What are the differences between different the implemented flavours of Van der Waals exchange-correlation functional in SIESTA? Which of them is recommended to use?
All DRSLL, LMKLL and KBM consist in a reparametrization of a van der Waals Density Functional of Dion et al Phys.Rev. Lett. 92, 246401 (2004). However, VV (Vroy and Van Voorhis) is formally different. It consists of a non-local functional of both the density and the gradient (instead of just the reduced gradient).
We recommend to the use DRSLL functional. In this case, the implementation of Van der Waals Density Functional of Dion et al has probed to be very efficient. The method is described in G. Román-Pérez and J. M. Soler Phys. Rev. Lett. 103, 096102, (2009).