What a property package does

Every unit operation calculation rests on the same handful of questions: at this temperature, pressure and composition, what phases exist, what is in each, and what are their density, enthalpy, entropy and transport properties? A property method is the set of models that answers them. Choosing it is the first and most consequential decision in a simulation, because a wrong method gives wrong answers with perfect convergence.

Reference equations of state

For a well-measured pure fluid the best model is a multiparameter Helmholtz equation: a fitted expression for the Helmholtz energy a(T, ρ) from which every other property follows by differentiation. The IAPWS-95 water formulation, the Span–Wagner CO₂ equation and the refrigerant equations behind REFPROP are all of this kind. They reproduce density to about 0.1% and carry dedicated viscosity and thermal-conductivity correlations.

Their limit is coverage. Each fluid needs its own equation fitted to a large data set, so only about 120 fluids have one. Mixtures of them exist only for a few families (natural gas, refrigerant blends).

Cubic equations of state

A cubic EOS needs only the critical temperature, critical pressure and acentric factor of each component, which exist for tens of thousands of substances. Peng-Robinson:

Mixtures use van der Waals mixing rules, a = ΣΣ xᵢxⱼ √(aᵢaⱼ)(1 − kᵢⱼ), where the binary interaction parameter kᵢⱼ is fitted to measured phase-equilibrium data. Both phases come from the same equation (the φ–φ approach), so the method is consistent through the critical region and at high pressure. It is the standard choice for hydrocarbons, natural gas, air and refinery streams.

The known weakness is liquid density, typically 5–10% low without volume translation, and polar liquids (water, alcohols, acids), where a single kᵢⱼ cannot describe hydrogen bonding.

Activity-coefficient models

At low pressure the vapour is close to ideal and all the non-ideality sits in the liquid. The γ–φapproach writes equilibrium as modified Raoult's law:

The activity coefficient γᵢ comes from an excess-Gibbs-energy model. NRTL uses two energy parameters and a non-randomness factor per binary, regressed from data; it handles strongly non-ideal systems and liquid–liquid splits. UNIFAC builds γᵢ from the molecule's functional groups (CH₂, OH, COOH …) with group–group parameters, so it predicts pairs nobody has measured, at some cost in accuracy. Azeotropes appear naturally from γᵢ > 1 (positive deviation, ethanol–water) or γᵢ < 1 (negative deviation, acetone–chloroform).

These models need a vapour pressure for every component, so they cannot describe a component above its critical temperature (dissolved nitrogen, CO₂). That is the job of a cubic EOS or a Henry's-law term.

K-values, bubble and dew points

The equilibrium ratio Kᵢ = yᵢ / xᵢ summarises how a component partitions. A mixture at (T, P) is a liquid if Σ zᵢKᵢ ≤ 1, a vapour if Σ zᵢ/Kᵢ ≤ 1, and two-phase in between, where the vapour fraction V/F solves the Rachford–Rice equation Σ zᵢ(Kᵢ − 1) / (1 + (V/F)(Kᵢ − 1)) = 0.

The bubble point is the state where the first vapour forms (Σ zᵢKᵢ = 1) and the dew point where the first liquid condenses (Σ zᵢ/Kᵢ = 1). Sweeping composition at fixed pressure traces the T-x-y diagram; at fixed temperature, the P-x-ydiagram. Where the bubble and dew curves touch away from the pure ends is an azeotrope.

Choosing a method

Reference states

Enthalpy and entropy have no absolute zero; every table picks one. Steam tables set liquid water at the triple point to zero, refrigerant tables (IIR) set saturated liquid at 0 °C to 200 kJ/kg, cubic-EOS results here take the ideal gas at 25 °C. Differences between two states in the same method are what carry physical meaning. Never subtract an enthalpy from one method from an enthalpy from another.

What this calculator does and does not replace

The equations are the same ones a commercial simulator solves. The difference is the parameter bank: a licensed package ships thousands of regressed kᵢⱼ and NRTL pairs, electrolyte models and polymer models. This tool carries the open ChemSep set and says so whenever a pair is missing. For a feasibility study, a sanity check or teaching, that is enough. For a guarantee on a column, regress the binaries against your own data first.