# ORCHID data

> ORCHID data: 73,542 literature values (invariant reactions, phase transitions, phase compositions, thermochemical and structural data) from 1,321 papers, 2,722 calculated binary phase diagrams and DFT/CALPHAD convex hulls; CC BY 4.0.

Evidence classes are never mixed: experimental | model (CALPHAD, linked to the TDB by SHA-256) | dft | unspecified.
Formation enthalpies are in kJ/mol-atom when the per-atom basis is known; temperatures in K; compositions in mole fraction.

## Agent use: fitting, identity, and evaluation
- Extraction acceptance is not fitting eligibility. Check flags.espei_ready, flags.espei_certified, gate_failures, and flags.specification_gaps before fitting; MCP summaries expose the corresponding fields directly.
- A human_review_correction_requires_downstream_revalidation specification gap blocks fitting. Preserve corrections with before/after values, source DOI, page, and adjudicator. AI-assisted source adjudication is not independent human re-review, and corrected records do not replace the frozen review sample used for accuracy estimates.
- Deduplicate observations using duplicate_of and evidence_identity.canonical_id. Treat alternate-unit representations as one observation; merged_ids and merged_pids retain aliases. These identifiers are not evidence of independent measurements.
- Use fold_lso for leave-system-out and fold_lpo for leave-paper-out evaluation. Keep duplicate and alternate-unit aliases in the same split, including aliases across papers; a paper split alone does not prove absence of leakage.
- Keep experimental, model, dft, unspecified, and assessed_phase_diagram evidence separate. Materials Project and OQMD use different energy references. Use uncertainty_budget.sigma_effective with its documented basis.
- Pin the data commit and record model hashes and the software environment. Cite original sources as well as ORCHID; a Zenodo DOI containing PENDING is not a published identifier.
- The orchid-mcp package provides tools, resources, and citation guidance. Local use requires ORCHID_DATA_DIR and ORCHID_SITE_DATA_DIR; use fields=full for complete provenance. A working local server does not imply a public HTTP endpoint.

## Per system (SYS = element symbols, uppercase, sorted alphabetically, joined by '-', e.g. AL-NI)
- Markdown page: https://data.orchid-project.org/systems/<SYS>.md (summary, calculated diagrams, invariants, formation enthalpies with DFT, other properties, sources)
- HTML page: https://data.orchid-project.org/systems/<SYS>.html
- CSV: https://data.orchid-project.org/systems/<SYS>.csv (all records incl. not accepted; column extraction_accepted)
- Full JSON records: https://data.orchid-project.org/literature/systems/<SYS>.json (large systems: also <SYS>.part2.json ...; see literature/index.json)
- Calculated phase diagrams: https://data.orchid-project.org/phase-diagrams/index.json -> phase-diagrams/sgte/<file>.json
- Phases (binary systems): https://data.orchid-project.org/phases/pages/<SYS>.md (also .html): every phase with ids, names, structure tier/basis, prototype, Strukturbericht, Pearson, space group, stability, verbatim TDB PHASE/CONSTITUENT statements, example DOIs
- Phase registry JSON (names, models with tdb_definition, stability, references, structure when the stored evidence supports an assignment under the stated tier rules): https://data.orchid-project.org/phases/systems/<SYS>.json; compact index of all phases https://data.orchid-project.org/phases/registry.json, fields https://data.orchid-project.org/phases/SCHEMA.md

## Indexes
- [Systems](https://data.orchid-project.org/systems/index.md)
- [Compounds (formation enthalpies, literature vs Materials Project / OQMD)](https://data.orchid-project.org/compounds/index.md); per compound: https://data.orchid-project.org/compounds/<FORMULA>.md
- [Reporting papers](https://data.orchid-project.org/papers/index.md)
- [Phase registry, binary systems](https://data.orchid-project.org/phases/pages/index.md); structure types https://data.orchid-project.org/phases/structure_types.json (Parquet: phases/structure_types.parquet), all phases https://data.orchid-project.org/phases/registry.csv / registry.parquet / registry.jsonld
- [All systems with one-line summaries](https://data.orchid-project.org/llms-full.txt)
- [Field reference](https://data.orchid-project.org/literature/SCHEMA.md)
- [Assessed phase assemblages](https://data.orchid-project.org/assessed-phase-assemblages/README.md): one (T, x, P) point per named field of expert-drawn binary diagrams, evidence class assessed_phase_diagram (not experimental, not CALPHAD); https://data.orchid-project.org/assessed-phase-assemblages/records.jsonl.gz (also .csv, .parquet), ESPEI ZPF datasets espei-zpf/<DIAGRAM>.json, fields https://data.orchid-project.org/assessed-phase-assemblages/SCHEMA.md
- [ESPEI datasets from the literature](https://data.orchid-project.org/espei/README.md): ESPEI 0.9.1 JSON per evidence class (experimental/, model/, invariants/, dft/literature, dft/mp, dft/oqmd), every value with its ORCHID observation id, source DOI, uncertainty and weight, plus optional tagged phase constraints (derived/invariant-brackets/, derived/liquid-stability/, assessed/); one row per dataset https://data.orchid-project.org/espei/index.csv; validation https://data.orchid-project.org/espei/VALIDATION.md; uncertainties https://data.orchid-project.org/espei/UNCERTAINTY_SUMMARY.md; DFT sigma calibration https://data.orchid-project.org/espei/dft/CALIBRATION.md
- [ORCHID bibliography](https://data.orchid-project.org/bibliography/README.md): every paper ORCHID knows (9,570; roles, evidence counts, typed relations). One paper: https://data.orchid-project.org/bibliography/papers/<key>.json (key = "p" + sha256("doi:<lowercase doi>")[:16]; lookup table in bibliography/index.json); one system: https://data.orchid-project.org/bibliography/systems/<SYS>.json (assessments and the data papers they used); tables: bibliography/papers.parquet, bibliography/edges.parquet; field reference: bibliography/SCHEMA.md; dropped DOI matches: bibliography/DOI_MISMATCHES.md

