| 4:23:38 PM PST - Tue, Jan 30th 2018 |
|
Chemhunter
The modification to the input file I suggested you increase greatly the radial size of the grid.
https://github.com/nwchemgit/nwchem/wiki/Density-Functional-Theory-for-Molecules#grid----n...
The number of radial grid point I have chosen are around three times larger than the xfine grid available in NWChem 6.8 (and earlier version). I did not suggest any change to the angular grid.
I reached the conclusion that the cause for negative frequencies ad to be traced back to the inadequacy of the radial grid after exploring a series of computational parameters that are involved in the evaluation of the DFT hessian. I had observed similar results with negative frequencies with some of my own calculations, but your test case were smaller and, therefore, better for debugging. The negative frequency issue seems more likely to appear the basis set has a combination of diffuse functions and high angular momentum functions (Therefore, once can get cleaner frequency by adopting a more modest basis set).
I have introduced a new grid (labeled as "huge") in the master branch. I am testing this new feature and it might be released as part of the 6.8.1 maintenance release.
PS Please do NOT try to use the input line "grid huge" in 6.8 or earlier release since a bug in the input code might make NWChem accept the input line without really adopting the huge grid itself.
|
|
|