Faraday Waves

The intuition

Pour a thin layer of water into a dish and set it on a loudspeaker cone. Drive the speaker straight up and down — no sideways motion at all — and start cranking up the volume. For a while nothing happens: the surface stays a flat mirror. Then, at a sharp threshold, the whole surface suddenly comes alive with a regular grid of standing ripples. Stripes, a checkerboard of squares, or a honeycomb of hexagons appear from nowhere and sit there, throbbing.

The strange part is that you only ever pushed the fluid vertically, yet the waves run horizontally. Nothing in the forcing told the water to ripple sideways — it organized those crests entirely on its own. And if you watch carefully with a strobe, the crests bob up and down at exactly half the frequency you're shaking the dish. Drive at 60 Hz, the ripples beat at 30 Hz.

This is the Faraday instability, first reported by Michael Faraday in 1831. It is one of the cleanest examples in all of physics of spontaneous pattern formation: a uniform, symmetric system (flat surface, vertical shaking) breaks its own symmetry and selects a structured state once you drive it hard enough.

The parametric mechanism — why it's a subharmonic

The key word is parametric. When you vibrate the container, you are not pushing the surface waves directly. Instead you are periodically changing the effective gravity that each surface mode feels. In the accelerating frame of the dish, the restoring acceleration is no longer a constant g but a wobbling g_eff(t) = g + a·cos(ω·t), where a = A·ω² is the peak vibration acceleration (amplitude A, drive angular frequency ω).

That puts a time-varying coefficient inside the equation governing the amplitude ζ_k(t) of a surface mode with wavenumber k. To leading order each mode obeys the Mathieu equation:

d²ζ_k/dt²  +  2γ_k · dζ_k/dt  +  ω_k² · [ 1 − (a·k/ω_k² · tanh(k·h)) · cos(ω·t) ] · ζ_k  =  0

Here ω_k is the mode's natural surface-wave frequency, γ_k ≈ 2·ν·k² is the viscous damping rate (ν = kinematic viscosity), and the bracketed cosine is the parametric pump. Compare it to a child pumping a swing by standing and crouching: you change a parameter (the pendulum length / your moment of inertia) twice per swing, not push the swing sideways. The most efficient resonance of a parametrically driven oscillator happens when you modulate the parameter at twice the oscillator's natural frequency.

Turn that around: when you pump at ω, the mode that resonates is the one with ω_k ≈ ω/2. The surface therefore responds at half the drive frequency:

f_response = f_drive / 2          (the principal subharmonic)

This subharmonic response is the diagnostic fingerprint of Faraday waves. (Higher "tongues" of the Mathieu equation also exist: a weaker harmonic response at f_drive itself can win for very high viscosity or for certain two-frequency forcing, but the standard, easily-seen Faraday wave is the f/2 subharmonic.) Benjamin and Ursell put this on rigorous footing in 1954, deriving exactly this Mathieu structure from the linearized fluid equations.

Threshold and stability tongues

The damping term 2γ_k is what makes the threshold sharp. The parametric pump tries to grow the mode; viscosity tries to kill it. The mode only grows when pumping beats dissipation. In the Mathieu picture, each mode lives inside or outside a resonance tongue in the (frequency, forcing-strength) plane. The tongue for the subharmonic response has its tip at zero forcing for an inviscid fluid; damping lifts the tip up to a finite critical forcing.

The result is a critical acceleration a_c. Below it the surface is flat; above it a band of modes is unstable and one wins. For weak damping the threshold scales roughly as:

a_c  ≈  (4 · γ_k · ω_k) / (k · tanh(k·h))   →   a_c  ∝  γ_k  ∝  ν   (low-viscosity limit)

Three consequences fall out immediately:

• More viscous fluid → higher threshold. Glycerin needs far more violent shaking than water, because γ_k = 2νk² grows with viscosity, and it especially penalizes short-wavelength (large-k) modes.
• Surface tension and depth set the selected wavelength. The mode that goes unstable first is the one whose natural frequency ω_k equals ω/2, fixed by the gravity-capillary dispersion relation below.
• Sharp on/off switch. Crossing a_c flips the surface from flat to patterned over a tiny change in drive amplitude — the hallmark of a supercritical bifurcation (occasionally weakly subcritical, giving a little hysteresis).

