I want to calculate UV/Vis for an organic dye, and I found that increasing the "nstates" in tddft calculations, changes the value of first excitation energy,
for example when I change nstate from 20 to 50, first excitation energy shifts about 0.2 eV,
Is there any clear rule to set nstate in TDDFT calculations?
Yes, but do that carefully. Depending on the HOMO and LUMO character, you can have different ways to converge the results. By using Gaussian and Gamess (the attached figure, obtained by Gabedit interface - this should happen with NWChem also), I show you our results for glycine, which the excitation character is rather different to the Benzene I have mentioned above, so that you can use less unoccupied states. By looking again to these results, I can infer that the inclusion of gaussian functions to create the absorption curves plays a significant role for the position of the first peak. However, if the energy position of the first peak does not agree with the experimental data, you should consider if there is a charge transfer in your system as discussed earlier in the literature (see D. J. Tozer, J. Chem. Phys. 119, 12697 (2003)).
It depends on what software you are using and on what system (molecules or solids). I've been working on Light emitting polymers and, when I treated them as molecules(oligomers), I have to increase "nstates" until the position of the first peak stabilizes at a given value. For instance, by using octopus code on the excited states of benzene, I have to include up to 30 unoccupied states. Im some biological molecules, this amount was reduced to 20 ones. So, from these experience, I can only speculate that if you are dealing with pi-orbitals, you have to increase the amount of unoccupied states in order to have a good description of the lowest excitation peaks. On solids, I have found that you need less occupied states than in the molecular case.
Dear Professor Alves,
If I understand truly, we should check convergence of first peak by increasing nstate?
I am using GAMESS and NWchem to do this calculations,
and thanks for your consideration and good comments.
Yes, but do that carefully. Depending on the HOMO and LUMO character, you can have different ways to converge the results. By using Gaussian and Gamess (the attached figure, obtained by Gabedit interface - this should happen with NWChem also), I show you our results for glycine, which the excitation character is rather different to the Benzene I have mentioned above, so that you can use less unoccupied states. By looking again to these results, I can infer that the inclusion of gaussian functions to create the absorption curves plays a significant role for the position of the first peak. However, if the energy position of the first peak does not agree with the experimental data, you should consider if there is a charge transfer in your system as discussed earlier in the literature (see D. J. Tozer, J. Chem. Phys. 119, 12697 (2003)).
Dear Professor Alves,
Thanks for your kindly attention, I have seen similar results by increasing the nstates,
Also, I have another question about these figures,
In many papers they rescales their result to match the peak values with experimental results, what is the physics behind this?
Kindly please check my previous question about this.
https://www.researchgate.net/post/Are_the_absolute_extinction_coefficient_epsilon_values_obtained_from_convolution_of_TDDFT_results_physically_meaningful
thanks so much again for your time and consideration.
- Badges
- Science topic