LAUSR

Ideal thermoelectric materials should possess low thermal conductivity $$\kappa $$κ along with high electrical conductivity $$\sigma $$σ. Thus, strategies are needed to impede the propagation of phonons mostly responsible for thermal conduction while only marginally affecting charge carrier diffusion. Defect engineering may provide tools to fulfill this aim, provided that one can achieve an adequate understanding of the role played by multiple morphological defects in scattering thermal energy carriers. In this paper, we study how various morphological defects such as grain boundaries and dispersed nanovoids reduce the thermal conductivity of silicon. A blended approach has been adopted, using data from both simulations and experiments in order to cover a wide range of defect densities. We show that the co-presence of morphological defects with different characteristic scattering length scales is effective in reducing the thermal conductivity. We also point out that non-gray models (i.e. models with spectral resolution) are required to improve the accuracy of predictive models explaining the dependence of $$\kappa $$κ on the density of morphological defects. Finally, the application of spectral models to Matthiessen’s rule is critically addressed with the aim of arriving at a compact model of phonon scattering in highly defective materials showing that non-local descriptors would be needed to account for lattice distortion due to nanometric voids.