Plateau-Rayleigh Instability

The intuition — area is the enemy of a cylinder

Turn a tap to a thin stream and watch the bottom: a few centimeters down, the smooth column dissolves into a chain of separate drops. Nothing pushed them apart. The stream tore itself apart, and the culprit is surface tension — the same force that pulls a droplet into a sphere.

Surface tension stores energy in every square meter of liquid surface. A liquid always wants to shrink that surface area to its smallest possible value for the volume it holds. A long thin cylinder is a remarkably inefficient shape by this measure: it has a lot of skin for not much volume. Gather that same volume into a row of beads and the total surface area drops. So the cylinder is sitting on an energy hill, and the slightest nudge sends it rolling toward beads.

The "nudge" is just thermal and mechanical noise — a microscopic pinch here, a microscopic bulge there. What makes this an instability rather than a wiggle that smooths out is that surface tension amplifies the right pinches instead of erasing them. That self-amplification is the whole story.

How the pinch feeds itself: Laplace pressure

The amplification engine is the Young-Laplace pressure inside a curved surface. For a liquid cylinder the surface has two principal curvatures — one around the circumference (radius R) and one along the axis (set by how the radius varies down the stream). The internal pressure jump is:

Δp = γ (1/R₁ + 1/R₂)

where γ is the surface tension and R₁, R₂ are the two principal radii of curvature. For an unperturbed cylinder Δp = γ/R is the same everywhere. Now imagine a small sinusoidal squeeze: a neck (smaller R, higher 1/R, higher pressure) next to a bulge (larger R, lower pressure).

Pressure pushes liquid from high to low — from the neck toward the bulge. That drains the neck, making it thinner, which raises its pressure even more, which drains it faster. The neck collapses to zero radius and pinches off; the drained liquid swells the bulge into a drop. A runaway feedback loop:

neck thins → higher Laplace pressure → liquid flees neck → neck thins more → … → pinch-off

There is a subtlety the simple picture hides: the axial curvature of a bulge actually opposes the circumferential term. The neck wins the tug-of-war only when the bulge is broad enough — which is exactly the wavelength condition below.

The cutoff: only long wavelengths grow (λ > 2πR)

Whether a given perturbation grows or decays is decided purely by whether it lowers the surface area. Take a cylinder of mean radius R and impose a small axisymmetric ripple of wavelength λ = 2π/k and amplitude ε:

r(z) = R + ε·cos(kz)

Holding the enclosed volume fixed (liquid is conserved) and expanding the surface area to second order in ε gives a change in area proportional to:

ΔA  ∝  (kR)² − 1

Read that off: the area decreases (ΔA < 0, unstable) only when (kR)² < 1, i.e. kR < 1. In wavelength terms:

unstable  ⇔  λ > 2πR   (the circumference of the stream)

This is the elegant heart of the result: the threshold wavelength is exactly the circumference. Squeeze the stream with bumps closer together than its own circumference and you'd be adding surface area, so surface tension irons them out. Space the bumps farther apart and breaking up wins. A 1 mm-radius stream is therefore stable to ripples shorter than about 6.3 mm and unstable to anything longer.

The growth rate and the fastest mode

Being unstable isn't enough — the stream picks the wavelength that grows fastest, because that mode outruns all the others from the same background noise. Each unstable mode grows exponentially, ε(t) = ε₀·e^ωt, and Rayleigh (1878) derived the growth rate ω(k) for an inviscid jet by balancing surface-tension energy against fluid inertia:

ω² = (γ / (ρ R³)) · (kR) · [1 − (kR)²] · [I₁(kR) / I₀(kR)]

Here ρ is the liquid density and I₀, I₁ are modified Bessel functions of the first kind. The bracket [1 − (kR)²] is the same area criterion — it goes negative (stable, ω imaginary) for kR > 1. Maximizing ω over k gives the dominant mode:

kR ≈ 0.697     →     λ_max ≈ 2πR / 0.697 ≈ 9.01·R
ω_max ≈ 0.343 · √(γ / (ρ R³))

So the beads space themselves about nine radii apart, and the natural breakup timescale is τ ≈ 1/ω_max ≈ 2.91·√(ρR³/γ). For a 1 mm-radius water jet (γ ≈ 0.072 N/m, ρ = 1000 kg/m³) that timescale is roughly 11 ms — fast enough to finish within a few centimeters of fall.

