Opacity and Absorption Coefficients

Opacity is the single number that tells you how strongly matter fights radiation. Send a beam of specific intensity $I_\nu$ into a slab of gas and after a path $\mathrm{d}s$ it has lost a fraction $\kappa_\nu\,\rho\,\mathrm{d}s$ of its energy to absorption and scattering. That definition, $\mathrm{d}I_\nu = -\kappa_\nu\rho I_\nu\,\mathrm{d}s$ , is the whole subject in one line: $\kappa_\nu$ is the mass absorption coefficient, the linear coefficient $\alpha_\nu=\rho\kappa_\nu$ has units of inverse length, and its reciprocal is the mean free path — the distance a photon travels before it is stopped. Integrate $\alpha_\nu$ along a ray and you have the optical depth, the natural coordinate of the radiative-transfer equation. What makes opacity rich is that it is a sum over microscopic channels, each with its own signature. Bound-bound transitions between discrete atomic levels carve the sharp absorption lines of a spectrum; bound-free photoionization removes every photon above a threshold and leaves an ionization edge; free-free (inverse bremsstrahlung) lets a free electron in an ion's field soak up a photon; and scattering — Thomson off free electrons, Rayleigh off bound ones — redirects photons without ever thermalizing them. Because the channels depend on frequency, temperature, density, and composition, a star cannot use a single $\kappa_\nu$ : in its opaque interior, where radiation diffuses rather than streams, the correct average is the Rosseland mean, which hands the flux to whichever frequencies are most transparent. This article defines the coefficients, works each channel, derives the Rosseland mean, and pins down the analytic limits — electron scattering, Kramers' law, the H⁻ floor — that stellar astrophysics leans on daily.

Key takeaways

Opacity $\kappa_\nu$ (m² kg⁻¹) is the energy a beam loses to matter per unit mass column; the linear coefficient $\alpha_\nu=\rho\kappa_\nu$ (m⁻¹) sets the mean free path $\ell_\nu=1/\alpha_\nu$ and, integrated, the optical depth.
Four channels add up: bound-bound (lines), bound-free (edges), free-free (inverse bremsstrahlung), and scattering (electron/Thomson, Rayleigh) — only the first three truly absorb.
Electron scattering gives a flat floor $\kappa_{\rm es}\approx0.02(1+X)$ m² kg⁻¹, independent of $\nu$ and $T$ ; bound-free and free-free follow Kramers' law $\bar\kappa\propto\rho\,T^{-7/2}$ .
In an opaque interior radiation diffuses, and the flux is governed by the harmonic Rosseland mean $\bar\kappa_{\rm R}$ , which weights the most transparent frequencies most.
In local thermodynamic equilibrium the same absorptive coefficient that attenuates the beam also emits: Kirchhoff's law $j_\nu/\alpha_\nu^{\rm abs}=B_\nu(T)$ .
Key idea: opacity is the coupling constant between radiation and matter — $\alpha_\nu=\rho\kappa_\nu$ fixes both the attenuation of a beam (optical depth) and, in LTE, the thermal re-emission (source function).

The core insight

Opacity is the coupling constant between radiation and matter: $\alpha_\nu=\rho\kappa_\nu$ sets both the attenuation of a beam (optical depth) and, in local thermodynamic equilibrium, the thermal re-emission (source function).

Where this lives today

Kramers' law is a teaching tool; real modelling runs on enormous computed opacity tables. The two standard sets — OPAL from Livermore (Rogers and Iglesias) and OP from the Opacity Project (Seaton and collaborators) — tabulate the Rosseland mean over grids of density, temperature, and composition, summing millions of iron-group bound-bound lines that Kramers' smooth power law misses. Those lines matter dramatically: the "iron opacity bump" near $T\approx2\times10^5$ K, absent from older tables, was what finally reproduced the pulsation of Cepheid and $\beta$ Cephei variables through the $\kappa$ -mechanism, in which a layer that grows more opaque on compression traps heat and drives the oscillation. Opacity is now a frontier of laboratory astrophysics too: in 2015 the Sandia Z-machine measured iron opacity at solar-interior conditions 30–400% higher than models predicted — bearing directly on the "solar abundance problem," the tension between revised solar abundances and the sound-speed profile helioseismology measures. Codes like MESA read these tables at every zone and timestep. And in cosmology the same Thomson opacity that once locked photons to baryons is now a measured number: the optical depth to reionization from Planck, $\tau\approx0.054$ , encodes the universe's free-electron history after the first stars reionized it.

