Vapor-Liquid Equilibrium (VLE) Prediction with NRTL & UNIFAC - Chemical Engineering Knowledge Platform

Industry Applications

Pharmaceutical purification, biofuel dehydration, petrochemical fractionation, solvent recovery

Key Standards

AIChE DDBST, NIST Standard Reference Database 103, ISO 8573-1 (for trace moisture in compressed air streams)

Typical Scale

Lab (10 mL), pilot (100 L), industrial (10⁴–10⁶ kg/h feed)

⚠️ Why It Matters

1

Inaccurate VLE predictions

↓

2

Wrong distillation column stage count or reflux ratio

↓

3

Poor product purity or yield

↓

4

Excessive energy consumption or off-spec product

↓

5

Plant startup delays or regulatory non-compliance

↓

6

Capital overdesign or operational instability

📘 Definition

Vapor-liquid equilibrium (VLE) describes the thermodynamic state at which a multicomponent liquid phase coexists in equilibrium with its vapor phase at fixed temperature, pressure, and composition. Activity coefficient models such as NRTL (Non-Random Two-Liquid) and UNIFAC (UNIQUAC Functional-group Activity Coefficients) quantify non-ideal interactions to predict phase compositions and bubble/dew points. These models are foundational for rigorous simulation of separation processes in chemical process design.

🎨 Concept Diagram

AI-generated illustration for visual understanding

💡 Engineering Insight

Never trust a UNIFAC prediction without verifying the group assignment logic — misclassifying a 'carboxylic acid' as 'alcohol' changes Δgₘₙ by >150 J/mol and shifts azeotrope composition by ±0.15 mol-fraction. Always cross-check UNIFAC with NRTL fitted to at least one binary pair in your system.

📖 Detailed Explanation

VLE prediction begins with Raoult’s law — the ideal case where partial pressure equals vapor pressure times liquid mole fraction. Real mixtures deviate due to molecular size, polarity, and hydrogen bonding; activity coefficients (γᵢ) correct for this. NRTL models these deviations using local composition concepts and three adjustable parameters per binary pair: τᵢⱼ (energy), τⱼᵢ, and αᵢⱼ (non-randomness).

NRTL excels when experimental binary VLE data exist — it fits τ and α to minimize objective functions like RMSD in y₁ or T. UNIFAC, by contrast, avoids fitting by decomposing molecules into functional groups (e.g., CH₃, OH, COOH) and using pre-regressed group–group interaction parameters. Its strength lies in predicting new mixtures, but accuracy drops sharply for associating fluids or electrolytes.

Advanced practice demands hybrid approaches: use UNIFAC to initialize NRTL regression, constrain αᵢⱼ within ±0.05 of literature values, and apply consistency tests (e.g., Van Ness area test) to ensure γᵢ data satisfy Gibbs–Duhem. For industrial applications involving trace impurities (<0.1 mol%), residual activity coefficient uncertainty must be propagated through column simulation to assess distillate specification risk — a step routinely omitted in early-stage design but critical for FDA-submitted pharmaceutical separations.

🔄 Engineering Workflow

Step 1

Step 1: Identify components, operating P/T range, and required purity specs

→

Step 2

Step 2: Screen available experimental VLE data (DDBST, NIST ThermoData Engine, DECHEMA)

→

Step 3

Step 3: Select model (NRTL if binary data exists; UNIFAC if only structure known)

→

Step 4

Step 4: Regress NRTL parameters or apply UNIFAC group contributions with consistent database version

→

Step 5

Step 5: Validate against independent data (e.g., isobaric VLE at 1 atm, azeotrope location, salt effect if present)

→

Step 6

Step 6: Integrate into flowsheet simulator (Aspen Plus, CHEMCAD) with convergence settings tuned for non-ideality

→

Step 7

Step 7: Perform sensitivity analysis on key parameters (αᵢⱼ, Δgₘₙ) and document uncertainty bands

