Geometry optimization with DFT-D3

Clicked A Few Times

6:59:11 AM PDT - Wed, Sep 2nd 2015

Hello everyone,

I ran into trouble doing geometry optimizations using D3-vdW corrections.

Examplary case: deprotonated Thymine

geometry noautoz noautosym
N       28.244018096  33.246369348  11.861178719
C       26.957713719  33.685034935  11.926387900
H       26.823102260  34.757893956  11.867853974
C       25.930007044  32.837449783  12.064908038
C       24.513432935  33.276081116  12.167416770
H       24.425223267  34.360732076  12.102674481
H       23.909994128  32.821313542  11.381993139
H       24.078438912  32.937259852  13.107471042
C       26.234367695  31.397900833  12.147964689
O       25.298061166  30.602638634  12.306878919
N       27.525290642  30.994407595  12.048677282
C       28.536759287  31.851707150  11.897827184
O       29.728553304  31.547621191  11.786332588
C       29.336853247  34.156507367  11.653137720
H       30.142174569  33.932485013  12.347562797
H       28.984684813  35.174718687  11.805840101
H       29.744704919  34.063888929  10.645354658
end

charge -1
scratch_dir temp
start

basis
 C library Def2-SVP
 O library Def2-SVP
 N library Def2-SVP
 H library Def2-SVP
end

memory 5000 mb

dft
 xc xpbe96 cpbe96
 disp vdw 3
 direct
 iterations 1000
 noprint "final vectors analysis"
end

task dft optimize

It does not crash as such but it just creates totally weird geometries that destroy the molecule. I noticed that already in the first electronic SCF, the estimated vdW Energies are very large.

    Dispersion correction =       -1.517049126987

I did a comparative run with Turbomole and there it is ~-0.02 and also the optimization finishes without problem.

Anything obvious I should change?

Cheers
B

Forum Vet

12:10:59 PM PDT - Wed, Sep 2nd 2015
Bob You have just hit a bug. As explained in the documentation, NWChem 6.5 does have vdw 3 parameters only for a restricted number of functionals -- and PBE96 is not one of them. Unfortunately, the code does not stops under the conditions just mentioned and keeps going, instead, with memory garbage used as vdw 3 parameters. http://nwchemgit.github.io/index.php/Release65:Density_Functional_Theory_for_Molecules#DISP_... Here is the patch needed to prevent this erroneous behavior http://nwchemgit.github.io/images/Vdw3_nwchem65.patch.gz

Clicked A Few Times

12:45:54 PM PDT - Wed, Sep 2nd 2015
Thanks for the explanation, Edoapra. I'm just a bit confused because in the documentation - which I checked before running - it says: Quote: Functionals for which DFT-D3 is available in NWChem are BLYP, B3LYP, BP86, Becke97-D, PBE96, TPSS, PBE0, B2PLYP, BHLYP, TPSSH, PWB6K, B1B95, SSB-D, MPW1B95, MPWB1K, M05, M05-2X, M06L, M06, M06-2X, and M06HF So there is PBE96 mentioned. Is it then available only after the patch or does the patch just throw a proper error? If so, where can I find a list of supported functionals? Thanks again! Bob

Clicked A Few Times

1:11:26 AM PDT - Thu, Sep 3rd 2015
Ok, I got it from looking at the patch. Do I see it right that one can now specify "pbe96" as a xc keyword. I wasn't aware of this, thought one had to make that xpbe96 cpbe96 line. Because when you do, the patched version indeed throws and error - but it runs through with the pbe96 keyword.

Forum Vet

2:35:37 PM PDT - Thu, Sep 3rd 2015
Thank you very much for catching this! The pbe96 parameters are there. I will modify the code accordingly (and the cpbe96 xpbe96 combination will work, too)

Forum >> NWChem's corner >> Running NWChem