Optical Depth

Optical depth is the single most powerful dimensionless number in radiative transfer, and it earns that status by answering one question at a stroke: can I see through this? Radiation crossing matter is thinned by absorption and scattering; the local rate of removal per unit path is the absorption coefficient $\alpha_\nu$ , an inverse length. Optical depth accumulates that removal along the ray, so $\tau_\nu$ is literally the number of photon mean free paths piled up between two points. Because it is dimensionless and rides in the exponent of the attenuation law, $\tau_\nu$ sorts every medium into two regimes with a single yardstick — whether $\tau_\nu$ is small or large compared with unity — no matter whether the medium is a stellar photosphere, an interstellar cloud, or the reionized intergalactic gas. The crucial subtlety is that "thin" and "thick" are not properties of a substance but of a frequency, a geometry, and a column. A cloud transparent in the radio continuum can be jet-black at the centre of a resonance line, where the opacity spikes by orders of magnitude — and that frequency dependence of $\tau_\nu$ is the whole engine of spectroscopy. Optical depth is also the change of variable that tames the transfer equation, turning it into $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu = I_\nu - S_\nu$ whose formal solution weights each layer's source function by $e^{-\tau_\nu}$ . That weighting is why we see down to $\tau_\nu\sim1$ and no deeper, why the diffuse Sun has a knife-edge limb, and — through the Eddington–Barbier relation — why the emitting surface sits precisely at $\tau_\nu\approx2/3$ . The same three-word statement carries all the way to cosmology: $\tau$ counts interactions, $e^{-\tau}$ is the survival probability, and $\tau\sim1$ is the wall between transparency and opacity.

Key takeaways

Optical depth $\tau_\nu=\int\alpha_\nu\,\mathrm{d}s$ is the dimensionless integral of the absorption coefficient along a ray; it counts how many mean free paths a photon must cross.
A medium is optically thin when $\tau_\nu\ll1$ (photons escape, we see through it) and optically thick when $\tau_\nu\gg1$ (photons are reprocessed, we see only a skin).
In pure absorption the intensity attenuates as $I_\nu=I_{\nu,0}\,e^{-\tau_\nu}$ ; the factor $e^{-\tau_\nu}$ is the Poisson probability that a photon crosses without a single interaction.
Optical depth is the natural variable of radiative transfer: it turns the transfer equation into the Schwarzschild form $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu=I_\nu-S_\nu$ , solved as an $e^{-\tau_\nu}$ -weighted average of the source function.
The photosphere — the effective radiating surface of a star — is the $\tau_\nu\approx2/3$ surface, fixed by the Eddington–Barbier relation.
The identical idea governs cosmology: the Thomson optical depth to reionization $\tau\approx0.054$ (Planck) records CMB photons rescattered by reionized electrons.
Key idea: optical depth is the dimensionless integral $\tau_\nu=\int\alpha_\nu\,\mathrm{d}s$ counting photon mean free paths; $e^{-\tau_\nu}$ is the survival probability, and the $\tau_\nu\sim1$ surface is where the emergent radiation is born.

The core insight

Optical depth is the dimensionless integral $\tau_\nu=\int\alpha_\nu\,\mathrm{d}s$ counting photon mean free paths; $e^{-\tau_\nu}$ is the survival probability, and the $\tau_\nu\sim1$ surface is where the emergent radiation originates.

Where this lives today

Optical depth is the quiet workhorse behind a startling range of modern measurements. Every stellar-atmosphere code — the ones that turn a spectrum into a temperature, gravity, and composition for millions of stars in surveys such as Gaia, APOGEE, and GALAH — integrates the Schwarzschild equation over a $\tau$ grid to synthesize spectra and locate the $\tau=2/3$ photosphere. In the interstellar medium, three-dimensional dust maps built from billions of stellar distances and reddenings are, at bottom, tomographic reconstructions of $\tau_\lambda$ across the Galaxy. Reionization sits at the frontier: the Thomson $\tau$ from Planck's large-scale polarization pins the epoch when the first galaxies ionized the intergalactic gas, and it remains one of the more stubborn cosmological parameters — degenerate with the primordial fluctuation amplitude and a prime target for next-generation CMB experiments such as the Simons Observatory and LiteBIRD. In each arena the calculation is identical: integrate the opacity, exponentiate, and ask whether $\tau$ has crossed one.

