5
$\begingroup$

I have troubles with obtaining spectra for selected molecules in different DFT functionals.

I calculate IR spectra for my target molecules (gold complexes with C/H/N/O/S/Au composition) to find how different levels of theory fit experimental data. In most cases everything looks fine, but for some unknown reason for two of them I have issues. In both cases the problem is the same: after geometry optimization is finished and I move to frequency job the Hessian appeared to have only -nan values:

Hessian of the SCF Energy
            1           2           3           4           5           6
    1        -nan        -nan        -nan        -nan        -nan        -nan
    2        -nan        -nan        -nan        -nan        -nan        -nan
    3        -nan        -nan        -nan        -nan        -nan        -nan
    4        -nan        -nan        -nan        -nan        -nan        -nan
    5        -nan        -nan        -nan        -nan        -nan        -nan
    6        -nan        -nan        -nan        -nan        -nan        -nan

And I get an error message after that:

    # Finishing finite difference calculation for IDERIV #
 ######################################################
 ######################################################

  ***ERROR*** Hessian Appears to have all zero or negative eigenvalues

For one molecule it is with B3LYP, for the other one with BPBE. Same molecules in other functionals work smoothly.

I had the same problem a few times for other complexes, but it was always resolved when I distorted the initial structure and re-performed optimization+frequency calculation. Calculation of frequencies alone with a given structure does not solve the issue either.

What might be the reason and possible solution for that?

$\endgroup$
8
  • $\begingroup$ -nan implies you might be dividing by zero or dividing by inf (both can give undefined results). Are you doing analytical frequency calculations or finite difference/numerical. $\endgroup$ Oct 25 '20 at 1:57
  • $\begingroup$ @TristanMaxson, This one is "Calculating analytic Hessian of the SCF energy". $\endgroup$ Oct 25 '20 at 5:03
  • $\begingroup$ My recomedation is to start with geometries near a minimum. I normally do a geometric optimization using Molecular Mechanics, then semi-empiric and finally DFT. Negative frequencies indicates that the geometry is far from equilibrium. $\endgroup$
    – Camps
    Oct 25 '20 at 22:17
  • $\begingroup$ Thank you for the suggestion. I will try semi-empirical first and see if it helps with further DFT job. $\endgroup$ Oct 26 '20 at 22:31
  • $\begingroup$ @DmitryEremin How did it go? Did the suggestion to try semi-empirical first work? $\endgroup$ Nov 29 '20 at 1:48