Skip to content

Latest commit

 

History

History
1200 lines (988 loc) · 74.7 KB

full-eos.md

File metadata and controls

1200 lines (988 loc) · 74.7 KB

Full EOS scientific specification

This document owns symbols, bases, source applicability and EOS conventions. It states required behavior, not installed availability. The canonical equation definitions are edited once; repeated derivations below explain their meaning. Algorithms own numerical methods. Correct mathematics is not weakened to match an implementation defect.

1. Sources and evidence classes

ID Source Use
Gross2001 Gross and Sadowski, Ind. Eng. Chem. Res. 2001, 40, 1244-1260, DOI 10.1021/ie0003887 PC-SAFT hard-chain, dispersion, pressure, chemical potential, fugacity, and residual properties
Gross2006 Gross and Vrabec, AIChE J. 2006, 52, 1194-1204, DOI 10.1002/aic.10683 Optional dipole-dipole term, Eqs. (7)-(16) and Table 1 constants; Table 2 PCP-SAFT parameters (ADR 0011)
Gross2005 Gross, AIChE J. 2005, 51, 2556-2568, DOI 10.1002/aic.10502 Optional quadrupole-quadrupole term, Eqs. (8)-(19) and Table 2 constants; Table A1 PCP-SAFT parameters (ADR 0014)
Gross2002 Gross and Sadowski, Ind. Eng. Chem. Res. 2002, 41, 5510-5515, DOI 10.1021/ie010954d Association mixing for associating systems
Chapman1990 Chapman et al., Ind. Eng. Chem. Res. 1990, 29, 1709-1721, DOI 10.1021/ie00104a021 Wertheim association energy and site-fraction equations
Bulow2019 Bülow et al., Fluid Phase Equilibria 2019, 492, 26-33, DOI 10.1016/j.fluid.2019.03.010 Debye-Huckel contribution with composition-dependent dielectric treatment
Figiel2025 Paul Figiel, Gangqiang Yu, and Christoph Held, Ind. Eng. Chem. Res. 2025, 64(18), 9406-9418, DOI 10.1021/acs.iecr.5c00475 Fitted Born diameter, SSM+DS Born term, solvation factor, dielectric rule, fugacity relations, and electrolyte reference-state equations
Ji2012 Ji, Held, and Sadowski, Fluid Phase Equilibria 2012, 335, 64-73, DOI 10.1016/j.fluid.2012.05.029 Fully dissociated ionic-liquid model with dispersion among all ion pairs
Figiel2025SI Official ACS supporting information, DOI 10.1021/acs.iecr.5c00475.s001, Figshare article 28888514, file 54030209, MD5 4c3d3964d6e61b908c3ce3b17bb4e6b0 Comparison of direct, SSM, and SSM+DS formulations and retained validation evidence
Figiel2026Correction Figiel, Yu, and Held, Ind. Eng. Chem. Res. 2026, DOI 10.1021/acs.iecr.6c01415 Authoritative corrected text of Figiel2025 Eq. (7): restores the missing bracket and replaces the second epsilon_r,bulk with epsilon_r,ion
Zuber2014 Zuber et al., Fluid Phase Equilibria 2014, DOI 10.1016/j.fluid.2014.05.037 Ion-specific dielectric suppression, Eqs. (8)--(10) and Table 2
Yang2024 Yang, Kontogeorgis, and de Hemptinne, Fluid Phase Equilibria 2024, 583, 114103, DOI 10.1016/j.fluid.2024.114103 Schreckenberg/Uematsu--Frank density-and-temperature dielectric model, Eqs. (1a), (2a), and (2b), and solvent parameters
Rueben2024 Rueben et al., J. Chem. Eng. Data 2024, DOI 10.1021/acs.jced.4c00347 Partial-density perturbation-theory dielectric model, Eqs. (4)--(6), and PT2 ion parameters
Perdomo2025 Perdomo et al., Comput. Chem. Eng. 2025, 194, 108977, DOI 10.1016/j.compchemeng.2024.108977 Read-only NaCl--water vapor/liquid endpoint diagnostic; not an ePC-SAFT parameter source
Held2008 Held, Cameretti, and Sadowski, Fluid Phase Equilibria 2008, 270, 87-96, DOI 10.1016/j.fluid.2008.06.010 Source-complete water plus all 29 Table 2 ion parameters and fully dissociated water-electrolyte formulation; package reference domain 298.15 K and 1 bar
Bulow2021 Bülow, Ascani, and Held, Fluid Phase Equilibria 2021, 537, 112989, DOI 10.1016/j.fluid.2021.112989 Linear ion-mole permittivity rule (§2.3, Table 5) and Born charging work (Eq. 24) for salts in an ionic liquid, with no neutral species
Held2014 Held et al., Chem. Eng. Res. Des. 2014, 92, 2884-2897, DOI 10.1016/j.cherd.2014.05.017 Affine k_ij(T) laws of Tables 1-2 (pp. 2888-2889), temperature-law falsifiers; the text after Eqs. (21)-(22), the ion diameter from zero like-ion dispersion (§4)
Uyan2015 Uyan et al., Fluid Phase Equilibria 2015, 393, 91-100, DOI 10.1016/j.fluid.2015.02.026 Quadratic k_ij(T) at 298 K, Eqs. (26)-(27), and the Wolbach--Sandler volume correction outside the square root, Eq. (9) and Table 5; Eqs. (10)-(11), the ion diameter from zero like-ion dispersion (§4)
Bulow2021MDEA Bülow et al., Ind. Eng. Chem. Res. 2021, 60, 6327-6336, DOI 10.1021/acs.iecr.1c00176 Cubic water--MDEA k_ij(T), Eq. (26) and Table 3, temperature-law falsifier only
Ascani2022 Ascani, Sadowski, and Held, J. Chem. Eng. Data 2022, 67, 1972-1984, DOI 10.1021/acs.jced.1c00866 Case Study 2 component, association, k_ij, l_ij, and k_ij^hb records at 298.15 K
AscaniHeld2021 Ascani and Held, Z. Anorg. Allg. Chem. 2021, 647, 1305-1314, DOI 10.1002/zaac.202100032 Original Born equation and combined solvent-mass/ion-mole dielectric rule
Hong2027 Hong et al., Sep. Purif. Technol. 2027, 422, 140522, DOI 10.1016/j.seppur.2026.140522, with supporting information Table S1 Non-ionic pathway-weighted reactive cross-association: the symmetric reactive-site correction, Eqs. (7)-(10); the two-pathway blend, Eqs. (11)-(12); Tables 1-3 parameters as test declarations
HamerWu1972 Hamer and Wu, J. Phys. Chem. Ref. Data 1972, 1, 1047-1100, DOI 10.1063/1.3253108; official NIST PDF SHA-256 8399084b29b4db4ce35ab47ed679446dbedb9bf1686f99f43fce7b2073ee4beb Practical molal osmotic-coefficient definition and aqueous NaCl comparison value

Each equation below has one status:

  • source equation: a direct mathematical transcription from the cited source;
  • exact derivation: algebra or thermodynamic differentiation of named source equations;
  • implementation convention: an explicit indexing, unit, or interface choice;
  • bounded empirical relation: a source correlation used only in its retained domain.

Archive distillation anchors

The clean implementation was distilled from the read-only lab at tannerpolley/ePC-SAFT-lab; it was not independently reconstructed from this specification alone. The principal archived owners are:

Clean family Archived owner or evidence
Hard-chain and dispersion packages/epcsaft/src/epcsaft/native/eos/residual/cppad/hard_chain_dispersion.h
Association packages/epcsaft/src/epcsaft/native/eos/residual/contributions/association.cpp and the archived association tests
Debye--Huckel and dielectric model archived residual Helmholtz/electrolyte owners plus the checked local equation mirror distilled from the legacy LaTeX provenance
Corrected Born SSM+DS packages/epcsaft/src/epcsaft/native/eos/residual/contributions/born.cpp, the correction paper, and the archived Figiel evidence
Pressure and density closure archived state/property owners and branch tests; the clean solver deliberately removes unsafe warm-start relabelling

For the three retained methane/ethane neutral anchors, the clean and archived kernels agree within 5e-7 relative at identical T, rho_m, x, and parameters. The remaining systematic difference is the archive's N_A = 6.022140857e23 mol^-1 versus the exact SI value 6.02214076e23 mol^-1 used by the clean EOS. The clean goldens retain the exact SI value rather than preserving an obsolete constant for bitwise parity. Association, Debye--Huckel, Born, property, and density tests additionally use closed-form, thermodynamic-identity, finite-difference, or branch oracles that do not call the archived runtime.

