Kármán Vortex Street

The idea — a fluid that can't make up its mind

Push a steady stream of water or air past a blunt object — a piling, a cable, a chimney, a finger held in a creek — and the wake behind it does not stay still. Above a threshold speed, the object starts shedding vortices first off one side, then the other, then the first side again, in a clockwork rhythm. Those vortices drift downstream in two offset rows, spinning in opposite senses, like a zip-line of alternating whirlpools. That pattern is a Kármán vortex street.

The key word is blunt (engineers say bluff). A streamlined teardrop lets the flow close gently behind it. A blunt body forces the flow to separate — the boundary layer peels off the surface and rolls up into discrete vortices instead of reattaching. The result is not random turbulence; it is a startlingly periodic, self-sustaining oscillation that the fluid generates entirely on its own from a perfectly steady oncoming stream.

You have heard it. The hum of telephone wires in the wind, the whistle of a car aerial, the "singing" of a rope on a flagpole — all are vortex streets driving the body (or the air around it) at the shedding frequency, audible when that frequency lands in the kilohertz range.

How the alternation gets started

As fluid wraps around the front of a cylinder, the pressure rises toward the rear (an adverse pressure gradient). The thin boundary layer can't push uphill against that gradient forever, so it separates from the surface on both sides, forming two free shear layers loaded with vorticity.

A perfectly symmetric pair of vortices would be neat — but it is unstable. One vortex inevitably grows a touch faster, and as it does it draws fluid across the wake centerline. That cross-flow severs the feeding shear layer of the other vortex, which then detaches and washes downstream. Now the surviving side starts a new vortex, the asymmetry flips, and the process repeats on the opposite side. The body becomes a metronome, alternately releasing a vortex left, right, left, right.

Theodore von Kármán's 1911–1912 stability analysis (at Göttingen, after his colleague Karl Hiemenz kept finding his cylinder wakes oscillating) showed that only the staggered double row is dynamically stable, and only at one spacing ratio:

h / a = (1/π)·sinh⁻¹(1) ≈ 0.281

Here a is the streamwise spacing between successive vortices in one row and h is the cross-stream distance between the two rows. The symmetric (side-by-side) arrangement has no stable spacing at all — which is exactly why nature picks the offset pattern.

The governing numbers — Reynolds and Strouhal

Two dimensionless groups control everything. The first sets whether a street forms at all; the second sets how fast it ticks.

Reynolds number compares inertial to viscous forces:

Re = ρ·U·d / μ = U·d / ν

where ρ is fluid density, U the free-stream speed, d the body's cross-stream width, μ the dynamic viscosity, and ν = μ/ρ the kinematic viscosity. See Reynolds number for the full story.

Strouhal number is the dimensionless shedding frequency:

St = f·d / U     ⟹     f = St·U / d

The remarkable empirical fact — established by Anatol Roshko in the 1950s — is that for a circular cylinder St ≈ 0.2 and barely moves across the entire range Re ≈ 300 to 100,000. A useful fit in the laminar range is:

St ≈ 0.198·(1 − 19.7/Re)        (Roshko, 50 ≲ Re ≲ 150)

Because St is essentially a geometric constant, the shedding frequency is simply proportional to wind speed. Double the wind, double the rhythm. That single fact powers vortex flowmeters and threatens slender structures in equal measure.

The forces it produces

Each time a vortex is released, it pulls the body toward the side it left from. Because the vortices alternate sides, the body feels a side (lift) force that oscillates at the shedding frequency f. The streamwise (drag) force, by contrast, gets a kick on every shedding event regardless of side, so it oscillates at twice the frequency, 2f — a detail that surprises people the first time they see the force traces.

The unsteady lift is far from negligible. The fluctuating lift coefficient on a cylinder can reach C′_L ≈ 0.4 to 1.0 in the lock-in regime — comparable to the mean drag coefficient itself. If f lands near a structural natural frequency, those alternating side loads can pump a slender body into large-amplitude, sometimes destructive, vibration. This is vortex-induced vibration (VIV).

Wake regimes vs. Reynolds number

Reynolds number (cylinder)	Wake behavior
Re < ~5	Creeping flow, no separation; streamlines close behind the body
~5 to ~47	Two steady, symmetric recirculation bubbles attached to the rear
~47 to ~190	Laminar, 2D, periodic vortex street — the textbook pattern
~190 to ~260	Wake becomes three-dimensional; spanwise vortex distortion sets in
~260 to ~1,000	Transition to turbulence in the wake; shedding still periodic
~1,000 to ~200,000	Subcritical: laminar separation, turbulent wake, St ≈ 0.2, strong shedding
~200,000 to ~3.5×10⁶	Drag crisis: boundary layer goes turbulent, separation delayed, drag drops sharply, shedding disorganized
> ~3.5×10⁶	Supercritical: organized shedding re-establishes at higher St ≈ 0.27

Worked numbers — frequencies you can hear and fear

Situation	d	U	f = 0.2·U/d
Telephone wire singing in wind	4 mm	10 m/s	≈ 500 Hz (audible hum)
Finger in a stream	15 mm	0.5 m/s	≈ 6.7 Hz
Industrial chimney	5 m	15 m/s	≈ 0.6 Hz
Marine riser / pipeline	0.5 m	1.5 m/s (current)	≈ 0.6 Hz
Suspension-bridge cable	0.6 m	20 m/s	≈ 6.7 Hz
Tacoma Narrows deck (1940)	~2.4 m (effective)	~19 m/s	~0.2 Hz torsional mode

The danger is always the same: when the calculated f lands on a structural natural frequency, the body and the wake synchronize and the amplitude explodes. That synchronization has a name — lock-in.

