Fluid Flow & Aerodynamics
why things fly, and why the usual explanation is wrong
What is taught wrongly

The story about air meeting up again

The most repeated explanation of lift says that air parting at the nose must rejoin at the tail, so the longer upper path forces a higher speed. The premise is false, and the speed it predicts is wrong by a factor of twenty.
15 min read 7 figures Taught wrongly, everywhereLift is circulationA smooth picture proves nothing

Here is the explanation, stated as fairly as it can be.

A wing has a curved upper surface and a flatter lower one, so the path over the top is longer than the path underneath. Air arriving at the leading edge splits, half going over and half going under. Since the two halves must meet again at the trailing edge, the air going over has further to travel in the same time, so it must go faster. Faster air has lower pressure, by Bernoulli. Lower pressure above than below is lift.

It is a good story. It is in a great many textbooks, on a great many museum panels, and in the briefing every private pilot gets. Every step of it after the second is fine.

The second step is false.

Two parcels released together do not arrive togetherThe most repeated explanation of lift says that air parting at the leading edge must meet again at the trailing edge, so the longer upper path forces a higher speed. Released into the solved field, the upper parcel arrives long before the lower one — the premise is simply false, and the real speed difference is larger than it would require.the upper surface is 1.7% longerso equal transit needs the flow over it 1.7% fasterit is actually 76.5% fasterthe premise is false and the number it predicts is wrongsurface lengths and speeds measured on the solved field6° incidence
Fig. 1 Two parcels released together upstream, carried by the solved velocity field, and tracked past the section. They do not arrive together and nothing requires them to.

Nothing makes them meet

There is no conservation law about parcels of air rejoining their former neighbours. Air is not a sheet being cut and rejoined; it is a continuum, and two bits of it that happen to be adjacent upstream have no obligation to be adjacent again downstream.

Ask what would enforce it and there is no answer. Mass conservation does not: it constrains the total flux through any surface, not the identity of the parcels crossing it. Momentum does not. The boundary conditions do not.

The claim is simply invented, and the reason it survives is that it sounds like a conservation law without being one.

What the flow actually does

Released together and tracked, the upper parcel arrives at the trailing edge substantially before the lower one. On a lifting section it is not close.

That is the opposite of the story’s implicit picture — the story imagines the upper parcel struggling to keep up over its longer path — and it is worse for the story than mere inequality would be. The upper air is not just fast enough to keep up. It is far faster than keeping up would require.

The two ratios the claim equates

Transit time turns out to be an awkward quantity to measure, and it is worth explaining why, because the reason is instructive.

A parcel released exactly on the stagnation streamline never arrives at all: it slows to a stop at the nose. A parcel released a hair off it takes a very long time to get past that region, and the closer to the streamline it starts, the longer it takes. So the transit time depends on an arbitrary choice of release point and can be made almost anything.

What the claim actually asserts is a proportion, and both sides of it are well defined without choosing any release point:

speed over the topspeed underneath=length of the upper surfacelength of the lower surface\frac{\text{speed over the top}}{\text{speed underneath}} = \frac{\text{length of the upper surface}}{\text{length of the lower surface}}

Both are properties of the solved flow and the geometry. Measure them.

The measurement

For the cambered section drawn here:
the upper surface is longer by 1.7%
so equal transit needs the upper flow faster by 1.7%
the upper flow is actually faster by 39% at 2°, 77% at 6°, 107% at 10°

The claim is not slightly off. At six degrees of incidence it predicts a speed difference more than forty times smaller than the one the flow actually has.

And the direction of the error matters. If equal transit time were the mechanism, the lift would be proportional to how much longer the upper surface is — so a thin, barely cambered wing would barely lift, and a symmetric one would not lift at all at any angle.

Both consequences are false. A flat plate lifts perfectly well at incidence, and a symmetric section is what most aerobatic aircraft use precisely so that it works the same way upside down.

The consequence nobody checks

If the story were right, it would make a prediction that can be tested in a sentence.

The upper surface of a typical section is one or two percent longer than the lower. A speed difference of one or two percent produces, through Bernoulli, a pressure difference of roughly twice that — three or four percent of the dynamic pressure. Integrate over the wing and the lift is a few percent of 12ρU2S\tfrac12 \rho U^2 S; that is, a lift coefficient of a few hundredths.

