Im currently trying to find a stable geometry for a small organic dye for its first excited state.
# opt TD=singlets B3LYP/6-31+G(d,p)
One of my most possible structure guess is obtained from TDDFT geometry scan, but the structure is hard to converge when I simply run TDDFT/opt on it (l9999.exe). Even scf=xqc is not helping.
Somehow, I shifted whole TDDFT method to CIS, and all of them converged easily (scf=tight)
1.Can anyone please tell me what makes TDDFT/opt so hard to converge?
2.Is it appropriate that I replace my method from TDDFT to CIS for excited state geometry optimization? Just because TDDFT failed the convergence?
Mr. Po-Wen ChenMr Po-Wen Chen Chen,
th ere is not such relation.tendency
Hallo, Bryan!
The convergence depends on the molecule optimized. In some cases
TD DFT is better, in other CIS. The convergence depends on DFT functional
and basis sets as well. My personal experience with 6-31+G** is not
good.
In principle, both TD and CIS treatments should lead to the same excited states (maybe except the extra high ones) within the same DFT functionals and basis sets.
Best wishes
mb
Can you provide the input/output files?
Without an input and log file to look at, it's hard to say. Could be that the order of electronic states is switching as you try to do the optimization, or something completely different. TDDFT has a lot of issues, so again, without those files, we're simply guessing at what the issue might be.
Thank you Martin Breza for the information.
To Jared Hansen, perhaps Im thinking the issue too easy, will pay more effort on output file, thanks
There are many possible reasons for this, hard to say without further info about your system.
First of all, which is the character of the excites state you are following? How much does the TDDFT energy oscillates during the optimisation (i.e. does it go down slowly but constantly or does it jump). Same consideration for the geometry that you are updating: does it follow a specific trend or not? Also, the guess from which you do the two optimization is the same?
Eventually, CIS and TDDFT compute the energy and the forces in different ways, and what you observe is strongly system-dependent.
In general, you should find good benchmark studies for excited states to decide the + and - of each method. Excited-state geometries are indeed a delicate matter, on which I also worked. Give a look to this paper if you have no idea where to start
Article Ground- and Excited-State Geometry Optimization of Small Org...
There might be a problem with triplet instabilities. In that case the optimizations would fail during the response calculations. As a check, you might try to make a stability analysis, which would say the reference state is unstable.
To alleviate the problem you might try the Tamm-Danncoff-approximation (TDA). Thats what youre doing anyway, since CIS is equivalent to TDHF invoking the TDA.
For some more details see (especially 3.2.2)
https://pdfs.semanticscholar.org/5350/c15e51fccbfc8908c9e3458667ed92e7c11c.pdf
- Badges
- Science topic
- Similar topics
- Condensed Matter Physics
- Zero