Massively Parallel Fragment-Based Quantum Chemistry for Large Molecular Systems: The Serestipy Software

We describe the Serestipy software, which is an add-on to the quantum-chemistry program Serenity. Serestipy is a representational-state transfer-oriented application programming interface written in the Python programming language enabling parallel subsystem density-functional theory calculations. We introduce approximate strategies in the context of frozen-density embedding time-dependent density-functional theory to make parallel large-scale excited-state calculations feasible. Their accuracy is carefully benchmarked with calculations for large assemblies of porphine molecules. We apply this framework to a theoretical model nanotube consisting of rings of porphine monomers, with 12,160 atoms (or 264,960 basis functions) in total. We obtain its electronic structure and absorption spectrum in less than a day of computation time.