The physical picture

Picture a beam of photons entering fog. Each metre steals a fixed fraction of the beam; that fraction-per-metre is the linear absorption coefficient $\alpha_\nu$ , and its reciprocal is how far a typical photon flies before it is scattered or swallowed — the mean free path $\ell_\nu$ . Opacity $\kappa_\nu$ is the same quantity per unit mass, an intensive property: what a kilogram of the stuff does to radiation, independent of how much you pile up. Because the beam loses a constant fraction per unit optical depth, the intensity falls exponentially, $I_\nu(s)=I_\nu(0)\,e^{-\tau_\nu}$ — the first figure, with the real numbers of a stellar interior attached.

Figure

A collimated beam of specific intensity I_ν attenuates exponentially through a slab of gas — each optical-depth interval removes a fixed fraction, so I_ν(s) = I_ν(0) exp(−κ_ν ρ s). For fully ionized solar-composition gas at the Sun's centre, κ_es ≈ 0.034 m² kg⁻¹ and ρ ≈ 1.5×10⁵ kg m⁻³, giving a photon mean free path ℓ = 1/(κρ) ≈ 0.2 mm

Opacity is fiercely frequency-selective: enormous at a line centre or just above an ionization edge, small in the transparent windows between them. That is why a stellar spectrum is a near-blackbody continuum stamped with dark lines — at line frequencies the mean free path is tiny, so those photons escape only from the cool outermost layers. Two different physical acts hide inside "extinction": true absorption, which converts a photon's energy into heat of the gas, and scattering, which merely changes its direction. Both feed the optical depth, but only true absorption couples radiation to the gas temperature and appears in Kirchhoff's law. The second figure sorts the four channels across frequency and shows why, deep in a star, the flux escapes through the transparent gaps.

Figure

The monochromatic opacity κ_ν versus frequency, decomposed into channels: a flat electron-scattering floor, a bound-free/free-free continuum that falls like ν⁻³ and jumps upward at each ionization edge, and forests of bound-bound lines. The Rosseland weight ∂B_ν/∂T (dashed) picks out where the flux flows — through the transparent windows between lines and edges where κ_ν sinks to the scattering floor

The core equations

The starting point is the operational definition: opacity is the constant of proportionality between the fractional energy a beam loses and the mass column it crosses. Everything else computes $\kappa_\nu$ from microphysics or averages it for a geometry.

\mathrm{d}I_\nu = -\kappa_\nu\,\rho\,I_\nu\,\mathrm{d}s \qquad (1)

where: $I_\nu$ — specific intensity at frequency $\nu$ (W m⁻² Hz⁻¹ sr⁻¹); $\kappa_\nu$ — opacity / mass absorption coefficient (m² kg⁻¹); $\rho$ — mass density (kg m⁻³); $\mathrm{d}s$ — path element (m). Reading: the beam loses a fraction $\kappa_\nu\rho\,\mathrm{d}s$ per step, so in a uniform slab $I_\nu$ decays exponentially. It is often cleaner to work with the linear coefficient $\alpha_\nu$ , which folds density into the opacity and has the plain meaning of inverse length — microscopically a number density times a cross section, with reciprocal the mean free path.

\alpha_\nu \equiv \rho\,\kappa_\nu = n\,\sigma_\nu, \qquad \ell_\nu = \frac{1}{\alpha_\nu} \qquad (2)

where: $\alpha_\nu$ — linear (volumetric) absorption coefficient (m⁻¹); $n$ — number density of absorbers/scatterers (m⁻³); $\sigma_\nu$ — cross section per particle (m²); $\ell_\nu$ — photon mean free path (m). Reading: per mass, per volume, or per particle — the same physics, three ways. Total opacity sums the microscopic channels, whose frequency signatures the second figure displays.

\kappa_\nu = \kappa_\nu^{\rm bb} + \kappa_\nu^{\rm bf} + \kappa_\nu^{\rm ff} + \kappa_\nu^{\rm sc} \qquad (3)

where: $\kappa_\nu^{\rm bb}$ — bound-bound (spectral lines); $\kappa_\nu^{\rm bf}$ — bound-free (photoionization edges); $\kappa_\nu^{\rm ff}$ — free-free (inverse bremsstrahlung); $\kappa_\nu^{\rm sc}$ — scattering (Thomson off free, Rayleigh off bound electrons). Reading: the first three destroy photons and thermalize; scattering only redirects them. Integrating the linear coefficient along the ray converts geometric path into optical depth, the dimensionless coordinate of the transfer equation.

