I am running excited state AIMD using the following $rem section:
At one specific time step, Q-Chem changes the number of roots to be calculated (initially just 2) and it prints the following string in the output:
NRoots was altered as:2 -->3
Calculating indeed 3 roots for singlet and 3 roots for triplet states.
I am curious to know the cause of this behavior. Is this something to be worried about?
I’ve noticed the same thing recently. It doesn’t happen often, but when it does I suspect it is because the root being added is close in energy to the last root requested (e.g., in your case, is root 3 almost degenerate with root 2?). My guess is that in this case root switching would make the convergence difficult, and increasing NRoots addresses the problem.
The TD-DFT excited state energies are not really sensitive to the NRoots, so as long as you are monitoring state 1 and 2 energies during the AIMD there should be nothing to worry about. However, since state 3 may be energetically close, it is not be a bad idea to run the full AIMD with NRoots=3 to monitor the behavior of the third excited state throughout the full dynamics run.
I haven’t had time to look at this specific case but I suspect Samer is right. A fairly recent change (v. 5.3 or 5.4) increases the number of roots in cases where MOs are degenerate. This was a result of some testing that indicated that the TDDFT solver could miss certain roots if one MO out of a degenerate pair was included in the guess but the other wasn’t, due to the user not realizing that there were possible degeneracies. Unless calculating that one extra state is a serious problem for your application, then there is nothing to worry about.
Thank you both for your help!