Every value the Codex publishes carries a cited source. This is that list — generated from one committed CSV, so the page cannot drift from the pipeline that reads it.
An abundance is only as good as the atomic data underneath it. A log gf measured in a laboratory and a log gf fitted so that the Sun comes out right are not interchangeable — the second one makes a solar abundance a restatement of its own input. So the Codex tracks provenance per source, and this page publishes it.
Read the badges. PDF means the metadata was read off the paper itself; DOI means the DOI record was fetched and matched to this row; UNCONFIRMED means no DOI or stable URL could be confirmed. Those rows are shown without a link and without a guessed DOI. Most are pre-DOI literature that no registry carries; a plausible-looking invented identifier would be a fabricated provenance record, so there isn’t one.
GOLD marks a reference standard: a scale the Codex compares against rather than re-derives.
The abundance scales every Codex measurement is referenced against, and the solar spectra used as the observational anchor.
GOLD reference standard. The 3D non-LTE solar abundance scale every Codex element is compared against, and the per-element physics-regime authority.
Open access (A&A).
The predecessor solar scale; still the comparison point for pre-2021 literature values quoted in the Codex.
LOCAL FILE IS MISNAMED: 'Microturbulence.pdf' is Asplund+2009, not a microturbulence paper. Identified from the PDF first page.
Telluric-corrected IAG solar atlas (0.5-1.0 um); the telluric control for the optical/near-IR solar reference.
Independent 3D solar abundance scale; the CO5BOLD anchor used where the Codex quotes a non-Stagger 3D value.
Crossref issue date is 2010 (online-first); the volume year is 2011.
Legacy solar composition scale; appears as a cited abundance source behind older log gf normalisations in the line lists.
Ticket text said 'Grevesse & Sauval 1999'; the record is 1998 (Space Sci. Rev. 85, 161). Year corrected against Crossref.
Space-borne IR solar atlas (700-4430 cm^-1) — the telluric-free IR solar comparison used when validating ground-based IR atlases.
GOLD reference standard. Canonical 28-element list and the analysis quality index; the meteoritic half of the solar scale.
On-disk copy is the accepted manuscript (accepted 2025-02-07); the DOI resolves to the published Space Sci. Rev. article.
Reference solar flux atlas (405-2300 nm) used as a wavelength-scale control for the solar anchor.
Iron-group solar abundances and the 3D corrections the Codex uses for Ti and Cr.
Crossref issue date is 2014 (online-first); the volume year is 2015.
Cross-instrument solar data comparison; context for combining HARPS solar products across spectrographs.
Where each log gf comes from. The distinction that matters is PRIMARY LABORATORY (branching fractions from a Fourier-transform spectrometer times a measured radiative lifetime, with no reference to the solar spectrum) versus semi-empirical or solar-fitted values, which cannot referee a solar abundance without restating their own input.
GES code BKK (8 canonical_gf rows). The earlier Bard-Kock laboratory Fe I set.
Pre-2000 A&A volume; no DOI is registered. Link is the ADS scanned article, downloaded and confirmed to be a real PDF of the published length. Bibcode 1991A&A...248..315B verified two ways: the ADS scanned-article gateway returns 200 for it and 404 for a fabricated bibcode, and SIMBAD's bibliographic record returns exactly these authors, this title and these pages.
GES code BK (33 canonical_gf rows). Laboratory Fe I gf; one of the reference sets expected to sit near zero offset in the RYA-760 GES-vs-VALD comparison.
Pre-2000 A&A volume; no DOI is registered. Link is the ADS scanned article, downloaded and confirmed to be a real PDF of the published length. Bibcode 1994A&A...282.1014B verified two ways: the ADS scanned-article gateway returns 200 for it and 404 for a fabricated bibcode, and SIMBAD's bibliographic record returns exactly these authors, this title and these pages. The title first drafted for this row from memory ('...for lines of astrophysical interest') was WRONG — that is the 1991 paper's title. SIMBAD corrected it.
PRIMARY laboratory Fe I gf over 213-1033 nm. Third of the three lab-gf sub-pool sources and the one that reaches the near-UV.