The physical picture

Walk toward a lamp through fog. Optical depth is the number of "fog thicknesses" between you and the lamp: at $\tau=1$ the lamp is dimmed to $1/e\approx37\%$ , at $\tau=3$ to $5\%$ , at $\tau=5$ to under $1\%$ . The quantity you accumulate along the path is the expected number of scatterings or absorptions a photon suffers, and the chance it reaches you untouched is the Poisson probability of zero such events, $e^{-\tau}$ . Optical depth is, quite literally, a probability ledger written in the exponent. The first figure turns that ledger into a curve — the survival probability $e^{-\tau}$ against $\tau$ , with the three landmark values every astronomer carries in their head. The transition is not sharp but exponential: nothing special happens at $\tau=1$ except that the medium changes character from a window to a wall.

Figure

Survival probability against optical depth. The transmitted fraction is exp(−τ), passing through the anchor values 0.513 at τ = 2/3 (the photosphere), 0.368 at τ = 1 (one mean free path), and 0.050 at τ = 3

Optical depth also tells you how deep you can see. Because the emergent radiation is an $e^{-\tau}$ -weighted average of the source function, the layers that actually reach your eye are bunched around $\tau\sim1$ : deeper layers hide behind an exponential wall, shallower ones hold too little material to matter. For a source function rising linearly with depth the contribution peaks at $\tau=1$ , and averaging over the visible disc sharpens the effective surface to $\tau\approx2/3$ — the photosphere, which is why the diffuse Sun shows a razor-sharp edge. The second figure is that picture: a plane-parallel atmosphere, the contribution function peaking near one mean free path, and the emergent intensity set by the Eddington–Barbier rule.

Figure

A plane-parallel atmosphere seen from outside. Optical depth grows downward from τ = 0; the layers you actually see are weighted by the contribution function τ exp(−τ), which peaks near τ = 1, and the disc-averaged emergent flux equals π S at the photosphere τ = 2/3

The frequency dependence of $\tau_\nu$ is the whole game in spectroscopy. Where the opacity spikes — at a spectral line — the $\tau_\nu=1$ level is reached higher and cooler, so at that frequency we see a darker surface: an absorption line. Put a hot emitting layer above a faint background instead, and the identical frequency shows up in emission — the sign of $S_\nu-I_{\nu,0}$ decides which.

The core equations

Optical depth is defined differentially as the opacity integrated over an infinitesimal step. By convention it is measured inward from the observer, so it increases as we look deeper into the medium and $\mathrm{d}\tau_\nu$ carries a minus sign relative to the direction the photon actually travels.

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

where: $\tau_\nu$ — optical depth at frequency $\nu$ (dimensionless), measured inward from the observer; $\alpha_\nu$ — absorption coefficient (opacity per unit length, units $\mathrm{m^{-1}}$ ); $\kappa_\nu$ — opacity per unit mass (units $\mathrm{m^2\,kg^{-1}}$ ); $\rho$ — mass density; $\mathrm{d}s$ — path-length element along the ray. Reading it: each step $\mathrm{d}s$ adds $\alpha_\nu\,\mathrm{d}s$ mean free paths to the tally. Integrating along the sightline from a point $s$ inside the medium out to the observer at $s_0$ gives the total optical depth to that point — the accumulated column of opacity.

\tau_\nu(s) = \int_{s}^{s_0} \alpha_\nu(s')\,\mathrm{d}s' \qquad (2)

where: $s_0$ — the observer's position (where $\tau_\nu=0$ ); $s$ — a point along the ray inside the medium. Reading it: $\tau_\nu$ is a running count of opacity between here and the surface; double the column and you double $\tau_\nu$ . With no emission, the transfer equation collapses to pure geometric attenuation. This is the law that gives optical depth its name and its intuition — the survival probability of a photon is the exponential of minus the interaction count.

