Hi everyone,
When doing optimization and frequency computation on molecules, it's common to have imaginary frequencies, sometimes more than one, many times undesired.
Sometimes tight requirements on the optimization and scf convergence help, especially with fine integration grid.
However... I've witnessed some examples when all that is simply not enough. You end with one or two spurious frequencies you can't supress.
After asking many times, the usual answer I get is 'just ignore them, it's numerical noise'.
So my questions are: What's this numerical noise? What is its origin and its meaning? Is it really safe to ignore it? How to tell when it's nois and when is an unwanted mode?
I may be naïve, but if I get two undesired imaginary frequencies, that is two degrees of freedom I'm not correctly accounting for in the calculation of entropy, and thus the free energies I get from this computation are meaningless...
Can someone give some guidance on this?
Thanks in advance.