balance cannot be provided by machine learning potentials when comparing performance with Matsui potential on the same computing infrastructure using a single Graphical Processing Unit. For non-stoichiometric alumina, the variable charge potentials, in particular ReaxFF, exhibit an impressive concordance with density functional theory calculations. However, our NequIP potentials trained on a small fraction of the ab initio database easily surpass ReaxFF in terms of both accuracy and computational performance. This is achieved without large overhead in terms of potential fitting and fine-tuning, often associated with the classical potential development process as well as training of standard deep neural network potentials, thus advocating for the use of data-efficient machine learning potentials like NequIP for complex cases of non-stoichiometric amorphous oxides.

Generating Minimal Training Sets for Machine Learned Potentials.

This Letter presents a novel approach for identifying uncorrelated atomic configurations from extensive datasets with a nonstandard neural network workflow known as random network distillation (RND) for training machine-learned interatomic potentials (MLPs). This method is coupled with a DFT workflow wherein initial data are generated with cheaper classical methods before only the minimal subset is passed to a more computationally expensive abinitio calculation. This benefits training not only by reducing the number of expensive DFT calculations required but also by providing a pathway to the use of more accurate quantum mechanical calculations. The method's efficacy is demonstrated by constructing machine-learned interatomic potentials for the molten salts KCl and NaCl. Our RND method allows accurate models to be fit on minimal datasets, as small as 32 configurations, reducing the required structures by at least 1 order of magnitude compared to alternative methods. This reduction in dataset sizes not only substantially reduces computational overhead for training data generation but also provides a more comprehensive starting point for active-learning procedures.

Thermodynamic stability and ionic conductivity in lithium–germanium binary system

Lithium–germanium binary compounds are promising anode materials for secondary lithium-ion batteries due to their high capacity, low operating voltage, and high electronic conductivity of lithiated Ge. For their successful application in batteries, it is essential to know the temperature stability of different Li–Ge phases and the variation of their ionic conductivity depending on the operating temperatures of the batteries. This work aims to comprehensively study the thermodynamic stability and ionic conductivity in Li–Ge binary compounds using a combination of first-principle computations and machine-learning interatomic potentials. We calculated convex hulls of the Li–Ge system at various temperatures and a temperature–composition phase diagram was obtained, delineating stability fields of each phase. Our calculations show that at temperatures higher than 590 K, LiGe undergoes a I41/a–P4/mmm transition, which leads to a change in the ionic conductivity. We show that all stable and metastable Li–Ge compounds have high ionic conductivity, but LiGe and Li7Ge12 have the lowest lithium diffusion. Trajectories of diffusion and Ge arrangements depend on lithium concentration. Based on advanced theoretical approaches, this study provides insights for the development of Li–Ge materials in lithium-ion and lithium-metal battery applications.

Machine-Learned Potential Energy Surfaces for Free Sodium Clusters with Density Functional Accuracy: Applications to Melting

Gaussian Process Regression based Gaussian Approximation Potential has been used to develop machine learned interatomic potentials having density functional accuracy, for free sodium clusters. The training data was generated from a large sample of over 100,000 data points computed for clusters in the size range of N = 40–200, where N denotes the number of atoms in the cluster, using the density functional method as implemented in the VASP code. Two models have been developed, model M1 using data for N = 55 only, and model M2 using additional data from larger clusters. The models are intended for computing thermodynamic properties using molecular dynamics. Hence, particular attention has been paid to improve the fitting of the forces. Although it was possible to obtain a good fit using the data of Na55 only, additional data points from larger clusters were needed to get better accuracies in energies and forces for larger sizes. Surprisingly, the model M1 could be significantly improved by adding about 50 data points per cluster from the larger sizes. Thus, it turns out that the best fit can be obtained by carefully selecting a small number of data points viz. 1,900 and 1,300 configurations for the two models M1 and M2, respectively. These two models have been deployed to compute the heat capacities of Na55. The heat capacities of Na147 and about 40 isomers for larger clusters of sizes N = 147, 200, 201, and 252, have been obtained using the final model M2. There is an excellent agreement between the computed and experimentally measured melting temperatures. The geometries of these isomers have been further optimized by density functional theory. The mean absolute error with respect to DFT energies is found to be about 7 meV/atom or less. The errors in the lowest interatomic bond lengths are below 2% in almost all the cases.