\mathrm{d}\tau_\nu = -\alpha_\nu\,\mathrm{d}s = -\kappa_\nu\,\rho\,\mathrm{d}s \qquad (4)

where: $\tau_\nu$ — optical depth (dimensionless), conventionally measured inward along the ray. Reading: one mean free path is one unit of optical depth; a medium is "thick" or "thin" as $\tau_\nu\gtrless1$ . The scattering floor is set by the Thomson cross section, a pure constant of the electron — an opacity that ignores frequency and temperature, the reference against which every other channel is measured.

\sigma_{\rm T} = \frac{8\pi}{3}\left(\frac{e^2}{4\pi\varepsilon_0 m_e c^2}\right)^2 = 6.652\times10^{-29}\ \mathrm{m^2},\qquad \kappa_{\rm es}=\frac{n_e\sigma_{\rm T}}{\rho}=\frac{\sigma_{\rm T}(1+X)}{2m_H}\approx 0.02\,(1+X)\ \mathrm{m^2\,kg^{-1}} \qquad (5)

where: $\sigma_{\rm T}$ — Thomson cross section per free electron; $e,\,m_e,\,\varepsilon_0$ — electron charge, mass, vacuum permittivity; $X$ — hydrogen mass fraction; $m_H$ — hydrogen mass. Reading: a fully ionized H+He mix has $(1+X)/2$ electrons per nucleon mass, giving a floor $\approx0.034$ m² kg⁻¹ at $X=0.7$ (using $m_H=1.67\times10^{-27}$ kg). The true-absorption continuum is dominated by free-free absorption, whose coefficient carries the characteristic $\nu^{-3}$ frequency law, a $T^{-1/2}$ from the electron thermal speed, and a stimulated-emission correction.

\alpha_\nu^{\rm ff} \propto Z^2\, n_e\, n_i\, T^{-1/2}\, \nu^{-3}\,\bar{g}_{\rm ff}\left(1 - e^{-h\nu/k_BT}\right) \qquad (6)

where: $Z$ — ion charge number; $n_i$ — ion number density; $\bar{g}_{\rm ff}$ — velocity-averaged Gaunt factor (a quantum correction of order unity); $(1-e^{-h\nu/k_BT})$ — net-absorption factor after stimulated emission. Reading: bound-free absorption shares the same $\nu^{-3}$ tail above each edge, so the two group together in Kramers' law. Averaged over frequency, the bound-free and free-free continua collapse to a single power law — Kramers' opacity, the workhorse for stellar envelopes.

\bar\kappa_{\rm Kr} \approx \kappa_0\,\rho\,T^{-7/2} \qquad (7)

where: $\kappa_0$ — composition-dependent constant (different for bound-free and free-free); $\bar\kappa_{\rm Kr}$ — Rosseland-mean opacity in the bound-free/free-free regime. Reading: opacity climbs steeply as temperature falls, peaking in the cool ionization zones of an envelope. In local thermodynamic equilibrium the same absorptive coefficient that dims the beam also emits, and their ratio is forced to be the Planck function — Kirchhoff's law, the bridge to the source function.

\frac{j_\nu}{\alpha_\nu^{\rm abs}} = B_\nu(T) \qquad (8)

where: $j_\nu$ — emission coefficient (emissivity); $\alpha_\nu^{\rm abs}$ — purely absorptive (not scattering) part of the linear coefficient; $B_\nu(T)$ — Planck function at the local temperature. Reading: only the absorptive part appears — scattering, which does not thermalize, is absent. In an optically thick interior the net radiative flux obeys a diffusion law whose conductivity is fixed by the Rosseland mean.

F_{\rm rad} = -\frac{c}{3\bar\kappa_{\rm R}\rho}\,\frac{\mathrm{d}(aT^4)}{\mathrm{d}r},\qquad a=7.566\times10^{-16}\ \mathrm{J\,m^{-3}\,K^{-4}} \qquad (9)

where: $F_{\rm rad}$ — radiative energy flux; $\bar\kappa_{\rm R}$ — Rosseland mean opacity; $a$ — radiation constant; $c$ — speed of light. Reading: high opacity throttles the flux and forces a steep temperature gradient — deciding where a star is radiative or convective.

Derivation AThe Rosseland mean opacity from the diffusion limit

The goal is to show that deep inside a star, where the radiation field is nearly isotropic and photons random-walk rather than stream, the net flux is governed not by any single monochromatic opacity but by a specific frequency average — the harmonic, $\partial B_\nu/\partial T$ -weighted Rosseland mean. We expand the intensity to first order in the small anisotropy, integrate the flux over angle and frequency, and demand a grey diffusion form; that requirement defines the mean, inverse-weighted so transparent frequencies dominate.

