Flow Lab

Syllabus / Module 11: Fluid Dynamics: Navier–Stokes and Beyond

Lesson 11.6

Vorticity Dynamics & Circulation

The vorticity equation, vortex stretching, Kelvin's circulation theorem, Biot–Savart, vortex methods and vorticity confinement.

Written byClaude Opus

🌊
Why this matters for simulation: Wingtip vortices and vortex confinement in Fedkiw–Stam–Jensen smoke.

By the end of this lesson you can…

  • Compute circulation as a loop integral and as vorticity flux, and check numerically that they agree (Stokes).
  • Derive the vorticity equation from the curl of Navier–Stokes, and explain why pressure drops out and why 2D flow cannot stretch vortices.
  • Show with the Lamb–Oseen vortex that viscosity spreads a core like $\sqrt{4\nu t}$ but keeps its circulation.
  • State Kelvin's and Helmholtz's theorems, sketch Kelvin's derivation, and explain the starting and trailing vortices of a wing.
  • Integrate point vortices with RK4, verify the pair speed $\Gamma/(2\pi d)$, and measure the drift of the invariants.
  • Estimate a light aircraft's bound circulation ($18.75$ m$^2$/s) and its wake's sink rate ($0.35$ m/s).
On this page
·
  1. Vorticity and circulation
  2. The vorticity equation
  3. Vorticity diffuses: the Lamb–Oseen vortex
  4. Kelvin’s circulation theorem
  5. Biot–Savart and point vortices
  6. Where a wing’s circulation comes from
  7. Computing it
  8. Concept checks
  9. Problems
  10. Key Takeaways

Last lesson ended with a claim: at high Reynolds number viscosity lives in thin layers, and what they carry and spread is vorticity. Vorticity is also the answer to an aviator’s first question. In 11.3 Momentum: Pressure & the Euler Equations a cylinder had zero lift until we added a circulation Γ\Gamma by hand; then the lift was ρUΓ\rho U\Gamma. Where does a wing get circulation, from air that started at rest? The starting vortex, Kelvin’s theorem and the trailing vortices are one bookkeeping rule.

Tools: 2.8 Curl & Vorticity, 2.9 Line Integrals, Circulation & Flux, 2.10 Green, Stokes & the Divergence Theorem, 11.4 Viscosity & the Navier–Stokes Equations and 4.3 Runge–Kutta Methods & Butcher Tableaux.

▶ Video Pressure drops out of the curl of Navier–Stokes, a core spreading with its circulation fixed, Kelvin's material loop, point vortices leapfrogging, and the starting vortex a wing leaves on the runwaycaptions only
Transcript
  1. A wing sits on the runway in still air, so a big loop of air around it has zero circulation. In flight, lift is rho U times circulation. Where does that come from? By Kelvin the loop must still read zero, so as the wing accelerates, an opposite vortex peels off the trailing edge and stays behind: the starting vortex.
  2. Circulation is the loop integral of velocity, and by Stokes it equals the vorticity flux inside. A Rankine vortex has a solid core of radius one half metre and circulation ten: vorticity twelve point seven three two four, rim speed three point one eight three. A small circle in the core gives two point five. A square of side two around it gives ten point oh one oh six, ten point oh oh oh one, ten point oh oh oh oh oh one, as points per side grow from ten to a thousand. A square beside the core gives zero: no vorticity enclosed.
  3. Take the curl of Navier-Stokes. Time gives omega t. Pressure gives p y x minus p x y: zero, so pressure drops out. Viscosity becomes nu Laplacian omega, and advection leaves a divergence of u, also zero. In two dimensions vorticity is simply carried and diffused. In three, omega dot grad u stretches and tilts vortex lines, like a skater pulling in her arms. Two dimensions cannot stretch.
  4. Without a flow, vorticity obeys the heat equation, and its point-source solution is the Lamb-Oseen vortex. For a light-aircraft wingtip vortex, circulation eighteen point seven five in air, the core spreads from sixty-six point four to two hundred ten millimetres between one minute and ten, and the peak swirl falls from thirty-two point two to ten point two metres per second. But the circulation stays eighteen point seven five. Viscosity moves vorticity; it never destroys it.
  5. Kelvin: in inviscid flow every material loop keeps its circulation. And vorticity sets the velocity through Biot-Savart: each point vortex swirls Gamma over two pi r around itself. So N vortices are two N ODEs for RK4. A pair of opposite sign translates at Gamma over two pi d, zero point one five nine one five five. Two of the same sign waltz, with period nineteen point seven three nine two.
  6. Pairs one behind the other leapfrog. The faster inner pair shoots ahead, and the squeezed outer pair speeds up and slips through. The swaps come at two point eight four, five point six six, eight point four eight: every two point eight three time units, the group drifting at zero point two one one one.
  7. Point vortices are Hamiltonian. RK4 keeps circulation and impulse to round-off, since they are linear. But H is not, and RK4 is not symplectic, so H drifts: six point three seven times ten to the minus five at step zero point two, two point one two times ten to the minus six at zero point one. A factor of thirty.
  8. Now the aircraft. Bound circulation is half U c C L: eighteen point seven five square metres per second, lifting twelve thousand six hundred thirty-three newtons over eleven metres of span. The starting vortex waits on the runway, and by Helmholtz the bound vortex trails off each tip as a pair eight point six three nine metres apart, sinking at Gamma over two pi b prime: zero point three four five four metres per second. A heavy jet: six hundred eighty, sinking two point three oh.
  9. Away from boundary layers and the wake, vorticity is zero. Then velocity is a gradient, phi obeys Laplace's equation, and that is linear: flows can simply be added. Next lesson: potential flow, a cylinder, our vortex, and a wing from one complex function.

