phonon scattering

If phonons carry heat, then whatever knocks them off course is what limits how well a ceramic conducts. Picture a crowd of runners trying to cross a field: obstacles slow them down and send them off in random directions. For phonons the obstacles are other phonons, stray impurity atoms, grain boundaries, and pores. The more obstacles a phonon meets, the shorter its mean free path, the lower the thermal conductivity — which, flipped around, is exactly why disordered and impure ceramics make such good insulators.

There are several distinct scattering mechanisms, and they add up like resistances (roughly 1/l_total equals the sum of 1/l for each mechanism). First, phonons scatter off each other through anharmonic bonds — the Umklapp process — which dominates at high temperature and makes the conductivity of a clean crystal fall off roughly as 1/T. Second, phonons scatter off point defects and mass disorder: an impurity or a substituted atom of different mass is a tiny bump, and this Rayleigh-type scattering hits short-wavelength (high-frequency) phonons hardest, scaling with the square of the mass difference. Third, grain boundaries and pores scatter phonons, which matters most at low temperature or in very fine-grained material where the mean free path would otherwise be long. Each mechanism dominates in its own temperature range.

This is the physics engineers dial to order. In a thermal-barrier coating you want the lowest possible conductivity, so yttria-stabilized zirconia is deliberately loaded with dopant atoms and oxygen vacancies to scatter phonons hard, pinning its conductivity near 2 W/m/K. Nanostructuring adds a forest of grain boundaries for the same reason, and is the basis of high-performance thermoelectrics. It also explains the humble glass: with no long-range order, its mean free path is already down at the interatomic spacing, so it insulates about as well as a solid possibly can. Scattering is the lever between a heat spreader and a heat shield.