Hi everyone:
I've been struggling with some bimolecular TSs on gaussian.
The problem seems to be that they are too early or late, so they oscillate for days, even weeks, around a given structure and eventually land on reactants or products. I'm using the def2-TZVP basis set.
I wonder if there is a way to control or define the size of the "steps" for the optimizer. I have the feeling that they are too large for the systems I'm dealing with, so they oscillate eternally until they die. Opt=tight and Int=ultrafine have not helped.hanks in advance for your help.
Mauricio
Try to perform a frequency calculation of one of the lower energy oscillating structures. Usually the frequencies are positive and what happens is that the structure is oscillating on a minimum. Optimization continues indefinitely because the convergence criteria are not met.
Dear Daniel Moreno:
Thanks for your reply.
I've tried this. There are 8-10 imaginary eigenvalues depending on the structure I test. One of these frequencies definitely corresponds to the TS I'm looking for, but can't get rid of the rest.
That's why I wonder if it is possible for gaussian to optimize through "smaller steps"...
Man. Have you tried the old trick that used to work with MS8???
Taking some of the imaginary frequencies and rotating them some degrees, even to the point of eclipsed conformations, and then used that as input for a further optimization???
I've done that on MS8 and it has worked, and in Gaussian for structures with 1-2 imaginary frequencies, and it alsohas worked. It might take longer for your structures, but while we look for an answer, that might provide a little help.
Greetings.
What makes you so sure that you're very close to the TS you're looking for with one of the degrees of freedom? Eleven negative frequencies doesn't sound that close to me to a saddle point.
If I were you I'd put the coordinates in something like chemcraft and add a small random displacement to all of the positions of the atoms and start over with your optimization. If you don't want to do that, then you could try opt=newton. It'll take longer but it might discover a way out from where ever you're stuck.
You could also try using a different coordinate system like modredundant or z-matrix. It seems weird if you've never looked under the hood, but the way the optimization steps are done can be a lot easier (that is, converge more smoothly) if you use a different coordinate system.
Good luck!
Dear Oscar,
I tried it, but I can't get rid of some small imaginary frequencies. I've asked and the most common answer I get is 'numerical noise'. I guess it's time to open a new question. :-)
Thanks a lot!
El OGA!
- Badges
- Science topic