Apparent bug in Hartree-Fock energies for selected cases with diffuse basis sets


Click here for full thread
Forum Regular
The issue here is the handling of (nearly) linear dependent basis functions. The default threshold for linear dependence in NWChem is 10^{-5}, while it is 10^{-6} in Gaussian (and I assume the same in the other codes you mention). With the default threshold, NWChem flags 3 vectors as linearly dependent. Changing the threshold to 10^{-6}, no vectors are flagged as linearly dependent, and I get an energy of -230.72835824 Hartree. To change the linear dependence threshold to 10^{-6} add

set lindep:tol 1.e-6

to your input.