This work presents Neural Equivariant Interatomic Potentials (NequIP), an E(3)-equivariant neural network approach for learning interatomic potentials from ab-initio calculations for molecular dynamics simulations. While most contemporary symmetry-aware models use invariant convolutions and only act on scalars, NequIP employs E(3)-equivariant convolutions for interactions of geometric tensors, resulting in a more information-rich and faithful representation of atomic environments. The method achieves state-of-the-art accuracy on a challenging and diverse set of molecules and materials while exhibiting remarkable data efficiency. NequIP outperforms existing models with up to three orders of magnitude fewer training data, challenging the widely held belief that deep neural networks require massive training sets. The high data efficiency of the method allows for the construction of accurate potentials using high-order quantum chemical level of theory as reference and enables high-fidelity molecular dynamics simulations over long time scales.

A simultaneously accurate and computationally efficient parametrization of the potential energy surface of molecules and materials is a long-standing goal in the natural sciences. While atom-centered message passing neural networks (MPNNs) have shown remarkable accuracy, their information propagation has limited the accessible length-scales. Local methods, conversely, scale to large simulations but have suffered from inferior accuracy. This work introduces Allegro, a strictly local equivariant deep neural network interatomic potential architecture that simultaneously exhibits excellent accuracy and scalability. Allegro represents a many-body potential using iterated tensor products of learned equivariant representations without atom-centered message passing. Allegro obtains improvements over state-of-the-art methods on QM9 and revMD17. A single tensor product layer outperforms existing deep MPNNs and transformers on QM9. Furthermore, Allegro displays remarkable generalization to out-of-distribution data. Molecular simulations using Allegro recover structural and kinetic properties of an amorphous electrolyte in excellent agreement with ab-initio simulations. Finally, we demonstrate parallelization with a simulation of 100 million atoms.

Recently, machine learning (ML) has been used to address the computational cost that has been limiting ab initio molecular dynamics (AIMD). Here, we present GNNFF, a graph neural network framework to directly predict atomic forces from automatically extracted features of the local atomic environment that are translationally-invariant, but rotationally-covariant to the coordinate of the atoms. We demonstrate that GNNFF not only achieves high performance in terms of force prediction accuracy and computational speed on various materials systems, but also accurately predicts the forces of a large MD system after being trained on forces obtained from a smaller system. Finally, we use our framework to perform an MD simulation of Li7P3S11, a superionic conductor, and show that resulting Li diffusion coefficient is within 14% of that obtained directly from AIMD. The high performance exhibited by GNNFF can be easily generalized to study atomistic level dynamics of other material systems.

We consider the transient non-equilibrium electronic distribution that is created in a metal nanoparticle upon plasmon excitation. Following light absorption, the created plasmons decohere within a few femtoseconds, producing uncorrelated electron-hole pairs. The corresponding non-thermal electronic distribution evolves in response to the photo-exciting pulse and to subsequent relaxation processes. First, on the femtosecond timescale, the electronic subsystem relaxes to a Fermi-Dirac distribution characterized by an electronic temperature. Next, within picoseconds, thermalization with the underlying lattice phonons leads to a hot particle in internal equilibrium that subsequently equilibrates with the environment. Here we focus on the early stage of this multistep relaxation process, and on the properties of the ensuing non-equilibrium electronic distribution. We consider the form of this distribution as derived from the balance between the optical absorption and the subsequent relaxation processes, and discuss its implication for (a) heating of illuminated plasmonic particles, (b) the possibility to optically induce current in junctions, and (c) the prospect for experimental observation of such light-driven transport phenomena.

This work presents Neural Equivariant Interatomic Potentials (NequIP), a SE(3)-equivariant neural network approach for learning interatomic potentials from ab-initio calculations for molecular dynamics simulations. While most contemporary symmetry-aware models use invariant convolutions and only act on scalars, NequIP employs SE(3)-equivariant convolutions for interactions of geometric tensors, resulting in a more information-rich and faithful representation of atomic environments. The method achieves state-of-the-art accuracy on a challenging set of diverse molecules and materials while exhibiting remarkable data efficiency. NequIP outperforms existing models with up to three orders of magnitude fewer training data, challenging the widely held belief that deep neural networks require massive training sets. The high data efficiency of the method allows for the construction of accurate potentials using high-order quantum chemical level of theory as reference and enables high-fidelity molecular dynamics simulations over long time scales.

No abstract

In DIII-D experiments, rapid termination by Ar pellet injection sometimes produces a posttermination runaway electron (RE) current plateau, but this effect is highly non-reproducible on a shot-to-shot basis, particularly for diverted target plasmas. A set of DIII-D discharges is analyzed with two MHD codes to understand the relationship between the current profile of the target plasma and the amplitude of the RE current plateau. Using the linear stability code GATO, a correlation between the radial profile of the unstable mode just after Ar pellet injection and the observed appearance of an RE plateau is identified. Nonlinear NIMROD simulations with RE test-particle calculations directly predict RE confinement times during the disruption. With one exception, NIMROD predicts better RE confinement for shots in which higher RE currents were observed in DIII-D. But, the variation in confinement is primarily connected to the saturated mode amplitude and not its radial profile. Still, both sets of analyses support the hypothesis that RE deconfinement by MHD fluctuations is a major factor in the shot-to-shot variability of RE plateaus, though additional factors such as seed current amplitude can not be ruled out.

Neural network force field (NNFF) is a method for performing regression on atomic structureforce relationships, bypassing expensive quantum mechanics calculation which prevents the execution of long ab-initio quality molecular dynamics simulations. However, most NNFF methods for complex multielement atomic systems indirectly predict atomic force vectors by exploiting just atomic structure rotation-invariant features and the network-feature spatial derivatives which are computationally expensive. We develop a staggered NNFF architecture exploiting both rotation-invariant and covariant features separately to directly predict atomic force vectors without using spatial derivatives, thereby reducing expensive structural feature calculation by ~180-480×. This acceleration enables us to develop NNFF which directly predicts atomic forces in complex ternary and quaternary-element extended systems comprised of long polymer chains, amorphous oxide, and surface chemical reactions. The staggered rotation-invariant-covariant architecture described here can also directly predict complex covariant vector outputs from local physical structures in domains beyond computational material science.