Step 1. Write the plane-parallel transfer equation $\mu\,\partial I_\nu/\partial\tau_\nu = I_\nu - S_\nu$ , with $\mu=\cos\theta$ and, in deep-interior LTE, source function $S_\nu = B_\nu(T)$ . Why: LTE is excellent in the interior because collision rates dwarf radiative rates, so the local matter and radiation share one temperature.

Step 2. Because the medium is optically thick, $I_\nu$ departs only slightly from $B_\nu$ ; keep first order in the gradient, $I_\nu \approx B_\nu - \mu\,(\kappa_\nu\rho)^{-1}\,\partial B_\nu/\partial r$ . Why: this is the diffusion ansatz — the isotropic Planck term plus a small forward-peaked correction proportional to how fast $B_\nu$ changes over a mean free path.

Step 3. Form the monochromatic flux $F_\nu = \int I_\nu\,\mu\,\mathrm{d}\Omega = -\dfrac{4\pi}{3\kappa_\nu\rho}\dfrac{\partial B_\nu}{\partial r}$ . Why: the isotropic $B_\nu$ carries zero net flux; only the $\mu^2$ term survives, and $\int\mu^2\,\mathrm{d}\Omega = 4\pi/3$ gives the diffusion factor.

Step 4. By the chain rule, since $B_\nu$ depends on radius only through $T(r)$ : $F_\nu = -\dfrac{4\pi}{3\kappa_\nu\rho}\dfrac{\partial B_\nu}{\partial T}\dfrac{\mathrm{d}T}{\mathrm{d}r}$ . Why: it separates the atomic physics from the one structural driver $\mathrm{d}T/\mathrm{d}r$ .

Step 5. Integrate over frequency, $F = -\dfrac{4\pi}{3\rho}\dfrac{\mathrm{d}T}{\mathrm{d}r}\displaystyle\int_0^\infty \dfrac{1}{\kappa_\nu}\dfrac{\partial B_\nu}{\partial T}\,\mathrm{d}\nu$ . Why: each channel is weighted by its own $1/\kappa_\nu$ , so transparent frequencies carry more flux.

Step 6. Demand the grey form $F = -\dfrac{4\pi}{3\bar\kappa_{\rm R}\rho}\dfrac{\mathrm{d}T}{\mathrm{d}r}\displaystyle\int_0^\infty\dfrac{\partial B_\nu}{\partial T}\,\mathrm{d}\nu$ ; matching the two lines gives the boxed harmonic average. Why: we choose $\bar\kappa_{\rm R}$ precisely so one number reproduces the true multi-frequency flux.

\frac{1}{\bar\kappa_{\rm R}} = \frac{\displaystyle\int_0^\infty \frac{1}{\kappa_\nu}\,\frac{\partial B_\nu}{\partial T}\,\mathrm{d}\nu}{\displaystyle\int_0^\infty \frac{\partial B_\nu}{\partial T}\,\mathrm{d}\nu} \qquad (\text{Rosseland mean})

Using $\int_0^\infty(\partial B_\nu/\partial T)\,\mathrm{d}\nu = acT^3/\pi$ reduces the grey flux to $F=-\tfrac{c}{3\bar\kappa_{\rm R}\rho}\,\mathrm{d}(aT^4)/\mathrm{d}r$ , equation (9). Verified: the transparency-weighted harmonic form makes spectral windows dominate the energy leak.

Go deeper · Derivation B · 1 minBound-bound opacity of a spectral line

Derivation BBound-bound opacity of a spectral line

The aim is to connect the sharply peaked line contribution to $\kappa_\nu$ with the atomic constants a spectroscopist tabulates — the oscillator strength and the level populations — and to show where the stimulated-emission correction comes from. We take the classical sum rule for the integrated cross section, spread it over the line with a normalized profile, multiply by the absorbing-level population, and subtract the downward stimulated transitions. The final bracket, in LTE, reproduces the same $1-e^{-h\nu/k_BT}$ factor that governs free-free absorption.

Step 1. The classical oscillator-strength sum rule fixes the integrated cross section, $\displaystyle\int\sigma_\nu\,\mathrm{d}\nu = \frac{\pi e^2}{4\pi\varepsilon_0 m_e c}\,f_{\ell u}$ , with $f_{\ell u}$ the dimensionless oscillator strength of the lower-to-upper transition. Why: it packages all the quantum detail of the transition into one measured number, leaving only the line shape to be specified.

