The Equation of Radiative Transfer

Almost everything we know about the Universe beyond the Solar System arrives as light, and light is not a passive messenger — it is edited by every atom, ion, and dust grain it passes on the way to the telescope. The equation of radiative transfer is the bookkeeping that tracks that editing. Its subject is the specific intensity $I_\nu$ , defined so that the energy crossing an area $\mathrm{d}A$ into solid angle $\mathrm{d}\Omega$ in band $\mathrm{d}\nu$ during $\mathrm{d}t$ at angle $\theta$ to the normal is $\mathrm{d}E = I_\nu\,\cos\theta\,\mathrm{d}A\,\mathrm{d}\Omega\,\mathrm{d}\nu\,\mathrm{d}t$ . In empty space $I_\nu$ is conserved along a ray — the fact that lets a detector on Earth speak about the surface brightness of a star. Matter breaks that conservation in exactly two ways: it removes photons from the beam (extinction, set by the absorption coefficient $\alpha_\nu$ ) and it adds photons to the beam (emission, set by the emission coefficient $j_\nu$ ). Balancing the two over a path $\mathrm{d}s$ gives the whole equation, $\mathrm{d}I_\nu/\mathrm{d}s = -\alpha_\nu I_\nu + j_\nu$ . The single change of variable to optical depth $\mathrm{d}\tau_\nu = \alpha_\nu\,\mathrm{d}s$ and the source function $S_\nu \equiv j_\nu/\alpha_\nu$ turns this into the deceptively simple $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = S_\nu - I_\nu$ : the intensity is driven toward the local source function on an $e$ -folding scale of one optical depth. Whatever the microphysics — stellar photospheres, interstellar clouds, accretion flows, or the last scattering of the cosmic microwave background — only the interpretation of $\alpha_\nu$ and $S_\nu$ changes; the equation does not.

Key takeaways

The transfer equation tracks how the specific intensity $I_\nu$ of a beam changes as it moves through matter that absorbs, emits, and scatters photons.
Over a path $\mathrm{d}s$ the change is a competition: emission adds $j_\nu\,\mathrm{d}s$ , extinction removes $\alpha_\nu I_\nu\,\mathrm{d}s$ , giving $\mathrm{d}I_\nu/\mathrm{d}s = -\alpha_\nu I_\nu + j_\nu$ .
In terms of optical depth $\tau_\nu$ and the source function $S_\nu = j_\nu/\alpha_\nu$ it collapses to the canonical form $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = S_\nu - I_\nu$ .
Its formal solution is an attenuated background beam $I_\nu(0)e^{-\tau_\nu}$ plus a source-function integral weighted by $e^{-(\tau_\nu-t)}$ along the ray.
In local thermodynamic equilibrium Kirchhoff's law sets $S_\nu = B_\nu(T)$ , so an optically thick, isothermal region radiates a pure blackbody.
The photons we observe emerge from the layer near $\tau_\nu \approx 1$ , so the emergent spectrum is essentially the source function at that depth — the content of the Eddington-Barbier relation.
Key idea: along any ray the specific intensity is pushed toward the local source function on a scale of one optical depth, $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = S_\nu - I_\nu$ ; we see the source function at the depth where $\tau_\nu \approx 1$ .

The core insight

Along any ray the specific intensity is pushed toward the local source function on a scale of one optical depth: $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = S_\nu - I_\nu$ . The photons we observe emerge from the layer where $\tau_\nu \approx 1$ , so the emergent spectrum is essentially the source function at that depth.

Where this lives today

The transfer equation is not a historical relic — it is the computational engine inside almost every quantitative interpretation of astronomical light. Model-atmosphere codes such as ATLAS, MARCS, PHOENIX, and TLUSTY solve it (dropping LTE where needed) over millions of frequency points to build synthetic spectra that, matched against APOGEE, GALAH, and Gaia data, deliver temperatures, gravities, and abundances for millions of stars. In cosmology the same equation, promoted to the photon Boltzmann equation, is integrated by solvers like CAMB and CLASS to predict the CMB power spectrum measured by COBE, WMAP, and Planck; the visibility function $g(\tau) = e^{-\tau}\,\mathrm{d}\tau/\mathrm{d}\eta$ that peaks at recombination is exactly the emergent-intensity kernel of equation (4) recast in cosmic time, and it is why the CMB is a near-perfect blackbody at $T_0 = 2.725$ K. The same solvers underpin protoplanetary disks and dusty galaxies imaged by ALMA and JWST, supernova light curves, and the general-relativistic transfer behind Event Horizon Telescope images of M87 and the Galactic Centre — photons traced through curved spacetime with $\alpha_\nu$ and $j_\nu$ from a hot accretion flow. The physics Schwarzschild wrote for the Sun now runs from the first minutes after the Big Bang to the edge of an event horizon.

