How much circulation is too much

A spinning cylinder lifts, and the amount is $\rho U \Gamma$ with nothing in the formula to stop it. Doubling the circulation doubles the lift; there is no coefficient that saturates, no angle that stalls, no term that turns over.

The flow, however, does not simply go round faster. At one particular circulation it changes character, and the change is visible in a feature nobody thinks to watch: the points where the air is at rest.

Where the flow stops, and why there are two of them #

With no circulation at all, the flow past a cylinder has a stagnation point dead ahead and another dead behind — the front and back of the body, at 180° and 0°. The picture is symmetric in every direction anybody could ask about, and there is no drag precisely because of it.

Adding circulation breaks the up-down symmetry. The flow over the top is now the stream plus the circulation and the flow underneath is the stream minus it, so the underside is slower, and the place where it reaches zero moves. The surface speed of a cylinder with circulation is

$$q = 2U\sin\theta + \frac{\Gamma}{2\pi a}$$

and setting it to zero gives the condition

$$\sin\theta = -\frac{\Gamma}{4\pi U a}$$

which is worth reading before it is used. The right-hand side is dimensionless. It compares the circulation with the product of the speed, the size, and $4\pi$ — and since a sine cannot exceed one in magnitude, there is a value of $\Gamma$ beyond which the equation has no solution at all.

That value is $4\pi U a$, and for the cylinder drawn here, with $a = 1$ and $U = 1$, it is 12.5664.

What the solver computed, and how it was checked #

The site’s habit is to refuse a demonstration of something it has already assumed, so the stagnation points are not placed using the formula above. They are found: the surface speed is evaluated at eight hundred points in a neighbourhood, and the minimum is taken. Only then is the answer compared with the closed form.

The sine of each measured angle agrees with the ratio in the second column to within the tolerance the build asserts, and the speed at each point is required to be below five thousandths of the free stream — otherwise the search has found a minimum that is not a stagnation point, which is exactly what would happen if the field were wrong in a way the picture concealed.

The threshold, and what is on the other side #

At $\Gamma = 4\pi U a$ exactly, the two points meet at the bottom of the cylinder, at 270°. Past it the surface speed is nowhere zero: the circulation is strong enough that even at the slowest point of the surface the flow is still moving.

The stagnation point has not vanished. It has left the body.

The detached point is found the same way — by walking down the axis of symmetry looking for a sign change in the horizontal velocity, then bisecting — and it is required to lie outside the body, which is the check that catches a root-finder that has wandered inside.

Below it, the streamline through the stagnation point closes on itself, and the fluid inside that closed curve never leaves. It is trapped, going round with the cylinder forever, in a flow that has no mechanism for letting it out. That trapped region is the most obviously unphysical thing an ideal flow produces, and it is worth dwelling on: nothing here is wrong, and the answer is nonsense.

What the pressure is doing while this happens #

The stagnation points are a feature of the velocity field, and the force is a feature of the pressure field, so it is worth putting the two side by side.

The asymmetry in that picture is the entire mechanism. Over the top the stream and the circulation add, so the flow is fast and the pressure is low; underneath they subtract, so it is slow and the pressure is high. Nothing about the body’s shape enters — the cylinder is as symmetric as an object can be — and the force comes out perpendicular to the stream rather than along it, which is the statement that there is still no drag.

As the circulation rises, the suction peak on top deepens and the high-pressure region underneath spreads and moves round towards the back. The two stagnation points are the boundaries of that region: they are where the pressure coefficient reaches its maximum of exactly 1, and watching them converge is watching the high-pressure patch shrink onto a single point at the bottom of the body.

Past the threshold there is no patch at all. Nowhere on the surface does the coefficient reach 1, because nowhere is the flow at rest, and the maximum pressure on the body is strictly less than the stagnation pressure of the oncoming stream.

Moving the number rather than reading about it #

The figures on this page are one generator called at four circulations, and the first of them carries a slider. Dragging it walks the circulation from zero to eighteen, redrawing the flow at each stop — and every frame is the same solver output as the static pictures, generated at build time rather than interpolated in the browser.

