Designing Materials That Defy Heat Flow
Linking simple physical rules, machine learning and neutron spectroscopy to discover crystals with record-low thermal conductivity
Why heat is so hard to control
Heat is everywhere and controlling it is a persistent challenge. Electronics overheat, engines waste energy as heat and thermal insulating materials are pushed to their limits.
In crystalline solids, heat is carried by the collective vibrations of the atoms, known as phonons. Because in these materials atoms are arranged in a periodic and ordered pattern, the phonons can travel efficiently through the crystal, allowing heat to move easily. This heat transport through atomic vibrations is quantified by the lattice thermal conductivity (κL).
Glasses and amorphous materials behave very differently. Unlike crystals, their atoms are arranged in a more random and irregular way. This irregularity prevents vibrations/phonons from travelling smoothly through the material, resulting in much lower thermal conductivity.
Creating a crystal that blocks heat like a glass has long been a goal for scientists, but one that has proven extremely difficult.
Schematic diagram of high-throughput screening process.
Starting from more than 150,000 known compounds in the Materials Project database, researchers applied a series of physical and chemical filters and used machine-learning calculations to estimate heat transport properties. This step-by-step selection narrowed the search to a few hundred promising candidates with extremely low thermal conductivity.Credit: Nature Communications (2025).
A surprisingly simple rule
Traditionally, discovering materials with very low lattice thermal conductivity requires expensive calculations or time-consuming experiments. While empirical models exist, they rely on numerous fitted parameters and are not well suited for rapid, large-scale screening.
In this study, researchers identified a simple universal descriptor designed to estimate lattice thermal conductivity using only two basic properties of a crystal:
- how many atoms make up its repeating unit (the number of atoms in the primitive unit cell)
- how fast vibrations/phonons travel through the material (the sound velocity in the material).
To test their idea, the researchers gathered experimental measurements of thermal conductivity at room temperature (~ 27 °C) for more than 60 different crystalline materials, ranging from very good heat conductors to extremely poor ones. They found a clear and consistent linear relationship (on a logarithmic scale) between thermal conductivity and their proposed descriptor.
This simple indicator makes it possible to quickly identify materials that are likely to conduct heat very poorly, without having to perform complex calculations of how vibrations move through the crystal.
From 153,902 materials to 2 candidates
Armed with this insight and machine-learning calculations, the team rapidly screened tens of thousands of known materials. Among them, two metal halides — CsAg₂I₃ and CsCu₂I₃ stood out.
When the materials were synthesised and measured in the laboratory, the result was striking. Lattice thermal conductivity is exceptionally low in these materials, among the lowest ever measured for a crystalline solid, and nearly constant over a wide temperature range. (CsAg₂I₃ exhibits a lattice thermal conductivity of 0.15 – 0.16 W·m⁻¹·K⁻¹ between 170 and 400 K and CsCu₂I₃: 0.18 – 0.20 W·m⁻¹·K⁻¹ between 300 and 523 K)
Notably, CsAg₂I₃ crystal behaves thermally more like a glass than like a crystal displaying a nearly temperature-independent thermal conductivity across a wide temperature range.
Neutrons probe the microscopic dynamics
Finding such a material is only half the story. Understanding why it behaves this way requires seeing how atoms actually move, something only a few experimental techniques can do.
This is where neutron scattering plays a decisive role. At the Institut Laue-Langevin (ILL) in Grenoble, scientists used inelastic neutron scattering measurements performed on the PANTHER time-of-flight spectrometer to observe atomic vibrations inside the crystal as temperature changed (from 1.5 to 300 K).
They proved that the microscopic vibrations, i.e. phonons, of the modeled materials are correct. Instead of travelling smoothly across the material as well-defined waves, these vibrations interact strongly with each other and quickly lose their collective nature. As a result, heat does not flow in a wave-like manner but spreads more diffusively, a behaviour more typical of glasses than crystals.
Although the atomic structure remains perfectly ordered, the way the atoms move prevents heat from travelling efficiently.
A new strategy for designing thermal materials
By linking simple measurable quantities to the way atoms move inside a crystal and validating this connection through neutron spectroscopy, this study demonstrates that heat transport in crystals can be predicted, identified and understood in a systematic way. Beyond this specific crystal, the implications are wide-ranging. Materials with very low thermal conductivity play a crucial role in thermoelectric devices, thermal barrier coatings and energy-efficient technologies, where controlling heat flow is essential for performance and sustainability.
More broadly, the work reflects a shift in how new materials are discovered, moving away from isolated, trial-and-error findings toward data-driven and theory-guided exploration.
References:
X.Shen, J. Zheng, M. M. Koza, P. Levinsky, J. Hejtmanek, P. Boullay, B. Raveau, J. Wang, J. Li, P. Lemoine, C.Candolfi & E. Guilmeau, Accelerated discovery of crystalline materials with record ultralow lattice thermal conductivity via a universal descriptor. Nature Communications (2025). (https://www.nature.com/articles/s41467-025-67333-z)
ILL instruments: PANTHER
ILL Contact Person: Michael Marek Koza
Institutions involved in the research: CRISMAT, CNRS, Caen-Normandy University,Northwestern Polytechnical University, Dartmouth College, Czech Academy of Science, ShanghaiTech University, University of Lorraine