I_\nu(\tau_\nu) = I_{\nu,0}\,e^{-\tau_\nu} \qquad (3)

where: $I_{\nu,0}$ — incident specific intensity before the medium; $e^{-\tau_\nu}$ — survival probability, the fraction transmitted without a single interaction. Reading it: one mean free path ( $\tau_\nu=1$ ) dims the beam to $37\%$ ; five mean free paths leave under $1\%$ . Restoring emission and dividing the transfer equation through by $\alpha_\nu$ recasts it with $\tau_\nu$ as the independent variable. In this form — the Schwarzschild equation — the messy, material-dependent path is hidden inside $\tau_\nu$ , and only one new quantity appears, the source function $S_\nu$ .

\frac{\mathrm{d}I_\nu}{\mathrm{d}\tau_\nu} = I_\nu - S_\nu, \qquad S_\nu \equiv \frac{j_\nu}{\alpha_\nu} \qquad (4)

where: $S_\nu$ — source function, the ratio of emission to absorption coefficient; $j_\nu$ — emission coefficient (emissivity per unit length and solid angle). Reading it: the intensity is pushed toward $S_\nu$ at a rate set by optical depth; wherever $I_\nu=S_\nu$ the radiation stops changing. Solving equation (4) for a semi-infinite atmosphere gives the formal solution — the emergent intensity as the source function averaged over the ray, each layer weighted by its own survival factor $e^{-\tau_\nu'}$ .

I_\nu(0) = \int_0^{\infty} S_\nu(\tau_\nu')\,e^{-\tau_\nu'}\,\mathrm{d}\tau_\nu' \qquad (5)

where: $I_\nu(0)$ — emergent intensity at the surface $\tau_\nu=0$ ; $e^{-\tau_\nu'}$ — the weight, suppressing each layer by its optical depth to the surface. Reading it: layers near $\tau_\nu'\sim1$ dominate the sum; deeper ones are exponentially silenced. The scale on which all of this happens is the photon mean free path, the reciprocal of the absorption coefficient. Optical depth is nothing but path length measured in units of this length.

\ell_\nu = \frac{1}{\alpha_\nu} = \frac{1}{\kappa_\nu\,\rho}, \qquad \tau_\nu = \int\frac{\mathrm{d}s}{\ell_\nu} \qquad (6)

where: $\ell_\nu$ — mean free path between interactions; $\tau_\nu=1$ is one mean free path. Reading it: a thick medium is simply one that is many mean free paths deep. The same machinery, carried to cosmology, gives the Thomson optical depth of the reionized intergalactic medium to the cosmic microwave background — the integral of the free-electron density times the Thomson cross-section along the line of sight back to reionization.

\tau = \int_0^{z_{\rm re}} n_e(z)\,\sigma_T\,\frac{c\,\mathrm{d}z}{(1+z)\,H(z)} \qquad (7)

where: $n_e$ — free-electron number density; $\sigma_T=6.652\times10^{-29}\,\mathrm{m^2}$ — Thomson cross-section; $z_{\rm re}$ — redshift of reionization; $H(z)$ — Hubble parameter. Reading it: the same $\tau$ , the same $e^{-\tau}$ , now counting scatterings of CMB photons off reionized electrons.

Derivation AThe formal solution of the transfer equation

The goal of this derivation is to convert the Schwarzschild equation $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu=I_\nu-S_\nu$ into an explicit expression for the light that actually leaves the medium, and in doing so to expose why the emergent intensity is an exponentially weighted average of the source function rather than a simple sum. We use nothing more than the integrating-factor trick for a first-order linear ordinary differential equation, plus the boundary condition that a semi-infinite atmosphere leaks no radiation in from infinite depth.

Step 1. Rearrange equation (4) into standard linear form, $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu - I_\nu = -S_\nu$ , with the unknown $I_\nu$ and its derivative on the left and the driving term on the right. Why: this is the canonical shape for which an integrating factor exists, so the whole equation can be written as a single total derivative.

Step 2. Multiply through by the integrating factor $e^{-\tau_\nu}$ ; the left-hand side collapses to a perfect derivative, $\mathrm{d}(I_\nu e^{-\tau_\nu})/\mathrm{d}\tau_\nu = -S_\nu e^{-\tau_\nu}$ . Why: bundling $I_\nu$ and its derivative into one derivative is exactly what makes the equation integrable in closed form.