What the slider makes obvious, and a set of static frames does not, is how little warning the transition gives. The pattern deforms continuously; the two points slide together at an accelerating rate as the ratio approaches one, because the inverse sine is steep there; and then they are gone. There is no frame in which anything looks unusual, which is the honest reason this threshold is rarely mentioned: it does not announce itself in the picture unless the stagnation points are marked.

Why an aerofoil never gets there #

The same arithmetic applies to a wing, and the reason no wing meets this threshold is instructive.

For a Joukowski section the circulation is not free. It is whatever the Kutta condition demands, and that works out at $\Gamma = 4\pi a U \sin(\alpha + \beta)$ — the threshold value multiplied by the sine of the effective incidence. So a wing reaches the critical circulation only at $\alpha + \beta = 90°$, which is to say with the section broadside to the stream.

That is a striking statement of what the Kutta condition does. It does not merely pick a circulation out of infinitely many; it picks one that keeps the rear stagnation point pinned to the trailing edge, and pinning it there is what keeps the flow pattern in the well-behaved family. The cylinder has no sharp edge and therefore no such condition, so the circulation is imposed from outside — by spinning it — and can be pushed anywhere.

What actually limits a rotor #

Real Flettner rotors — cylinders spun to make lift on ships and, occasionally, aircraft — do not run into this threshold either, and the reasons are all outside the model.

The circulation a spinning cylinder actually generates is not $2\pi a$ times the surface speed. The no-slip condition drives the fluid at the surface, but that momentum has to diffuse outward through a boundary layer, and how much circulation ends up in the outer flow depends on the spin ratio, the Reynolds number, and whether the layer separates. Measured lift coefficients rise with spin ratio and then flatten out, at values well short of what the ideal formula would give.

There is also the drag, which the ideal theory says is zero and which is in practice substantial — a spinning cylinder is a bluff body with a wide wake — plus the power to keep it turning against that wake. The Magnus effect is real and usable and has been demonstrated at sea more than once; the reason it is rare is arithmetic that this model has nothing to say about.

The threshold read as a maximum lift, and the measurements that pass it #

The threshold has been turned into a design claim, and it is worth following that step because the claim is checkable and turns out to be false.

Convert the critical circulation into a lift coefficient. With $L = \rho U\Gamma$ per unit span and the diameter as the reference length,

$$C_L = \frac{\rho U\Gamma}{\tfrac12\rho U^2 (2a)} = \frac{\Gamma}{Ua},$$

so at $\Gamma = 4\pi U a$ the coefficient is exactly $4\pi = 12.57$. Prandtl proposed in 1925 that this is the maximum lift coefficient a rotating cylinder can achieve, reasoning that a circulation past the threshold detaches the stagnation point and traps a body of fluid which then rotates with the cylinder — so that the effective object is larger and slower, and no more circulation is delivered to the outer flow.

It is a good argument and it is a potential-flow argument, which is the flaw. What sets the circulation a real spinning cylinder produces is the boundary layer on it and the wake behind it, and neither of those knows anything about the topology of an inviscid streamline pattern.

Careful experiments exceed it. Measurements at high spin ratios, with the ends of the cylinder properly closed off, have reached lift coefficients well above $4\pi$ — half as much again and more. The end plates are the load-bearing detail: without them the pressure difference between the two sides drives a flow round the ends, exactly as it does at a wing’s tips, and the circulation bleeds away three-dimensionally. Most of the older measurements that appeared to confirm Prandtl’s limit were limited by that relief rather than by anything in the flow round the section.

There is a second thing the rotation does that the ideal model cannot see, and it works in the rotor’s favour. Above a spin ratio near two the vortex shedding stops — the alternating wake that gives a stationary cylinder its drag is suppressed by the rotation — so the drag falls before it rises again at higher spin. A Flettner rotor is therefore operating in a regime where the bluff-body wake it would otherwise pay for has been switched off.

So the threshold in this essay is exact and its most famous consequence is not. That is a fair description of the whole relationship between potential flow and the Magnus effect: the mechanism is right, the force law is right, and every bound derived from the streamline pattern belongs to the model rather than to the fluid.

A note on the sign, which is where the mistakes are #

Every number on this page is quoted in the aerodynamic convention, where positive circulation lifts the body upwards. The solver’s internal convention is the opposite — a positive vortex strength circulates anticlockwise, which pushes a body down in a left-to-right stream — and the figures negate on the way out.

