NeqSim uses Henry-law reference states for neutral solutes in aqueous excess-Gibbs models. This page records the implemented convention, source provenance, qualification status, and the evidence boundary for brines and multiphase calculations.

Engineering question and stop boundary

This increment asks whether common gases can use a reproducible, attributed pure-water Henry reference instead of invalid or undocumented database placeholders, without changing electrolyte interactions, reaction constants, EOS mixing rules, flash topology, or non-aqueous model paths.

It implements the IAPWS G7-04 H2O correlation for He, Ne, Ar, Kr, Xe, H2, N2, O2, CO, CO2, H2S, CH4, C2H6, and SF6. It does not add a pressure/Poynting correction, a vapor-liquid distribution-constant correlation, brine salting-out parameters, reactive speciation, or new electrolyte-EOS parameters.

Convention and equations

IAPWS defines the mole-fraction Henry constant on the liquid-water saturation boundary as

\[k_H(T)=\lim_{x_i\rightarrow0}\frac{f_i}{x_i}.\]

With $T_r=T/T_c$, $\tau=1-T_r$, and water saturation pressure $p^*$,

\[\ln\left(\frac{k_H}{p^*}\right) =\frac{A}{T_r}+\frac{B\tau^{0.355}}{T_r} +C T_r^{-0.41}\exp(\tau).\]

The implementation returns $k_H$ in bar and provides the analytical $\mathrm{d}\ln k_H/\mathrm{d}T$ required by fugacity derivatives. It reports whether a requested state is inside the species-specific fitted range, a guideline extrapolation, unsupported, or outside the liquid-water correlation domain. A supported species outside the domain fails closed.

Pitzer neutral activities use a molality standard state. At infinite dilution,

\[H_m=k_H M_{\mathrm{water}},\]

where $M_{\mathrm{water}}=0.01801528$ kg/mol. The existing $m_i/x_i$ conversion then maps the Pitzer expression back to the common mole-fraction/fugacity kernel. The constant conversion leaves $\mathrm{d}\ln H/\mathrm{d}T$ unchanged.

The coherent family is identified as iapws-g7-04-water-gases-v1. Ordinary runtime evaluation accepts only the published species-specific fitted range. getHenryCoefficientBarAllowExtrapolation is deliberately named for reproducing guideline check values outside that range; GE and Pitzer fugacity paths never call it.

Analytical derivative contract

For the implemented GE reference states,

$\phi_i=\frac{\gamma_i H_i}{P}$

for a Henry solute, or the corresponding solvent vapor-pressure expression. NeqSim therefore exposes the constant-composition analytical derivatives

$\left(\frac{\partial\ln\phi_i}{\partial T}\right)_{P,\mathbf{x}}=\frac{\partial\ln\gamma_i}{\partial T}+\frac{\partial\ln H_i}{\partial T}$

and

$\left(\frac{\partial\ln\phi_i}{\partial P}\right)_{T,\mathbf{x}}=\frac{\partial\ln\gamma_i}{\partial P}-\frac{1}{P}$

SystemInterface.init(2) and init(3) publish these values through getdfugdt(), getdfugdp(), and the Java/JPype getProperty("logfugdT"|"logfugdP", component, phase) adapter. Pressure is in bar, so the pressure derivative is in 1/bar. The -1/P term is the derivative of the explicit fugacity-coefficient denominator; it is not a Poynting correction. The IAPWS reference remains defined at water saturation, and no pressure-dependent partial-molar-volume model is implied.

Versioned coefficient family

The following values are transcribed from IAPWS G7-04 Tables 2 and 5. A, B, and C are dimensionless; temperature limits are in kelvin; RMS is the published root-mean-square deviation in ln(kH).

