# Convection **Convection** is the transport of [[Heat|heat]] — and of anything else the fluid carries — by the bulk motion of the fluid itself. Where [[Thermal_conduction|conduction]] passes energy from one stationary molecule to its neighbour and [[Thermal_radiation|radiation]] sends it across empty space, convection picks up a parcel of warm fluid and moves it bodily somewhere else. That makes it the fastest of the three modes wherever a fluid is free to move, and it makes it structurally different: the [[Temperature|temperature]] field and the [[Velocity|velocity]] field are coupled, because temperature sets the [[Density|density]] that sets the [[Buoyancy|buoyancy]] that sets the motion that redistributes the temperature. In the microsim below — the Rayleigh–Bénard cell reused from the Thury Hydrodynamics set — the reader heats a thin horizontal layer of fluid from beneath and watches nothing happen, then keeps heating. Below a threshold, [[Viscosity|viscosity]] and [[Thermal_diffusivity|thermal diffusion]] together damp out every disturbance and the heat crosses the layer by conduction alone, in a straight-line temperature profile. Above it the layer breaks into rolls that turn over steadily, carrying warm fluid up one side of each roll and cool fluid down the other. The dimensionless group that decides which regime you are in is the [[Rayleigh–Bénard_convection|Rayleigh number]], Ra = g·β·ΔT·d³/(ν·κ), and Rayleigh's own 1916 calculation puts the threshold for a layer between two stress-free surfaces at Ra_c = 27·π⁴/4 ≈ 657.5.[^rayleigh1916] The sim plays a pre-computed pseudo-spectral Boussinesq solution and offers a spacetime view of the roll pattern's history alongside the live cell.[^wt-bake] On the [[Energy]] flagship this article is the child of Part VII — Energy transfer, section *Convection* (row E58). It is the one section of that part whose sim is not rebuilt: fluid dynamics is placed from the Thury Hydrodynamics spine, and the dense child that carries the detail is [[Rayleigh–Bénard_convection]]. ## History The word is younger than the phenomenon. Eighteenth- and nineteenth-century writers spoke of "caloric" carried by currents in air and water; the recognition that fluid motion is a distinct mode of heat transport, alongside conduction through solids and radiation across a vacuum, belongs to the same nineteenth-century settlement that produced the [[First_law_of_thermodynamics|first law]] and identified heat as energy in transit. The quantitative subject dates from the beginning of the twentieth century, when Henri Bénard observed that a thin layer of liquid heated uniformly from below does not warm smoothly but organises itself into a tiling of polygonal cells. Lord Rayleigh set out to explain those experiments and in 1916 published the linear stability analysis that still carries his name, showing that the layer is stable to infinitesimal disturbances until a single dimensionless combination of gravity, thermal expansion, layer depth, viscosity and thermal diffusivity exceeds a critical value, and that the disturbance which grows first has a definite horizontal wavelength comparable to the layer depth.[^rayleigh1916] That result — a continuous parameter crossing a threshold and a pattern of definite scale appearing — became a template far beyond fluids, and Rayleigh–Bénard convection is still the standard laboratory system for studying [[Self-organization|pattern formation]] and the transition to [[Turbulence|turbulence]]. ## Terminology In fluid mechanics the word carries two meanings that are worth separating. *Convection* in the general sense is transport by bulk motion, whatever drives it; *advection* is sometimes reserved for the transport of a quantity by a flow that is imposed from outside, with *convection* kept for motion the temperature field drives itself. The same distinction appears in engineering as **natural** (or free) convection, in which buoyancy drives the flow, against **forced** convection, in which a pump or a fan drives it and buoyancy is a correction. Mathematically the distinction is invisible: both appear as the convective term in the acceleration of a fluid parcel, a = ∂v/∂t + (v·∇)v, the nonlinear piece that makes the [[Navier–Stokes_equations|Navier–Stokes equations]] hard.