2. Symbols, bases, and units

The EOS uses absolute temperature T [K], molar density rho_m [mol m^-3], SI number density rho_N = N_A rho_m [m^-3], and the Angstrom-basis number density rho_hat = rho_N / 10^30 [A^-3]. The identity

rho_m d/d(rho_m) = rho_N d/d(rho_N) = rho_hat d/d(rho_hat)

holds because the conversions are constant.

a^r = A^r/(nRT) is the dimensionless molar residual Helmholtz energy. Mole fractions are ordered by Parameters.component_ids and satisfy x_i >= 0 and sum_i x_i = 1. Ordinary ionic states also satisfy sum_i x_i z_i = 0 within the declared state tolerance.

Component inputs use:

  • segment count m_i [-], segment diameter sigma_i [A], and dispersion energy u_i/k_B [K];
  • association site topology, epsilon_AB/k_B [K], and kappa_AB [-];
  • integer charge z_i and source-formulation-specific dielectric and Born records; and
  • pair values k_ij for dispersion energy and, where the selected formulation requires it, explicit l_ij for cross diameter.

The clean native kernel converts every length to one basis before combining it with number density. It does not mix SI number density with Angstrom cubed.

3. Equation and implementation ownership

The equation catalogue gives each defining relation a stable ID and source locator. Its YAML index is generated, not independently edited. Historical _native paths and pure-version interfaces do not describe current ownership.

ADR 0006 retires residual point-multipole behavior while preserving the separate dipolar-density-perturbation relative-permittivity formulation; ADR 0011 readmits the optional dipole-dipole term of section 4.1 and ADR 0014 the optional quadrupole-quadrupole term of section 4.2. The dipole-quadrupole cross term stays retired. The common Helmholtz definition owns thermodynamic derivatives. Independent finite-difference formulas belong to numerical verification, not an alternate production equation owner.

See Figiel input identity and calibration evidence for the adopted packet and its source limits.

4. Neutral PC-SAFT

The neutral model uses the Gross2001 temperature-corrected diameters, mixture moments, hard-sphere Helmholtz energy, hard-chain contact correction, and two dispersion perturbation terms. The coefficient arrays a_0..a_6 and b_0..b_6 are the Gross2001 universal constants retained as A0, A1, A2, B0, B1, B2 in engine/native/contributions/helmholtz.hpp, transcribed from Gross2001 Table 1. The catalogue defines their interpolation; it does not duplicate the arrays.

For numerical evaluation of Gross2001 A.6--A.8, first form the composition moments $M_k=\sum_i x_i m_i d_i^k$ independently of density. With $\eta=(\pi/6)\widehat\rho M_3$, $u=1-\eta$, $A=M_1M_2/(M_0M_3)$ and $B=M_2^3/(M_0M_3^2)$, evaluate $a^{hs}=3A\eta/u+B\eta/u^2+(B-1)\log1p(-\eta)$. For $t_{ij}=d_i d_j\zeta_2/(d_i+d_j)$ the contact departure is $D_{ij}=\eta/u+3t_{ij}/u^2+2t_{ij}^2/u^3$. The chain uses $\log1p(D_{ii})$ directly; other contact consumers use $1+D_{ij}$. These are the same mixture equations at positive and zero series density, with density powers cancelled before arithmetic.

The pair rules consumed by the admitted bundles are

Defining relation: EqID: cross_segment_diameter, EqID: cross_dispersion_energy, EqID: same_sign_ionic_dispersion.

The l_ij = 0 Gross rule remains an implementation identity for bundles that do not admit that family. The Ascani2022 bundle instead requires an explicit l_ij record for every unlike pair and uses its Eq. S4 (1-l_ij) sign convention. A missing applicable unlike-component k_ij or Ascani l_ij record is an input error. For electrolyte families declaring same-sign ion exclusion, same-sign dispersion is equation-level NOT_APPLICABLE, not an implicit zero source record. Families declaring all-ion-pair dispersion retain the Gross combining rule for those pairs. The diagonal identities k_ii = l_ii = 0 are not pair records. An associating neutral subset selected from the Ascani bundle retains its explicit source l_ij; only bundles with no selected l_ij family use the zero-matrix identity.

For charged components, the electrolyte model family declares its source rule. Figiel2025 sets cation--cation and anion--anion dispersion energies to zero; Ji2012 retains dispersion among all ionic species. An ion declaring fixed.charged_molecule_ion is a charged molecule: it keeps dispersion with itself and with every other declared charged molecule, and takes the neutral Gross2001 diameter with its own u_i instead of a packing diameter. Held2014 and Uyan2015 derive the fixed ion diameter from zero like-ion dispersion; that a kept like-ion dispersion restores the temperature-dependent diameter is our inference from that derivation, which the sources state only in the zero-dispersion direction. Ji2012 is precedent for kept like-ion dispersion, not for the ion diameter, which its text does not state. The declaration is opt-in with no default. Opposite-sign and solvent--ion pairs always retain the Gross combining equation with their explicit source-backed k_ij. These ionic rules are part of the common scalar kernel, so Helmholtz values and CppAD composition derivatives consume the same definition.

4.1 Optional dipole-dipole term

A model that declares polar: dipole-dipole adds the Gross2006 third-order Padé dipole-dipole contribution to the residual Helmholtz energy. Every component supplies dipole_moment (debye) and dipole_count n_μ; zero marks a nonpolar species. With the reduced moment written q_i = n_μ,i μ_i^2/(4π ε_0 k T m_i σ_i^3), ε_i appears only in the J2 energy ε_ij = (ε_i ε_j)^0.5, which carries no k_ij. The chain lengths in the J integrals are capped at 2 (source Eqs. 15-16). The term uses the PC-SAFT packing fraction and segment diameters of the same state.

Defining relation: EqID: dipolar_helmholtz, EqID: dipolar_second_order, EqID: dipolar_third_order, EqID: dipolar_integrals.

When every dipolar species is absent, A2 = A3 = 0 and the Padé factor takes its limit 1. The value and the first composition derivative then equal the analytic limit a_dd ≈ A2 ∝ x_μ², so the absent species' dipolar chemical-potential share is zero. Higher composition derivatives along an absent dipolar direction are outside the claim; public actions report a support change there. The constants are fitted to two-centre Lennard-Jones plus point-dipole simulations at 0.55 ≤ T/T_c ≤ 0.99 with elongation L* ≤ 1. Polar parameters are valid only with their own fitted m, σ, ε/k and k_ij (ADR 0011).

4.2 Optional quadrupole-quadrupole term

A model that declares polar: quadrupole-quadrupole adds the Gross2005 third-order Padé quadrupole-quadrupole contribution instead. Every component supplies quadrupole_moment (debye angstrom) and quadrupole_count n_Q; zero marks a nonquadrupolar species. The reduced moment is q_i = n_Q,i Q_i^2/(4π ε_0 k T m_i σ_i^5), with the same debye bridge as the dipole, and the J integrals use the uncapped m_ij = (m_i m_j)^0.5 and m_ijk = (m_i m_j m_k)^(1/3) of source Eqs. 18-19. The two-body third-order integral is zero (Eq. 14).

Defining relation: EqID: quadrupolar_helmholtz, EqID: quadrupolar_second_order, EqID: quadrupolar_third_order, EqID: dipolar_integrals.

The absent-species limit and the support-change status are those of section 4.1. The constants are fitted to two-centre Lennard-Jones plus point-quadrupole simulations at elongations 0 ≤ L* ≤ 0.8 (m up to 1.8447) and 0.5 ≤ T/T_c ≤ 0.99; larger m, such as benzene's 2.2463, is the source's own extrapolation. The source finds its mixing rule unsuitable for real mixtures with more than one quadrupolar component. A model declares either the dipole or the quadrupole term, never both.

5. Association

The version-1 parameter document declares exactly one choice for each base, association, electrolyte, and permittivity family. Unknown or duplicate family kinds are input errors. Native admission accepts base pc-saft and association none or general-site. Association none requires no selected sites; contradictory topology is rejected. General-site with no selected sites has the exact zero-association limit. Changing a choice does not establish that an existing fit transfers to the new formulation.

