Mixing-Length Theory

Below the visible surface of the Sun, the gradient a star would need to carry its luminosity by radiation alone grows steeper than the gas can bear. The Schwarzschild and Ledoux criteria announce the instability, and the gas answers with turbulence: packets of plasma lift off, rise, cool, and dissolve, ferrying heat as bulk motion. Computing that turbulence honestly would mean solving three-dimensional Navier–Stokes flow coupled to radiation across a range of scales no one-dimensional model can hold. Mixing-length theory is the pragmatic escape. In the form Ludwig Prandtl sketched for laboratory flow and Erika Böhm-Vitense forged into a stellar tool in 1958, it replaces the turbulent cascade with a single caricature: a parcel rises one distance ℓ — the mixing length — almost adiabatically, then merges with its surroundings and gives up its excess heat. The contrast built over ℓ drives both the parcel's buoyant acceleration and the enthalpy it carries, so a few lines of algebra convert it into a convective velocity and a flux. The whole apparatus rides on one number, α = ℓ/H_P. It cannot be derived within the theory; it is calibrated — usually α ≈ 1.5–2.0 — by demanding that a one-solar-mass model reach the observed solar radius and luminosity at the solar age, then carried across the Hertzsprung–Russell diagram. What MLT delivers, at every point of a convection zone, is the actual gradient ∇ between ∇_ad and ∇_rad, closing the stellar-structure equations. It ignores the eddy spectrum, overshoot, and non-locality, and works anyway — which is why every stellar-evolution code still has it at its heart.

Key takeaways

Mixing-length theory (MLT) is a local, phenomenological closure for the convective flux once the Schwarzschild or Ledoux criterion signals instability, standing in for the intractable hydrodynamics of stellar turbulence.
A parcel rises or sinks one mixing length ℓ = α H_P nearly adiabatically before dissolving and depositing its heat; α ≈ 1.5–2.0 is calibrated against the Sun, not derived.
The convective flux depends only on the superadiabatic excess ∇ − ∇_ad, and it does so as a steep three-halves power, F_conv ∝ (∇ − ∇_ad)^{3/2}.
In stellar interiors the density is so high that an excess of order 10⁻⁶ carries the entire luminosity, so ∇ ≈ ∇_ad and deep convection is essentially adiabatic and nearly independent of α.
MLT closes the structure equations by supplying ∇, interpolating between ∇_rad (inefficient, near-surface) and ∇_ad (efficient, deep) through a cubic in the velocity.
Key idea: model turbulent convection as parcels that rise one mixing length ℓ = α H_P nearly adiabatically, then dissolve; balancing buoyancy against inertia fixes a convective velocity and flux that depend only on the superadiabatic excess ∇ − ∇_ad.

Where this lives today

Every modern stellar-evolution code — MESA, the Geneva and Padova grids, YREC, GARSTEC — carries mixing-length theory at its core, because nothing else is both cheap enough to run millions of models and good enough to match real stars. The frontier is not whether to use MLT but how to calibrate it. Three-dimensional radiation-hydrodynamics simulations from the Copenhagen and Stagger groups now resolve the granulation MLT only caricatures; matching them to one-dimensional envelopes yields an effective α that varies across the Hertzsprung–Russell diagram rather than holding at the solar value. Asteroseismology has sharpened the stakes: Kepler and TESS measure frequencies to parts in $10^4$ , and the near-surface layers MLT treats crudely produce the notorious "surface-effect" offsets that must be corrected before precise ages and radii emerge. Meanwhile its blind spots — no overshoot beyond formal boundaries — remain a leading uncertainty in core masses, massive-star lifetimes, and the white-dwarf initial–final mass relation, driving work on non-local and time-dependent models that aim to discipline α.

The physical picture

Watch a pot of water just before it boils. Heat does not crawl up molecule by molecule; the fluid organizes into rolls, hot packets rising and cool sinking, and the transport rate jumps by orders of magnitude once convection switches on. Inside a star the same happens wherever the radiative gradient the luminosity demands exceeds the adiabatic gradient. MLT poses one blunt question: how far does a rising packet travel before it blends in? Its one-word answer — about a mixing length ℓ — is the whole theory, because everything follows from what the packet does over it. What the packet does is stay hot. As it rises into lower pressure it expands and cools, but along the shallower adiabat ∇_ad rather than the steeper actual gradient ∇, so it stays a little warmer than its surroundings — by an amount that grows with the excess ∇ − ∇_ad and with height. A warmer packet is lighter, buoyancy pushes it on, and the excess heat it carries is the convective flux. The first figure freezes that ascent and the contrast it accumulates.

