problems with periodic MP2

[Matt W]

Hi Katarina,

it is not the MP2 that is causing the problem (yet), but the Hartree Fock 
calculation that proceeds it - yours starts to go badly wrong at step 5 of 
the SCF loop:

     1 OT DIIS     0.15E+00  224.1     0.05110270      -406.0514831350 
-4.06E+02
     2 OT DIIS     0.15E+00   12.0     0.03105298      -419.6718508407 
-1.36E+01
     3 OT DIIS     0.15E+00   12.0     0.31947531      -423.6123120015 
-3.94E+00
     4 OT SD       0.15E+00   12.2     0.58812263      -425.0262271621 
-1.41E+00
     5 OT DIIS     0.15E+00   12.0   776.49566382    -38869.7170605659 
-3.84E+0

It seems you chose a bad preconditioner for OT, or maybe your integral 
setting are a little loose. Still, a best strategy would be to run a quick 
DFT calculation to get a reasonable wavefunction to start with, then start 
the HF/MP2 calc with that.

The MP2 can only work on top of a well converged Hartree Fock calculation.

HTH,

Matt

On Wednesday, May 11, 2016 at 7:22:29 AM UTC+1, katarína stančiaková wrote:
>
> Dear CP2K users,
> I am trying to do a single point calculations of a zeolite system using 
> MP2. However, after 300 steps SCF run did not converge (which itself I do 
> not consider as a problem), but the total energies does not seem right for 
> me.  They are in order of magnitude 10 000 a.u. and the fluctuations 
> between iteration steps are of the same magnitude. I expect the total 
> energy of the system to be around 450 a.u., which is a way lower as those 
> numbers. As I have no experience with MP2, I have not a very much idea what 
> can be wrong with my setup, or, if this kind of behavior is normal for MP2. 
> I did the same calculation, but with the water molecule in the given unit 
> cell, and there everything seems to be ok.
>
> I have attached the input and output file.
>
> Thank you for any comments!
> Katarina
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20160511/827aa54f/attachment.htm>