Accuracy flag for convergence criterion with SCC iterations

[gavinwinter]

Does the float specified with the accuracy flag in the set() function of the tblite calculator here exactly correspond with the units for energy convergence found in the --acc flag in the original xtb code here?

Thanks!

[awvwgk]

Yes the thresholds correspond closely to the ones used in xtb, however I'm not exactly matching them in every detail. As in xtb, a lower value for the accuracy results in tighter convergence thresholds and larger integral cutoffs.