I am trying to conduct a frequency calculation on an already optimized molecular geometry by changing its charge and applying an electric field.
For most field values I try, I get convergence errors by exceeding the maximum number of cycles. I've set the maximum cycles for SCF convergence at 1000 but it does not converge still.
I've faced many convergence errors for geometry optimizations and resolved them using quadratic convergence, level shifting and so forth.
Will the same techniques work for frequency calculations as well? Or is there anything else I should look into?
Thank you.
Dilhan
Dear Dilhan,
The following link contains some hints in order to overcome convergence problems when using G 09:
http://glab.cchem.berkeley.edu/glab/faqs/gauss_conv.html
Hoping this will be helpful,
Rafik
Dear Rafik,
Thank you sir for the link provided.
However, it provides suggestions to overcome convergence errors in geometry optimizations. I am getting convergence errors in a frequency calculation of an already optimized geometry.
Sometimes, it helps to think about the physical nature of your problem, first. The fact, that you change charge and apply an electric field seems to be some alteration of the "already optimized" original problem, unless you applied an electrical field before as well for the nonzero total charge. Are you certain, that for a nonzero charge an electric field of a certain strength can yield a convergence at all? Shouldn't the charged molecule start to accelerate or dissociate if the field strength is high enough?
Hi Dilhan,
I also think that as Herbert H H Homeier said above your system may be non-physical after changing charges/ adding E field. Therefore, it is better to check the convergence with more strict criteria and make sure your molecule achieve convergence. If the molecule is in local minima it should give you non-negative frequencies/frequency. Then proceed with electric field and check again for convergence. Because it may seems that your have convergence issue on the first place.
As Herbert suggested, your system is geometry-optimized only for a particular set of values (including charge and the lack/presence of external electric field) and changing any of them requires, in general, a reoptimization of the geometry. It's quite a normal thing that e.g. a molecule and its ion (different charge) have different geometries.
In general, however, any keywords helping in scf convergence may help here - unless at this particular geometry the scf won't ever converge for this particular charge and electric field, which also may happen.
Dear Herbert, Jeffrey and Bartosz,
Thank you for your valuable suggestions and sorry for the delay in response. I am doing a calculation using Koopmans' theory and therefore need to add a negative charge to the geometry optimized neutral structure. I need the electric field to move the frontier orbitals to degeneracy.
All these I'm doing with the approximation of minimal geometry change and as you've pointed out, it doesn't really have to be so!
I used Opt=VeryTight and ran an frequency calculation as Jeffrey suggested and it showed all frequencies as positive values. However, with that geometry, I was able to get convergence for many of the field values which I previously could not.
Thank you very much.
- Badges
- Science topic