Ghost in geometry block with just ECP for DFT calculation
Clicked A Few Times
2:20:53 AM PDT - Fri, Aug 3rd 2012
Hi,

Is there a method to specify a ghost atom (Bq/XX) in the geometry block for a DFT calculation without having to put a basis set on it?

I would like to put a zero electron ECP on a Bq, and whilst this is permissible for HF it won't pass gridrtrunc in DFT mode due to this error:

------------------------------------------------------------------------

gridrtunc: no basis on ghost atom?        2

------------------------------------------------------------------------

Of course, a work around is to put some heavily localised basis function in on the Bq, but if I have lots of Bqs then I am just increasing my compute time just to find a work around...

Any input appreciated. Cheers, Andy

Forum Vet
2:44:11 PM PDT - Fri, Aug 3rd 2012
Andy,
A way to avoid putting a basis set on the ECP (and keep the DFT happy) is to put a tiny charge on the BQs as in the example below.
Let me know if this solution is adequate for you.
Cheers, Edo

start nw

geometry
O -0.9890 0.6203 0.0827
H -0.3665 1.1207 -0.4459
H -1.6444 1.259 0.3632
bq -0.67775 0.8705 -0.1816 charge 1d-18
bq -1.3167 0.93965 0.22295 charge 1d-18
end

basis
* library 6-31G*
end
ecp
bq library Mg "LANL2DZ ECP"
bq nelec 0
end

task dft

Clicked A Few Times
10:53:58 AM PDT - Sat, Aug 4th 2012
Hi Edo,

That works fine, thanks. Performance wise there is little difference between this and putting in the localised basis sets as I described above... but I can be much more confident about the influence an approach like you have described is having in a physical sense.

Thanks for all your help.

For anyone else who reads this, there are limits on the range for the "tiny charge".

For charge >= 1e-8, an error occurs due for a non-integral number of electrons
For charge <= 1e-21, the ghost atom error persists.

Andy


Forum >> NWChem's corner >> NWChem functionality