How to create Dummy centres with a grid but no basis
Clicked A Few Times
3:30:41 AM PDT - Thu, Oct 9th 2014
Hi Guys,

Is there a simple trick to getting a dummy centre in an NWChem DFT calculation that has a density grid but does not have an active basis?

If I put an XX centre in my calculation, I get a centre with no grid, and if I put in a BQ then a basis is expected by default.

Thanks in advance for your answer (and apologies if this is a duplicate question).

Andy

Clicked A Few Times
5:25:33 AM PDT - Thu, Oct 16th 2014
Hi,

Any ideas on this? I am aware of the no_prune option to make sure that density points are not clipped off if I want to keep them in the calculation, but as yet have found no way to include extra grid points.

The reason I ask is I have programmed frozen density embedding into my SVN version of the code (energy and potentials only at this point, no gradients) and I wanted to check that the exchange correlation for a dimer model matches that of the same system divided in to two parts, including the non-additive contributions. Same goes for Ts, which I have trivially coded up.

Also, why I am on this note, I coded up the Thomas-Fermi Ts functional based on the Dirac XC routine, but notice there is an additional factor of two in the 2nd and 4 in the 3rd derivatives. Can anyone tell me where this originates from? It doesn't seem to fall out when I do the derivation but perhaps I'm making a schoolboy error....

Thanks,

Andy

Clicked A Few Times
8:14:09 AM PDT - Tue, Oct 21st 2014
I don't see how this could be done, as NWChem doesn't allow creation of neither ghost atoms without basis (gridrtunc: no basis on ghost atom?) nor real atoms with zero charge (grid_atom_type_info: non-bq center with zero charge).

You might try and see what happens if you just skip the "call errquit" statement when a "non-bq center with zero charge" is found. I haven't tried it, so it might be a silly thing to do and perhaps breaks things, though :-D The check is found in src/nwdft/grid/grid_atom_type_info.F

If skipping the quit causes no problems, then perhaps a geometry/basis spec like this could work:

geometry units au
   O1     0.00000000    0.00000000    0.00000000
   H      0.00000000    1.43042809   -1.10715266
   H      0.00000000   -1.43042809   -1.10715266
   O2     2.00000000    0.00000000    0.00000000 charge 0
 end
 basis
   H  library 6-31g
   O1 library 6-31g
 end

Clicked A Few Times
6:57:09 AM PDT - Wed, Oct 29th 2014
Thanks for the suggestions!.... but....

Quote:Mpjohans Oct 21st 8:14 am


You might try and see what happens if you just skip the "call errquit" statement when a "non-bq center with zero charge" is found. I haven't tried it, so it might be a silly thing to do and perhaps breaks things, though :-D The check is found in src/nwdft/grid/grid_atom_type_info.F



Unfortunately, after commenting out this and the other check that exists in src/nwdft/grid/grid_acc_def.F, there are insurmountable errors that appear, which look like they are due to bad indices allocations:
nga_get_common:cannot locate region: [1:9 ,14:22 ]

I also get this problem if I try to getting rid of the error checks for XO and XH species - the grids are defined, but eventually after some effort I get to the same error as above before the SCF cycle starts.

This makes me think this is not going to be an easy task to debug (not least as I don't know where this error propagates from!) so I will just stop and hope that my non-additive contributions are correct - at least for the functionals I have successfully compare the kinetic energy functionals with previous benchmarks for the TF and VW implementations.

Thanks again for yourhelp. If anyone else thinks of an alternative smart solution please let me know!

Andy

Forum Vet
5:52:22 PM PDT - Fri, Oct 31st 2014
Andy
I have created a patch that you can apply to nwchem 6.5 to be able to have a grid on bq's
The URL for the patch is
http://nwchemgit.github.io/images/Bqgrid.patch.gz
In order to be able to have grid on the bq's,
please add the following line to your input file just before the task line
set dft:bqdontcare .true.
Please be aware that this modification has not gone through the same thorough QA testing we use in a released version

Here is an example of a small input file
geometry noautoz
 h  0. 0.  0.5
 h  0. 0. -0.5
 bq 0. 0.  0.
end

basis
 * library 6-31g*
end

set dft:bqdontcare .true.

task dft


Forum >> NWChem's corner >> NWChem functionality