Nanoscale device modeling using a conserving analytic continuation technique

Abstract

We propose an alternative approach to self-consistency and conservation laws in the theory of nonequilibrium Green's functions (NEGF's), which provides an infinite family of conserving but, generally, non-self-consistent approximations. Within any $\ensuremath{\Phi}$-derivable approximation the associated Born series for the NEGF is shown to be conserving. Expectation values calculated from the Born series are then used to build a Pad\'e table of approximations, while conservation laws are naturally preserved. We implement this technique for the $\ensuremath{\Phi}$-derivable self-consistent Born approximation (SCBA), for which we obtain a recursion relation that yields the Born series for the NEGF up to any desired order. The expectation values calculated from the Born series are then postprocessed to build a Pad\'e table of conserving approximations. The calculation of the SCBA photocurrent in a biased molecular junction model provides an example where, in addition to conservation laws, a substantial convergence acceleration relative to standard techniques is achieved. The present reformulation of the SCBA might aid convergence to the fully self-consistent results in a wide variety of problems.