[phosphorix]

I have got a question concerning low frequencies in Gaussian:

A small organic molecule (19 atoms) only containing C, H, O (input drawn and cleaned in GaussView 4.1) is subjected to a structure optimization (B3LYP, 6-311++G(2d,p)) applying a solvent model (SCRF). The optimization was conducted with tight convergence criteria and an ultrafine grid (keywords OPT=(gdiis,verytight) nosym int=ultrafine).

A subsequent frequency analysis shows the presence of several negative frequencies. Although the latter are not flagged explicitly as imaginary by Gaussian, negative values are never acceptable, especially not in such a case as described (light atoms only, small molecule).

I have searched the internet extensively for possible solutions, however, in many cases the answers boil down to

1. use tighter convergence criteria and a finer grid OR
2. visualize the negative frequencies, slightly move the atoms involved and re-optimize from this geometry OR
3. look for the atomic coordinates below the low frequency and manually alter the strongest-moving atoms' coordinates accordingly.

As I have already the tightest convergence criteria in place, the first point is not an option. The second point fails as log files created by Gaussian 09 cannot be visualized by means of GaussView 4.1 (why not, actually, and can that be fixed?). The third point does not really work as the output in my log file does not provide for coordinates for these low frequency vibrations - see the respective section of output:

####
.

Full mass-weighted force constant matrix:
Low frequencies --- -14.2377 -4.7523 -0.0008 -0.0006 -0.0005 8.2981
Low frequencies --- 45.0788 118.6498 224.1491
Diagonal vibrational polarizability:
31.1676073 115.6507661 21.3064080
Harmonic frequencies (cm**-1), IR intensities (KM/Mole), Raman scattering
activities (A**4/AMU), depolarization ratios for plane and unpolarized
incident light, reduced masses (AMU), force constants (mDyne/A),
and normal coordinates:
1 2 3
A A A
Frequencies -- 45.0175 118.6215 224.1479

.
####

Even applying different starting geometries does not change the overall picture. What can I do in this case to get rid of the low frequencies?

Thank you to everybody who might come up with a nice solution!

[Irlanur]

How can you have negative Frequencies? the frequencies are calculated from the square root of the diagonalized hessian, how can they be negative? imaginary ok, but negative. (I don't know Gaussian, I work with TURBOMOLE...) (that's a real question)

what you could try:
-what's the solution without the solvent model?
-use a better basis set!

what's the molecule after all?

[curiouscat]

I think they are imaginary. It just represents them as negative I thought.

Sometimes the low imaginary freqa, (say below 100 cm-1) can be neglected. Those are modes along spurious translations or rotations. As far as I remember. If you work on changing the geometry slightly & re-relaxing it might get rid of those modes but often the effort isn't worth it.

[rebeccabmhill]

As far as I understand, the "low frequencies" are not what you need to worry about being negative, the "Frequencies" are, and your frequencies are positive "Frequencies -- 45.0175 118.6215 224.1479".

[adhikary]

I think you can check the intermediate geometries and find the closest one you expect, make a very little change in it and use it as your input structure. Hope problem will be solved, but remaining very low magnitude (I usually consider up to -10) of imaginary frequencies can be avoided.