Repository navigation
Verify: Quantum ESPRESSO checks only that the namelists and blocks exist, and marks almost no lines #349
Description
Found 2026-10-15 (Linux, test-912 build, silicon SCF input). Verify for Quantum ESPRESSO reports that &CONTROL, &SYSTEM, &ELECTRONS, ATOMIC_SPECIES and ATOMIC_POSITIONS are present and that the namelists close. In the listing it marks one line (line 29, the first atom) and leaves every keyword, the cell and the k-points unmarked. That makes it look as if nothing was checked.
Expected, as for ECCE-QM and MOPAC: per-line marks for every keyword line with a known keyword and a valid value, plus checks that nat matches the atom count, ntyp matches ATOMIC_SPECIES, every species in ATOMIC_POSITIONS has a pseudopotential, CELL_PARAMETERS is present when ibrav = 0, the K_POINTS form, and the ecutrho/ecutwfc ratio. Not release-critical.
Activity
ohlincha commentedon Oct 15, 2026
Agreed scope (2026-10-15), one piece of work after 2026-10-18:
1. Verify for QE, computed from the generated input plus the local UPF files:
natequals the atom count;ntypequals the ATOMIC_SPECIES entries; every species has a UPF file; CELL_PARAMETERS is present when ibrav = 0; the K_POINTS form is valid; the ecutrho/ecutwfc ratio is 8 for US/PAW and 4 for NC.2. Automatic choices in Theory Details. The dialog does not know the structure, so ai.qe resolves these at input generation and the values appear in Final Edit:
Automatic (from the cell). Gamma only for a molecule in a box, otherwise the 25 Å / 50 Å rule, depending on the occupations.Recommended for these elementstick box, giving the highest recommended cutoff of the elements present. Use the SSSP recommended values (embedded table), and fall back to the UPF header's suggested minimum (wfc_cutoff) for non-SSSP pseudopotentials. For Si the two differ: SSSP gives 30 Ry, the pslibrary UPF header 44 Ry.Not planned: per-field hint text in the dialog (it would need the structure passed to every Details dialog).
Help: help/src/qe-choosing-settings.md (2c38e6e) describes the same rules and must follow any change to them.
ohlincha commentedon Oct 15, 2026
Addendum (2026-10-15): the quiet hints are wanted after all, to be designed rather than dropped. Small grey text beside a field, shown only when the value differs from the recommendation, for example "recommended: 60 Ry (H)" or "suggested: 5×5×5 for this cell". The prerequisite is giving the Details dialogs the structure: elements, cell, atom count and charge. Today globals.py passes only the name, theory, runtype, counts and symmetry group. Design that once, for all codes. Likely users: QE, CP2K (cutoff, k-points, box) and GROMACS (box size against the cut-offs, time step against constraints). Unclear whether the molecular codes (ORCA, NWChem, Gaussian, MOPAC) gain anything. Decide per code.