Ticket text said 'aa8bd9'; the DOI in the PDF metadata is 10.3847/1538-4357/aa8cd3. Corrected from the PDF.
GES code BDMQ (1,040 canonical_gf rows).
GES codes CB (1,836 rows) and CBcor (661 rows). Arc-emission gf; superseded for most species but still live in the pool.
The empirical S I line data behind A(S) = 7.486.
Pure laboratory Fe II gf in the UV/blue — the independent referee for the Fe II scale-offset investigation.
Laboratory Mn I/Mn II gf; the adjudicated Mn source behind A(Mn) = 5.466. GES code DLSSC; cited verbatim in canonical_gf.loggf_reference.
PRIMARY laboratory Fe I gf. Second of the three lab-gf sub-pool sources. GES code GESHRL14.
GES code FMW (492 canonical_gf rows). A NIST compilation — checking it against NIST ASD is circular, which is why RYA-760 could not use ASD as referee.
JPCRD supplement volume: not in Crossref, and no stable publisher URL was found — doi and url are deliberately BLANK. The TITLE, however, is confirmed from a primary source: the reference list of NIST's own 2006 Fe compilation (doi:10.1063/1.2218876) cites it verbatim as 'Atomic transition probabilities — Iron through nickel', J. Phys. Chem. Ref. Data 17, Suppl. 4 (1988).
Cited in canonical_gf by bibcode 2006JPCRD..35.1669F. The modern NIST Fe compilation; supersedes FMW for Fe but carries the same non-independence.
GES code GARZ (27 canonical_gf rows).
Pre-2000 A&A volume; no DOI is registered. Link is the ADS scanned article, downloaded and confirmed to be a real PDF of the published length. Bibcode 1973A&A....26..471G verified two ways: the ADS scanned-article gateway returns 200 for it and 404 for a fabricated bibcode, and SIMBAD's bibliographic record returns exactly these authors, this title and these pages.
The published account of the GES line list that seeds 125,958 canonical_gf rows (seed_source synth(GES)).
THE DECODER for canonical_gf.loggf_reference: Table 4 maps every short code (BWL, RU, PQWB, MC, K07...) to its bibliographic source. Verified 2026-08-17.
Survey documentation, no DOI. URL confirmed HTTP 200 on 2026-08-17.
The Ni I 6300.34 gf that sets the [O I] 6300 blend correction — the whole solar oxygen gate depends on it. seed_source 'Johansson2003' in canonical_gf.
Cited verbatim in canonical_gf for an Al/Si grade-C+ line.
Cited verbatim in canonical_gf for Na/Mg grade-D and grade-E lines.
PAGE CORRECTION: two canonical_gf rows cite 'JPCRD 37 p1285 (Na/Mg)'. p1285 is the SILICON compilation (doi:10.1063/1.2734566); Na/Mg is p267. Flagged to RYA-853, not fixed here.
GES code KSG (6 canonical_gf rows) — the Schnabel-coauthored source the line list actually cites.
Grade-bearing gf compilation queried through astroquery.nist. NOT independent of FMW/MFW, which are themselves NIST compilations — it cannot referee them.
Database DOI; resolves to the NIST ASD landing page.
GES codes K99/K03/K04/K06/K07/K08/K09/K10/K11/K12/K13/K14 — semi-empirical calculations, ~91,000 canonical_gf rows and the dominant non-lab gf source. The RYA-161 firewall class.
Living online database, no DOI. URL confirmed HTTP 200 on 2026-08-17.
GES code KP (4,044 canonical_gf rows). Semi-empirical; firewall class.
No DOI exists for this SAO Special Report; the link is the ADS scanned report. Bibcode 1975SAOSR.362.....K confirmed by SIMBAD (exact title and authors) and by CDS VizieR catalogue VI/10, the machine-readable table derived from it (https://cdsarc.cds.unistra.fr/viz-bin/cat/VI/10). CAVEAT, stated because it is a real gap: unlike the four A&A scans this 64 MB PDF was NOT downloaded in full — the gateway resolves HTTP 200 but the transfer timed out, so the file itself is unread and the identification rests on SIMBAD and VizieR.
