Introduction
Open any commercial process simulator — Aspen Plus, ProMax, UniSim, or others — and you face the same first decision: which property package? The dropdown lists dozens — Peng-Robinson, SRK, NRTL, UNIQUAC, GERG-2008, CPA, PC-SAFT, Kent-Eisenberg, Acid Gas, Amine, Glycol, Sour PR, electrolyte NRTL, and on. Pick correctly and the simulation will faithfully represent the physics. Pick incorrectly and every flow rate, every duty, every dewpoint downstream is wrong — confidently and precisely wrong, which is the most dangerous kind.
This post is a service-by-service selection guide. It does not defend any single equation; it tells you when each is the right one. (For a deep dive on Peng-Robinson specifically, see Why Peng-Robinson EOS is a Good Default Choice.)
The Three Families
Every thermodynamic model falls into one of three families:
1. Cubic Equations of State (cubic EOS)
Cubic in molar volume. Solve for two coefficients (typically a for attraction and b for repulsion). Fast, robust, and the workhorse for non-polar and mildly polar fluids. Examples: SRK, PR, PR-Twu, PR with Peneloux volume correction.
2. Activity Coefficient Models
Compute the deviation of liquid phase from ideal mixing through an activity coefficient γ. Used when the liquid phase is non-ideal — polar mixtures, alcohol-water, hydrocarbons-water-glycols, electrolytes. Examples: NRTL, UNIQUAC, Wilson, UNIFAC (predictive variant).
3. Specialty Models
Built for specific physics: dense gas (GERG), associating fluids (CPA), chain molecules (PC-SAFT), electrolyte solutions (electrolyte NRTL).
The key insight: cubic EOS handle most of upstream and midstream oil & gas. The other families exist because cubic EOS fail in specific situations — and recognising those situations is the whole game.
Cubic EOS Family
| Model | Year | Best for | Limitations |
|---|---|---|---|
| van der Waals | 1873 | Historical only | Not used for modern engineering |
| Redlich-Kwong (RK) | 1949 | Simple gas-phase calculations | Inaccurate liquid density |
| Soave-Redlich-Kwong (SRK) | 1972 | General hydrocarbons, pre-PR plants | Liquid density 5–10% low |
| Peng-Robinson (PR) | 1976 | Modern hydrocarbon default | Critical region, polar systems |
| PR-Twu | 1991 | Higher-MW hydrocarbons, asymmetric mixtures | Same intrinsic limits as PR |
| Peneloux volume correction | 1982 | Adds ~3% liquid density accuracy to SRK | Not strictly thermodynamically rigorous |
| Sour PR | — | High H₂S/CO₂ content (simulator-specific variant) | Calibrated for sour service only |
Rule of thumb: SRK and PR are interchangeable for ~80% of gas processing applications. The differences are in the 5th decimal of K-values for most light hydrocarbon systems. Where they diverge:
- PR is more accurate at and near the critical point — important for dense-phase pipelines and supercritical CO₂ injection.
- PR predicts liquid density slightly better out of the box. SRK + Peneloux closes that gap.
- SRK was the original "modern" cubic and remains the European refining default.
For new oil & gas projects today, PR is the default. For brownfield work where the original simulation used SRK, stay with SRK to maintain comparability.
Activity Coefficient Models
When the liquid phase contains polar species at non-trivial concentrations, the cubic EOS approach (which lumps polarity into binary interaction parameters) breaks down. Switch to an activity coefficient model:
| Model | Best for | Limitations |
|---|---|---|
| Wilson | Miscible polar mixtures (alcohols, ketones, water) | Cannot predict LLE; struggles near critical |
| NRTL | Polar mixtures with possible LLE (water-hydrocarbon) | More parameters to fit |
| UNIQUAC | Strongly non-ideal mixtures, asymmetric size molecules | Parameter quality matters |
| UNIFAC | Predictive — no binary interaction params needed | Less accurate than tuned NRTL/UNIQUAC |
The vapour phase is still typically calculated with a cubic EOS (PR or SRK). This is the so-called "γ-φ" approach — activity coefficient for liquid, fugacity coefficient for vapour. In Aspen Plus, this is the NRTL-RK or UNIQUAC-PR combination.