A real wing in cruise runs at a lift coefficient around 0.5, and near the stall above 1.5. The story under-predicts lift by a factor of twenty or more, which is enough that an aircraft designed on it would not leave the ground.

Nobody notices, because nobody carries the story as far as a number. It is offered as a qualitative account and accepted as one, and the arithmetic that would kill it in a line is never done.

Flat plates, and the end of the argument

The cleanest refutation needs no computation at all.

A flat plate has an upper surface and a lower surface of exactly the same length. Equal transit time therefore predicts precisely zero lift from a flat plate at any angle whatsoever.

Flat plates lift. A sheet of plywood held at an angle out of a car window lifts hard enough to be difficult to hold. Balsa gliders, kites, sails, paper aeroplanes and the tail surfaces of many aircraft are flat or very nearly so, and all of them work.

That single observation disposes of the mechanism completely, and it does not require anybody to solve anything.

The Kutta condition picks the circulationIdeal flow round an aerofoil admits any circulation at all, and each gives a different lift. Only one value lets the flow leave the sharp trailing edge without turning a corner at infinite speed, and that is the one nature selects.too little — the flow whips round the edgeΓ = 0.93the Kutta value — it leaves smoothlyΓ = 2.67too much — the rear stagnation point is on topΓ = 4.81ideal flow — three admissible solutions, one physical8° incidence
Fig. 2 The mechanism that does explain a flat plate, and every other section: the sharp trailing edge selects one circulation out of infinitely many, and the selection depends on the angle rather than on the shape of the surfaces.

What actually sets the speed

The speed difference comes from the circulation, which is fixed by the requirement that the flow leave the trailing edge smoothly. The upper and lower speeds are two halves of one flow pattern, and neither is caused by the geometry of the paths.

That also explains why the ratio grows with incidence while the path lengths do not change at all. Tilt the section and the surface lengths are identical — it is the same shape — but the circulation rises with sin⁡α\sin\alpha and so does the speed difference. A theory based on path length has nothing to say about angle of attack, which is the single most important variable there is.

A Joukowski aerofoil at 2°A cambered aerofoil in a uniform stream. The circulation is not chosen: it is whatever value makes the flow leave the sharp trailing edge smoothly, and that single condition fixes the lift.Γ = 1.487C_L = 0.737ideal flow with the Kutta condition applied2° incidence
Fig. 3 Two degrees. The suction over the upper surface is already far larger than a 1.7% length difference could account for.
A Joukowski aerofoil at 10°A cambered aerofoil in a uniform stream. The circulation is not chosen: it is whatever value makes the flow leave the sharp trailing edge smoothly, and that single condition fixes the lift.Γ = 3.390C_L = 1.681ideal flow with the Kutta condition applied10° incidence
Fig. 4 Ten degrees, the same section with the same surface lengths, and much more lift. Nothing about the path lengths changed.

What Bernoulli is doing here

The story’s third step — faster air has lower pressure — is correct, and it is worth saying why, because the refutation above should not leave the impression that everything in the account is suspect.

Along a streamline in steady, inviscid, incompressible flow,

p+12ρU2=constantp + \tfrac12 \rho U^2 = \text{constant}

so where the speed is higher the pressure is lower. Over the upper surface of a lifting section the speed genuinely is higher and the pressure genuinely is lower, and integrating that pressure difference over the section genuinely does give the lift.

Every one of those steps survives. What does not survive is the reason offered for the speed difference. The story has the right physics attached to a fabricated premise, which is exactly the combination that makes a misconception durable — anybody who checks the conclusion finds it correct.

There is a separate and very common error in the use of Bernoulli, which is applying it between points that are not on the same streamline, or across regions where the hypotheses fail. That is worth its own essay.

Where the story came from

The account appears to have entered aviation teaching in the 1920s and 1930s, in textbooks aimed at pilots rather than engineers, and it has been remarkably durable since.

