Open-Shell Systems are Cursed

The-program-that-shall-not-be-named just won't handle them, it seems. Or something.

This here is basically as a log to keep track of what I've tried so far.

Goal: geometry optimization, followed by a frequency calculation. Your typical sequence of events.

The input structure is the optimized structure of a molecule that looks exactly the same except the charge is different by +1. Hence the open shell.

Things that I've tried from here (all of which failed):

• Pick a smaller basis set. Mind you, my current basis set is plenty small as it is. Then I tried using STO-3G and it still won't work. Semi-empirical methods do not apply because I have a metal in my system. Maybe I should try DFT instead of HF.
• Changing the initial structure slightly (minor bond length change). Nope.
• Screw the geometry optimization. Ran a single-point energy calculation (which worked), and a frequency calculation (which didn't).

Now I am trying to force the convergence, and added a couple of thousand of cycles to compensate. Let's see if that works, although I am not optimistic in that regard (the log file is not looking good).

I should note that the error isn't "I ran out of cycles/memory". More like "something weird's happening and so I keep repeating the same procedure - which doesn't work - until I run out of cycles".

Actually, I should also try shifting the MO energies. And turning off DIIS extrapolation. Maybe later, when the current run is finished.