Component IDs must be nonempty strings and unique in the supplied document before subset selection. Repeated records are ambiguous physical inputs, not parameter overrides. Every selected component supplies fixed.charge_number, including zero for a neutral species. Python transports that record; native admission rejects omission by component identity. No charge is inferred from a component name, source label, or absent field. A charged component may supply fixed.charged_molecule_ion with magnitude 1 (§4); admission refuses it on a neutral component, beside a packing_diameter, or with another magnitude. The version-3 Gross/DEEA packets state previously implicit neutral charges without changing coefficients.

For each explicit site A on component i, define the site fraction X_iA by the fixed-topology mass-action residual

Defining relation: EqID: association_mass_action.

The association strength is

Defining relation: EqID: association_strength.

These are source equations from Chapman1990 Eqs. (19) and (18), with explicit mixture/site indexing as an implementation convention. g_ij^hs is the Gross2001 hard-sphere contact function. Chapman1990 Eq. 18 carries d^3; the strength here carries sigma_ij^3, the PC-SAFT convention written explicitly by Nasrifar and Tafazzol 2010 (DOI 10.1021/ie901181n, Eq. 34). A parameter set fitted with d_ij^3 is a different model: its volume must be converted in the packet, and no constant conversion is exact because (d/sigma)^3 depends on temperature. See association source conversions for packet-specific evidence. Every admitted site pair is an explicit topology edge whose source is either a direct coefficient record or one of two named closed-registry rules. For unlike components i and j, both rules resolve

epsilon_iA,jB/k_B = [(epsilon_iA,iB/k_B) + (epsilon_jA,jB/k_B)] / 2
kappa_iA,jB = sqrt(kappa_iA,iB kappa_jA,jB).

arithmetic-energy-geometric-volume uses these values unchanged. wolbach-sandler is Gross2002 Eqs. 2--3, which multiply the volume by the size factor [sqrt(sigma_i sigma_j) / ((sigma_i + sigma_j)/2)]^3. Because Delta already carries sigma_ij^3 = [(sigma_i + sigma_j)/2 (1 - l_ij)]^3, the size factor is applied exactly by using [sqrt(sigma_i sigma_j)(1 - l_ij)]^3 in Delta for that edge. The state-evaluated sigma(T) keeps the temperature dependence and its derivatives, which matter for the water correlation. The size factor is 1 for like pairs.

Both inputs must be donor--acceptor self-edges for the same two components and must contain the output sites. The output must itself connect one donor and one acceptor. Resolved outputs are derived, non-selectable coordinates; their recorded derivative chain is 1/2 for each energy input and kappa_iA,jB/(2 kappa_input) for each volume input. This lets later exact parameter derivatives propagate through the rule without treating its output as an independent fit coefficient.

See association source conversions for the immutable packet history. An omitted unlike-site pair remains inactive, not inferred. String schemes, inferred donor/acceptor pairs, implicit combining, and default insertion are forbidden.

Every selected pair of components that both carry sites must therefore declare its cross association: at least one unlike-site edge (explicit or a named rule), or one topology.no_cross_association item naming component_id_a, component_id_b and its provenance. The declaration makes every unlike-site edge of that pair inactive; it preserves a model, not evidence that the physical interaction is absent. build_parameters refuses a missing declaration before any evaluation, naming both components, and refuses a duplicate, a self or unknown pair, a nonzero native marker and a declaration contradicted by a resolved active edge. A declaration outside the selection is ignored. The native entry is no_cross_association on the unordered component pair with a zero dimensionless marker, so Python, direct C++ and ASL records share this owner.

The Ascani2022 water--1-butanol selection likewise stores both cross directions explicitly. Their energy is the SI Eq. S6 arithmetic mean multiplied by 1-k_ij^hb with the source value k_ij^hb=0.026; their volume is evaluated once from the cited Wolbach--Sandler rule. The distinct k_ij^hb provenance is therefore retained without adding a runtime pair matrix or overloading k_ij.

Temperature laws. A pair k_ij or l_ij record, an explicit edge's energy_over_k or volume, and a combining-rule edge's volume_correction carry exactly one of value or temperature_law:

{"identity": "uyan2015/water-co2/k_ij", "family": "k_ij",
 "temperature_law": {"form": "polynomial-plus-reciprocal-temperature",
   "reference_temperature": {"magnitude": 298.0, "unit": "kelvin"},
   "terms": [{"power": 0, "coefficient": {"identity": "uyan2015/water-co2/a0", "value": {"magnitude": -0.022, "unit": "dimensionless"}}},
             {"power": 1, "coefficient": {"identity": "uyan2015/water-co2/a1", "value": {"magnitude": 0.00042, "unit": "1 / kelvin"}}}],
   "reciprocal_temperature_slope": {"identity": "...", "value": {"magnitude": 0.0, "unit": "kelvin"}}}}

Defining relation: EqID: pair_parameter_temperature_law. Powers are unique integers 0-3 and the power-0 term is required; omitted terms are absent. T_r > 0 is fixed, not fitted. Coefficient a_p has unit kelvin ** (e - p) and s unit kelvin ** (e + 1), with e = 1 for the association energy and e = 0 otherwise, spelled dimensionless, kelvin, kelvin ** n, 1 / kelvin or 1 / kelvin ** n. Each coefficient is one temperature_law_coefficient active slot selected by its identity; a law owns its symmetric pair in both directions, and its stored constant and k_ij reciprocal slope are unconsumed. A record repeated under one identity (for example one volume correction on both cross edges of a component pair) is one law and one set of slots. Source validity ranges remain the caller's. An evaluated k_ij must be finite, l_ij < 1, association energy and volume nonnegative, at every state: a law leaving that domain is refused with its target and temperature, never clamped. The same finite l_ij < 1 rule admits constant pairs and active fitting trials, preserving positive unlike diameters.

A combining-rule edge that declares a volume_correction, or whose self edges carry a law, is evaluated at each temperature from its explicit self edges: EqID: association_volume_correction, the Uyan2015 Eq. (9) convention with (1 - c) outside the square root and the size factor applied once by wolbach-sandler. Its self-edge constants keep their derivative chain; its own energy and volume are not coordinates. Rule edges without these declarations keep construction-time resolution. Held2014's corrected water/1-butanol cross energy (2420.53610 K) is representable as an explicit energy; adopting that whole edge, with its Eq. (18) volume, is not.

The association Helmholtz contribution is

Defining relation: EqID: association_helmholtz.

Pathway blocks. Hong2027 blends two simultaneous association solutions of one mixture. A topology site may declare "pathway": "weighted" or "pathway": "complement", and a reactive site may declare "symmetric_reactive_site_correction": true. The declared sites form one augmented site vector: no edge crosses blocks (admission refuses one), each block keeps its own mass action, and an undeclared site has weight 1.

Defining relations: EqID: pathway_weighted_association_helmholtz, EqID: symmetric_reactive_site_correction, EqID: pathway_weight and EqID: pathway_weighted_cross_dispersion.

The top-level pathway_weight record declares the loading components (loading.numerator, loading.denominator) and three dimensionless coefficients with families intercept, loading_slope and temperature_slope; a pair may add a k_ij_complement temperature law, the complement pathway's k_ij. A weight is required exactly when a block or a complement law uses it. w is a state function of T and x, so partial w/partial n_i and partial w/partial T enter every property and derivative; each phase's State reports its w. A zero loading_slope and temperature_slope is the fixed pathway weight: exact for the declared w, but resetting w between states (the Hong2027 Table S1 convention) is not one fluid model. A nonzero or active loading_slope is refused at a zero loading-denominator amount. Which reactive-partner site bonds the reactive sites is a declaration, never inferred from species.

The corrected term is not stationary in X: the site chain below is complete for it, and the energy's error estimate uses the signed site gradient (algorithms.md, ALG-ASSOCIATION).

The value solver runs outside CppAD recording. For independent variables q, CppAD supplies the partial Jacobians of H(X,q) and the code solves

dX/dq = -(partial H/partial X)^-1 (partial H/partial q).

For q,r drawn from T, rho_m, and independent composition coordinates, the second sensitivity consumed by a_TT, a_T,rho, and a_rho,rho is

X_qr = -H_X^-1 [H_qr + H_qX X_r + H_Xr X_q + H_XX[X_q,X_r]].