Figure

A parcel forms in pressure balance with its surroundings, then rises a mixing length ℓ = α H_P. It follows the shallow adiabat ∇_ad while the background follows the steeper gradient ∇, so it arrives hotter by ΔT = (T divided by H_P) times (∇ − ∇_ad) times ℓ over 2. Being hotter it is lighter, and the buoyancy g δ ΔT over T drives it onward.

The efficiency of this transport is a tug of war between holding heat and leaking it. Deep inside the Sun the density and opacity are enormous, the parcel is optically thick and adiabatic, and a whisper of superadiabaticity moves a river of flux. Near the surface the gas thins, the parcel radiates freely, and only a large contrast pushes the same energy — convection turns inefficient and ∇ climbs toward ∇_rad. That contrast is the most important quantitative fact in the theory, and the second figure makes it concrete. The excess ∇ − ∇_ad is one part in a million through the bulk of the convection zone and detonates to order unity only in a wafer-thin surface layer. Everything MLT gets right about stellar radii is decided in that spike; everything it gets right about deep interiors follows from the flatness elsewhere, where the answer barely knows what α is.

Figure

The superadiabatic excess ∇ − ∇_ad through the solar convection zone on a logarithmic scale. Across the deep bulk from the base at r ≈ 0.71 R_sun it sits near 10⁻⁶, so the gas is adiabatic to six digits and α is irrelevant. Only in the thin surface layer near r = R_sun does it climb to order unity, where convection is inefficient and α fixes the star's radius.

The core equations

Convection is discussed in gradients, not temperatures. The actual, adiabatic, and radiative logarithmic gradients are the three characters, and MLT is a rule for placing the first between the other two.

\nabla\equiv\frac{\mathrm{d}\ln T}{\mathrm{d}\ln P},\qquad \nabla_{\mathrm{ad}}=\left(\frac{\partial\ln T}{\partial\ln P}\right)_{s},\qquad \nabla_{\mathrm{rad}}=\frac{3}{16\pi acG}\frac{\kappa L P}{m\,T^4} \qquad (1)

where: $\nabla$ — actual logarithmic temperature gradient; $\nabla_{\mathrm{ad}}$ — adiabatic gradient at constant entropy $s$ (≈ 0.4 for an ideal monatomic gas); $\nabla_{\mathrm{rad}}$ — gradient radiation alone would require; $\kappa$ — Rosseland opacity; $L$ — local luminosity; $m$ — interior mass; $a=4\sigma/c$ . Convection switches on when $\nabla_{\mathrm{rad}}>\nabla_{\mathrm{ad}}$ (Schwarzschild), and MLT then finds $\nabla$ between them. The mixing length needs a yardstick, and the only natural one is the pressure scale height — the distance over which pressure falls by a factor $e$ . It follows in one line from hydrostatic equilibrium.

H_P=-\frac{\mathrm{d}r}{\mathrm{d}\ln P}=\frac{P}{\rho g},\qquad g=\frac{Gm}{r^2} \qquad (2)

where: $H_P$ — pressure scale height; $\rho$ — local mass density; $g$ — local gravity. Because thermodynamic quantities change appreciably over $H_P$ , a parcel can keep its identity for at most a scale height — which is why $\ell$ is written as a multiple of it. The theory's one free parameter enters here, and only here — all of MLT's ignorance about the turbulent cascade compressed into the dimensionless ratio α.

\ell=\alpha\,H_P,\qquad \alpha\approx1.5\text{–}2.0\ \ (\text{solar-calibrated}) \qquad (3)

where: $\ell$ — mixing length, the distance a parcel travels before merging; $\alpha$ — mixing-length parameter, fixed by matching the Sun. It is a fitted number, not a prediction — the honest confession that we cannot compute $\ell$ from first principles. As a parcel rises through $\ell$ it separates thermally from its surroundings at a rate set by the excess. Integrating over the ascent, with a mean displacement of half a mixing length, gives the driving temperature contrast.