[^likharev-cm] A third usage, common in [[Geophysical_fluid_dynamics|geophysics]] and astrophysics, calls any overturning driven by an unstable density stratification "convection" regardless of whether the density contrast comes from heat, from salt, or from composition — which is why one speaks of [[Earth's_internal_heat_budget|mantle convection]] in a solid. ## Mechanisms What all convection has in common is a body force acting on a fluid whose density is not uniform. Change the force or the reason for the density difference and the mechanisms below follow. ### Natural convection In natural convection the density difference is thermal. A fluid warmed at a surface expands, becomes lighter than its surroundings and rises; cooler fluid takes its place and is warmed in turn. The engineering correlation is written as a heat transfer coefficient h in Newton's form Q = h·A·ΔT, but h is not a material property: it depends on the geometry, the orientation, the fluid and the temperature difference itself, typically as a fractional power of ΔT, so natural convection is nonlinear in a way conduction is not. In a soil–atmosphere energy balance the resulting sensible heat flux is the dominant term on a sunny day, peaking near 200 W/m² in Ochsner's measured Iowa cycle against a net radiation of about 300 W/m².[^ochsner-seb] ### Gravitational or buoyant convection Strip away the requirement that heat cause the density difference and the same mechanism operates on any unstable stratification: fresh water over brine, a lighter gas under a heavier one, sediment-laden water plunging beneath clear. This is the general buoyant case, and it includes the inverted arrangement — dense fluid above light — that goes unstable immediately rather than at a threshold. Salinity-driven and temperature-driven contributions can also oppose each other, which is the setting for the double-diffusive regimes seen in the ocean. ### Solid-state convection in ice Given enough time, solids flow. Ice at the base of a thick sheet, or in the mantles of the icy moons, is warm relative to its melting point and creeps under stress, so a layer heated from below can overturn on a timescale of thousands to millions of years. The Rayleigh criterion still applies; what changes is the viscosity, which is larger by twenty or more orders of magnitude than water's and strongly temperature-dependent, so the convecting layer selects its own thickness. The observable signature is a surface that has been resurfaced without melting. ### Thermomagnetic convection A magnetic fluid loses magnetisation as it warms, so in a field gradient the cold fluid is pulled harder than the hot and a circulation develops that needs no gravity at all. The effect cools loudspeaker voice coils and high-field magnets with ferrofluids, and it is one of the few ways to drive convection in free fall. The dielectric analogue, driven by the temperature dependence of permittivity in a strong electric field, works the same way. ### Combustion A flame is a convection engine with its own fuel supply. [[Combustion|Combustion]] releases heat into the gas, the hot gas expands and rises, and the rising column draws fresh air in at the base — so the buoyancy that the reaction creates supplies the oxidiser that sustains it. Remove gravity and a candle flame becomes a dim, nearly spherical ball limited by [[Diffusion|diffusion]] instead of a bright teardrop, which is the cleanest demonstration that the familiar shape of a flame is a convective, not a chemical, fact. ## Examples and applications Convection sets the pace of most of the energy transfers a person meets in a day, and of several that operate on geological and stellar timescales. ### Convection cells Above the threshold the flow organises into cells of a definite size, set by the layer depth, rather than into motion at every scale. Rayleigh's analysis predicts the wavelength of the first mode to grow; the sim shows the resulting rolls, and the spacetime panel shows that once established they are remarkably persistent, drifting and merging only slowly.[^rayleigh1916][^wt-bake] Bénard's polygons, granulation on the Sun and the cellular cloud patterns seen from orbit are the same instability at wildly different scales. ### Atmospheric convection The [[Atmosphere_of_Earth|atmosphere]] is heated from below, at the surface, and cooled by radiation aloft, which is the Rayleigh–Bénard arrangement on a rotating sphere. The energy that drives it is the sensible and latent heat flux leaving the surface, and the mechanical result is [[Wind_power|wind]]. Wind speed grows with height roughly as v(h) = v₀·(h/h₀)^(1/7) in Kerlin's engineering form, so the power in the flow, which goes as v³, grows as (h/h₀)^(3/7) — a factor 2.7 between 10 ft and 100 ft.