Limitations

Before choosing an analysis, check what your file contains and what the method supports. Reading a format, running a method and validating its numerical results are different checks. Keep the warnings with your results.

• Wavefunction analysis needs the required basis, orbital and density records, regardless of format. The included IOData 1.0.1 adapter reads the 25 listed formats; available analyses depend on the records in each file. Ingestion validation is fixture-scoped, not a claim of universal coverage across program versions or every job type.
• The cross-format water checks compare files derived from one originating FCHK calculation. They test conversion consistency, not independent scientific agreement with external calculations or experiment.
• Some ORCA, Q-Chem and GAMESS interoperability fixtures are constructed program-output examples with geometry/energy fields; incomplete orbitals are not interpreted as a wavefunction. Broader native program coverage and independent reference corpora remain open work.
• Multi-record trajectories are not yet a sequence API: one record is loaded, and multi-frame XYZ is rejected rather than silently concatenated.
• Periodic cells, volumetric grids, and integral-only records can be ingested, but isolated-molecule frontier, population, and density analyses are not inferred from those components.
• Binary CHK files require Gaussian formchk and are not decoded internally.
• Gaussian real spherical 5D, 7F, 9G, and 11H shells are supported; pure spherical shells above H (l > 5) are rejected explicitly.
• ECP and ghost centers are handled with effective nuclear charges from the FCHK Nuclear charges record when present, and regression tests cover those semantics. This is source-faithful handling, not broad independent validation of every ECP family, basis set, or ghost-center workflow.
• For post-HF calculations, openWFN currently warns when an analysis is using the available SCF density. A warning naming the SCF density does not mean a correlated/post-SCF density was parsed or selected; post-SCF density support should not be assumed.
• Density integration and cube validation depend on the numerical grid. The default 0.15 bohr spacing is an accuracy/performance tradeoff and may be insufficient for tightly localized core density, heavy atoms, diffuse tails, or unusually demanding quantitative targets. Converge spacing and padding for the system being reported.
• Density integration and density/MO/NCI cube exports evaluate coordinates and fields in bounded chunks. APIs that return a complete scalar grid still retain that field in memory. Fine spacing, large boxes and large bases can remain expensive; the default molecular-grid safety limit is two million points.
• Molecular-summary bonds and fragments are inferred with a covalent-radius heuristic. This may not represent unusual coordination, transition states, stretched bonds, metals, or other nonstandard bonding situations.
• Gaussian-integral electronic and total point ESP is Validated for the documented eleven-case reference set. The explicit grid-based electronic/total ESP quadrature pathway remains Experimental and requires system-specific convergence.
• Embedded coarse workbench grids are visualization data unless separately converged and validated; they are not final quantitative integration evidence.
• Independent qc-iodata==1.0.1 comparisons currently cover parsed total energies and alpha frontier orbitals for six cases. They do not validate Mulliken, Löwdin, ESP, or density algorithms.
• Multiwfn same-wavefunction comparisons validate frontier orbitals, Mulliken charges, and Löwdin charges only for restricted water and unrestricted LiH; broader chemical coverage remains pending.
• Synthetic/special-case regression fixtures for ECP, ghost, UHF, ROHF, and post-HF behavior protect implementation contracts but are not a substitute for an independently reproduced validation corpus.
• XYZ files provide geometry only.
• Structure-only and periodic summaries are partial by design: missing effective charges, electron counts, orbitals, and energies are not guessed. A known zero effective nuclear charge identifies a ghost center; an absent charge remains unknown. XYZ/PDB/MOL/SDF exports cannot retain ghost/ECP effective charges and emit explicit loss warnings.
• Batch processing accounts for supported, partial, failed, and unsupported discovered inputs. An unsupported file makes the top-level batch fail, even if other files succeed. Format hints help with ambiguous filenames; they cannot make a damaged or unrecognized scientific record valid.
