Find transition states using Gaussian software: some tips

Here, I would like to summarize my past month's work about transition state searching with Gaussian program package.

1. QST2, QST3, and OPT(TS) are three good ways to identify TS structure. I personally prefer the OPT(TS) because it can generate the imaginary frequency as a criteria, while QST method can not. But in general it is really hard to find the TS structure with human guess and TS could be very sensitive to DFT functional, basis sets, guess structures, step size, etc. Anything wrong could give a big headache. So QST3 or QST2 is very useful for initial TS structure guess. And of course, after QST method you need run OPT-TS again to get imaginary frequency.

2. I like B3LYP/6-31+G* method. HF is good for an initial guess and fast calculation. But HF or BLYP (pure functional) can be totally wrong for TS guess, or totally miss the TS.

3. Sometimes you get imaginary frequency from OPT-TS, but be careful that you need run IRC reverse and forward calculations to confirm that this is a real TS structure. It could be totally wrong.

4. Some people suggested after you get multiple imaginary frequencies you could add the coordinate matrix of corresponding largest imaginary frequency multiplied by some factor (0.1-0.5) with the initial TS coordinate matrix. The new coordinate can be used to eliminate the imaginary frequency.

5. Final check for TS: After you run IRC forward and backward calculations, make sure that your TS energy is always the highest, i.e. if you draw energy profile pictures for both calculations, the energy curves should always go down. Otherwise, it may be something wrong. At this time, you should choose the highest peak point as initial guess for TS again.

6. If the electron spin multiplicity is not consistent before and after reactions, you have to specify two multiplicity values for OPT(TS). Also do the IRC testing too.