Wavelength selection — the gravity-capillary dispersion relation

Which wavelength wins? The one whose free surface-wave frequency matches the subharmonic ω/2. For a fluid of density ρ, surface tension σ, depth h, that frequency follows the gravity-capillary dispersion relation:

ω_k²  =  ( g·k  +  (σ/ρ)·k³ ) · tanh(k·h)

set  ω_k = ω_drive / 2   →   solve for k   →   λ = 2π / k

The two terms compete: g·k is the gravity restoring force (dominant for long waves), and (σ/ρ)·k³ is the capillary (surface-tension) restoring force (dominant for short waves). They cross over at the capillary length λ_c = 2π·√(σ/(ρg)), about 1.7 cm for water.

For the drive frequencies people actually use (tens to hundreds of Hz), the selected ripples are far shorter than 1.7 cm, so the capillary term dominates: ω_k ≈ √(σ/ρ)·k^(3/2), which gives λ ∝ f_drive^(−2/3). Plainly: shake faster, get a finer grid. That is exactly the trend the visualization above shows — doubling the drive frequency crowds the crests closer together.

Pattern selection — stripes, squares, hexagons

Linear theory tells you the magnitude of k that grows, but it is blind to direction: every orientation of the wave-vector grows at the same rate (the flat surface has no preferred direction). So linear theory predicts a wavelength but a degenerate jumble of orientations. The actual pattern — one set of lines, two perpendicular sets, or three at 120° — is chosen by nonlinear interactions between the growing modes, captured by amplitude (Landau-type) equations.

Pattern	Wave-vectors	Symmetry	Typically favored when
Stripes (rolls)	1 (a single k)	1-D translational	Strong cross-mode suppression; some single-frequency regimes
Squares	2, at 90°	4-fold (tetragonal)	Low-to-moderate viscosity, deep water, single-frequency drive near threshold
Hexagons	3, at 120°	6-fold	Broken up/down symmetry — e.g. two-frequency forcing, or a quadratic resonance
Quasi-patterns	8, 10, 12-fold	No translational order	Two-frequency forcing with incommensurate spatial scales (Edwards & Fauve, 1994)
Oscillons	localized	radially symmetric blob	Granular layers and some fluids — a single bouncing peak
Spatiotemporal chaos	many, defect-ridden	none (disordered)	Far above threshold — patterns roam, merge, and break

Squares are the classic deep-water single-frequency result near threshold; the system "wants" two perpendicular subharmonic modes to coexist. Hexagons need an additional ingredient that breaks the up–down symmetry of the surface (a quadratic term in the amplitude equation), most easily arranged with a two-frequency drive. Push well past threshold and the orderly lattice gives way to defects, drifting domains, and ultimately spatiotemporal chaos.

Numbers — water shaken at 60 Hz

Take a shallow dish of clean water: ρ = 1000 kg/m³, σ = 0.072 N/m, depth h = 5 mm, kinematic viscosity ν = 1.0 × 10⁻⁶ m²/s. Drive it at f_drive = 60 Hz, so the surface responds at f_response = 30 Hz, meaning ω_k = 2π·30 ≈ 188 rad/s.

Quantity	Symbol / formula	Value
Drive frequency	f_drive	60 Hz
Response frequency	f_drive / 2	30 Hz
Selected wavenumber (capillary)	k ≈ (ρ·ω_k²/σ)^(1/3)	≈ 7.9 × 10² rad/m
Crest-to-crest spacing	λ = 2π/k	≈ 8 mm
Capillary length	2π·√(σ/ρg)	≈ 17 mm (> λ ⇒ capillary regime)
Viscous damping rate	γ = 2·ν·k²	≈ 1.2 s⁻¹
Threshold acceleration (order of magnitude)	a_c ≈ 4·γ·ω_k / (k·tanh kh)	~ 0.5–1 g (often below g)

The takeaways: at 60 Hz on water you get a few-millimeter grid, the response is firmly capillary, and you need to shake the dish at a peak acceleration of order g — for clean water at tens of Hz, experimentally a fraction of g (around 0.5–1 g) — before anything appears. Switch to glycerin (ν ≈ 1000× water) and γ — and therefore a_c — jumps by orders of magnitude: you have to shake it brutally hard to see any waves at all.

