Short Answer
If you model a compressor, check the vapor outlet of a flash drum, or estimate how dense a gas stream becomes at pressure, you will eventually face a property-model choice between two closely related equations of state: Peng–Robinson (PR) and Soave–Redlich–Kwong (SRK). For pure-component gas predictions — compressibility factors, gas densities, vapor fugacities, and vapor pressures — the honest short answer is that the two usually perform almost identically. They share the same underlying architecture, the same required input data, and the same anchoring to critical-point behavior. The differences that do matter show up mainly in saturated-liquid density (where Peng–Robinson has a long-standing advantage), in derived properties such as heat capacities, and in how faithful each version is near the critical region. This article explains what separates the two models, what large-scale testing has found, and how to choose between them for gas-phase work.
What a Cubic Equation of State Does
An equation of state is a relationship between pressure, molar volume, and temperature that lets engineers calculate density, fugacity, and derived thermodynamic properties without new measurements at every condition. The cubic family gets its name from the resulting algebra: when written in terms of the compressibility factor Z (defined as Z = Pv/RT), the equation becomes a cubic polynomial. At a given temperature and pressure it can therefore produce up to three roots. The smallest root corresponds to a liquid-like state, the largest to a vapor-like state, and the middle root is mathematically unstable. This built-in ability to represent both phases is what makes cubic equations of state useful for vapor–liquid equilibrium (VLE), a capability the ideal gas law lacks entirely.
PR and SRK both descend from the van der Waals equation, which improved the ideal gas law by recognizing that molecules occupy a finite volume (the covolume parameter b) and attract one another (via the attraction parameter a). In both SRK and PR, a and b are not fitted per component; they are fixed by forcing the equation to satisfy the critical-point conditions of the pure substance. The only pure-component inputs required are the critical temperature, the critical pressure, and the acentric factor ω — a dimensionless measure of how non-spherical a molecule is, obtained from its reduced vapor-pressure curve. The temperature dependence of the attraction parameter is handled by an α function, typically anchored so that α equals 1 at the critical temperature.
Shared Roots: Redlich–Kwong to Soave to Peng–Robinson
The Redlich–Kwong equation (1949) improved on van der Waals by making the attraction term strongly temperature dependent. Soave’s insight in 1972 was to parameterize that temperature dependence using the acentric factor and fit it to hydrocarbon vapor pressures. The resulting Soave–Redlich–Kwong equation markedly improved vapor–liquid equilibrium predictions and was rapidly adopted in the refinery and gas-processing industries. Peng and Robinson reworked both the parameter constants and the shape of the attraction term in 1976, addressing a well-known weakness of RK-type models: saturated-liquid molar volumes. SRK’s critical compressibility factor of 1/3 (≈ 0.333) is well above the values typical of real fluids (roughly 0.2–0.3), one of the root causes of its over-expanded predicted liquids, while PR’s value of 0.307 sits closer to those of common hydrocarbons.
Structural Differences at a Glance
Because the two models share a template, their differences can be tabulated compactly:
| Feature | Soave–Redlich–Kwong (SRK) | Peng–Robinson (PR) |
|---|---|---|
| Introduced | 1972 (Soave’s revision of Redlich–Kwong) | 1976 (Peng and Robinson) |
| Attraction term form | a α / [v(v + b)] | a α / (v² + 2bv − b²) |
| Covolume parameter Ωb (b = ΩbRTc/Pc) | 0.08664 | 0.07780 |
| Attraction parameter Ωa | 0.42747 | 0.45724 |
| Critical compressibility factor Zc | 1/3 ≈ 0.333 | 0.307 |
| α-function slope parameter | m = 0.480 + 1.574ω − 0.176ω² | κ = 0.37464 + 1.54226ω − 0.26992ω² (with adjustments recommended for very heavy compounds) |
| Typical edge | Competitive on some vapor-pressure fits; long-standing gas-industry standard | Better saturated-liquid molar volumes (densities) |
Everything that follows flows from these modest-looking changes: a rearranged attraction term, slightly different parameter ratios, and different α-function correlations.
Where Predictions Differ for Pure Components
Vapor pressure and the saturation curve
Both equations were calibrated to reproduce vapor pressures — Soave’s α function was fitted to them directly, and Peng and Robinson anchored their correlation the same way — so the pure-component saturation curve is where the two agree best. A comparison by Twu, Coon, and Bluck in 1998 noted that PR and SRK are probably the most widely used cubic equations of state for vapor–liquid equilibrium prediction in the refinery and gas-processing industries, a status earned largely through decades of acceptable saturation-curve behavior. Notably, the choice of α function matters as much as the choice of underlying EOS: a 2024 study in ACS Omega examined four combinations of Peng–Robinson-type and Redlich–Kwong-type EOS with Soave and Twu α functions precisely because the combination — not just the EOS — governs accuracy and thermodynamic consistency for pure components.
Gas-phase density and compressibility
For vapor compressibility factors and gas densities away from the critical region, the two equations give similar answers. Both reduce to ideal-gas behavior at low pressure, and both handle moderately dense vapors competently because their repulsive (covolume) behavior is similar. Above the critical temperature a pure component has a single fluid root, so both models produce smooth supercritical density estimates, but accuracy degrades as conditions approach the critical region for either model. If near-critical gas accuracy is important, results should be checked against experimental data or high-accuracy reference correlations rather than trusted from any cubic EOS.