Vorticity and circulation

Vorticity is ω=∇×u\boldsymbol\omega = \nabla\times\mathbf u, in 2D the scalar ω=vx−uy\omega = v_x - u_y: twice the spin rate of a tiny paddle wheel (11.1). The circulation around a closed curve CC is the line integral of the velocity along it,

Γ=∮Cu⋅dl=∬Sω⋅n dA,\Gamma = \oint_C \mathbf u\cdot d\mathbf l = \iint_S \boldsymbol\omega\cdot\mathbf n\,dA ,

and the second equality is Stokes’ theorem: circulation is the vorticity flux through any surface the loop bounds, measured from the boundary alone.

A Rankine vortex has a solid-body core of radius a=0.5a = 0.5 m, circulation Γ=10\Gamma = 10 m2^2/s, and an irrotational swirl outside:

uθ=Γr2πa2  (r<a),uθ=Γ2πr  (r≥a).u_\theta = \frac{\Gamma r}{2\pi a^2}\ \ (r \lt a), \qquad u_\theta = \frac{\Gamma}{2\pi r}\ \ (r \ge a) .

Inside, ω=(1/r) d(ruθ)/dr=Γ/(πa2)=12.7324\omega = (1/r)\,d(r u_\theta)/dr = \Gamma/(\pi a^2) = 12.7324 s−1^{-1}; outside, ruθr u_\theta is constant, so ω=0\omega = 0. The rim spins fastest, 3.1833.183 m/s. The lab integrates:

  • A circle of radius 0.250.25 inside the core: Γ=2.500000=ω⋅π(0.25)2\Gamma = 2.500000 = \omega\cdot\pi(0.25)^2, the enclosed flux.
  • A square of side 22 around the core, midpoint rule with NN points per side: 10.010610.0106, 10.000110.0001, 10.00000110.000001 for N=10,100,1000N = 10, 100, 1000, second-order convergence to Γ\Gamma. The loop’s shape does not matter, only the vorticity inside.
  • A unit square beside the core, centred at (2,0)(2, 0): −4.8×10−6-4.8\times10^{-6}, zero. The air there moves in circles, yet encloses no vorticity.

The vorticity equation

Take the curl of Navier–Stokes. In 2D, the components are

ut+uux+vuy=−1ρpx+ν∇2u,vt+uvx+vvy=−1ρpy+ν∇2v,u_t + uu_x + vu_y = -\tfrac1\rho p_x + \nu\nabla^2u, \qquad v_t + uv_x + vv_y = -\tfrac1\rho p_y + \nu\nabla^2v ,

differentiate the second in xx, the first in yy, and subtract. Term by term:

  • Time: ∂t(vx−uy)=ωt\partial_t(v_x - u_y) = \omega_t.
  • Pressure: −1ρ(pyx−pxy)=0-\tfrac1\rho(p_{yx} - p_{xy}) = 0: mixed partials commute, ∇×∇p=0\nabla\times\nabla p = 0 (2.11 The Laplacian & Vector Identities). A gradient pushes a parcel without twisting it.
  • Viscosity: derivatives commute with ∇2\nabla^2, so it becomes ν∇2ω\nu\nabla^2\omega.
  • Advection: the product rule gives uωx+vωyu\omega_x + v\omega_y plus (ux+vy) ω(u_x + v_y)\,\omega, and the last bracket is ∇⋅u=0\nabla\cdot\mathbf u = 0.

