Interatomare potentials for amorphous solids from second principles
Final Report Abstract
In this project MLPs for the material systems Cu-Zr and Si-O-C were developed, extensively tested and then applied to study various materials properties. The major results are: • A GAP and an ACE potential were developed for the Si-O system. For the first time, these potentials allow the simultaneous description of amorphous silica, high pressure phases, silicon monoxide and various defects in the material. Furthermore, a new approach for the extraction of small samples from large scale simulations feasible for active learning workflows was presented. • The Si-O dataset was extended to Si-O-C using an active learning based approach and an ACE potential was fitted to the system. Using the potential, details of the nanostructure of SiOC ceramics (siliconoxycarbides) were investigated showing that two opposing microstructure models established in the literature (based on experimental data) most likely correspond to differently relaxed samples. Additionally, structure-property relations were established for the Young’s modulus and the silica and free carbon volume fractions, as well as the C content in mixed SiOx C4−x tetrahedra. • An ACE potential for the Cu-Zr system was fitted to a large DFT dataset [B]. This potential was employed to calculate the phase diagram of the system, showing that it is capable of an accurate description of thermodynamics over the whole compositional range. Furthermore, the potential is capable of an accurate description of the glass structure. The mechanical properties of a glass-crystal composite were investigated by tensile tests, showing a martensitic phase transformation from the B2 to the B19’ phase along the shear band path. In addition to the various MLPs, tools for the reproducible development of classical and machine learning potentials were implemented and integrated into the pyiron framework.
Publications
-
A machine-learned interatomic potential for silica and its relation to empirical models. npj Computational Materials, 8(1).
Erhard, Linus C.; Rohrer, Jochen; Albe, Karsten & Deringer, Volker L.
-
General purpose potential for glassy and crystalline phases of Cu-Zr alloys based on the ACE formalism. Physical Review Materials, 8(4).
Leimeroth, Niklas; Rohrer, Jochen & Albe, Karsten
-
Modelling atomic and nanoscale structure in the silicon–oxygen system through active machine learning. Nature Communications, 15(1).
Erhard, Linus C.; Rohrer, Jochen; Albe, Karsten & Deringer, Volker L.
