Why does Re anti-segregate from the grain boundaries in Ni?

<title>Abstract</title> In virtually every polycrystalline material with more than one elemental constituent, grain boundaries (GBs) offer preferential sites for minority elements to occupy; the phenomenon of GB segregation is ubiquitous. Here, in stark contrast, we report a rare observation of “antisegregation” in a clean binary alloy of Ni-Re. In addition to being unusual, this observation also runs counter to prior density functional theory predictions, which suggest that rhenium should segregate to GBs in Ni. From first principles, we demonstrate that this discrepancy originates from a magnetic effect. Whereas the zero-Kelvin calculations that are typical for assessing GB segregation energies are appropriate for the low-temperature, ferromagnetic state of Ni, GB segregation is a high temperature phenomenon where the system is paramagnetic instead. This magnetic change turns out to flip the sign of the segregation tendency on average, and indeed predicts GB solvent enrichment, rather than typical solute enrichment. With a multiscale, quantum-accurate model, we apply machine learning methods to address the full, complex case of a polycrystalline system. The analysis shows that when this magnetic effect is accounted for, the experimental antisegregation is quantitatively predicted with accuracy. Our findings call for a broader focus on magnetic disorder as a critical factor in GB segregation science and high-temperature alloy design.