[^kerlin-wind] That single exponent is why [[Wind_turbine|wind turbines]] sit on towers hundreds of feet tall. ### Oceanic circulation The [[Ocean_current|ocean]] convects on two clocks. Wind-driven surface circulation turns over in years; the [[Thermohaline_circulation|thermohaline circulation]], driven by cold, salty water sinking at high latitudes, turns over in centuries to a millennium and carries a substantial share of the planet's poleward heat transport. Because both temperature and salt set the density, and because heat diffuses about a hundred times faster than salt, the ocean exhibits double-diffusive layering that a purely thermal analysis cannot produce, visible as the stacked steps in a [[Thermocline|thermocline]]. ### Mantle convection The Earth's rocky mantle convects as a solid, at a few centimetres a year, driven by the [[Earth's_internal_heat_budget|internal heat budget]] — primordial heat plus radioactive decay. That circulation is what moves the plates at the surface and what delivers heat to the crust, and it is the ultimate source of the [[Geothermal_energy|geothermal]] resource; extraction, though, is limited not by the mantle but by [[Thermal_conduction|conduction]] through the last few kilometres of rock into a well.[^kerlin-rock] ### Stack effect A heated building is a chimney. Warm indoor air is lighter than the outdoor air at the same level, so the pressure difference across the envelope grows with height and with ΔT, pushing air out near the top and drawing it in near the bottom. In a tall building this can dominate the ventilation rate and defeat the design intent of a mechanical system; in a chimney it is the design intent, and it is the same buoyant column that makes a fire draw. ### Stellar physics A [[Star|star]] carries its energy outward by radiation where the material is transparent enough and by convection where it is not. The Sun's outer third is a convection zone, and the granulation visible on its surface is the top of that circulation, cells roughly a thousand kilometres across turning over in minutes. Which mechanism operates where is the central question of [[Stellar_structure|stellar structure]], because convection also mixes composition and so changes how long a star can burn. ### Nuclear reactors Convection is a safety strategy as well as a heat-transfer mechanism. A reactor core coolable by natural circulation — hot coolant rising through the core, cooling in a heat exchanger above, falling back — keeps removing decay heat with no pump and no power, the principle behind passive safety systems.[^kerlin-nuclear] The reverse case is what makes loss-of-coolant accidents serious: with the circulation broken, conduction alone cannot carry away even a few percent of full power, and the fuel heats until it damages itself. ## Mathematical models of convection Convection is modelled by solving the [[Navier–Stokes_equations|Navier–Stokes equations]] together with an advection–diffusion equation for temperature, coupled through a buoyancy term. The standard simplification is the Boussinesq approximation: treat the fluid as incompressible everywhere except in the buoyancy force, where the density is allowed to vary linearly with temperature.[^wt-bake] That single move removes acoustic waves from the problem, keeps the physics that matters, and makes the system tractable by spectral methods. ### Onset The onset problem is the one the sim illustrates, and it is a [[Bifurcation_theory|bifurcation]]. Linearise about the motionless conducting state, look for disturbances of the form exp(i·k·x + σ·t), and ask for which Rayleigh number some wavenumber k has a growth rate σ > 0. Rayleigh found that for a layer bounded by two stress-free surfaces the answer is Ra > 27·π⁴/4 ≈ 657.5, attained at a horizontal wavenumber of π/(d·√2), so the first rolls are about 2.8 layer depths wide — wider than they are tall.