| 10:46:04 AM PDT - Tue, Jul 30th 2013 |
|
Hi Ohlincha,
As part of the provision for analytical gradients in COSMO there were a couple of additional changes that fixed problems related to geometry optimizations:
1. The sum of COSMO charges formally should equal the total charge in the cavity. Previously the COSMO charges were "corrected" to force this relationship to hold. However, doing that violated the variational nature of the total energy resulting in incorrect gradients. Therefore this "correction" has been removed.
2. In geometry optimizations problems arise when a pair of COSMO charges get very close. This leads to inaccurate gradients and strange steps in the geometry optimizations. Setting Rsolv to 0.5 screens away COSMO charges that are closer together than 0.5 Angstrom and eliminates this problem.
So you are right that rsolv has a new default in the NWChem-6.3. This has been updated the documentation for NWChem-6.3 at http://nwchemgit.github.io/index.php/Release62:COSMO_Solvation_Model as far as I can see. Please let me know if you find an occurance that I have missed.
I hope this resolves any confusion about this issue.
Best wishes,
Huub
|