Learning Grain Boundary Segregation Vibrational Spectra from Ni-Based Polycrystals

Polycrystalline systems consist of a broad range of planar defect configurations that cannot be represented by simplified boundary models such as small symmetric tilt boundaries. The multiple-site nature of grain boundaries calls for a spectral approach as opposed to a single-site McLean model. However, evaluating solute segregation thermodynamic quantities at every site in large polycrystals is an exhaustive process, especially for the segregation vibrational entropy. Here we show an attempt to integrate a data science approach with computational materials modeling for computational tractability. A modified local harmonic method is applied in combination with a statistical approach to quantify grain boundary segregation vibrational spectra of embedded dilute Ni-based binary alloy pairs. The database allows more rigorous predictions of grain boundary enrichment at a function of concentration, grain size and temperature which are critical for solute segregation-stabilized nanocrystalline alloys.