[^rayleigh1916] Rigid, no-slip plates stiffen the layer and push the threshold higher. The physical content of the criterion is a competition of timescales: buoyancy tries to lift a warm parcel, while [[Viscosity|viscous]] drag slows it and thermal diffusion erases the temperature excess that made it buoyant, and Ra is the ratio of the destabilising rate to the product of the two stabilising ones. Reading the sim this way makes the threshold visible rather than asserted. Below Ra_c the flux across the layer is exactly the conductive flux, k·ΔT/d, and the [[Heat_transfer|Nusselt number]] — the ratio of actual to conductive flux — is 1. Above it the Nusselt number rises steeply, and the layer has found a way of moving heat that costs kinetic energy but beats diffusion by orders of magnitude. ### Turbulence Push the Rayleigh number several orders of magnitude past the threshold and the steady rolls give way to time-dependent, then chaotic, then fully turbulent convection: a well-mixed interior at nearly uniform temperature with all the gradient squeezed into thin boundary layers at the plates, and buoyant plumes detaching from them. The route there is the general route to [[Turbulence|turbulence]] in shear and buoyant flows; in pipe flow the analogous transition happens near a [[Reynolds_number|Reynolds number]] of about 2,100, and in a plane channel linear theory gives 5,772.[^likharev-cm] The practical consequence is that turbulent convection is much better at moving heat and much harder to predict from first principles, which is why engineering leans on empirical correlations. ### Behavior Across all these regimes the heat flux is usually written as a power law in the governing dimensionless numbers, Nu = C·Ra^n, with n near 1/4 in the laminar regime and near 1/3 when the flow is turbulent — the second exponent implying a flux independent of the layer depth, because the boundary layers set the answer and the interior is along for the ride. ILLUSTRATIVE: those exponents are display fits to measured data over limited ranges, not derived laws, and a correlation quoted outside the range it was fitted over can be badly wrong. ## Natural convection from a vertical plate The vertical heated plate is the textbook case because it is the simplest geometry in which buoyancy and shear act along the same axis. Fluid next to the plate warms, rises, and forms a [[Boundary_layer|boundary layer]] that thickens with height while the velocity profile inside it goes from zero at the wall, through a maximum, back to zero at the edge of the still fluid — a shape no forced flow produces. Because the layer thickens upward, the local heat transfer coefficient is largest at the leading edge and falls with height, so a tall plate transfers less per unit area than a short one at the same ΔT. Beyond a height that depends on the fluid and the temperature difference, the rising layer itself becomes unstable and turns turbulent, and the coefficient stops falling. ## Pattern formation Convection is the canonical laboratory for [[Self-organization|self-organisation]]: a system driven steadily away from equilibrium selects a spatial pattern with a length scale that no one put in by hand. Ilya Prigogine used exactly this system as the type specimen of a [[Dissipative_system|dissipative structure]], a pattern that exists because of, and not despite, the [[Entropy_production|entropy it produces]].[^prigogine-nobel] The sequence — uniform state, threshold, rolls, secondary instabilities that make the rolls wavy, then chaos — recurs in [[Reaction–diffusion_system|reaction–diffusion systems]], in granular layers and in [[Nonlinear_system|nonlinear]] optics, and the reason is that the linear stability problem near threshold has the same mathematical form in all of them. What the microsim shows is the first step of that ladder, the one place where a clean number, Ra_c, separates "nothing happens" from "structure".[^rayleigh1916] ## See also - [[Rayleigh–Bénard_convection]] - [[Heat_transfer]] - [[Thermal_conduction]] - [[Thermal_radiation]] - [[Heat_exchanger]] - [[Buoyancy]] - [[Turbulence]] - [[Thermohaline_circulation]] ## References [^rayleigh1916]: Rayleigh, Lord (Strutt, J. W.) (1916). "On convection currents in a horizontal layer of fluid, when the higher temperature is on the under side." *Philosophical Magazine*, Series 6, 32 (192): 529–546. The stress-free critical value 27·π⁴/4 = 657.51 and the corresponding wavenumber π/(d·√2), giving rolls about 2.83 layer depths wide, are computed from the paper's result. [^wt-bake]: Wikitube MICROSIM_GUIDE, `01_Core_Guide.md` ("The Kelvin–Helmholtz, Rayleigh–Taylor