Engineering heat transfer is critical for applications
in heat exchangers, semiconductor devices,
thermoelectrics and more. This demand motivates a
high-throughput computational methodology for
systematically designing materials with improved
thermal properties. One emerging method which can
rapidly and explicitly consider phonon-phonon
interactions even for highly anharmonic materials is
Compressive Sensing Lattice Dynamics (CSLD). In
fact, CSLD accurately predicts 2nd (harmonic) and
3rd+ order (anharmonic) interatomic force
constants (IFCs) with orders of magnitude fewer
Density Functional Theory (DFT) calculations than
conventional methods. However, doing CSLD can be
an exhaustive, many-step process requiring an
intimate knowledge of its sensitivity to parameters.
Consequently, this work implements an automatic
CSLD workflow capable of obtaining the thermal
conductivity of potentially thousands of materials,
benchmarks the stability of the workflow against
materials with a range of anharmonicity, and begins
constructing a dataset of thermal conductivity values
and phonon dispersion curves to be stored in the Materials Project for public use.