I currently trying to solve the problem of imaginary frequency. Simply, I open the log file showing one or two imag. frq. and i do manual displacemt at the higher absolute value of imag. freq. and reoptimize this new coordinate structure. I only succeeded with one structure after 3 trials. However, other structures failed to show zero imag freq, sometimes the number of imag freqs increases. As u know, the freq calculation of excited state is highly cpu-costive. My question can I predict if the reoptimized structure will show zero imag freq or not by comparing the (maximum force, RMS force, maximum displacement, RMS displacement and net change in energy) of the first and latter optimized structure ??
Do you suggest other well-reliable tried-before procedure to eliminate imag freq problem ???
Thanks so much!
If you run a geometry optimization, which is converged (abs(grad) = 0) then it means that you sit on a saddle point (one or more imaginary frequencies). Have you tried breaking the molecular symmetry by displacement before running the optimization? When you end up on a saddle point, that usually means that the geometry optimization started in highly symmetric configuration.
Sometimes, when you have potential energy surfaces along a certain coordinate (for instance, the one regarding the above mentioned vibration), that are rather flat, the minimization routine might not be efficient. In this cases, the best thing is to do a tighter convergency, and to use a nonlinear type of minimization routine. This usually settles the problem.
Also, keep in mind that sometimes, basis functions have a MAJOR impact in the correct prediction of minimas. So, sometimes, albeit more costly, more complete basis sets are better on these jobs, specially if you have, as stated before, shallow PES.
Last, but not least - the level of calculation. Albeit there may be a lot of debate on this issues, sometimes DFT vs MP2 optimizations give results that are contradictory.
So, the key here seems to be 'variety'.
Just my two cents.
Keep in touch. If you share the problem you are facing, and the structure, one might give you better help, or even collaborate with you.
Sorry Muhammad, it is not quite clear to me what you are trying to do and what it is that fails. You use 'excited state' a couple of times. In order to get excited state vibrational frequency information you will have to optimize the electronically excited state first (using B3LYP or whatever). A vibrational frequency calculation on the optimized ground state does not tell you very much (if anything) about the vibrations of the excited state.
Can you visualize the modes? Does the imaginary frequencies relate to methyl group rotation? What is (are) the molecule(s)? Do they have really high symmetry? Can you provide an input file?
- Badges
- Science topic