Value for H radius does not agree with other codes

[wavefunction91]

We take the value of the H radius (0.25 pm) from Slater, J, . J. Chem. Phys. 41, 3199, 1964. Pyscf and NWChem use 0.35 pm and ChronusQ uses the Bohr radius (0.529 pm, note factor of 2 in source). I have no strong opinion for what is "correct" here, but we need to expose the ability to make this flexible to ensure agreements between codes.

N.B. This only affects MHL for the time being

Thanks @ajaypanyala, @dmejiar, @elambros, and @aodongliu for pointing this out

[dmejiar]

I have no idea where the 35 pm value came from (there is a citation to J.C. Slater, Symmetry and Energy Bands in Crystals, Dover, N.Y. 1972, page 55, but that table has the same 25 pm as the 1964 reference). The Bohr radius value might come from Gill, Johnson, and Pople, Chem. Phys. Lett. 209, 506, 1993 as that value was used to compare SG-1 to MHL.

The H radius is not the only source of discrepancy between codes. For example, molgrid_defaults.cxx scales by 0.5 all non-H Slater radii but this is not done in NWChem and, if I am reading correctly, was also not done by MHL):

We have found that a good value for \alpha is the Bragg radius [21], although for neon a value is not given in this reference, we can extrapolate to give 0.45 A. This contrasts with Becke [12] who suggests one half the Bragg radius for a related parameter.

[wavefunction91]

Ah, OK, so the half came from Becke's paper, not MHL. Good to know, its been awhile since I've thought about this.

[susilehtola]

The Molpro definitions courtesy of @pjknowles are in https://gist.github.com/pjknowles/e25c5c12542da9a0fa25ad6cab4b459a

Looks like they use a Bragg radius of 0.25 Å for H and He, which appears to be in agreement with Slater's value.

That snippet of code also has tabulated values for the Mura-Knowles scheme, and the Treutler-Ahlrichs scheme, and Becke's modification of halving the Bragg radius discussed above.