Non-real instabilities when using COSMO

[simon]

Dear all,
I recently discovered, that my calculations using the b3-lyp or bh-lyp functional and COSMO have non-real instabilities ($scfinstab non-real).(despite successful singlet excitation calculations)
Using better SCF and density convergence criteria and a better grid had no effect. The problem occurs in both Turbomole 6.6 and 7.2

Further testing revealed, that this instabilities do not occur, when a vacuum single point calculation on the same geometry is performed indicating a problem with COSMO.

Furthermore, I tested this behaviour using CH4 and THF, as test molecules, and obtained the following results: Both molecules also have this instability. In case of THF I used 1.0, 1.01, 1.1, 2.379, 7.52 and 36.64 as epsilon value.  Only the epsilon=1.0 calculation had no instability.

In case of methane, I tested the effect of basis set: I used the def2-TZVP, cc-pVTZ, aug-pVTZ and cc-pV6Z basis set in all cases. All tests revealed non-real instabilities. The imaginary/negative eigenvalue seems to increase with increasing the basis set.

This non-real instabilities consist of a  large to very large amount of relatively small  contributions. The coefficient of the contribution decreases, the number of contributions increases with increasing the system size. All contributions consist of low lying occupied orbitals and very high lying unoccupied orbitals(see example below (CH4 and def2-TZVP basis)).

I therefore wonder:
Is the non-real instability test compatible with COSMO?
Can the excitation energies, which correspond quite well  to the experiment,  trusted despite these findings?
What could be a cause and potential solution for this problem be?

Kind regards,
Simon                           

escf.out for CH4(def2-TZVP/COSMO(epsilon=7.52)):

Code: [Select]

1st a eigenpair



 SCF energy hessian eigenvalue:          -164.2170926565011   


                    WARNING! NON-REAL INSTABILITY DETECTED!


 Dominant contributions:

    occ. orbital  energy / eV   virt. orbital  energy / eV   |coeff.|^2*100
       2 a         -18.77          13 a          10.78           11.4
       2 a         -18.77          44 a          81.78            8.4
       5 a         -10.57          52 a         114.32            5.7
       4 a         -10.57          51 a         114.27            5.3
       3 a         -10.57          53 a         114.33            5.1
       5 a         -10.57          13 a          10.78            5.1
       4 a         -10.57          41 a          70.20            4.7
       5 a         -10.57          40 a          70.19            4.2
       3 a         -10.57          39 a          70.18            4.0
       2 a         -18.77           6 a           1.50            3.7
       4 a         -10.57          22 a          25.44            3.7
       5 a         -10.57          21 a          25.43            3.2
       2 a         -18.77          30 a          53.28            2.7
       3 a         -10.57          25 a          40.39            2.4
       5 a         -10.57          23 a          40.36            2.3
       3 a         -10.57          20 a          25.41            2.1
       4 a         -10.57          24 a          40.37            1.9
       3 a         -10.57          21 a          25.43            1.5
       5 a         -10.57          20 a          25.41            1.5
       2 a         -18.77          26 a          40.48            1.3
       2 a         -18.77          17 a          14.47            1.2
       5 a         -10.57          24 a          40.37            1.1
       2 a         -18.77          54 a         121.70            1.1
       5 a         -10.57           6 a           1.50            1.0
       5 a         -10.57          26 a          40.48            0.9
       4 a         -10.57          20 a          25.41            0.9
       3 a         -10.57          22 a          25.44            0.9
       5 a         -10.57          10 a           5.47            0.8
       3 a         -10.57          23 a          40.36            0.8
       4 a         -10.57          25 a          40.39            0.8
       4 a         -10.57          23 a          40.36            0.8


 

[uwe]

Hi,

it is a bit unfortunate that escf does not print a warning in such cases, but an instability calculation is a ground-state property. COSMO is implemented for "vertical excitations and polarizabilities for TDDFT, TDA and RPA" (to cite the documentation).

If you want to calculate ground-state properties with escf, please remove the $cosmo keyword before calling escf (see COSMO section in the manual, although this is usually done if the 'fast term' should be neglected).

I guess a closing remark on this issue should be given by an expert for TDDFT (not me)...

Regards,

Uwe

[simon]

Thank you for your answer.

Just to make sure I understand your answer correctly:

Including COSMO in an escf calculation makes only sense for a excitation energy calculation (e.g. $scfinstab rpas ), whereas for a stability analysis I should take the mos obtained from a COSMO SCF-calculation (e.g. DSCF) and perform a "vacuum" escf calculation on these orbitals by removing the cosmo keyword.
The non-real instabilities I obtained are therefore due to using the stability analysis in the wrong way.

Thank you in advance and kind regards,
Simon