Step 3. Integrate from the surface $\tau_\nu=0$ to the deep interior $\tau_\nu\to\infty$ , giving $I_\nu(0) - I_\nu(\infty)e^{-\infty} = \int_0^\infty S_\nu e^{-\tau_\nu'}\,\mathrm{d}\tau_\nu'$ . Why: choosing these limits ties the mathematics to the physical geometry — we integrate along the ray from where we stand to where the medium becomes infinitely deep.

Step 4. Drop the interior term: for a semi-infinite atmosphere $I_\nu(\infty)$ is finite while $e^{-\tau_\nu}\to0$ , so their product vanishes and $I_\nu(0)=\int_0^\infty S_\nu(\tau_\nu')\,e^{-\tau_\nu'}\,\mathrm{d}\tau_\nu'$ . Why: this is equation (5), and it makes the central claim concrete — no matter how bright the deep interior, its light is erased by its own optical depth, so we only ever see the outer few mean free paths.

Go deeper · Derivation B · 1 minThe uniform slab and the birth of spectral lines

Derivation BThe uniform slab and the birth of spectral lines

This derivation treats the simplest non-trivial geometry — a homogeneous slab of constant source function $S_\nu$ and total optical depth $\tau_\nu$ , backlit by an intensity $I_{\nu,0}$ — and extracts from it the single result that decides whether a spectral feature appears in emission or absorption. The payoff is that one compact formula smoothly interpolates between a transparent window and an opaque blackbody wall, and its optically thin expansion is Kirchhoff's law written at the level of individual photons.

Step 1. Start from the formal solution over a finite depth, $I_\nu = I_{\nu,0}e^{-\tau_\nu} + \int_0^{\tau_\nu} S_\nu\,e^{-t}\,\mathrm{d}t$ : the backlight attenuated across the slab, plus the slab's own emission weighted by each layer's escape factor. Why: separating the transmitted background from the internally generated light is what lets us see the two competing contributions cleanly.

Step 2. Pull the constant $S_\nu$ out of the integral and evaluate the elementary exponential, $\int_0^{\tau_\nu}e^{-t}\,\mathrm{d}t = 1-e^{-\tau_\nu}$ , to obtain the closed form $I_\nu = I_{\nu,0}\,e^{-\tau_\nu} + S_\nu\bigl(1-e^{-\tau_\nu}\bigr)$ . Why: with a uniform slab the integral is trivial, and the result is the workhorse formula for every homogeneous emitting or absorbing cloud.

Step 3. Take the optically thin limit $\tau_\nu\ll1$ , expand $e^{-\tau_\nu}\approx1-\tau_\nu$ , and collect terms to get $I_\nu\approx I_{\nu,0} + (S_\nu-I_{\nu,0})\,\tau_\nu$ . Why: this is the decisive result — the feature rises above the background if $S_\nu>I_{\nu,0}$ (emission) and dips below it if $S_\nu<I_{\nu,0}$ (absorption), which is Kirchhoff's rule stated microscopically.

Step 4. Take the optically thick limit $\tau_\nu\to\infty$ , where $e^{-\tau_\nu}\to0$ erases the backlight entirely and $I_\nu\to S_\nu$ . Why: a thick slab forgets whatever was behind it and radiates only its own source function — in local thermodynamic equilibrium that is the Planck function $B_\nu(T)$ , and the slab glows as a blackbody.

Go deeper · Derivation C · 1 minThe Eddington–Barbier relation and the two-thirds surface

Derivation CThe Eddington–Barbier relation and the two-thirds surface

Here we pin down the most-quoted number in stellar atmospheres — that the emergent flux of a star is set at optical depth $\tau_\nu\approx2/3$ — and show it is not a fudge but the exact consequence of averaging the emergent intensity over the visible disc. The goal is to derive the photosphere: the effective radiating surface whose temperature we quote as a star's, from a source function assumed linear in optical depth, $S_\nu(\tau_\nu)=a_\nu+b_\nu\tau_\nu$ .

Step 1. Write the formal solution along a ray inclined at angle $\theta$ to the normal, with $\mu=\cos\theta$ : the slant path makes the optical depth along the ray $\tau_\nu/\mu$ , so $I_\nu(0,\mu)=\int_0^\infty S_\nu(\tau_\nu)\,e^{-\tau_\nu/\mu}\,\mathrm{d}\tau_\nu/\mu$ . Why: a tilted sightline traverses more material per unit vertical depth, and that geometric factor $\mu$ is exactly what will produce the two-thirds.

