How can I overcome negative frequencies that occur during Gaussian calculations?.
There are several ways to achieve this. As a negative vibrational frequency (I guess this is what you mean with imaginary frequency) indicates a transition state, it should connect to local minima on the potential-energy surface of your system. The easiest way to overcome this is to elongate your atomic positions along the normal coordinates of your imaginary frequency. Do not elongate too far, as it is sufficient to get off the transition state within numerical accuracy (0.1 Angstroms is typically sufficient). You should go into both directions of the normal coordinate vector so that you can evaluate both minima. I would also recommend to reduce the trust radius for the optimizer, i.e. MaxStep=1.
Another way would be a random distortion of the geometry, i.e. changing the coordinates of all atoms in your molecule randomly by a maximum of 0.1 Angstroms and re-optimization of the resulting structure.
What could also work is to pick only the atom with the maximum displacement in the imaginary normal coordinates and disturb only that atom's position, and subsequent re-optimization.
Finally, if you have a symmetry constraint in your calculation (any symmetry except C1) you should check if the imaginary frequency is within a irreducible representation which, if removed, would lead to the next lesser symmetry. For example, if the frequency would run orthogonal to a mirror plane, which point group would result if you remove this mirror plane? Then, apply the less symmetric point group and run the optimization again. Sometimes this helps to solve the problem as well.
If none of the aforementioned advice helps, it would be quite helpful if you could give some more information on your problem, especially on the point group of your molecule and the irreducible representation of the imaginary frequency, unless you have only C1 symmetry.
There are several ways to achieve this. As a negative vibrational frequency (I guess this is what you mean with imaginary frequency) indicates a transition state, it should connect to local minima on the potential-energy surface of your system. The easiest way to overcome this is to elongate your atomic positions along the normal coordinates of your imaginary frequency. Do not elongate too far, as it is sufficient to get off the transition state within numerical accuracy (0.1 Angstroms is typically sufficient). You should go into both directions of the normal coordinate vector so that you can evaluate both minima. I would also recommend to reduce the trust radius for the optimizer, i.e. MaxStep=1.
Another way would be a random distortion of the geometry, i.e. changing the coordinates of all atoms in your molecule randomly by a maximum of 0.1 Angstroms and re-optimization of the resulting structure.
What could also work is to pick only the atom with the maximum displacement in the imaginary normal coordinates and disturb only that atom's position, and subsequent re-optimization.
Finally, if you have a symmetry constraint in your calculation (any symmetry except C1) you should check if the imaginary frequency is within a irreducible representation which, if removed, would lead to the next lesser symmetry. For example, if the frequency would run orthogonal to a mirror plane, which point group would result if you remove this mirror plane? Then, apply the less symmetric point group and run the optimization again. Sometimes this helps to solve the problem as well.
If none of the aforementioned advice helps, it would be quite helpful if you could give some more information on your problem, especially on the point group of your molecule and the irreducible representation of the imaginary frequency, unless you have only C1 symmetry.
Dear Andreas Funk,
Thank you very much for your reply. Your answer gives alot information regarding the negative frequency solvation. The initial assumption of my problem was Cs symmetry and the first frequency alone was negative.
Thank you...
I'm curious to how you performed your geometry optimisation. I find that when I'm working purely with organics e.g., the centre of a protein enzymatic active site, that I never run into negative frequencies. But if I include a transition metal ion cofactor then optimising becomes a bit trickier. I find that a gradual stage of optimisation really helps, for example in steps I would:
opt=(calcfc,loose) wb97xd/3-21+g
opt=(readfc,tight) int=ultrafine wb97xd/3-21+g
opt=(readfc,loose) wb97xd/6-31+g*
opt=(readfc,tight) int=ultrafine wb97xd/6-31+g*
etc... etc...
- Badges
- Science topic