A machine learning interatomic potential for high entropy alloys

High entropy alloys (HEAs) possess a vast compositional space, providing exciting prospects for tailoring material properties yet also presenting challenges in their rational design. Efficiently achieving a well-designed HEA often necessitates the aid of atomistic simulations, which rely on the availability of high-quality interatomic potentials. However, such potentials for most HEA systems are missing due to the complex interatomic interaction. To fundamentally resolve the challenge of the rational design of HEAs, we propose a strategy to build a machine learning (ML) interatomic potential for HEAs and demonstrate this strategy using CrFeCoNiPd as a model material. The fully trained ML model can achieve remarkable prediction precision (>0.92 R2) for atomic forces, comparable to the ab initio molecular dynamics (AIMD) simulations. To further validate the accuracy of the ML model, we implement the ML potential for CrFeCoNiPd in parallel molecular dynamics (MD) code. The MD simulations can predict the lattice constant (1 % error) and stacking fault energy (10 % error) of CrFeCoNiPd HEAs with high accuracy compared to experimental results. Through systematic MD simulations, for the first time, we reveal the atomic-scale deformation mechanisms associated with the stacking fault formation and dislocation cross-slips in CrFeCoNiPd HEAs under uniaxial compression, which are consistent with experimental observations. This study can help elucidate the underlying deformation mechanisms that govern the exceptional performance of CrFeCoNiPd HEAs. The strategy to establish ML interatomic potentials could accelerate the rational design of new HEAs with desirable properties.

Thermoelectric properties of lead-free anti-perovskites X3BN (B = Bi, Sb, X = Mg, Ca, Sr): A theoretical study based on first-principles calculations and machine learning interatomic potential

The search for non-toxic and low-cost perovskites with high thermoelectric performance is still a challenge despite low thermal conductivity. The thermoelectric properties of nitride anti-perovskites X3BN (B = Bi, Sb, X = Mg, Ca, Sr) with a cubic structure were investigated using density functional theory and machine learning interatomic potential. The low Debye temperature and thermal conductivity were obtained due to strong lattice anharmonicity, and the phonon vibration modes were also analyzed. The high band degeneracy and suitable bandgap lead to a large power factor. The maximum power factor is 7.54 mW/mk2 for Mg3BiN at 900 K. We obtained a maximum ZT of 1.49 for p-type Sr3BiN, and it is 1.22 for n-type doping at 900 K. The ZT for Mg3BiN is 1.18 and 1.19 for p-type and n-type doping, respectively. Our calculations reveal that these anti-perovskites are excellent materials for non-toxic and low-cost thermoelectric applications.

Neural network molecular dynamics study of LiGe2(PO4)3: Investigation of structure

Machine learning interatomic potential for LiGe2(PO4)3 in a form of a set of neural networks (DeePMD-model) was trained on DFT data. DFT simulations in GGA PBE approximation were performed for ordered and disordered LiGe2(PO4)3 supercell with 432 atoms which allowed the developed potential to describe crystalline, molten and glassy states. The developed DeePMD-potential was verified using literature and our experimental data demonstrating good agreement with it. Neural network molecular dynamic simulations indicate that P atoms have tetrahedral oxygen environment in both crystalline and glassy state. Coordination environment of Ge atoms is more complex. They have octahedral oxygen environment in crystalline state and mixed environment in glassy state: 4-, 5- and 6-coordinated Ge atoms were found. It was shown by calculation of orientational order parameters that in glassy LiGe2(PO4)3 4-coordinated Ge atoms have tetrahedral oxygen environment, while 6-coordinated Ge atoms have octahedral oxygen environment. Analysis of O-Ge-O angle distribution for 5-coordinated Ge atoms in glassy LiGe2(PO4)3 demonstrates that the environment of Ge atoms in this case is represented by various structures: square pyramids, trigonal bipyramids and transitional structures from 5-coordinated Ge atoms to tetrahedrons.