Step 2. Insert the linear source function and evaluate with the standard integrals $\int_0^\infty e^{-x}\mathrm{d}x=1$ and $\int_0^\infty x\,e^{-x}\mathrm{d}x=1$ after substituting $x=\tau_\nu/\mu$ , giving $I_\nu(0,\mu)=a_\nu+b_\nu\mu=S_\nu(\tau_\nu=\mu)$ . Why: the emergent intensity along any ray equals the source function evaluated at $\tau_\nu=\mu$ — you see one mean free path down along that particular line of sight.

Step 3. Average over the stellar disc to get the flux, $F_\nu=2\pi\int_0^1 I_\nu(0,\mu)\,\mu\,\mathrm{d}\mu = \pi\bigl(a_\nu+\tfrac{2}{3}b_\nu\bigr)=\pi\,S_\nu(\tau_\nu=2/3)$ . Why: the extra factor of $\mu$ weights the disc toward its centre, and the ratio $\int_0^1\mu^2\,\mathrm{d}\mu/\int_0^1\mu\,\mathrm{d}\mu=(1/3)/(1/2)=2/3$ is precisely where the famous number comes from.

Step 4. Read off the physics: the emergent flux equals $\pi$ times the source function at $\tau_\nu=2/3$ , so the layer whose temperature the star wears on its sleeve is the $\tau_\nu\approx2/3$ surface — the photosphere. Why: this is what makes "the surface of a star" a well-defined idea for a body that is gas all the way down, and it is why line photons (higher opacity, hence $\tau_\nu=2/3$ reached higher and cooler) appear dark.

Limiting cases

The entire behaviour of a medium is decided by where $\tau_\nu$ sits relative to unity — and, in the cosmological case, by a specific measured value.

Case	Result	Meaning
Optically thin	$\tau_\nu\ll1$	Photons pass with little chance of interaction; $I_\nu\approx I_{\nu,0}+(S_\nu-I_{\nu,0})\tau_\nu$ . We see through the gas and observe a signal proportional to the column of emitters — the regime of optically thin nebular lines.
Optically thick	$\tau_\nu\gg1$	Photons are absorbed and re-emitted many times; the emergent intensity saturates at $I_\nu\to S_\nu$ . In local thermodynamic equilibrium this is $B_\nu(T)$ — the surface radiates as a blackbody.
Photosphere	$\tau_\nu\approx2/3$	The Eddington–Barbier surface: the disc-averaged emergent flux equals $\pi S_\nu(2/3)$ . This defines the effective radiating surface of a star and the temperature we assign it.
Reionization (cosmology)	$\tau\approx0.054$	The Thomson optical depth to reionization measured by Planck: only about $5\%$ of CMB photons rescatter off reionized electrons, so the late Universe is optically thin even though it was thick before recombination.

Common trapA medium is not simply "thin" or "thick"

The commonest error is to attach thinness or thickness to a substance. Optical depth is frequency-dependent and geometry-dependent: the very same cloud can be transparent in the radio continuum and opaque at the centre of a resonance line, or transparent face-on and opaque edge-on. Always ask "thick at which frequency, along which path?" Two further traps. First, the sign convention: $\tau_\nu$ is measured inward from the observer, so $\mathrm{d}\tau_\nu=-\alpha_\nu\,\mathrm{d}s$ along the direction of photon travel — get the sign wrong and your attenuation becomes amplification. Second, do not confuse the three related quantities: optical depth $\tau_\nu$ is dimensionless, the absorption coefficient $\alpha_\nu$ is per unit length, and the opacity $\kappa_\nu$ is per unit mass, with $\alpha_\nu=\kappa_\nu\rho$ .

Worked examples

1 · Dust extinction toward a background star (fully worked)

A star lies behind a dust cloud whose visual optical depth is $\tau_V=2.0$ . Find the transmitted fraction of its visual light, the extinction in magnitudes, and — using the standard Galactic reddening law $A_B/A_V=1.32$ — the colour excess $E(B-V)$ and the ratio of total to selective extinction.

