<div dir="ltr">Hi, I doubt this is a DFTD3 problem. I couldn't check the input file yet (stupid computer problems) but my guess is more like it has to do with the cutoff in the calculations. By changing the cell dimensions you change the properties of the grid for the fft's. It is known in cp2k that low and even medium cutoffs will not be 100% translational invariant wrt the relative positions of the grid point and the coordinates. The problem has been discussed at some place in this forum, e.g. here https://groups.google.com/forum/#!searchin/cp2k/vibrational$20$20cutoff/cp2k/DVCV0epl7Wo/p-_KFKQaYnkJ Could you try to run with a higher cutoff or apply some of the smoothing options and report whether the problem remains? Cheers Flo On Monday, October 6, 2014 4:16:56 PM UTC+2, Leopold Talirz wrote:<blockquote class="gmail_quote" style="margin: 0;margin-left: 0.8ex;border-left: 1px #ccc solid;padding-left: 1ex;"><div dir="ltr"><div dir="ltr"> Dear all, a colleague in my lab found an unexpected dependence of the DFT-D3 dispersion energy on the amount of vacuum in a slab calculation, which could indicate a bug in the DFT-D3 implementation. Here the message from Sasha:I have a slab geometry for a slab that is ~20 A thick (less than 20) and I perforrm a fully periodic calculation. To test the convergence of the vacuum region i performed ENERGY calculations with a cell of 40 A (so 20 A of vacuum) and, afterwards, the same ENERGY calculation with a cell thickness of 60A (so 40 A of vacuum). I did define R_CUTOFF for the VDW as 8 A, i.e. less than half of the vacuum region. I was therefore expecting the dispersion energy to be exactly the same (?). The ENERGY calculations for the SAME geometry (file p.xyz) give rise to a difference of 1.8eV in the dispersion energy term for the DFT-D3 calculation, when i change the cell size. I repeated the same calculations with DFT-D2 and in this case there is no difference in the dispersion energy (as I would have expected). In the tar file you can find the input that I used for the four different calculations (DFT-D3 with 20 and 40 A of vacuum and DFT-D2 same two cases) as well as output files and the geometry p.xyz. Does anybody have an idea on what could be wrong? Thanks in advance for your help Kind regards, Sasha (and Leopold) P.S. I used cp2k revision svn:14377</div></div></blockquote></div>