GES code LGWSC (504 canonical_gf rows). Laboratory Ti I gf.
GES code LSCI (519 canonical_gf rows). Laboratory Ce II gf.
GES code MFW (632 canonical_gf rows). Companion NIST compilation to FMW; same circularity caveat.
JPCRD supplement volume: not in Crossref, no stable publisher URL found, and — UNLIKE its Fe-Ni companion fuhr1988 — the title could not be confirmed from any primary source held here. The GES v5 report and Heiter et al. 2021 both list it without a title. Left blank rather than guessed; authors, year and venue are substantiated. Secondary sources give 'Atomic transition probabilities — Scandium through manganese', which is why Suppl. 3 is recorded, but that is not primary evidence and is not promoted into the title field.
GES code MC (3,620 canonical_gf rows). Arc-emission intensities converted to gf; the oldest and least reliable large source in the pool.
No DOI confirmed: Crossref holds only the 1961 NBS Monograph 32 parts. Title and imprint are verbatim from the GES v5 line-list report reference list. Link is the Internet Archive scan of Part 1 (arranged by elements), confirmed by page content; Part 2 (arranged by wavelengths) is the companion item tablesofspectral1452megg. Public domain — a US Government publication.
Fe II gf set. FIREWALLED: multiplets are globally normalised on laboratory data but individual values are partly solar-fitted, so it must never referee a solar abundance (RYA-161 class).
On-disk copy is the arXiv v2 preprint (arXiv:0901.4451).
GES/VALD code BWL (258 canonical_gf rows). Laboratory Fe I lifetimes and transition probabilities.
CODE CORRECTION: the ticket listed 'BWL/Blackwell-Whitehead-Lawler' and 'O'Brian 1991' as two sources. The GES v5 line-list report decodes BWL as O'Brian et al. (1991) — they are ONE source, not two.
GES code PQWB (5,522 canonical_gf rows). Theoretical Ce II gf.
The independent Mg I gf cross-check quoted in two canonical_gf provenance strings during the RYA-592 re-verification.
Fe I level energies and line identifications; the term-label authority behind level keying (RYA-823) and the IR jK levels.
GES code PTP (35 canonical_gf rows). Erratum: ApJS 138, 247 (2002), doi:10.1086/323910.
SPECIES CORRECTION: the ticket listed Pickering 2001-02 under 'Fe II experimental'. The paper is Ti II, not Fe II.
GES code QPB (1,680 canonical_gf rows). Theoretical lanthanide gf.
GES code RPU (600 canonical_gf rows). Theoretical Co II gf.
GES code RU (8,473 canonical_gf rows) — orthogonal-operator theoretical gf, the second-largest single source after VALD3/Kurucz.
Pre-2000 A&A volume; no DOI is registered. Link is the ADS scanned article, which was downloaded and read: its own first page gives this title, these authors and 'Astron. Astrophys. 340, 300-304 (1998)'. NOTE THE SUBJECT — this is a solar-Fe-abundance paper; it derives A(Fe) = 7.59 from its own gf. Those gf are theoretical, so the abundance is an OUTPUT and not a fit input — this is NOT RYA-161 firewall class — but 8,473 canonical_gf rows carry it, so the distinction is worth stating rather than assuming.
GES code NIST10 (974 canonical_gf rows) — the frozen ASD v4.0.0 snapshot the GES list was built against, distinct from today's live ASD.
Version-specific snapshot; the DOI resolves to the current ASD landing page, not to v4.0.0 specifically.
PRIMARY laboratory Fe I gf (FTS branching fractions x radiative lifetimes). One of exactly three sources admitted to the lab-gf sub-pool.
Ticket text said 'stu520'; the DOI printed on the paper is 10.1093/mnras/stu780. Corrected from the PDF.
VALD3 — the largest single loggf_reference in canonical_gf (40,515 rows) and the source of every vald_*_raw delivery.
Named in RYA-854 as an expected Fe II source. NOT currently cited by any canonical_gf row — the Schnabel paper that IS cited is Kling et al. (2001).
