{
  "id": "convection",
  "title": "Convection",
  "lang": "en",
  "source_lang": "en",
  "summary": "Convection transports heat and other properties through fluid motion, shaping weather, oceans, planetary interiors, stars, and engineered systems.",
  "category": "science",
  "subcategory": "physics",
  "type": "concept",
  "aliases": [
    "Convective transport"
  ],
  "generated_at": "2026-11-03",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/convection/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/convection/",
      "json": "https://aiwiki.page/zh/convection.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Convection is the transport of heat, dissolved substances, or other properties by the collective motion of a fluid. In heat-transfer engineering, the term commonly describes heat exchange between a surface and a moving fluid, a process that combines fluid motion with molecular conduction. In atmospheric science, oceanography, and planetary physics, it often refers more specifically to motion generated by buoyancy: differences in density cause some material to rise and other material to sink in a gravitational field.\n\nConvection connects processes across an extraordinary range of scales. It circulates air above a heated surface, organizes thunderstorms, helps ventilate the deep ocean, moves the slowly deforming rock of Earth’s mantle, and carries energy through parts of stars. Its behavior depends on the competition among buoyancy, inertia, viscosity, and thermal or compositional diffusion. Although the governing conservation laws are well established, their nonlinear interactions produce complex patterns and turbulence that remain central subjects of experimental and theoretical research."
    },
    {
      "heading": "Physical meaning and forms of convection",
      "text": "The defining feature of convection is transport by macroscopic motion. This distinguishes it from heat conduction, which transfers energy through microscopic interactions without requiring bulk displacement of matter, and from electromagnetic radiation, which can transfer energy through a vacuum. These mechanisms often operate simultaneously. A hot radiator, for example, emits radiation while also warming adjacent air; that air becomes buoyant and moves upward, drawing cooler air toward the surface.\n\nTerminology varies between disciplines. Advection is the transport of a property by a specified velocity field. In fluid mechanics, convection may be used broadly for this transport or for the combination of advection and molecular diffusion. In meteorology, however, convection usually means buoyantly driven motion with an important vertical component, whereas horizontal transport by the large-scale wind is normally called advection. Consequently, a statement that a process is “convective” must be interpreted in its disciplinary context.\n\nThree engineering categories distinguish how motion is produced. Natural convection, also called free convection, occurs when density differences in a body-force field generate the flow. Air rising above a warm wall is a standard example. Forced convection occurs when a fan, pump, moving surface, or externally imposed flow drives the fluid. Air blown over an electronic heat sink and water pumped through a cooling channel are examples. Mixed convection occurs when imposed motion and buoyancy are both dynamically important. Buoyancy may assist the imposed flow, oppose it, or generate secondary circulation across it.\n\nTemperature differences are common drivers of natural convection, but they are not essential. Differences in dissolved salt, chemical composition, or suspended material can also create density contrasts. Cooling at the top of a liquid can drive sinking plumes just as heating from below can drive rising plumes. Density depends on temperature, pressure, and composition through an equation of state; the relevant dependence may be nonlinear. Fresh water, for instance, reaches its maximum density near \\(4\\,^\\circ\\mathrm{C}\\) at atmospheric pressure, so the assumption that warmer liquid is always lighter fails near freezing.\n\nConvection also need not involve a material that appears liquid on human timescales. Mantle rock behaves predominantly as a solid during an earthquake but can deform viscously over millions of years. Conversely, an ordinary fluid does not necessarily convect simply because it is heated. A stabilizing density arrangement, a sufficiently small container, strong viscous resistance, or weak heating may prevent buoyant motion from developing."
