Author Topic: Elimination of imaginary vibrational frequencies in COSMO optimisation  (Read 7651 times)

pandaaka

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
Hi,

I have performed a gas phase optimisation of a molecule, which has no imaginary frequencies. I have then taken this structure and reoptimised using default cosmo parameters (except for the dielectric constant which is set to 37.5). However, after numerical frequency analysis (using NumForce -central -ri -cosmo) I find several imaginary frequencies. This is not necessarily surprising, but if I use screwer to distort the geometry with respect to the corresponding modes and reoptimise, I end up back with my initial structure, no matter which of the imaginary modes I choose to use in screwer. this suggests that there are no energetic minima with similar geometries. Could the imaginary frequencies be an artifact of the model, and if so, can anyone suggest how I might eliminate them?

Thanks,

Andy
 

Hauke

  • Full Member
  • ***
  • Posts: 37
  • Karma: +0/-0
Re: Elimination of imaginary vibrational frequencies in COSMO optimisation
« Reply #1 on: February 18, 2013, 09:22:24 PM »

If you type in "NumForce -help" you'll get (at least if you are using TM 6.3.1) the following description of the "-cosmo" option
Quote
  -cosmo         using cosmo with frozen orientational part (recommended for freq. calc.)
                 The hessian (e.g. for optimizations) should be calc. without this option !!!
If you want to check if your structure is a minimum on the potential energy surface you are interested the Hessian, and therefore I would recommend to run NumForce -ri -central (and Numforce -ri -central -cosmo in a different folder afterward if you are interested in the frequencies)
I so far never had the case that one but not the other gave imaginary modes but maybe it helps in your case.

You could of course also try a finer grid (m4 or m5), and tighter convergence criteria, for example $scfconv 8 and jobex -ri -gcart 4 . Especially if you have heavier elements in your molecules also give weight derivatives in the DFT section a try; see the section on Prerequisites for the Calculation of Vibrational Frequencies in the manual.





pandaaka

  • Jr. Member
  • **
  • Posts: 10
  • Karma: +0/-0
Thanks for the reply Hauke, I thought that this question had been forgotten! I'll do as you suggest...