Work

Composition–structure–property relationships and finite-temperature magnetic trends in Co2MnTi1-xAlx full-Heusler alloys: A first-principles and external-dataset-trained GPR study

Year: 2026

Type: article

Abstract: The composition-dependent structural, lattice-dynamical, mechanical, electronic, and magnetic properties of Co 2 MnTi 1-x Al x (x = 0, 0.25, 0.50, 0.75, and 1) full-Heusler alloys were investigated using spin-polarized first-principles calculations. The previously studied Co 2 MnTi and Co 2 MnAl end members were used as validation benchmarks, while the analysis focused primarily on the intermediate ordered compositions. Ti/Al ordering at x = 0.50 permits an I4/mmm tetragonal description with c/a ≈ 1.40. Because the cubic-equivalent body-centered tetragonal ratio is √2, the normalized tetragonality is approximately 0.99, indicating chemical-order-induced symmetry lowering rather than a large spontaneous strain. The calculated elastic constants satisfy the relevant mechanical-stability conditions. However, the Poisson ratios indicate predominantly brittle or limited-ductility behavior, although x = 0.25 and 0.50 are relatively less brittle according to Pugh's criterion. The spin-resolved electronic structures reveal metallic majority-spin states and strongly suppressed, but finite, minority-spin states at the Fermi level. The alloys are therefore classified as highly spin-polarized metals with minority-spin pseudogaps rather than ideal half-metals. The total magnetic moment decreases from 4.95 to 4.02 μ B /f.u. with increasing Al content. An independent Heusler-family Gaussian-process regression model, trained on 246 structural entries from 144 composition groups, gives grouped-cross-validation errors of MAE = 131.86 K and RMSE = 177.45 K, with R 2 = 0.599. After endpoint calibration, the intermediate Curie-temperature estimates are 905 ± 198, 840 ± 195, and 774 ± 194 K. These values describe broad model-conditional trends rather than precise material-specific predictions. Experimental validation remains necessary for the ordered intermediate compositions, and chemical disorder may cause measured properties to differ from those calculated for the ideal ordered models.

Source: Solid State Communications

DOI: 10.1016/j.ssc.2026.116590