    },
    {
      "heading": "Governing equations and transport measures",
      "text": "The mathematical description of convection combines conservation of mass, momentum, and energy. For a constant-density, incompressible fluid, mass conservation gives\n\n\\[\\nabla\\cdot\\mathbf{u}=0,\\]\n\nwhere \\(\\mathbf{u}\\) is velocity. The Navier–Stokes equations describe the acceleration of that fluid under pressure gradients, viscous stresses, and body forces. The temperature equation, for constant properties and no internal heating, is\n\n\\[\\frac{\\partial T}{\\partial t} +\\mathbf{u}\\cdot\\nabla T =\\alpha\\nabla^2T,\\]\n\nwhere \\(\\alpha=k/(\\rho c_p)\\) is thermal diffusivity, \\(k\\) is thermal conductivity, \\(\\rho\\) is density, and \\(c_p\\) is specific heat capacity at constant pressure. The advective term couples the evolving temperature field to fluid motion.\n\nA widely used simplification is the Boussinesq approximation. Density is treated as constant except where its small variations generate buoyancy. If density decreases approximately linearly with temperature, the buoyancy acceleration relative to a reference state is\n\n\\[\\mathbf{b}=g\\beta(T-T_0)\\,\\hat{\\mathbf{z}},\\]\n\nwith upward unit vector \\(\\hat{\\mathbf{z}}\\), gravitational acceleration \\(g\\), and volumetric thermal expansion coefficient \\(\\beta\\). This approximation isolates the feedback in which temperature differences drive motion and motion redistributes temperature. It is unsuitable for flows spanning very large density contrasts, such as convection through much of a star.\n\nDimensionless numbers express the competition among transport processes. The Rayleigh number is\n\n\\[Ra=\\frac{g\\beta\\Delta T L^3}{\\nu\\alpha},\\]\n\nwhere \\(L\\) is a characteristic length and \\(\\nu\\) is kinematic viscosity. It measures buoyant driving relative to viscous and thermal diffusion. The Prandtl number, \\(Pr=\\nu/\\alpha\\), compares momentum diffusion with thermal diffusion. Near room conditions, air has \\(Pr\\) about \\(0.7\\), water about \\(7\\), and liquid metals often much less than one. These differences strongly affect the relative thicknesses of velocity and temperature boundary layers.\n\nFor imposed flows, the Reynolds number, \\(Re=UL/\\nu\\), compares inertia with viscosity, while the Péclet number, \\(Pe=UL/\\alpha=Re\\,Pr\\), compares advection with thermal diffusion. The Nusselt number, commonly written \\(Nu=hL/k\\), measures heat transfer relative to a conductive reference. For a horizontal layer with fixed temperature difference, it can be defined as\n\n\\[Nu=\\frac{\\langle q_z\\rangle L}{k\\Delta T}.\\]\n\nHere \\(\\langle q_z\\rangle\\) is the averaged upward heat flux. Pure conduction gives \\(Nu=1\\) in that configuration; convective transport generally increases it. These numbers do not by themselves determine a flow: geometry, boundary conditions, rotation, and material-property variations must also be specified."
    },
    {
      "heading": "Buoyancy, stability, and organized flow",
      "text": "Buoyancy describes the force arising when a parcel has a different density from the surrounding material in a gravitational field. A parcel displaced upward will continue rising if it is lighter than its new surroundings, but return toward its original position if it is heavier. This distinction defines unstable and stable stratification. The familiar expression “hot material rises” is therefore incomplete: motion depends on a density contrast with the surroundings, not on temperature alone. In a compressible fluid, a rising parcel also expands and cools, so its changing state must be considered.\n\nThe canonical laboratory problem is Rayleigh–Bénard convection: a horizontal fluid layer heated from below and cooled from above. In an ideal, horizontally unbounded Boussinesq layer with rigid, fixed-temperature upper and lower boundaries, the motionless conductive state becomes unstable at a Rayleigh number of approximately \\(1707.76\\). Below this threshold, sufficiently small disturbances decay. Above it, disturbances can extract enough gravitational potential energy from the unstable thermal arrangement to overcome viscous and diffusive losses. Different mechanical or thermal boundary conditions change the critical value.\n\nNear onset, convection often forms organized rolls with horizontal dimensions comparable to the layer depth. Other geometries and fluid properties permit hexagonal cells, spirals, or more complicated arrangements. These patterns are not universal geometric signatures of convection. Surface-tension gradients can produce visually similar cells through Marangoni convection, in which flow is driven along an interface toward regions of higher surface tension. In thin liquid layers with a free surface, thermocapillary forces may compete with or dominate buoyancy.\n\nAs driving increases, steady patterns can become time-dependent, interacting plumes appear, and the flow may become turbulent. Thin thermal boundary layers next to heated and cooled surfaces supply buoyant fluid to the interior. A rising plume entrains surrounding material, broadens, and changes temperature as it travels. Large-scale circulation can coexist with irregular small-scale motion, and its direction may fluctuate or reverse. The transition is not a single universal sequence: it depends on the Prandtl number, container shape, sidewalls, and imposed constraints.\n\nRotation and magnetic forces introduce further organizing effects. Rapid rotation makes the Coriolis effect important and can favor elongated structures aligned approximately with the rotation axis. In electrically conducting fluids, motion can interact with a magnetic field, which may suppress some motions while allowing others. These effects are essential in planetary cores and many astrophysical settings. They also demonstrate why convection cannot be characterized by a temperature difference alone: the same thermal forcing can produce very different flows under different mechanical conditions."