Real-world Faraday waves

• Cymatics and the speaker-dish demo. The viral videos of water or cornstarch dancing on a subwoofer are Faraday waves. (When the medium is a vibrating solid plate with sand on top, you're seeing Chladni nodal patterns instead — a related but distinct phenomenon; see below.)
• Ultrasonic atomization. Drive a liquid film at tens of kHz past threshold and the crests sharpen until droplets pinch off from the antinodes. This is the working principle of cool-mist humidifiers, medical nebulizers, fuel-injection studies, and some inkjet and spray-coating systems. The droplet size tracks the Faraday wavelength, so higher frequency makes a finer mist.
• Bouncing droplets / hydrodynamic quantum analogs. Below the Faraday threshold, a droplet placed on the bath bounces indefinitely and "walks," guided by the wave it makes. Couder and Fort's walking droplets reproduce single-particle diffraction-like statistics — a striking macroscopic pilot-wave analog studied since 2005.
• Ripple tanks and wave-pattern teaching. Vibrated trays are a standard demo of standing waves, dispersion, and symmetry breaking; Faraday's hard threshold makes the onset dramatic to watch.
• Sloshing and engineering. Vertical vibration of fuel tanks, reactor vessels, and transport containers can parametrically excite Faraday-type surface instabilities — a sloshing-control concern in aerospace and process engineering.
• Pattern-formation research. Faraday waves are the go-to lab system for testing theories of nonlinear pattern selection, quasicrystalline order, and spatiotemporal chaos, because the control parameters (frequency, amplitude, viscosity, depth, drive waveform) are all easy to tune.

Faraday waves vs Chladni patterns

They are the two great "vibration makes patterns" demos, and they're constantly confused. The difference is what's vibrating and what the pattern marks.

Feature	Faraday waves	Chladni patterns
Medium	Liquid surface	Solid elastic plate
What you see	Antinodal crests — fluid piles up where motion is largest	Nodal lines — sand collects where the plate is still
Frequency response	Subharmonic: f_drive / 2	Harmonic: at the plate's resonant eigenfrequency
Mechanism	Parametric instability (Mathieu equation)	Resonant standing modes of the plate (eigenmodes)
Onset	Hard acceleration threshold a_c	Gradual — grows as you tune toward a resonance
Sideways forcing	None — purely vertical drive, surface self-organizes	Plate boundary conditions fix the mode shapes
Restoring force	Gravity + surface tension	Plate bending stiffness (elasticity)
Wavelength trend	Shorter as f rises (capillary, λ ∝ f^(−2/3))	Higher modes (more nodal lines) as f rises

Bottom line: Chladni patterns are the resonant eigenmodes of a driven solid; Faraday waves are a parametric instability of a fluid surface with no direct horizontal forcing and a subharmonic, threshold-gated response.

Misconceptions and edge cases

• "The waves respond at the drive frequency." No — the principal, easily-seen response is the subharmonic at f_drive / 2. A harmonic (f_drive) response exists in the Mathieu tongue structure and can dominate for very high viscosity or specially tuned two-frequency forcing, but the canonical Faraday wave is f/2.
• "You need to push the fluid sideways." No. The forcing is strictly vertical. The horizontal pattern is self-organized; the drive only modulates the effective gravity (a parameter), which is why it's a parametric, not a directly forced, resonance.
• "Any shaking, however gentle, makes waves." No — below the threshold acceleration a_c, viscosity wins and the surface stays flat. The onset is a genuine bifurcation, not a gradual ramp.
• "It's the same as sloshing or capillary waves you blow on." Ordinary sloshing and wind-blown ripples are directly forced. Faraday waves are parametrically forced and subharmonic — a different beast, even though all three live on a fluid surface.
• "The pattern is fixed by the container shape." The boundary nudges orientation and can quantize allowed modes in small dishes, but the wavelength is set by the dispersion relation at f/2, and the lattice symmetry (stripes/squares/hexagons) is chosen by nonlinear mode competition — not by the dish.
• "Faraday waves and the Faraday rotation/Faraday's law are related." Same person, three different phenomena. Faraday waves are fluid mechanics; Faraday's law is electromagnetic induction; Faraday rotation is magneto-optics. Don't conflate them.