Aerospace
Prandtl-Meyer Expansion Fan: How Supersonic Flow Accelerates Around a Corner
Turn a stream of Mach 2 air by just 10° around a convex corner and it does something impossible for a subsonic flow: it speeds up, dropping its static pressure by roughly 45% and its temperature by about 16% in a fan of expansion waves only a few millimetres thick. There is no shock, no entropy jump, no loss — the acceleration is smooth, reversible, and governed by a single closed-form equation Ludwig Prandtl and Theodor Meyer wrote down in 1908.
This Prandtl-Meyer expansion fan is the mirror image of the oblique shock. It is what lets an over-expanded rocket nozzle keep working, what forms the diamond shock-cell pattern in an afterburner plume, and what the leeward surface of every supersonic wing does with its flow. Get the turn angle and the ν(M) function right and you can predict the exit Mach number to three decimals with a pocket calculator.
- Governing equationν(M) = √((γ+1)/(γ−1))·atan√((γ−1)(M²−1)/(γ+1)) − atan√(M²−1)
- ProcessIsentropic (Δs = 0), reversible
- Turn effectM↑, p↓, T↓, ρ↓ across convex corner
- ν max (γ=1.4)≈ 130.45° as M → ∞
- Mach wave angleµ = asin(1/M), e.g. 30° at M=2
- Used inNozzles, supersonic airfoils, method of characteristics
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why supersonic flow speeds up around a corner
In subsonic flow, a convex corner would let information travel upstream and the streamlines would smoothly bend with a local pressure adjustment. In supersonic flow (M > 1) disturbances cannot outrun the stream, so a corner communicates itself only through Mach waves — infinitesimally weak pressure lines inclined at the Mach angle µ = asin(1/M) to the local velocity. When the wall turns away from the flow (a convex, or expansive, corner), the flow must expand to fill the widening space.
The key result is counter-intuitive: for M > 1 the area-velocity relation dA/A = (M² − 1)·dV/V flips sign relative to subsonic flow. An expanding stream tube therefore demands higher velocity, lower pressure, lower density, and lower temperature. Each incremental Mach wave turns the flow by a tiny dθ and raises the Mach number by dM; the corner is resolved by a centred fan of these waves — the Prandtl-Meyer expansion fan. Because every wave is infinitesimally weak, the process crosses no shock and generates no entropy: it is isentropic and fully reversible. Run the same air back through a matching compression corner and you recover exactly the upstream state.
The Prandtl-Meyer function ν(M)
Integrating the incremental turning dθ = √(M²−1)·dV/V across the fan yields the Prandtl-Meyer function, the angle through which a flow starting at M = 1 must expand to reach Mach number M:
ν(M) = √((γ+1)/(γ−1)) · atan√((γ−1)(M²−1)/(γ+1)) − atan√(M²−1)
with γ the ratio of specific heats (1.4 for air, ~1.2 for rocket exhaust). ν is measured in degrees or radians and is a monotonically increasing function of M. Its value at M = 1 is exactly 0, and as M → ∞ it asymptotes to a finite maximum:
- ν_max = (π/2)·(√((γ+1)/(γ−1)) − 1) = 130.45° for γ = 1.4.
- This ceiling means a flow can only be turned about 130.45° before, in the limit, reaching infinite Mach number and zero pressure and temperature — a vacuum. Turn beyond that and the flow simply cannot follow the wall; it separates.
The design rule for a finite corner of angle Δθ is simply ν(M₂) = ν(M₁) + Δθ. You look up ν(M₁), add the geometric turn angle, and invert ν to get M₂. Because ν has no closed-form inverse, M₂ is found from tables, a chart, or a two-line Newton iteration.
Sizing an expansion: a Mach 2 corner worked end to end
Take air (γ = 1.4) at M₁ = 2.00, static pressure p₁ = 100 kPa, static temperature T₁ = 250 K, expanding around a Δθ = 10° convex corner. The procedure:
- Step 1 — upstream ν: ν(2.00) = 26.38°.
