Heat travels through rutile germanium dioxide 1.46 times faster along one crystal axis than another. Along [001], 47.5 W/m/K. Along [110], 32.5 W/m/K. This anisotropy — directional dependence of thermal transport — is expected in crystals with asymmetric unit cells. What isn't expected is how it depends on temperature.
Standard phonon transport theory predicts thermal conductivity falls as T^(-1) — the hotter the crystal, the more phonon-phonon collisions, the lower the conductivity. Rutile GeO2 falls as T^(-1.4). The steeper decline means something beyond three-phonon scattering is at work. Four-phonon processes — where four phonons interact simultaneously rather than three — contribute additional scattering that increases faster with temperature than the three-phonon channel alone.
Note (2026-11-15; read against the paper's full text): the paper does not attribute the T^-1.4 dependence to four-phonon scattering. It says the effective exponent can reflect the combined influence of multiple scattering mechanisms and anharmonic effects, including thermal expansion and higher-order phonon-phonon scattering, and cites four-phonon scattering reducing thermal conductivity in other solids as an example. Its measurements run from 80 to 350 K; I found no statement about 100 K, and what it reports for cooling is that depopulation of high-frequency phonons reduces the anisotropy (the cumulative conductivity saturates near 8 THz at 80 K versus about 12.5 THz at 300 K, and the two directions look much more similar at 80 K). The essay's 'below about 100 K' threshold, its 'only the low-frequency phonons survive' and the device-layout implication are the essay's own.
The anisotropy itself has a dual origin. Phonons traveling along [001] are faster (higher group velocities) AND live longer (longer lifetimes) than those along [110]. In many materials, one effect dominates — either the speed or the lifetime sets the conductivity. Here, both contribute. The two effects are not independent: the crystal structure that creates faster propagation paths also creates less-scattered propagation paths, and both happen to align with the [001] direction.
Upon cooling, the anisotropy shrinks. High-frequency phonons — which contribute to the anisotropy — depopulate at low temperatures. Below about 100 K, the thermal transport becomes increasingly isotropic as only the low-frequency, long-wavelength phonons survive.
The practical implication: rutile GeO2 is being developed for power electronics, where self-heating kills devices. Knowing that heat escapes 46% faster in one direction means device layouts can be oriented to exploit the crystal's thermal highway. The material property that matters for engineering isn't the average conductivity but the maximum conductivity along the best direction.