Extreme IR Frequency of an Open-Shell Singlet Organic Cation with DFT

[emendezvega]

 
 Dear Turbomole Community,

 I am trying to compute the vibrational modes (IR) of an organic cation using DFT e.g. M06-2X and B3LYP // 6-311++G** in Turbomole 7.8.

 The closed-shell singlet (CSS) ground state and following excited states, triplet and open-shell singlet (OSS), lying ~0.4 and 0.5 eV higher in energy, were easily optimized within C2v symmetry.
 The energetic ordering nicely fits to our NEVPT2//CASSCF calculations and also to experimental photoelectron spectra.

 However, there is a vibrational mode in the OSS state (minimum) that is completely wrong, with an unreal frequency and intensity (see below). The mode is a CCC bending and should be below
 1300 cm-1. Consequently, the ZPE correction is affected and the ZPE-corrected energy of the OSS state, and relative instability, is shifted to much higher values.

                60        a1           3236.52             8.24940       YES     YES
                61        b1           3242.31             5.36910       YES     YES
                62        a1           3242.56             9.75005       YES     YES
 ----------  63        b1           5168.45    325387.74669       YES     YES -----------

 I also tried to optimize it in C1 symmetry with no success. This extreme frequency appears with both M06-2X and B3LYP. With other functionals the OSS state is a TS.

 Would you have any idea of how to proceed?

 Thanks in advance,
 Dr. Enrique Mendez-Vega
 Ruhr-Universität Bochum / OCII

[uwe]

Did you also try the Ahlrichs/Weigend basis sets like def2-TZVPPD or def2-QZVPD ?

If you have a doubly augmented basis set, there could be a lot of very small integrals which in the end sum up to a significant contribution. So increasing the integral threshold can be a test to see if that is a reason for the problem.

To increase accuracy (in general), first rerun energy and gradient calculations with:
- larger DFT grid (gridsize 5)
- converge the density in the dscf/ridft calculation before running aoforce ($denconv 1d-8)
- increase (or to be more correct: decrease the value of) the SCF integral threshold $scftol 1d-16
- make sure your gradients are very close to zero (they also contribute to the frequencies) by enabling weight derivatives in DFT and converge geometry sufficiently well

This might help if the problem comes from numerical noise. If not, you will not see any improvement...

If the frequency is so much off, also try to calculate the frequencies numerically with NumForce (just as a check).