That is not pedantry. This site shipped an aerofoil whose circulation carried the wrong sign, and every other check passed: the flow was divergence-free, the surface was a streamline, the measured circulation agreed exactly with the value it had been handed, and the lift agreed with Kutta–Joukowski. All true, all self-consistent, and the wing was flying downwards. It surfaced only when a separate figure measured the flow over the upper surface as slower than the flow underneath, which is not a wing.

The stagnation points are the cheapest available guard against a repeat. Their positions are a directional statement: for lift upwards they must move onto the underside, at angles between 180° and 360°, and a sign error puts them on top. Every table on this page would look equally plausible with all the angles reflected, and only comparing them against the direction of the force catches it.

What the picture cannot show #

The trapped region past the threshold is the clearest case on this site of a solution that is mathematically perfect and physically impossible. In a real fluid, however small the viscosity, that region would not survive: the fluid in it would be spun up, would leak, and would exchange momentum with the surroundings. Ideal flow has no mechanism for any of that, so the region persists forever in a picture that satisfies every equation it is asked to.

The figures also cannot show the surface. A stagnation point marked on a cylinder in ideal flow sits on a surface where the tangential velocity is whatever the solution says; in a real flow the tangential velocity at the surface is zero everywhere, by no-slip, and the stagnation point is instead a point where the streamline pattern in the layer just outside changes topology. The two definitions coincide in the outer flow and are not the same statement.

Where the model stops #

The threshold at $4\pi U a$ is exact within potential flow and has no counterpart in any real experiment, because the flow separates long before the circulation gets that large. It matters here for a different reason: it is a limit that emerges from a linear theory with no nonlinearity in sight.

Nothing in the equations saturates. The lift is linear in the circulation right through the transition and beyond, at 12.000 below it and 14.000 above it, and the force notices nothing at all. What changes is the topology of the streamline pattern, and topology is not something a force coefficient can report.

That is a useful general lesson about reading results from a model. A quantity can be perfectly smooth across a change that alters the whole character of the solution, and a study that plotted only the lift would have no way of knowing anything had happened.

The threshold as a ratio, and what it is a ratio of #

$4\pi U a$ is a strange-looking quantity to be a limit, and it repays being read as a comparison rather than as a formula.

Circulation has the units of a speed times a length. So does $U a$. The threshold is therefore a statement that the circulation matters relative to the product of the free stream and the size of the body — the same kind of statement as the Reynolds number makes about inertia and viscosity, and with the same consequence: what decides the behaviour is a ratio, not a magnitude.

That makes the threshold a scaling law rather than a number. A cylinder ten times larger reaches it at ten times the circulation; the same cylinder in a stream twice as fast needs twice as much. What the picture depends on is the dimensionless group $\Gamma / 4\pi U a$, and the table above is indexed by exactly that.

The customary version of the same ratio, in the literature on rotor ships, is the spin ratio — the surface speed of the cylinder divided by the free-stream speed. A cylinder whose surface moves at $V$ drives a circulation of at most $2\pi a V$, so the threshold in those terms is $V/U = 2$: the surface must be moving at twice the free stream before the stagnation points leave. Real rotors run at spin ratios between about 2 and 4, so they are past this threshold in principle — and it means nothing in practice, because the circulation they actually achieve is a fraction of the ideal value.

Who found it, and when #

The Magnus effect is named for Gustav Magnus, who investigated it in 1852 to explain the deflection of artillery shells, though Newton had described the same effect on tennis balls in 1672 and Robins had measured it on musket balls in 1742. The potential-flow analysis, including the threshold described here, belongs to the same generation as Rankine’s work in the 1860s and is standard in Lamb’s Hydrodynamics by 1879.

Anton Flettner built two rotor ships in the 1920s, one of which crossed the Atlantic. The idea has been revived several times since, most recently as a fuel-saving device on cargo vessels, and each revival rediscovers the same arithmetic: the lift is real, and so is the wake.

Where the ladder goes next #

Below this rung, lift with no wing at all establishes that circulation and not shape is the cause. Beside it, the Kutta condition is what removes the freedom that this essay exploits.

Above it, the question becomes what sets the circulation when nobody is spinning anything — which is the whole of thin-aerofoil theory, and, once a hinge is added to the section, what a flap does to it.