Time-dependent i-DFT exchange-correlation potentials with memory: Applications to the out-of-equilibrium Anderson model

TL;DR

This paper extends steady-state i-DFT to the time domain for the Anderson model, deriving memory-dependent exchange-correlation potentials and analyzing their impact on transient current simulations.

Contribution

It introduces a time-dependent i-DFT framework with memory effects for the Anderson model, including a reverse engineering method for xc potentials at various temperatures.

Findings

Time-dependent i-DFT potentials depend only on the first time-derivative of density.

Numerical simulations show the importance of history dependence in transient current.

Empirical extension of potentials below the Kondo temperature improves accuracy.

Abstract

We have recently put forward a steady-state density functional theory (i-DFT) to calculate the transport coefficients of quantum junctions. Within i-DFT it is possible to obtain the steady density on and the steady current through an interacting junction using a fictitious noninteracting junction subject to an effective gate and bias potential. In this work we extend i-DFT to the time domain for the single-impurity Anderson model. By a reverse engineering procedure we extract the exchange-correlation (xc) potential and xc bias at temperatures above the Kondo temperature $T_{\rm K}$. The derivation is based on a generalization of a recent paper by Dittmann et al. [arXiv:1706.04547]. Interestingly the time-dependent (TD) i-DFT potentials depend on the system's history only through the first time-derivative of the density. We perform numerical simulations of the early transient current and investigate the role of the history dependence. We also empirically extend the history-dependent TD i-DFT potentials to temperatures below $T_{\rm K}$. For this purpose we use a recently proposed parametrization of the i-DFT potentials which yields highly accurate results in the steady state.