Documentation for AtomicStructure.jl
AtomicStructure.jl (and the whole JuliaAtoms cluster of libraries) is research grade software. This means, that although we try our best to ensure that the individual components behave as they should, and we do have a fair amount of complete calculations that we compare against values from the literature, we give no guarantees. Furthermore, AtomicStructure.jl is not an expert system, in that if the calculation does not converge, it will not try another route (such as e.g. ATSP or Grasp would); it also does not adapt the grid to the problem, this has to be set by the user before the calculation starts.
Some things to check:
- Is the grid large enough? Too small grids will force all eigenvalues to increase.
- Is the grid spacing small enough? Too coarse grids will not be able to resolve the finer details of the orbitals, which may or may not be a problem, depending on your use-case (i.e. the energies may be good enough, but matrix elements between orbitals not, etc).
- If it seems that the optimizer cannot converge to the minimum, but instead just "cycles" a few energies, it could be worthwhile to try another line-search algorithm, e.g. More–Thuente instead of the default Hager–Zhang.
- Try with another basis set, increase or decrease polynomial order if using B-splines or FE-DVR, try with different knot-sets or placement of finite-elements.
- For point-charge nuclei, there is a singularity at the origin (for $\ell=0$) which needs to be accounted for (in terms of fulfilling the boundary conditions). For the finite-differences, CompactBases.jl provides one-point fixes that approximately satisfy the boundary conditions, for B-splines one can place the intervals densely close to the origin, for FE-DVR one can increase the polynomial order of the first element.