Rule of thumb: if your system has more than ~5 mol% water, methanol, glycol, or amines in the liquid phase, you almost certainly need an activity coefficient model. Cubic EOS will give the wrong answer.
Specialty Models
GERG-2008
The reference equation of state for natural gas, developed by GERG (Groupe Européen de Recherches Gazières) for custody-transfer-grade accuracy. Uses 21 components including all major natural gas species. Inputs and outputs match cubic EOS but the underlying form is multi-parameter (not cubic).
Use GERG-2008 when: high-pressure gas pipelines, dense-phase CO₂ pipelines, custody transfer metering calculations, LNG facilities. Expect ±0.1% density accuracy vs typical 1–2% for cubic EOS.
Don't use GERG when: the system contains components outside its 21-species table (most chemical processes), or when computational speed is critical (it's roughly 5× slower per flash than PR).
CPA — Cubic Plus Association
A standard cubic EOS (typically SRK) with an additional term for hydrogen bonding. Calibrated for water-hydrocarbon-alcohol-glycol systems where association dominates phase behaviour.
Use CPA when: high-pressure water-hydrocarbon equilibria, glycol or methanol injection at deep-water tieback pressures, hydrate inhibition modelling. CPA is the de facto standard for subsea hydrate inhibitor sizing.
PC-SAFT — Statistical Associating Fluid Theory
Models molecules as chains of segments with explicit attractive and association terms. More physically rigorous than cubic EOS for long-chain hydrocarbons, polymers, and asphaltenes.
Use PC-SAFT when: heavy oil characterisation, asphaltene precipitation modelling, polymer process simulation. Increasingly used in upstream for modelling reservoir fluids with very heavy ends.
Don't use PC-SAFT when: a faster cubic EOS works — it is computationally heavy.
Electrolyte Models (ENRTL, Mixed-Solvent Electrolyte)
For aqueous solutions containing ionised species — sour gas treating with amines, sulphuric acid plants, brine systems. The non-ideality from charged species is dominant and conventional models give nonsense answers.
Use electrolyte NRTL when: amine gas sweetening (MDEA, MEA, DEA), sour water stripping, oxide leaching. Most commercial process simulators bundle a dedicated Acid Gas / Chemical Solvents package built on electrolyte NRTL.
Service-Specific Recommendations
This is the practical heart of the post. Look up the service, find the right package.
Upstream Oil & Gas — Production Separation, Compression, Pipelines
| Service | Recommended | Notes |
|---|---|---|
| Production separators (gas-oil-water) | PR EOS | Default; tune binary interactions for high-CO₂ wells |
| Wellstream choke modelling | PR EOS | Accept ±5% on multiphase flow; specialist correlations may give better |
| Gas compressors | PR EOS | Verify against vendor curves at design point |
| Dense-phase CO₂ pipeline | GERG-2008 or PR with tuned BIPs | Cubic EOS errors near critical point can be 5%+ |
| Multiphase flowlines | PR EOS + multiphase correlations | Use the EOS for properties, dedicated flow correlations for hydraulics |
| Oil characterisation (heavy ends) | PR EOS with characterised pseudo-components | Assay characterisation matters more than EOS choice |
Gas Processing
| Service | Recommended | Notes |
|---|---|---|
| TEG dehydration | Glycol property package or NRTL-PR | Cubic EOS alone will not predict water dewpoint |
| Amine sweetening (MDEA, MEA, DEA) | Acid Gas / Chemical Solvents package or electrolyte NRTL | Non-negotiable; cubic EOS gives garbage |
| Mol sieve dehydration | PR EOS for bulk; sieve-specific correlations for adsorption | Adsorption is not phase equilibrium |
| Cryogenic NGL extraction | PR EOS with low-T tuning, or GERG-2008 | At -100°C the EOS choice starts to matter |
| LNG liquefaction | GERG-2008 preferred; PR with tuned BIPs acceptable | Custody-grade calculations need GERG |
Refining
