When the wedge is too blunt

Each curve on the θ–β–M chart rises from the Mach angle, reaches a maximum, and comes back down to ninety degrees. The maximum is a real boundary rather than a feature of the plot: past that deflection, no shock at any angle turns the flow that far.

What the solver has to do about it #

This is a place where a numerical routine has to make a physical statement, and getting it wrong would be invisible.

maxDeflection finds the maximum by ternary search on the θ(β) curve between the Mach angle and ninety degrees, which is smooth and unimodal there. At Mach 2 it returns 22.974°, at $\beta = 64.67°$. obliqueShock then compares any requested deflection against that value and throws if it exceeds it, with a message naming the detachment angle.

The alternative is what a careless implementation does: clamp to the maximum, or return whichever bracket end the bisection happens to converge on. Either produces a number. That number would be drawn as a neat attached shock on a forty-degree wedge at Mach 2 — a picture that is smooth, captioned, correctly coloured, and describes nothing that has ever happened.

assertDetaches is the check, and it is written to be asked in both directions. Given a deflection past the maximum it verifies that obliqueShock really does refuse. Given one within reach it itself throws, on the grounds that the check has been asked the wrong question — a detachment test applied to an attached case is a test that will pass forever and mean nothing.

The numbers, which are not intuitive #

The maximum deflection rises with Mach number, which is the opposite of what one might guess about a flow becoming harder to influence.

At Mach 1.2 a wedge of more than about seven degrees included angle already detaches, which is why transonic and low-supersonic aircraft need genuinely sharp leading edges. At Mach 5 a 40° half-angle wedge still carries an attached shock.

And there is a ceiling. As the Mach number goes to infinity the limiting deflection approaches 45.6° and stops, so there exists a turn no attached shock can ever accomplish, at any speed whatever. A body blunter than that is guaranteed a detached shock however fast it goes.

What appears instead #

The flow does not become impossible. It stops being attached.

A shock forms standing off in front of the body, curved, with a shape that is part of the solution rather than given by the geometry. Along that curve, the local shock angle varies continuously: near the axis it is essentially normal to the flow, and far out to the sides it weakens and asymptotes to a Mach line.

So every point on a bow shock is a solution of the θ–β–M relation, and the body’s job is only to decide where the curve goes. Near the axis the flow crosses a nearly normal shock and becomes subsonic; there is a pocket of subsonic flow between the shock and the nose, and within that pocket the flow can and does sense the body’s shape and turn gradually around it.

That subsonic pocket is how the flow solves an otherwise impossible problem. It cannot turn 40° abruptly, so it stops being supersonic first, and then turns as much as it likes.

Which is why the mathematics gets much harder #

The elegance of the previous rung came entirely from the flow being supersonic and hyperbolic: influence travels one way, faces can be solved in order, and the answer is algebraic.

A detached shock destroys all of that. The subsonic pocket is elliptic, the supersonic region outside it is hyperbolic, and the sonic line between them is part of the unknown — which is exactly the structural difficulty the transonic range has, arriving here at high speed rather than low.

Worse, the shock is curved, so each streamline crosses it at a different angle, receives a different entropy rise, and emerges with a different total pressure. By Crocco’s relation that entropy gradient is vorticity, so the flow behind a bow shock is rotational in a fluid with no viscosity and no velocity potential exists for it.

The blunt-body problem was one of the hardest open problems in aerodynamics through the 1950s, and it was solved by computers rather than by analysis. That is not a failure of the analysts.

The deliberate bluntness of a re-entry vehicle #

Here is the result that makes detachment a design tool rather than a hazard, and it is one of the best pieces of applied reasoning in the subject.

A vehicle returning from orbit arrives at enormous speed and has to lose essentially all of its kinetic energy as heat. The heat is going to be produced whatever shape it is; the only question is where it ends up.

A sharp nose carries an attached, weak, oblique shock. Little entropy is produced, so the flow retains most of its total pressure and most of its energy remains as ordered kinetic energy — right up against a thin, hot boundary layer on a slender body with almost no volume to absorb heat. The sharp tip melts.

A blunt nose forces a detached shock that is nearly normal on the axis. Most of the total pressure is destroyed there, which is to say most of the kinetic energy is converted to thermal energy in the gas, out in front of the vehicle, and then swept away round it. The stand-off distance keeps the hottest gas away from the surface, and there is enough volume behind the nose to carry ablative material. The vehicle also acquires a great deal of pressure drag in the bargain, which for something trying to slow down is not a cost — a rare case on this site where drag is the objective rather than the enemy.

