FAIRmat-NFDI / nomad-simulations / 32727898389

87%

Build Type

push

github

Committed by web-flow

Commit Message

QC1: `MolecularOrbitals` semantic and scientific cleanup

Introduce `MolecularOrbitals` as a standalone `PhysicalProperty` (no longer subclassing `ElectronicEigenvalues`), describing molecular-orbital eigenstates in an atom-centered AO basis, and expose it as a repeating `molecular_orbitals` sub-section on `Outputs`.

Schema:
- `value` holds the orbital energies (eigenvalues), mirroring `ElectronicEigenvalues`; `occupations` is a companion quantity typed as an `m_float_bounded` interval `[0, 2]` with a small numerical-noise slack.
- AO->MO coefficient matrix as `coefficients` (plus `coefficients_im` for complex wavefunctions), stored as `HDF5Dataset`, with a `basis_set_ref` to the atom-centered basis set.
- Per-orbital classification `role` (core/inactive/active/virtual/deleted) and `symmetry`, whole-set `kind` (canonical/natural/localized), and `spin_channel` for the spin representation.
- Frontier orbitals split into provenance `_parsed` fields and purely derived `_normalized` fields (`homo`/`lumo`/`homo_lumo_gap`): resolved only for explicitly `canonical` orbitals with an occupied/unoccupied boundary, with no fallback to the parsed values, and the gap taken from the derived pair so it stays consistent.

Normalization:
- Infer `n_mo`/`n_ao` from the coefficient matrix; validate coefficient shapes, per-orbital lengths, `n_mo <= n_ao`, and `spin_channel`.

Docs:
- Export `MolecularOrbitals`, add it to the electronic-properties vertical, and regenerate the schema docs.
- Document the `value`-vs-`Variables` convention (dependent quantity vs independent axes, overridden per subclass) in a new schema-development page and the `PhysicalProperty` base docstring.

Tests: coefficient-matrix (HDF5) handling and parametrized frontier-resolution, occupation-bound, and dimension-inference tests.

Follow-ups: #467 (gap as a derived `ElectronicBandGap` using `is_derived`), #468 (interval datatype with configurable slack and error/log mode).

Coverage Stats

160 of 167 new or added lines in 5 files covered. (95.81%)

1 existing line in 1 file now uncovered.

10129 of 11630 relevant lines covered (87.09%)

2.61 hits per line