Bénard Convection: Order from a Heated Fluid Layer

Heat a thin layer of fluid from below

Take a shallow layer of fluid, heat it uniformly from below and cool it from above, and for small temperature differences nothing visible happens — heat simply conducts upward through the still fluid. Push the temperature difference past a critical threshold, though, and the fluid spontaneously self-organises into a regular pattern of counter-rotating convection rolls or hexagonal cells: warm, less dense fluid rises in some regions, cool, denser fluid sinks in others, in a strikingly ordered pattern that emerges from a completely uniform starting state. This is Rayleigh-Bénard convection, one of the classic examples of spontaneous pattern formation in physics.

The Rayleigh number: the single dial that decides everything

Whether convection happens at all is governed by a single dimensionless quantity, the Rayleigh number, which compares the buoyancy forces driving overturning motion against the two effects that resist it — viscosity, which damps fluid motion, and thermal diffusion, which erases temperature differences before they can drive flow:

Ra = ( g * alpha * DeltaT * L^3 ) / ( nu * kappa )
  g      = gravitational acceleration
  alpha  = thermal expansion coefficient
  DeltaT = temperature difference top-to-bottom
  L      = layer depth
  nu     = kinematic viscosity
  kappa  = thermal diffusivity

Ra < Ra_c (~1708 for rigid boundaries) -> pure conduction, no motion
Ra > Ra_c                              -> convection rolls appear

The critical value Ra_c ≈ 1708 for a layer with two rigid, no-slip boundaries was first derived analytically by the physicist Lord Rayleigh in 1916, building on the experimental observations of Henri Bénard, who in 1900 had heated a thin layer of spermaceti wax and photographed the resulting hexagonal cells. Below Ra_c, conduction always wins — any incipient convective motion is damped out faster than buoyancy can sustain it. Cross the threshold and the fluid abruptly reorganises into rolls, because that convective flow pattern moves heat upward far more efficiently than conduction alone.

Why rolls, and why hexagons

Just above Ra_c, the preferred pattern in an idealised layer with fixed boundary temperatures is a set of parallel convection rolls: fluid rises on one side of a roll, travels across the top, sinks on the other side, and returns along the bottom, with adjacent rolls counter-rotating so their shared edge has fluid moving in the same direction. If instead the upper surface is free (open to air) and surface tension is significant, the same instability more commonly organises into hexagonal cells — Bénard's original wax cells were surface-tension-driven (technically the related Marangoni effect) rather than pure buoyancy-driven convection, a distinction only clarified decades after his original experiments.

The Boussinesq approximation and Navier-Stokes

Simulating this properly means solving the incompressible Navier-Stokes equations coupled to a heat-transport equation, but with one simplifying trick: the Boussinesq approximation treats the fluid's density as constant everywhere except in the buoyancy term, where small density changes from thermal expansion are what actually drive the flow. This keeps the momentum equation's mass-conservation constraint simple (incompressible flow, zero divergence) while still letting temperature-driven buoyancy do its job.

Boussinesq momentum equation (2D, simplified):
du/dt = -grad(p)/rho0 + nu * laplacian(u) + alpha * g * (T - T0) * y_hat
dT/dt = kappa * laplacian(T) - u . grad(T)
div(u) = 0                                    // incompressibility retained exactly

Beyond onset: chaos at high Rayleigh number

Push Ra far beyond Ra_c and the orderly rolls themselves become unstable — first developing wavy, oscillating distortions, then time-periodic and eventually chaotic motion, a route to turbulence that Edward Lorenz's famous three-variable truncation of these very equations (the Lorenz system) was originally derived to study. At very high Rayleigh number the flow becomes fully turbulent convection, the regime relevant to the Sun's convective zone, Earth's mantle, and the atmosphere's convective boundary layer — the same instability mechanism operating across vastly different scales and fluids.

What the simulation shows

This simulation solves a 2D Boussinesq Navier-Stokes model of a heated fluid layer, letting you raise the effective Rayleigh number past the critical threshold and watch conduction give way abruptly to organised convection rolls, then watch those rolls destabilise into more complex, time-varying flow as the driving temperature difference increases further.

Frequently asked questions

Try it live

Everything above runs in your browser — open Bénard Convection and change the parameters while it is running. Nothing is installed, nothing is uploaded, the whole model lives in one tab.