These are implicit-function identities, not a claim that every displayed derivative is implemented. An equivalent stationary elimination must reproduce them on its regular domain. The EOS rejects X outside (0,1], a nonfinite residual, failed convergence, or an ill-conditioned association Jacobian.

6. Bulk dielectric and Debye-Huckel

The active record includes charged-slot Debye-Huckel and packing diameters in Angstrom and dimensionless ion-suppression coefficients. The scalar-generic permittivity and screening owners consume that record directly. Pressure-closed permittivity columns include the exact density chain, including the density topology, and ratios differentiate both states. Model choices, charges, masses and dielectric source correlations stay fixed. Parameter admission and fitting execution belong to ALG-PARAMETERS and ALG-FIT.

The relative-permittivity topology is selected by a complete, source-bearing parameter family. It is not a caller option and there is no phase label or runtime fallback. Mixed or incomplete topology records fail during model resolution.

The solvent-mole-weighting declaration selects neutral component identifiers in solvents and requires normalization equal to unnormalized or solvent-normalized. It consumes the relative-permittivity laws of the declared solvents. Charged, missing, duplicate or unknown solvent identifiers and unconsumed dielectric laws are refused. An actually neutral-only component selection retains the removed-electrolyte convention. Defining relation: EqID: solvent_mole_weighted_bulk_permittivity. Bülow2021a Eq. (19) supplies the unnormalized relation; normalization is the explicitly requested alternative. Existing mass and density rules retain their arithmetic.

Declaration Equation Source
Component constant or temperature polynomial Component polynomial, with the sum empty for a constant Engine component-correlation convention
solvent-mole-weighting, either normalization Solvent mole weighting Bülow et al., Eq. (19), with solvent normalization as an Engine convention
solvent-only Salt-free solvent mass weighting Figiel et al., Eq. (11) at zero suppression
ion-fraction-suppression or ion-specific-suppression Common or species-specific suppression Figiel et al., Eq. (11), or Zuber et al., Eqs. (8)–(10)
component-permittivity-mixing Solvent-mass and ion-mole weighting Ascani and Held, Eq. (13)
solvent-density-temperature Separate parameter averages in the density law Yang et al., Eqs. (1a), (2a), and (2b)
solvent-density-temperature-ion-specific-suppression Density law divided by the suppression factor Engine combination of Yang and Zuber relations
solvent-density-temperature-ion-specific-susceptibility-suppression Susceptibility divided by the suppression factor Engine relation selected in #78
dipolar-perturbation Dipolar response Rueben et al., Eqs. (4)–(6)
none Electrolyte contribution removed Engine convention

solvent-volume-weighting uses the Novak solvent-volume rule. Declare solvent_molar_volume component coefficients in liter / mole for exactly the neutral species with dielectric laws. The permittivity record requires a common reference_temperature in kelvin and reference_pressure in pascal, each a quantity with magnitude and unit. The component coefficients are pure-liquid molar volumes on this reference basis. The state evaluation uses the supplied volumes as constants and requires positive total selected solvent volume when evaluating the rule. The weights always use solvent normalization.

full-species-mole-weighting uses the linear species sum. Declare a relative-permittivity law for each neutral component and an explicit relative_permittivity value for each ion. All weights are total species mole fractions.

The component correlation form constant-plus-polynomial-and-reciprocal-in-temperature accepts the polynomial fields and a reciprocal_temperature_coefficient with an identity and a value in kelvin. Its denominator is absolute temperature. The reciprocal-temperature equation gives the water coefficients and the temperature interval for their use.

6.1 Figiel composition-only and Ascani mole-fraction topologies

For the bounded Figiel2025 solvent/electrolyte model, first form salt-free solvent mass fractions from component molar masses:

Defining relation: EqID: salt_free_solvent_permittivity.

Figiel2025 Eq. (11) then gives

Defining relation: EqID: ion_suppressed_bulk_permittivity, with alpha_i = C_epsilon for every ion.

C_epsilon = 7.01 is a source-bearing model record. The relation is empirical for water and methanol correlations, was checked against ethanol, uses data up to about x_ion = 0.38, and was evaluated by Figiel2025 at 298.15 K. Other temperatures require source-backed pure-solvent dielectric correlations and do not broaden the empirical ion-suppression evidence.

For the bounded Ascani2022 Case Study 2 model, AscaniHeld2021 Eq. 13 instead defines

Defining relation: EqID: component_mixed_bulk_permittivity.

The source records set epsilon_r,i = 8 for the admitted alkali-halide ions. That mixture quantity is not Figiel's radial epsilon_r,ion: the Ascani bundle contains neither the Figiel radial-region model record nor a dielectric suppression coefficient. Its water and 1-butanol source correlations are evaluated exactly at the bundle's sole admitted temperature, 298.15 K. CppAD differentiates the displayed composition rule in the common owner.

With every species charged (an ionic liquid plus a salt), the solvent sum is empty and the rule is epsilon_r,bulk = sum_i x_i epsilon_r,i (Bulow2019 Eq. 7; Bulow2021 §2.3, Table 5). Both component mixing and full-species mole weighting admit this charge-only parameter set. The Debye-Huckel term and the Born charging work from vacuum (Bulow2021 Eq. 24) need epsilon_r,bulk > 0. Stability admits ionic trials for these forms on the condensed side of their own composition's loop (ALG-STABILITY).

6.2 Schreckenberg density topology

Yang2024 Eqs. (1a), (2a), and (2b) specify the Schreckenberg/Uematsu--Frank topology, admitted as the permittivity choice solvent-density-temperature with component records dielectric_volume (liter / mole) and dielectric_temperature (kelvin). With S the neutral-solvent set,

Defining relation: EqID: density_correlation_mixed_parameters, EqID: density_correlated_bulk_permittivity.

solvent-density-temperature-arithmetic consumes the same component records and uses the Perdomo arithmetic rule. It evaluates each solvent's temperature law before averaging the resulting coefficients on the salt-free solvent mole basis. The Engine multiplies the resulting average by its solvent partial density, the solvent moles per volume of salted solution. Both forms coincide for a single solvent.

The source water comparison uses d_V = 0.3777 dm3 mol^-1 and d_T = 1403 K from Yang2024. Because solvent partial density is explicit, epsilon_r,bulk -> 1 as rho_m x_solv -> 0. Every neutral solvent requires both parameters; missing records, records on an ion, or a constant solvent permittivity record under this choice fail closed.

The choice solvent-density-temperature-ion-specific-suppression retains those solvent records and requires one dimensionless ion_specific_suppression_coefficient per charged species, including explicit zeros. Each coefficient is an effective value for the declared solvent system. Signed coefficients are admitted where the denominator and resulting permittivity are finite and positive and the neutral solvent fraction is finite and positive.

Defining relation: EqID: ion_suppressed_density_correlated_bulk_permittivity. The fixed-composition zero-density limit is $1/D$. Charged trials use the own-composition liquid-side root rule in ADR 0012.

The choice solvent-density-temperature-ion-specific-susceptibility-suppression uses the same records and positivity requirements. It divides the solvent susceptibility $F-1$ by $D$, giving $1+(F-1)/D$ and a fixed-composition vacuum limit of 1 for every admitted $D$. The denominator must remain positive at zero density.

Defining relation: EqID: susceptibility_suppressed_density_correlated_bulk_permittivity.

6.3 Zuber ion-specific topology

For a single source-supported solvent, Zuber2014 Eq. (9) is

Defining relation: EqID: ion_suppressed_bulk_permittivity, with one alpha_i,s per ion.

The water/Na+/Cl- comparison uses the Table 2 coefficients alpha_Na+,water = 3.62 and alpha_Cl-,water = 6.80. These parameters are ion--solvent specific and are not replacements for Figiel's universal C_epsilon = 7.01. Zuber Eq. (8) requires solvent volume fractions for a mixed-solvent system. The current topology therefore admits exactly one neutral solvent and rejects a mixed-solvent selection rather than inventing volume data or a mixing rule.