The counter-intuitive statement, put plainly: the shape that wastes the most energy is the one that survives. Every crewed re-entry vehicle ever flown has been blunt, and Harvey Allen’s 1951 argument for it was initially disbelieved on the reasonable-sounding ground that streamlining is good.

Detachment is not stall, and the difference is instructive #

There is a temptation to file this beside separation, since both are a flow refusing to follow a surface past some limiting angle. The comparison is worth making carefully, because the two mechanisms could hardly be less alike.

Separation is a viscous phenomenon. The boundary layer runs out of momentum against an adverse pressure gradient, reverses at the wall, and the outer flow leaves the surface. It depends on the Reynolds number, on whether the layer is laminar or turbulent, on surface roughness, and on history — and it is notoriously hard to predict.

Detachment has no viscosity in it anywhere. It is a statement that a set of algebraic equations has no root, and the boundary is a number computed exactly from the Mach number and $\gamma$: 22.974° at Mach 2, and not 22.9° or 23.1°. It does not depend on the Reynolds number, on the surface, or on anything about the fluid except its ratio of specific heats.

The pair is a good illustration of a distinction this site keeps returning to. Some limits in fluid mechanics are properties of a real fluid’s messy behaviour, and some are properties of the equations themselves. The second kind can be computed exactly and refused exactly, and it is the kind that belongs in an assertion.

Where else this boundary is met #

Detachment is not only a nose problem, and two other places it appears are worth naming.

Intake ramps. An external-compression intake turns the flow through a series of ramps, and each ramp’s turn must stay within the detachment angle at the local Mach number — which falls after every shock. So the later ramps have less turning available to them than the first, and the design is a sequence of shrinking allowances. Exceed one and the intake unstarts: the shock system pops out in front of the inlet, spilling flow, with a violent and sometimes destructive transient.

Wing leading edges. A supersonic wing’s leading edge is subsonic or supersonic depending on the component of the free stream normal to it. If it is swept far enough back that the normal component is subsonic, a round leading edge is acceptable and even beneficial, exactly as at low speed. If not, the edge must be sharp or it will carry a detached shock and pay for it. That is the aerodynamic argument for wing sweep at supersonic speed, and it is a different argument from the subsonic one.

Why a cone gets away with so much more #

The table above is a wedge’s allowance, and a cone of the same half-angle stays attached to nearly twice it — 41° against 22.97° at Mach 2. That is not a small correction and the reason for it is worth having, because it says what “three-dimensional relief” actually means.

Past a wedge, the flow behind the shock runs parallel to the surface and goes on running parallel to it forever: in two dimensions there is nowhere else for it to go. So the entire turn must be accomplished at the shock, in one step, and the shock has to be strong enough to do all of it.

Past a cone, the streamlines can spread sideways round the axis as they travel aft, so the area available to them grows. The flow therefore does not have to turn the full cone angle as it crosses the shock. It turns part of the way there, and the rest gradually and isentropically in the region between the shock and the surface, where it is still being compressed but no longer discontinuously.

Two things follow. The shock is weaker than the wedge’s for the same body angle, so less total pressure is lost and the wave drag is lower. And because only part of the turn is asked of it, the shock stays attached to body angles the wedge could never manage.

The price is that the answer stops being algebraic. The flow behind a cone’s shock is not uniform — every property varies with polar angle between the shock and the surface — so there is no jump relation to evaluate and no closed form to quote. What there is instead is an ordinary differential equation in that angle, derived by Taylor and Maccoll in 1933 and integrated numerically, which was among the first pieces of aerodynamics to be settled by arithmetic rather than by algebra. Nothing on this page computes it.

What a body does about it, if it can #

A designer who has been told the deflection limit has one obvious move: take the turn in stages.

A single wedge of 30° at Mach 2 detaches. A wedge of 15° followed by a further 15° does not — the first shock turns the flow through 15° and leaves it at Mach 1.45, and at Mach 1.45 the available deflection is 10.6°, which is not enough for the second. So a naive two-stage turn detaches at the second stage instead of the first, and the fix is unequal stages: 15° then 8°, then whatever the falling Mach number still allows.