## Bulk files
- https://data.orchid-project.org/literature/orchid-literature.jsonl.gz (all records, one JSON object per line)
- https://data.orchid-project.org/literature/orchid-literature.csv.gz (all records, flattened)
- https://data.orchid-project.org/literature/dft_parity.json (literature formation enthalpies vs DFT)

## Licence and citation
- Status: Version 1, machine-extracted: values were read by language-model readers constrained to quote their source and checked automatically, and the release was then audited on the printed pages by adversarial audits, an independent second review of fit-facing records and a preliminary human review of a frozen sample. Flags such as accepted and ESPEI-ready record screening results, not certification; known issues are listed in the release notes.
- Zenodo release v2026.10 (cite the dataset you use and the umbrella record):
  - ORCHID data: umbrella record of the ORCHID datasets: https://doi.org/10.5281/zenodo.23736065
  - ORCHID-Values: thermodynamic and phase-equilibrium values read from the literature: https://doi.org/10.5281/zenodo.23735569
  - ORCHID-Diagrams: calculated binary phase diagrams: https://doi.org/10.5281/zenodo.23735859
  - ORCHID-Hulls: binary convex hulls (Materials Project, OQMD, assessed CALPHAD): https://doi.org/10.5281/zenodo.23735891
  - ORCHID-Parity: DFT-vs-experiment formation-enthalpy benchmark: https://doi.org/10.5281/zenodo.23735943
  - ORCHID-Registry: phase registry of binary systems: https://doi.org/10.5281/zenodo.23735970
  - ORCHID-Assemblages: assessed phase assemblages of binary systems: https://doi.org/10.5281/zenodo.23735996
  - ORCHID ESPEI datasets: literature and database-derived thermodynamic fitting inputs: https://doi.org/10.5281/zenodo.23735788
  - ORCHID-Bibliography: papers and typed citation relations: https://doi.org/10.5281/zenodo.23736021
- Every JSON file carries a top-level cite_as object (text, BibTeX, DOI); Parquet files carry orchid.cite_as metadata; literature/CITATION.txt sits beside the CSV.
- Derived data CC BY 4.0. Cite the ORCHID release (https://data.orchid-project.org/CITATION.cff) and each value's source paper (DOI given per value).
- Quotes: one short verbatim excerpt per value, kept for verification; publishers' copyright, not relicensed (LICENSES.md). Source papers, page images and full text are not redistributed.
- Use values with extraction_accepted = true unless you handle disputes; for weighting use sigma_effective (uncertainty_budget), not uncertainty.value: a stated_default or reading_resolution uncertainty.basis is not a measured uncertainty.
- Every value carries uncertainty_budget (components printed / reading / method floor / composition / scatter / conversion kept separate; sigma_effective = max(sqrt(max(printed, reading, method)^2 + composition^2 + conversion^2), scatter) is the recommended sigma; floors: https://data.orchid-project.org/literature/UNCERTAINTY_CALIBRATION.md).
- Project site: https://orchid-project.org/ (interactive explorer https://orchid-project.org/data-explorer/#<SYS>)

## Reporting errors and contributing
- A wrong value: https://github.com/orchid-calphad/orchid-data/issues/new?template=report-value-error.yml&record_id=<ORC id> opens a pre-filled GitHub issue (record id, system, DOI, page, value only; the person submits it). orchid-mcp returns this link as report_issue_url with every value.
- Papers to read, corrections and new data (types A-E): https://github.com/orchid-calphad/orchid-data/blob/main/CONTRIBUTING.md and contributions/SCHEMA.md; plain-English guide https://orchid-project.org/contribute/