6.4 Rueben perturbation topology

Rueben2024 Eqs. (4)--(6) are specified in the source's Angstrom number-density basis, admitted as the permittivity choice dipolar-perturbation with component records dipole_scaling (debye ** 2), polarizability_scaling (angstrom ** 3) and correlation_integral_parameter (dimensionless). Let rho_hat_i = rho_m N_A x_i 10^-30 in particles per Angstrom cubed:

Defining relation: EqID: effective_dipolar_density, EqID: dipolar_correlation_integral, EqID: dipolar_perturbation_bulk_permittivity.

mu2_i, alpha_i, and a2_i are the source-backed dipole-squared, polarizability, and correlation-integral records. The water/Na+/Cl- comparison uses water mu2=5.199 D2, alpha=0 A3, a2=0.1276, and Rueben2024 PT2 ion values Na+ (5.407 D2, 0 A3, 0.215) and Cl- (2.563 D2, 0 A3, 0.138). All three records are required for every selected component. The equation gives epsilon_r,bulk -> 1 as all partial densities approach zero.

6.5 Common electrostatic owner

The selected epsilon_r,bulk is evaluated once by the scalar-generic native owner and is consumed unchanged by both Debye--Huckel and the selected Born equation. Temperature, density, composition, and parameter dependence therefore remain inside the common CppAD tape. The parameter record identifies the generic bundle-defined formulation, and State.bulk_relative_permittivity reports the evaluated state value.

See permittivity comparisons on the Figiel parameter base for the retained endpoint and liquid-property evidence.

The defining Debye screening quantity is

Defining relation: EqID: debye_screening_parameter.

Only charged components enter the following sum. Their finite-size DH diameter is a separately declared record. Equality to a packing or Born diameter is allowed only when the selected source prescribes it. Source-specific examples:

Figiel: d_i^DH(T) = 0.88 sigma_i,
Ascani: d_i^DH(T) = sigma_i.

The Figiel relation follows its Eqs. (3)--(4) and admitted u_ii = 0 convention. AscaniHeld2021 explicitly uses sigma_i in the hard-chain, Debye--Huckel, and Born terms. The DH owner consumes its declared d_i^DH; the packing owner consumes d_i^pack and Born consumes d_i^Born. A source relation between these records must be retained explicitly, not inferred universally. Here sigma_i and d_i^DH use the same length basis and d_i^DH is converted to metres in the SI DH equation. For each charged component,

Defining relation: EqID: debye_finite_size_chi.

The residual contribution is

Defining relation: EqID: debye_huckel_helmholtz.

When sum_i x_i z_i^2 = 0, the EOS defines kappa = 0 and a^DH = 0 without evaluating chi_i; this is the exact neutral limit. The fitted Born diameter is used only in the Born term. CppAD differentiates the defining relations above on the positive-ionic-strength branch; the lab dkappa/dx_i catalogue entry is not a production owner.

7. Canonical Born source reconciliation

Charged-slot Born diameters belong to the same scalar active record; the corrected Born owner consumes them directly. Source correction factors and solvent solvation factors stay fixed. A declared native hydration hold treats the diameter as a dependent coordinate through EqID: dependent_born_diameter_derivative; the observed target uses the single-ion x-process J/mol convention. The hold is not a high-weight residual. Stored-diameter acceptance is separate from the inner root solve. No cross-family diameter tie or coupled hold is qualified here.

Figiel2025 defines a fitted d_i^Born and the shell thickness

f_mix = sum_k x_k f_k / sum_k x_k,
Delta d_i = [(f_mix - 1)/|z_i|] d_i^Born,
D_i = d_i^Born + Delta d_i.

For the admitted systems the sum contains only solvents and ions and therefore sum_k x_k = 1. Ion values f_i = 1 are source records, not defaults.

Corrected Eq. (7)

The literal rendered Eq. (7) on article p. 9408 places (1 - 1/epsilon_r,bulk) in both radial terms. That form simplifies to the direct Born expression at d_i^Born; the shell thickness cancels. It therefore contradicts all of the following evidence in the same source:

  • Figure 1 shows epsilon_r,ion between the ion and shell and epsilon_r,bulk outside the shell;
  • the preceding prose assigns epsilon_r,ion = 8 from cited salt data;
  • the following prose says SSM+DS depends separately on d_i^Born and D_i; and
  • Figiel2025SI compares SSM+DS with direct and SSM models as distinct calculations and selects s-SSM+DS-f.

Figiel2026Correction confirms that the original Eq. (7) omitted one bracket and incorrectly named the second dielectric constant epsilon_r,bulk instead of epsilon_r,ion. The required canonical primary-source equation is:

Defining relation: EqID: born_outer_shell_diameter, EqID: born_residual_helmholtz.

Every d_i^Born and D_i in this equation is converted from angstrom to metre before use with the SI factor e^2/(4 pi epsilon_0 k_B T). This equation is a corrected source equation. It matches the radial two-region model, retains both limiting diameters, and matches the archived candidate kernel. epsilon_r,ion = 8 enters through a Figiel/Buelow source-bearing model record. The native source must not embed 8 or 7.01.

The fixed pairs (c_shell, c_dielectric) = (0,0), (1,0), (1,1) recover the uncorrected, shell-only, and shell-plus-dielectric limits. The redundant (0,1) pair is rejected. Source packets retain the literature-specific diameter values and provenance, but all resolve through this single equation; no source or bundle selects a different runtime Born owner.

8. Common derivatives and state properties

CppAD records the common a^r(T,rho_m,x) equations for fixed component and site topology. Association uses the implicit sensitivities in Section 5.

Define

Defining relation: EqID: compressibility_factor, EqID: pressure.

The unconstrained composition partials used below follow the Gross2001 constant-T, constant-density mixture convention. For component i,

Defining relation: EqID: residual_chemical_potential, EqID: log_fugacity_coefficient.

The chemical-potential convention and fugacity coefficient must be explicit. It does not publish the lab's nonunique contribution allocation of ln Z.

The exact residual caloric identities at the state density are

Defining relation: EqID: residual_internal_energy, EqID: residual_enthalpy, EqID: residual_heat_capacity_cv.

State values and actions evaluate residual enthalpy as residual internal energy plus the direct residual pressure-volume term, without subtracting rounded ideal terms from the compressibility factor.

With Z_T = (partial Z/partial T)_(rho_m,x) and Z_rho = (partial Z/partial rho_m)_(T,x),

Defining relation: EqID: residual_heat_capacity_cp.

The -1 removes the ideal-gas c_p-c_v = R contribution. The EOS also reports the same-density residual entropy

Defining relation: EqID: residual_entropy.

If it reports the departure entropy relative to an ideal gas at the same T,P, the result is named separately and equals s_V^r/R + ln Z. No result named only residual_entropy may switch reference convention by state route.

For the current central state result, c_p^r is available only when D = Z + rho_m Z_rho = P_rho/(RT) is finite and positive and the resulting quotient is finite. Positivity is the existing mechanically stable branch rule; no additional universal slope cutoff is adopted. A numerical accuracy claim at that state additionally requires a recorded fixed-pressure enthalpy-derivative ladder on the same branch. A near-zero positive D is retained as conditioning stress outside the qualified claim, not silently rejected or treated as accurate. P4 owns general per-action conditioning and unavailable behavior. Explicit-density evaluation does not reject an otherwise finite state solely because the slope is negative; it reports the other mathematical EOS results without certifying stability and leaves c_p^r unavailable.

On a derived_properties request, State also assembles fixed-composition response and partial molar properties from the same T/V Hessian columns, with no further derivative evaluation:

Defining relation: EqID: isothermal_compressibility, EqID: isobaric_expansivity, EqID: heat_capacity_ratio, EqID: isentropic_compressibility, EqID: speed_of_sound, EqID: joule_thomson_coefficient, EqID: partial_molar_volume, EqID: partial_molar_enthalpy, EqID: partial_molar_entropy.

The same positive-D rule admits every one of them. Values that need total c_p or c_v, and the partial enthalpy and entropy, follow the thermochemistry availability of Section 9. Unavailable values stay unavailable; no floor or residual-only substitute is returned. The six named residual contributions, dipolar included, sum to residual_helmholtz.

9. General phase energy and caloric basis

The phase formulation defines amount-volume coordinates for mixtures, association and electrolytes. Pure nonassociating states are a restriction of that definition, not a separate thermodynamic owner. Observable derivative orders and chain rules are specified in algorithms.