Listed for completeness. DOI confirmed via Crossref; aanda.org returns HTTP 403 to automated agents, which is a bot block, not a dead link.
GES code SPN (5 canonical_gf rows). This — not the 1999 FERRUM paper — is the Sikstrom source actually cited by the line list.
Named in RYA-854 as an expected Fe II source. NOT currently cited by any canonical_gf row — recorded here so the claim is checkable, not assumed.
Listed for completeness; canonical_gf's Sikstrom code (SPN) resolves to the 2001 Mo II paper instead.
GES code SLS (122 canonical_gf rows). Laboratory Cr I gf.
Cited verbatim in canonical_gf for a NIST-grade-A forbidden line.
The extraction TOOL the raw deliveries came from; cite alongside ryabchikova2015 whenever a delivery is reproduced.
Database interface, not a publication — no DOI exists.
GES code WBb (854 canonical_gf rows).
Collisional broadening theory behind the van der Waals damping the line lists carry.
The ABO theory behind the van der Waals broadening the line lists carry (VALD code BPM, Barklem-Piskunov-O'Mara).
Workshop proceedings; no DOI. The link is the proceedings host at the University of Colorado, and it is the DEFINITIVE identification: the served file is BYTE-IDENTICAL to the local copy (cmp, 96,017 bytes), so the on-disk PDF demonstrably came from there. ADS bibcode 2003csss...12.1097O.
The model atoms and departure-coefficient grids that move an abundance off its LTE value — including the papers that define where a grid does NOT reach.
3D non-LTE O I grids ([O I] 6300/6363, O I 777) behind the solar oxygen leg.
Engine A 3D non-LTE Fe. The MLP correction surface — in-domain only where the training set reaches, which excludes the IR (RYA-817).
On-disk copy is arXiv v3 (arXiv:2209.13449); the DOI is the published A&A version.
The C I / O I 3D- and 1D-NLTE correction tables applied in the CNO leg.
Cu I NLTE model atom; one of the three Cu NLTE treatments compared for A(Cu).
On-disk copy is the arXiv v1 preprint (arXiv:1709.08619).
The 1D/<3D> NLTE Fe grid served by the MPIA web service — Engine A for Fe wherever the Amarsi MLP is out of domain.
The NLiTE tool and NLTE-corrected SAGA catalogue; an external comparison for grid-to-grid NLTE spread.
On-disk copy is the arXiv v1 preprint (arXiv:2505.13607).
3D non-LTE Ca II reference; the frontier check for the Ca leg.
Ticket inventory listed no authors for this file; identified from PDF metadata as Lagae, Amarsi & Lind (2025).
Third Cu NLTE treatment held on disk; the LTE-vs-NLTE size check for Cu I.
On-disk copy is the arXiv v2 preprint (arXiv:1806.02225).
The Zn NLTE grid that made Zn the 28th canonical element.
Cu NLTE with quantum-mechanical Cu+H collision rates (Belyaev+2021) — the modern replacement for the Drawin approximation.
On-disk copy is the arXiv v1 preprint (arXiv:2208.11812).
Radiative-transfer codes that solve statistical equilibrium during synthesis.
Engine B. The NLTE-capable Turbospectrum and the departure-coefficient grids the Codex runs on Sirius.
The 1D and 3D atmospheric structures the synthesis runs on top of.
MARCS. The atmosphere Engine B requires — MARCS.GES, not ATLAS9, was the blocker RYA-798 found.
ATLAS9/ATLAS12 and SYNTHE — the 1D-LTE atmosphere family behind the Engine-A baseline grids.
The 1D->3D Fe correction source. RYA-819 showed the correction is Elo-driven, so the published -0.05 scalar is wrong in magnitude AND shape.
docs/references.md cites 'A&A 557, A26' for Magic 2013; the file on disk is paper II, A&A 560, A8. Both papers exist — the one the Codex uses is A560/A8.
Independently determined temperatures, gravities, metallicities and microturbulences — parameters the Codex adopts rather than fits.
Microturbulence and fundamental parameters for bright solar-type stars — the Procyon xi anchor.
