Skip to main content
. 2023 Jun 19;9(1):106. doi: 10.1038/s41524-023-01033-4

Fig. 17. First-principles computational cost as a function of model’s size.

Fig. 17

Blue is the computational cost for computing second-order IFC using DFPT as implemented in Quantum ESPRESSO115,118,142, green is the computational cost for computing third-order IFC using the finite-differences as implemented in ShengBTE89 (interfaced with Quantum ESPRESSO), and red is the computational cost for computing anharmonic linewidths at q = 0 using Phono3py91 and a Gaussian broadening ℏσ = 2 cm−1 for the collision operator (with cutoff 3ℏσ). The calculation of the linewidths requires the second- and third-order IFC as inputs, and its computational cost is independent from the technique (DFT or interatomic potential) used to calculate the IFC.