\Delta T=\left(\left.\frac{\mathrm{d}T}{\mathrm{d}r}\right|_{\mathrm{par}}-\left.\frac{\mathrm{d}T}{\mathrm{d}r}\right|_{\mathrm{env}}\right)\frac{\ell}{2}=\frac{T}{H_P}\,(\nabla-\nabla_{\mathrm{ad}})\,\frac{\ell}{2} \qquad (4)

where: $\Delta T$ — mean temperature excess of parcel over surroundings; $\nabla-\nabla_{\mathrm{ad}}$ — superadiabatic excess. The parcel follows $\nabla_{\mathrm{ad}}$ , the background the steeper $\nabla$ , and the gap accumulates with height — no excess, no convection. A hotter parcel is lighter, so buoyancy accelerates it. Equating the buoyant work over the mixing length to the kinetic energy — half lost to drag — fixes the convective speed.

v_{\mathrm{conv}}^2\simeq g\,\delta\,(\nabla-\nabla_{\mathrm{ad}})\,\frac{\ell^2}{8\,H_P} \qquad (5)

where: $v_{\mathrm{conv}}$ — characteristic convective speed; $\delta=-(\partial\ln\rho/\partial\ln T)_P$ — thermal-expansion exponent ( $\delta=1$ for an ideal gas). The speed grows only as the square root of the excess, far more gently than the temperature contrast. The convective flux is the excess enthalpy each parcel carries, $c_P\Delta T$ , times the mass flux $\rho\,v_{\mathrm{conv}}$ . Since $\Delta T\propto(\nabla-\nabla_{\mathrm{ad}})$ and $v_{\mathrm{conv}}\propto(\nabla-\nabla_{\mathrm{ad}})^{1/2}$ , their product carries the three-halves power.

F_{\mathrm{conv}}=\rho\,c_P\,v_{\mathrm{conv}}\,\Delta T=\rho\,c_P\,T\sqrt{g\delta}\,\frac{\ell^2}{4\sqrt{2}\,H_P^{3/2}}\,(\nabla-\nabla_{\mathrm{ad}})^{3/2} \qquad (6)

where: $F_{\mathrm{conv}}$ — convective energy flux; $c_P$ — specific heat at constant pressure. The $(\nabla-\nabla_{\mathrm{ad}})^{3/2}$ dependence is why a minuscule excess suffices in dense gas — the prefactor $\rho c_P T$ is enormous deep down. Finally, the closure. Radiation and convection together carry all the luminosity crossing $r$ , the radiative part driven by the same gradient $\nabla$ . Setting the sum equal to the required flux is one nonlinear equation for $\nabla$ .

F_{\mathrm{rad}}+F_{\mathrm{conv}}=\frac{L}{4\pi r^2},\qquad F_{\mathrm{rad}}=\frac{4acG}{3}\frac{T^4 m}{\kappa P r^2}\,\nabla,\qquad \nabla_{\mathrm{ad}}<\nabla<\nabla_{\mathrm{rad}} \qquad (7)

where: $F_{\mathrm{rad}}$ — radiative flux, linear in $\nabla$ ; $L/4\pi r^2$ — total flux to be transported. Solving this for $\nabla$ is the entire job of MLT inside a convection zone, and the answer is guaranteed to fall between the two references.

Derivation AThe convective velocity and the three-halves flux law

This derivation turns the picture of a buoyant, heat-carrying parcel into the two results MLT feeds to a stellar code: the convective velocity and the flux. We start from the buoyancy on a parcel that has risen against the superadiabatic background, extract its speed from a work–energy balance over one mixing length, and multiply the enthalpy it carries by that speed. The reward is the theory's fingerprint — the flux depends on the excess not linearly but as its three-halves power, which is why deep convection needs almost no superadiabaticity at all.

Step 1. A parcel displaced upward is hotter by $\Delta T$ , hence less dense, $\Delta\rho/\rho=-\delta\,\Delta T/T$ , so the buoyancy per unit mass is $f_{\mathrm{buoy}}=g\,\delta\,\Delta T/T$ . Why: buoyancy drives convection, and it is set entirely by the density deficit the temperature excess produces.

