Preprint
Article

This version is not peer-reviewed.

A Low‐Lattice‐Thermal‐Conductivity Screen for High‐Entropy Semiconductor Thermoelectric Candidates Across the GNoME Space

Submitted:

25 September 2026

Posted:

28 September 2026

You are already at the latest version

Abstract
High configurational entropy is a route to ultralow lattice thermal conductivity: chemicallydisordered, multi-cation lattices scatter phonons strongly, and low κ_lattice is the singlehardest term to suppress in the thermoelectric figure of merit ZT = S²σT/κ. Mosthigh-entropy thermoelectric work, however, is metallic-alloy-centric and proceeds onecomposition at a time. Here we screen a large computational materials space — the~51,000 stability-filtered GNoME candidates for which we have computed a full anharmonicthermal-transport layer — for high-entropy semiconductor thermoelectric candidates: band-gap-bearing, multi-cation compounds selected by configurational entropy, anentropy-corrected stability criterion, and directly computed low lattice thermal conductivity,rather than by the metallic solid-solution rules (Yang Ω/δ, Miedema ΔH_mix) that do notapply to semiconductors. Of 44,110 four-or-more-element GNoME candidates, 26,890already have a Slack lattice thermal conductivity below 1 W m⁻¹K⁻¹. Imposing anear-equimolar configurational-entropy cut (ΔS_config ≥ 1.5R), a hard κ_L < 1 W m⁻¹K⁻¹screen, and entropy-corrected thermodynamic stability at 1000 K yields a focused shortlistof 922 quinary candidates, every one of which is entropy-stabilized at 1000 K and carriesa calibrated predicted ZT (median 0.45, up to 0.75) from our gradient-boosted quantilemodel. The shortlist is dominated by heavy multi-cation pnictides and chalcogenides (e.g.La₄Th₂Sc₂(BiSb)₅ at κ_L ≈ 0.02 W m⁻¹K⁻¹; Li₂TbPrBiPb at predicted ZT ≈ 0.75). Thecontribution is a differentiated, physically-grounded, pre-screened candidate set forhigh-entropy-semiconductor thermoelectrics, released openly for first-principles andexperimental follow-up.
Keywords: 
;  
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.