    },
    {
      "heading": "Atmospheric convection and cloud systems",
      "text": "In the atmosphere, pressure decreases with height, so rising air expands and cools approximately through an adiabatic process. Unsaturated air cools at the dry adiabatic lapse rate,\n\n\\[\\Gamma_d=\\frac{g}{c_p},\\]\n\napproximately \\(9.8\\ \\mathrm{K\\,km^{-1}}\\) under ordinary lower-atmospheric conditions. Whether a displaced parcel is buoyant depends on how its temperature and moisture compare with those of the environment at the same pressure. Potential temperature, the temperature a parcel would have after dry adiabatic adjustment to a reference pressure, is a convenient measure of dry static stability.\n\nMoisture changes this behavior substantially. When rising air reaches saturation, condensation releases latent heat, reducing its rate of cooling relative to unsaturated ascent. The saturated adiabatic lapse rate is not constant; values around \\(4\\)–\\(7\\ \\mathrm{K\\,km^{-1}}\\) are common in the lower and middle troposphere, depending on temperature and pressure. A layer can therefore resist unsaturated ascent yet support vigorous motion after saturation. Water vapor also affects density directly, so buoyancy calculations commonly use virtual temperature to represent the density effect of moist air.\n\nMeteorologists quantify positive parcel buoyancy using convective available potential energy, or CAPE:\n\n\\[\\mathrm{CAPE}= \\int_{z_{\\mathrm{LFC}}}^{z_{\\mathrm{EL}}} g\\frac{T_{v,p}-T_{v,e}}{T_{v,e}}\\,dz.\\]\n\nThe subscripts \\(p\\) and \\(e\\) denote parcel and environment; the integration extends from the level of free convection to the equilibrium level. CAPE has units of joules per kilogram. It is an idealized measure, not a direct prediction of storm intensity, because entrainment, water loading, pressure forces, and other processes modify real updrafts. Convective inhibition measures the negative buoyancy that must be overcome before sustained free ascent can occur.\n\nDaytime surface heating produces thermals and a turbulent atmospheric boundary layer, sometimes marked by shallow cumulus clouds. Deep moist convection can extend through most of the troposphere and generate heavy precipitation, lightning, and strong winds. Evaporation and melting of precipitation cool descending air, helping produce downdrafts and spreading surface cold pools. Their leading edges can lift additional warm air and initiate new cells. Vertical wind shear affects how updrafts and downdrafts interact and whether storms remain isolated or organize into longer-lived systems.\n\nConvection is also integral to climate. It redistributes energy and moisture vertically, influences cloud cover, and helps maintain the tropical atmospheric temperature profile. Large-scale circulations associated with the intertropical convergence zone and monsoons both organize and respond to convective heating. Individual updrafts are too small to be explicitly represented in many global models, making their collective effects an important source of uncertainty in simulations of rainfall and Earth’s energy budget."
