Machine learning and Calphad guided exploration of low-Neutron-cross-section refractory multi-principal element alloys

The development of Generation IV nuclear reactors demands cladding materials that combine high-temperature strength with low neutron cross-section. This study presents an integrated Machine Learning and CALPHAD framework to navigate the refractory multi-principal element alloy (MPEA) design space under constraints of Valence Electron Concentration (VEC ≤ 4.8) and thermal neutron cross-section (< 2.4 barns). Random Forest analysis identified Mo and Al as the dominant solid-solution strengthening elements in BCC MPEAs, with the Al strengthening effect saturating at approximately 7 at.%, and guided subsequent high-throughput CALPHAD screening of four candidate compositions. Composition #C3, Al9.48-Mo4.68-Nb20.22-Ti4.61-Zr61.01 (at. %), achieved a chemically homogeneous, single-phase BCC structure in the as-cast condition, consistent with Scheil-Gulliver solidification predictions, and exhibited a Vickers hardness of 441 HV and a localized estimated yield strength of approximately 1970 MPa from nanoindentation reverse analysis. While the neutron cross-section of #C3 (0.77 barns) is approximately one-third that of FeCrAl alloys, it remains roughly four times higher than commercial Zircaloy-4 (∼0.195 barns). This penalty is partially offset by a yield strength approximately five times greater than Zircaloy-4, motivating the use of a strength-to-neutron-penalty ratio as the primary performance benchmark, on which #C3 achieves parity with Zircaloy-4. Comprehensive evaluation of bulk ductility, fracture toughness, oxidation resistance, and irradiation behavior remains necessary before nuclear cladding qualification and constitutes essential future work.