Version v1.2 · 2026-10-14
Pátek & Klomfar (1995) published five explicit correlations regressed from their own Gibbs free-energy function: bubble-point temperature (Eq. 6), dew-point temperature (Eq. 7), equilibrium vapour composition y(p, x) (Eq. 8), liquid enthalpy (Eq. 9) and vapour enthalpy (Eq. 10), with x, y ammonia mole fractions, p in MPa, T in K. Validity: Eq. (6) 0.002–2 MPa over the full composition range; Eq. (7) 0.02–2 MPa; Eq. (8) x > 0.05, p ≤ 2 MPa. The authors report 95% of bubble points within 1.5 K, 95% of dew points within ±2 K, and RMS deviations of 0.9 kJ/kg (liquid enthalpy) and 4.5 kJ/kg (vapour enthalpy). The reference state is liquid water at its triple point (273.16 K). Recomputation in this survey gives, at 0.101325 MPa (1 atm), boiling points of −33.42 °C (ammonia) and 99.75 °C (water); for x > 0.2 the three temperature/composition equations are mutually consistent to 0.88% at most. At 1 MPa and z = 0.5 the set gives a bubble point of 65.46 °C, a dew point of 153.21 °C, and hence a true glide of 87.75 K. Two boundary behaviours are on record: at 2 MPa the dew-point equation returns 210.36 °C at the water end, 2.04 K below the true value; and in the corner x < 0.1, p > 1 MPa the mutual deviation reaches about 12%.
Conde (2004), an engineering booklet, supplies the transport properties by a corresponding-states construction with per-property mixing corrections; its §4 reproduces the Pátek–Klomfar equations (6)–(10) verbatim — a coefficient-by-coefficient comparison in this survey found the five tables identical. For pure ammonia it reports a mean deviation of 0.0252% (maximum 1.15%) for specific heat and 0.527% (maximum 6.630%) for thermal conductivity against its own data sources.
Schulz (1971) constructed a Gibbs free-energy equation of state for the system (0.001–2.5 MPa, 200–450 K per the transcribing notes), from which phase equilibrium, enthalpy, entropy and specific volume follow by differentiation. The original text is not in hand; the coefficient tables survive in three complementary thesis transcriptions, but the equation faces lost symbols in transcription and are incomplete [to be verified]. The set is registered, not yet programmable.
A pure-ammonia endpoint comparison illustrates why provenance matters even at the boundary of a mixture model. At the ammonia endpoint, the Pátek–Klomfar mixture fit gives latent heats deviating from CoolProp by −21.1% (−30 °C), −14.7% (−20 °C), −4.2% (10 °C) and −1.2% (35 °C), whereas the polynomial set of the Ferreira lineage (Section 3.1.3), a fit to pure ammonia itself, deviates by −0.57%, −0.39%, +0.43% and −0.51% at the same four temperatures. The two routes — endpoint extrapolation of a mixture fit versus a pure-fluid fit — are kept in separate accounts in the library.
Castilla (2014), Appendix B of a doctoral thesis, is a compilation of published fits to the measurements of Libotean et al.: bubble pressure ln P = pa(x) + pb(x)/T (2007 data), liquid enthalpy after Vallès & Salavera (2011), specific heat, density and viscosity after Libotean et al. (2008), thermal conductivity after Cuenca et al. (2013), and a ten-segment crystallisation polynomial after Infante Ferreira et al. (1984); x is the ammonia mass fraction. Validity: bubble pressure, specific heat and density 293.15–353.15 K, x 0.35–0.65; enthalpy 273.15–353.15 K; conductivity 303.15–353.15 K, x 0.35–0.60. The reference state is H = 0 at 0 °C and x = 0.5 (the library's check gives hsol(273.15 K, 0.5) = 0.0024 kJ/kg). This is the set built into the library: C and Python implementations agree on 1715 check points to 4.3×10⁻¹⁰ at most, with spot values pbubble(313.15 K, 0.5) = 0.50471 MPa, ρ(313.15 K, 0.45) = 1067.9 kg/m³, cp(313.15 K, 0.45) = 2.9288 kJ/(kg·K). The appendix's viscosity equation is typographically damaged; a reconstruction in the Python module is marked experimental and unverified, and is excluded from the C/VBA builds.
Cai Dehua (2015) fitted the same Libotean data (293.15–353.15 K, 0.2 < w < 0.6) as ln p = A + B/T + C·ln T + D·Tᴱ, with A–E cubic in w, in two variants — the double-effect variant sets the temperature-nonlinear terms to zero to restrain extrapolation toward 200 °C and 9.0 MPa. Reported maximum relative errors: vapour pressure 2.488% (single-effect) and 0.875% (double-effect); specific heat (linear in T and w) 0.875%. The enthalpy is assembled as h(T, w) = h₀°C(w) + ∫cp dT, the 0 °C segment taken from Infante Ferreira's piecewise correlation; the reference state sets the enthalpy of solid salt and of saturated liquid ammonia at 0 °C both to zero. The set is transcribed in reading notes, not yet built; several equation faces are OCR-damaged.
