Introduction

Density functional (DF) theory is widely used in materials and nano science because it offers an efficient approach for a computational characterization of structure and response. In its first-principle formulations it has no adjustable parameters and can thus work also in a a predictive mode, in the absence of prior measurements. In traditional semilocal DFT the descriptions were limited to materials having a dense electron distributions but failed, for example, to account for binding between molecules.

The Chalmers-Rutgers van der Waals density functional (vdW-DF) method provides an approach for crafting nonempirical DF that extend the reach of first-principle DFT also to systems with a sparse electron distribution, for example, as is found between two molecules. The existing vdW-DF versions systematically include also truly nonlocal correlation and can thus capture van der Waals forces at the same footing as DFT describes other types of interactions. The recent vdW-DF versions, such as the consistent-exchange vdW-DF-cx, yield a general-purpose materials description, giving accurate predictions also when there is a need to simultaneously describe binding in systems with a genaral electron disctribution.

The Roman-Soler algorithm for evaluating the nonlocal correlation energy functional of vdW-DF means a significant speed up because it allow us to relay on a FFT evaluation. At the same time it is clear that it makes it difficulat to reach very large systems size unless special attention is given to the implementation. Here we prsent a library that rests upon a highly scaleable FFT evaluation and hence allow us to ensure that the vdW-DF can be evaluated at very large systems sizes.