    },
    {
      "heading": "Oceanic convection and double-diffusive processes",
      "text": "Ocean convection is controlled mainly by temperature, salinity, and pressure. Cooling usually increases seawater density, while adding salt also makes it denser. Surface cooling can therefore create water that sinks and mixes with the underlying ocean. Sea-ice formation can contribute because much of the dissolved salt is excluded from the growing ice, increasing the salinity of nearby seawater. Freshening from rainfall, river input, or melting ice often has the opposite effect, creating a buoyant surface layer that resists vertical mixing.\n\nDeep convection occurs intermittently in particular high-latitude regions, including parts of the Labrador, Irminger, and Nordic seas. The depth reached depends on winter heat loss, prior stratification, salinity, and the ocean’s circulation. Convection may ventilate layers hundreds of metres deep and, during strong events in suitable locations, extend beyond a kilometre. Around Antarctica, dense shelf waters and their subsequent descent along continental slopes also contribute to deep and bottom-water formation. These processes transport oxygen and dissolved carbon into the interior and alter the ocean’s capacity to store heat.\n\nLocal convective sinking is only one part of the global overturning circulation. Wind-driven transport, turbulent mixing, boundary currents, and density changes at the surface all contribute. The common image of a simple global conveyor belt captures some connections but obscures the three-dimensional pathways and multiple timescales involved. The carbon cycle is affected because physical transport works together with biological uptake and the sinking and decomposition of organic matter.\n\nA distinct class of instability, double-diffusive convection, arises when two density-controlling properties diffuse at different rates. Heat diffuses molecularly through seawater roughly one hundred times faster than dissolved salt. Warm, salty water over cooler, fresher water can be stable in its total density distribution yet unstable to narrow descending and ascending structures called salt fingers. A small downward displacement of warm, salty water loses its excess heat more rapidly than its excess salt; it consequently becomes denser and can continue descending.\n\nThe opposite arrangement—colder, fresher water over warmer, saltier water—can support a different double-diffusive regime, often termed diffusive convection. It may produce well-mixed layers separated by thin interfaces, creating a staircase-like vertical temperature and salinity profile. Such structures occur in several oceanic settings and influence vertical heat transport. Their formation illustrates a broader principle: static stability based on total density does not guarantee stability when different contributions to density are transported at unequal rates. Related effects occur in industrial solutions, magma, and stellar interiors."
    },
    {
      "heading": "Convection inside planets and stars",
      "text": "Inside Earth, convection operates in materials with radically different mechanical properties. The mantle consists mostly of solid silicate rock, but high temperature and sustained stress allow deformation by creep. Over geological timescales, this makes large-scale circulation possible. Heat comes from the planet’s gradual cooling, radioactive decay, and exchange with the core. Cold lithosphere entering subduction zones forms negatively buoyant downwellings, while hotter material rises elsewhere, sometimes in localized mantle plumes.\n\nMantle convection and plate tectonics are dynamically coupled rather than independent mechanisms. Tectonic plates form the cold upper boundary of the convecting system, and the sinking of dense slabs is an important driver of their movement. Mantle viscosity varies strongly with temperature, pressure, composition, and deformation mechanism. Mineral phase changes, especially near depths of about 410 and 660 kilometres, also influence flow. Seismic imaging shows that some slabs penetrate into the lower mantle, while others stagnate or spread near the transition zone; neither a completely isolated two-layer system nor uniform whole-mantle circulation describes every region.\n\nEarth’s outer core is a liquid iron-rich alloy. Its convection involves both thermal and compositional buoyancy. As the solid inner core grows, the remaining liquid becomes enriched in lighter components, providing a source of buoyancy. Rotation organizes the motion, and the movement of electrically conducting liquid sustains the geomagnetic field through the geodynamo. The balance among core cooling, heat conduction, and compositional driving remains important to reconstructions of Earth’s magnetic history.\n\nIn stars, convection occurs where radiation alone cannot carry the outward energy flux without producing an unstable temperature gradient. For a chemically uniform region, the Schwarzschild criterion states that instability arises when the logarithmic temperature gradient required for radiative transport exceeds the adiabatic gradient. Composition gradients modify this criterion because a displaced parcel can differ in mean molecular weight as well as temperature. Stellar convection therefore depends on opacity, energy generation, chemical structure, and the equation of state.\n\nThe Sun has an outer convection zone extending from approximately \\(0.71\\) solar radii to the visible surface. Near the surface, rising hot plasma and sinking cooler plasma create granulation: bright cells typically around a thousand kilometres across that evolve over several minutes. Other stars have different structures. Many low-mass main-sequence stars are fully convective, whereas massive main-sequence stars commonly have convective cores. Convection transports energy and mixes chemical elements, influencing stellar lifetimes and evolution. Most one-dimensional stellar models represent it using approximate prescriptions such as mixing-length theory rather than resolving the full flow."
