The sharp edge decides

Here is an uncomfortable fact about the exact theory of flight: it does not, by itself, say how much lift a wing has.

Put a section in a stream at a given angle, write down the equations of ideal flow, and solve them. There is not one solution. There is a one-parameter family of them — one for every value of the circulation — and every single one satisfies the equations exactly.

That is a problem. A theory that admits every answer has not explained anything.

What is missing #

The equations of ideal flow are, at bottom, a statement that the flow is incompressible and irrotational, plus a boundary condition saying nothing flows through the surface.

Those conditions are satisfied by adding any amount of circulation at all. Circulation is a solution of the equations with no body in it — a free vortex is a perfectly good irrotational flow everywhere except at its centre, and the centre can be hidden inside the wing. So the circulation can be dialled up or down at will and nothing in the mathematics objects.

Physically, that corresponds to a wing that might be generating any amount of lift, including negative, while sitting still at a fixed angle. Real wings do not behave that way. Something is selecting.

Look at the trailing edge #

The selection is visible, and it is entirely local.

In the first panel above, the flow comes round the trailing edge from underneath and continues onto the upper surface before turning back. In the third, the reverse. In the middle, the flow leaves the edge cleanly and the two streams rejoin downstream.

The first and third demand something impossible. A trailing edge is a corner — in the idealised section, a corner of zero radius — and getting round a corner of zero radius requires an infinite velocity. Bernoulli then requires an infinitely negative pressure to go with it.

Nothing in the ideal equations forbids that, which is why those solutions exist mathematically. Real air forbids it immediately: viscosity, which the theory has thrown away, cannot sustain an infinite velocity gradient for any length of time at all.

So of the infinite family, the physical solution is the one where the flow does not have to turn the corner: the circulation is whatever value places a stagnation point exactly at the trailing edge.

That is the Kutta condition.

It comes from outside #

Worth being clear about the logical status, because textbooks often are not.

The Kutta condition is not derived from the equations of ideal flow. It cannot be — it is a statement that certain of their solutions are unphysical, and the equations have no notion of physicality. It is imported.

What justifies it is viscosity, which the ideal theory discarded on the first line. So the structure of the argument is: throw away viscosity to make the problem solvable, discover the problem is now underdetermined, and put back one carefully chosen consequence of viscosity to determine it again.

That is a slightly disreputable manoeuvre and it works extremely well. It is also why the condition can be stated in several equivalent ways — finite velocity at the edge, a stagnation point at the edge, the two surface flows leaving with equal pressure — which look like different conditions and are the same one.

What the sharp edge is for #

An immediate consequence, and it changes how a wing looks.

If a trailing edge is rounded, there is no corner the flow cannot get round, no impossible solution to rule out, and therefore no selection. The circulation becomes genuinely ambiguous, and in practice settles somewhere unhelpful. Lift is poor and unpredictable.

Sharpness is not neatness. It is the mechanism that converts an ambiguous flow into a definite one.

This is why trailing-edge damage matters more than leading-edge damage of the same size, why control surfaces are placed at the trailing edge, and why a wing with its trailing edge deliberately blunted — as some are, for structural or acoustic reasons — pays for it in lift.

The condition in numbers #

It is worth seeing what the condition actually fixes, because “a stagnation point at the edge” sounds qualitative and is not.

For a Joukowski section the circulation that satisfies it is

$$\Gamma = -4\pi a U \sin(\alpha + \beta)$$

where $a$ is the radius of the generating circle, $U$ the free-stream speed, $\alpha$ the angle of attack, and $\beta$ a small angle set by the camber. Everything on the right is geometry and incidence. Nothing is fitted, nothing is measured, and nothing is free.

Feed that into Kutta–Joukowski and the lift follows immediately. For the section drawn here at six degrees, the circulation comes out at 1.967 and so does the lift, in the units the solver works in.

The point is not the formula but its shape: lift is proportional to the sine of the angle, which for small angles is the angle itself. That is why the lift curve is straight, and it is a prediction rather than an observation.

