A proper ballistic calculation of tunneling conductance for real junctions

Abstract

Employing an ab initio Screened Korringa-Kohn-Rostoker (SKKR) band structure method for a metal-vacuum-metal junction, we find that the tunnel conductance is different when it is calculated across the barrier and far from it. We attribute this difference to an artefact of the ballistic approach which overestimates the role of specular reflections, and its inability to pick up contributions from localized interface states. To reconcile the ballistic approach with experiment, we propose that the tunnel conductance should be calculated as if it is measured directly across the barrier. In this case the predicted tunneling magnetoresistance is larger.