What survives:

DωDt=ωt+uωx+vωy=ν∇2ω.\frac{D\omega}{Dt} = \omega_t + u\omega_x + v\omega_y = \nu\nabla^2\omega .

Vorticity is carried and diffused like heat, nothing more; with ν=0\nu = 0 each parcel keeps its own, so the largest ∣ω∣|\omega| can never grow.

In 3D the same steps, with two vector identities (stated: (u⋅∇)u=∇12∣u∣2−u×ω(\mathbf u\cdot\nabla)\mathbf u = \nabla\tfrac12|\mathbf u|^2 - \mathbf u\times\boldsymbol\omega and the curl of u×ω\mathbf u\times\boldsymbol\omega), give

DωDt=(ω⋅∇)u+ν∇2ω.\frac{D\boldsymbol\omega}{Dt} = (\boldsymbol\omega\cdot\nabla)\mathbf u + \nu\nabla^2\boldsymbol\omega .

The new term, the change of velocity along the vortex line, does two things:

  • Stretching. In the straining flow u=(−αx/2,−αy/2,αz)\mathbf u = (-\alpha x/2, -\alpha y/2, \alpha z) a vortex tube along zz gets (ω⋅∇)u=ωz ∂u/∂z=(0,0,αωz)(\boldsymbol\omega\cdot\nabla)\mathbf u = \omega_z\,\partial\mathbf u/\partial z = (0, 0, \alpha\omega_z), so ωz=ω0eαt\omega_z = \omega_0 e^{\alpha t}. The tube lengthens like eαte^{\alpha t}, its cross-section shrinks like e−αte^{-\alpha t}, and its circulation ωA\omega A stays fixed: a skater pulling in her arms spins faster.
  • Tilting. Where the vertical speed ww varies with xx, ωx ∂w/∂x\omega_x\,\partial w/\partial x feeds ωz\omega_z: vorticity is turned to point another way.

2D has neither. There ω=(0,0,ω)\boldsymbol\omega = (0, 0, \omega) and nothing depends on zz, so (ω⋅∇)u=ω ∂u/∂z=0(\boldsymbol\omega\cdot\nabla)\mathbf u = \omega\,\partial\mathbf u/\partial z = 0. Stretching drives the 3D energy cascade of 11.9; no 2D simulation can do it.

Vorticity diffuses: the Lamb–Oseen vortex

With no flow carrying it, ωt=ν∇2ω\omega_t = \nu\nabla^2\omega is the heat equation: 11.4’s momentum diffusion, as vorticity. Its axisymmetric point-source solution is the Lamb–Oseen vortex:

ω=Γ4πνt e−r2/(4νt),uθ=Γ2πr(1−e−r2/(4νt)).\omega = \frac{\Gamma}{4\pi\nu t}\,e^{-r^2/(4\nu t)}, \qquad u_\theta = \frac{\Gamma}{2\pi r}\Big(1 - e^{-r^2/(4\nu t)}\Big) .

The circulation inside radius rr is 2πruθ=Γ(1−e−r2/(4νt))2\pi r u_\theta = \Gamma\big(1 - e^{-r^2/(4\nu t)}\big), which climbs to Γ\Gamma as r→∞r \to \infty at every time: viscosity moves vorticity outward without destroying it. The core spreads like 4νt\sqrt{4\nu t}; the swirl peaks where (1−e−x2)/x(1 - e^{-x^2})/x is largest, x=1.1209x = 1.1209, so the core radius is rc=1.12094νtr_c = 1.1209\sqrt{4\nu t} and the peak swirl 0.63817 Γ/(2π4νt)0.63817\,\Gamma/(2\pi\sqrt{4\nu t}).