The physical picture

Picture photons marching down a ray as if through a stack of thin veils. Every veil does two things at once: it dims whatever comes from behind it (extinction, $-\alpha_\nu I_\nu\,\mathrm{d}s$ ) and it glows with its own light (emission, $+j_\nu\,\mathrm{d}s$ ). The transfer equation is nothing more than the arithmetic of one veil — subtract what is absorbed, add what is emitted — accumulated along the ray. Optical depth $\tau_\nu$ counts how many veils deep you are: after one, only $e^{-1} = 0.37$ of the original beam survives. The first figure freezes a single slab and labels the two competing terms and the ruler that measures them.

Figure

A ray of intensity I_ν(0) enters a slab: extinction removes α_ν I_ν ds while emission adds j_ν ds, and optical depth τ_ν = ∫ α_ν ds counts the slabs — after one optical depth the incoming beam is attenuated to e⁻¹ = 0.37 of its value, leaving the emergent intensity governed by dI_ν/dτ_ν = S_ν − I_ν

The ratio $S_\nu = j_\nu/\alpha_\nu$ is the intensity the beam would relax to if it sat inside the medium long enough. If $I_\nu > S_\nu$ the slab is a net absorber and the beam dims; if $I_\nu < S_\nu$ it is a net emitter and brightens; the intensity always chases $S_\nu$ . That relaxation is quantitative: for an isothermal slab with constant $S_\nu$ , $I_\nu(\tau_\nu) = I_\nu(0)e^{-\tau_\nu} + S_\nu(1-e^{-\tau_\nu})$ , an exponential approach to $S_\nu$ over one optical depth. The second figure plots it for a beam entering at $I_\nu(0) = 0.2\,S_\nu$ : by $\tau_\nu = 1$ it has climbed to $0.71\,S_\nu$ , and by $\tau_\nu \approx 3$ it is within a few percent of $S_\nu$ whatever its start — the beam has forgotten its history.

Figure

Relaxation of intensity toward the source function for an isothermal slab, I_ν(τ) = I_ν(0)e⁻τ + S_ν(1−e⁻τ) with I_ν(0) = 0.2 S_ν: the beam climbs to 0.71 S_ν at τ = 1 and reaches 0.96 S_ν by τ = 3, forgetting its starting value on a scale of one optical depth

Because emergent radiation comes from $\tau_\nu \approx 1$ , and because opacity varies strongly with frequency, we see to different physical depths at different colours. At a spectral line where $\alpha_\nu$ spikes, $\tau_\nu = 1$ is reached high in a cooler layer, producing a dark absorption line; in the neighbouring continuum we see deeper, hotter gas — the whole basis of stellar spectroscopy. In cosmology the same equation runs with $\alpha_\nu$ set by Thomson scattering off free electrons and $S_\nu$ the local blackbody: as the Universe recombines, $\tau$ drops below unity, the photons last-scatter, and they stream freely ever after — the free-space limit where $I_\nu$ is conserved.

The core equations

The transfer equation begins as a balance sheet written along the ray: over a path $\mathrm{d}s$ the beam loses a fraction $\alpha_\nu\,\mathrm{d}s$ of itself to extinction and gains $j_\nu\,\mathrm{d}s$ from emission. This is the physical statement from which every other form follows.

\frac{\mathrm{d}I_\nu}{\mathrm{d}s} = -\alpha_\nu I_\nu + j_\nu \qquad (1)

where: $I_\nu$ — specific intensity at frequency $\nu$ (erg s $^{-1}$ cm $^{-2}$ Hz $^{-1}$ sr $^{-1}$ ); $s$ — path length along the ray; $\alpha_\nu$ — absorption (extinction) coefficient, cm $^{-1}$ , equal to $\kappa_\nu\rho$ for mass opacity $\kappa_\nu$ ; $j_\nu$ — emission coefficient (erg s $^{-1}$ cm $^{-3}$ Hz $^{-1}$ sr $^{-1}$ ). The first term is proportional to $I_\nu$ (you can only remove what is there); the second is independent of it (the medium glows on its own). The equation simplifies enormously once we stop measuring distance in centimetres and start measuring it in optical depths — the natural, dimensionless yardstick of how opaque the medium is — and package the emission-to-extinction ratio into a single quantity, the source function.

\mathrm{d}\tau_\nu = \alpha_\nu\,\mathrm{d}s, \qquad S_\nu \equiv \frac{j_\nu}{\alpha_\nu} \qquad (2)

where: $\tau_\nu$ — optical depth (dimensionless), measured along the ray; $S_\nu$ — source function, the ratio of emission to extinction, carrying the same units as $I_\nu$ . A region with $\tau_\nu \ll 1$ is transparent (optically thin); a region with $\tau_\nu \gg 1$ is opaque (optically thick). Dividing equation (1) by $\alpha_\nu$ and substituting these definitions yields the canonical form — arguably the most important single line in the article — whose right-hand side reads directly as a restoring force pulling $I_\nu$ toward $S_\nu$ .