Why camber shifts the zero #

The $\beta$ in that formula is worth a paragraph, because it explains something every pilot knows.

A symmetric section has $\beta = 0$, so it generates no lift at zero incidence — obvious enough, by symmetry. A cambered section has $\beta > 0$, so at zero incidence the sine is not zero and the wing is still lifting.

To get no lift at all from a cambered section, it must be tilted nose-down by the angle $\beta$. That is the zero-lift angle, it is negative, and it is why a wing at what looks like level incidence is already working.

Camber, in other words, does not change the slope of the lift curve. It slides the whole line sideways. That is a much more specific claim than “curved wings lift better”, and it is testable: the computed slope for a cambered and an uncambered section of the same thickness is the same to within a percent.

What the solver computed #

The three panels are three genuine solutions, not sketches of what different circulations might look like.

lib/flow.js builds the flow past a circle with a vortex at its centre and maps it to an aerofoil with the Joukowski transform. The circulation is a free parameter of that construction. The Kutta value is $\Gamma = -4\pi a U \sin(\alpha + \beta)$, where $\beta$ is set by the camber; the outer panels use a scaled version of it and are exact solutions of the same equations with a different constant.

For the middle panel the site checks that the flow really does leave the edge smoothly, by requiring the surface velocity to be tangent everywhere within a stated neighbourhood of the trailing edge. It holds to three parts in a hundred thousand over the whole surface, with a small region round the cusp itself excluded because the mapping is singular there and the residual is conditioning rather than physics.

The sign of that circulation is the site’s most instructive bug. Taking the formula with the textbook’s positive sign, in the convention this solver uses, produced an aerofoil generating lift downwards — and it passed every check it had. Mass was conserved, the surface was a streamline, and the measured circulation agreed exactly with the Kutta value it had been given. Everything was self-consistent and the wing was upside down.

It surfaced only when a different figure measured the flow over the upper surface as forty percent slower than the lower, which is not a wing. There is now a check that computes the lift twice by unrelated routes and refuses a sign disagreement.

What the unphysical solutions look like #

The outer panels of the first figure are worth returning to, because they are not sketches of a mistake — they are exact solutions, and looking at them carefully is the fastest way to see what the condition is doing.

In the under-circulated panel, the rear stagnation point is not at the trailing edge; it is on the upper surface, some way forward of it. The flow from below therefore has to run past the edge, round the corner, and forwards along the top to reach it. That corner is where the velocity would be infinite.

In the over-circulated panel the stagnation point has moved onto the lower surface, and the flow from above has to make the same impossible turn in the other direction.

Between them, as the circulation is varied continuously, the stagnation point slides from one surface to the other and passes through the trailing edge exactly once. That crossing is the Kutta value. There is precisely one, which is why the condition determines the lift rather than merely constraining it.

How the wing acquires its circulation #

A question the steady picture cannot answer: if the wing has circulation and the air did not, where did it come from?

Circulation is conserved in an inviscid fluid — Kelvin’s theorem — so the total round any large loop enclosing both the wing and the air it has passed through must still be zero. If the wing has acquired some, an equal and opposite amount must have been left behind.

It has. When a wing starts moving, the flow initially does whip round the trailing edge, exactly as the unphysical panels show. That state lasts a fraction of a second: the sharp edge sheds the offending vorticity as a starting vortex, which floats away downstream, and the wing is left with the equal and opposite circulation bound to it.

So the Kutta condition is not merely a selection rule; it describes the end state of a real process that takes a few chord lengths of travel to complete. An aircraft accelerating down a runway is leaving its starting vortex at the point where it began to move.

Three statements, one condition #

The condition turns up in several forms and they can look like different requirements. They are not.

Finite velocity at the trailing edge. The original statement, and the one that connects most directly to why unphysical solutions are unphysical.

A stagnation point at the trailing edge. What that finite velocity has to be, for a genuinely sharp edge: zero. This is the version easiest to see in a picture.

