Skip to content

Kelvin–Helmholtz Instability

Intuition

Wind blowing over water makes waves. Two clouds layers sliding past each other roll up into perfect billows. Wherever two fluid streams shear past one another, the interface between them is a coiled spring: any tiny ripple deflects the flow, the deflection changes the pressure (Bernoulli: faster = lower), and the pressure difference amplifies the ripple. Shear + interface = instability.

Setup and result

Two inviscid, irrotational streams: density \(\rho_1\), speed \(U_1\) above; \(\rho_2\), \(U_2\) below, gravity \(g\) down, interface at \(y = 0\) perturbed as \(h \sim e^{ikx + st}\). Away from the interface the flow stays potential (\(\nabla^2\phi_{1,2} = 0\), perturbations decaying as \(e^{\mp ky}\)); matching the kinematic condition and pressure continuity (perturbed Bernoulli) yields the dispersion relation:

\[s = -ik\,\frac{\rho_1 U_1 + \rho_2 U_2}{\rho_1 + \rho_2} \;\pm\; \sqrt{\frac{k^2\rho_1\rho_2 (U_1 - U_2)^2}{(\rho_1+\rho_2)^2} + \frac{kg(\rho_1 - \rho_2)}{\rho_1 + \rho_2}}\]

(Full derivation on the dispersion-relation page.) Instability ⇔ the square root is real and positive, i.e.

\[\boxed{\,k^2 \rho_1\rho_2 (U_1 - U_2)^2 > k g\,(\rho_2^2 - \rho_1^2)\,}\]

(with \(\rho_2\) the lower/denser fluid).

Reading the physics out of the formula

  • Any shear destabilizes short waves. If \(U_1 \neq U_2\), the \(k^2\) shear term beats the \(k\) gravity term at large \(k\) — without extra physics, arbitrarily small wavelengths grow arbitrarily fast.
  • What tames it in reality: viscosity (damps large \(k\)) and surface tension \(\gamma\), which adds \(-\gamma\kappa^3/(\rho_1+\rho_2)\) under the square root and stabilizes short waves. With both gravity and surface tension, stability holds while $\((U_1 - U_2)^2 < \frac{2(\rho_1+\rho_2)}{\rho_1\rho_2}\sqrt{\gamma g (\rho_2 - \rho_1)}\)$
  • Wind over water: \(U_2 = 0\), \(\rho_2 \gg \rho_1\) gives onset at \(k > g\rho_2/(U_1^2\rho_1)\). With real air/water numbers the critical wind is \(U_1 \approx 6.6\ \text{m/s}\) with first wavelength \(\lambda \approx 1.7\ \text{cm}\) — ripples appear on a pond at a stiff breeze, as observed.
  • 3-D perturbations (\(e^{ikx + ilz}\)) are less unstable than 2-D ones with the same \(k\) (Squire's theorem in spirit): the most dangerous billows are two-dimensional.

Where you see it

Wind waves · billow clouds (Kelvin–Helmholtz clouds) · smoke rising into crosswind · the flanks of Jupiter's Great Red Spot · magnetized versions in plasmas · mixing layers in jet engines.

Common mistakes

  • Forgetting the vortex sheet. The unperturbed interface carries infinite vorticity (velocity jump); "irrotational" applies only away from it.
  • Concluding "everything is always unstable". The inviscid, tension-free result is a limiting idealization; real cutoffs matter and give the observed onset thresholds.
  • Confusing KH with Rayleigh–Taylor: KH is shear-driven (can occur with stable stratification); RT is buoyancy-driven (no shear needed).

Knowledge graph position

Prerequisites: Hydrodynamic stability, Potential flow, Bernoulli. Leads to: Turbulence, mixing-layer theory.

Quiz

Q1 (conceptual). Explain in one sentence, using Bernoulli, why a bump on the interface between two shearing streams grows.

Answer

The stream deflected over the bump speeds up, lowering pressure above it (and the one below slows, raising pressure beneath), so the net pressure difference pulls the bump further out.

Q2 (computational). For air over water with \(U_1 = 10\) m/s, is a 1.7 cm ripple unstable? (\(\rho_1 = 1.22\), \(\rho_2 = 1000\ \text{kg/m}^3\), \(\gamma = 0.074\) N/m)

Answer

Critical speed is 6.6 m/s; at 10 m/s > 6.6 m/s the combined gravity–capillary restoring force is beaten and ripples near that wavelength grow. Yes — unstable.

Q3 (multiple choice). In the pure KH limit (\(g = 0\), no surface tension), the growth rate scales with wavenumber as:

  • (a) \(s \propto k^{1/2}\) (b) \(s \propto k\) (c) \(s \propto k^2\) (d) independent of \(k\)
Answer

(b). \(s = k\,|U_1 - U_2|\sqrt{\rho_1\rho_2}/(\rho_1+\rho_2)\) — shortest waves grow fastest, which is why regularization physics is essential.