Its persistence is not really a failure of physics education so much as a demonstration of what makes an explanation spread: it is short, it is visual, it requires no mathematics, it reaches a correct conclusion, and the reader has no obvious way to test it. The correct account requires a line integral and a boundary condition at a singularity, neither of which fits on a museum panel.

Attempts to displace it have been made for decades. NASA maintains a page on it. It remains in circulation, and it will probably outlive this site.

What the solver computed

The parcel tracks are integrated through the exact velocity field, and the surface quantities are measured on it.

Getting this figure right took three attempts, all of which are instructive about measuring the wrong thing.

The first released parcels at fixed offsets from the leading-edge point and reported the upper one taking twice as long — an artefact of releasing it deep in the slow region near the nose. The second released both a hair either side of the stagnation streamline upstream and had them arriving within a tenth of a percent of each other, which would have made equal transit correct; that was an artefact too, since the shared crawl past the stagnation point swamped everything that happened over the body.

The third abandoned transit time as the measurement and compared the two ratios above, which depend on no arbitrary choice. Those are the numbers in the table.

The build asserts that the ratios differ by more than two percent. If they ever agreed, the figure would be making a false claim about a true explanation, and it refuses to render rather than do that.

Why it persists

Three reasons, and they are worth separating.

It gets the answer right. The upper flow is faster and the pressure is lower. A story that reaches a true conclusion by a false route is much harder to dislodge than one that is simply wrong.

It is easy to draw. Two arrows, one longer path, one shorter. The correct account requires circulation, which is a line integral, and the Kutta condition, which is a statement about a singularity.

It flatters the reader. It offers a complete mechanical picture in four sentences, and the honest answer — that the flow pattern is a single object fixed by a boundary condition at a sharp edge — sounds evasive by comparison.

What to say instead

The shortest honest replacement is not much longer.

Air flowing round a wing at an angle has net circulation round it, because the sharp trailing edge forbids the flow from whipping round the back. Circulation means faster flow over the top and slower underneath, which by Bernoulli means lower pressure above than below. The pressure difference, integrated over the wing, is the lift.

Four sentences, no false premise, and it survives every case the old story fails: symmetric sections, flat plates, inverted flight, and spinning cylinders with no wing at all.

The other version, with a fan

There is a second form of the story that deserves the same treatment, because it is offered as the sophisticated correction to the first.

It says: the parcels do not meet, granted, but the upper one still goes faster because the wing acts like a constriction — the air over the top is squeezed between the wing and the undisturbed air above, so by continuity it must accelerate, as it would through a nozzle.

This is better, in that it appeals to a real conservation law. It is still wrong, for a specific reason: there is no upper wall. The “undisturbed air above” is not a boundary and does not constrain anything; the streamtube over the wing is free to expand upwards, and whether it does is exactly what the solution has to determine rather than assume.

In fact the streamlines above a lifting wing are displaced upwards, over a distance of several chords, which is the opposite of a constriction. The acceleration is not geometric. It comes from the circulation, as everything else does.

How to test an explanation

The general lesson is worth stating, because the subject is unusually full of plausible accounts.

An explanation of lift has to survive four cases, and they are easy to remember: a symmetric section, which has no longer surface; a flat plate, which has two identical surfaces; inverted flight, which turns the curvature the wrong way; and a spinning cylinder, which has no wing at all.

Equal transit time fails all four. The constriction story fails the last two. The circulation account handles every one, because none of them changes the fact that the flow has net circulation round the body.

Lift from a spinning cylinderA circular cylinder with circulation round it. There is no aerofoil section, no camber and no sharp trailing edge, and it lifts — which rules out shape as the explanation and leaves circulation as the thing that matters.lift 3.393= ρUΓ = 3.400drag 1.5e-16 — still zeroideal flow — no shape, only circulationΓ = -3.4
Fig. 5 The fourth case. No section, no camber, no trailing edge, no surfaces of any length — and lift, in exactly the amount the circulation predicts.

What a solved flow settles

Every claim in this essay was checked against a computed field rather than argued from a diagram, and it is worth being explicit about which claims those were.

The surface lengths are measured on the section’s own outline. The surface speeds are sampled from the solved velocity field just outside the body. The parcel tracks are integrated through that same field. None of it is drawn by hand, and the build refuses to render the figure if the two ratios ever come out equal.

