A Local Variant of the Conductor-Like Screening Model for Fragment-Based Electronic-Structure Methods

Due to steadily rising computational power and sophisticated modeling approaches, increasingly larger molecular systems can be modeled with ab initio methods. An especially promising approach is subsystem methods, where the total system is broken down into smaller subunits that can be treated individually. If an implicit solvent environment such as the conductor-like screening model (COSMO) is included in the description, then additional environmental effects can be incorporated at relatively low cost. For very large systems described with subsystem methods, however, the solution of the COSMO equations can actually become the bottleneck of the calculation. A prominent example in this area is biomolecular systems such as proteins, which can, for instance, be described with frozen density embedding (FDE), especially the related 3-FDE approach. In this article, we present an alternative COSMO variant, which exploits the subsystem nature of the underlying electronic description and has been implemented in a development version of the Amsterdam Density Functional program suite (Adf). We show that the computational cost for the solvent model can be reduced dramatically while retaining the accuracy of the regular description. We compare several schemes for density and surface charge updates and assess the effect of the single tuning parameter.