Step 2. Insert the mean contrast from equation (4) to get $f_{\mathrm{buoy}}=\tfrac{1}{2}\,g\,\delta\,(\nabla-\nabla_{\mathrm{ad}})\,\ell/H_P$ . Why: this recasts the force in terms of the excess and geometry, eliminating the unknowable $\Delta T$ .

Step 3. Apply work–energy over the ascent, $\tfrac12 v_{\mathrm{conv}}^2=\tfrac12\langle f_{\mathrm{buoy}}\rangle\,\ell$ , keeping half the buoyant work as kinetic energy and losing the rest to drag. This gives $v_{\mathrm{conv}}^2\simeq g\,\delta\,(\nabla-\nabla_{\mathrm{ad}})\,\ell^2/(8H_P)$ . Why: the one-half drag factor is an order-unity fudge MLT hides inside α, absorbed by calibration.

Step 4. The flux is the excess enthalpy $c_P\Delta T$ times the mass flux $\rho v_{\mathrm{conv}}$ : $F_{\mathrm{conv}}=\rho c_P v_{\mathrm{conv}}\Delta T$ . Since $\Delta T\propto(\nabla-\nabla_{\mathrm{ad}})$ and $v_{\mathrm{conv}}\propto(\nabla-\nabla_{\mathrm{ad}})^{1/2}$ , the product scales as $(\nabla-\nabla_{\mathrm{ad}})^{3/2}$ . Why: this steep power is the load-bearing result — a $10^{-6}$ excess carries the full solar luminosity deep down. *The exponent 3/2 = 1 (from ΔT) + 1/2 (from v) follows directly by substitution, and the full prefactor is confirmed by direct substitution.*

Go deeper · Derivation B · 1 minThe cubic that closes the stellar structure equations

Derivation BThe cubic that closes the stellar structure equations

The aim is to show how the flux balance of equation (7) becomes, after non-dimensionalization, a single cubic whose real positive root is the convective velocity — and hence ∇. This is the form that lives inside stellar-evolution codes: not a picture of blobs, but one algebraic equation solved at every mesh point in every timestep. We introduce a dimensionless velocity, an inefficiency measuring the parcel's radiative losses, and a driving term for how superadiabatic the radiative gradient wants to be, then assemble the standard cubic and read off its two limits.

Step 1. Define reduced variables: a scaled velocity $\xi\propto v_{\mathrm{conv}}$ , an inefficiency $\zeta\propto\Gamma^{-1}$ ( $\Gamma$ = heat retained over heat radiated), and a driving term $W\propto(\nabla_{\mathrm{rad}}-\nabla_{\mathrm{ad}})$ measuring the available superadiabaticity. Why: dimensionless variables strip out the stellar quantities and leave one numerical relation of the same shape in every star.

Step 2. Write the two fluxes in these variables: $F_{\mathrm{conv}}\propto\xi^3$ (velocity squared times $\Delta T$ , both tied to the excess), while the radiative flux carries the remainder, $F_{\mathrm{rad}}\propto(\nabla_{\mathrm{rad}}-\nabla)$ . Why: routing both channels through $\xi$ lets a single equation, not a coupled pair, fix the gradient.

Step 3. Impose $F_{\mathrm{rad}}+F_{\mathrm{conv}}=L/4\pi r^2$ and eliminate $\nabla$ for $\xi$ ; collecting terms gives the standard MLT cubic

\xi^3+\frac{\xi^2}{\zeta}+\frac{\xi}{\zeta}\Big(1+\frac{1}{\zeta}\Big)-\frac{W}{\zeta}=0

whose unique real positive root gives the convective velocity. Why: a cubic is the lowest-order polynomial that smoothly links the efficient and inefficient regimes with one physical root, which is why this form recurs from Böhm-Vitense to modern codes.