    },
    {
      "heading": "Engineering heat transfer and technological applications",
      "text": "Engineering treatments of convection commonly describe the heat-transfer rate between a surface and a fluid through\n\n\\[\\dot Q=hA(T_s-T_\\infty),\\]\n\nwhere \\(A\\) is surface area, \\(T_s\\) is surface temperature, \\(T_\\infty\\) is an appropriate bulk-fluid or ambient temperature, and \\(h\\) is the convective heat-transfer coefficient. This relationship is often called Newton’s law of cooling. Unlike thermal conductivity, \\(h\\) is not an intrinsic material property: it depends on velocity, geometry, fluid properties, orientation, and the developing flow. It can also vary substantially over one surface.\n\nAt an ordinary impermeable solid wall, fluid has no velocity normal to the surface, and the no-slip condition usually makes its tangential velocity zero as well. Heat passes through the immediately adjacent fluid by molecular conduction. Bulk motion enhances overall transfer by bringing cooler fluid toward a hot surface and carrying warmed fluid away, steepening the near-wall temperature gradient. The boundary layer concept connects this local conductive process to the surrounding moving fluid. Velocity and thermal boundary layers need not have the same thickness.\n\nEmpirical and theoretical correlations express \\(Nu\\) in terms of parameters such as \\(Re\\), \\(Pr\\), and \\(Ra\\). For example, fully developed laminar flow in a circular tube has \\(Nu=3.66\\) for constant wall temperature and \\(Nu=4.36\\) for uniform wall heat flux, under standard idealizations including constant properties and negligible axial conduction. Entrance regions, roughness, variable properties, and buoyancy can alter these results. In turbulent pipe flow, enhanced mixing generally increases heat transfer, but pumping power also rises because of frictional pressure losses.\n\nApplications range from building ventilation to power generation. Passive electronic cooling uses fins and buoyant airflow, while forced-air or liquid-cooling systems increase transport when heat loads are larger. Heat exchangers place flowing streams on opposite sides of a conductive wall, allowing energy transfer without intentional mixing. Their performance depends not only on convective coefficients but also on wall resistance, fouling, flow arrangement, and the changing temperature difference along the device. Natural circulation loops can move coolant through density differences without a continuously operating pump.\n\nBoiling and condensation require additional distinctions because phase change carries latent heat. Bubble formation and motion can greatly enhance heat removal, but a continuous vapor film can instead reduce contact between liquid and a hot surface. Industrial furnaces combine convection with radiation, while casting and crystal growth couple fluid circulation to solidification and solute transport. In microgravity, the greatly reduced buoyancy drive suppresses ordinary natural convection, making imposed flow and surface-tension-driven motion relatively more important. Thus, successful heat-transfer models must identify the mechanism actually controlling the transport rather than assigning a single universal “convection coefficient.”"
    },
    {
      "heading": "Historical development and research methods",
      "text": "The scientific description of convection developed from studies of heat, fluid motion, and atmospheric circulation. In 1701, Isaac Newton published experiments associated with the cooling relation that now bears his name, although modern heat-transfer analysis distinguishes more carefully among radiation, conduction, and convection. In the nineteenth century, investigations of fluid motion and heat transport established the conservation equations and material laws that made quantitative analysis possible. James Thomson used the term “convection” in its scientific heat-transfer sense in 1834.\n\nAround 1900, Henri Bénard investigated cellular patterns in thin liquid layers heated from below. In 1916, Lord Rayleigh published a stability analysis of a horizontal fluid layer driven by buoyancy. Later work clarified that surface-tension gradients played an important role in Bénard’s original free-surface experiments. The distinction between buoyancy-driven Rayleigh–Bénard convection and thermocapillary Bénard–Marangoni convection became central to interpreting apparently similar cellular patterns.\n\nConvection also contributed to the development of modern ideas about nonlinear dynamics. In 1963, Edward Lorenz introduced a simplified three-variable model derived from equations for thermal convection. Its irregular, nonperiodic behavior demonstrated how deterministic equations can exhibit sensitive dependence on initial conditions. The Lorenz system is not a detailed weather model, but it became a foundational example in chaos theory and helped clarify why exact long-range prediction can fail even when the underlying equations are deterministic.\n\nLaboratory experiments use temperature-controlled plates, carefully characterized fluids, and controlled container geometry. Thermometers and heat-flux measurements determine global transport, while optical methods reveal flow structure. Shadowgraph and schlieren techniques visualize refractive-index variations associated with temperature or composition. Particle image velocimetry estimates velocity from the displacement of tracer particles between images. Liquid crystals, fluorescent dyes, and infrared imaging provide complementary temperature information, although each method has limitations related to calibration, access, response time, or disturbance of the flow.\n\nNumerical methods range from linear stability calculations to three-dimensional simulations. Direct numerical simulation resolves the dynamically important velocity and temperature scales without a turbulence model, but its computational cost grows rapidly as the separation between large and small scales increases. Large-eddy simulation resolves larger motions and models smaller ones, while engineering approaches often solve averaged equations with closure models. In planetary and stellar applications, the achievable parameters can remain far from natural conditions. Confidence therefore comes from comparisons among theory, laboratory experiments, simulations, and indirect observations rather than from any single method."