\frac{\mathrm{d}I_\nu}{\mathrm{d}\tau_\nu} = -I_\nu + S_\nu \qquad (3)

where: $\tau_\nu$ — the independent variable, increasing along the ray; $S_\nu$ — the local source function that the intensity relaxes toward. When $I_\nu = S_\nu$ the derivative vanishes and the beam is in radiative balance with the matter. Treating (3) as a first-order linear ODE and integrating gives the formal solution — the emergent intensity written as the incident beam, attenuated, plus a sum of source-function contributions each attenuated over the path that remains ahead of it.

I_\nu(\tau_\nu) = I_\nu(0)\,e^{-\tau_\nu} + \int_0^{\tau_\nu} S_\nu(t)\,e^{-(\tau_\nu - t)}\,\mathrm{d}t \qquad (4)

where: $I_\nu(0)$ — incident intensity at $\tau_\nu = 0$ ; $t$ — dummy optical-depth variable; $e^{-\tau_\nu}$ — attenuation of the background beam; $e^{-(\tau_\nu-t)}$ — attenuation of emission released at depth $t$ over the remaining path. It is "formal" because $S_\nu$ may itself depend on the radiation field. For a stratified atmosphere it is natural to measure optical depth vertically, inward from the surface, and to look out along a ray tilted by $\theta$ from the vertical. Two changes appear — a factor $\mu = \cos\theta$ and a sign flip — giving the working equation of stellar-atmosphere theory.

\mu\,\frac{\mathrm{d}I_\nu}{\mathrm{d}\tau_\nu} = I_\nu - S_\nu \qquad (5)

where: $\mu = \cos\theta$ — cosine of the angle between the ray and the outward normal; $\tau_\nu$ — vertical optical depth, measured inward from the surface. The sign is opposite to (3) because $\tau_\nu$ now increases in the direction opposite to the outgoing ray. The source function is where all the physics hides. In pure absorption it is the Planck function; in pure scattering it is the mean intensity $J_\nu$ ; in general it is a weighted average, and the weight $\varepsilon_\nu$ decides how thermalized the radiation is.

S_\nu = (1-\varepsilon_\nu)\,J_\nu + \varepsilon_\nu\,B_\nu(T) \qquad (6)

where: $\varepsilon_\nu$ — thermalization parameter, the fraction of extinction due to true absorption (as opposed to scattering); $J_\nu = \frac{1}{4\pi}\oint I_\nu\,\mathrm{d}\Omega$ — mean intensity; $B_\nu(T)$ — Planck function at the local temperature. Setting $\varepsilon_\nu = 1$ recovers LTE ( $S_\nu = B_\nu$ ); setting $\varepsilon_\nu = 0$ gives coherent scattering ( $S_\nu = J_\nu$ ), which couples the equation to itself. Finally, integrating (5) outward through a semi-infinite atmosphere and taking a source function that varies linearly with depth yields the Eddington-Barbier relation — a compact statement of exactly where in the atmosphere the light we see is born.

I_\nu(0,\mu) \approx S_\nu(\tau_\nu = \mu) \qquad (7)

where: $I_\nu(0,\mu)$ — intensity emerging at the surface along direction cosine $\mu$ ; $S_\nu(\tau_\nu=\mu)$ — source function evaluated at vertical optical depth equal to $\mu$ . Looking straight down ( $\mu = 1$ ) we sample $\tau_\nu = 1$ ; looking near the limb ( $\mu \to 0$ ) we sample shallower, cooler layers — the origin of limb darkening.

Derivation AFormal solution of the transfer equation

The goal is to convert the canonical differential form $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = S_\nu - I_\nu$ into an explicit expression for the intensity at any optical depth, so that the emergent radiation can be read off directly once the run of the source function along the ray is known. The strategy is the standard integrating-factor method for a first-order linear ODE: put the equation in the form $y' + Py = Q$ , multiply by a factor that turns the left-hand side into an exact derivative, integrate, and interpret the two resulting terms physically. What emerges is the single most useful equation in the subject — an attenuated background beam plus a weighted integral over local emission.

Step 1. Rearrange the canonical form (3) into standard linear form, $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu + I_\nu = S_\nu$ , identifying $P = 1$ and $Q = S_\nu$ . Why: this exposes the structure that guarantees an integrating-factor solution exists, however complicated $S_\nu(\tau_\nu)$ is.