Step 1. The transmitted fraction is the survival probability, $e^{-\tau_V}=e^{-2.0}=\mathbf{0.135}$ : only about $14\%$ of the star's visual light reaches us; the cloud removes the rest by absorption and scattering off dust grains.

Step 2. Convert optical depth to magnitudes. Since a flux ratio $e^{-\tau_\lambda}$ corresponds to $\Delta m=-2.5\log_{10}e^{-\tau_\lambda}=2.5(\log_{10}e)\,\tau_\lambda=1.086\,\tau_\lambda$ , the visual extinction is $A_V=1.086\times2.0=\mathbf{2.17\ mag}$ . The factor $1.086$ is just $2.5\log_{10}e$ and is worth memorizing.

Step 3. The blue-band extinction follows from the reddening law: $A_B=1.32\,A_V=1.32\times2.17=2.87$ mag. The colour excess — the reddening — is the difference $E(B-V)=A_B-A_V=2.87-2.17=\mathbf{0.70\ mag}$ . Starlight is not merely dimmed but reddened, because $\tau_\lambda$ is larger in the blue.

Step 4. The ratio of total to selective extinction is $R_V=A_V/E(B-V)=2.17/0.70=\mathbf{3.1}$ , the canonical diffuse-interstellar-medium value. Consistency here is the check: a sightline that returned $R_V$ far from $3.1$ would signal unusual grain sizes rather than an arithmetic slip. Transmitted 13.5%, $A_V=2.17$ mag, $E(B-V)=0.70$ mag, $R_V=3.1$ .

2 · The Thomson optical depth to reionization (fully worked)

Estimate the CMB optical depth through a Universe reionized instantaneously at $z_{\rm re}=7.7$ , taking a present-day free-electron density $n_{e,0}\approx0.20\ \mathrm{m^{-3}}$ (hydrogen and helium once-ionized), Hubble length $c/H_0=1.37\times10^{26}\ \mathrm{m}$ , and matter and dark-energy densities $\Omega_m=0.315$ , $\Omega_\Lambda=0.685$ . Compare with Planck's measured $\tau\approx0.054$ .

Step 1. Write equation (7) with the physical electron density scaling as $n_e(z)=n_{e,0}(1+z)^3$ . The volume-element factor $c/[(1+z)H(z)]$ cancels one power of $(1+z)$ , leaving $\tau=\sigma_T\,n_{e,0}\,(c/H_0)\int_0^{z_{\rm re}}\dfrac{(1+z)^2\,\mathrm{d}z}{E(z)}$ , with $E(z)=\sqrt{\Omega_m(1+z)^3+\Omega_\Lambda}$ .

Step 2. Evaluate the dimensionless integral. Matter dominates over almost the whole range, so $E(z)\approx\sqrt{\Omega_m}\,(1+z)^{3/2}$ and the integrand behaves as $(1+z)^{1/2}/\sqrt{\Omega_m}$ ; carrying the full $E(z)$ numerically gives $\int_0^{7.7}(1+z)^2\,\mathrm{d}z/E(z)\approx\mathbf{28}$ .

Step 3. Multiply the pieces: $\tau=(6.652\times10^{-29})(0.20)(1.37\times10^{26})(28)$ . Group them as $6.652\times10^{-29}\times0.20=1.33\times10^{-29}$ , then $\times1.37\times10^{26}=1.82\times10^{-3}$ , then $\times28=\mathbf{0.051}$ .

Step 4. Compare and interpret. The estimate $\tau\approx0.05$ sits right on Planck's $\tau\approx0.054$ , remarkable given the crude sudden-ionization assumption. Physically only about $1-e^{-0.05}\approx5\%$ of CMB photons rescatter after reionization, which is why the effect shows up as a modest $e^{-2\tau}$ damping of the temperature anisotropies plus a large-scale polarization bump — small, but one of the cleanest handles we have on when the first sources lit up. $\tau\approx0.05$ , matching Planck's $0.054$ .

Physics of the portrait

Optical depth as we use it — the independent variable of the transfer equation, the yardstick of the photosphere — was forged in a single remarkable paper on the Sun's atmosphere, written by a man better remembered today for a black hole.

