I was going through some old (digital) notes and saw one about MO calculation speed.
We go through the loop of calculating MO contributions for points in space that are far from a given atom (e.g. >8 Å). Most of these points naturally contribute negligible values to an MO surface.
I’m not going to merge this yet - it needs to be smarter.
The good news is that it seems to make calculating the LUMO of C60 about 2x faster.
I think the right strategy would be to loop through the basis exponents and angular momenta and calculate the distance threshold for S / P / D / F functions in whatever basis you used (e.g. did you use some diffuse S functions or something).
The other parameter that matters is the isosurface value (e.g., 0.03 or something else) … so I’d guess you want to ignore values that would contribute to 1/100 times smaller than the isovalue or a similar threshold.
Thoughts about “how far is too far?” when you can safely ignore orbital contributions for rendering?