Take a light-aircraft wingtip vortex, Γ=18.75\Gamma = 18.75 m2^2/s (derived below), in air, ν=1.4607×10−5\nu = 1.4607\times10^{-5} m2^2/s. The lab checks the PDE by central differences at r=5r = 5 cm, t=60t = 60 s: ωt=−3.98940=ν∇2ω\omega_t = -3.98940 = \nu\nabla^2\omega. Integrating ω\omega gives 18.7500018.75000 at 6060 s and 600600 s, while the core grows from 66.466.4 to 210210 mm and the peak swirl falls from 32.232.2 to 10.210.2 m/s. Molecular viscosity alone would leave a wake vortex swirling at 3232 m/s after a minute; real ones decay far faster, through turbulence and pair instabilities (from memory).

◆ Interactive Lamb–Oseen vortex: the core spreads like sqrt(4 nu t) while the circulation inside a large radius never changes; finite volumes against the exact solution

Loading simulation…

The sim runs 300300 finite volumes from 4νt=0.05\sqrt{4\nu t} = 0.05 m (t=42.8t = 42.8 s) to 0.50.5 m: 71.371.3 minutes. Each face flux leaves one cell and enters the next, so the total circulation, 18.76565218.765652 on this grid, changes by about 10−1410^{-14} in 99009900 steps. At the end the core radius is 0.56050.5605 m exact, 0.5600.560 from the volumes; the peak swirl 3.8093.809 against 3.8123.812 m/s.

Kelvin’s circulation theorem

Follow a material loop C(t)C(t), made of fluid particles.

The mechanism: each element dld\mathbf l is a material fibre, which changes at the velocity difference across it (11.1), D(dl)/Dt=duD(d\mathbf l)/Dt = d\mathbf u. Then

dΓdt=∮CDuDt⋅dl+∮Cu⋅du=∮C(−∇pρ−∇Φ)⋅dl+∮Cd(12∣u∣2)=0.\frac{d\Gamma}{dt} = \oint_C \frac{D\mathbf u}{Dt}\cdot d\mathbf l + \oint_C \mathbf u\cdot d\mathbf u = \oint_C\Big(-\frac{\nabla p}{\rho} - \nabla\Phi\Big)\cdot d\mathbf l + \oint_C d\big(\tfrac12|\mathbf u|^2\big) = 0 .

Euler’s equation (gravity potential Φ\Phi) gave the middle step; every term integrates an exact differential (∇p/ρ=∇ ⁣∫dp/ρ\nabla p/\rho = \nabla\!\int dp/\rho when ρ\rho depends on pp alone) around a closed loop, giving zero. Only what was excluded can change it: viscosity, density not set by pressure alone, non-conservative forces.

From Kelvin come the Helmholtz vortex theorems (stated, same assumptions):

  1. Vortex lines are material: fluid on one stays on one. Mechanism: a small loop drawn on the side of a vortex tube encloses no vorticity flux (ω\boldsymbol\omega is tangent there), so its circulation is zero; by Kelvin it stays zero, so the loop stays on the tube.
  2. A vortex tube’s circulation is the same at every cross-section (∇⋅ω=0\nabla\cdot\boldsymbol\omega = 0) and constant in time.
  3. A vortex tube cannot end in the fluid: it closes, ends on a boundary, or runs to infinity.
◆ Interactive Point vortices: a material loop around vortex 1 keeps its circulation; presets for the pair, co-rotation, leapfrogging and the ground

Loading simulation…

The green loop, 400400 tracers refined as it stretches, starts at 0.999960.99996 (chords cut corners) and reads 0.999810.99981 at t=30t = 30 on 11311131 points. In the leapfrog preset it hits its 30003000-point cap near t=5t = 5, still at 0.999750.99975; then the polygon fails, not Kelvin.

Biot–Savart and point vortices

Since ∇×u=ω\nabla\times\mathbf u = \boldsymbol\omega and ∇⋅u=0\nabla\cdot\mathbf u = 0, vorticity determines velocity (plus boundary conditions) through the Biot–Savart law, borrowed from magnetism. In 3D, stated: u(x)=14π∭ω(x′)×(x−x′)/∣x−x′∣3 dV′\mathbf u(\mathbf x) = \frac{1}{4\pi}\iiint \boldsymbol\omega(\mathbf x')\times(\mathbf x - \mathbf x')/|\mathbf x - \mathbf x'|^3\,dV'. In 2D each bit of vorticity induces a swirl Γ/(2πr)\Gamma/(2\pi r) around itself.

For NN point vortices Γj\Gamma_j, the velocity sums their swirls:

u=−∑jΓj (y−yj)2πrj2,v=∑jΓj (x−xj)2πrj2.u = -\sum_j \frac{\Gamma_j\,(y - y_j)}{2\pi r_j^2}, \qquad v = \sum_j \frac{\Gamma_j\,(x - x_j)}{2\pi r_j^2} .