That matters because this particular argument has been had for a century in words, and words are where it stalls: the story is plausible, its conclusion is right, and disputing it without numbers sounds like pedantry. With numbers it takes one line — 1.7% against 39% — and there is nothing left to discuss.

It is the same discipline that produced the exact theory’s zero drag as a computed 10−1610^{-16} rather than a quoted result, and that caught a circulation whose sign was reversed while every other check passed.

A Joukowski aerofoil at 6°A cambered aerofoil in a uniform stream. The circulation is not chosen: it is whatever value makes the flow leave the sharp trailing edge smoothly, and that single condition fixes the lift.circulation round this loop: -2.445Γ = 2.445C_L = 1.212ideal flow with the Kutta condition applied6° incidence
Fig. 6 The quantity that does explain the speed difference, measured round a loop enclosing the section. It grows with incidence; the surface lengths do not change at all.
The Kutta condition picks the circulationIdeal flow round an aerofoil admits any circulation at all, and each gives a different lift. Only one value lets the flow leave the sharp trailing edge without turning a corner at infinite speed, and that is the one nature selects.too little — the flow whips round the edgeΓ = 0.93the Kutta value — it leaves smoothlyΓ = 2.67too much — the rear stagnation point is on topΓ = 4.81ideal flow — three admissible solutions, one physical8° incidence
Fig. 7 And what fixes that quantity. The trailing edge admits one circulation out of infinitely many, which is a mechanism the path-length story has no equivalent of.

Where the model stops

This is an inviscid solution. The speeds are computed just off the surface; inside the real boundary layer the velocity falls to zero at the wall, and the “surface speed” here means the speed just outside that layer.

The section is a Joukowski aerofoil, not a specific aircraft’s. The percentages depend on the section; the conclusion — that the two ratios are nowhere near equal — does not.

It refutes one claim, not every popular claim. The momentum account, that a wing pushes air down, is correct and is not what is being refuted here.

A note on refuting things fairly

Two habits are worth keeping when a site sets out to say that a widely taught explanation is wrong.

State it in its strongest form. The version at the top of this essay is the version a careful teacher gives, not a straw one. It is worth checking that the account being refuted is one somebody would defend.

Refute it by measurement, not by assertion. Saying equal transit time is wrong is not an argument; producing the two numbers it equates and showing they differ by a factor of twenty is. The rule this site follows is that a refutation must be a computation, and if the computation cannot be done then the claim stays out.

Both habits cost effort and both are what separate this from a list of grievances about textbooks. They are also why the momentum account is treated as correct here rather than lumped in with the rest: it survives the same tests.

The ladder from here

Nearby: where Bernoulli’s equation genuinely applies, which is the other half of the standard confusion; the momentum account of lift, done properly; and the history of how this explanation entered the textbooks.

Then across to what does hold a wing up, and to the sharp edge that decides it.

Where this goes next

Γ = 2.445C_L = 1.212ideal flow with the Kutta condition applied6° incidence What actually holds a wing up Not the shape, and not the story about air meeting up again behind. A wing lifts because there is circulation round it, and the sharp trailing edge is what decides how much. Bernoulli holds…along one streamline, steady, inviscid, incompressiblethe theorembetween two different streamlinesneeds irrotational flow as wellthrough a fan, pump or propellerwork is being done on the fluidinside a boundary layerviscosity is the whole story thereacross a shockentropy rises; total pressure does not survivethe hypotheses, not the algebra, are what fail Where Bernoulli's equation applies The equation is right. Its hypotheses are strict, and almost all misuse is a correct formula carried somewhere it does not hold — across streamlines, through a fan, or into the one layer where friction is the whole story. too little — the flow whips round the edgeΓ = 0.93the Kutta value — it leaves smoothlyΓ = 2.67too much — the rear stagnation point is on topΓ = 4.81ideal flow — three admissible solutions, one physical8° incidence The sharp edge decides Ideal flow round a wing admits infinitely many solutions, each with a different lift, and all of them exact. One extra requirement — that the air leaves the trailing edge instead of whipping round it — picks a single one.