Lattice-to-Total Thermal Conductivity Ratio: A Phonon-Glass Electron-Crystal Descriptor for Data-Driven Thermoelectric Design

TL;DR

This paper introduces a data-driven framework using machine learning to identify and optimize thermoelectric materials by targeting an optimal lattice-to-total thermal conductivity ratio near 0.5, enhancing energy harvesting efficiency.

Contribution

The study develops a novel machine learning approach to predict thermal conductivities and identifies a key ratio for thermoelectric performance, guiding materials design.

Findings

High-$ZT$ materials cluster near a $$ ratio of 0.5.

The framework screened over 104,000 compounds, finding 2,522 ultralow-$$ candidates.

Doping strategies can shift materials toward the optimal PGEC regime.

Abstract

Thermoelectrics (TEs) are promising candidates for energy harvesting with performance quantified by figure of merit, $ZT$. To accelerate the discovery of high-$ZT$ materials, efforts have focused on identifying compounds with low thermal conductivity $\kappa$. Using a curated dataset of 71,913 entries, we show that high-$ZT$ materials reside not only in the low-$\kappa$ regime but also cluster near a lattice-to-total thermal conductivity ratio ($\kappa_\mathrm{L}/\kappa$) of approximately 0.5. This optimal ratio provides a quantitative descriptor for the well-known phonon-glass electron-crystal (PGEC) design concept. Building on this insight, we construct a framework consisting of two machine learning models for the lattice and electronic components of thermal conductivity that jointly provide both $\kappa$ and $\kappa_\mathrm{L}/\kappa$ for screening and guiding the optimization of TE materials. By applying this framework to 104,567 inorganic compounds, we identify 2,522 ultralow-$\kappa$ candidates while simultaneously evaluating their proximity to the optimal PGEC regime. A follow-up case study on chemical doping demonstrates how the framework can qualitatively provide optimization strategies that shift pristine materials toward the ideal $\kappa_\mathrm{L}/\kappa$ $\approx$ 0.5 target. Ultimately, by integrating rapid screening with PGEC-guided optimization, our data-driven framework takes a critical step towards closing the gap between materials discovery and performance enhancement.