• Native vibrational spectroscopy currently targets the documented Gaussian harmonic frequency text-output grammar. Other programs, anharmonic jobs, arbitrary Gaussian output variants, thermochemistry, and experimental spectral assignment are outside the current typed parser scope.
• IR Inten is preserved as source IR intensity. Raman Activ is preserved as Raman activity; openWFN does not silently convert it to experimental Raman intensity because excitation frequency, temperature/population treatment, and convention-dependent factors are not known from the activity alone.
• Imaginary frequencies remain signed source data and are omitted from the broadened physical IR/Raman curve with a warning. Their physical interpretation is calculation-dependent.
• Normal-mode animation arrows require source Cartesian displacement vectors. Missing vectors are reported unavailable and are never synthesized. Workbench amplitude changes only visualization scale.
• The current spectroscopy fixtures, analytic broadening tests, and CLI/Python/MCP/report/Workbench parity tests establish implementation consistency but not independent scientific validation across methods, molecules, or program versions. Spectroscopy remains Experimental.
• Excited-state parsing currently targets documented source patterns from Gaussian, ORCA, and Q-Chem. The typed model is intentionally method-general, but recognizing and preserving a source state does not independently validate the underlying TDDFT/TDA, CIS, ADC, EOM, STEOM, spin-flip, multireference, core-excited, ΔSCF/MOM, local-correlation, or other quantum-chemical method.
• Excited-state source jobs remain distinct. Repeated root numbers in separate Link1/jobs/blocks are not globally flattened. Analyses that require one job need an explicit one-based job selector when more than one excited-state job is present.
• Missing oscillator strength or transition dipole remains unavailable; openWFN does not infer it. A negative source oscillator strength is preserved and excluded from the simulated optical curve rather than clamped. Dark f=0 states remain valid source data.
• The UV–Vis profile is a deterministic Gaussian oscillator-strength visualization in energy space, with a default 0.20 eV FWHM. It is not absorbance, molar extinction coefficient, solvent/instrument broadening, or an experimental line-shape model. The wavelength-domain array uses the energy-to-wavelength Jacobian rather than simply relabelling the x axis.
• NTO analysis requires a complete transition matrix with a supported convention and compatible orbital mappings. It is implemented for that restricted input scope, not for arbitrary printed amplitudes. TDDFT X/Y vectors, EOM left/right vectors, ADC/CI quantities and percentages are not interchangeable.
• The current excited-state fixtures, analytic UV–Vis/Jacobian tests, and CLI/Python/MCP/report/Workbench parity tests establish implementation consistency but not independent scientific validation across all programs, methods, molecules, or versions. Excited states and UV–Vis remain Experimental.

Always record the openWFN version, source-file checksum, grid spacing, padding, units, warnings, execution status, and validation status in research outputs.

Everyday-QC validation scope

MO cube, orbital composition, Mayer, DOS and PDOS are Validated for the documented eleven-molecule reference scope; that scope is not universal validation across all elements, high angular momentum, ECPs, correlated densities or programs. Löwdin partitions depend on AO representation; Mulliken fractions/PDOS may be signed. Mayer row sums are bonded-valence diagnostics, and correlated improved Mayer definitions are absent. DOS is finite-molecule orbital-energy broadening with one count per supplied spatial orbital/channel, without occupancy weighting or periodic bands. Failed source, conditioning, normalization or convergence diagnostics remain failed or partial/Experimental. Hirshfeld is validated for its separate H/C/N/O scope. Native typed Gaussian vibrational/IR/Raman workflows and Gaussian/ORCA/Q-Chem excited-state/UV–Vis workflows are implemented but remain Experimental pending broader independent validation. NTO, density derivatives, QTAIM critical-point searches, ELF/LOL and NCI are implemented with their documented restrictions and remain Experimental. QTAIM basin integration is not implemented. See the completion and gate record, vibrational spectroscopy semantics, and excited-state/UV–Vis semantics.