Step 4. Recover the gradient, $\nabla=\nabla_{\mathrm{ad}}+(\nabla_{\mathrm{rad}}-\nabla_{\mathrm{ad}})\,f(\xi)/[1+f(\xi)]$ . Efficient limit $\xi\to\infty$ gives $\nabla\to\nabla_{\mathrm{ad}}$ ; inefficient limit $\xi\to0$ gives $\nabla\to\nabla_{\mathrm{rad}}$ . Why: these limits are the sanity check — the deep interior must be adiabatic and the surface must revert to radiation, and the cubic delivers both. *The two limiting roots — $\nabla\to\nabla_{\mathrm{ad}}$ in the efficient limit and $\nabla\to\nabla_{\mathrm{rad}}$ in the inefficient limit — are confirmed by taking $\zeta\to0$ and $\zeta\to\infty$ respectively in the cubic.*

Go deeper · Derivation C · 1 minWhy the deep interior is adiabatic to one part in a million

Derivation CWhy the deep interior is adiabatic to one part in a million

This derivation puts a number on the theory's most surprising claim: that through the bulk of the solar convection zone the gas is superadiabatic by only one part in a million, so ∇ = ∇_ad to six decimals and α scarcely matters there. The argument is a single inversion: demand that convection carry the full solar flux at the base of the zone, solve the three-halves law for the excess, and plug in interior conditions. The tininess is not an approximation we impose — it is forced by the enormous prefactor ρ c_P T √(g H_P) that multiplies the excess.

Step 1. At the base of the solar convection zone ( $r\approx0.71\,R_\odot\approx4.9\times10^{8}\,\mathrm{m}$ ) essentially the whole luminosity is convective, so $F_{\mathrm{conv}}\approx L_\odot/(4\pi r^2)\approx1.2\times10^{8}\,\mathrm{W\,m^{-2}}$ . Why: the flux the gas must carry is the boundary condition that pins the required superadiabaticity.

Step 2. Invert equation (6): $(\nabla-\nabla_{\mathrm{ad}})^{3/2}=F_{\mathrm{conv}}\,/\,[\rho c_P T\sqrt{g\delta}\,\ell^2/(4\sqrt{2}\,H_P^{3/2})]$ . Why: this isolates the excess and shows it is suppressed as the two-thirds power of a tiny flux-to-enthalpy ratio.

Step 3. Insert base-of-zone values ( $\rho\approx200\,\mathrm{kg\,m^{-3}}$ , $T\approx2\times10^{6}\,\mathrm{K}$ , $g\approx530\,\mathrm{m\,s^{-2}}$ , $c_P\approx3.4\times10^{4}\,\mathrm{J\,kg^{-1}K^{-1}}$ , $H_P\approx5.6\times10^{7}\,\mathrm{m}$ , $\delta=1$ , $\alpha=1.5$ ). The prefactor is about $9.5\times10^{17}\,\mathrm{W\,m^{-2}}$ . Why: this gigantic prefactor swallows the flux and leaves almost nothing for the excess to supply.

Step 4. Therefore $(\nabla-\nabla_{\mathrm{ad}})^{3/2}\approx1.3\times10^{-10}$ , so $\nabla-\nabla_{\mathrm{ad}}\approx3\times10^{-7}$ , of order $10^{-6}$ . Why: at this level $\nabla$ is indistinguishable from $\nabla_{\mathrm{ad}}$ , so deep stratification is set by thermodynamics alone and is blind to α — which is why α is calibrated once at the surface and trusted throughout. *Direct evaluation with these inputs gives $\nabla - \nabla_{\mathrm{ad}} = 2.6\times10^{-7}$ ; the result is sensitive to the assumed base values but robustly of order $10^{-6}$ or smaller.*

Limiting cases

MLT is best understood through its limits, governed by the efficiency Γ — heat a parcel retains over what it radiates in a mixing length. The theory interpolates ∇ between the two references as Γ runs from large to small.

Case	Result	Meaning
Efficient, deep interior	$\Gamma\gg1\ \Rightarrow\ \nabla\to\nabla_{\mathrm{ad}}$	High density and opacity trap the heat; convection carries nearly the whole luminosity with an excess of order $10^{-6}$ . Stratification is adiabatic and almost independent of α.
Inefficient, surface layers	$\Gamma\ll1\ \Rightarrow\ \nabla\to\nabla_{\mathrm{rad}}$	Low density lets parcels radiate their heat away fast, so convection is weak and $\nabla$ reverts toward the radiative value. This thin layer is where α bites and the radius is decided.
Marginal stability at a boundary	$\nabla_{\mathrm{rad}}\to\nabla_{\mathrm{ad}}\ \Rightarrow\ F_{\mathrm{conv}}\to0$	At the zone edge the excess vanishes and the velocity goes smoothly to zero. MLT predicts an abrupt edge with no penetration — a known failing, patched with ad hoc overshoot.
Highly subsonic flow	$v_{\mathrm{conv}}\ll c_s$	Interior speeds are $10^{-3}$ – $10^{-4}$ of the sound speed, justifying the neglect of kinetic energy relative to enthalpy transport in the standard flux formula.

