Important issues facing model-based approaches to tunneling transport in molecular junctions

TL;DR

This paper critically examines model-based approaches to tunneling transport in molecular junctions, demonstrating that simple cubic current-voltage expansions are insufficient for accurate quantitative analysis at relevant biases, and emphasizes careful model parameter interpretation.

Contribution

It provides detailed analytical formulas, discusses model limitations, and highlights the importance of proper parameter estimation for interpreting tunneling transport data in molecular junctions.

Findings

Cubic expansions are inadequate for quantitative description at biases above transition voltage.

Uncritical parameter fitting can lead to invalid model interpretations.

Spatial potential profile effects are negligible at biases just above transition voltage.

Abstract

Extensive studies on thin films indicated a generic cubic current-voltage $I-V$ dependence as a salient feature of charge transport by tunneling. A quick glance at $I-V$ data for molecular junctions suggests a qualitatively similar behavior. This would render model-based studies almost irrelevant, since, whatever the model, its parameters can always be adjusted to fit symmetric (asymmetric) $I-V$ curves characterized by two (three) expansion coefficients. Here, we systematically examine popular models based on tunneling barrier or tight-binding pictures and demonstrate that, for a quantitative description at biases of interest ($V$ slightly higher than the transition voltage $V_t$), cubic expansions do not suffice. A detailed collection of analytical formulae as well as their conditions of applicability are presented to facilitate experimentalists colleagues to process and interpret their experimental data by obtained by measuring currents in molecular junctions. We discuss in detail the limits of applicability of the various models and emphasize that uncritically adjusting model parameters to experiment may be unjustified because the values deduced in this way may fall in ranges rendering a specific model invalid or incompatible to ab initio estimates. We exemplify with the benchmark case of oligophenylene-based junctions, for which results of ab initio quantum chemical calculations are also reported. As a specific issue, we address the impact of the spatial potential profile and show that it is not notable up to biases V somewhat larger than V_t, unlike at higher biases, where it may be responsible for negative differential resistance effects.