I have a relatively symmetric structure. I ran an excited state freq/opt/freq/opt/freq process and at the end had a structure that looked good and had no imaginary frequencies. Later calculations for the ground state failed for various reasons.

So to restart my vibronic calculation, I did the following: I took the geometry optimized excited state structure, the one that showed 0 imaginary frequencies, and repeated just the frequency calculation. I used the same settings from the first time through. However, I was shocked that the frequency calculation ended up with 10 imaginary frequencies! I really don’t understand at all how this happened. I’ve double checked my input file and the coordinates I used for the optimized structure and they are correct.

How do I go from a structure where I am able to perform a frequency calculation with no problem, but then on repeating the same calculation, I get 10 imaginary frequencies?

Thanks,

Forrest

That is surprising. My questions are:

(1) Are you sure that your subsequent optimization is locating and moving on the same electronic state as the original one? Do you have any near-degeneracies in the excitation spectrum?

(2) How large are the imaginary frequencies and what is the nature of the corresponding normal modes?

Hi John,

Thank you for your input.

(1) Are you sure that your subsequent optimization is locating and moving on the same electronic state as the original one?

I believe so. I’ve triple checked that the optimized geometry I used were from the coordinates found from the first set of calculations where everything just seemed to work.

Do you have any near-degeneracies in the excitation spectrum?

I don’t believe so, but I may not be properly “calibrated” on this. Would # 5 and 6 be considered degenerate?

1 - 0.9563 (eV)

2 - 1.0906

3 - 2.0321

4 - 2.5488

5 - 3.159

6 - 3.180

7 - 3.433

8 - 3.684

9 - 3.742

10 - 3.816

(2) How large are the imaginary frequencies and what is the nature of the corresponding normal modes?

**The frequencies range from 509.56i to 4842.79i.

The 509.56i mode appears to be a kind of methyl partial rotation back and forth.

The 4842.79i mode appears to be 2 hydrogens on an aromatic structure and the hydrogens are “wagging” up and down out of the plane of the aromatic structure. The 2 hydrogens are moving up and down , but are effectively 90 degrees out of phase (while one is going up, the other is going down).

As an example…

John,

We would have to maybe confirm something about NDAs, but I may be able to send you some files offline.

Forrest

States 5/6 are definitely near-degenerate but that only matters if you are optimizing on one of those states. The magnitude of the imaginary frequencies is not consistent with numerical issues such as thresholds or grids; 4800 cm-1 looks like bond-breaking and even 500 cm-1 is too large to be a trivial artifact of a flat potential surface. Something is wrong and I don’t think I can diagnose it without additional info. If you’re in contact with Q-Chem staff then I suggest reaching out to them.

I will contact Shannon. I’ve communicated this to her previously, but I did not realize that such a large imaginary frequency meant more than a flat potential surface.

Thank you.

Forrest

For future reference, the resolution of this problem was that an SCF calculation had failed to converge during the optimization, therefore the subsequent frequency calculation was being performed at a non-minimized geometry, hence the imaginary frequencies. It’s always a good idea to very (by inspecting the Q-Chem output file) that the optimization has successfully converged to a minimum.