Lock-in — when the wake captures the structure

Near resonance, the vortex street stops obeying f = St·U/d and instead locks onto the body's natural frequency, holding there over a band of wind speeds (often ±20–40%). During lock-in three things spike together: the vibration amplitude, the spanwise coherence of the vortices (they shed in unison along the whole length instead of in patches), and the unsteady force. This is why a chimney rattles over a range of wind speeds rather than at one razor-thin value, and why "it's just resonance, avoid the one bad speed" is the wrong mental model.

Engineers fight lock-in with helical strakes — the spiral fins you see wrapped around tall steel stacks and offshore risers. The strakes deliberately disrupt spanwise coherence so vortices shed out of phase along the length, killing the organized side force. Perforated shrouds, tuned-mass dampers, and streamlined fairings do the same job by other means.

JavaScript — predicting and detecting shedding

// Vortex shedding frequency from the Strouhal relation
// St ≈ 0.2 for a circular cylinder over Re ≈ 300..1e5
function sheddingFrequency(U, d, St = 0.2) {
  return St * U / d;            // hertz
}

// Reynolds number (kinematic form): nu in m^2/s
// air ~1.5e-5, water ~1.0e-6 at room temperature
function reynolds(U, d, nu = 1.5e-5) {
  return U * d / nu;
}

// Roshko's laminar-range fit for St(Re), 50 <= Re <= 150
function strouhalRoshko(Re) {
  return 0.198 * (1 - 19.7 / Re);
}

// Telephone wire: d = 4 mm, wind U = 10 m/s
const d = 0.004, U = 10;
const Re = reynolds(U, d);                 // ~2670 -> subcritical, St ~ 0.2
console.log(`Re = ${Re.toFixed(0)}`);      // Re = 2667
console.log(`f  = ${sheddingFrequency(U, d).toFixed(0)} Hz`); // f = 500 Hz (audible hum)

// Resonance / lock-in risk check against a structural mode
function lockInRisk(U, d, fNatural, band = 0.3) {
  const f = sheddingFrequency(U, d);
  const ratio = f / fNatural;
  // lock-in window is roughly (1 - band) .. (1 + band) of resonance
  return { f, fNatural, ratio, atRisk: ratio > (1 - band) && ratio < (1 + band) };
}

// Chimney: d = 5 m, first bending mode ~0.6 Hz, design wind U = 15 m/s
console.log(lockInRisk(15, 5, 0.6));
// { f: 0.6, fNatural: 0.6, ratio: 1, atRisk: true }  --> needs strakes/damper

// Detect the dominant shedding frequency from a sensor time series
// (naive single-bin Goertzel-style power estimate)
function powerAtFrequency(samples, fs, fProbe) {
  let re = 0, im = 0;
  const w = 2 * Math.PI * fProbe / fs;
  for (let n = 0; n < samples.length; n++) {
    re += samples[n] * Math.cos(w * n);
    im += samples[n] * Math.sin(w * n);
  }
  return Math.hypot(re, im) / samples.length;  // amplitude at fProbe
}

Where it shows up

• Civil engineering. Chimneys, towers, masts, suspension-bridge cables, and skyscrapers all face vortex-induced vibration; helical strakes and tuned-mass dampers are standard countermeasures.
• Offshore and subsea. Drilling risers, pipelines, and mooring lines in ocean currents shed vortices; VIV fatigue is a primary design driver, often mitigated with strakes or fairings.
• Flow measurement. Vortex flowmeters count shed vortices off a fixed shedder bar; since St is constant, frequency reads velocity directly to about ±1%.
• Heat exchangers and boilers. Tube banks in cross-flow can shed in resonance with acoustic or structural modes, causing loud "singing" and tube fatigue.
• Aeroacoustics. The "Aeolian tone" of wind over wires, antennas, and cables; cooling-fan and pantograph noise on trains.
• Geophysics and atmosphere. Cloud streets trailing downwind of islands (Guadalupe, Jan Mayen, the Canary Islands) are atmospheric Kármán streets tens of kilometers wide, visible from orbit.
• Biology. Trout hold station in the low-energy wake behind rocks, slaloming between shed vortices in a "Kármán gait" to swim upstream for almost free.

Common misconceptions and edge cases

• "It's just turbulence." No — a vortex street is highly ordered, periodic motion. It can exist in completely laminar flow (Re ≈ 50) where there is no turbulence at all.
• "Drag oscillates at the shedding frequency." The lift oscillates at f; the drag oscillates at 2f, because every shed vortex (either side) pulls backward.
• "Tacoma Narrows was simple resonance." The fatal mechanism was torsional aeroelastic flutter, a self-excited instability. Shedding contributed but the runaway torsional mode was flutter — a distinction engineers stress.
• "Faster wind always means worse vibration." The worst vibration is at lock-in, near resonance. Far above or below that band, even high winds may excite the structure far less.
• "Only round cylinders shed streets." Any bluff body does — square bars (St ≈ 0.12), flat plates, H-shaped bridge decks, even buildings. The fixed separation points of sharp-edged bodies actually make shedding more regular.
• "Streamlining stops it." Streamlining delays separation and weakens the street, but a truly bluff trailing edge will still shed. Splitter plates placed along the wake centerline are a direct fix — they block the cross-wake communication that drives the alternation.