Have you tried to distort the input geometry? Just look at the animation for that imaginary mode. Then, distort the geometry of the molecular site which gives rise to that mode, and optimize again. Generally it works.

I've already tried but i failed to obtain stable structure using gauss view. Is there any option to obtain stable structure?? If you know anything, please let me know

Maybe it's the basis set. Sometimes, using a smaller basis set may help to get rid of negative frequency. Another way you can follow is to do your initial calculations at a very basic level, ie MM or semi-empirical. Then you may employ higher detailed levels. Besides, using ultrafine grid option may help for DFT calculations.

I've already performed the optimization with opt=tight and int=ultrafine option but it didn't give stable structure. I'll try to perform the optimization with small basis set as you told me. Thanks ^^

It sounds like you have tried the right things, so it's a little hard to help further without some more details. What is the magnitude of the imaginary frequency? If it is very low, then you may get some success but increasing the convergence threshold of the energy in the TD calculation. I am not certain, but I believe that numerical frequency calculations use the single-point thresholds, which are quite low. You might try specifying (TD=Conver=6) or even 7, to tighten the convergence threshold.

If your imaginary frequency is quite large, then it is likely that your geometry is a transition state on the excited state surface. If I understood your earlier answers correctly, it sounds like you have tried distorting the structure along the normal coordinate for the imaginary frequency, and then restarting the excited state optimization? If you have done that, and it still isn't working, then it's hard to know what's going on without further details. One possibility is that you are at a conical intersection, and so your system has Born-Oppenheimer breakdown, and you cannot use single-determinant methods to address the excited states. Is your system Jahn-Teller active?

Thanks to your reply. Actually, my system is not jahn-teller active. Also, the imaginary frequency is below than 95 cm-1. it is small frequency value. I'll also try to calculate with tight convergence as you comment me. Thanks^^

To be completely clear, I was suggesting that you first *check* the convergence tolerance for the energy and wavefunction for the numerical frequency calculations. If I remember correctly, they are set to the same low values as for the single point calculations (i.e. 10-4 for the energy and 10-2 for the wavefunction). If I am correct about that, then it may produce different results if you repeat the frequency calculations using the tighter tolerances.