Vorticity is carried (2D, ν=0\nu = 0), so each vortex moves with the velocity the others induce at it: 2N2N ODEs for RK4. This is the simplest vortex method, vorticity on moving particles with no grid at all (viscosity can enter as a random walk, Chorin’s method, from memory).

The pair. +Γ+\Gamma above −Γ-\Gamma, distance dd apart. Each induces Γ/(2πd)\Gamma/(2\pi d) at the other, in the same direction, so the pair translates rigidly at

U=Γ2πd=0.159155(Γ=d=1).U = \frac{\Gamma}{2\pi d} = 0.159155 \quad (\Gamma = d = 1).

The velocities never change, so RK4 is exact: the sim reads 0.1591550.159155.

Same sign. Two +Γ+\Gamma vortices each move at Γ/(2πd)\Gamma/(2\pi d) on a circle of radius d/2d/2, so Ω=Γ/(πd2)\Omega = \Gamma/(\pi d^2) and the period is 2π2d2/Γ=19.73922\pi^2 d^2/\Gamma = 19.7392; the sim measures 19.739219.7392.

Leapfrogging. Pairs at y=±1y = \pm1 and ±0.5\pm0.5: the faster inner pair shoots ahead, where the outer pair’s flow spreads and slows it while it squeezes the outer pair, which speeds up and passes through. The sim logs the swaps at t=2.84,5.66,8.48,…t = 2.84, 5.66, 8.48, \dots (first step after each): every 2.832.83 time units, the group drifting at 0.21110.2111. Smoke rings do it in 3D (from memory).

The ground. An image vortex −Γ-\Gamma at (x,−y)(x, -y) per real one cancels the flow through the wall y=0y = 0. A pair started at height 33 descends at 0.15490.1549 (free: 0.15920.1592), slows, and runs apart along the ground at nearly Γ/(4πy)\Gamma/(4\pi y), about 0.1610.161. Problem 2 shows each follows 1/x2+1/y2=1/x^2 + 1/y^2 = const, here 4.1111114.111111, so the height tends to 0.49320.4932; the sim holds the constant to 4×10−124\times10^{-12}.

Where a wing’s circulation comes from

A wing sits on the runway in still air; a large material loop around it has Γ=0\Gamma = 0. The wing accelerates and lifts. Outside the thin boundary layers the air is inviscid, so by Kelvin the loop, which never touches the skin, still has Γ=0\Gamma = 0. Yet a lifting wing has circulation. The books balance because a vortex of opposite circulation peels off the sharp trailing edge and stays behind: the starting vortex. With flow left to right and lift up, the bound circulation is clockwise, the starting vortex counterclockwise, summing to zero. Viscosity in the boundary layer makes the vorticity; the sharp trailing edge decides how much (the Kutta condition, 12.1).

How much? Equate Kutta–Joukowski’s lift per unit span ρUΓ\rho U\Gamma (11.3) with 12ρU2c CL\tfrac12\rho U^2 c\,C_L:

Γ=12 Uc CL=12⋅50⋅1.5⋅0.5=18.75 m2/s\Gamma = \tfrac12\,U c\,C_L = \tfrac12\cdot50\cdot1.5\cdot0.5 = 18.75 \text{ m}^2/\text{s}

for 11.5’s light-aircraft wing, chord 1.51.5 m, 5050 m/s, CL=0.5C_L = 0.5. Over an 1111 m span that lifts 12 63312\,633 N, roughly a loaded four-seater (from memory). The starting vortex, +18.75+18.75 m2^2/s, sits near the start of the take-off roll.

By Helmholtz’s third theorem the bound vortex cannot stop at the wingtips: it trails behind each tip back to the starting vortex, one closed loop. The trailing legs are the wingtip vortices. Seen from behind they are a pair, left clockwise, right counterclockwise, and like any pair they translate, here downward. With the trailing sheet rolled up into cores b′=π4b=8.639b' = \tfrac\pi4 b = 8.639 m apart (elliptic loading, from memory), the sink rate is

w=Γ2πb′=18.752π⋅8.639=0.3454 m/s,w = \frac{\Gamma}{2\pi b'} = \frac{18.75}{2\pi\cdot8.639} = 0.3454 \text{ m/s},

and the wing preset measures 0.3454130.345413 m/s, 20.7220.72 m in 6060 s. (A crude model: ρUΓb′=9922\rho U\Gamma b' = 9922 N, π/4\pi/4 of the lift; good to twenty percent.)

