Data provenance¶

photonics-helper bundles a numerical material database (photonics_helper/materials.db) alongside the code. This page documents where that data comes from, the terms it is redistributed under, and how to audit it. The canonical notice also ships as NOTICE in the repository.

Code vs data¶

Component	Licence
`photonics_helper` source code	MIT (see `LICENSE`)
`photonics_helper/materials.db`	compilation of literature data — see below

Tables and provenance columns¶

Table	Rows	Contents	Provenance
`nk_data`	24 767	tabulated `n`, `k` vs λ (µm) for 12 materials	`source`, `citation` (24 760 / 24 767 populated), `license`
`sellmeier`	39	Sellmeier coefficients + validity window	`source` (39 / 39), `license`
`raman_specs`	44	Raman shift, linewidth, `f_R`, `n₂`, gain	`references` (44 / 44), `license`
`phonon_modes`	52	multi-mode phonon data for 10 crystals	`note`, `license`
`provenance`	30	central registry joinable via `nk_data.source`	`source_key`, `citation`, `doi`, `url`, `license`

Every data row also carries a license. Concrete identifiers are used only where a source declares one (nk_data is CC0-1.0); elsewhere the value is a documented sentinel (see-source-publication, see-references, see-note) pointing at the row's citation. See the database schema for the full model.

Inspect any row directly:

import sqlite3, importlib.resources

with importlib.resources.as_file(
    importlib.resources.files("photonics_helper") / "materials.db"
) as db:
    con = sqlite3.connect(db)
    for row in con.execute("SELECT material, source, citation FROM nk_data LIMIT 3"):
        print(row)

Where the data comes from¶

nk_data — collected from refractiveindex.info via its public sitemap (nk_datasets/collect.py, nk_datasets/manifest.json). The site publishes its compilation under CC0 1.0 Universal. Individual datasets originate in specific publications, so each row keeps a citation naming the original source; keep that column if you repackage the database. Seven rows are currently provenance-incomplete.
sellmeier — dispersion equations taken from the cited papers (Malitson 1965, Edwards & Lawrence 1984, Luke et al. 2015, …); source carries the attribution.
raman_specs — a literature compilation original to this project, with per-row references to the underlying paper or textbook.
phonon_modes — multi-mode phonon data whose canonical source is photonics_helper.phonon.PHONON_MATERIALS (per-mode literature notes).
provenance — a registry table joining on nk_data.source, backfilled from the tabulated data so the database is auditable without external files. Its doi column is populated by extracting the identifier from each citation, so consumers do not have to parse citation strings.

To list everything that ships (with wavelength range, DOI and licence), use material_catalog() / print_material_catalog() — see the database schema page.

Licensing caveat¶

CC0 applies to the refractiveindex.info aggregation. Because datasets are contributed from many publications, a small number may carry the terms of their original source. The citation column exists so a redistributor can verify each one. If you believe an entry cannot be redistributed, open an issue and it will be corrected or removed.

Status¶

The data-layer consolidation (Phase 2) is complete:

phonon_modes is populated and reachable through RamanDatabase and PhononResponse.from_material() — the same interface as n/k data.
Every row in every data table carries a license; the provenance registry is populated and queryable (RamanDatabase.get_provenance()).
The shipped database is verified against the canonical Python tables by drift tests, and key values are pinned by golden tests.

SQLite access is lazy: constructing a RamanDatabase has no filesystem side effects; the schema/seed step runs on first use.

Splitting the data into a separate distribution or a download-on-demand extra remains a future option; the database stays bundled for now.

REPORT.md (git-ignored local notes) tracks the wider "Foundation Backbone" roadmap.