Common trapConfusing a nearly-adiabatic gradient with a small convective flux

The convective flux scales as the excess $(\nabla-\nabla_{\mathrm{ad}})^{3/2}$ , not as $\nabla$ itself. Deep in a star that excess is one part in a million, so it is tempting to conclude the flux is tiny — but the opposite is true: convection there carries essentially the entire luminosity, because the enormous prefactor $\rho c_P T\sqrt{gH_P}$ turns a microscopic excess into a colossal flux. "Nearly adiabatic gradient" means the driving is barely superadiabatic, not that little heat moves. A related trap: α is a calibrated fudge, not a property of the gas — MLT quantifies efficiency but cannot give ℓ from first principles, and the α that fits the Sun needs adjustment elsewhere on the Hertzsprung–Russell diagram.

Worked examples

1 · Superadiabaticity and convective velocity at the base of the solar convection zone (fully worked)

At $r\approx0.71\,R_\odot$ take $\rho\approx200\,\mathrm{kg\,m^{-3}}$ , $T\approx2\times10^{6}\,\mathrm{K}$ , $g\approx530\,\mathrm{m\,s^{-2}}$ , $c_P\approx3.4\times10^{4}\,\mathrm{J\,kg^{-1}K^{-1}}$ , $\delta=1$ , $\alpha=1.5$ . Find the excess and convective speed needed to carry the solar luminosity, and compare with the sound speed.

Step 1. Fix the scale height and mixing length. With $P\approx6\times10^{12}\,\mathrm{Pa}$ , equation (2) gives $H_P=P/\rho g\approx5.6\times10^{7}\,\mathrm{m}$ , so $\ell=\alpha H_P\approx8.4\times10^{7}\,\mathrm{m}$ . H_P ≈ 5.6×10⁷ m (about 56,000 km).

Step 2. Fix the flux the gas must carry. Almost all of $L_\odot=3.83\times10^{26}\,\mathrm{W}$ is convective here, so $F_{\mathrm{conv}}=L_\odot/(4\pi r^2)\approx1.2\times10^{8}\,\mathrm{W\,m^{-2}}$ . F_conv ≈ 1.2×10⁸ W m⁻².

Step 3. Solve equation (6) for the excess. The prefactor $\rho c_P T\sqrt{g\delta}\,\ell^2/(4\sqrt2\,H_P^{3/2})\approx9.5\times10^{17}\,\mathrm{W\,m^{-2}}$ , so $(\nabla-\nabla_{\mathrm{ad}})^{3/2}\approx1.3\times10^{-10}$ and $\nabla-\nabla_{\mathrm{ad}}\approx2.6\times10^{-7}$ . The excess is ∇ − ∇_ad ≈ 3×10⁻⁷, of order 10⁻⁶ — the deep zone is adiabatic to six digits.

Step 4. Get the convective speed from equation (5): $v_{\mathrm{conv}}^2=g\delta(\nabla-\nabla_{\mathrm{ad}})\ell^2/(8H_P)\approx2.2\times10^{3}\,\mathrm{m^2\,s^{-2}}$ , so $v_{\mathrm{conv}}\approx47\,\mathrm{m\,s^{-1}}$ ; the sound speed is $c_s=\sqrt{\tfrac53 P/\rho}\approx2.2\times10^{5}\,\mathrm{m\,s^{-1}}$ . v_conv ≈ 47 m s⁻¹, Mach ≈ 2×10⁻⁴ — deeply subsonic, as MLT assumes. A part-in-a-million excess and a plodding 50 m/s flow move the Sun's entire power output, because the medium is dense, hot, and vast — which is why α barely matters deep down.

2 · The photospheric scale height and why α bites at the surface

Estimate the pressure scale height at the solar photosphere ( $T\approx5800\,\mathrm{K}$ , $g=274\,\mathrm{m\,s^{-2}}$ , $\mu\approx1.3$ for the mostly neutral gas) and the mixing length, and contrast the convective regime with the deep interior.

