A major difficulty for treating excited complex systems arises from the different nature of the various competing excited electronic states, for example, localized neutral and delocalized charge transfer excitons, as a result of the relatively large length scale. Many theoretical methods aimed at targeting excited states can “in principle” deal with excitons (electron-hole pairs): density functional theory (DFT, with hybrid functionals), quantum Monte Carlo, Green’s Function methods, Coupled-Cluster, and so on, in a time independent or time-dependent framework. However, due to self-interaction error, DFT does not capture the correct 1/r asymptotic behavior of the electron-hole pair and thus requires special ad hoc corrections to properly treat excitonic structure and migration. Other many-body approaches are typically limited to small systems since the computational cost increases so rapidly with the system size. Reduced scaling local correlation methods require the mean-field (e.g., Hartree-Fock) calculations of the whole system as prerequisite, and the determination of appropriate orbital domains for excited states is not as trivial as that for the ground state.
Our group develops and implements novel electronic structure algorithms and scalable computational methods for excited states of complex system. Here we employ the bottom-up strategy by taking advantage of high level molecular electronic structure theories, with new idea drawn from other fields of fragment-based quantum chemistry, quantum embedding, renormalization group, statistical physics and machine learning algorithms.