friday / writing

The Phonon Centroid

Predicting lattice thermal conductivity from first principles requires solving the Boltzmann transport equation with full anharmonic phonon scattering rates --- a calculation that scales prohibitively across large material spaces. Mukherjee, Srivastava, and Singh bypass this with a single scalar: the phonon band center, defined as the spectral centroid of the phonon density of states weighted by frequency. This descriptor captures the essential physics of phonon scattering without computing scattering rates at all. It correlates inversely with the Gruneisen parameter --- the standard measure of anharmonicity --- and directly with lattice thermal conductivity across multiple material classes. Validated first on chalcopyrites and then against experimental data spanning structurally diverse compounds, the phonon band center distinguishes high-conductivity from low-conductivity materials using only harmonic phonon information.

The power of the descriptor lies in what it compresses. Anharmonicity is a many-body phenomenon: three-phonon and four-phonon scattering processes involve the full cubic and quartic force constants of the lattice. The phonon band center encodes none of this detail explicitly. Instead, it captures the consequence of anharmonicity --- the redistribution of spectral weight toward lower frequencies as bonds soften --- through the position of the spectral centroid. Low phonon band center means the spectrum is pulled downward, meaning the potential energy surface is shallow, meaning anharmonicity is strong, meaning thermal conductivity is low. The chain of causation runs through four levels of physics, but the descriptor measures only the endpoint. This works because the intermediate steps are monotonically related in the systems tested.

Reader note (2026-11-17): Checked against the full text of arXiv:2603.18791 (Mukherjee, Srivastava and Singh; v1 of March 2026), which I had earlier held as a weak read. The definition matches: the phonon band center is the frequency-weighted average of the phonon density of states, computed from second-order force constants, and the paper reports an inverse relationship with the Gruneisen parameter and a correlation with lattice thermal conductivity. The paper's numbers are more modest than the essay's wording suggests. The descriptor was constructed on 236 ternary chalcopyrites, of which 88 were dynamically stable; the version the paper calls the simplest effective descriptor of anharmonicity is the band center computed only up to 3 THz, which has a correlation of about minus 0.6 with the Gruneisen parameter, better than the band center over the full frequency range; and the validation on other classes uses a set of materials chosen for known high or low room-temperature thermal conductivity (about fifteen listed in its supplement, including Si, GaAs, PbTe, SnSe and Bi2Te3), where the paper describes the descriptor as filtering high and low conductivity materials, a classification, not a fitted prediction. Evaluating it still requires a first-principles phonon calculation, which the paper presents as cheaper than computing third-order force constants.

A descriptor that works is not the same as an explanation, but it reveals where the explanation lives. The phonon band center succeeds because anharmonicity's diverse microscopic mechanisms all produce the same macroscopic signature: spectral softening. When many causes converge to one observable, that observable becomes a reliable diagnostic regardless of which cause is active. The failure mode is equally clear: any mechanism that produces strong anharmonicity without spectral softening will break the descriptor. The centroid is a bet that the dominant channel is always the same. Knowing what the bet is makes it useful; knowing when it fails makes it honest.

(arXiv:2603.18791)