Step 1. For an ideal gas $H_P=kT/(\mu m_H g)\approx(1.38\times10^{-23}\times5800)/(1.3\times1.67\times10^{-27}\times274)\approx1.35\times10^{5}\,\mathrm{m}$ . H_P ≈ 135 km — about 400× smaller than at the base of the zone.

Step 2. With $\alpha\approx1.6$ , $\ell=\alpha H_P\approx215\,\mathrm{km}$ , and observed solar granules are roughly $1000\,\mathrm{km}$ across — a few scale heights, consistent with the local $H_P$ setting the dominant cell size. ℓ ≈ 215 km, comparable to the observed granulation scale.

Step 3. Here the density is tiny ( $\sim10^{-4}\,\mathrm{kg\,m^{-3}}$ ), so parcels radiate heat almost as fast as they build it: convection is inefficient, the excess climbs to order unity, and granule velocities reach $1$ – $2\,\mathrm{km\,s^{-1}}$ at Mach a few tenths — near-sonic, unlike Example 1. Because ∇ − ∇_ad is large and α-dependent here, this wafer-thin layer, not the vast interior, is what calibrating α against the solar radius adjusts.

Physics of the portrait

Mixing-length theory carries two names: the engineer who invented the mixing length for pipes and jets, and the astronomer who made it the standard tool of stellar structure — the concept is Prandtl's, the stellar theory every code runs is Erika Böhm-Vitense's.

Origins · 1 minErika Böhm-Vitense and how this was discovered

Erika Böhm-Vitense

1923–2017 · the standard theory of stellar convection

Erika Böhm-Vitense was a German-American astrophysicist whose 1958 paper in the Zeitschrift für Astrophysik gave mixing-length convection the concrete, calculable form stellar-evolution codes have used ever since. The mixing length — a Mischungsweg, the distance a turbulent parcel travels before losing its identity — was Prandtl's, coined in 1925 for laboratory shear flow, and Biermann had applied convective ideas to stars in the 1930s. But it was Böhm-Vitense who wrote down the full local prescription: the parcel following the adiabat against a superadiabatic background, the velocity from a work balance, the flux as the three-halves power of the excess, and the radiative leakage that makes surface convection inefficient — all closed by α = ℓ/H_P. Her formulation made it possible, for the first time, to compute the gradient throughout a convective envelope and so predict the radii and effective temperatures of cool stars. Working for decades at the University of Washington, often with her husband Karl-Heinz Böhm, she also contributed to stellar atmospheres, cool-star chromospheres, and Cepheids. The parameter that bears the imprint of her theory remains, sixty years on, the most-quoted free parameter in all of stellar modeling.

History

1925

— Prandtl introduces the mixing length (Mischungsweg) for turbulent shear flow

1932

— Biermann applies mixing-length convective transport to stellar interiors

1958

— Böhm-Vitense publishes the standard local MLT still used by stellar codes

1968

— Cox and Giuli codify the MLT cubic in Principles of Stellar Structure

1991

— Canuto and Mazzitelli propose a full-spectrum-of-turbulence alternative to single-scale MLT

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §10.4 · Kippenhahn, Weigert & Weiss, Stellar Structure and Evolution, 2nd ed., Ch. 7 · Hansen, Kawaler & Trimble, Stellar Interiors, 2nd ed., §5.1 · Böhm-Vitense, E. 1958, Zeitschrift für Astrophysik, 46, 108 · Cox & Giuli, Principles of Stellar Structure, Ch. 14. Key self-consistency checks: (i) $H_P=P/\rho g$ follows from hydrostatic equilibrium; (ii) the $(\nabla-\nabla_{\mathrm{ad}})^{3/2}$ scaling of $F_{\mathrm{conv}}$ is the product of $\Delta T\propto(\nabla-\nabla_{\mathrm{ad}})$ and $v_{\mathrm{conv}}\propto(\nabla-\nabla_{\mathrm{ad}})^{1/2}$ ; (iii) the base-of-zone excess is $\approx3\times10^{-7}$ and the convective speed $\approx47\,\mathrm{m\,s^{-1}}$ ; (iv) the cubic yields $\nabla\to\nabla_{\mathrm{ad}}$ in the efficient limit and $\nabla\to\nabla_{\mathrm{rad}}$ in the inefficient limit. The calibrated $\alpha\approx1.5$ – $2.0$ is a fitted range, not a prediction.