Step 2. Multiply through by the integrating factor $e^{\int P\,\mathrm{d}\tau_\nu} = e^{\tau_\nu}$ , so the left side becomes an exact derivative: $\frac{\mathrm{d}}{\mathrm{d}\tau_\nu}\!\left(I_\nu\,e^{\tau_\nu}\right) = S_\nu\,e^{\tau_\nu}$ . Why: collapsing two terms into one derivative makes the equation integrable in closed form.

Step 3. Integrate from $0$ to $\tau_\nu$ with $t$ as the dummy variable: $I_\nu(\tau_\nu)e^{\tau_\nu} - I_\nu(0) = \int_0^{\tau_\nu} S_\nu(t)\,e^{t}\,\mathrm{d}t$ . Why: the definite integral carries the boundary value $I_\nu(0)$ explicitly — the light entering the region.

Step 4. Multiply through by $e^{-\tau_\nu}$ to reach $I_\nu(\tau_\nu) = I_\nu(0)e^{-\tau_\nu} + \int_0^{\tau_\nu} S_\nu(t)\,e^{-(\tau_\nu-t)}\,\mathrm{d}t$ . Why: the first term is the background dimmed by all the optical depth in front of it; the second sums photons emitted at each depth $t$ , each dimmed only over the remaining path $\tau_\nu - t$ .

Go deeper · Derivation B · 1 minThe isothermal slab and the origin of Kirchhoff's laws

Derivation BThe isothermal slab and the origin of Kirchhoff's laws

The aim here is to specialize the formal solution to the simplest non-trivial medium — a slab at a single temperature, so that $S_\nu = B_\nu(T)$ is constant along the ray — and to show that this one case already contains the optically thick blackbody, the optically thin emission line, and the absorption line, i.e. all three of Kirchhoff's empirical laws. We pull the constant source function out of the integral, evaluate the elementary exponential, and then read off the two limits, watching how the sign of $B_\nu(T) - I_\nu(0)$ decides between an emission and an absorption feature.

Step 1. With $S_\nu = B_\nu(T)$ constant, factor it out of the formal solution (4): $I_\nu(\tau_\nu) = I_\nu(0)e^{-\tau_\nu} + B_\nu(T)\int_0^{\tau_\nu} e^{-(\tau_\nu-t)}\,\mathrm{d}t$ . Why: constancy is the only assumption needed to make the integral elementary, and isothermal regions model single clouds and thin shells well.

Step 2. Evaluate the integral, $\int_0^{\tau_\nu} e^{-(\tau_\nu-t)}\,\mathrm{d}t = 1 - e^{-\tau_\nu}$ , giving $I_\nu(\tau_\nu) = I_\nu(0)e^{-\tau_\nu} + B_\nu(T)\bigl(1 - e^{-\tau_\nu}\bigr)$ . Why: this is the exact intensity for any slab thickness — the curve in the second figure, a weighted average of background and Planck function with weight $e^{-\tau_\nu}$ .

Step 3. Take the optically thick limit $\tau_\nu \gg 1$ : the background term is suppressed and the second saturates, so $I_\nu \to B_\nu(T)$ independent of $I_\nu(0)$ . Why: after many absorptions and re-emissions the photons thermalize and forget their origin — a deep opaque isothermal body radiates a blackbody (Kirchhoff's first law).

Step 4. Take the optically thin limit $\tau_\nu \ll 1$ , expanding to $I_\nu \approx I_\nu(0)(1-\tau_\nu) + B_\nu(T)\,\tau_\nu$ : the slab imprints an absorption line if $B_\nu(T) < I_\nu(0)$ and an emission line if $B_\nu(T) > I_\nu(0)$ . Why: the sign of $B_\nu(T) - I_\nu(0)$ alone decides emission versus absorption — the content of Kirchhoff's other two laws, and why the same cloud can show a line either way.

Go deeper · Derivation C · 1 minThe Eddington-Barbier relation from a linear source function

Derivation CThe Eddington-Barbier relation from a linear source function

The purpose of this derivation is to answer the practical question every observer asks — "where does the light I measure actually come from?" — by evaluating the emergent intensity from a semi-infinite atmosphere for the simplest realistic depth dependence of the source function, a linear one. We start from the emergent-intensity integral obtained by integrating the plane-parallel equation (5) outward, insert $S_\nu(\tau_\nu) = a_\nu + b_\nu\tau_\nu$ , and use two standard exponential-moment integrals. The remarkably clean result, $I_\nu(0,\mu) = S_\nu(\tau_\nu = \mu)$ , says the emergent beam is a direct readout of the source function one optical depth (in $\mu$ ) below the surface.