| Service | Recommended | Notes |
|---|---|---|
| Crude distillation (atmospheric / vacuum) | PR or SRK with API characterisation | Pseudo-component characterisation is the dominant accuracy lever |
| Reformer / hydroprocessing | PR or SRK | Reactor kinetics dominate; EOS is secondary |
| Sour water stripper | electrolyte NRTL or Sour Water property package | Cubic EOS will not predict NH₃/H₂S/CO₂ stripping |
| Aromatics extraction | NRTL or UNIQUAC with PR vapour | Liquid-liquid extraction needs activity coefficients |
Petrochemicals & Specialty
| Service | Recommended | Notes |
|---|---|---|
| Olefins cracking | PR EOS | Specialty packages (e.g., Aspen Olefins) wrap PR |
| Polymer reactors | PC-SAFT or POLYNRTL | Conventional EOS cannot handle polymers |
| Sulphuric acid | electrolyte NRTL with sulphate equilibria | Specialised; use vendor-recommended package |
| Distillation of azeotropes | NRTL or UNIQUAC | Activity coefficient required |
| Solid-liquid systems | eNRTL or specialty | Most simulators struggle |
Property Package Validation Workflow
Whatever package you select, validate it before trusting the results:
- Identify three or four critical operating points in your flowsheet — typically the toughest separation, the highest pressure, the lowest temperature, the most polar mixture.
- Look up experimental data at those conditions from published sources, vendor data, or pilot-scale measurements.
- Run the package at those conditions — single flash, no flowsheet — and compare the predicted properties (density, K-values, dewpoint, enthalpy) against the data.
- Tune binary interaction parameters if the discrepancy is material. Most simulators allow this in the property package definition.
- Document the validation — which model, which conditions, which data, and what the residual error is.
A simulation without a validated property package is a calibration exercise pretending to be engineering.
Common Pitfalls
- Treating PR as universal. It is the right starting point for hydrocarbons but materially wrong for amines, glycols, polymers, and electrolytes. The PR-EOS-only project will eventually produce a result that surprises everyone.
- Mixing models inconsistently across the flowsheet. If the absorber uses Acid Gas and the regenerator uses PR, the recycle convergence will produce nonsense. Use one model everywhere — or use simulator-managed model switches at unit-block boundaries with care.
- Skipping characterisation for crude / heavy oil systems. The choice of EOS matters less than how you characterised the C7+ pseudo-components. Get the characterisation wrong and PR vs SRK is irrelevant.
- Ignoring binary interaction parameters (BIPs). For systems with significant CO₂, H₂S, or N₂, default BIPs from the simulator may be ten years stale. Check vendor recommendations or recent literature.
- Using GERG-2008 in flowsheets it cannot handle. GERG is restricted to 21 components. The moment you add a heptane+ pseudo or trace mercaptan, GERG either falls back to defaults or refuses to converge.
- Not validating against experimental data. "I used PR" is not validation. PR is wrong by varying amounts depending on system, and the only way to know is to compare against measured behaviour.
Conclusion
The right property package is service-specific. PR for hydrocarbons, NRTL/UNIQUAC for polar liquids, GERG for custody-grade gas, electrolyte NRTL for amines and sour water, CPA for subsea hydrate inhibition. There is no universal package — and the first job of any simulation engineer is to recognise the boundary between "PR works fine" and "PR will give wrong answers".
The discipline is in the validation. A simulation built on an unvalidated property package is a confident-looking source of bad numbers. A simulation built on a validated, service-appropriate model is engineering. The difference shows up six months later when the plant either meets spec or doesn't.