Species A B C Tmin Tmax RMS
He -3.52839 7.12983 4.47770 273.21 553.18 0.0341
Ne -3.18301 5.31448 5.43774 273.20 543.36 0.0577
Ar -8.40954 4.29587 10.52779 273.19 568.36 0.0443
Kr -8.97358 3.61508 11.29963 273.19 525.56 0.0434
Xe -14.21635 4.00041 15.60999 273.22 574.85 0.0363
H2 -4.73284 6.08954 6.06066 273.15 636.09 0.0517
N2 -9.67578 4.72162 11.70585 278.12 636.46 0.0372
O2 -9.44833 4.43822 11.42005 274.15 616.52 0.0377
CO -10.52862 5.13259 12.01421 278.15 588.67 0.0039
CO2 -8.55445 4.01195 9.52345 274.19 642.66 0.0528
H2S -4.51499 5.23538 4.42126 273.15 533.09 0.0408
CH4 -10.44708 4.66491 12.12986 275.46 633.11 0.0386
C2H6 -19.67563 4.51222 20.62567 275.44 473.46 0.0259
SF6 -16.56118 2.15289 20.35440 283.14 505.55 0.0505

Model semantics

Runtime path IAPWS reference decision Electrolyte-interaction decision
Generic aqueous GE use the mole-fraction reference for a supported neutral gas in water, including H2/He/Ar rows historically marked as solvents retain the GE model’s own activity coefficient
Pitzer, no ionic component topology use the explicit molality conversion above no salting-out claim; this is the pure-water limit
Pitzer, ions and a coverage-checked neutral family use the molality reference apply the explicitly selected lambda, zeta, mu, and eta family
Pitzer, ions without a qualified neutral family preserve the established database/capping reference for the complete calculation do not silently combine IAPWS with a unit neutral activity coefficient
SRK, PR, CPA, and electrolyte EOS do not dispatch through this GE reference retain EOS fugacity, association, and mixing-rule semantics

The Pitzer decision depends on component topology, not the instantaneous ionic strength. Consequently a reaction solve cannot change reference models merely because trace ions appear or disappear. Pitzer $\lambda$, $\zeta$, $\mu$, and $\eta$ interactions remain distinct parameter families. CO2 and H2S also require a qualified neutral family before Pitzer changes their reference, even before reaction species have been added. This prevents initialization order from seeding a reactive calculation with a temporarily different standard state.

Current-master gap reproduced before adoption

The audited component table contained a methane overflow sentinel, zero coefficient rows for nitrogen and oxygen, and incomplete or undocumented rows for several other gases. Those rows could not serve as qualified common-gas water correlations. The IAPWS dataset is selected by explicit species and water topology; unrelated database correlations remain available for unsupported species.

Public evidence and source comparison

Source Species/system Convention and range Licensing/reuse Decision
Fernandez-Prini et al. (2003), DOI 10.1063/1.1564818 14 gases in H2O and 7 gases in D2O (k_H=f/x) at water saturation; triple-point region toward critical conditions; species-specific fitted ranges and RMS residuals primary JPCRD article available through NIST adopted lineage and independent equation cross-check for the 14 H2O gases
IAPWS G7-04, Henry guideline He, Ne, Ar, Kr, Xe, H2, N2, O2, CO, CO2, H2S, CH4, C2H6, SF6 in H2O same equation and standard state; Table 2 fitted ranges; Table 6 check values at 300/400/500/600 K guideline states that publication in whole or in part is permitted with attribution adopted coefficients, ranges, equations, and 56 published check values
Sander (2023), DOI 10.5194/acp-23-10901-2023 water-solvent compilation standardized alternative Henry definitions and conversions article and supplement are CC BY 4.0 terminology and conversion cross-check; no values adopted
NIST Chemistry WebBook SRD 69, CO2, CH4, and N2 Henry data three gases in water near ambient temperature solubility-form values and temperature parameters; conventions require conversion before comparison compilation is copyright protected read-only independent screening index; no values copied
USGS PHREEQC 3.9.0, pitzer.dat provenance and license ionic and neutral Pitzer interaction families molality scale, PHREEQC six-term temperature functions, explicit tuple topology USGS release is public domain; exact release/commit/blob are recorded in the linked matrix confirms that a pure-water Henry constant is not a substitute for missing neutral-ion terms; no Pitzer value adopted here
Kaasa (1998), stable National Library item oil-recovery-brine Pitzer binary, same-sign, ternary, and neutral families Appendix F is a provenance index with its own coefficient order and row-level source lineage scan-table redistribution terms are unresolved; original pages were not retrievable from the public item in this run metadata/provenance index only; no value inferred, OCRed, copied, or adopted

The IAPWS high-temperature root-mean-square deviations in $\ln k_H$ range from 0.0039 for CO to 0.0577 for Ne. These source residuals describe the correlation fit; they are not NeqSim regression tolerances or independent held-out validation.