Step 1. Integrating (5) from the surface into a semi-infinite atmosphere gives $I_\nu(0,\mu) = \int_0^\infty S_\nu(\tau_\nu)\,e^{-\tau_\nu/\mu}\,\frac{\mathrm{d}\tau_\nu}{\mu}$ . Why: the kernel $e^{-\tau_\nu/\mu}/\mu$ is a normalized weight that concentrates the contribution near $\tau_\nu \approx \mu$ , so most emergent light is born in a thin range of depths.

Step 2. Insert the linear source function $S_\nu(\tau_\nu) = a_\nu + b_\nu\tau_\nu$ : $I_\nu(0,\mu) = \int_0^\infty (a_\nu + b_\nu\tau_\nu)\,e^{-\tau_\nu/\mu}\,\frac{\mathrm{d}\tau_\nu}{\mu}$ . Why: a linear expansion is the first correction beyond a constant, accurate over the one optical depth from which the light emerges, so the result is more general than it looks.

Step 3. Use the exponential moments $\int_0^\infty e^{-\tau/\mu}\frac{\mathrm{d}\tau}{\mu} = 1$ and $\int_0^\infty \tau\,e^{-\tau/\mu}\frac{\mathrm{d}\tau}{\mu} = \mu$ to get $I_\nu(0,\mu) = a_\nu + b_\nu\mu$ . Why: these are the zeroth and first moments of the kernel, and the $\mu$ in the second places the effective formation depth at $\tau_\nu = \mu$ .

Step 4. Recognize $a_\nu + b_\nu\mu = S_\nu(\tau_\nu = \mu)$ , the Eddington-Barbier relation. Why: a face-on view ( $\mu = 1$ ) samples $\tau_\nu = 1$ while a limb view ( $\mu \to 0$ ) samples shallower, cooler layers — directly explaining limb darkening and why line cores, formed high where opacity is large, appear dark.

Limiting cases

The transfer equation collapses into a handful of physically transparent regimes set by the optical depth and by the nature of the extinction:

Case	Result	Meaning
Optically thin ( $\tau_\nu \ll 1$ )	$I_\nu \approx I_\nu(0) + \tau_\nu S_\nu = I_\nu(0) + \int j_\nu\,\mathrm{d}s$	the medium barely attenuates; the emergent intensity is the background plus the integrated emission along the path — emission nebulae, the solar corona, optically thin plasmas
Optically thick ( $\tau_\nu \gg 1$ )	$I_\nu \to S_\nu$	the beam forgets its history and equals the local source function; in LTE this is $B_\nu(T)$ , so a deep isothermal body radiates a blackbody
Local thermodynamic equilibrium	$S_\nu = B_\nu(T) = \dfrac{2h\nu^3/c^2}{e^{h\nu/k_{\mathrm B}T}-1}$	Kirchhoff's law fixes the source function to the Planck function, mapping temperature structure directly onto the spectrum
Pure scattering (coherent, isotropic)	$S_\nu = J_\nu = \frac{1}{4\pi}\oint I_\nu\,\mathrm{d}\Omega$	the source function equals the mean intensity, coupling emission to the field itself; the transfer equation becomes integro-differential with no closed-form solution in general
Free space ( $\alpha_\nu = j_\nu = 0$ )	$I_\nu = \text{const}$	with no matter to edit the beam, specific intensity is conserved along the ray — the reason surface brightness is distance-independent

Common trapOptical depth is not geometric depth, and S_ν is not always the Planck function

Two related errors trip up nearly everyone. First, $\tau_\nu$ is measured along the actual ray path, not straight down into the medium. For a plane-parallel atmosphere viewed at angle $\theta$ from the vertical, the vertical optical depth $\tau_\nu$ and the ray are related through $\mu = \cos\theta$ , so the transfer equation reads $\mu\,\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = I_\nu - S_\nu$ — a factor of $\mu$ and a sign flip relative to the ray form, because $\tau_\nu$ conventionally increases inward while the ray travels outward. A slanted sightline pierces more material per unit height ( $\mathrm{d}s = \mathrm{d}z/\mu$ ), which is exactly why the solar limb, seen at small $\mu$ , looks darker and cooler than disk centre. Second, $S_\nu = B_\nu(T)$ holds only in LTE, where collisions dominate and Kirchhoff's law $j_\nu = \alpha_\nu B_\nu$ applies. The moment scattering matters, $S_\nu = (1-\varepsilon_\nu)J_\nu + \varepsilon_\nu B_\nu$ couples the source function to the radiation field and the problem must be solved self-consistently. Assuming $S_\nu = B_\nu$ universally is the classic beginner's mistake.

Worked examples

1 · Reading optical depth through a dust cloud (confidence builder)

An interstellar dust cloud has optical depth $\tau_V = 1.5$ at visual wavelengths along the line of sight to a background star. What fraction of the star's light gets through, and what is the extinction in magnitudes?

