# Element reference completion — Lab 0.14.0

Date: 21 September 2026. This release expands the current crystal table using
source-backed data and keeps every earlier geometry and optical input intact.

## Coverage

| Layer | Previous | Current |
|---|---:|---:|
| Selectable elements | 118 | 118 |
| Typed general-property fields per element | — | 15 |
| Populated general-property values | — | 1,544 / 1,770 |
| Complete selected reference crystal cells | 67 | 90 |
| Additional complete phase options | 0 | 2 |
| Elements with visible optical constants | 11 | 30 |

The 226 null property fields include 34 elements for which CIAAW assigns no
standard atomic weight; these are marked **not defined**, not zero mass or an
ordinary missing measurement. Other blanks mean this source snapshot does not
supply a value. All 118 identities and property records are present, but this
is not a claim that all properties of all elements have been measured.

The 15 fields are standard atomic weight, reported atomic mass, electron
configuration, electronegativity, van der Waals radius, first ionization
energy, electron affinity, oxidation states, reference state, melting point,
boiling point, density, category, discovery, and CPK illustration color.
The last is explicitly not optical color and never enters the optical model.

## Reference sources and their limits

[CIAAW Standard Atomic Weights 2024](https://ciaaw.org/atomic-weights.htm)
supplies dimensionless recommended values; isotope intervals and uncertainties
are preserved verbatim. [PubChem's periodic table](https://pubchem.ncbi.nlm.nih.gov/periodic-table/)
supplies the other reference fields. Both snapshots were retrieved on
21 September 2026 and are bundled with hashes. The compact PubChem export does
not consistently specify pressure, temperature, phase or measurement
uncertainty. Consequently source-reported thermal and density values are not
automatically certified ambient-pressure measurements. Explicit predictions,
helium/carbon/arsenic thermal cautions, and Lr/Ds/Rg/Og source conflicts are
retained. The old engine electron configurations are not overwritten.

New selected complete atomic structures are **He, B, N, O, F, P, S, Cl, Mn,
Ga, Se, Br, Te, I, La, Pr, Nd, Ra, Pa, U, Np, Pu and Am**. Raw CIFs,
experimental citations, metric parameters, full symmetry-expanded coordinates,
conditions, and source caveats are archived in `data/completion_*`.
These include alpha-boron's 12-site primitive motif, nitrogen's 8 sites,
sulfur's 32 sites, alpha-manganese's 29 sites, and alpha-plutonium's 16 sites.
The selected alpha-boron phase is an explicit allotrope, not an assertion
that it is the unique stable room-temperature form.

Graphite is an experimental-reference alternative to the existing diamond
carbon cell. Hydrogen has a separate **DFT-predicted C2/c phase at 300 GPa**,
from [Pickard and Needs (2007)](https://doi.org/10.1038/nphys625); this does not
increase experimental element coverage. The main phase-I hydrogen record
contains a sourced molecular-center lattice. Its orientational disorder does
not justify inserting fixed atomic orientations. The two molecular centers
are not represented as two hydrogen atoms.

The remaining 28 selected atomic-coordinate gaps are H, Pm, At, Rn, Fr, Cm,
Bk, Cf, Es and Fm through Og. For Pm/Rn/Cm/Bk/Cf/Es, the primary numeric-source
ingestion gap is distinguished from an unmeasured physical property. No
fictitious atomic basis is assigned to short-lived superheavy elements.

Two incomplete COD Pr/Nd transcriptions omitted the 2a sites, giving half
the correct site density. This release uses complete 2a+2c dhcp bases from
the independently corroborated records at the same metrics; the rejected
transcriptions and an explicit origin-shift comparison remain archived.
An obsolete chlorine prototype is likewise not silently used in place of the
22 K experimental molecular structure.

## Geometry and display

Conventional single-element cells are reduced only by translations that
actually permute every atomic site. The reduction verifies the lattice index,
site multiplicities, periodic round trips and unchanged number density. A
left-handed-input metadata error found during the independent audit was
corrected: the reported transformation includes the initial axis permutation.
No atomic displacement, guessed basis or forced symmetry is introduced.

The default crystal-motif view shows the full primitive basis for complex
and molecular crystals, and local neighbors for simple structures. Local,
primitive and 27-cell views remain available. The local cutoff is 1.10 times
the shortest site distance from basis site 0; lines are distance guides, not
chemical bonds. Inequivalent sites may have different environments. The full
lattice matrix and fractional coordinates are inspectable. A primitive cell
is a repeating unit, not the smallest stable isolated crystallite. Fractional
sites at opposite boundaries reconnect through periodic copies.

The 4D mode remains an explicitly defined embedding and mathematical
projection. These source coordinates do not establish a physical extra
dimension or a unique hidden trajectory.

## Optical expansion

The 19 additions are **Al, Bi, Cs, Ge, K, Li, Mg, Mo, Na, Nb, Pb, Pt, Rb, Rh,
Si, Ta, W, Zn and Zr**. Raw tabulations are pinned to refractiveindex.info
database commit `c5c2f188e848453def5970e347399d653df2ffc2` (4 September 2026).
The original 11 optical records are unchanged. Primary references include
[Aspnes and Studna (1983)](https://doi.org/10.1103/PhysRevB.27.985),
[Werner et al. (2009)](https://doi.org/10.1063/1.3243762), and
[Palm et al. (2018)](https://doi.org/10.1021/acsphotonics.8b01243).

Each accepted source spans 380–780 nm, avoiding extrapolation. The source
samples and inference methods differ: ellipsometry, reflectance or REELS
derived constants, prepared films, and a compiled experimental Rh tabulation
are labeled separately. Si/Ge samples are lightly doped. Scalar inputs for
anisotropic materials are not orientation-resolved dielectric tensors.
Identical duplicate ultraviolet rows in Nb/W were removed only from the
normalized interpolation arrays; original files and all differing values
remain unchanged. Conflicting duplicate wavelengths would fail ingestion.

The existing Fresnel/CIE pipeline reconstructs ideal smooth, opaque,
semi-infinite surface reflection under a 5772 K Planck approximation or D65.
It does not predict optical constants from a crystal shape. It omits oxide,
roughness, coatings, finite-cluster response and sample-specific uncertainty.
Changing the displayed phase does not create phase-resolved optical data.

## Verification and next work

All **259 unit tests passed**, including the 228 previous tests. Independent
raw-CIF reconstructions checked 10 light-region phase records (including the
two alternatives), nine middle-region cells and six heavy-region cells.
The middle audit additionally confirms 1,793 periodic site comparisons in the
integrated data. Independent scalar optics reproduced **7,619 reflectances,
15,238 interpolated n/k values and 38 Sun/D65 colors**. All hex colors matched;
maximum reflectance discrepancy was 6.67e-16. All 11 original optical records
and 67 original crystal geometries remained unchanged.

The actual view script passed **14,105 interaction and data checks**. Native
Canvas motif/comparison previews and the scientific poster were inspected.
Full browser CSS/layout QA was unavailable and is not claimed. The complete
JSON, raw inputs, scripts, independent audits, source hashes, old results and
previous paper remain in the runnable package.

Next: retrieve the remaining primary numeric crystal records for Pm, Rn and
transplutonium metals; then add phase-matched optical constants and uncertainty
before comparing reconstructed colors against independent measured spectra.
This release compiles external measurements and performs checked computation;
it adds no new physical measurement and does not close earlier detector,
fixed-grid, identifiability or dimensional-evidence gaps.