Gaussian calculation - PCM model for IRC compared to gas phase?

Question

Finding a reaction pathway thanks to Gaussian calculations is tricky exercise requiring a lot of resources for each step of this time-consuming process. Thus, man needs to wonder if each calculation has to be performed or not in order to gain as much time and resource as possible.

After achieving a 3-step mechanistic path for my reaction in gas, meaning finding all optimized structures, TS, verifications by frequencies of the stationary points and IRC, I now need to use the PCM model in my calculations to provide a solvent correction to my mechanism. However, I was wondering which calculations were required for this purpose.

From the 2 options I have in front of me, which is more consistent to avoid wasting time and resources ?

• Perform new optimizations for the minima and TS in my pathway using the PCM model. Then calculate frequencies to be sure of my stationary points.
• Perform new optimizations of the minima and TS AND perform new IRC calculations from the fresh new TS (after solvent correction) with PCM model.

The first option means that I perform an optimization from the reverse and forward structures (already further optimized) obtained from my IRC calculation in gas. The second option means that I calculate a new IRC to obtain new forward and reverse structures that I have to optimize again.

Answer 1

Option 2 is the most scientific and rigorous. You should re-run the IRC in solvent again because the reaction path can be different from what you got in the gas phase.

Now, will it be different in solvent? Unless you are dealing with some exotic system, different reaction path with implicit solvation is unlikely (provided you have the same TS and minima with PCM). However, the reaction barriers may significantly change, which can cause two subsequent steps to become one single step (the TS between them forming a shoulder in the IRC). I have seen this happen.

What matters though is that different reaction paths can happen, so you should run the IRCs in solvent if you want to be sure of your result (also because you want to publish it).