Equal pressure on the two surfaces at the edge. The two streams arrive from opposite sides and leave together, so they must arrive at the same pressure; if they did not, there would be a jump across a surface with nothing in it.

For a cusped trailing edge — where the two surfaces meet tangentially, as in the Joukowski sections here — all three coincide exactly. For a finite-angle edge they differ slightly, and the third is the one that generalises best. A reader meeting a form they do not recognise is almost always meeting one of these three.

What it does not fix #

The condition determines the circulation, and therefore the lift. It does not determine several things a designer needs.

It says nothing about drag, which in this model is zero regardless.

It says nothing about the pressure distribution’s shape, only its integrated effect. Two sections with the same lift can have very different distributions, and the difference is what decides whether one of them stalls gently or abruptly.

It says nothing about when it stops applying. The condition is satisfiable up to some incidence and then it is not, and the model contains no hint of where that is. Finding out requires the viscous flow.

It says nothing about the boundary layer that justifies it. The argument is circular in a benign way — viscosity is invoked to rule out solutions and then dropped again — and closing that circle properly is a substantial piece of twentieth-century work.

What it buys, in the end #

The condition looks like a technicality and it is the reason aerodynamics is a predictive science rather than a catalogue.

Without it, the theory says a wing might have any lift at all. With it, the lift follows from the shape and the angle by a chain with no free parameters: geometry gives $a$ and $\beta$, incidence gives $\alpha$, the condition gives $\Gamma$, and Kutta–Joukowski gives the force.

That chain is what makes a lift curve computable in advance rather than measurable afterwards, and it is what let aircraft design stop being a matter of building something and finding out.

The one thing it cannot do is tell anybody when it stops applying, and that limit is viscous.

Where the model stops #

It is imported, not derived. A first-principles account needs the viscous flow near the edge.

It assumes the flow is attached. Above the stall the flow has separated from the upper surface, there is no clean departure from the trailing edge to enforce, and the condition simply stops applying — which is why this model has no stall in it.

It assumes steady flow. During the first moments of motion, and during rapid manoeuvres, the circulation is still changing and the condition holds only approximately.

A truly sharp edge is an idealisation. Real edges have a small radius, and the condition is an excellent approximation rather than an exact statement.

Who found it, and when #

Martin Wilhelm Kutta stated the condition in 1902, in work on the flight of gliders — Otto Lilienthal’s, whose death in 1896 had made the problem urgent. Nikolai Zhukovsky arrived at essentially the same idea independently, and the lift theorem that carries his name followed in 1906.

Both were working on a problem that already had a practical solution: the Wright brothers flew in 1903 without any of it. The theory is not what made flight possible; it is what made flight designable, which is why the aeroplanes of 1910 look like guesses and those of 1935 do not.

Sharpness, and how sharp is sharp #

A practical question the idealisation dodges: real trailing edges have a finite radius, so how sharp must one be for the condition to hold?

The relevant comparison is with the boundary layer’s thickness at that point. If the edge radius is much smaller than the layer, the flow cannot distinguish it from a true cusp and the condition applies cleanly. If the radius is comparable to or larger than the layer, the flow can negotiate the corner and the circulation becomes ambiguous again.

For an airliner wing the layer at the trailing edge is a few centimetres and the edge radius is under a millimetre, so the condition holds with room to spare. For a model aircraft, where the layer is proportionally much thicker, a trailing edge that looks sharp to the eye may not be sharp to the flow, and lift suffers accordingly.

This is one of the ways model aerodynamics differs from full-size aerodynamics for reasons that have nothing to do with the model being small and everything to do with the Reynolds number being different.

The ladder from here #

Next rungs: the starting vortex and Kelvin’s theorem, worked properly; the lift curve that this condition generates, and its slope; thin-aerofoil theory, where the whole construction is carried out for a general camber line; and the stall, which is the moment the condition can no longer be satisfied.

Then across to the thin viscous layer that justifies the condition in the first place, and which the ideal theory had to discard to get started.