GPUMD: A package for constructing accurate machine-learned potentials and performing highly efficient atomistic simulations.

We present our latest advancements of machine-learned potentials (MLPs) based on the neuroevolution potential (NEP) framework introduced in Fan et al. [Phys. Rev. B 104, 104309 (2021)] and their implementation in the open-source package gpumd. We increase the accuracy of NEP models both by improving the radial functions in the atomic-environment descriptor using a linear combination of Chebyshev basis functions and by extending the angular descriptor with some four-body and five-body contributions as in the atomic cluster expansion approach. We also detail our efficient implementation of the NEP approach in graphics processing units as well as our workflow for the construction of NEP models and demonstrate their application in large-scale atomistic simulations. By comparing to state-of-the-art MLPs, we show that the NEP approach not only achieves above-average accuracy but also is far more computationally efficient. These results demonstrate that the gpumd package is a promising tool for solving challenging problems requiring highly accurate, large-scale atomistic simulations. To enable the construction of MLPs using a minimal training set, we propose an active-learning scheme based on the latent space of a pre-trained NEP model. Finally, we introduce three separate Python packages, viz., gpyumd, calorine, and pynep, that enable the integration of gpumd into Python workflows.

Non-Bonded Radii of the Atoms Under Compression.

We present quantum mechanical estimates for non-bonded, van der Waals-like, radii of 93 atoms in a pressure range from 0 to 300 gigapascal. Trends in radii are largely maintained under pressure, but atoms also change place in their relative size ordering. Multiple isobaric contractions of radii are predicted and are explained by pressure-induced changes to the electronic ground state configurations of the atoms. The presented radii are predictive of drastically different chemistry under high pressure and permit an extension of chemical thinking to different thermodynamic regimes. For example, they can aid in assignment of bonded and non-bonded contacts, for distinguishing molecular entities, and for estimating available space inside compressed materials. All data has been made available in an interactive web application.

Beyond Magic Numbers: Atomic Scale Equilibrium Nanoparticle Shapes for Any Size.

In the pursuit of complete control over morphology in nanoparticle synthesis, knowledge of the thermodynamic equilibrium shapes is a key ingredient. While approaches exist to determine the equilibrium shape in the large size limit (≳10-20 nm) as well as for very small particles (≲2 nm), the experimentally increasingly important intermediate size regime has largely remained elusive. Here, we present an algorithm, based on atomistic simulations in a constrained thermodynamic ensemble, that efficiently predicts equilibrium shapes for any number of atoms in the range from a few tens to many thousands of atoms. We apply the algorithm to Cu, Ag, Au, and Pd particles with diameters between approximately 1 and 7 nm and reveal an energy landscape that is more intricate than previously suggested. The thus obtained particle type distributions demonstrate that the transition from icosahedral particles to decahedral and further into truncated octahedral particles occurs only very gradually, which has implications for the interpretation of experimental data. The approach presented here is extensible to alloys and can in principle also be adapted to represent different chemical environments.