SKH Research Group

Vapor–liquid–liquid equilibrium

Three phases.
One tangent.

Put vapor and liquid on the same Gibbs-energy reference. Find the globally stable phases, then separate their compositions from their amounts.

01 Choose the liquid model02 Set the pressure03 Check all phases04 Explore possible amounts

Global phase equilibrium

Pressure / kPa
Selected vapor fraction
Phases with nonzero amountFor the selected fraction assignment

03 · Interpret the equilibrium

Verify all phase candidates Fugacity equality and global stability

Equality checks use every phase on the selected coexistence line, including a phase with zero selected amount at a boundary. Global support tests both the liquid and vapor Gibbs curves over the full composition interval. Small equality residuals alone are insufficient.

Why are three phase fractions not unique? Two balances for three unknown amounts
fᴸᵅ + fᴸᵝ + fⱽ = 1
z₁ = fᴸᵅx₁ᵅ + fᴸᵝx₁ᵝ + fⱽy₁
fᴸᵅ, fᴸᵝ, fⱽ ≥ 0

At a binary three-phase state, fixed T fixes the equilibrium P and all phase compositions. The phase rule concerns intensive variables; it does not supply a missing balance for the amounts. For a feed strictly inside the span of the phase compositions, the feasible fractions generally form a line segment. An additional independent constraint, such as total enthalpy or volume with suitable property models, is needed to choose a unique point.

The slider interpolates between the two feasible endpoints, ordered by vapor fraction. It is a freely chosen assignment, not a calculated heat input, flash path or time evolution. Every point has the same total Gibbs energy and satisfies the balances. At the endpoints, one phase can disappear. For a feed at an extreme phase composition, the feasible set can collapse to a single point.

If the vapor composition lies outside the two liquid compositions, some feeds require a nonzero minimum vapor fraction. This is why the feasible interval is calculated from all three phase compositions. At a homogeneous azeotrope, liquid and vapor have the same composition, so their fractions can also be indeterminate at saturation.

One Gibbs reference for both phase types Ideal vapor and an activity-coefficient liquid
gᴸ(x) = x ln x + (1−x) ln(1−x) + gᴱ/(RT)
gⱽ(y) = y ln y + (1−y) ln(1−y)
+ y ln(P/P₁sat) + (1−y) ln(P/P₂sat)
fᵢᴸ = xᵢγᵢPᵢsat; fᵢⱽ = yᵢP

Here g is molar Gibbs energy divided by RT, after subtracting the same pure-liquid standard-state linear contribution from both phases. This vapor curve is not just ideal mixing entropy: the pressure-dependent standard-state offsets are essential. The model assumes ideal vapor fugacity coefficients and neglects liquid pressure corrections (Poynting factors).

For a globally supporting liquid common tangent, define aᵢ=xᵢᵅγᵢᵅ=xᵢᵝγᵢᵝ. Then:

P* = a₁P₁sat + a₂P₂sat
y₁* = a₁P₁sat / P*
g′ᴸ(xᵅ) = g′ᴸ(xᵝ) = g′ⱽ(y*)

For partially miscible liquids, P* is generally not P₁sat+P₂sat. The latter is the limiting result for completely immiscible liquids with unit activities in their respective pure-liquid phases. Above P*, the vapor lies above this liquid tangent. Below P*, a vapor state lies below it and destabilizes that LLE construction; the actual equilibrium is recalculated for the selected feed.

Margules conventions and liquid spinodal calculations are inherited from Lab 04. Equal A₁₂=A₂₁ recovers the symmetric model. NRTL uses dimensionless τ₁₂, τ₂₁ and symmetric α at the selected fixed temperature. All distinct globally supporting liquid gaps are retained; select a gap to explore its three-phase pressure.

How the solver finds the stable answer Stationary points, not a plotted-grid minimum

At the chosen pressure, all liquid–vapor candidates solve P=xγ₁P₁sat+(1−x)γ₂P₂sat. Every spinodal and homogeneous azeotrope partitions this bubble function into monotone intervals. For Margules, the azeotrope condition is quadratic. For NRTL, the derivative of ln(γ₁/γ₂) has at most one internal zero, obtained analytically from its positive cubic denominators. These stationary points bracket every isolated azeotrope, including a tangent root. Each bracket is searched, then any candidate whose tangent fails global support is discarded. The liquid–liquid tangent is retained only if the vapor cannot undercut it.

For a line ℓ(w)=μ₂/RT+(μ₁−μ₂)w/RT, the liquid-distance minimum is evaluated at both endpoints and every root g′ᴸ(w)=ℓ′. The ideal-vapor minimum is analytical:

min[gⱽ(w)−ℓ(w)] = ln[P / Σᵢ Pᵢsat exp(μᵢ/RT)]

The lowest feasible Gibbs energy across single phases and supported coexistence lines determines the result. Support tolerance is 10⁻⁹ in g; pressures within 10⁻¹⁰ in ln(P/P*) are treated as numerically coincident with VLLE. The diagram samples these verified constructions; it does not determine phase compositions by grid interpolation. Near-degenerate liquid states that cannot be verified are reported as unresolved.

Pure feeds are handled with their own saturation condition. A full-mixture tangent/TPD is not evaluated there because the absent-component chemical potential diverges. The present component's fugacity equality still applies.

Numerical curves & sources Reproduce the study

All curves, supported phase candidates, feasible fraction endpoints and fugacity checks are retained in the JSON export. Calculation source · Regenerate figures (Python).

Background: LearnChemE: immiscible-liquid phase diagrams; Howard DeVoe: Gibbs phase rule.

This is a synthetic, isothermal model study. No experimental vapor pressures, physical temperature sweep, energy balance, volume closure, kinetics or distillation design is claimed.