Step 1. With no significant emission from cold dust in the visible, the cloud only attenuates: the transmitted fraction is $e^{-\tau_V} = e^{-1.5} = 0.223$ . So about 22 percent of the starlight reaches us and 78 percent is removed.

Step 2. Extinction in magnitudes is defined from the transmitted flux ratio, $A_V = -2.5\log_{10}(e^{-\tau_V}) = 1.086\,\tau_V$ . Numerically $A_V = 1.086 \times 1.5 = \mathbf{1.63\ mag}$ .

Step 3. The handy rule of thumb $A_\lambda \approx 1.086\,\tau_\lambda$ means one optical depth costs almost exactly one magnitude. Since $\tau_V$ scales with the dust column, measuring the reddening and dimming of background stars maps the dust — the working principle behind extinction surveys. Transmission 22 percent, extinction 1.63 mag, and one optical depth ≈ one magnitude.

2 · Emission line or absorption line? An isothermal radio cloud (fully worked)

A warm interstellar cloud has (Rayleigh-Jeans) brightness temperature $T_c = 100\ \mathrm{K}$ and optical depth $\tau_\nu = 0.5$ at a radio line. In the Rayleigh-Jeans regime intensity is proportional to brightness temperature, so the isothermal-slab solution becomes $T_b = T_{\rm bg}\,e^{-\tau_\nu} + T_c\,(1-e^{-\tau_\nu})$ . Find the observed brightness temperature (a) with no background source and (b) against a 300 K continuum source, and say in each case whether the line appears in emission or absorption.

Step 1. Compute the attenuation factors once: $e^{-0.5} = 0.6065$ and $1-e^{-0.5} = 0.3935$ . These are the weights the slab gives to the background and to its own emission, respectively.

Step 2. Case (a), no background ( $T_{\rm bg} = 0$ ): $T_b = 0 \times 0.6065 + 100 \times 0.3935 = \mathbf{39.3\ K}$ . Against the empty (cold) sky this is a positive feature — an emission line of amplitude 39.3 K, since the cloud is warmer than what lies behind it.

Step 3. Case (b), against a 300 K source: $T_b = 300 \times 0.6065 + 100 \times 0.3935 = 181.96 + 39.35 = \mathbf{221.3\ K}$ . This is below the 300 K continuum, so the line appears in absorption, with a depth of $300 - 221.3 = 78.7\ \mathrm{K}$ .

Step 4. The same cloud, same $T_c$ and $\tau_\nu$ , produces emission or absorption depending only on whether $T_c$ is greater or less than $T_{\rm bg}$ — exactly the $B_\nu(T) \gtrless I_\nu(0)$ criterion from the isothermal-slab derivation, and the microscopic content of Kirchhoff's laws. Emission (+39.3 K) with no background, absorption (−78.7 K) against a hotter one.

3 · Where the continuum forms: the grey atmosphere

In the grey (frequency-independent opacity) approximation with the Eddington approximation, a stellar atmosphere has the temperature structure $T^4(\tau) = \tfrac{3}{4}T_{\rm eff}^4(\tau + \tfrac{2}{3})$ . For the Sun ( $T_{\rm eff} = 5772\ \mathrm{K}$ ), find the temperature at the surface ( $\tau = 0$ ) and at $\tau = \tfrac{2}{3}$ , and connect the result to the Eddington-Barbier relation.

Step 1. At $\tau = \tfrac{2}{3}$ : $T^4 = \tfrac{3}{4}T_{\rm eff}^4(\tfrac{2}{3}+\tfrac{2}{3}) = \tfrac{3}{4}\cdot\tfrac{4}{3}\,T_{\rm eff}^4 = T_{\rm eff}^4$ , so $T = T_{\rm eff} = \mathbf{5772\ K}$ . The effective temperature is the actual gas temperature at $\tau = \tfrac{2}{3}$ .

Step 2. At the surface $\tau = 0$ : $T^4 = \tfrac{3}{4}\cdot\tfrac{2}{3}\,T_{\rm eff}^4 = \tfrac{1}{2}T_{\rm eff}^4$ , so $T = (\tfrac{1}{2})^{1/4}T_{\rm eff} = 0.841 \times 5772 = \mathbf{4854\ K}$ . The topmost visible layers are markedly cooler than $T_{\rm eff}$ .

Step 3. By the Eddington-Barbier relation the disk-centre continuum ( $\mu = 1$ ) forms near $\tau \approx \tfrac{2}{3}$ (the flux-weighted mean, close to $\tau = 1$ ), where $T \approx T_{\rm eff}$ — so defining $T_{\rm eff}$ through the Stefan-Boltzmann law $F = \sigma T_{\rm eff}^4$ is physically the temperature of the layer we actually see. $T(\tfrac{2}{3}) = 5772\ \mathrm{K} = T_{\rm eff}$ and $T(0) = 4854\ \mathrm{K}$ ; the continuum is born near $\tau \approx \tfrac{2}{3}$ .

