Density-functional theory (DFT) and ab initio molecular dynamics (AIMD) calculations are used to build machine-learning interatomic potentials (MLIPs), which enable large-scale atomistic simulations of ceramics, intermetallics and metals with ab initio accuracy. In the coming year, the project focuses on
(i) Monte-Carlo (MLIP-based) discovery of new nitride phases that tend to form Guiner-Preston zones for high-temperature hardness
(ii) stress-intensity-controlled (K-controlled) fracture simulations, including fatigue-crack initiation in tungsten with point defects, crack-tip plasticity and toughness of V-Mo-N alloys, and fracture of MAB (layered-borides) phases;
(iii) surface energies from the random-phase approximation, with a convergence study with respect to the slab thickness;
(iv) catalytic reactions of H2O and CO2 molecules on two-dimensional gold (goldene), trilayer goldene (+graphene taken as reference chemically-inert system)
(v) finite-temperature phase stability, order-disorder transitions and Ag-ion transport in superionic silver chalcohalide antiperovskites (Ag3SX, X = Cl, Br, I), studied with fine-tuned MLIPs and free-energy calculations. The DFT and AIMD calculations provide the training and reference data for the MLIPs, which are used for molecular dynamics simulations of diffusion, nucleation, phase transitions and the mechanical behavior of defect-free and defective crystals. All theoretical predictions are subject to experimental validation through collaborations with synthesis and characterization groups.
Using NAISS resources (acknowledged), during 2024-present we published 19 peer-reviewed articles, including Nature Synthesis (2024), npj Computational Materials (2024), Science Advances (2025), Communications Materials (2025) and Acta Materialia (2024 & 2025), and seven further manuscripts are submitted or under review. The availability of NAISS resources has been fundamental for our publications. The full list of publications and academic achievements can be found in the report.