Massively Parallel Simulations on Light-Induced Charge Transfer in Molecules

The development of environmentally friendly, inexpensive, lightweight solar cells would significantly enhance both sea- and land- based DoD operations and directly enhance global security through a reduction of competition for carbon-based fuels. For example, capturing one out of every 10000 photons from the sun would provide all the power the world currently uses. Approximately 75% of the solar radiation striking the upper atmosphere (1368 W/m² = 8555 eV/sec/nm² ) reaches the surface of the earth and most of this is in the form of photons with more than 1 eV of energy since water vapor and other atmospheric constituents effectively absorb energy below this threshold. In order to produce a one micron photovoltaic molecular film (composed of approximately 1000 layers of a molecule) with 10-20 percent efficiency, each molecule would have to create 1-3 eV/sec. Since the available solar radiation is in the 1-4 eV range, this implies ~1-3 electron-hole pairs per second which biological systems currently achieve. As such, various organic molecules, some of which are biologically inspired, have been proposed as alternative building blocks for solar energy materials. In the first phase of our challenge project, we have performed calculations on a molecular triad composed of a fullerene, a porphyrin and a carotenoid polyene. By calculating electronic structures, approximate excited states and respective dipole transition rates, we have simulated charge transfer dynamics in a collection of molecules exposed to an appropriate bath of solar photons. The resulting time constants associated with capture of solar radiation in the form of a charge-separated state have been determined. In the first phase of this challenge project, only electronic excitations were considered in the kinetic Monte Carlo modeling. Furthermore, the low symmetry in the molecular triad made it relatively easy to identify charge transfer excitations. In this phase of the project we report four new computationally intensive investigations that are aimed at validating and extending the formalism. First, to justify the use of an approximate excited state formalism, we have performed analogous calculations of excited states on a much larger selection of molecules and atoms and compared these results to experiment. Second we have performed calculations entirely analogous to our earlier molecular-triad results on highly idealized and high-symmetry fullerene tubules. Third we have used and extended a recently developed method to calculate all electron-hole-phonon interactions in these carbon nanotubes as well as the light-harvesting molecular triad. Fourth, in analogy to symmetry-breaking methods used for density-functional treatments of ferro- and ferri- magnetic ordering in molecular magnets, we have developed and tested a massively parallel computational method, employing coarse and fine-grain strategies, for identifying self trapped ferroelectric excited states in highly symmetric carbon nanotubes. These techniques allow us to determine charge-transfer excitations and to determine whether excited-states are to be viewed within a Franck-Condon or Marcus-Hush picture