Number of shells generated by AutoAux function

[sunqm]

For basis set 6-31G** of H atom, manip.autoaux_basis produces two d shells:

manip.autoaux_basis(api.get_basis('6-31G**', 'H'))

...
    {'function_type': 'gto_spherical',
     'region': '',
     'angular_momentum': [2],
     'exponents': ['2.200000e+00'],
     'coefficients': [['1.0']]},
    {'function_type': 'gto_spherical',
     'region': '',
     'angular_momentum': [2],
     'exponents': ['4.840000e+00'],
     'coefficients': [['1.0']]}]}},

The 6-31G** basis has one p-shell with exponent 2.2 . The generated auxiliary basis set should contain only one d shell with exponent = 2.2 . The shell with exponent=4.84 seems caused by the rounding error in

basis_set_exchange/basis_set_exchange/manip.py

Line 901 in eacdd02

effective_exponents = [2 * k_value**2 / (pi * rexp**2) for rexp in rvec]

The effective_exponents is 2.200000000000002 in this case. If this number is 2.2 , the function only generates one d shell.

[susilehtola]

Maybe the BSE should contain some function to remove near-duplicate exponents? Not sure what the right fix would be otherwise.

[sunqm]

Or add a small penalty term here to tolerate certain rounding errors

basis_set_exchange/basis_set_exchange/manip.py

Line 969 in eacdd02

if current_exponent >= amax_aux[laux]: