What actually holds a wing up
A wing weighing three hundred tonnes stays up because the air around it is going round it slightly faster over the top than underneath. That much is not in dispute. The question is why, and the answer given in most classrooms is wrong.
The right answer is a single word — circulation — and the rest of this essay is what that word means and why nothing else will do.
The shape is not the reason
Start by killing the most durable idea in the subject: that lift comes from the shape of the section, from the curved top and the flat bottom.
It does not, and the counterexample is a cylinder. A circular cylinder has no top, no bottom, no camber and no trailing edge, and if it is spinning it lifts perfectly well.
That is the Magnus effect, and it is what makes a sliced tennis ball dip and a curveball curve. No aerofoil, no camber, plenty of lift. Whatever explains lift has to explain that too, and “the top is longer than the bottom” does not.
Nor does the shape argument survive an aeroplane flying upside down, which happens at every airshow. If the curvature of the upper surface were doing the work, inverted flight would push the aircraft into the ground.
What circulation is
Take a closed loop drawn in the air, encircling the wing. Walk round it, and at every step add up how much the air is moving along the direction of travel. The total is the circulation, written .
For a wing generating lift, that number is not zero. Air on the upper path is moving with the loop and air on the lower path is moving against it less strongly, so the sum comes out non-zero. That imbalance is the lift, by a result so tidy it looks like a coincidence:
Density times free-stream speed times circulation. Nothing about shape, nothing about the length of the upper surface, nothing about where the air ends up.
Where the faster air comes from
The circulation account raises an obvious question: if the air over the top is faster, what is speeding it up?
Nothing is. That is the wrong way round.
Circulation is not a mechanism that acts on the air; it is a description of the flow that is already there. The stream arrives, the wing sits in it at an angle, the sharp trailing edge forbids one whole family of possible flows, and what is left is a flow with net circulation round the section. The faster upper flow and the slower lower flow are two halves of one thing, and neither causes the other.
Asking what accelerates the upper air is like asking which half of a rotation is doing the turning. The honest answer is that the flow pattern is a single object, fixed by the shape, the incidence and the edge condition, and it happens to have more speed above than below.
The pressure is where the force is
Circulation gives the lift in one line, but a wing does not feel circulation. It feels pressure, on every square millimetre of its surface, and nothing else.
Those two facts have to be reconciled, and the reconciliation is the second calculation this site runs. Take the solved velocity field, apply Bernoulli along each streamline to get the pressure at the surface, multiply by the local area, resolve perpendicular to the stream, and add up all the way round. That is a force, arrived at without ever mentioning circulation.
It comes to the same number. At six degrees of incidence the circulation route gives 1.967 and the pressure route gives 1.936 — under two percent apart, and the remainder is the discretisation of the surface integral rather than a disagreement about physics.
Two routes to one answer is worth more than either route alone, and the reason is in the section below.
Why one value and not another
Here is the difficulty that makes this subject interesting rather than trivial.
The equations of ideal flow do not fix the circulation. Every value is a perfectly good solution: a wing with a lot, a wing with a little, a wing with none, all of them satisfying the equations exactly and all of them looking plausible when drawn. If the mathematics admits every answer, something outside the mathematics has to choose.
Look at the trailing edge in each panel. In the first, the flow comes round the sharp corner from below — which, at a genuinely sharp edge, means turning through a right angle at infinite speed. In the third, the same thing happens the other way. In the middle, the flow simply leaves.
Real air cannot turn a sharp corner at infinite speed. Viscosity, which the ideal theory has thrown away, will not permit it. So of all the solutions the equations allow, the physical one is the single value of circulation for which the flow leaves the trailing edge smoothly.
That is the Kutta condition, and it deserves an essay of its own. Its role here is that it converts an underdetermined problem into a determined one: given the shape and the angle, the circulation follows, and so does the lift.
The sharp edge is the point
It follows that the trailing edge is the most important part of a wing, which is not where anybody looks.
A wing with a rounded trailing edge has no Kutta condition to apply, because there is no corner the flow cannot get round. Its circulation is genuinely ambiguous and its lift is poor. The sharpness is not aerodynamic tidiness; it is the mechanism.
This also explains why damage to a trailing edge matters more than the same damage to a leading edge, and why the flaps and ailerons — which change the effective trailing edge — are such effective controls.
What lift costs
A wing does not get lift for nothing, and the accounting is worth doing because the popular version of it is also wrong.
The popular version says the wing pushes air down, and by Newton’s third law the air pushes the wing up. That is true, and it is not an alternative explanation — it is the same explanation, seen from a control volume drawn round the whole wing rather than from the surface. A flow with circulation round the section leaves the region with a net downward momentum, and the rate at which that momentum is created is exactly the lift. The two accounts have to agree, because they are the same flow.
Where the popular version goes wrong is in the picture that usually accompanies it: air striking the underside of the wing and being deflected, like water off a plank. Very little of a wing’s lift comes from the lower surface. Most of it comes from the reduced pressure above, which is why an aerofoil at a few degrees of incidence lifts far better than a flat plate at the same angle, and why the top surface is the one designers worry about.
Two dimensions, and what a real wing adds
Everything above is a section — an infinite wing, with no ends.
A real wing has ends, and that changes the accounting in a way worth knowing about now even though it belongs to a later essay. Circulation cannot simply stop at a wingtip: a theorem of Helmholtz’s says a vortex line cannot end in the fluid. So the circulation bound to the wing turns and trails downstream from each tip, and those trailing vortices are what an aircraft leaves behind it.
They are also what makes a real wing cost more than a section. The trailing vortices induce a downwash over the wing itself, tilting the effective oncoming flow slightly downward, which tilts the lift vector slightly backwards. The backwards component is induced drag, and it is a consequence purely of having ends. A very long wing has little of it, which is why gliders look the way they do.
What the picture cannot show
Worth stating plainly, since the figures here are persuasive.
They show a section, not a wing. No tips, no trailing vortices, no induced drag, no span.
They show one instant of a steady flow. A wing that has just started moving has no circulation at all; it acquires it over the first few chord lengths of travel, shedding an equal and opposite starting vortex behind it. That process is invisible here.
They show no boundary layer. The surface in these figures is a perfect streamline with the air sliding along it. Real air sticks to a surface, and the transition from stuck to free happens in a layer far too thin to draw at this scale — and that layer is where drag and stall come from.
They stop before stall. Nothing in this model prevents the incidence being raised to forty degrees and the lift going on rising. Real sections give up around fifteen.
What the solver computed
Every number in these figures came out of lib/flow.js and none was quoted from a reference.
The section is a Joukowski aerofoil, produced by mapping a circle with the transform . The circle must pass through the critical point for the trailing edge to be sharp; the Kutta condition then fixes the circulation, and Kutta–Joukowski converts it to lift.
The check that matters is that the lift is computed twice, by unrelated routes. Once from the circulation, . Once by integrating the pressure over the surface and resolving it perpendicular to the stream — a completely different calculation that never mentions circulation. The build asserts the two agree, and they do, to within 0.8% at zero incidence and 3.6% at twelve degrees.
That check was not decoration. Without it the site shipped an aerofoil whose circulation carried the wrong sign, so the section was generating lift downwards — and every other check passed. Divergence was zero. The surface was a streamline. The measured circulation agreed exactly with the Kutta value it had been handed. All true, all self-consistent, and the wing was upside down. It surfaced only when a different figure measured the flow over the upper surface as slower than the lower, which is not a wing.
A wrong flow field is beautiful. That is the whole reason this site solves rather than sketches.
Why two routes, and not one
The double calculation deserves its own paragraph, because it is the site’s method in miniature and the reason it exists.
A single calculation can be self-consistent and wrong. The circulation route computes from the Kutta condition and then reports ; if the sign convention inside that chain is reversed, every step still agrees with every other step and the answer comes out confidently backwards. Nothing in the chain can detect it, because the error is in the chain.
The pressure route shares almost nothing with it. It samples the velocity field near the surface, squares it, subtracts from the free-stream value, resolves along a normal and sums. Different quantities, different geometry, different arithmetic. For the two to agree by accident, two unrelated calculations would both have to be wrong in exactly compensating ways.
So the check is not “is the number plausible” — which no reader and no author can judge — but “do two independent routes land on the same number”. That is the difference between a figure that looks right and one that has been tested, and on this site it is the whole point of solving rather than sketching.
Where the model stops
This is inviscid theory, and it is missing something enormous: it predicts no drag at all. The lift it gives is close to right and the drag it gives is zero, which is a strange combination and is the subject of a separate essay.
There is no stall here. The lift-curve slope continues for ever in this model. Real wings stop lifting somewhere around fifteen degrees because the flow separates, and separation is a viscous phenomenon this model cannot represent.
Two dimensions only. A real wing has ends, and the circulation has to go somewhere at them — trailing vortices, and the induced drag that comes with them.
The Kutta condition is an input, not a derivation. It is a physically motivated selection rule imported from outside the equations. Deriving it properly requires the viscosity that the ideal theory discarded.
The three accounts, and how they relate
There are three explanations of lift in circulation, and readers often meet them as competitors. They are not; they are one explanation described from three distances.
The surface account. The pressure over the upper surface is lower than over the lower, and the integral of that difference is the lift. This is what the wing feels, and it is what the pressure figures here show.
The circulation account. The flow has net circulation round the section, fixed by the trailing edge, and . This is the one that generalises — to spinning cylinders, to wings of any section, to a starting vortex left behind on the runway.
The momentum account. Air leaves the region behind the wing with net downward momentum, and the rate at which that momentum is created is the lift. This is the control-volume view, and it is the one that satisfies anybody worried about Newton’s third law.
All three describe the same flow and all three give the same number. Arguing about which is “really” the cause is a category error: the flow is a single object, and these are three quantities computed from it.
What is not among them is the equal-transit story, which is not a third view of the same flow but a claim about it that is false.
Who worked it out, and when
Frederick Lanchester had the circulation idea around 1894 and could not get it published; his Aerodynamics appeared in 1907. Martin Kutta wrote the condition down in 1902, and Nikolai Zhukovsky — Joukowski, in the usual transliteration — gave the lift theorem in 1906.
It is worth noticing what they were up against. The Wright brothers flew in 1903 on the strength of measurement and persistence, not theory; the theory that explains what they did arrived afterwards. And the wrong explanation, the one about air meeting up behind, was already in circulation and has outlived every attempt to kill it.
The ladder from here
Next rungs: the Kutta condition in detail, and why a sharp edge is worth so much; the starting vortex, which is where a wing’s circulation is paid for; the lift curve, computed rather than measured; and lift with no wing at all.
Then outward: what the equal-transit-time story actually claims, and by how much it is wrong; where Bernoulli’s equation applies and where using it is an error; and the thin viscous layer that this entire essay pretended did not exist.