◆ Interactive A light aircraft's wake, seen from behind: two trailing vortices, 18.75 m^2/s each, 8.64 m apart, sinking at G/(2 pi b')

Loading simulation…

A heavy jet on approach, 300300 t at 7575 m/s with a 6060 m span: Γ=W/(ρUb′)=680\Gamma = W/(\rho U b') = 680 m2^2/s, sinking at 2.302.30 m/s, 275275 m in two minutes. Hence the several miles, or two to three minutes, that controllers put behind a heavy aircraft (from memory). Near the ground (the ground preset) the pair stops sinking at about half its spacing and runs outward at roughly its old sink rate, toward a parallel runway.

Computing it

🐍 PythonLab: Rankine circulation as loop integrals (A), Lamb–Oseen in air (B), the wing and the wake (C)
⌘/Ctrl + Enter
🧪 Coding exerciseExercise: Biot–Savart velocities and an RK4 step for N point vortices; verify the pair speed and measure the invariant drift
⌘/Ctrl + Enter

Graded: velocities to 10−1210^{-12} (three unequal vortices catch sign slips), the pair at t=10t = 10 to 10−910^{-9}, the impulse to 10−1210^{-12}, HH drift below 10−510^{-5} at h=0.1h = 0.1 and ten times below h=0.2h = 0.2. Total circulation is trivially kept here; on a grid 11.11 must earn it.

Concept checks

✅ Concept check

A light aircraft's trailing vortices each have circulation 18.75 m$^2$/s and are 8.639 m apart. At what speed, in m/s, does the pair sink?

✅ Concept check

Why does pressure vanish from the vorticity equation?

✅ Concept check

Which statement about inviscid 2D flow is true?

✅ Concept check

A wing at rest in still air accelerates and develops a bound circulation of 18.75 m$^2$/s (clockwise, in the wing's frame with flow left to right). What is the total circulation around a large material loop enclosing both the wing and everything it has shed?

Problems

✏️ Pen & paperThe pair and the waltz●○○

Two point vortices are a distance dd apart. (a) For ±Γ\pm\Gamma, find the translation speed. (b) For two +Γ+\Gamma, find the period of rotation. Evaluate for Γ=d=1\Gamma = d = 1.

Show worked solution

(a) Each vortex feels the other’s swirl, Γ/(2πd)\Gamma/(2\pi d), perpendicular to the line joining them. For opposite signs both point the same way: the pair translates at Γ/(2πd)=0.159155\Gamma/(2\pi d) = 0.159155, unchanged forever. (b) For equal signs they are opposite: each moves at Γ/(2πd)\Gamma/(2\pi d) on a circle of radius d/2d/2, so Ω=Γ/(2πd)÷(d/2)=Γ/(πd2)\Omega = \Gamma/(2\pi d)\div(d/2) = \Gamma/(\pi d^2) and the period is 2π/Ω=2π2d2/Γ=19.73922\pi/\Omega = 2\pi^2d^2/\Gamma = 19.7392. The sim’s pair and co-rotation presets print 0.1591550.159155 and 19.739219.7392.

✏️ Pen & paperA wake meets the ground●●●

A pair, −Γ-\Gamma at (−x,y)(-x, y) and +Γ+\Gamma at (x,y)(x, y), descends toward a wall at y=0y = 0. Use images and the conservation of HH to show that 1/x2+1/y21/x^2 + 1/y^2 stays constant, and find the final height when the pair starts at x=0.5x = 0.5, y=3y = 3.

Show worked solution

The images, +Γ+\Gamma at (−x,−y)(-x, -y) and −Γ-\Gamma at (x,−y)(x, -y), make a free four-vortex system that stays a rectangle and conserves HH. The six pairs: two at distance 2x2x (real pair, image pair) with product −Γ2-\Gamma^2; two at distance 2y2y (each vortex with its image) with product −Γ2-\Gamma^2; two diagonals at 2r2r, r=x2+y2r = \sqrt{x^2 + y^2}, with product +Γ2+\Gamma^2. So

H=Γ22π(2ln⁡2x+2ln⁡2y−2ln⁡2r)=Γ2πln⁡2xyr.H = \frac{\Gamma^2}{2\pi}\big(2\ln 2x + 2\ln 2y - 2\ln 2r\big) = \frac{\Gamma^2}{\pi}\ln\frac{2xy}{r} .