    },
    {
      "heading": "Turbulence, scaling laws, and unresolved problems",
      "text": "A central research problem is predicting how much heat a convecting system carries. For an idealized Rayleigh–Bénard cell, this means finding the dependence\n\n\\[Nu=f(Ra,Pr,\\text{geometry},\\text{boundary conditions}).\\]\n\nExperiments and simulations often find approximate power-law behavior over limited parameter ranges, but no single exponent describes every regime. Changes in boundary layers, plume emission, large-scale circulation, and material properties can all affect measured scaling. Sidewall conduction and imperfectly controlled plate temperatures can also influence apparent transport laws.\n\nAt very large Rayleigh number, a proposed “ultimate” regime involves turbulent boundary layers as well as a turbulent interior. Predictions associated with Robert Kraichnan and subsequent theories suggest stronger growth of heat transport with thermal forcing than in many lower-\\(Ra\\) regimes, with logarithmic corrections in some formulations. Identifying the transition experimentally remains demanding. Reported changes in scaling must be separated from effects of roughness, non-Boussinesq behavior, finite plate conductivity, and differences in aspect ratio. There is no universally applicable transition threshold independent of the system.\n\nTurbulence creates a closure problem because the mean transport depends on correlations between fluctuations. In a Boussinesq fluid, the vertically transported thermal energy associated with fluctuating motion contains the term\n\n\\[F_{\\mathrm{conv}}=\\rho c_p\\langle w'T'\\rangle,\\]\n\nwhere \\(w'\\) and \\(T'\\) are vertical-velocity and temperature fluctuations. Knowing the mean temperature and mean velocity is not enough to determine this correlation. Hot upflows and cold downflows both usually contribute positive upward heat transport, but coherent plumes, intermittent events, and local countergradient fluxes complicate simple diffusion-like representations.\n\nExtrapolation is especially difficult in natural systems. Mantle flow includes strongly temperature-dependent viscosity and deforming plates; atmospheric convection includes moisture, precipitation, and radiation; stellar convection includes compressibility, ionization, rotation, and magnetic forces. These are not small corrections to one universal laboratory problem. Reduced models must preserve the mechanisms controlling the quantity of interest, whether that quantity is rainfall, planetary cooling, chemical mixing, or magnetic-field generation.\n\nAnother active problem is how small-scale convection interacts with larger-scale organization. Storms can aggregate into extensive cloud systems; oceanic mixing modifies the stratification that controls subsequent convection; and convective turbulence can generate or alter large-scale flows. In numerical models, unresolved motions are represented through parameterization, whose assumptions may change validity as model resolution changes. Current research combines high-resolution calculations, targeted observations, laboratory measurements, and statistical methods to determine which relationships remain transferable across these different scales and physical regimes."
    }
  ],
  "links": [
    "heat",
    "density",
    "ocean",
    "viscosity",
    "electromagnetic-radiation",
    "advection",
    "diffusion",
    "temperature",
    "pressure",
    "equation-of-state",
    "water",
    "mass",
    "momentum",
    "energy",
    "navier-stokes-equations",
    "thermal-conductivity",
    "heat-capacity",
    "prandtl-number",
    "reynolds-number",
    "coriolis-effect",
    "magnetic-field",
    "atmosphere-of-earth",
    "adiabatic-process",
    "condensation",
    "latent-heat",
    "climate",
    "earth-energy-budget",
    "carbon-cycle",
    "earth",
    "plate-tectonics",
    "sun",
    "boundary-layer",
    "isaac-newton",
    "chaos-theory",
    "turbulence",
    "parameterization"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}