Step 2. Distribute that strength over frequency with a normalized profile, $\sigma_\nu = \dfrac{\pi e^2}{4\pi\varepsilon_0 m_e c}\,f_{\ell u}\,\phi(\nu)$ , where $\int\phi(\nu)\,\mathrm{d}\nu=1$ . Why: $\phi(\nu)$ carries the broadening — natural, Doppler, and pressure — while the line area stays fixed by the oscillator strength.

Step 3. Multiply by the lower-level number density, $\alpha_\nu = n_\ell\,\sigma_\nu$ , with $n_\ell$ set by the Boltzmann and Saha equations. Why: opacity is a bulk property, so the cross section must be scaled by how many absorbers sit in the lower level.

Step 4. Subtract stimulated emission from the upper level, giving $\alpha_\nu^{\rm net} = n_\ell\sigma_\nu\left(1-\dfrac{g_\ell n_u}{g_u n_\ell}\right)$ , which in LTE becomes $n_\ell\sigma_\nu(1-e^{-h\nu/k_BT})$ . Why: downward stimulated transitions add photons back to the beam ("negative absorption"), and in LTE the population ratio collapses to the Boltzmann factor — matching Kirchhoff's law.

\alpha_\nu^{\rm bb} = \frac{\pi e^2}{4\pi\varepsilon_0 m_e c}\,f_{\ell u}\,n_\ell\left(1 - \frac{g_\ell n_u}{g_u n_\ell}\right)\phi(\nu) \qquad (\text{line opacity})

Verified: with $n_u/n_\ell=(g_u/g_\ell)e^{-h\nu/k_BT}$ the bracket reduces to $1-e^{-h\nu/k_BT}$ , so line, free-free, and continuum absorption all share one thermodynamic factor.

Go deeper · Derivation C · 1 minWhy bound-free and free-free opacity follow Kramers' T to the minus seven-halves law

Derivation CWhy bound-free and free-free opacity follow Kramers' T to the minus seven-halves law

The goal is to reconstruct the famous $\bar\kappa\propto\rho\,T^{-7/2}$ scaling from the microphysics, so that it reads as a consequence rather than a fit. The cross sections for bound-free and free-free absorption fall as $\nu^{-3}$ ; free-free adds $T^{-1/2}$ from the electron thermal velocity; the per-mass conversion supplies one power of density; and the Rosseland average over the Planck weighting converts $\nu^{-3}$ into an extra power of temperature. These four ingredients give the classic exponent — and show why the law is only a smooth approximation to the jagged real opacity of the previous figure.

Step 1. Take the monochromatic coefficient $\kappa_\nu \propto \rho\,\nu^{-3}\,T^{-1/2}$ : the $\nu^{-3}$ is the shared bound-free/free-free frequency law, the $T^{-1/2}$ the electron-speed factor, and one power of $\rho$ survives because $n_e n_i\propto\rho^2$ divided by $\rho$ per mass. Why: this isolates the intrinsic dependence before any averaging.

Step 2. Form the Rosseland average, $\bar\kappa_{\rm R} \propto \rho\,T^{-1/2}\Big[\displaystyle\int \nu^{3}\,\partial_T B_\nu\,\mathrm{d}\nu \big/ \int\partial_T B_\nu\,\mathrm{d}\nu\Big]$ ; the $1/\kappa_\nu$ weighting brings the $\nu^{+3}$ upstairs. Why: the harmonic mean of a $\nu^{-3}$ opacity is set by the transparent high-frequency side of the Planck weight.

Step 3. The weighting peaks at $\nu\sim k_BT/h$ , so $\langle\nu^3\rangle\propto T^3$ , giving $\bar\kappa\propto\rho\,T^{-1/2}\,T^{-3}=\rho\,T^{-7/2}$ . Why: the thermal shift of the Planck peak converts a frequency power into a temperature power, completing Kramers' exponent.

\bar\kappa_{\rm Kr} \propto \rho\,T^{-7/2} \qquad (\text{Kramers' law})

Verified: the exponent decomposes as $-1/2$ (free-free thermal speed) $-3$ (Rosseland-averaged $\nu^{-3}$ ); the same reasoning with the bound-free cross section gives an identical scaling with a different $\kappa_0$ .

Limiting cases

Which channel wins depends on temperature, density, and composition; the four regimes below cover essentially every stellar and cosmological setting.

