Efficient computation of the second-Born self-energy using tensor-contraction operations

: Tuovinen R., Covito F., Sentef M.

Publisher: American Institute of Physics Inc.

: 2019

: Journal of Chemical Physics

: 151

: 17

: 0021-9606

DOI: https://doi.org/10.1063/1.5121820

: https://research.utu.fi/converis/portal/detail/Publication/44073739

In the
nonequilibrium Green’s function approach, the approximation of the
correlation self-energy at the second-Born level is of particular
interest, since it allows for a maximal speed-up in computational
scaling when used together with the generalized Kadanoff-Baym ansatz for
the Green’s function. The present day numerical time-propagation
algorithms for the Green’s function are able to tackle first principles
simulations of atoms and molecules, but they are limited to relatively
small systems due to unfavorable scaling of self-energy diagrams with
respect to the basis size. We propose an efficient computation of the
self-energy diagrams by using tensor-contraction operations to transform
the internal summations into functions of external low-level linear
algebra libraries. We discuss the achieved computational speed-up in
transient electron dynamics in selected molecular systems.ACKNO

JCP19-AR-02913.pdf