Baard Kaasa’s 1998 thesis remains a provenance index for Pitzer specific-ion-interaction parameters. No Kaasa table was copied and no Pitzer coefficient was adopted here: Henry constants and Pitzer interaction coefficients are different model families.

Regression evidence

IapwsHenryLawTest reproduces all 56 rounded IAPWS Table 6 values, compares the analytical derivative with centered finite differences for all 14 gases, checks fitted/extrapolated/unsupported/out-of-domain diagnostics, verifies mole-fraction-to-molality mapping, and exercises fail-closed behavior.

ComponentGEHenryDerivativeTest retains the legacy database-correlation regressions, including the dimensional identity, corrects the historic H2/He/Ar solvent classification in aqueous GE phases, and verifies that an ionic Pitzer topology without a qualified neutral family remains on one deterministic compatibility path even at trace ionic strength. It also checks the system-level init(2) derivative dispatch against an IAPWS analytical temperature derivative and a centered pressure finite difference, including changed-state, return-state, clone, and Java/JPype property access. The system-level Pitzer regression verifies the explicit molality conversion, including the fixed IAPWS water molar mass. The dimensional identity is

\[\frac{\mathrm{d}\ln H}{\mathrm{d}T}=\frac{1}{H}\frac{\mathrm{d}H}{\mathrm{d}T}.\]

These tests are implementation and source-table regressions. They are separate from independent VLE/VLLE or brine validation. In particular, six held-out CO2-Na2SO4 bubble pressures from Bermejo et al. (2005), DOI 10.1016/j.fluid.2005.10.006, are underpredicted by 32.6–43.8% even though the phase-boundary calculations close. SystemPitzer.requirePitzerDatasetValidationFor(GAS_AQUEOUS_VLE) therefore rejects that parameter subset; aqueous-activity and water-property qualification are unaffected. The Pitzer provenance page records the NIST ThermoML checksum, preprocessing, uncertainty, and composition convention.

Scientific and performance controls

The repository’s complete reactive-H2S benchmark remains on the established legacy Pitzer reference because its diagnostic reports missing H2S|H2S, H2S|H3O+, H2S|HS-, and H2S|H3O+|HS- neutral interactions. At 298.15 K, the repaired branch gives maximum absolute reaction ln(Q/K) residual 4.2633e-14, maximum elemental residual 1.1927e-13, charge -1.0825e-14 mol, and normalized charge residual 1.5922e-10. HS- molality increases from 3.3995324e-5 mol/kg at 298.15 K to 4.3530791e-5 mol/kg at 318.15 K. These are compatibility, closure, and nearby-state controls; they are not treated as new independent H2S-solubility validation.

PitzerCatalogPerformanceBenchmark uses nine fixed-work batches after warmup. On OpenJDK 17 in the same runner, the affected combined IAPWS value/derivative kernel decreased from 1140 ns on exact pre-repair PR head a3acb09cf to 573 ns on the repaired tree. The complete reactive-H2S median was 40.76 ms versus 50.25 ms on that head, with the compatible chemical state above. Absolute wall-clock timings remain environment-sensitive; the comparison is a regression screen, not a portable throughput guarantee. Neutral SRK/PR/CPA paths do not dispatch through this helper and remain covered by their existing tests.

Remaining validation boundary

Publication-quality aqueous/electrolyte predictions still require:

  1. independently sourced neutral-ion interactions for the selected Pitzer dataset, including complete binary and ternary coverage;
  2. held-out pure-water and brine solubility over temperature, pressure, ionic strength, and salt composition;
  3. reactive CO2/H2S material balance, electroneutrality, and $\max|\ln(Q/K)|$;
  4. VLE/VLLE fugacity closure, stable topology, normalized non-negative phases, and nearby-state trends;
  5. deterministic repeated and changed-state execution, stale-state protection, clone/serialization/thread checks, and Java/JPype composability;
  6. complete-calculation and kernel benchmarks, including neutral EOS controls.

The next parameter increment is therefore a complete, redistributable neutral-gas/ion interaction family with held-out brine validation. Missing interactions must remain explicit rather than silently defaulting to zero.