Ab-initio Adaptive Density Embedding for Mesoscale Systems

Density embedding [1] relies on a divide-and-conquer description of the electronic structure of large systems splitting them into smaller interacting subsystems. It is emerging as a powerful electronic structure method for large-scale simulations of molecular condensed phases and interfaces [2]. However, due to limitations of the employed non-additive density functionals, to date density embedding has been limited to approach weakly interacting subsystems. Additionally and more severely, when a single subsystem is very large (as in the case of interfaces of mesoscopic size), the computational cost is dominated by one of the large subsystems resulting in little overall gain compared to a full-fledged Kohn-Sham DFT simulation. We will show that these problems can be elegantly resolved. We devised an adaptive density embedding method prescribing subsystem merging/splitting events whenever subsystems interact too strongly/weakly redistributing work and data in an efficient way[3]. We will also show that by making judicious use of orbital-free DFT as a solver for metallic subsystems, mesoscopic molecule-metal interfaces can be modeled with an accuracy that is virtually identical to a Kohn-Sham DFT simulation of the supersystem[4]. The resulting object-oriented Python implementations[5] constitute a black-box, flexible and efficient electronic structure software for mesoscale systems.