Advanced chemical engineering thermodynamics
One mixture.
Two phases. One criterion.
Predict the split. Change the conditions. Follow equal fugacity from a phase diagram to a flash calculation.
Loading thermodynamic models…
02 · Read the phase diagram
Pressure reveals the split.
Check the solution Balances & equilibrium
A small numerical residual verifies the equations solved. It does not establish experimental accuracy or validate a model outside its assumptions.
Follow the calculation Equations & iterations
Ideal vapor, pure-liquid reference states, and negligible Poynting correction give:
1 parameter: ln γ₁ = Ax₂², ln γ₂ = Ax₁²
2 parameters: gᴱ/(RT) = x₁x₂(A₁₂x₂ + A₂₁x₁)
ln γ₁ = x₂²[A₁₂ + 2(A₂₁ − A₁₂)x₁]
ln γ₂ = x₁²[A₂₁ + 2(A₁₂ − A₂₁)x₂]
x₂ = 1 − x₁
Pbubble = Σ xᵢγᵢPᵢˢᵃᵗ
zᵢ = (1 − β)xᵢ + βyᵢ
Bubble pressure is evaluated directly. Dew pressure is found by bracketing the liquid composition whose equilibrium vapor matches the specified y₁.
For binary TP flash, first bracket all roots of Pbubble(x₁) − P = 0. Activity coefficients are recalculated at every trial x₁. Retain tie lines that contain the feed; then solve Rachford–Rice with the converged Kᵢ = γᵢPᵢˢᵃᵗ/P.
Composition-root relative-pressure tolerance: 10⁻¹³. Rachford–Rice tolerance: 10⁻¹⁴. Independent component balance and equilibrium checks must each be ≤10⁻⁸. Near an azeotrope, β is poorly conditioned; at exact azeotropic or pure-component saturation, T and P do not determine β.
Current calculation
Temperature at the phase boundary
At the selected pressure, using x₁ = z₁ for bubble T and y₁ = z₁ for dew T. These are separate boundary calculations.
Inspect curve data Accessible table
Every fifth point of the selected model is shown. Download CSV for the complete curve. Missing T–x–y values mean the correlation range does not bracket a root.
Data, models & limits Sources and reproducibility
Antoine form: log₁₀(P/bar) = a − b/(T/K + c). Synthetic form: ln(P/Pref) = −ΔHvap/R · (1/T − 1/Tref).
Pressure inputs are restricted to 1–200 kPa as a teaching scope, not an accuracy guarantee. Temperature limits are the intersection of both component correlation ranges. Curves are clipped to this domain.
The engine checks that h(x₁) = x₁(1−x₁) d²(gL/RT)/dx₁² stays positive across the entire composition interval. For the two-parameter model, the minimum of this cubic is evaluated at both endpoints and every internal stationary point. Non-convex or near-critical parameter sets are rejected because this VLE-only calculation does not solve LLE. Equal A₁₂ = A₂₁ = A recovers the one-parameter model. The engine searches its binary vapor–liquid tie lines; it does not implement a general multicomponent TPD stability test, VLLE, or energy balances.
Model equations and parameter convention: Harvey Mudd College, Excess Gibbs Energy model table. Parameters are held constant with temperature in this teaching model.
Export includes inputs, diagnostics, property records and sampled curves. Inspect the engine · Property records · Regenerate a publication figure (Python)
Continue the argument