Peritheos 0.8.0

This release expands the model and dataset APIs and revises
the scientific catalog. It contains 210 material documents, 209 executable
materials, and 573 primary-source-validated EOS records. Source validation
checks the published parameterization; it does not imply independent refit
parity or a recommendation for every quantitative application.

Added

• Typed observation loading through Material.get_dataset(), Dataset,
  DatasetColumn, DatasetResource, and PressureVolumeData. Embedded and
  packaged CSV data expose read-only arrays, verified resource checksums,
  column metadata, unit conversion, and pressure-volume views with reported
  uncertainties. DatasetError and DatasetLookupError provide typed failures;
  the existing raw Material.datasets mappings remain available.
• EulerianFiniteStrainAcoustic, fit_acoustic_finite_strain(), and
  AcousticFitResult for joint density/P-wave/S-wave velocity fitting,
  including optional density errors, bounds, robust losses, covariance,
  adjusted observations, and JSON-safe diagnostics.
• Python and Rust Vinet3 and the Fratanduono et al. (2020) 298 K Cu reference
  record, completing the implementation described prematurely in the 0.7.0
  notes. Same-run Cu volumes now support Shen–Smith pressure reconstruction;
  reduced isentropic states are not treated as independent 298 K fit data.
• Thermal model APIs and .eosmat registrations for Dewaele2006,
  DebyeQuadraticThermalPressure, HollandPowellThermalPressure,
  AsymptoticPowerLawMieGruneisenDebyeExcess, and
  SoundVelocityDebyeHelmholtz. These cover single-Debye iron scales with
  anharmonic/electronic terms, Fei's quadratic Debye pressure, a composable
  Holland–Powell thermal term, a quadratic excess free energy, and a
  sound-velocity-derived Debye model. Native implementations support the first
  four; the sound-velocity model provides Python thermodynamic and shock-state
  calculations. Dewaele2006 includes the exact gamma_inf=0 limit.
• MieGruneisenDebye reference-isentrope subtraction and optional Cvmax,
  with native support for the BM3 shock-static construction. Tange2009Debye
  exposes constrained_reference_parameters() for the source's coupled ambient
  elastic and heat-capacity constraints.
• Cubic compressibility (bulk_modulus_law="reciprocal_cubic") and a
  logarithmic temperature correction to K0_prime in ThermalReferenceStateEOS,
  including native evaluation and fitting. The existing linear modulus law
  remains the default; these laws define a mechanical P-V-T surface.
• Complete caloric Helmholtz observables for DorogokupetsOganov2007, and
  ionic, anharmonic, and total entropy/internal energy for both double-Debye
  Helmholtz models, consistent with their reference-isotherm subtractions.
• The executable ruby_sokolova_2013 calibration in Python and Rust, with
  explicit provenance for the Sokolova pressure-scale reconstruction.
• Primary-source Fe, Fe0.9Ni0.1, and NaCl-B2 records from Brown (2000),
  Dubrovinsky (2000), Dewaele (2006), Yamazaki (2012), and Sakai (2014): 43
  published records and 348 observations, with qualified regression audits.
  Fei et al. (2016) adds two 300 K hcp-Fe BM3 alternatives, its thermal pressure
  surface, and all 96 supplementary observations; remaining source and partial
  refit discrepancies are documented.
• Hirose et al. (2008) Au: the 300 K isotherm, both thermal fits, and all 21
  paired Au/MgO observations. Sakai et al. (2018) adds the hcp-Re Vinet scale
  on Yokoo Pt, all 58 Figure 10 vector markers, and a 26-point RP01 refit with
  coefficient and uncertainty parity; alternate selections remain diagnostics.
• Source-owned MgSiO3 records from Wang (2004), Zhou (2014), and Komabayashi
  (2008), MgAl2O4 records from Irifune (2002), Ono (2006), and Sueda (2009),
  and high-/low-spin ferropericlase records from Fei (2007) and Marquardt
  (2009). Source-fit alternatives, published errors, staged protocols, and
  independently derived refits are distinguished in their provenance.
• Separate Ye et al. (2017) Au/Pt/MgO 300 K standards and Zhu et al. (2025)
  thermal P-V-T records. The corrected Ye supplements reproduce the printed
  fits and errors; Zhu's released v3 optimizer/property coefficients and
  bundled thermal inputs support the thermal reconstruction. Differences
  from the preprint and standalone calculators remain explicit.
• Fei et al. (2004) Au/Pt pressure standards, a separately identified Fu
  CaSiO3 literature-data refit, and Kuznetsov (2002) fcc/hcp Pb reference
  isotherms. The Pb isotherms do not establish a refit of the unpublished
  full thermal observation set.

Corrected

• Chen et al. (2018) tetragonal CaSiO3 now uses the source-stated Vinet EOS,
  replacing ca_perovskite_tetragonal_chen_2018_bm2 with
  ca_perovskite_tetragonal_chen_2018_vinet. All seven table rows, the Z=4
  volume normalization, two-sigma errors, and Pt calibration are preserved.
  Numerical agreement remains qualified because source weights and pressure
  errors are unavailable; no separate refit record is introduced.
• Zhao et al. (1997) jadeite now includes its complete 300–1280 K thermal
  relation and the fitted reference coefficient V0=403 ų, replacing the
  measured 403.32(8) ų table value. The 31 hydrostatic states and Decker
  NaCl lineage support a qualified refit with unresolved source weighting.
• Restored the complete Fischer et al. (2011) B1/B8 FeO thermal relations,
  Anzellini et al. (2025) Ir BM3 plus Holland–Powell thermal EOS, and
  Mosenfelder et al. (2009) MgSiO3 post-perovskite shock-static EOS. Ir uses
  the mean pyrometry temperature for its 122 supplementary observations;
  Mosenfelder's reconstruction uses 48 static states and six PPv shock states.
• Holmes et al. (1989) Pt now maps the exact published universal-equation
  coefficients to Vinet parameters instead of using their rounded elastic
  interpretations. Its approximate thermal extension still uses the published
  rounded modulus; the seven shock observations qualify the curve rather than
  independently reproducing the unavailable theoretical fit.
• The Au/Pt/MgO *_zhu_2025_vinet_300k names now resolve as aliases of the
  source-correct *_ye_2017_vinet_300k identifiers. Zhu's thermal models have
  separate *_zhu_2025_pvt identifiers.
• Source-only cards are skipped by executable catalog enumeration. Coesite-V
  remains accessible through the document API, but cannot be constructed as
  a Material without an executable EOS.