Saturated-liquid molar volume
This is the classic dividing line between the two models. SRK systematically over-predicts liquid molar volumes — that is, it underestimates liquid densities — while PR was built partly to correct exactly this flaw, with its lower critical Zc producing denser predicted liquids. Neither is exact, since both models’ critical Zc values remain higher than real-fluid values, but the gap is smaller for PR. The standard remedy for either model is volume translation, a correction introduced in the early 1980s that shifts predicted molar volumes without changing fugacities or phase-equilibrium results; modern EOS variants incorporate translation by design. Comparative property-method studies aimed at helping process engineers choose a model — which typically evaluate molar-volume predictions of SRK, PR, PRSV (a PR modification by Stryjek and Vera), and other methods against experiment — have long used liquid molar volume as one of the headline differentiators.
Derived properties: heat capacities and related quantities
Properties such as isobaric heat capacity, speed of sound, and the Joule–Thomson coefficient depend on temperature derivatives of the EOS, particularly of the α function. Classical Soave-type α functions do not satisfy thermodynamic-consistency constraints at the critical point: their compound-specific fitted slopes break the continuity required of the attraction parameter as it crosses the critical temperature. Twu-style α functions improve pure-component fits but introduce a similar consistency concern at the critical point. A practical consequence is that classical SRK and PR tend to describe equilibrium-type properties (vapor pressures, enthalpies of vaporization) far better than derivative properties, and their predictions of saturated-liquid heat capacity are notably weak. This tension between accuracy and consistency has motivated much of the recent benchmarking work.
What Large-Scale Benchmarking Shows
To replace scattered, compound-by-compound anecdotes with a statistically fair picture, Piña-Martinez, Privat, and Jaubert built a benchmark of more than 300,000 pseudo-experimental data points — benchmark values generated from high-accuracy reference correlations — covering vapor pressures, saturated-liquid densities, enthalpies of vaporization, and liquid heat capacities over about 1,800 pure fluids (AIChE Journal, 2021). They used it to assess four models: the classical SRK and PR EOS and their modernized translated and consistent counterparts, tc-RK and tc-PR.
The broad lessons match the structural analysis above. The classical cubic models handle equilibrium-type properties usefully but perform poorly on derivative properties, with liquid heat capacities being a prominent weak point; combining a volume translation with a thermodynamically consistent α function substantially improves performance, especially for those derivative properties. The authors concluded that the translated-consistent Peng–Robinson variant, tc-PR, is the best overall model among those tested, while differences among the strongest variants remain property-dependent. Alongside the 2024 ACS Omega study of EOS/α-function combinations, and the long-running industry comparisons of PR and SRK begun decades ago, the current evidence points to a clear hierarchy: consistent, translated modernizations outperform both classical models, and between the two classics, PR is generally the better default when liquid-phase density matters.
Choosing Between PR and SRK for Gas-Phase Work
- General pure-component gas predictions (compressibility, gas density, vapor fugacity): Either model is defensible; away from the critical region the practical difference is usually small.
- Saturated-liquid densities matter: Prefer PR, or better, a volume-translated variant; the liquid-volume advantage is the most reproducible difference between the two classics.
- Heat capacities, speed of sound, or Joule–Thomson data must come from the EOS: Avoid relying on classical SRK or PR; use translated-consistent variants such as tc-PR, or dedicated correlations.
- Polar or strongly non-ideal compounds: Both models lose accuracy because their acentric-factor correlations were developed largely around hydrocarbon-like behavior; consider more appropriate property methods for such substances.
- Component data are available: Regressing parameters (especially the α function) against measured vapor pressures beats choosing between the two unmodified classics.
Common Misconceptions
Practical Takeaways
For pure-component gas predictions, PR and SRK are near-interchangeable: same inputs, same structure, similar vapor-phase results. Reach for PR — ideally a volume-translated or translated-consistent version — when liquid densities or derivative properties matter, keep SRK mainly where legacy processes and conventions favor it, and treat both classical models as first approximations rather than precision tools near the critical region. When accuracy claims matter for a specific compound or condition, validate against data or reference correlations; the modern benchmarking literature exists precisely because such checks, done systematically, changed the historical ranking.
FAQ
Which equation of state gives more accurate gas densities for a pure component?
Away from the critical region, Peng–Robinson and SRK usually differ only slightly for vapor-phase densities and compressibility factors because they share the same architecture and input data. Near the critical region or for dense supercritical gases, accuracy degrades for both, so results should be checked against experimental data or high-accuracy reference correlations.
Why is Peng–Robinson preferred for liquid densities?
PR's critical compressibility factor of 0.307 is closer to the values typical of real fluids than SRK's 1/3, so its saturated-liquid molar volumes are much closer to measured values. Volume translation can improve either model's liquid volumes without altering phase-equilibrium results, and modern translated variants build this in.
Do PR and SRK require different input data?
No. Both need the same three pure-component inputs: critical temperature, critical pressure, and acentric factor. Their parameters are fixed from critical-point conditions, with the acentric factor entering through the α function.
What are tc-PR and tc-RK?
They are modernized 'translated and consistent' counterparts of PR and SRK that combine a volume translation with an α function satisfying thermodynamic-consistency constraints at the critical point. A 2021 benchmark over about 1,800 fluids with more than 300,000 data points found them superior to the classical models, especially for saturated-liquid heat capacities, with tc-PR identified as the best overall model.
Is pure-component performance enough to pick an EOS for a process mixture?
Only partly. Mixture predictions also depend strongly on mixing rules and binary interaction parameters, so pure-component comparisons should be supplemented with validation against mixture data before committing to a property package.
Leave a Reply