Constant HH means constant x2y2/(x2+y2)x^2y^2/(x^2 + y^2), that is, constant 1/x2+1/y21/x^2 + 1/y^2. At the start that is 4+1/9=4.1111114 + 1/9 = 4.111111. As the vortices run apart, x→∞x \to \infty and y→1/4.111111=0.4932y \to 1/\sqrt{4.111111} = 0.4932. The ground preset holds the constant to 4×10−124\times10^{-12} and shows the height 0.49330.4933 at x=29.4x = 29.4.

✏️ Pen & paperLamb–Oseen: what spreads and what stays●●○

(a) Find the Lamb–Oseen circulation inside radius rr. (b) Show the swirl peaks at rc=1.12094νtr_c = 1.1209\sqrt{4\nu t}. (c) Find rcr_c at t=60t = 60 s in air.

Show worked solution

(a) ∫0rω 2πs ds=Γ4πνt⋅2π⋅2νt (1−e−r2/(4νt))=Γ(1−e−r2/(4νt))\int_0^r \omega\,2\pi s\,ds = \frac{\Gamma}{4\pi\nu t}\cdot2\pi\cdot2\nu t\,\big(1 - e^{-r^2/(4\nu t)}\big) = \Gamma\big(1 - e^{-r^2/(4\nu t)}\big), using ∫0rs e−s2/ads=a2(1−e−r2/a)\int_0^r s\,e^{-s^2/a}ds = \tfrac a2(1 - e^{-r^2/a}); it tends to Γ\Gamma at every tt. (b) With x=r/4νtx = r/\sqrt{4\nu t}, uθ∝(1−e−x2)/xu_\theta \propto (1 - e^{-x^2})/x; its derivative vanishes when 2x2e−x2=1−e−x22x^2e^{-x^2} = 1 - e^{-x^2}, at x=1.1209x = 1.1209 (numerically). (c) 4⋅1.4607×10−5⋅60=0.05921\sqrt{4\cdot1.4607\times10^{-5}\cdot60} = 0.05921 m, so rc=0.0664r_c = 0.0664 m, as the lab prints.

Key Takeaways

  • Circulation is the loop integral of velocity, equal to the vorticity flux inside (Stokes): 2.52.5, 1010 and 00 for loops inside, around and beside a Rankine core.
  • The curl of Navier–Stokes removes pressure: Dω/Dt=ν∇2ωD\omega/Dt = \nu\nabla^2\omega in 2D, plus stretching and tilting (ω⋅∇)u(\boldsymbol\omega\cdot\nabla)\mathbf u in 3D only.
  • Viscosity diffuses vorticity without destroying it: the Lamb–Oseen core grows like 4νt\sqrt{4\nu t}, the circulation stays 18.7518.75 m2^2/s.
  • Kelvin: inviscid barotropic flow keeps the circulation of every material loop. Helmholtz: vortex lines move with the fluid and cannot end in it.
  • Biot–Savart turns vorticity into velocity; point vortices make it ODEs. Pairs move at Γ/(2πd)\Gamma/(2\pi d), equal vortices waltz with period 2π2d2/Γ2\pi^2d^2/\Gamma, coaxial pairs leapfrog. RK4 keeps the impulse exactly and lets HH drift: 6.4×10−56.4\times10^{-5} at h=0.2h = 0.2, 2.1×10−62.1\times10^{-6} at h=0.1h = 0.1.
  • A wing’s bound circulation, 12Uc CL=18.75\tfrac12 U c\,C_L = 18.75 m2^2/s, is balanced by the starting vortex on the runway, joined to it by trailing vortices sinking at Γ/(2πb′)=0.35\Gamma/(2\pi b') = 0.35 m/s (a heavy jet’s: 2.32.3 m/s).
  • Next, 11.7: outside boundary layers and wakes the vorticity is zero, so u=∇ϕ\mathbf u = \nabla\phi with ∇2ϕ=0\nabla^2\phi = 0. That is linear: superpose a stream, a cylinder and our point vortex, map the cylinder to an airfoil with one complex function, and Laplace’s equation alone gives a wing its lift.
Finished this lesson?

Mark it complete to track your progress (saved in this browser). Reading every section does this for you.

Where this leads

These later lessons build directly on this one: