Learning from small data: Mechanistic understanding of B2/BCC-induced strengthening in FeCrNiCoAl based high-entropy alloys at elevated temperatures

This work proposed an efficient machine learning (ML) methodology to predict the composition and properties of a Al–Cr–Co–Fe–Ni based HEAs with high fidelity using a limited data set. Alloying effects were predicted to provide guidance for material design. For experimental validation, the microstructure and mechanical behavior at elevated temperatures of a Al_1.1CrCoFe_0.9Ni_0.9 HEA were systematically investigated. It consisted of polygonal grains with intragranular dendritic segregation. This alloy was composed of B2 and BCC phase, and the phase structure remained stable as the temperature increased. High nanohardness of 7.93 GPa and elastic modulus of 229.1 GPa, respectively, were obtained. This alloy exhibited good compressive properties at elevated temperatures. The yield strength and fracture strength at 400 °C were 1162 MPa and 1843 MPa, with plastic strain of 24%, respectively. The yield strength decreased to 968 MPa at 600 °C, and showed superior specific yield strength compared with some reported HEAs. The work hardening capability can be retained up to 600 °C, maintaining high mechanical strength and resistance to deformation at elevated temperatures. The deformation mechanisms at 400 °C included local lattice rotation and dislocation-based mechanisms. It exhibited a transition from dislocation slip and twinning to co-deformation of the BCC and B2 phases. The texture intensity decreased with increasing temperatures, accompanied by increased ductility and the ability to accommodate larger deformations without failure. This temperature-dependent mechanism, along with stable phase structures and interface strengthening, ensured good mechanical performance at elevated temperatures. This work demonstrated that a data-driven ML model could be successfully used for property prediction and develop high entropy alloys with attractive elevated temperature mechanical properties.