disp dimer ch4 QA case needs update for dispersion correction

From NWChem

Viewed 1663 times, With a total of 3 Posts
Jump to: navigation, search

Gets Around
Threads 18
Posts 133
I was running a subset of the tests included in doNightlyTests.mpi, the ones that looked tractable on a laptop computer. The only one that notably deviated from reference output was disp_dimer_ch4. After a careful comparison of the current and reference output I believe that the deviation is due to a new correction factor incorporated in the calculation rather than a defective NWChem binary.

Here is the energy section currently in the reference output file, which was generated with NWChem 6.0:

Total DFT energy =      -81.003753211886
One electron energy =     -188.758861459904
Coulomb energy =       80.106514620115
Exchange-Corr. energy =      -13.739492494534
Nuclear repulsion energy =       41.390622275475


Here is the energy section in my local output file, which was generated with the 6.3 snapshot released on May 4 (svn revision 25564):

Total DFT energy =      -81.006289363570
One electron energy =     -188.758858931976
Coulomb energy =       80.106511730437
Exchange-Corr. energy =      -13.739492131429
Nuclear repulsion energy =       41.390622275475

Dispersion correction =       -0.002536153039


If I remove the dispersion correction from the total DFT energy, the total energy matches the reference to 8 decimal places. It appears that either the dispersion correction needs to be disabled in from the input file, if that is possible, or the reference output file needs to be updated to a run including dispersion correction.

Forum Vet
Threads 10
Posts 1632
Mernst
What NWChem source code have you been using and what compiler/platform?

Gets Around
Threads 18
Posts 133
I have been compiling/testing the revision 25564 snapshot: http://www.nwchem-sw.org/download.php?f=Nwchem-6.3.revision25564-src.2014-05-03.tar.gz

My platform is LINUX64.

I have built the code and run the tests with:

Debian Wheezy with gcc/gfortran 4.7.2
Ubuntu 14.04 with gcc/gfortran 4.8.2

On both platforms I am using OpenBLAS 0.28 configured with the 64 bit integer interface as my BLAS.

According to svn log, the reference output QA/tests/disp_dimer_ch4/disp_dimer_ch4.out was last changed at revision 21012 in August 2011.

It looks like these files started incorporating/reporting dispersion correction later than revision 21012:
src/nwdft/so_dft/dft_scf_so.F
src/nwdft/scf_dft_cg/dft_cg_solve.F
src/nwdft/scf_dft_cg/dft_roks_nr_solve.F
src/nwdft/scf_dft/dft_scf.F

The reference QA/tests/disp_dimer_ch4_cgmin/disp_dimer_ch4_cgmin.out was generated with a newer NWChem revision that incorporates/prints the dispersion correction, but QA/tests/disp_dimer_ch4/disp_dimer_ch4.out was generated using an older NWChem 6.0, old enough that it does not report revision number under job information.

Forum Vet
Threads 10
Posts 1632
The bug is no longer there in the current svn repository

Thanks for the feedback


Forum >> NWChem's corner >> NWChem functionality



Who's here now Members 0 Guests 1 Bots/Crawler 0


AWC's: 2.5.10 MediaWiki - Stand Alone Forum Extension
Forum theme style by: AWC