Residual enthalpy and internal energy differ by the residual pressure-volume term; total enthalpy and internal energy differ by the full pressure-volume term. Missing reference thermochemistry means total calorics are unavailable, not equal to residual calorics. Standard-pressure and standard-density bases must be converted consistently; see EqID: standard_pressure_density_conversion. A reaction-consistent reference gauge alone does not determine absolute enthalpy, entropy or heat capacity. An empty record list declares a species without ideal-gas data; its Gibbs standard and consumed derivatives are completed from the reaction chain only when supplied records leave nothing but the ionic charge gauge (EqID: reference_species_calorics). Caloric completion retains the supplied h/s anchors, independently of the chemical rows' conserved gauge. Fixed source pressure stays fixed when system pressure moves; system-pressure references move with it. Their frozen internal energy and heat capacity use the same Gibbs-conjugate volume (EqID: gibbs_conjugate_molar_volume), separately from EOS mechanical volume. A regular local volume chart does not establish a globally stable reacting minimum.

10. Electrolyte reference states

Figiel2025 Eqs. (17), (19), (20)–(22) define solvation, transfer, single-ion reference activities and stoichiometric mean ionic activity. Hamer–Wu1972 Eqs. (3.4′), (3.5) define the practical molal osmotic coefficient. The canonical definitions are EqID: infinite_dilution_solvation_energy, EqID: transfer_free_energy, EqID: single_ion_activity_coefficient, EqID: mean_ionic_activity_coefficient, EqID: molality_activity_conversion and EqID: practical_osmotic_coefficient. The scalar solvent difference uses the extensive residual Helmholtz Euler identity at prescribed pressure before the molality division; its exact parameter derivative uses the shared CppAD fugacity columns. This protects dilute cancellation without changing the thermodynamic definition or reporting floor.

Declare solvent composition, component order, charges, formula-unit stoichiometry, molar masses, concentration basis and reference pressure. Dilute along an exactly electroneutral path with fixed salt-free solvent composition. Single-ion quantities are path/gauge dependent; charge-neutral stoichiometric contractions carry the physical comparison. The neutral basis is EqID: neutral_reference_stoichiometric_basis; its existence does not establish current public availability. Mixed-solvent MIAC does not imply a unique scalar osmotic coefficient. Multi-salt paths require a separately specified reference convention.

Properties supplies the raw lower reference partials C_T, C_TT, C_theta, C_Ttheta and C_TP and the bounded higher-order reference columns in EqID: reference_mixed_temperature_pressure_derivatives, using EqID: reference_single_insertion_taylor. The F8/F9 notebook (private research repository) qualifies the stated finite-terminal derivatives and observed dilution changes. It does not extend reacting caloric/action consumers by implication.

The analytic first insertion is EqID infinite_dilution_reference. One ion-free solvent root and the existing association solution supply its values and exact T/P/parameter derivatives. Born remains nonzero; DH has zero first insertion. Neutral solvent mixtures use the same amount insertion. Charged backgrounds are refused, even if net neutral. The former EqIDs reference_molality_sequence and reference_limit_terminal_pair are retained qualification calculations. No terminal-change error is measured by the analytic method; retain actual density and site diagnostics without assigning a zero error estimate.

A neutral solute in one pure solvent of an all-neutral model has the fugacity-based Henry constant of EqID: henry_constant_px, read from one unit reaction row of the same reference owner and analytic solvent insertion. It is lim f/x, not the partial-pressure volatility p/x. Normalized infinite-dilution activity coefficients tend to one and carry no absolute fugacity scale.

Thermochemical standard states and reaction correlations belong to the common Engine property/formulation owners; they do not modify the residual EOS. Their coefficients require complete independent source evidence. Reference pressure is not the trial pressure unless the declared transformation says so.

11. Density closure

Use EqID: density_residual, EqID: pressure_density_derivative and EqID: mechanical_stability. The root-selection algorithm and failure meaning are ALG-DENSITY. A mathematical fixed-density state can exist on a mechanically unstable branch. A specified-pressure physical phase must meet the root's mechanical criterion; neither choice establishes global equilibrium.

12. Domain and rejection rules

The optional cation_mixing declaration is one record or an array of records. Each record adds the empirical solvent-mass pair/triple term in EqID cation_pair_triple_mixing_helmholtz and names one neutral solvent, two distinct cations and one declared anion, with both constant coefficients explicit. Records add independently; an ion in several records receives their sum. No three-cation cross term is included. Repeated unordered cation pairs are refused. An identical record shared across merged packets is accepted once. Admission orders records lexicographically by the two sorted cation component ids, independently of component or input-record order. The term is evaluated in every phase on the molality basis of that declared neutral solvent, with positive solvent mass and the declared anion retained. Additional anions are admitted: $\theta_{ij}$ is independent of the anion, while only the declared $\psi_{ijk}$ is nonzero; every other anion triple coefficient is zero as a model assumption (Pitzer and Kim, 1974, Eqs. 8, 9 and 11, doi:10.1021/ja00825a004; Peiper and Pitzer (1982), J. Chem. Thermodyn. 14, 613–638, Eqs. (7)–(8), Zotero X8GEDSJ2). The unequal-charge electrostatic $E\theta$ terms are excluded. Transferring chloride-fitted coefficients to sulfate is unvalidated. Other neutral components enter this contribution only through mole fractions and total-mole normalization; their masses are not added to the declared solvent mass. There is no phase selector or automatic deactivation in an organic phase. Omitting either cation from a component selection omits the term; a zero amount in a retained component set keeps its insertion derivatives when the declared solvent amount is positive. With no declaration, arithmetic is unchanged. The cation-mixing qualification records the retained domain and extrapolation limits.

At exact $x_s=0$, if $x_i=0$ or $x_j=0$, define the contribution as zero on that fixed support. Every derivative within that support, including thermal, volume and active-coefficient derivatives, is exactly zero. Dense arrays store zero transverse mixing entries; these are defined storage, not physical insertion derivatives or limits of the aqueous Hessian. In general $F_{ij}=2\theta/W+\psi n_k/W^2$ diverges as $W\to0$, so this definition is not a continuous full-composition extension. Dry states with both cations present remain refused even with zero coefficients or an underflowing cation product. The individual exact-zero comparisons introduce no water threshold; positive-solvent arithmetic and its positive-mass refusal are unchanged.

Existing IonicInsertion refusals for absent-ion actions and for ionic values in an uncharged phase remain. In a dry phase containing one participating cation, the absent partner's scalar chemical potential and fugacity, including fugacity actions, are also unavailable with IonicInsertion. Separate scalar insertions when both cations are absent do not authorize mixed insertion actions. Component support signatures distinguish dry and wet tapes; these stored zeros cannot authorize continuation across the dry boundary. This admits neutral pure-organic reference evaluations without qualifying organic ions or predicting Lithium extraction. Other neutral components retain their existing effects on the other EOS contributions.

Finite temperature bounds attached to the selected source records are intersected by the resolver and enforced for every topology. A record whose source domain is explicitly unknown contributes no invented numerical bound.

See the Figiel and iodide input domain for the source-specific selection and pressure restriction.

Relative-permittivity behavior is selected by one source-neutral Model Family choice in the canonical Parameters document. Section 6 lists its choices and equations. A choice consumes only its own records; any other permittivity record is rejected. Citations to the papers that support a choice and its coefficients remain in provenance; paper names do not select runtime implementations. Parameters.from_mapping(...) and Parameters.from_json(...) parse the same strict document, which must provide every coefficient and one complete fixed topology needed by the selected model. No model choice or topology broadens the temperature, pressure, composition, or fitted-parameter evidence recorded by its sources.

See Ascani Case Study 2 input selection for the source-specific pair classifications.

The Held2008 input selection and reference restriction retains its fail-closed source-state requirement.

The fully dissociated Debye--Huckel model family also accepts explicitly declared neutral--ion and opposite-charge k_ij records when a different source strategy requires those dispersion interactions. Same-sign ionic dispersion remains identically suppressed by the EOS and its declared pair record must be k_ij = 1. The Engine does not infer either strategy from a component identity or application.

Topology is determined by the selected components, not by unrelated records remaining in their source bundle. A selection containing no nonzero charge is neutral or associating according to its explicit sites; ionic contributions activate only when the selection contains both positive and negative charges.

