No abstract

We critically re-examine conductance in benzenedithiol (BDT)/gold junctions using real-time DFT simulations. Our results indicate a powerful influence of the BDT molecular charge on current, with negative charge suppressing electron transport. This effect occurs dynamically as the BDT charge and current oscillate on the femtosecond time scale, indicating that a steady-state picture may not be appropriate for this single molecule conducting device. Further, we exploit this effect to show that a gate voltage can be used to indirectly control the device current by adjusting the molecular charge. Thus, it appears that transport in even this simple molecular junction involves a level of sophistication not heretofore recognized.

We analyze spin-charge separation in molecular wires using a combination of real-time density-functional simulations and model Hamiltonian calculations. By considering the ab initio electron dynamics of positively charged ͑C 50 H 52 + ͒ and negatively charged ͑C 50 H 52 − ͒ polyacetylene chains under a chemical potential bias, we are able to extract information about the mobility of electrons, holes, and spins in these molecules. Our results indicate that charges move more rapidly than spins in these molecules. We further supplement our ab initio data with empirical calculations employing the Pariser-Parr-Pople ͑PPP͒ model Hamiltonian. Our modeling indicates that the degree of spin-charge separation responds very strongly to the nonlocal exchange interaction, while showing little sensitivity to Coulombic forces. In particular, in order to reproduce the B3LYP results within the PPP model, it is necessary to reduce the strength of the exchange interaction by ca. 50% in the latter. We therefore conclude that many of the features present in the B3LYP spin current response are a direct result of self-interaction error in the functional.

We study real-time electron dynamics in a molecular junction with a variety of approximations to the electronic structure, toward the ultimate aim of determining what ingredients are crucial for the accurate prediction of charge transport. We begin with real-time, all-electron simulations using some common density functionals that differ in how they treat long-range Hartree-Fock exchange. We find that the inclusion or exclusion of non-local exchange is the dominant factor determining the transport behavior, with all semilocal contributions having a smaller effect. In order to study non-local correlation, we first map our junction onto a simple Pariser-Parr-Pople (PPP) model Hamiltonian. The PPP dynamics are shown to faithfully reproduce the all-electron results, and we demonstrate that non-local correlation can be readily included in the model space using the generator coordinate method (GCM). Our PPP-GCM simulations suggest that non-local correlation has a significant impact on the I-V character that is not captured even qualitatively by any of the common semilocal approximations to exchange and correlation. The implications of our results for transport calculations are discussed.