Possible origin of thermoelectric response fluctuations in single-molecule junctions

Abstract

The thermoelectric response of molecular junctions exhibits large fluctuations, as observed in recent experiments (e.g. Malen J A et al 2009 Nano Lett. 10 3406). These were attributed to fluctuations in the energy alignment between the highest occupied molecular orbital (HOMO) and the Fermi level at the electrodes. By analyzing these fluctuations assuming resonant transport through the HOMO level, we demonstrate that fluctuations in the HOMO level alone cannot account for the observed fluctuations in the thermopower, and that the thermo-voltage distributions obtained using the non-equilibrium Green's function method, arguably the most common method at practice, are qualitatively different than those observed experimentally. We suggest that this inconsistency between theory and experiment is due to the level broadening that is inherently built-in to the calculation method, and smears out any variations of the transmission on energy scales smaller than the level broadening. We show that although this smearing only weakly affects the transmission, it has a large effect on the calculated thermopower. To amend this discrepancy, we introduce the method of open quantum systems and use it to calculate the thermopower and its fluctuations. We account for both the magnitude of the variations in the thermopower and the qualitative form of the distributions, and show that they arise not only from variations in the HOMO-Fermi level offset, but also from variations of the local density of states at the contact point between the molecule and the electrode.