Construction or evaluation fails on:

  • missing parameter, model, pair, site, source, unit, or domain records;
  • non-electroneutral ordinary ionic composition;
  • nonpositive dielectric values or Born/shell diameters;
  • association nonconvergence or an ill-conditioned association Jacobian;
  • packing-limit violation, density branch ambiguity, or mechanical instability of a specified-pressure root; and
  • a requested derivative or property whose defining Jacobian is singular.

13. Scope and derivative boundaries

EOS supplies residual thermodynamics. Equilibrium, reaction standard states, reference thermochemistry and fitting have separate responsibilities inside the same Engine. Their absence in the current implementation is a transfer gap, not a permanent ban on the accepted target. Solids, critical-point conditions as an augmented equation system, kinetics and Bjerrum pairing require separately justified scientific extensions. ADR 0011 admits the optional Gross–Vrabec dipole–dipole term, and ADR 0014 admits the optional Gross quadrupole–quadrupole term. A model selects either term, never both; the dipole–quadrupole cross term remains retired under ADR 0006.

Residual EOS parameters and reference thermochemistry are separately declared physical records. Public density/pressure state calls supply temperature and composition with exactly one of density or pressure; native amount-volume coordinates retain their defining meaning. Reference calculations additionally require the declared stoichiometry, basis and standard state. Physical-input resolution and unit arithmetic belong to the native admission owner.

The model/property owner converts declared reaction correlations to reference Gibbs offsets and their temperature derivatives; equilibrium consumes those values consistently. Request value and derivative orders explicitly. Missing thermochemistry or unsupported actions produce typed unavailable total fields while preserving valid residual properties. These are required semantics; class names and installed availability belong to the public interface and the limits page, not this mathematical specification.

On an ion-free support, Born and DH are identically zero before reciprocal permittivity, solvation mixing or screening operations are evaluated. Their supported temperature, density, neutral-composition and parameter actions, including mixed actions, are exactly zero. Pure CO₂ has no dielectric solvent in this model: its zero bulk-permittivity diagnostic supplies no ionic-insertion limit. The empty electrostatic sums require no replacement permittivity or floor.

With finite positive limiting solvent permittivity, a dilute ionic path has a linear Born leading term and a DH leading term proportional to ionic strength to the power 3/2. Born's insertion first derivative is generally nonzero; DH's insertion second derivative diverges. Neutral-restriction derivatives therefore do not establish unrestricted composition derivatives. Absent charged ordinary chemical-potential/fugacity slots and their parameter columns are unavailable with IonicInsertion; present neutral slots and regular scalar calorics remain available. Pressure closure preserves this admission. Full-species insertion planes on a zero-inventory face require separate qualification.

Zero amount of an associating species is not permission to add a positive composition floor. The unweighted mass-action Jacobian may remain regular where a composition-weighted stationary association Hessian becomes singular. Retain valid central states and report unsupported derivative directions separately. See ALG-ASSOCIATION.

Ordinary physical ionic states require electroneutrality and explicit charges. Unconstrained composition partials use a declared mathematical extension; they do not admit a charged physical feed. A neutral subset is selected from its own components/sites, not unrelated ionic records in its source packet. Source extrapolation must be identified separately from mathematical domain validity; convergence does not expand a source's applicability.

Central mixing-energy and entropy terms use the continuous limit of n_i log n_i at zero component amount. Positive total amount suffices for mole fractions; logarithmic coordinates and insertion chemical potentials require separate interior/limit treatment. An incipient phase at zero total amount uses the independent intensive chart in the equilibrium formulation.

Appendix: Special cases and numerical qualification

Figiel input identity and calibration evidence

The source-complete starting input is the immutable figiel-yu-held-2025/3 packet, with parameter JSON SHA-256 3d37235cd8217e056f55005bfdf2b65dcf7c4a8621e239bdd40a70c273a9be8c. Its packet fingerprint is retained in packet-locks.json. Supply the JSON to Parameters.from_json and an ordered component subset; no paper name selects physics. For the fixed-density equation control, the subset is (water, sodium-cation, chloride-anion), T = 298.15 K, rho_m = 56000 mol m^-3, and x = (0.98, 0.01, 0.01) in that order. The calculated pressure is about 17.1 MPa, outside the packet's reported 1-bar conditions; it tests admission and equations, not empirical accuracy.

The document declares base: pc-saft with the Gross reference and dispersion rules of §4; association: general-site with explicit water A/B sites and no ion sites (§5); electrolyte: born with Debye--Huckel, same-sign ionic dispersion exclusion, and corrected SSM+DS Born (§§6–7); and permittivity: ion-fraction-suppression with salt-free solvent mass averaging and source coefficient 7.01 (§6). Its explicit charge, pair, association, and correlation records supply units and source locators. Segment, ionic packing, Debye--Huckel, and fitted Born diameters are distinct inputs; their source relations are stated in §§4, 6, and 7. Equilibrium and reported activities use the existing reference conventions in §10. The packet's reported evidence domain is 298.15 K, 1 bar, and ion mole fraction up to about 0.38. It does not provide missing ideal caloric inputs.

The current-wheel admission control retains sampled values, derivative outputs, exact input identities, and limits.

Figiel parameter roles and calibration rows remain with the existing reproduction target ledger (private research repository): the canonical Data packet owns reusable model inputs, while the reproduction analysis owns paper-specific fitted targets and partitions. Its parameter-role ledger (private research repository) (SHA-256 932e8baa90fcefbaa8c3a8730cdeadd83a4c01f0a3b109f4e4cd0319aee9312b) covers 122 source cells from Tables 2–5, their association/segment-count footnotes, and Eq. 11. It classifies 57 cells as fitted in Figiel, 33 inherited from cited sources, 13 fixed or assumed, and 19 blank or unpublished. Explicit published zero interactions stay fitted records; blank cells remain absent. Table 1 solvent-permittivity values and the fixed ionic-region value are source-provided inputs in the version-3 Data packet, outside that census and not regression targets. The parameter document preserves compatible identities: Table 2 water f_k is component/water/solvation_factor; Table 3 sodium d_i^Born is component/sodium-cation/born_diameter; Table 4 sodium–chloride k_ij is pair/chloride-anion/sodium-cation/k_ij; Table 5 sodium–water k_ij is pair/sodium-cation/water/k_ij; the Eq. 11 coefficient 7.01 is model/relative_permittivity/ion_fraction_suppression_coefficient; and Table 1 epsilon_r,ion = 8 is model/relative_permittivity/ionic_region_relative_permittivity. The source 2B association choice is materialized as explicit A/B sites, and the Table 3 segment sigma remains distinct from the fitted Born diameter; packing and Debye–Huckel diameters use the source transformation d_i^pack = d_i^DH = 0.88 sigma_i.

The 407-row calibration-observation ledger (private research repository) (SHA-256 4d6ce439f732d44c6cee08d9d800cef6f844eb297a30ebf935daa1c65403c5e7) records each row's role, value and unit, species/solvent, temperature, pressure, composition basis/value, thermodynamic basis, source locator and row, extraction or construction, source hash, and uncertainty status. It retains the published order: SI Table S5 water solvation, NaBr MIAC, aqueous MIAC, Eq. 19 transfer construction, dielectric correlation, then SI Table S4 density and osmotic-coefficient data. The current five-ion Born-diameter tracer uses only five SI Table S5 reported averages as targets and their 27 underlying literature values as support: T = 298.15 K, P = 1 bar, x_ion -> 0, with the x-treatment from gas at 1 bar to a hypothetical dilute-ideal solution at 1 bar. The SI reports no pointwise uncertainty.

The ledger classifies 191 rows as direct, 47 table-derived, 90 digitized, 43 constructed, and 36 missing-primary-source. In particular, its 36 NaBr MIAC rows for methanol and ethanol remain explicitly missing-primary-source: each is dimensionless at 298.15 K and 1 bar on a formula-unit molality basis, but the retained lab CSVs have no primary-paper extraction result. Digitized and constructed values remain marked as such; Eq. 19 constructed solvation rows are not source observations. The ledger preserves calibration roles and conditions without asserting a complete replay of the paper's objective, weights, or uncertainty model. It is source-design evidence, not a new fit or an extension of the baseline's numerical claim.