- Step 2 — add the turn: ν(M₂) = 26.38° + 10° = 36.38°.
- Step 3 — invert ν: M₂ ≈ 2.385.
- Step 4 — apply isentropic relations (stagnation quantities are conserved because Δs = 0). With p₀/p = (1 + 0.2M²)^3.5 and T₀/T = (1 + 0.2M²): p₀/p₁ = 7.824, p₀/p₂ = 14.28, so p₂/p₁ = 7.824/14.28 = 0.548 → p₂ ≈ 54.8 kPa.
- Step 5 — temperature: T₀/T₁ = 1.80, T₀/T₂ = 2.137, so T₂/T₁ = 0.842 → T₂ ≈ 210 K, a 40 K drop.
The fan itself spans from the leading Mach wave at µ₁ = asin(1/2.00) = 30.0° to the trailing wave at µ₂ = asin(1/2.385) = 24.8°, both measured from the local flow direction. The whole expansion is contained in a wedge of open angle roughly (µ₁ − µ₂ + Δθ) ≈ 15°. Density falls with ρ₂/ρ₁ = (p₂/p₁)/(T₂/T₁) = 0.651 by the ideal-gas law.
Method of characteristics: turning theory into nozzle geometry
The single most important engineering use of ν(M) is the method of characteristics (MOC), the workhorse for designing shock-free supersonic nozzle and inlet contours. In a 2-D isentropic supersonic field the flow properties are constant along characteristic lines — Mach waves — and the Riemann invariants are conserved:
- Along a left-running characteristic (C⁻): θ + ν(M) = constant = K⁻.
- Along a right-running characteristic (C⁺): θ − ν(M) = constant = K⁺.
At any grid intersection, θ = ½(K⁻ + K⁺) and ν = ½(K⁻ − K⁺), which instantly gives the local Mach number by inverting ν. A minimum-length supersonic nozzle is built by first over-expanding the flow at the throat through a sharp-cornered expansion section whose total turn equals ½·ν(M_exit), then straightening it in a cancellation section where the wall is shaped so that each reflected characteristic is absorbed rather than reflected — producing a shock-free, uniform, parallel exit flow. This is exactly how a Mach 3 wind-tunnel nozzle or a rocket bell contour is laid out before CFD refinement.
Where expansion fans show up in real hardware
Prandtl-Meyer expansions are everywhere in high-speed flow:
- Supersonic airfoils and diamond/biconvex wings: the flow accelerating over the shoulder of a double-wedge section expands, dropping surface pressure and contributing to lift and wave drag. Linearized supersonic (Ackeret) theory gives surface pressure coefficient C_p = 2θ/√(M²−1); exact answers come from stringing together shocks on windward faces and expansion fans on leeward faces.
- Over-expanded and under-expanded nozzles: a rocket nozzle whose exit pressure exceeds ambient (under-expanded) dumps the pressure difference in a Prandtl-Meyer fan at the lip, producing the barrel-shocks and expanding plume you see on a Falcon 9 upper-stage or a launch at altitude.
- Shock-cell / diamond patterns: in an imperfectly expanded jet, expansion fans reflect off the constant-pressure free boundary as compression waves, coalesce into shocks, reflect again as fans, and repeat — the periodic shock-diamond pattern and its screech tone in afterburner plumes.
- Supersonic inlets and scramjets: corner expansions around cowl lips and isolator geometry are designed with MOC to avoid unstart.
- Wind-tunnel nozzles: every blow-down supersonic tunnel throat-to-test-section contour is an MOC expansion-cancellation design.
Trade-offs, limits, and failure modes
The expansion is beautifully clean, but the physics has hard boundaries:
- The ν_max ceiling: you cannot turn more than ν_max − ν(M₁). Ask a nozzle wall to turn the flow past this and the stream detaches from the surface, leaving a separated region and re-circulation — a real limit on how far an over-expanded plume or a leeward wing surface can follow geometry.