Case	Result	Meaning
Hot, fully ionized gas (stellar cores, pre-recombination universe)	$\kappa_{\rm es}\approx0.02(1+X)$ m² kg⁻¹	Electron scattering dominates; frequency- and temperature-independent. Sets the opacity floor and the photon–baryon coupling of the early universe.
Warm, partially ionized gas (stellar envelopes)	$\bar\kappa_{\rm bf,ff}\propto\rho\,T^{-7/2}$	Kramers' law: bound-free and free-free absorption dominate and rise steeply as $T$ falls, driving convective instability outward.
Cool gas ( $T<10^4$ K, atmospheres, molecular clouds)	$\kappa\propto T^{\sim4\text{–}10}$	H⁻ photodetachment (solar-type stars), molecular bands, and dust grains dominate; opacity climbs sharply with $T$ over a narrow range.
Optically thick diffusion limit	$\dfrac{1}{\bar\kappa_{\rm R}}=\dfrac{\int_0^\infty\kappa_\nu^{-1}\,\partial_T B_\nu\,\mathrm{d}\nu}{\int_0^\infty\partial_T B_\nu\,\mathrm{d}\nu}$	The Rosseland harmonic mean gives greatest weight to the most transparent frequencies — the ones through which energy actually escapes.

Common trapExtinction is not the same as absorption.

The tempting mistake is to treat every process that dims a beam as if it heats the gas. It does not. Scattering contributes fully to $\kappa_\nu$ and $\tau_\nu$ — a scattered photon leaves your line of sight just as surely as an absorbed one — but it converts no photon energy into heat, so it is absent from Kirchhoff's law $j_\nu/\alpha_\nu^{\rm abs}=B_\nu$ . This is why the source function in a scattering-dominated atmosphere is not Planckian: $S_\nu = (\alpha^{\rm abs}B_\nu + \sigma^{\rm sc}J_\nu)/(\alpha^{\rm abs}+\sigma^{\rm sc})$ , a mix of local temperature and mean intensity $J_\nu$ . A related trap: large opacity does not imply strong emission — only a true absorber in LTE emits at the blackbody rate, while a purely scattering fog can be perfectly opaque and radiate nothing.

Worked examples

1 · Electron-scattering opacity, mean free path, and the photon escape time from the Sun (fully worked)

Take fully ionized solar-composition gas ( $X=0.70$ ) at the Sun's centre, where $\rho\approx1.5\times10^5$ kg m⁻³ and $R_\odot=6.96\times10^8$ m. Find the electron-scattering opacity, the photon mean free path, and roughly how long a photon takes to random-walk out.

Step 1. Opacity floor from equation (5): $\kappa_{\rm es}=0.02(1+X)=0.02(1.70)=\mathbf{0.034\ m^2\,kg^{-1}}$ . This is a lower bound — bound-free and free-free add to it — but the cleanest single estimate for the hot core.

Step 2. Linear coefficient and mean free path: $\alpha=\kappa\rho=0.034\times1.5\times10^5=5.1\times10^3$ m⁻¹, so $\ell=1/\alpha=1/5100=\mathbf{2.0\times10^{-4}\ m\approx0.2\ mm}$ . A photon in the core travels about a fifth of a millimetre between scatterings.

Step 3. Random walk out. After $N$ steps of length $\ell$ the displacement is $\sim\ell\sqrt{N}$ , so escaping radius $R$ needs $N\approx(R/\ell)^2$ steps and time $t\approx N\ell/c = R^2/(\ell c) = (6.96\times10^8)^2/(2.0\times10^{-4}\times3.0\times10^8)=\mathbf{8\times10^{12}\ s\approx2.6\times10^5\ yr}$ .

Step 4. For perspective, an unobstructed photon would cross the Sun in $R_\odot/c=\mathbf{2.3\ s}$ ; opacity stretches that to of order $10^5$ years. (The true figure is a few $\times10^4$ yr with varying opacity, but the constant- $\kappa$ estimate captures the enormous factor.) Result: κ_es ≈ 0.034 m² kg⁻¹, ℓ ≈ 0.2 mm, escape time ~10⁵ yr versus a 2.3 s free crossing.

2 · Why the Rosseland mean tracks the transparent window, not the average (fully worked)

Model a gas with two regions carrying equal shares of the Planck weight $\partial B_\nu/\partial T$ : a clean window with $\kappa_A=0.01$ m² kg⁻¹ and a line-choked band with $\kappa_B=100$ m² kg⁻¹. Compare the arithmetic and Rosseland means.