Scientific evidence and limitations

• Consolidated recovered source tables, digitizations, deterministic audits,
  and record-level refit results. These include Sun (2016/2022) and Noguchi
  (2013) CaSiO3, Li (2006) acoustic MgO, Noguchi (1999) shock-reduced NiO,
  Redfern magnesite, Datchi diamond, and MgSiO3 staged fits. Akins (2004)
  liquid MgSiO3 retains a qualified similar source-equation check from three
  reduced shock states; it does not establish strict fit parity. Published values
  remain distinct from diagnostic refits. Noguchi's subscription-only table
  is represented by a row-free reproducibility artifact and local-input script.
• Dewaele (2019) metal evidence is consolidated without claiming completion of
  all 30 records. The common ledger classifies 24 records as similar using
  two-ruby-scale refits within the printed 95% intervals, without inferred
  refit covariance. Recovered Fe EPAPS observations additionally support both
  scale reductions in the dedicated audit, although the common ledger still
  reports discrepancies for Fe under its different fitting path.
• All 41 Lazicki Be helium-run rows are bundled, including the version-of-record
  a=2.162 Å at 24.6 GPa and the superseded manuscript token 2.262 Å in
  provenance. The two Be records nevertheless remain not_refittable in the
  current common ledger, and the dedicated audit has not integrated this
  recovery. The earlier claim that 26 records were linked to completed refits
  is therefore not the released validation state; neither is the old claim
  that the Be table is unavailable.
• hcp-Pb retains 17 tabulated Dewaele rows, 21 digitized Kuznetsov Figure 3
  markers (seven room-temperature and 14 thermally reduced), and a separate
  nine-marker selection recovered from Dewaele Figure 2. The 24-point
  room-temperature combination, 38-point all-marker combination, and 26-point
  Figure 2 combination are distinct diagnostics. The current common ledger
  evaluates the 21 Kuznetsov markers alone and reports parity_not_achieved;
  the dedicated metal audit fits the 17 Dewaele rows alone. No combined
  selection establishes the unpublished original rows, weights, pressure-scale
  handling, or full-fit parity. Published Pb coefficients are preserved.
• Anderson Au, Dorogokupets–Oganov Pt, Sokolova's global calibration, and
  Tange MgO retain qualified component/partial reconstructions rather than
  claims of complete global-refit parity. The Luo MgO full refit remains
  not_refittable; derived diamond composites use source_reconstruction.
  Funamori (1998) transformation-derived MgAl2O4 curves remain executable
  source constructions with no statistically independent regression to rerun.
• Sun (2019) liquid FeSiO3 audits distinguish the 40 liquid and six nonliquid
  table states and digitized thermal-fit inputs from an exact numerical source
  regression. Ismailova Fe0.88SiO3, Cohen–Lin FeSiO3, and Dorogokupets (2015)
  MgSiO3 retain explicit missing-input or fit-protocol limitations. Recovered
  checkpoints and curve agreement do not establish unavailable source data.

See the record-level refit ledger and
paper investigation ledger for the
qualified outcomes and links to individual reproductions.

Removed and compatibility

• Removed 92 Delta-project benchmark records (50 code-specific PBE curves and
  42 experimental-reference comparison curves), their benchmark-only cards,
  datasets, and import/reproduction tooling from the production distribution.
• Removed 220 Sun et al. (2010) universal-EOS benchmark parameterizations and
  49 phase-unresolved cards. The numerical source table, Morse/Sun-Morse
  implementations, and equation tests remain benchmark fixtures; the
  independently refitted AIP Handbook Cu isotherm retains its own card.
• Removed 14 Caracas et al. (2005) exploratory CaSiO3 tilt-branch records
  without diffraction-ready structures. The full nine-structure coefficient
  table remains an audit fixture; production retains the Pm-3m and I4/mcm pairs.
• Removed the Chantel (2012) composite bridgmanite EOS and Schoelmerich (2020)
  shock-corrected stishovite EOS from production, and demoted the one-anchor
  coesite-V reconstruction to non-executable provenance. Their available source
  evidence remains; Schoelmerich's fitted-curve digitization is removed because
  it cannot independently reproduce the unpublished shock correction.
• Removed identifiers and benchmark resources are no longer available through
  catalog lookup or package-resource access. Callers must update the Chen
  identifier and account for the distinction between material documents and
  executable materials. Ye/Zhu's former isotherm identifiers remain aliases.
• .eosmat stays at format 3, with additional model discriminators and thermal
  options; older consumers may not construct those models. A refit record
  still requires fit_provenance, but may now represent a source-data refit
  without derived_from_record. Existing default thermal conventions and
  raw dataset mappings are preserved. Rust callers must account for added
  model/reference enum variants in exhaustive matches and new fields when
  constructing affected thermal structs directly. The ruby calibration array
  now has eight entries, which affects fixed seven-element array types.