📋 Decision Guide

Rock/Field Condition	Recommended Design Action
Polar–nonpolar mixture (e.g., methanol–hexane), no binary data available	Use UNIFAC (Lyngby version, 2018 parameters); validate with batch distillation residue curve analysis
Strongly associated system (e.g., acetic acid–water) with high-boiling azeotrope	Prefer NRTL with temperature-dependent αᵢⱼ and ternary fit; avoid UNIFAC due to poor H-bond group representation
Hydrocarbon mixture (C₃–C₈) at near-ideal conditions (P < 10 bar, T < 150°C)	Use Wilson or simplified NRTL (αᵢⱼ = 0.3); UNIFAC overkill and introduces unnecessary uncertainty

📊 Key Properties & Parameters

Activity Coefficient (γᵢ)

0.1–10.0 (unitless)

Dimensionless factor quantifying deviation from Raoult’s law behavior for component i in a liquid mixture.

⚡ Engineering Impact:

Directly determines relative volatility and thus minimum reflux ratio and theoretical stage requirements.

Binary Interaction Parameter (αᵢⱼ)

0.0–0.47 (unitless)

Empirical parameter in NRTL representing non-randomness in local composition around component i and j.

⚡ Engineering Impact:

Controls shape of activity coefficient curves; errors > ±0.05 cause >5% dew point error in ethanol–water systems.

Group Contribution Parameter (Δgₘₙ)

−200 to +300 J/mol

UNIFAC-specific interaction parameter between functional groups m and n, derived from experimental data.

⚡ Engineering Impact:

Determines predictive accuracy for unseen mixtures; uncertainty > ±50 J/mol risks >10% vapor-phase mole fraction error in ketone–hydrocarbon systems.

Relative Volatility (αᵢⱼ)

1.05–50.0 (unitless)

Ratio of effective volatilities of components i and j, αᵢⱼ = (yᵢ/xᵢ)/(yⱼ/xⱼ), where y = vapor mole fraction, x = liquid mole fraction.

⚡ Engineering Impact:

Dictates feasibility of separation: α < 1.15 often requires extractive or azeotropic distillation.

📐 Key Formulas

NRTL Activity Coefficient

ln γᵢ = ∑ⱼ(xⱼτⱼᵢGⱼᵢ/∑ₖxₖGₖᵢ) + ∑ⱼ[xⱼ(Gᵢⱼ/∑ₖxₖGₖⱼ)(τᵢⱼ − ∑ₖxₖτₖⱼGₖⱼ/∑ₖxₖGₖⱼ)]

Calculates component i activity coefficient using local composition and energy parameters.

Variables:

Symbol	Name	Unit	Description
γᵢ	Activity coefficient of component i	dimensionless	Measure of deviation from ideal solution behavior for component i
xⱼ	Mole fraction of component j	dimensionless	Mole fraction of component j in the liquid phase
τⱼᵢ	Energy parameter for pair j-i	dimensionless	Binary interaction parameter representing excess energy between components j and i
Gⱼᵢ	Margules energy parameter for pair j-i	dimensionless	Local composition parameter related to the energy parameter τⱼᵢ and non-ideality
τᵢⱼ	Energy parameter for pair i-j	dimensionless	Binary interaction parameter representing excess energy between components i and j
Gᵢⱼ	Margules energy parameter for pair i-j	dimensionless	Local composition parameter related to the energy parameter τᵢⱼ and non-ideality

Typical Ranges:

Ethanol–water at 78°C

γ₁ = 1.2–2.1, γ₂ = 3.8–4.9

Acetone–chloroform at 50°C

γ₁ = 0.7–1.0, γ₂ = 1.0–1.3

⚠️ γᵢ < 0.1 or > 100 indicates model breakdown — verify phase stability or consider association models.

🔧 Open Calculator

UNIFAC Group Contribution

ln γᵢ = ln γᵢᶜ + ln γᵢˢ

Splits activity coefficient into combinatorial (c) and residual (s) terms based on group surface area and interaction parameters.

Variables:

Symbol	Name	Unit	Description
γᵢ	Activity coefficient of component i	dimensionless	Measure of deviation from ideal solution behavior for component i
γᵢᶜ	Combinatorial contribution to activity coefficient of component i	dimensionless	Part of activity coefficient accounting for differences in molecular size and shape
γᵢˢ	Residual contribution to activity coefficient of component i	dimensionless	Part of activity coefficient accounting for energetic interactions between molecular groups