We present a high-throughput investigation of all 1062 aliovalent double half-Heusler (DHH) compositions by combining ${\mathrm{r}}^2\mathrm{SCAN}$-based density functional theory, Boltzmann transport, and a hybrid machine learning interatomic potentials MACE-assisted Slack workflow for lattice thermal conductivity. Clear sublattice-dependent trends are identified, where $X$- and $Y$-splitting generally favor the tetragonal $I\overline{4}2d$ prototype and preserve the half-Heusler's (HH) hybridization mechanism, while $Z$-splitting tends to stabilize $Pmn{2}_1$ at the prototype level but strongly perturbs the $p–d$ backbone, narrows or collapses the band gap, and greatly reduces thermodynamic competitiveness. As a result, near-hull compounds are concentrated mainly in the $X$-split and $Y$-split families, whereas the $Z$-split family is largely unstable. The thermodynamic accessibility of DHHs is further shown to correlate with both the stability of the parent HH end points and the presence of a finite semiconducting gap, indicating that stronger bonding and larger $E_{\mathrm{g}}$ generally coincide with greater stability. Transport analysis reveals that stable DHHs can reach power factors comparable to high-performance HH under heavy doping, with the optimal carrier concentration shifting systematically to higher values with increasing temperature. At a given thermal-conductivity budget, $p$-type systems outperform $n$-type ones. At the same time, the lowest lattice thermal conductivities are found for heavy-element, structurally complex compounds, confirming that DHH chemistry provides a natural route to phonon suppression. Together, these results establish a physically motivated design map for DHH thermoelectrics and identify $X/Y$-split compositions as the most promising route toward stable, high-performance materials.

Close

PrevNext

View figure in article

• Thermoelectrics
• Half-Heusler alloy
• Boltzmann theory
• Convolutional neural networks
• Density functional theory