Improving the accuracy of spin-lattice simulations with machine-learning interatomic potentials.
Abstract not provided.
SEARCH · Search NASA
Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.
Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.
Abstract not provided.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Explore the source record for details and available documents.
Uranium-silicide (U-Si) fuels are being pursued as a possible accident tolerant fuel (ATF). This uranium alloy fuel benefits from higher thermal conductivity and higher fissile density compared to UO{sub 2}. In order to perform engineering scale nuclear fuel performance simulations, the material properties of the fuel must be known. Currently, the experimental data available for U-Si fuels is rather limited. Thus, multiscale modeling efforts are underway to address this gap in knowledge. In this study, a semiempirical modified Embedded Atom Method (MEAM) potential is presented for the description of the U-Si system. The potential accurately describes not only the primary phase of interest (U{sub 3}Si{sub 2}), but also a variety of U-Si phases across the composition spectrum.
In fusion applications, helium created by transmutation plays an important role in the response of reduced-activation ferritic/martensitic (RAFM) steels to neutron radiation damage. We have performed extensive atomistic simulations using the ORNL 3-body Fe He interatomic potential combined with three interatomic potentials for the iron matrix. Some of the results obtained are summarized in this review. Interstitial helium is very mobile and coalesces together to form interstitial clusters. We have investigated the mobility of these clusters. When an interstitial He cluster reaches sufficient size, it punches out an Fe interstitial, creating an immobile helium vacancy cluster. If more helium atoms join it, more Fe interstitials can be created; the He V defect is a nascent bubble. These mechanisms are investigated together in simulations that examine the nucleation of He defects. Mobile interstitial He clusters and helium bubbles 1 to 6 nm across are also simulated separately. Results are compared based on temperature and interatomic potentials used.
Abstract Availability of affordable and widely applicable interatomic potentials is the key needed to unlock the riches of modern materials modeling. Artificial neural network-based approaches for generating potentials are promising; however, neural network training requires large amounts of data, sampled adequately from an often unknown potential energy surface. Here we propose a self-consistent approach that is based on crystal structure prediction formalism and is guided by unsupervised data analysis, to construct an accurate, inexpensive, and transferable artificial neural network potential. Using this approach, we construct an interatomic potential for carbon and demonstrate its ability to reproduce first principles results on elastic and vibrational properties for diamond, graphite, and graphene, as well as energy ordering and structural properties of a wide range of crystalline and amorphous phases.
Here, a classical interatomic potential for iron/iron-fluoride systems is developed in the framework of the charge optimized many-body (COMB) potential. This interatomic potential takes into consideration the effects of charge transfer and many-body interactions depending on the chemical environment. The potential is fitted to a training set composed of both experimental and ab initio results of the cohesive energies of several Fe and FeF 2 crystal phases, the two fluorine molecules F 2 and the F 2 –1 dissociation energy curve, the Fe and FeF 2 lattice parameters of the ground state crystalline phase, and the elastic constants of the body centered cubic Fe structure. The potential is tested in an NVT ensemble for different initial structural configurations as the crystal ground state phases, F 2 molecules, iron clusters, and iron nanospheres. In particular, we model the FeF 2 /Fe bilayer and multilayer interfaces, as well as a system of square FeF 2 nanowires immersed in an iron solid. It has been shown that there exists a reordering of the atomic positions for F and Fe atoms at the interface zone; this rearrangement leads to an increase in the charge transfer among the atoms that make the interface and put forward a possible mechanism of the exchange bias origin based on asymmetric electric charge transfer in the different spin channels.
Machine-learned interatomic potentials (MLIPs) are becoming an essential tool in materials modeling. However, optimizing the generation of training data used to parametrize the MLIPs remains a significant challenge. This is because MLIPs can fail when encountering local environments too different from those present in the training data. The difficulty of determining a priori the environments that will be encountered during molecular dynamics simulation necessitates diverse, high-quality training data. Here, this study investigates how training data diversity affects the performance of MLIPs using the Ultra-Fast force field (UF 3 ) to model amorphous silicon nitride. We employ expert and autonomously generated data to create the training data and fit four force field variants to subsets of the data. Our findings reveal a critical balance in training data diversity: insufficient diversity hinders generalization, while excessive diversity can exceed the MLIP's learning capacity, reducing simulation accuracy. Specifically, we found that the UF 3 variant trained on a subset of the training data, in which nitrogen-rich structures were removed, offered vastly better prediction and simulation accuracy than any other variant. By comparing these UF 3 variants, we highlight the nuanced requirements for creating accurate MLIPs, emphasizing the importance of application-specific training data to achieve optimal performance in modeling complex material behaviors.
A new interatomic potential for the Mo-V system is introduced to facilitate the study of phase stability and mechanical properties at lower temperatures. This potential is based on a generalization of the embedded atom method and includes contributions from embedding energy, explicit two- and three-body interactions and nonlocal many-body interaction terms. The parameters of the potential are optimized by using data from ab initio density functional theory (DFT) calculations. The potential is rigorously validated across a range of physical properties, such as elastic constants, equation of states, phonon dispersion curves, point defect properties and melting temperatures for different compositions. Even though our potential is trained on a small dataset, its accuracy is comparable to available machine learning potentials for Mo and V. Furthermore, our results show that an ordered B2 phase is stable at low temperatures in alloys containing 50% V, but the solid solution phase is stable above 800 K. However, such long-range ordering is not observed in V-rich or Mo-rich alloys. In addition, our results show that V segregates to dislocation cores and grain boundaries.
The spectral features of the velocity of light-induced drift (LID) of lithium atoms in a neon buffer gas have been theoretically studied using four known ab initio calculated interatomic interaction potentials for the Li–Ne system of colliding particles. The calculations with each of these four potentials predict a strong temperature dependence of the spectral shape of the LID signal of Li atoms in the Ne atmosphere in various temperature intervals. The results allow the high-precision test of the interatomic interaction potentials in experiments on anomalous LID.
Based on five different nonempirical (calculated ab initio) interaction potentials for pairs of colliding Li – Ne particles and three different interaction potentials for pairs of colliding Li – Ar particles, we have theoretically investigated the spectral features of the light-induced drift (LID) rate of Li atoms in the buffer Ne gas in a mixture of buffer Ne + Ar gases. The calculations of LID of Li atoms in the buffer Ne gas for two interaction potentials predict anomalous LID and, as a result, strong sensitivity of the spectral shape of the LID lines to the differences in these potentials. For three other potentials (out of the five in question), the shape of the LID line of Li atoms in Ne is insensitive to the shape of the potential, since calculations with these potentials predict a normal LID effect. In this case, as it turned out, by adding a small fraction (approximately 10 %) of Ar to Ne, one can go from normal LID to anomalous LID and thereby radically increase the sensitivity of the LID line shape of Li atoms to the difference in these interaction potentials. The results obtained enable high-precision testing of the interatomic interaction potentials in experiments on anomalous LID. (nonlinear optical phenomena)