Physics of the portrait

The transfer equation as astrophysicists use it — radiative equilibrium, the source function, the emergent spectrum tied to conditions at $\tau \approx 1$ — was assembled in the first decade of the twentieth century, and one paper on the Sun's atmosphere did more than any other to put it there.

Origins · 1 minKarl Schwarzschild and how this was discovered

Karl Schwarzschild

1873–1916 · radiative equilibrium and the transfer equation in stellar atmospheres

Karl Schwarzschild was the rare theorist who was also a superb observer and instrument-builder, and he left a mark on nearly every branch of astrophysics he touched. In a landmark 1906 paper on the Sun's atmosphere he argued that energy in a star's outer layers is carried chiefly by radiation rather than convection, and he wrote down the condition of radiative equilibrium together with what is now recognized as the transfer equation — complete with the idea that the emergent intensity is governed by a source function evaluated near optical depth unity. This reframing, that the spectrum forms in a thin skin where photons can finally escape, is the conceptual core of modern stellar spectroscopy, and it explained solar limb darkening quantitatively for the first time. Schwarzschild had already, at 27, become director of the Göttingen observatory and later of the Potsdam Astrophysical Observatory, pioneering precise photographic photometry along the way. When the First World War broke out he volunteered despite being over forty, and it was in the trenches on the Russian front in 1915 that he read Einstein's newly published general relativity and derived the first exact solution of the field equations — the Schwarzschild metric that describes black holes — mailing it to Einstein from the front. He contracted a fatal illness there and died in 1916 at 42. The transfer equation of stellar atmospheres and the geometry of black holes trace to the same remarkable mind.

History

1860

— Kirchhoff states that the ratio of emission to absorption depends only on temperature, defining the source function in equilibrium

1905

— Schuster analyses radiation through a foggy atmosphere, separating scattering from true absorption

1906

— Schwarzschild formulates radiative equilibrium and the transfer equation for the solar atmosphere and explains limb darkening

1926

— Eddington's The Internal Constitution of the Stars codifies radiative transfer in stellar interiors and atmospheres

1950

— Chandrasekhar's monograph Radiative Transfer systematizes exact solutions by the discrete-ordinate method

Going further

Texts. Rybicki & Lightman, Radiative Processes in Astrophysics, Ch. 1 (transfer equation, formal solution Eq. 1.29, source function) · Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., Ch. 9 (Stellar Atmospheres: plane-parallel form, Eddington-Barbier, grey atmosphere) · Mihalas, Stellar Atmospheres, 2nd ed., Ch. 1–2 · Dodelson, Modern Cosmology, Ch. 4 (the Boltzmann/transfer equation for the CMB) · Ryden, Introduction to Cosmology, 2nd ed., Ch. 9 (recombination and last scattering) · Chandrasekhar, Radiative Transfer (Dover) for the classical exact solutions. The key results collected in this article are: (1) the integrating-factor formal solution; (2) the isothermal-slab result $I_\nu = I_\nu(0)e^{-\tau_\nu} + B_\nu(1-e^{-\tau_\nu})$ and its emission/absorption limits (Kirchhoff's laws); (3) the plane-parallel $\mu$ -factor and sign convention; (4) the linear-source-function Eddington-Barbier integrals $I_\nu(0,\mu) = a_\nu + b_\nu\mu$ ; (5) the grey-atmosphere structure $T^4(\tau) = \tfrac{3}{4}T_{\rm eff}^4(\tau + \tfrac{2}{3})$ giving $T(\tfrac{2}{3}) = T_{\rm eff} = 5772$ K and $T(0) = 4854$ K; (6) the radio isothermal-cloud brightness temperatures 39.3 K (emission) and 221.3 K (absorption against 300 K). The CMB temperature $T_0 = 2.725$ K is the COBE/FIRAS value; $A_\lambda = 1.086\,\tau_\lambda$ is the standard extinction relation.

Prerequisites. Specific Intensity and Radiative Flux · Opacity and Absorption Coefficients · Emission and the Source Function · Optical Depth. Leads to. Local Thermodynamic Equilibrium · Kirchhoff's Laws and Thermal Emission · Spectral Line Formation · Blackbody Radiation and the Planck Law.

Ask the tutor: ask me to work the isothermal-slab solution for any cloud and background, or to trace how the same transfer equation becomes the grey stellar atmosphere and the CMB last-scattering surface.

What is the difference between the absorption coefficient α_ν and the optical depth τ_ν?

