BSSE/ghost atom calculations with DFT-D3


Just Got Here
Hiya,

I'm getting some unusual results while computing the counterpoise correction for binding energies of a formic acid dimer. When including ghost atoms with the vdw 3 dispersion correction I get energies which are ~0.2 Hartrees lower than the same calculation without the ghost atoms! This is a much much larger difference than I expected. A similar computation with vdw 2 gives results that are much closer.

So I wonder if there is some issue with ghost atoms and the vdw 3 correction?

Attached is my input script if anyone cares to check it.

Thanks.

EDIT: forgot to mention this is version 6.3.




start 2hcoohII.vdwd3

geometry units angstroms print xyz autosym

  C                  0.13610421     1.91612312     0.00000000
H 0.20980015 3.01221077 0.00000000
O 1.13159962 1.19699064 0.00000000
O -1.12156508 1.51388537 0.00000000
H -1.16041424 0.51093133 0.00000000
bqC -0.13610421 -1.91612312 0.00000000
bqH -0.20980015 -3.01221077 0.00000000
bqO -1.13159962 -1.19699064 0.00000000
bqO 1.12156508 -1.51388537 0.00000000
bqH 1.16041424 -0.51093133 0.00000000

end

basis spherical
 H  library "DZVP (DFT Orbital)"
O library "DZVP (DFT Orbital)"
C library "DZVP (DFT Orbital)"
bqH library H "DZVP (DFT Orbital)"
bqO library O "DZVP (DFT Orbital)"
bqC library C "DZVP (DFT Orbital)"
end

dft
 xc b3lyp
disp vdw 3
end

task dft

Forum Vet
With NWChem 6.8 I am getting
-189.795307502748
-189.794085741354

Just Got Here
Thanks for checking. With 6.3 I have -189.971756 with the ghost atoms, a big difference! But your result seems sensible.

I get -189.7940857 without, which matches your result.

So I wonder which version this got fixed in? I think I can check with 6.5 on another machine.

EDIT: I match your result with 6.5! So, nothing to see here I guess :P Time to talk the supercomputer guys into upgrading from 6.3...


Forum >> NWChem's corner >> Running NWChem