Abstract
The configuration interaction method has been widely used to calculate electronic excitations in nanostructures, but it suffers from a slow rate of convergence with the number of configurations in the basis set and from the inability to select a priori the most important configurations. The optimized configuration interaction method presented here removes the limitations of the conventional approach by identifying at the outset the configurations that are most relevant for describing electronic excitations. We show that the ‘best’ configurations are remarkably different from the configurations that one would expect on the basis of the single-particle energy ladder, and that a small, optimized set of configurations predicts excitation energies with accuracy comparable to that for much larger, non-optimized sets of configurations. This approach opens the way to a new generation of configuration interaction methods where the configurations are pre-selected using heuristic search methods.