excellent news. It will make my life much easier in the following weeks!
BTW, I didn’t get around pull-requesting the MN15 definitions into the master repo - I am waiting for some other things to come through, before continuing with this work. It’s still on my to-do list, though.
Btw, do you do any cleanup of libxc’s output? I just noticed that e.g. SCAN may still give out some NaNs. It is a horribly numerically unstable functional… It might make sense to set any non-finite entries in the energy/potential to zero.
Currently we only sieve second derivates or higher based on input density. Its not too much effort to do the same for lower level derivates as well, we just haven’t encountered that issue yet.