Hi, I’m trying to build a program that helps me optimize the cutoff radii of my pseudopotentials. I’m using Atom to calculate them. My problem is that I don’t know what to optimize. For instance, in the transferability test of atom I have 2 options, the eigenvalues or the cross-excitations energies. My intuition says that the eigenvalues are probably a better reference. Regardless of what I choose, I still have the problem that I’m not being able to give a score to the pseudo because there are a lot of numbers. For example for the d valence orbital.
grep '&v' test-temporary/OUT | grep d outputs
ATM4.2.7 19-AUG-21 O Test – GS 2s2 2p4 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.8 2p4.2 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.6 2p4.4 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.4 2p4.6 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.2 2p4.8 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.0 2p5.0 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s2 2p3 3d1 &v&d
3d 0.0 1.0000 -0.05294417 0.11943286 -1.83484294 &v
----- End of series ---- spdfg &d&v
ATM4.2.7 19-AUG-21 O Test – GS 2s2 2p4 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.8 2p4.2 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.6 2p4.4 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.4 2p4.6 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.2 2p4.8 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s1.0 2p5.0 &v&d
ATM4.2.7 19-AUG-21 O Test – 2s2 2p3 3d1 &v&d
3d 0.0 1.0000 -0.05294654 0.11911674 -1.37558792 &v
----- End of series ---- spdfg &d&v
I intentionally choose the d orbital because is shorter. This output is the Ae calculation eigenvalues at the top and the Pt eigenvalues at the bottom. I should calculate the difference between them and try to make it as small as possible. The problem is that there are 4 values and, also, in this case, I have only 1 test, but you see that there are 7 tests and orbitals s and p will have eigenvalues for all those tests.
Should I calculate the difference of all the eigenvalues and add it all together and then calculate a mean value between the 4 orbitals ? There’s a representative eigenvalue?
Thanks.