On-disk copy is the arXiv v1 preprint (arXiv:1002.4268). '1002.4268v1.pdf' in the same folder is a BYTE-IDENTICAL duplicate of it (md5 match), not a second version. Crossref's record for this DOI carries no volume/page (early-view); volume 405 p1907 is the printed citation.
Fundamental Teff/log g for the benchmark stars — the adopted Procyon parameters.
On-disk copy is the arXiv v2 preprint (arXiv:1506.06095).
Benchmark [Fe/H] — the adopted alpha Cen A/B metallicity.
On-disk copy is the arXiv v2 preprint (arXiv:1309.1099).
Interferometric 55 Cnc parameters — the adopted Teff/radius for the flagship system.
External line-selection catalogues and instrument-specific abundance work the Codex cross-checks its own pools against.
The IR line-selection catalogue cross-checked against the Codex IR pool.
THIS ROW ACCOUNTS FOR THREE FILES in the reference library, checked by content rather than by name: 'aa59148-26 (1).pdf' is a BYTE-IDENTICAL duplicate of aa59148-26.pdf (md5 match), while '2602.14294v1.pdf' is a genuinely DIFFERENT file (4,385,306 vs 4,848,632 bytes) — the arXiv preprint arXiv:2602.14294 of this same paper. One row, cited to the published A&A version.
NIR abundance work on planet-hosting wide binaries; the closest external comparison for the Codex IR programme.
arXiv DOI — the A&A version had no DOI at the time of writing. arXiv abs page confirmed HTTP 200 on 2026-08-17.
The C/N/O literature behind the C/O ratio — the flagship derived product and the hardest measurement in the project.
Compares oxygen abundances from the [O I] 630 nm line against IR OH lines in bulge giants — the same two-indicator disagreement the Codex hits on its own solar oxygen, from a different stellar population.
The local file has NO extension; its name is the ADS bibcode. DOI is printed on the paper's own first page and confirmed against Crossref. Crossref dates the record 2008 (the symposium); the proceedings volume and the paper's own copyright line say 2009, which is the year used here.
Disk-to-planet C/O synthesis — why a stellar C/O ratio constrains planet composition at all.
On-disk copy is the 2024-06-19 AASTeX draft; the DOI resolves to the published ApJL version.
Disk carbon isotopic reservoirs — the formation-side context for the C/O programme.
The [O I] 6300 vs O I 6158 comparison — the external precedent for the Codex's own two-indicator oxygen disagreement.
Disk-midplane chemistry: the C/O of planet-building material is not the disk's initial C/O.
C I NLTE line formation — the size of the carbon NLTE correction the CNO leg must carry.
Ticket inventory listed this only as 'Carbon abundances'; identified from PDF metadata as Fabbian et al. 2006.
825-star O I 777 NLTE survey — the population-scale comparison for the Codex oxygen values.
On-disk copy is the arXiv v1 preprint (arXiv:1301.1582).
The 55 Cnc C/O controversy in one paper, including the Ni I 6300.34 blend sensitivity the Codex inherits.
The results that make a host-star abundance worth measuring at all.
The alpha-enhanced super-solar population — the precedent for the Codex's alpha-enhancement framing.
LOCAL FILE IS MISNAMED: 'Sousa et al 2011.pdf' is Adibekyan et al. 2011 (A&A 535, L11). Identified from PDF metadata.
The 1111-star HARPS abundance baseline — the population the Codex's HARPS solar and host-star work sits inside.
LOCAL FILE IS MISNAMED: 'Sousa et al 2012.pdf' is Adibekyan et al. 2012 (A&A 545, A32). Identified from PDF metadata.
THE motivating result: host-star composition and rocky-planet iron fraction correlate, but not 1:1.
NOT a pipeline input. Held in the reference library as Galactic-bulge abundance context (VLT-UVES / FLAMES / Gemini-PHOENIX abundance patterns in old bulge globulars). Recorded because the bibliography must account for every document in the library, and an unlisted document is an uncited one.
