Kondo-like phonon scattering in thermoelectric clathrates

TL;DR

This paper introduces an 'all phononic Kondo effect' to explain ultralow thermal conductivity in type-I clathrates, combining experimental measurements and ab initio simulations, and suggests new design strategies for thermoelectric materials.

Contribution

It proposes a novel phononic Kondo effect mechanism that explains low thermal conductivity in clathrates, supported by experimental and theoretical evidence.

Findings

Identification of the 'all phononic Kondo effect' as a key mechanism.

Experimental validation through thermodynamic and transport measurements.

Implications for designing better thermoelectric materials.

Abstract

Crystalline solids are generally known as excellent heat conductors, amorphous materials or glasses as thermal insulators. It has thus come as a surprise that certain crystal structures defy this paradigm. A prominent example are type-I clathrates and other materials with guest-host structures. They sustain low-energy Einstein-like modes in their phonon spectra, but are also prone to various types of disorder and phonon-electron scattering and thus the mechanism responsible for their ultralow thermal conductivities has remained elusive. While recent ab initio lattice dynamics simulations show that the Einstein-like modes enhance phonon-phonon Umklapp scattering, they reproduce experimental data only at low temperatures. Here we show that a new effect, an "all phononic Kondo effect", can resolve this discrepancy. This is evidenced by our thermodynamic and transport measurements on various clathrate single crystal series and their comparison with ab initio simulations. Our new understanding devises design strategies to further suppress the thermal conductivity of clathrates and other related materials classes, with relevance for the field of thermoelectric waste heat recovery but also more generally for phononic applications. More fundamentally, it may trigger theoretical work on strong correlation effects in phonon systems.