Dear Niri,
We have re-compiled with Intel 14, and now nwchem reproduces the QA test case SMD results. I have placed the config.log files for both our Intel 16 and Intel 14 installation here: https://github.com/tsenf/nwchem_SMD_compiler
Hopefully this will help you pin point the issue. We are also looking into compiling with Intel 15, with the debugging option turned on as suggested by Edo.
I have a few questions regarding how dispersion energies are reported in the final SMD results. When comparing DFT runs (outputs below) on a Py molecule with D2 dispersion turned on and solvation turned off, the difference in the reported "Total DFT energy" equals the dispersion correction as expected. When solvation is turned on, the difference in "Total DFT energy" equals the dispersion correction, however, when comparing the "total free energy in solvent" the energy difference equals twice the dispersion correction. Am I double counting the dispersion energy somewhere?
I have placed the input files for the following results here: https://github.com/tsenf/Py_Dispersion_SMD
DFT:
Total DFT energy = -248.280874615300
DFT+D2:
Total DFT energy = -248.286948144340
....
Dispersion correction = -0.006073528966
E(DFT+D2) - E(DFT) = E(Dispersion), as expected.
DFT+SMD:
Total DFT energy = -248.287389181435
.....
total free energy in solvent including G(SMD-CDS) = -248.2873891814
DFT+SMD+D2:
Total DFT energy = -248.293462707063
.......
total free energy in solvent including G(SMD-CDS) = -248.2995362360
.......
Dispersion correction = -0.006073528966
E(DFT+SMD+D2) - E(DFT+SMD) = E(Dispersion), as expected. G(DFT+SMD+D2) - G(DFT+SMD) = 2*E(Dispersion), unexpected factor of 2
Any help interpreting these results would be greatly appreciated.
Thanks and Best Regards,
Tom
|