Dual-channel phonon transport leads to low thermal conductivity in pyrochlore La2Hf2O7

The pyrochlore oxide La2Hf2O7 (LHO) has attracted extensive attention in thermal management applications. Accurate determination and comprehensive understanding of thermal conductivity (κ) stand as imperative prerequisites for the effective utilization and manipulation of LHO in various applications. In this study, with the assistance of machine learning interatomic potential, we investigate the microscopic mechanisms of phonon thermal transport in LHO crystals. Our findings show that the thermal transport in LHO defies a precise explication through the conventional framework of the Boltzmann transport equation (BTE) due to the emergence of diffusive phonon modes, engendered by strong anharmonicity. Using the state-of-the-art dual-phonon model incorporating both propagating and diffusive modes, we successfully reproduce the experimentally measured results in both magnitude and temperature dependence. It is also shown that the thermal conduction in LHO crystals is dominated by the propagating modes at low temperatures while by the diffusive modes at high temperatures, which synergistically lead to a low κ for LHO crystals. The novel insights into phonon thermal transport provide a stepping stone for regulating the thermal transport properties of LHO crystals.

Transferable Water Potentials Using Equivariant Neural Networks.

Machine learning interatomic potentials (MLIPs) have emerged as a technique that promises quantum theory accuracy for reduced cost. It has been proposed [J. Chem. Phys. 2023, 158, 084111] that MLIPs trained on solely liquid water data cannot accurately transfer to the vapor-liquid equilibrium while recovering the many-body decomposition (MBD) analysis of gas-phase water clusters. This suggests that MLIPs do not directly learn the physically correct interactions of water molecules, limiting transferability. In this work, we show that MLIPs using equivariant architecture and trained on 3200 liquid water structures reproduces liquid-phase water properties (e.g., density within 0.003 g/cm3 between 230 and 365 K), vapor-liquid equilibrium properties up to 550 K, the MBD analysis of gas-phase water cluster up to six-body interactions, and the relative energy and the vibrational density of states of ice phases. We show that potentials developed using equivariant MLIPs allow transferability for arbitrary phases of water that remain stable in nanosecond long simulations.

Deep-potential enabled multiscale simulation of gallium nitride devices on boron arsenide cooling substrates

High-efficient heat dissipation plays critical role for high-power-density electronics. Experimental synthesis of ultrahigh thermal conductivity boron arsenide (BAs, 1300 W m−1K−1) cooling substrates into the wide-bandgap semiconductor of gallium nitride (GaN) devices has been realized. However, the lack of systematic analysis on the heat transfer across the GaN-BAs interface hampers the practical applications. In this study, by constructing the accurate and high-efficient machine learning interatomic potentials, we perform multiscale simulations of the GaN-BAs heterostructures. Ultrahigh interfacial thermal conductance of 260 MW m−2K−1 is achieved, which lies in the well-matched lattice vibrations of BAs and GaN. The strong temperature dependence of interfacial thermal conductance is found between 300 to 450 K. Moreover, the competition between grain size and boundary resistance is revealed with size increasing from 1 nm to 1000 μm. Such deep-potential equipped multiscale simulations not only promote the practical applications of BAs cooling substrates in electronics, but also offer approach for designing advanced thermal management systems.

Mechanical Properties of Single and Polycrystalline Solids from Machine Learning

AbstractCalculating the elastic and mechanical characteristics of non‐crystalline solids can be challenging due to the high computational cost of ab initio methods and the low accuracy of empirical potentials. This paper proposes a computational technique for efficient calculations of mechanical properties of polycrystals, composites, and multi‐phase systems from atomistic simulations with high accuracy and reasonable computational cost. The calculated elastic moduli of polycrystalline diamond and their dependence on grain size are determined using a developed approach based on actively learned machine learning interatomic potentials (MLIPs). These potentials are trained on local fragments of the polycrystalline system, and ab initio calculations are used to compute forces, stresses, and energies. This technique allows researchers to perform extensive calculations of the mechanical properties of complex solids with different compositions and structures, achieving high accuracy and facilitating the transition from ideal (single crystal) systems to more realistic ones.