Prerequisites. Schwarzschild and Ledoux Criteria for Convection · Adiabatic Temperature Gradient · Radiative Temperature Gradient · Equation of State for Stellar Matter · Opacity and Radiative Transfer in Stars. Leads to. Convective Energy Transport · Equations of Stellar Structure · Hydrostatic Equilibrium in Stars.

Ask the tutor: Want to see how the MLT cubic is solved at a mesh point inside a stellar model, or how the calibrated α changes for a red giant versus the Sun?

What is the mixing length, physically?

It is the average distance a rising or sinking parcel travels before merging with its surroundings and giving up its excess heat — the stellar analogue of Prandtl's Mischungsweg, representing the scale of the dominant eddies. Because a stratified gas changes appreciably over a pressure scale height $H_P$ , the mixing length is written as a multiple of it, $\ell=\alpha H_P$ , with α of order one. No parcel truly travels a fixed distance and then dissolves; the mixing length is a modeling device that captures the right scaling while hiding our inability to compute real turbulence.

Why does the theory have only one free parameter, and is it universal?

All of MLT's ignorance about the turbulent cascade — eddy sizes, the asymmetry between up- and down-flows, the details of merging — is compressed into the single number α = ℓ/H_P, the theory's great virtue and its great limitation. α is fixed by requiring a one-solar-mass model to reproduce the Sun's radius and luminosity at the solar age, giving α ≈ 1.5–2.0. It is not universal: simulations and asteroseismic data show the effective α drifts across the Hertzsprung–Russell diagram, smaller for hotter, evolved, or metal-poor stars. Using the solar α everywhere is a convenient approximation, not a law.

If deep convection is nearly adiabatic anyway, why bother with MLT there?

Because "nearly" still has to be quantified, and the theory must join the adiabatic interior to the non-adiabatic surface as one solution. In the deep interior MLT gives $\nabla=\nabla_{\mathrm{ad}}$ to one part in a million — exactly what a structure code needs to integrate the temperature outward without resolving a superadiabatic term it cannot. Just as important, MLT supplies the convective velocity and turnover time, which set the mixing of composition and hence the fuel supply to nuclear burning. So even where the gradient is trivially adiabatic, the theory does quantitative work on efficiency, mixing, and boundaries.

What does standard MLT get wrong?

It replaces a spectrum of eddies with one length and one velocity, so it has no turbulent cascade, no distinction between broad gentle upflows and narrow fast downdrafts, and no memory — it is strictly local and time-independent. Its most consequential failure is at convective boundaries: it predicts convection stops abruptly where the excess vanishes, with no overshoot into the stable layers beyond. Real convection overshoots, mixing extra fuel into burning cores and changing lifetimes and remnant masses. These gaps are why codes bolt on ad hoc overshoot parameters and why the surface-effect offsets exist — MLT is a calibrated first approximation, to be corrected against data and simulations.

How exactly is α calibrated to the Sun?

You build a one-solar-mass model with the best opacities, equation of state, and composition, and evolve it for the solar age of 4.57 billion years. Its radius and luminosity depend on α mainly through the thin, inefficient, strongly superadiabatic layer below the photosphere, whose structure sets where the surface sits. You adjust α until the model reproduces the measured solar radius and luminosity together (with the surface metal-to-hydrogen ratio). The value that does this, typically α ≈ 1.7–1.9, is then adopted for other stars. It works because the deep interior is insensitive to α, so the calibration is effectively a one-parameter fit to the surface layer.

Why do state-of-the-art 3D simulations still keep MLT around?

Because three-dimensional radiation-hydrodynamics simulations can resolve convection in a small patch of surface for a limited time, but cannot evolve a whole star over billions of years. MLT remains the only convection model cheap enough to run inside the millions of one-dimensional models a grid needs. The simulations do not replace MLT but calibrate it: they provide the "true" mean stratification and velocity field for a given gravity and temperature, from which one reads off the effective α and the corrections that repair MLT's surface layers. The two are complementary — simulations for fidelity in a box, MLT for reach across a star's life.