Liang Xiao (2022) re-fitted the Libotean data to a single algebraic template, Y = A + B·T + C·T² + D·T³ with A–D cubic in the ammonia mass fraction, for seven properties. The thesis reports, for NH₃–LiNO₃: VLE RMSE 0.00653 MPa (relative 2.45%, R² = 0.9999), specific heat relative deviation 0.09%, density 0.06%, viscosity 2.1%, thermal conductivity 0.44%. Cai's and Liang's chapters originate from the same research group and are treated as related, not independent, until a coefficient-by-coefficient comparison is carried out [to be verified]. The upstream source of the whole family, Infante Ferreira (1984) — Solar Energy 32(2): 231–236, with Delft internal reports K 91/K 92 — is archived in the survey library but has not yet been opened for the arbitration it is needed for (Section 6) [to be verified].
Four sets are on record; all four descend from the same experimental literature (Chaudhari et al.; Blytas) assembled by Ferreira, so they are successive correlations of one data lineage, not four independent experiments.
The Ferreira lineage as transcribed by Yin Jiayuan (2015) (via Lu & Hu 1995 and Sun 1998) gives a bubble-point equation ln P = A + B/T, with A = 15.7266 − 0.298628·X and B = −2548.65 − 2621.92·(1 − X)³ (X ammonia mass fraction; P in kPa in the source formula), and an enthalpy cubic in t. The transcription is verified at the thesis's own design points: (X = 0.4562, T = 353.15 K) returns 1.3123 MPa against the printed value (1312 kPa), and (X = 0.5149, T = 311.15 K) returns 0.6139 MPa against the printed value (614 kPa). The enthalpy's leading polynomial A(X) is Ferreira's 0 °C integral heat-of-solution fit: its intercept 79.72 matches the original 79.7237, and the endpoint test gives A(1) = −0.05 ≈ 0. One coefficient remains open — the sign of the C polynomial's first term; the two readings differ by 1.65–1.87 kJ/kg at 38 °C and 7.33–8.27 kJ/kg at 80 °C, so the enthalpy equation serves for cross-checking only until the 1984 original arbitrates [to be verified].
A second, independent transcription chain of the same Ferreira set (an open-access paper, SCIRP paperid 99018) carries a misprint in the same A polynomial; the case is documented as TE-001 in Section 5, and that chain's enthalpy equation is excluded from use.
Cai Dehua (2015) fitted the vapour pressure to Chaudhari's data (253.1 K < T < 393.1 K, 0.35 < w < 0.9) in the same ln p form as §3.1.2, reporting maximum relative errors of 3.858% (single-effect) and 1.978% (double-effect), and specific heat (fitted to Sargent's data) at 2.5%. For the enthalpy, the 0 °C integral heat of solution is fitted to Blytas's data with the double-reciprocal model of Wang Xi et al. (a = 58.7178373, b = −0.076577614, c = −0.016925925; maximum relative error 1.47%) — a form chosen, the thesis states, because it satisfies h → 0 at both w → 1 and w → 0, where Ferreira's polynomial gives h = 79.7237 ≠ 0 at w → 0. A same-point comparison with the Yin chain is on record: enthalpy rises of 102.2 (Yin) versus 94.2 kJ/kg (Cai) at 38 °C, X = 0.4562, and 231.0 versus 244.9 kJ/kg at 80 °C, X = 0.5149, with specific heats differing by about 23%; the difference is registered at the source-data level [to be verified].
Liang Xiao (2022) re-fitted the Chaudhari data to the unified template of §3.1.2 (the upstream printed coefficients were discarded as carrying large errors): VLE RMSE 0.02650 MPa (relative 3.53%, R² = 0.9997), specific heat 0.46%, density 0.11%, viscosity 4.9%. The enthalpy datum i₀ is again an Infante Ferreira form, its unit and reference state unstated [to be verified].
A derivative system is registered separately: Sun Qiang (2020) measured 63 VLE points (293.15–373.15 K) for NH₃ over [Na(TX-7)]SCN, a NaSCN–surfactant complex absorbent, with an NRTL correlation (mean relative deviation 2.70%) and a Wilson excess-enthalpy chain; one parameter value (NRTL α = 2.65, outside the conventional 0.2–0.5 span) is flagged [to be verified]. Its parameters are not interchangeable with those of NH₃–NaSCN proper.
Ionic liquids. The census grades three NH₃–ionic-liquid pairs A (VLE and enthalpy both programmable): NH₃/[emim][SCN] and NH₃/[emim][EtSO₄] (VLE from Wang 2017; enthalpy via the extended-NRTL treatment of Karakostas 2025), and NH₃/[emim]Cu₂Cl₅ (Chen & Bai 2016), correlated with a modified UNIFAC (Dortmund) model plus an enthalpy construction h = u₁h₁ + (1 − u₁)h₂ + hᴱ; its composition window is narrow, xNH₃ = 0.89–0.96, and its interaction function Ψ uses a non-standard ordering, exp[−(aT² + bT + c)/T], not the standard Dortmund form. NH₃/[bmim]Zn₂Cl₅ (Chen et al. 2013, UNIFAC, 12 group-interaction parameters) is grade B — VLE only. Extraction notes for four further ionic-liquid pairs are in progress in the library's research notes; no model code exists for them yet.
References are numbered in Section 7 of the overview page.