The change of the radius doesn't necessarily improve geometry optimization convergence for systems containing Br atom, but as it solved your problem, it is acceptable. Using the SMD18 radius, even for PBE0, should be generally better (in the accuracy of energy aspect) than the original SMD radius.

Dear Prof. Lu,

I am calculating a toluene-solvated organometallic transition state that contain half-cleaved palladium-bromide bond. I was optimizing the geometry of this TS using PBE0-D3(BJ)/ma-SVP (PCM, toluene) level of theory, and observed oscillating behavior.

Out of curiosity, I modified the PCM radius of bromine atom to 2.60 by using modifysph command, which is what was done in Truhlar's SMD18 paper (https://doi.org/10.1002/chem.201803652).

Of course, what Truhlar did was optimizing the geometries of his compounds using M06-2X/def2-TZVP (def2-TZVPD for bromine). I knew this, but I still chose to optimize my compound using PBE0-D3(BJ)/ma-SVP (PCM, toluene) and only change the PCM radius of bromine.

By doing this, the TS optimization was successful without oscillating behavior. Do you think that this could actually be a generally better way to calculate bromine-containing molecules, even when using functionals and basis sets that are different from the original SMD18 paper?