The absorption coefficient $\alpha_\nu$ is a local property — units of inverse length, the fraction of a beam removed per centimetre right where you are. Optical depth $\tau_\nu = \int\alpha_\nu\,\mathrm{d}s$ is the accumulated total along a whole path, and it is dimensionless: think of it as the number of $e$ -foldings of attenuation a ray has suffered. A dense thin layer and a tenuous long one can share the same $\tau_\nu$ and attenuate a beam identically, because only the product of density, cross-section, and path length matters for what escapes. This is why astronomers quote optical depth rather than a coefficient — it directly controls the observable $e^{-\tau_\nu}$ .

Why does the emergent spectrum come from τ_ν ≈ 1 rather than the surface or the deep interior?

It is a compromise between two competing effects. Photons emitted very deep ( $\tau_\nu \gg 1$ ) are almost all reabsorbed before they can escape — the factor $e^{-\tau_\nu}$ crushes their contribution. Photons emitted right at the surface ( $\tau_\nu \to 0$ ) escape freely, but there is very little material there to emit them. The product of "amount of emitting matter" and "probability of escape" peaks where the two balance, at $\tau_\nu \approx 1$ , which the emergent-intensity kernel $e^{-\tau_\nu/\mu}/\mu$ makes precise; the Eddington-Barbier relation is the sharp statement that we effectively see one optical depth in. Because opacity varies with frequency, $\tau_\nu = 1$ sits at different physical depths at different colours, which is how a single object shows us a range of temperatures.

When exactly can I set the source function equal to the Planck function?

Only in local thermodynamic equilibrium (LTE): when collisions with the ambient thermal particles set the level populations and ionization to their equilibrium (Boltzmann-Saha) values at the local temperature. Then Kirchhoff's law $j_\nu = \alpha_\nu B_\nu(T)$ follows and hence $S_\nu = B_\nu(T)$ . LTE is an excellent approximation deep in stellar atmospheres and interiors, where densities are high and collisions frequent. It breaks down in low-density environments — the outer layers of hot stars, chromospheres and coronae, nebulae, the intergalactic medium — where radiative rates outrun collisional ones and the level populations are set by the radiation field itself. There you must solve the full non-LTE problem, in which $S_\nu = (1-\varepsilon_\nu)J_\nu + \varepsilon_\nu B_\nu$ couples the source function back to the mean intensity.

Why does scattering make the equation so much harder to solve?

Because scattering turns a local equation into a global one. With pure absorption and thermal emission, $S_\nu = B_\nu(T)$ is fixed by the local temperature, so you integrate ray by ray in a single pass. But when photons scatter, $S_\nu = J_\nu$ depends on the mean intensity — an angular average of $I_\nu$ over all directions, and $I_\nu$ is exactly what you are trying to find. The equation becomes integro-differential: the intensity at every point depends on the intensity everywhere else. In general there is no closed form, and one resorts to iterative schemes (Lambda iteration and its accelerated variants), Chandrasekhar's discrete-ordinate method, or Monte Carlo photon transport that follows individual packets through many scatterings.

How does this stellar equation become the equation for the cosmic microwave background?

The transfer equation is the photon-number balance along a ray, and the CMB obeys the same balance with the roles reassigned: $\alpha_\nu$ is the Thomson-scattering rate off free electrons ( $n_e\sigma_{\rm T}$ ), and the source function is the local blackbody because scattering keeps photons and plasma in equilibrium before recombination. Promoting the equation to an expanding universe and linearizing about the smooth background gives the photon Boltzmann equation, whose solution is a line-of-sight integral of exactly the form (4). The visibility function $g(\tau) = e^{-\tau}\mathrm{d}\tau/\mathrm{d}\eta$ — the probability that an observed photon last scattered in a given interval — is the emergent-intensity weighting of that solution, and it peaks sharply at recombination ( $z \approx 1100$ ). Before the peak the plasma is optically thick and the radiation is a perfect blackbody; after it, $\tau$ drops below unity and the photons free-stream to us, preserving that blackbody as the 2.725 K CMB.

What is the physical meaning of the two terms in the formal solution?

It splits the emergent light into two clean contributions. The first term is the background: whatever entered at $\tau_\nu = 0$ , dimmed by the full optical depth it had to traverse, hence the factor $e^{-\tau_\nu}$ . The second is the medium's self-emission: at each depth $t$ the matter adds light proportional to its source function $S_\nu(t)$ , but that light must then travel the remaining path from $t$ to $\tau_\nu$ , so it is attenuated only over that shorter distance, $e^{-(\tau_\nu-t)}$ . Summing all such contributions plus the surviving background gives the total — simply "what came in, dimmed" plus "what was made along the way, each piece dimmed by whatever lay in front of it."