Neural-Network Based Atomistic Modeling of Plastic Deformation Mechanisms of Crystalline Molybdenum

Numerical investigations of nano-mechanical testing for metals and alloys require accurate inter-atomic potentials that may predict configurational energies and interatomic forces, consistent with ab initio calculations. In this work, we investigate crystalline molybdenum (Mo), a viable candidate for extreme environments such as fusion reactors [1]. Mechanical properties of Mo, such as nanoin-dentation hardness, display non-trivial temperature dependence that requires further validation and deeper understanding, beyond classical force fields methods [2]. In this work, we create a Neural-network interatomic potential (NNIP) for nanoindentation and uniaxial tension of pure crystalline Mo to investigate mechanisms of dislocation nucleation and evolution at multiple temperatures, up to 1000K. Elastic constants, dislocation densities, strain maps and slip traces as a function of inden-tation depth of the system are compared with embedded atom method (EAM) potentials and the advantages and limitations of NNIPs over traditional potentials are reported [3].<br/><br/>[1] J. Byggmastar, A. Hamedani, et al. Phys. Rev. B, 100:144105, Oct 2019.<br/>[2] F.J. Dominguez-Gutierrez, S. Papanikolaou, et al. Materials Science and Engineering: A, 826:141912, 2021.<br/>[3] Amirhossein Naghdi et al. In preparation (2022).