Step 1. Arithmetic (flux-blind) mean: $\bar\kappa_{\rm arith}=\tfrac12\kappa_A+\tfrac12\kappa_B=\tfrac12(0.01)+\tfrac12(100)=\mathbf{50\ m^2\,kg^{-1}}$ . The opaque band dominates a straight average.

Step 2. Rosseland (harmonic) mean from the boxed formula, with equal weights: $\dfrac{1}{\bar\kappa_{\rm R}}=\dfrac12\dfrac{1}{\kappa_A}+\dfrac12\dfrac{1}{\kappa_B}=\dfrac12(100)+\dfrac12(0.01)=50.005$ , so $\bar\kappa_{\rm R}=\mathbf{0.0200\ m^2\,kg^{-1}}$ .

Step 3. The Rosseland mean sits essentially at the transparent-window value, a factor $50/0.02\approx\mathbf{2500}$ below the arithmetic mean — the flux does not care how opaque the worst band is; it pours through the one open window. Result: arithmetic 50 vs Rosseland 0.0200 m² kg⁻¹ — the harmonic mean is hostage to the most transparent frequency. (For contrast, Kramers' $T^{-7/2}$ means halving the temperature at fixed density raises bound-free/free-free opacity by $2^{7/2}\approx11$ , while the electron-scattering floor stays flat — which is why cooling outer envelopes become opaque enough to convect.)

Physics of the portrait

Opacity is where thermodynamics, atomic physics, and stellar structure meet, and one idea holds the interior together: in a diffusing medium the flux is set by a single harmonic-mean opacity. That idea is Svein Rosseland's.

Origins · 1 minSvein Rosseland and how this was discovered

Svein Rosseland

1894–1985 · radiative transfer and the internal constitution of the stars

Svein Rosseland was a Norwegian astrophysicist who, in a short 1924 note in the Monthly Notices, solved a problem that had stalled the theory of stellar interiors: how to compress the wildly frequency-dependent absorption of stellar matter into a single number a structure calculation could use. His insight was that in the deep, optically thick interior radiation diffuses, and diffusion is controlled by the frequencies through which energy most easily flows — the transparent windows — so the right average of $\kappa_\nu$ is a harmonic one, weighted by the temperature derivative of the Planck function. That "Rosseland mean" became, and remains, the opacity that appears in every stellar-evolution code. Trained under Bohr in Copenhagen and influenced by Eddington's contemporaneous work on radiative equilibrium, Rosseland founded the Institute of Theoretical Astrophysics in Oslo in 1934 and built one of Europe's early differential analysers for astrophysical computation. His name now labels not only the mean opacity but the optical-depth scale and the diffusion approximation that carry a star's luminosity from core to surface.

History

1859

— Kirchhoff relates a body's thermal emission to its absorption, j_ν/α_ν = B_ν

1906

— Karl Schwarzschild formulates radiative equilibrium in the solar atmosphere

1923

— Kramers derives the bound-free and free-free absorption coefficient, scaling as ν⁻³

1924

— Rosseland defines the harmonic-mean opacity governing radiative diffusion

1992

— the OPAL project delivers modern tabulated stellar opacities, resolving the Cepheid opacity crisis

Going further

Texts. Carroll and Ostlie, An Introduction to Modern Astrophysics, 2nd ed., ch. 9–10 · Rybicki and Lightman, Radiative Processes in Astrophysics, ch. 1 and 5 · Mihalas, Stellar Atmospheres, 2nd ed. · Kippenhahn, Weigert and Weiss, Stellar Structure and Evolution, 2nd ed., ch. 17 · Dodelson, Modern Cosmology (Thomson scattering, visibility function) · Mo, van den Bosch and White, Galaxy Formation and Evolution.

Provenance. Definitions and the channel decomposition follow Carroll and Ostlie ch. 9 and Rybicki and Lightman ch. 1. A verifier should recompute: (1) $\sigma_{\rm T}=6.652\times10^{-29}$ m² from $r_e=2.818\times10^{-15}$ m; (2) $\kappa_{\rm es}=\sigma_{\rm T}(1+X)/2m_H\approx0.02(1+X)$ with $m_H=1.67\times10^{-27}$ kg (0.034 at $X=0.7$ ); (3) $\ell=1/(\kappa\rho)\approx0.2$ mm and escape time $\sim10^5$ yr at solar-centre density; (4) the Rosseland harmonic mean and the toy-model factor $\approx2500$ ; (5) Kramers' $T^{-7/2}$ ; (6) $a=7.566\times10^{-16}$ J m⁻³ K⁻⁴. All are standard textbook values.

