Periodic calculations with non-periodic solutions

Matt W MattWa... at
Wed Apr 22 15:00:19 UTC 2009


Just a couple of comments:

The Na semicore states seem to require about 600 ry cutoff - so any
data with much less than that is just noise.

Typically the noisest part of the cp2k calc is in the XC derivatives
(see the main quickstep reference) - using a different smoothing
function (e.g. XC_DERIV SPLINE3) might help.

Any code that uses realspace grids will have some "eggbox" effect,
i.e. the system may not be perfectly translationally invariant ... but
I shouldn't think this would be significant with your setup.

Maybe the very high cuttoffs make some of the default settings not so
good, perhaps someone else can comment more fully.



More information about the CP2K-user mailing list