This is why supersonic bodies get more slender as their design Mach number falls rather than rises, which is backwards from most intuitions. At Mach 5 a great deal of turning is available; at Mach 1.3 almost none is, and every surface must be nearly aligned with the flow.

What the solver computes, and how it is checked #

Everything here rests on maxDeflection, which is a search rather than a formula. A closed form for the detachment angle exists — it is a root of a quartic — and the search was preferred because the same routine then supplies the bracket boundary for the weak and strong branches, so the boundary the figures draw and the boundary the assertion refuses against are guaranteed to be the same number.

Three rejection tests guard it. Asking for a shock at 30° at Mach 2 throws. Asking assertDetaches about 10° at Mach 2 throws, because that angle attaches. And asking for a branch that is neither “weak” nor “strong” throws rather than defaulting.

The maximum is not where the flow behind goes subsonic #

Two nearby boundaries on the θ–β chart are routinely confused, and the difference is small enough to be worth stating exactly.

The detachment condition is where θ is maximised: the largest turn a shock can produce. The sonic condition is where the flow behind the shock is exactly Mach one, which happens at a slightly smaller deflection.

At Mach 2, detachment is at 22.97° and the downstream flow goes sonic at about 22.8°. Between those two angles there is a sliver of solutions on the weak branch that are attached and leave the flow subsonic — genuinely weak shocks with subsonic flow behind them, which most descriptions of the weak branch quietly deny exist.

The gap is a fraction of a degree and no practical design lives in it. It is worth knowing because it explains a common statement that is nearly true — “the weak solution leaves the flow supersonic” — and shows exactly where the “nearly” is. This site’s solver reports the Mach number behind rather than a supersonic-or-not flag, precisely so that the sliver is visible rather than rounded away.

Where the model stops #

The bow shock’s shape is not computed. The curve drawn in the figure is indicative and the caption says so. Its stand-off distance, its curvature and the sonic line behind it are the blunt-body problem, which needs a full flow solve this site does not have.

Two dimensions. A cone detaches at a larger half-angle than a wedge — about 41° against 22.97° at Mach 2 — for the reason the section above works through. Nothing here computes the conical case, and no number on this page applies to one.

No real-gas effects. The re-entry argument above is qualitatively right and quantitatively wrong with $\gamma = 1.4$, because at those temperatures the gas dissociates, $\gamma$ falls, and the shock stands closer to the body than a perfect-gas calculation predicts. That stand-off distance matters for heating, so real-gas effects are not a refinement there but a central term.

What the picture cannot show #

The detached panel is drawing an absence, which is a hard thing to draw honestly.

There is no computed content in it. The bow shock’s curve was chosen to look like a bow shock, the stand-off distance is not a solved quantity, and the subsonic pocket is not drawn at all because nothing here can locate the sonic line. The caption and the regime note both say so, and that is the best available: a figure whose subject is that no solution exists cannot show the solution.

The alternative — omitting the panel — would leave the essay asserting that something happens without showing what. The alternative that is actually forbidden is drawing it as though it were solved, and the difference between the two is entirely in what the caption claims.

Who found it, and when #

The detachment condition falls out of Meyer’s 1908 analysis and was understood as soon as the θ–β–M relation was written down, since the maximum is a feature of the algebra.

Harvey Allen and Alfred Eggers at NACA Ames produced the blunt-body argument in 1951, in a report that was classified for several years. The reasoning is a single thermodynamic observation applied to a re-entry vehicle — put the energy into the gas, not into the structure — and it inverted the design of every ballistic vehicle then being considered.

The blunt-body flow field itself resisted computation until the 1960s. Moretti and Abbett’s time-marching solution in 1966 is usually taken as the point at which it was settled, and it settled it by giving up on the steady equations and marching an unsteady solution to equilibrium instead — which sidesteps the mixed elliptic-hyperbolic difficulty entirely — the same difficulty that makes a transonic wing hard, met at the other end of the speed range.

Where the ladder goes next #

Compression has now been followed to its limit: it costs entropy, it has a maximum, and past that maximum the flow rearranges itself entirely.

Turning the other way costs nothing at all. There is no maximum, no entropy, no detachment, and no discontinuity — an expansion of any angle is free, which is the deepest asymmetry in this subject and the one that makes supersonic aerofoils computable.