Efficient compression of first-principles electron-phonon interactions

First-principles calculations of electron-phonon (e-ph) interactions have advanced materials science and physics. Combined with Wannier interpolation of e-ph matrix elements, one can compute a wide range of e-ph physical effects in real materials, including phonon-limited transport, band structure renormalization, and superconductivity. However, Wannier interpolation remains a trial-and-error, material-specific computational step that currently hinders workflow automation and bottlenecks efficiency.

In this talk, we present a more efficient approach to parametrize e-ph interactions using a widely used compression technique, singular value decomposition (SVD). With proper improvements of the basic SVD scheme, we show a constrained SVD (c-SVD) method that achieves high accuracy for e-ph coupling and related properties (transport, renormalized band structure, superconducting Tc, etc) at significantly reduced computational cost compared to Wannier interpolation. This material-independent approach can systematically achieve a Pareto-optimal parametrization of e-ph interactions, leading to order-of-magnitude speed-ups. The talk will conclude by discussing future extensions to other interactions and incorporation of this computational tool in the Perturbo code [1].

[1] ] J.-J. Zhou, J. Park, I.-T. Lu, I. Maliyov, X. Tong, and M. Bernardi, PERTURBO: A software package for ab initio electron–phonon interactions, charge transport and ultrafast dynamics. Comput. Phys. Commun. 264,107970 (2021)