Non-random two-liquid model

The non‑random two‑liquid model (NRTL) is a thermodynamic model that predicts how the components of a liquid mixture interact at the molecular level. It was first proposed by Renon and Prausnitz in 1968 and has since become a staple in the calculation of phase equilibria for a wide range of industrial processes. The model belongs to the family of local‑composition approaches, which include the Wilson, UNIQUAC, and UNIFAC models. Unlike many classical equations of state, NRTL focuses on the activity coefficients of each component, offering a detailed description of how a compound’s local environment deviates from the bulk composition.

1. Historical Context

The 1960s were a period of rapid expansion in the field of thermodynamic modeling of mixtures. Prior to NRTL, the Wilson model (developed in the 1950s) and the UNIQUAC model (introduced in the 1970s) were already in use, but they suffered from limited accuracy for highly non‑ideal systems. Renon and Prausnitz introduced the NRTL model in 1968 as an improvement that explicitly accounted for the non‑randomness of local molecular arrangements. Their work was built upon Wilson’s hypothesis that the local concentration around a molecule differs from the bulk concentration, a concept that had been gaining traction in the 1950s and 1960s.

In 1976, Flemr published a critical assessment that highlighted an inconsistency in local‑composition models when applied to a single-fluid representation of a real mixture. The core of Flemr’s argument was that the assumption of independence between the local compositions around different species (i.e., the local composition around molecule i being independent of that around molecule j) is not valid. This insight paved the way for the two‑liquid interpretation that underpins the consistency of NRTL and related models.

2. Theoretical Foundations

2.1 Wilson’s Hypothesis

At the heart of the NRTL model lies Wilson’s hypothesis, which states that the local concentration surrounding a molecule i in a mixture is not equal to the bulk mole fraction xᵢ. Instead, it is influenced by the interaction energies between the central molecule and its neighbors. Two key energy terms are introduced:

• Uᵢᵢ: Interaction energy of molecule i with molecules of its own kind.
• Uᵢⱼ: Interaction energy of molecule i with molecules of a different kind j.

The difference Uᵢᵢ – Uᵢⱼ determines how strongly a molecule prefers to be surrounded by like or unlike molecules, thereby creating a non‑random local environment.

2.2 Local‑Composition Concept

The local‑composition approach assumes that the local mole fraction of component j around a central molecule i, denoted xᵢⱼ, differs from the bulk mole fraction xⱼ. The NRTL model uses this concept to derive the activity coefficient γᵢ of each component:

\[ \ln \gamma_i = \sum_{j} \tau_{ji} \frac{G_{ji}x_j}{\sum_k G_{ki}x_k} + \sum_{j} \frac{x_j G_{ij}}{\sum_k G_{ik}x_k}\left(\tau_{ij} - \frac{\sum_l \tau_{jl} G_{jl}x_l}{\sum_m G_{jm}x_m}\right) \]

where \( \tau_{ij} \) and \( G_{ij} \) are temperature‑dependent parameters that encode the energy differences and non‑randomness, respectively. The explicit form of the equation is not reproduced here to avoid unnecessary complexity; however, the key takeaway is that the NRTL model relates the activity coefficient to both the composition of the mixture and the energetic differences between interacting species.

3. Mathematical Formulation

The NRTL model expresses the activity coefficient \( \gamma_i \) of component i as a function of the mole fractions \( x_j \) of all components in the liquid phase. The model introduces two sets of binary interaction parameters:

• \( \tau_{ij} \): Dimensionless energetic parameters representing the relative interaction energy between species i and j.
• \( \alpha_{ij} \): Non‑randomness parameters that quantify the degree to which local composition deviates from a random mixture.

These parameters are typically fitted to experimental data, such as vapor–liquid equilibrium (VLE) measurements. The temperature dependence of the parameters can be incorporated through empirical correlations, allowing the NRTL model to predict phase behavior over a wide range of temperatures.

4. Parameter Estimation

While the source text does not detail the exact procedure for estimating NRTL parameters, it is standard practice to use experimental VLE data. The process usually involves:

1. Collecting experimental data for the mixture of interest over a range of temperatures and compositions.
2. Defining an objective function that quantifies the difference between measured and calculated properties (e.g., bubble point pressure, composition of vapor phase).
3. Optimizing the binary interaction parameters \( \tau_{ij} \) and \( \alpha_{ij} \) to minimize the objective function, often using nonlinear regression techniques.

Because NRTL is a local‑composition model, the number of parameters grows with the number of binary pairs in the system, making the estimation process computationally intensive for large mixtures.

5. Applications in Chemical Engineering

The NRTL model is frequently applied in the field of chemical engineering to calculate phase equilibria. Typical uses include:

• Design of distillation columns: Accurate prediction of vapor–liquid equilibrium curves is essential for determining column operating conditions.
• Extraction processes: Modeling the distribution of solutes between aqueous and organic phases.
• Solvent selection: Assessing how different solvent mixtures behave under varying temperatures and compositions.
• Process simulation: Integrating NRTL into process simulators (e.g., Aspen Plus, HYSYS) to provide reliable thermodynamic data for complex systems.

