Molecular tunnel junctions based on mixed SAMs: exponential correlation of the average metal–HOMO coupling with SAM/metal work function

We report the transport characteristics of molecular tunnel junctions based on binary mixed self-assembled monolayers (SAMs) of aromatic molecules using the conducting probe atomic force microscopy platform. The molecules include terphenyl dithiol and two derivatives, 2',3',5',6'-tetrafluoroterphenyl dithiol and 4,4'-(bicyclo[2.2.2]octane-1,4-diyl)dibenzenethiol, whose junction conductances differ from terphenyl dithiol by factors of 10 and 100, respectively. The junction conductance <i>G</i> varies exponentially with binary SAM composition, not linearly, as might be expected. By employing an analytical model for the off-resonant current-voltage (<i>I</i>-<i>V</i>) behavior, we extract the average electronic density of state parameters for junctions with Au tip and substrate contacts as a function of binary SAM composition. The average HOMO to Fermi level offset <i>ε</i>_h is weakly dependent on SAM composition, reflecting commonly observed HOMO level pinning. In stark contrast, and in correspondence with the conductance results, the average metal-HOMO coupling <i>Γ</i> depends exponentially on composition, not linearly as simple composition-based averaging predicts. On the other hand, Kelvin probe measurements of binary mixed SAM-metal work functions reveal that work function (or work function change Δ<i>Φ</i>) varies linearly with SAM composition; it is a simple sum of the single component SAM work functions weighted by their relative surface coverages. Thus, <i>G</i> and <i>Γ</i> are both exponentially correlated with Δ<i>Φ</i>; variation of Δ<i>Φ</i> by 360 meV (∼7%) tunes <i>G</i> by >100× and <i>Γ</i> by a factor of 10, with smaller work functions (and larger Δ<i>Φ</i> values) leading to larger <i>G</i> and <i>Γ</i> values. Qualitatively, a correlation between <i>Γ</i> and Δ<i>Φ</i>, which both reflect interfacial charge transfer, appears reasonable, but a fundamental understanding requires a first principles model that can also account for the simultaneous pinning of <i>ε</i>_h.