Conference proceedings; no DOI is registered. The link is the publisher's own copy at UNAM, and it is the DEFINITIVE identification: the served file is BYTE-IDENTICAL to the local copy (cmp, 105,455 bytes). CATEGORY CAVEAT: the RYA-854 vocabulary has no bucket for Galactic-population context, so this sits under motivation as the least-wrong fit — the role field above says what it actually is. NOTE a near-miss: a web search for this title surfaces a DIFFERENT Barbuy paper of almost the same name in ESO Astrophysics Symposia (2009, pp. 207-208). That is not this document.
Interior-composition inference from mass-radius — what a host-star abundance is ultimately FOR.
Uniform C/N/O and Mg/Si for 149 planet hosts — the direct comparison sample for the Codex's own ratios.
The [X/H]-Tcond slope debate: planet signature vs Galactic chemical evolution.
The software and methodological references the pipeline is built on.
iSpec — the framework every Codex EW and synthesis fit runs through.
Standing methodological reference for line formation, curve of growth, and macroturbulence conventions.
Why the Codex derives xi from OBSERVED equivalent widths rather than expected line strength, and where each method breaks down with S/N.
LOCAL FILE IS MISNAMED: 'Apslund2010.pdf' is Mucciarelli 2011 (A&A 528, A44). Identified from PDF metadata. This is the 'aa1581410.pdf' of the ticket inventory.
The MOOG code paper — the LTE line-analysis engine iSpec can drive.
The operating manual for the MOOG build compiled into the Codex iSpec install.
Software manual; no DOI. Link is Sneden's MOOG distribution page (content confirmed). Cite sneden1973 for the code itself.
Molecular line lists (CO, CH, CN) vendored for the IR and CNO synthesis legs.
On-disk copy is the arXiv v3 preprint (arXiv:1603.05890).
The project’s own methodology, glossary and architecture documents.
Where the project's question comes from, and the origin of the alpha-enhancement framing the Codex still uses: [Fe/H] alone is not the interesting axis. The 2010 ELODIE analysis of 55 Cnc A, HD 89307 and Gliese 581. Historical context — never a substitute for current pipeline provenance.
TWO EARLIER ROWS DESCRIBED THIS ONE DOCUMENT. It was carried both as 'schmitt_beyond_metallicity' (an unpublished essay, category motivation) and as 'schmitt2010_thesis' — they are BYTE-IDENTICAL (md5 36b8ce3883cbe75b...), and the thesis row additionally carried an INVENTED title, 'Elemental Abundances of Three Exoplanet Host Stars', which I had paraphrased from the About page's prose description instead of opening the PDF. The title above is the document's own first page. Merged to one row. Four byte-identical copies exist on disk: 'Archive docs/Beyond Metallicity.doc.pdf', 'Beyond Metallicity.doc (1).pdf', 'Archive docs/schmitt-2010-thesis.pdf' and the published site copy. Worked on 2007-2010; written and published through the Honors College at the University of Montana, with no formal journal publication — so the site copy is the only URL there is, and that is the complete answer, not a gap.
The published glossary; the vocabulary the site and the SPP share.
Project document; no DOI. Published copy is served from exoplanetcodex.org.
The published methodology document the site links from every results page.
Project document; no DOI. Published copy is served from exoplanetcodex.org.
The architecture document behind the two-engine floor, the three promotion gates, the freeze tiers, the honesty tripwire and the RYA-161 firewall — the reasoning every other row in this bibliography is serving.
FOUND, after an earlier pass wrongly recorded it as missing. The RYA-854 inventory named 'Exoplanet_Codex_Science_Architecture.docx', which genuinely does not exist under that name — but the document does, published on the site as v3.0 (PDF) with v2.0 (.docx) archived beside it. Title and subtitle read from the PDF's own first page; the live URL serves application/pdf. The lesson is that searching for the INVENTORY'S filename and concluding absence was the error: the absence was scoped to a name, not to the document.
GENERATED PAGE · source: data/refs/bibliography.csv · sha256 4d724042e82ad57b6ba565b142f1ce939efdbbb351ce12ae6186eac734282820 · 100 rows · generator: scripts/generate_sources_page.py · RYA-854. Never hand-edited: --check regenerates this file and fails the build on any difference.