and Rayleigh–Bénard movies are pseudo-spectral Boussinesq runs shipped the same way"), `03_Framework_Architecture.md` (pseudo-spectral Boussinesq on a 256 × 128 grid, RK4 with 2/3 dealiasing, shipped as 32 frames at 96 × 48 in u8, about 200 KB) and `LIBRARY_API.md` (`bake_spectral.py`, Rayleigh–Bénard by the mirror trick). The sim embedded on this page is the Thury Hydrodynamics build at `/thury/Rayleigh–Bénard_convection.html`, reused rather than rebuilt for this cluster. [^likharev-cm]: Likharev, Konstantin (2013). *Essential Graduate Physics, Part CM: Classical Mechanics*. Chapter 8 (fluid mechanics), pp. 193–222: the convective term in the acceleration of a fluid parcel, a = ∂v/∂t + (v·∇)v, at pp. 199–202; the Navier–Stokes equation at p. 209; the definitions of the drag coefficient and Reynolds number at p. 214; and the transition thresholds — pipe flow at Re ≈ 2,100 and plane channel flow at 5,772 — at pp. 216–217. Portal Book 074, https://open.umn.edu/opentextbooks/textbooks/part-cm-classical-mechanics [^ochsner-seb]: Ochsner, Tyson (2019). *Rain or Shine*. Chapter 12, "Surface Energy Balance and Evapotranspiration", pp. 275–296: the surface balance R_n = LE + H + G with sign conventions at p. 293, the measured Iowa diurnal cycle over corn residue with R_n from about −50 W/m² at night to +300 W/m² near noon and a sensible heat flux H peaking near +200 W/m² at p. 295, and conduction dominating heat transfer within the soil at p. 294. Portal Book 119, https://open.umn.edu/opentextbooks/textbooks/rain-or-shine [^kerlin-wind]: Kerlin, Thomas (2013). *Future Energy: Opportunities & Challenges*, pp. 220–222 — the wind-class table at 33 ft and 164 ft (p. 220), the shear relation v(h) = v₀·(h/h₀)^(1/7) stated as approximate and the derived power ratio (h/h₀)^(3/7) tabulated in Table 8-2, which gives 2.69 at 100 ft against 10 ft (p. 222), and the cubic power law P = ½·ρ·A·v³ with the Betz cap of 59 % (p. 221). Portal Book 048, https://open.umn.edu/opentextbooks/textbooks/future-energy-opportunities-challenges [^kerlin-rock]: Kerlin (2013), *Future Energy: Opportunities & Challenges*, Chapter 7, pp. 249–266 (geothermal extraction and heat transfer through rock; page to pin). Portal Book 048. [^kerlin-nuclear]: Kerlin (2013), *Future Energy: Opportunities & Challenges*, Chapter 8, pp. 267–353 (reactor types, decay heat and safety; page to pin). Murphy, Thomas (2021), *Energy and Human Ambitions on a Finite Planet*, Chapter 6 "Alternative Energy", p. 276, gives the thermal-to-electric ratio of a modern fission plant as 2.5 GW thermal to 1 GW electric. Portal Books 048 and 097. [^prigogine-nobel]: Nobel Prize Outreach. "The Nobel Prize in Chemistry 1977 — Ilya Prigogine." https://www.nobelprize.org/prizes/chemistry/1977/summary/ ## External links - [Future Energy: Opportunities & Challenges](https://open.umn.edu/opentextbooks/textbooks/future-energy-opportunities-challenges), Thomas Kerlin, open textbook (Portal Book 048) - [Rain or Shine](https://open.umn.edu/opentextbooks/textbooks/rain-or-shine), Tyson Ochsner, open textbook (Portal Book 119) - [Essential Graduate Physics, Part CM: Classical Mechanics](https://open.umn.edu/opentextbooks/textbooks/part-cm-classical-mechanics), Konstantin Likharev, open textbook (Portal Book 074) - The Wikipedia pair's external links list further open resources <!-- MATTERSIM:BEGIN g24 — Matter & Energy Cluster microsim (framework build, specs/sims/Convection.json); do not hand-edit inside --> **Microsim — three.js (Wikitube framework):** *Convection* <div class="wt-sim" data-src="https://wikitube-3d-microsims.netlify.app/thury/Rayleigh–Bénard_convection.html" data-title="Convection"></div> *Reused from the Thury Hydrodynamics set (Compendium §5, `bake_spectral.py`): fluids are placed from Thury, not rebuilt for this cluster. The dense child is [[Rayleigh–Bénard_convection]].* <!-- MATTERSIM:END --> ## Wikipedia : Wikitube **Strict pair:** [Wikipedia](https://en.wikipedia.org/wiki/Convection) : [Wikitube](https://en.wikitube.io/wiki/Convection) · pinned revision [1372226643](https://en.wikipedia.org/w/index.php?oldid=1372226643) · 2026-09-11 ## Previous hub tags Hubs: `Life_Physics`. Portals: [[PORTAL_Energy]]. --- *Matter & Energy Cluster child articles, wave 1 · 2026-09-11 · drafted · Energy row E58 · sim live (reused thury/Rayleigh–Bénard_convection.html; fluids placed from Thury, not rebuilt).*