Figiel and iodide input domain

The first electrolyte model is bounded to the Figiel source records for water, methanol, ethanol, Li+, Na+, K+, Cl-, and Br- at T = 298.15 K and 0 <= x_ion <= 0.38. A distinct aqueous bundle admits source-complete NaI and KI at the same conditions without changing established model fingerprints. LiI remains excluded because Figiel Table 4 publishes no Li+/I- interaction. Each admitted MIAC calculation selects one solvent composition, one cation, and one anion. A mixed-solvent selection additionally requires every solvent pair and cross-association record. The ordinary-state composition is electroneutral. The retained Figiel records report p = 1 bar; evaluation at another pressure is a mathematical EOS result, not source-backed physical validation at that pressure. The dielectric suppression relation is not evidence for another solvent, temperature, component topology, or larger ion fraction. Solver convergence does not expand that evidence.

Association source conversions

The MEA nine-species packet from version 5 converts the Fakouri Baygi 2015 (DOI 10.1016/j.cherd.2014.07.017, Eq. 4) MEA 2B volume this way; the conversion analysis (private research repository) owns the evidence.

The Figiel2025, Figiel–Yu–Held 2025, Khudaida2026 and MEA nine-species packets name wolbach-sandler from their current versions onward, with its pure-component self-edge inputs and its provenance for both admitted cross directions. Their earlier versions name arithmetic-energy-geometric-volume and stay unchanged as immutable records.

Permittivity comparisons on the Figiel parameter base

The bounded comparison uses Perdomo2025 Table 3 at T=298.15 K and P=2508 Pa, with explicit-species compositions in EOS order (water,Na+,Cl-):

x_v = (1-2(1.41063e-116), 1.41063e-116, 1.41063e-116),
x_l = (0.826112, 0.086944, 0.086944).

At the corrected retained vapor volume V=0.9841 m3 mol^-1, Figiel gives epsilon_r=78.09, while Schreckenberg and Rueben give approximately 1.00142 and 1.000324. Thus the density topologies repair the dilute-vapor limit at the equation level. The retained SAFT-gamma Mie liquid volume 1.787e-5 m3 mol^-1 is not an ePC-SAFT pressure root: direct ePC-SAFT pressures are about -2.08e8 to -2.16e8 Pa across the four topologies. Comparisons of chemical potentials therefore use each topology's own pressure-certified root, not the retained foreign-EOS liquid volume.

The diagnostic changes only the permittivity records on the explicit Figiel2025 ePC-SAFT/SSM+DS parameter base. Its immutable fingerprints are:

Figiel          sha256:79faf0da75ddb6dfe44d0282f93c5c9497941bee82227d8d76dd1149618838ef
Zuber           sha256:bb00c286a3d36a14485b96ae72577716a3a8813708d7ec5907f1da7cc03bcc1f
Schreckenberg   sha256:07d3034eb210e89c6a2ecb736735c85fafeb6161c03c4febd473428c8883d47a
Rueben          sha256:9d7ae94f2cda937ea541a3d65da3d04c66d62cca0af8fff23896456f3602cbf2

The vapor root is V=0.9862436645 m3 mol^-1 for every topology. The liquid root and electrostatic results are:

topology liquid rho_m / mol m^-3 epsilon_v epsilon_l a_l^DH a_l^Born mean-ion gap / RT water gap / RT
Figiel composition-only 59380.0341 78.0900 35.1922 -0.350489 -24.5552 -269.7235 0.0170
Zuber ion-specific 59364.9061 78.0900 40.9716 -0.292775 -24.6290 -269.5921 0.0089
Schreckenberg density 59466.9217 1.001419 69.7591 -0.154498 -24.8146 -170.8193 -0.0177
Rueben perturbation 59456.7965 1.000323 74.7311 -0.142058 -24.8321 -170.6547 -0.0231

Here the reported phase-equilibrium diagnostic is evaluated from EOS fugacity coefficients,

Delta_mu_ion/(RT) = (1/2) sum_(i=Na+,Cl-)
                     ln[(x_i^v phi_i^v)/(x_i^l phi_i^l)],
Delta_mu_water/(RT) = ln[(x_water^v phi_water^v)/
                          (x_water^l phi_water^l)].

All displayed roots have positive dP/d(rho_m) and pressure closure within 3e-6 Pa. Zuber's explicit ion coefficients barely change the approximately 270 RT mismatch. Schreckenberg and Rueben improve it by about 99 RT, but a remaining 171 RT mean-ion gap is still far from equilibrium. The water gap is already within 0.024 RT, which localizes this endpoint disagreement to ionic solvation rather than solvent phase equality.

As a liquid-property cross-check at 298.15 K, 1 bar, and 1 mol kg^-1 aqueous NaCl, HamerWu1972 Table 16 reports gamma_+/-=0.657. With all other Figiel parameters unchanged, the EOS returns 0.70721 for Figiel, 0.66422 for Zuber, and 0.55220 for Rueben. Their corresponding practical osmotic diagnostics are 0.97101, 0.94640, and 0.87190; no acceptance claim is assigned to those three osmotic numbers here. The Schreckenberg reference path, which formerly failed on a lone stable root, now completes and gives 0.56348. The common-data comparison (engine/tests/cases/dielectric-alternatives (private research repository)) supersedes these single-point diagnostics for water/NaCl at 298.15 K and 1 bar.

Since the retained ePC-SAFT ion, Born, and interaction parameters were not fitted with the density topologies, this is equation and topology evidence, not VLE validation or authority to change a caller-owned default. The next scientific step is a parameter regression study (sensitivity and refit) with liquid dielectric-decrement, density, MIAC, and osmotic anchors before cross-package equilibrium validation. The density-based topology is worth pursuing; changing only the Zuber suppression coefficients is not supported as the next priority by this comparison.

Cation-mixing qualification

The coefficients are qualified only for retained aqueous water/chloride-salt cases at 298.15 K and 1 bar; the Yao calibration covers ionic strength through 8.3 mol/kg. It gives no vapor, dry-salt, mixed-solvent, nonaqueous, temperature-law, caloric or stability qualification. In particular, it has no ideal-gas residual limit at fixed composition. Occupied additional neutrals and organic-phase effects are unqualified extrapolation.

For an illustrative, electroneutral organic-rich composition, take water $x_s=10^{-3}$, Li and Mg $x_i=x_j=10^{-6}$, chloride $x_k=3\times10^{-6}$, and the remainder neutral organic. Substituting $M_s=0.01801528$ kg/mol, $\theta=-0.2205$ kg/mol and $\psi=0.024$ kg²/mol² gives $\Delta a^r\simeq-2.426\times10^{-8}$, about $-6.0\times10^{-5}$ J/mol of phase at 298.15 K. The Li chemical-potential contribution is nevertheless $\Delta\mu_i/(RT)=2\theta m_j+\psi m_jm_k\simeq-0.02426$, about $-60$ J/mol of inserted Li or $-0.0105$ decade in its activity contribution. These rounded formula substitutions are neither an Engine run nor a predicted phase or physical validation. At fixed water molalities the molar energy scales with $x_s$ while ion chemical potentials stay finite. At fixed nonzero ion fractions the pair and triple terms instead grow as $x_s^{-1}$ and $x_s^{-2}$ as $x_s\to0$; there is no universal small-organic-phase bound.

Ascani Case Study 2 input selection

The Ascani Case Study 2 bundle is a separate exact selection of water, 1-butanol, Na+, K+, and Cl- at T = 298.15 K and p = 1 bar. It requires all ten unordered k_ij and l_ij pair classifications explicitly. The same-sign Na+/K+ dispersion pair is source-marked not applicable; its equation-level zero cannot be used as source evidence for another pair. The bundle also carries the source-derived water/1-butanol cross-association energy and volume while preserving the distinct k_ij_hb provenance. Its source records preserve the combined-dielectric and uncorrected-Born provenance, while it and the Figiel records resolve through the same canonical born equation with explicit fixed coefficients. Those source distinctions remain packet provenance and do not create runtime variants or alter packet fingerprints.

Held2008 input selection and reference restriction

The Held2008 bundle is a source-complete selection of water plus all 29 Table 2 ions at T = 298.15 K and p = 1 bar. It requires explicit water--ion k_ij = 0 and ion--ion k_ij = 1, so no ion--ion dispersion is silently introduced. It uses the source's fully dissociated Debye--Huckel formulation with concentration-independent water permittivity and has no Born term, ion-pairing, or reaction-equilibrium behavior. Evaluation or the standard-reference evaluator outside the recorded source state is not source-backed, and the evaluator fails closed outside that state.