BLYP water: recommended XC_DERIV and XC_SMOOTH_RHO smoothing

[Oliver Gittus]

Dear all,

I would like to simulate MD systems containing organic molecules (C,H,O 
only) and water using BLYP (doing some thermodynamic integration). I 
eventually intend to scale up to systems containing hundreds of atoms, and 
so am looking at a moderate cutoff of 400 Ry or so, in the NVT ensemble so 
no flexible cell. I cannot find a clear evaluation of the options for 
XC_DERIV and XC_SMOOTH_RHO. Are there any accepted combination(s)?

NN50 smoothing seems to be recommended for water in doi: 10.1063/1.3651474, 
but as far as I can tell, they do not specify XC_DERIV. Is smoothing 
required for XC_DERIV?

The smoothing was first introduced in doi:10.1016/j.cpc.2004.12.014. In 
this paper, S^10 – D^6(S^6)^(-1) performed quite well. Am I correct in 
thinking this corresponds to XC_SMOOTH_RHO NN50 (nearest neighbour 
smoothing operator S^50) and XC_DERIV NN6  (D^6(S^6)^(-1)). 

What are the qualitative differences between there different options for 
XC_DERIV and XC_SMOOTH_RHO?

Thanks,

Oliver 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20180505/01f396c9/attachment.htm>