Prerequisites. Specific Intensity and Radiative Flux · The Equation of Radiative Transfer · Blackbody Radiation and the Planck Law. Leads to. Optical Depth · Emission and the Source Function · Local Thermodynamic Equilibrium · Spectral Line Formation.

Ask the tutor: ask me to walk through why the Rosseland mean is harmonic, or to estimate the opacity, mean free path, and photon escape time for any star you name.

What is the difference between the opacity κ and the absorption coefficient α?

They are the same physics in different bookkeeping. The opacity (or mass absorption coefficient) $\kappa_\nu$ is per unit mass (m² kg⁻¹); the linear coefficient $\alpha_\nu=\rho\kappa_\nu$ is per unit length (m⁻¹), with $1/\alpha_\nu$ the mean free path. Astronomers favour $\kappa_\nu$ because it is intensive — it does not change when you compress the gas — whereas $\alpha_\nu$ and the optical depth do. A third form, the cross section $\sigma_\nu=\alpha_\nu/n$ , is opacity per particle, the natural unit for a single scatterer like the electron.

Why is electron-scattering opacity independent of frequency and temperature?

Because the Thomson cross section is a pure constant of the electron. In the non-relativistic limit a free electron oscillates in the wave's field and re-radiates with a cross section $\sigma_{\rm T}=6.652\times10^{-29}$ m² that contains no reference to the photon's frequency — every colour scatters equally. The opacity $\kappa_{\rm es}=n_e\sigma_{\rm T}/\rho$ then depends only on the free electrons per unit mass, which for a fully ionized plasma is fixed by composition ( $X$ ) alone. The constancy breaks only when photons reach $h\nu\sim m_ec^2$ (Compton recoil) or the gas is not fully ionized so that $n_e$ itself depends on $T$ .

If scattering does not destroy photons, why does it count as opacity at all?

Opacity measures removal from a particular beam, not destruction of energy. A photon Thomson-scattered out of your line of sight is gone from that beam just as completely as one absorbed, so it contributes fully to $\kappa_\nu$ , the optical depth, and the mean free path. The distinction only matters for the source function: absorbed energy is thermalized and re-emitted at the local temperature (Kirchhoff), whereas scattered energy is merely redirected. That is why a scattering atmosphere can be extremely opaque yet emit almost nothing of its own.

Why is the Rosseland mean a harmonic average rather than a straight one?

Because in the diffusion limit the flux is proportional to $1/\kappa_\nu$ , not to $\kappa_\nu$ . Energy leaks out through whichever frequencies are most transparent, so the average governing the net flux must give those small- $\kappa_\nu$ windows their proper large influence — exactly what averaging $1/\kappa_\nu$ (weighted by $\partial B_\nu/\partial T$ ) does. A straight arithmetic mean would let one opaque band dominate and wildly overestimate how well the star traps radiation, as the worked-example toy model shows: arithmetic 50 versus Rosseland 0.02 for the same gas, a factor of 2500 apart.

Why is H⁻ the dominant opacity source in the Sun's photosphere rather than hydrogen itself?

In the $\sim5800$ K solar photosphere almost all hydrogen is neutral and in the ground state, so hydrogen bound-free absorption needs ultraviolet photons and does little in the visible. The negative hydrogen ion H⁻ — a proton with two bound electrons — binds its extra electron by only 0.75 eV, so visible and near-infrared photons easily photodetach it, producing a smooth continuum across the band where the Sun radiates most. The trick is that H⁻ exists at all: the second electron comes from metals (iron, silicon, magnesium) partially ionized at photospheric temperatures — which is why solar continuum opacity is so sensitive to temperature and metallicity, and why H⁻ is the workhorse opacity of cool atmospheres.

How do modern opacity tables differ from Kramers' law?

Kramers' $\rho\,T^{-7/2}$ law is a smooth fit to the bound-free and free-free continua only, ignoring the millions of bound-bound lines of partially ionized metals. Modern tables — OPAL and the Opacity Project — compute those lines explicitly, and the difference is not cosmetic: they reveal an iron-group opacity bump near $2\times10^5$ K that Kramers' law omits, and that bump drives the pulsation of classical variable stars through the $\kappa$ -mechanism. Kramers' law survives as an intuition-builder and an analytic limit for hot, metal-poor gas, but no serious stellar model uses it in place of a computed table.