Because the model captures non‑ideal behavior, it is particularly valuable for systems with strong interactions, such as hydrogen bonding, dipole–dipole forces, or significant size and shape disparities between components.

6. Consistency Issues and the Two‑Liquid Hypothesis

Local‑composition models, including NRTL, are not thermodynamically consistent when applied to a single‑fluid representation of a real mixture. This inconsistency arises from the assumption that the local composition around molecule i is independent of that around molecule j. Flemr’s 1976 study demonstrated that this assumption fails for real systems.

A remedy is the hypothetical two‑liquid model. In this framework, the mixture is conceptualized as consisting of two interpenetrating liquids, each with its own local composition. When the two liquids are treated separately, the NRTL equations become thermodynamically consistent. In practice, however, the two‑liquid interpretation remains a conceptual tool rather than a physically observable phenomenon.

Other models that maintain consistency between bulk and local concentrations include COSMO‑RS and COSMOSPACE, both of which employ quantum‑chemical calculations to predict interaction energies and local composition effects.

7. Comparison with Other Local‑Composition Models

While all these models share the local‑composition concept, NRTL distinguishes itself by explicitly incorporating a non‑randomness parameter, enabling it to capture subtle deviations from ideal behavior in many industrially relevant mixtures.

8. Practical Implementation

Implementing the NRTL model in a process simulator involves several steps:

1. Database selection: Many commercial simulators provide built‑in NRTL parameter sets for common binary systems. For new mixtures, users must supply their own parameters.
2. Parameter fitting: Experimental VLE data are entered into the simulator’s parameter estimation routine to derive \( \tau_{ij} \) and \( \alpha_{ij} \).
3. Model validation: The simulator’s predictions are compared against additional experimental data (e.g., ternary VLE) to ensure reliability.
4. Process simulation: With validated parameters, the NRTL model is used to compute equilibrium properties at each step of the process (e.g., column stages, extraction contacts).

Because NRTL requires a separate set of parameters for each binary pair, the effort to develop a comprehensive database can be significant. Nonetheless, the model’s predictive power often justifies the investment, especially for processes where accurate phase behavior is critical to safety, efficiency, or product quality.

9. Case Studies

9.1 Ethanol–Water System

The ethanol–water mixture is a classic example of a strongly non‑ideal system. NRTL parameters fitted to experimental data reproduce the well‑known azeotrope at 95.6 % ethanol by weight. The model accurately predicts the vapor composition at various pressures and temperatures, enabling the design of efficient distillation columns for ethanol purification.

9.2 Aromatic–Aliphatic Solvent Mixtures

In processes that involve the separation of aromatic hydrocarbons (e.g., toluene) from aliphatic solvents (e.g., hexane), the NRTL model captures the large interaction energy differences due to π‑electron interactions. Accurate VLE predictions are essential for designing solvent‑based extraction steps or for optimizing solvent recovery.

9.3 Ionic Liquid–Water Systems

Ionic liquids often exhibit highly non‑ideal behavior with water due to strong electrostatic interactions. NRTL has been employed to model the solubility of water in various ionic liquids, guiding the selection of ionic liquids for applications such as CO₂ capture or biomass processing.

10. Future Directions

Research into local‑composition models continues to evolve. Some emerging trends include:

• Hybrid models that combine NRTL with machine‑learning techniques to predict parameters for novel mixtures.
• Integration with quantum‑chemical calculations to derive \( \tau_{ij} \) from first principles, reducing reliance on experimental data.
• Enhanced consistency frameworks that address the limitations highlighted by Flemr, potentially leading to new two‑fluid or multi‑fluid formulations.

As computational power grows and data availability increases, the NRTL model remains a cornerstone of thermodynamic modeling, offering a balance between accuracy and practicality.

11. Conclusion

The non‑random two‑liquid model (NRTL) is a powerful tool for predicting the activity coefficients of components in liquid mixtures, thereby enabling accurate calculation of phase equilibria. Introduced by Renon and Prausnitz in 1968, the model builds on Wilson’s hypothesis of local concentration differences and explicitly incorporates non‑randomness at the molecular level. Although local‑composition models face consistency challenges—highlighted by Flemr in 1976—they retain practical value when applied within a two‑liquid framework. Compared to other local‑composition models, NRTL’s explicit treatment of non‑randomness makes it particularly suitable for systems with strong non‑ideal interactions. Its widespread use in chemical engineering—from distillation to solvent extraction—underscores its enduring relevance.

FAQ

What is the primary purpose of the NRTL model? The NRTL model is used to predict activity coefficients of components in liquid mixtures, which in turn allows calculation of phase equilibria such as vapor–liquid equilibrium (VLE) curves.

How does the NRTL model differ from the Wilson model? While both are local‑composition models, NRTL includes an explicit non‑randomness parameter that accounts for deviations from random mixing, giving it greater accuracy for strongly non‑ideal systems.

**