GBSA model gave different energies if used with ParamFile or a hsd input.

[KStrutynski]

The GBSA method produces different energies if one uses ParamFile option or use hsd input in the dftb_in.hsd. The GFN2 parameters for water taken from Grimme github:

Solvation = GeneralizedBorn {
    ParamFile = "param_gbsa_h2o.txt"
}

or explicitly stating the contents of hsd file

Solvation = GeneralizedBorn {
  Solvent = fromConstants {
    Epsilon = 80.20000000
    MolecularMass [amu] = 18.00000000
    Density [kg/l] = 1.00000000
  }
...

The different approaches result in different energies for SP step of dftb_in.hsd.txt.

hsd.log:Total Energy:                      -27.0386890418 H         -735.7602 eV
txt.log:Total Energy:                      -26.9614063373 H         -733.6572 eV

The problem was spotted then explicit hsd produced exploding SCC after some steps of geometry optimization.

Tested on several versions, including (commit: 4334d64b, base: 23.1) and 23.1 from conda on mac.

It is not clear fore me which method of GBSA parameters are preferred (similar to issue 1279. As I understand two files should result in the same energies.