# opt(maxsteps=500) TD(singlets,nstates=3) B3LYP/6-31G(d,p) scf=qc
-----------------------------------------------------------
I am currently trying to calculate first excited geometry optimization, and its quite different from ground state. In TDDFT, l9999.exe always happened, but opt=calcall seems to be not avaliable for TD mode. The optimization trajectory ends up like steep fluctuations, so RMS just never reached thresholds.
# opt(maxsteps=100) CIS/6-31G(d,p)
---------------------------------------------------------------------
Afterward, I change to CIS(singlets), and it is easier to reach energy RMS threshold and easy to converge as well.
Supposed that same compound of different excited state geometries were obtained from CIS, is the value comparison reliabe? at least comparison? And do I publish that?
Hi Bryan,
I assume you are using Gaussian. I suggest you restart the calculation using the geometry that corresponds to the lowest RMS with the keword opt(maxstep=15) or less (default 30). Do NOT mix up the keywords maxstep and maxcycles. By the way, G16 seems to support the opt=calcall in the TDDFT calculations.
The CIS and TD use the same linear-response scheme. The only difference is that the B3LYP uses the hybrid XC (20% HF with 80% fxc) while the CIS uses 100% HF. So the B3LYP will provides much more delocalization than the CIS does. This means the B3LYP will give better description of the excited state geometries of the organic pi-conjugated chromophores. On the other hand, the CIS or range-separated functionals are good at describing the excited state geometries of charge transfer compounds.
Hope it helps,
Chao
Thank you so much, Mr. Wang. I tried maxsteps, TD just likes to fluctuate in a range. Will check calcall in G16
Dear Bryan
You can also try turning off the symmetry use in the TD-DFT optimization. Moreover, you need more excited states in the calculation i.e. nstates=10; 3 states appear to be low for well convergence. Otherwise, the use of scf=qc is only recommended in extremely difficult scf cases, not in geometry problems.
Best regards!
Thanks, Diego Corté-Arriagada. Scf=qc is required for convergence in my case. The problem seems to be Max RMS/force never reached the threshold in TD, so i shifted to CIS.
- Badges
- Science topic