Search

Machine-learned interatomic potentials (MLIPs) are becoming an essential tool in materials modeling. However, optimizing the generation of training data used to parameterize the MLIPs remains a significant challenge. This is because MLIPs can fail when encountering local enviroments too different from those present in the training data. The difficulty of determining \textit{a priori} the environments that will be encountered during molecular dynamics (MD) simulation necessitates diverse, high-quality training data. 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., Comment: 6 pages, 4 figures

The structure and interactions of coated silica nanoparticles have been studied in water using molecular dynamics simulations. For 5 nm diameter amorphous silica nanoparticles we studied the effects of varying the chain length and grafting density of polyethylene oxide (PEO) on the nanoparticle coating's shape and on nanoparticle-nanoparticle effective forces. For short ligands of length $n=6$ and $n=20$ repeat units, the coatings are radially symmetric while for longer chains ($n=100$) the coatings are highly anisotropic. This anisotropy appears to be governed primarily by chain length, with coverage playing a secondary role. For the largest chain lengths considered, the strongly anisotropic shape makes fitting to a simple radial force model impossible. For shorter ligands, where the coatings are isotropic, we found that the force between pairs of nanoparticles is purely repulsive and can be fit to the form $(R/2r_\text{core}-1)^{-b}$ where $R$ is the separation between the center of the nanoparticles, $r_\text{core}$ is the radius of the silica core, and $b$ is measured to be between 2.3 and 4.1., Comment: 20 pages, 6 figures

To prevent the flocculation and phase separation of nanoparticles in solution, nanoparticles are often functionalized with short chain surfactants. Here we present fully-atomistic molecular dynamics simulations which characterize how these functional coatings affect the interactions between nanoparticles and with the surrounding solvent. For 5 nm diameter silica nanoparticles coated with poly(ethylene oxide) (PEO) oligomers in water, we determined the hydrodynamic drag on two approaching nanoparticles moving through solvent and on a single nanoparticle as it approaches a planar surface. In most circumstances, acroscale fluid theory accurately predicts the drag on these nano-scale particles. Good agreement is seen with Brenner's analytical solutions for wall separations larger than the soft nanoparticle radius. For two approaching coated nanoparticles, the solvent-mediated (velocity-independent) and lubrication (velocity-dependent) forces are purely repulsive and do not exhibit force oscillations that are typical of uncoated rigid spheres., Comment: 4 pages, 3 figs

We describe a Continuous Hugoniot Method for the efficient simulation of shock wave fronts. This approach achieves significantly improved efficiency when the generation of a tightly spaced collection of individual steady-state shock front states is desired, and allows for the study of shocks as a function of a continuous shock strength parameter, $v_p$. This is, to our knowledge, the first attempt to map the Hugoniot continuously. We apply the method to shock waves in Lennard-Jonesium along the $<100>$ direction. We obtain very good agreement with prior simulations, as well as our own benchmark comparison runs., Comment: 4 pages, 3 figures, from Shock Compression of Condensed Matter 2005