Numbers: water, the math, and the drops you see

Quantity	Symbol / formula	1 mm-radius water jet
Surface tension	γ	0.072 N/m
Density	ρ	1000 kg/m³
Stability cutoff	λ_min = 2πR	6.28 mm
Fastest wavelength	λ_max ≈ 9.01·R	9.0 mm
Growth rate	ω_max ≈ 0.343·√(γ/ρR³)	≈ 92 s⁻¹
Breakup time	τ ≈ 1/ω_max	≈ 11 ms
Drop volume	V ≈ πR²·λ_max	≈ 28 µL
Drop radius (sphere)	r_drop = (3V/4π)^(1/3)	≈ 1.9 mm

Notice the payoff in the last rows: a length λ_max of the 1 mm cylinder holds about πR²λ_max ≈ 28 microliters, which rounds up into a sphere roughly 1.9 mm in radius — nearly twice the stream's radius. That's why the drops always look fatter than the thread that made them.

When viscosity enters: the Ohnesorge number

Rayleigh's clean result assumes an inviscid jet. Real fluids resist the internal flow that feeds the bulge, and the relevant dimensionless group is the Ohnesorge number, comparing viscous forces to inertia and surface tension:

Oh = μ / √(ρ γ R)

Viscosity never restores stability — area still wants to fall — but it slows growth and pushes the dominant wavelength longer. In the viscous-dominated limit (Oh ≫ 1), the Chandrasekhar/Tomotika analysis gives a dominant wavelength that grows with Oh:

λ_max ≈ 2πR · √(2 + 3√2 · Oh)

Fluid (thin jet)	Ohnesorge regime	Behavior
Water	Oh ≪ 1 (inertial)	Tight beads ~9R apart; fast pinch-off
Glycerol / light oil	Oh ~ 1	Longer necks; beads farther apart
Honey	Oh ≫ 1 (viscous)	Very long thinning threads; slow, widely spaced beads
Molten polymer	Oh ≫ 1 + elasticity	Beads-on-a-string; threads can persist (elasticity)

Polymer solutions add a twist: stretched polymer chains resist the final pinch and leave the classic "beads-on-a-string" morphology, where tiny droplets hang on a thread that refuses to break. That elastic stabilization is what makes ink and paint behave differently from water at the nozzle.

Where it shows up — and what it costs to control

• Inkjet printing. Every drop-on-demand printer deliberately triggers a Plateau-Rayleigh pinch-off so each ejected ligament collapses into a single, repeatable droplet. Stray satellite drops from the same instability are the main print-quality defect engineers fight.
• Spray nozzles and fuel injection. Atomizers in engines, agricultural sprayers, and inhalers rely on jet breakup to set droplet size — which controls combustion efficiency and how deep a medicine reaches in the lungs.
• Microfluidics. Lab-on-a-chip devices generate millions of identical microdroplets per second by squeezing one fluid into another, tuning the same instability to make uniform reaction vessels for drug screening and single-cell sequencing.
• Spider silk and dew. The sticky coating on a spider's capture thread, and morning dew on it, both bead by Plateau-Rayleigh — regularly spaced glue droplets that make the web catch prey.
• Glass and metal fibers. Drawing optical fiber or making metal powder by atomization both must outrun the instability (fast cooling/solidification) or live with it (deliberate spheroidization for solder balls).
• Your faucet. The few-centimeter "unbroken length" before the stream beads up is the distance the dominant mode needs to grow from noise to full pinch.

Common misconceptions and edge cases

• "Gravity pulls the stream apart." No. Gravity stretches and thins the stream (which actually helps a little, by shrinking R), but the breakup happens in zero gravity too. Astronauts' floating water blobs bead by the same instability. The driver is surface tension, not weight.
• "Air resistance does it." Aerodynamics matters only for very fast or very thin jets (a separate Kelvin-Helmholtz route to atomization). The classic dripping-tap breakup is purely a free-surface, surface-tension effect.
• "All wavelengths break up." Only λ > 2πR grow. Anything shorter than the circumference is stable and is smoothed away. The stream is genuinely selective.
• "More surface tension means slower breakup." Backwards — higher γ means a stronger restoring/driving force, so ω_max ∝ √γ and the stream breaks up faster. Soapy water (lower γ) actually beads a touch more slowly than clean water.
• "Each bump becomes exactly one drop." Usually, but the thin necks between big drops often collapse into small "satellite" droplets too. Suppressing satellites is a real engineering headache in inkjet.
• "It only happens to water in air." It happens to any liquid column bounded by a different fluid — oil in water, water in oil, molten metal in gas — wherever an interface with surface (or interfacial) tension can lower its area by beading.