Speaker
Description
Titanium aluminium nitride (TiAlN) is a benchmark protective coating. Incorporating oxygen can delay thermally induced phase transformations by up to 300 °C, but this enhanced thermal stability comes at the cost of a lower elastic modulus. Understanding how chemical disorder, vacancies, and local atomic reconstructions control this trade-off and influence high-temperature mass transport requires simulations beyond the length and time scales accessible to density-functional theory. We therefore developed an atomic cluster expansion machine-learning potential for cubic TiAlON through iterative active learning focused on structural, mechanical, and diffusion-related configurations. The potential reproduces DFT energies and forces with RMSE values of 4.08 meV/atom and 0.178 eV/Å, respectively. Independent validation gives normalized mean absolute errors of 2.2–3.5% for the elastic moduli, an energy RMSE of 5.8 meV/atom, and an RMSE of 0.25 eV for point-defect formation and reconstruction energies. Monte Carlo simulations show that chemically random TiAlON configurations are both higher in energy and softer than structures exhibiting moderate Ti/Al partitioning. Warren–Cowley parameters connect the developing chemical short-range order with an increase in elastic modulus. At higher vacancy concentrations, this ordering becomes coupled to metal-vacancy clustering and the displacement of predominantly Al atoms towards tetrahedral sites, producing tetrahedrally reconstructed, vacancy-rich aggregates. Nanosecond-scale molecular-dynamics simulations at 2000 K reveal that these reconstructed motifs remain dynamic, forming transient aggregates and enabling Al motion between neighbouring tetrahedral sites. Together, the results identify chemical short-range order and vacancy-mediated reconstruction as coupled mechanisms governing the mechanical response and high-temperature atomic transport in cubic TiAlON.