search

baoilleach / qmspeedtest

Quantum Chemistry speed test
BSD 2-Clause "Simplified" License
11 stars 10 forks source link

NWChem-6.3 B3LYP initial Hessian from restart #5

Closed mbanck closed 10 years ago

mbanck commented 10 years ago

The B3LYP geometry calculation has not been started from a clean slate. While the HF geometry calculation says "Using diagonal initial Hessian", the B3LYP says "Using old Hessian from previous optimization".

This is problematic and I assume that the results are skewed, I saw this problem as well when I got a suspiciously low number of geometry steps for the HF calculation and on inspection it turned out that the Hessian was also from a restart.

baoilleach commented 10 years ago

Drat. Sorry about this. Will redo with an empty folder. I'm not used to this sort of automagic behaviour.

mbanck commented 10 years ago

I fixed it on a branch (https://github.com/mbanck/qmspeedtest/tree/mbanck-nwchem-initial-hessian) on my repo (see above comment), but my git/github-foo isn't strong enough to figure out a pull request, as my repo (and my branch) contains also the additional input/output files from the other calculations I ran.

So maybe cherry-picking that commit (6cb7520) would be possible? Otherwise, let me know what I should do.

baoilleach commented 10 years ago

I've updated the input file. Now I need to redo the NWChem calculations...