- The isentropic assumption breaks with viscosity: real expansions still form a boundary layer; a strong favourable pressure gradient thins it, but downstream re-compression (shock–boundary-layer interaction) can separate the flow. The inviscid fan is only the core solution.
- Condensation shocks: because expansion drops T sharply, moist or vapour-laden flow can cross saturation and condense, releasing latent heat and forming a condensation shock — the visible white cone over a transonic wing, and a genuine error source in humid wind tunnels.
- Non-centred vs centred fans: a sharp corner gives a centred fan (all waves from one point, infinite local gradient); a gradual convex curve spreads the same total turn over a finite arc, gentler on the boundary layer but geometrically longer.
- Best practice: chain expansions and shocks correctly in sequence — an oblique shock is not the reverse of an expansion fan because the shock adds entropy. Turning a flow out and back by equal angles through a shock then a fan (or vice-versa) does not return the original state; total pressure is permanently lost at the shock.
| Property | Expansion fan (convex turn) | Oblique shock (concave turn) |
|---|---|---|
| Corner geometry | Flow turns away from itself | Flow turns into itself |
| Mach number | Increases 2.00 → 2.39 | Decreases 2.00 → 1.64 |
| Static pressure ratio p₂/p₁ | ≈ 0.55 (drop) | ≈ 1.71 (rise) |
| Static temperature | Falls ~40 K | Rises ~43 K |
| Entropy change Δs | 0 (isentropic) | > 0 (loss, non-isentropic) |
| Wave structure | Continuous fan of Mach waves | Single thin discontinuity |
Frequently asked questions
Why does supersonic flow accelerate through an expansion fan instead of slowing down?
Above Mach 1 the area-velocity relation dA/A = (M²−1)·dV/V changes sign, so an expanding stream tube demands higher velocity. Around a convex corner the flow area effectively grows, and the only way to conserve mass and momentum isentropically is to speed up while pressure, density, and temperature all fall. This is the exact opposite of subsonic (Bernoulli-style) behaviour in an expanding duct.
Is a Prandtl-Meyer expansion really lossless?
In the inviscid ideal-gas model, yes — it is isentropic with Δs = 0 and stagnation pressure fully conserved, because it is built from infinitely many infinitesimally weak Mach waves, none of which is a shock. In reality a boundary layer, heat transfer, and possible condensation introduce small losses, but the core flow is far cleaner than any shock, which always destroys total pressure.
How is an expansion fan different from an oblique shock?
An oblique shock forms at a concave corner where the flow turns into itself: it is a single thin discontinuity that compresses the flow (M drops, p and T rise) and increases entropy. An expansion fan forms at a convex corner where the flow turns away from itself: it is a continuous, spread-out, isentropic fan that accelerates the flow (M rises, p and T drop). They are not reverses of each other because the shock is irreversible.
How do you calculate the downstream Mach number for a given turn angle?
Use ν(M₂) = ν(M₁) + Δθ. Look up the Prandtl-Meyer function at the incoming Mach number, add the geometric turn angle in degrees, then invert ν to get M₂ from tables or a Newton iteration. Because ν has no closed-form inverse, this last step is numerical — but for M₁ = 2 and Δθ = 10°, ν goes 26.38° → 36.38°, giving M₂ ≈ 2.39.
What is the maximum angle a supersonic flow can turn through?
The Prandtl-Meyer function asymptotes to ν_max = (π/2)(√((γ+1)/(γ−1)) − 1) = 130.45° for air (γ = 1.4). Starting from M = 1 you can turn at most that much; from a higher initial Mach number the available turn is ν_max − ν(M₁). Beyond it the flow would need infinite Mach number (a perfect vacuum), so instead it separates from the wall.
How does γ change the result for rocket exhaust versus air?
The whole ν(M) curve depends on γ. Combustion gases have γ ≈ 1.2 rather than 1.4, which raises ν_max to about 208° and lets the flow turn and expand much further for a given geometry — one reason high-area-ratio rocket bells can extract so much expansion. You must use the correct, temperature-dependent γ for the exhaust species, not the air value.