Origins · 1 minKarl Schwarzschild and how this was discovered

Karl Schwarzschild

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

Karl Schwarzschild is famous for the exact solution of Einstein's field equations he found in 1916, from a trench on the Russian front months before his death — the metric that defines a black hole's horizon. But a decade earlier he had already reshaped a different subject. In his 1906 paper on the equilibrium of the solar atmosphere, Schwarzschild argued that energy near the visible surface of the Sun is carried by radiation rather than convection, and to make the idea quantitative he introduced the machinery we still teach: optical depth as the natural coordinate, and the transfer equation in the form $\mathrm{d}I_\nu/\mathrm{d}\tau_\nu=I_\nu-S_\nu$ that now bears his name. From radiative equilibrium he derived the outward temperature decline of the photosphere and explained solar limb darkening as its signature — the limb, seen along a slanted ray, reaches $\tau_\nu=2/3$ higher and cooler and so looks fainter. Arthur Eddington built the theory of stellar interiors on the same foundation through the 1920s, and with Daniel Barbier sharpened the emergent-flux rule to the $\tau_\nu=2/3$ surface. Schwarzschild's optical depth turned stellar atmospheres from description into calculable physics — and the same variable, unchanged, now measures the reionized Universe's transparency to the microwave background.

History

1905

— Schuster models radiation escaping a foggy, scattering stellar atmosphere

1906

— Schwarzschild introduces optical depth and radiative equilibrium for the Sun and explains limb darkening

1926

— Eddington founds the radiative theory of stellar interiors in "The Internal Constitution of the Stars"

1943

— Barbier sharpens the emergent-flux rule to the τ = 2/3 photosphere

2018

— Planck measures the Thomson optical depth to reionization, τ ≈ 0.054

Going further

Texts. Rybicki and Lightman, Radiative Processes in Astrophysics, Ch. 1 (definition, formal solution, uniform slab) · Carroll and Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §9.2 · Mihalas, Stellar Atmospheres, 2nd ed., Ch. 2 (Eddington–Barbier, limb darkening) · Dodelson, Modern Cosmology, 2nd ed., Ch. 3 (Thomson optical depth) · Ryden, Introduction to Cosmology, 2nd ed., Ch. 9 (reionization).

Provenance. The differential definition, the exponential attenuation law, the uniform-slab result $I_\nu=I_{\nu,0}e^{-\tau_\nu}+S_\nu(1-e^{-\tau_\nu})$ , and the formal solution follow Rybicki and Lightman Ch. 1 and Carroll and Ostlie §9.2. The Eddington–Barbier $\tau=2/3$ surface is standard (Mihalas Ch. 2); the two-thirds arises from the disc-average moments $\int_0^1\mu^2\,\mathrm{d}\mu=1/3$ and $\int_0^1\mu\,\mathrm{d}\mu=1/2$ , ratio $2/3$ . Numerical anchors: Thomson cross-section $\sigma_T=6.652\times10^{-29}\,\mathrm{m^2}$ (CODATA); reionization optical depth $\tau\approx0.054$ (Planck 2018); the extinction relation $A_\lambda=1.086\,\tau_\lambda$ from $2.5\log_{10}e$ ; the reionization worked example reproduces $\tau\approx0.05$ from equation (7) with $z_{\rm re}=7.7$ , $n_{e,0}=0.20\ \mathrm{m^{-3}}$ , matching Planck within the crudeness of the sudden-reionization assumption.

Prerequisites. Opacity and Absorption Coefficients · Specific Intensity and Radiative Flux · The Equation of Radiative Transfer. Leads to. Emission and the Source Function · Spectral Line Formation · Kirchhoff's Laws and Thermal Emission · Local Thermodynamic Equilibrium.

Ask the tutor: Ask me why the photosphere sits at $\tau=2/3$ rather than $\tau=1$ , or to reproduce Planck's reionization optical depth from the electron density.

Is optical depth the same thing as opacity?

No, though the words are often blurred. Opacity is a local property of the material: $\kappa_\nu$ per unit mass, or the absorption coefficient $\alpha_\nu=\kappa_\nu\rho$ per unit length, telling you how strongly a small parcel removes radiation. Optical depth is the integral of that opacity along a whole path, $\tau_\nu=\int\alpha_\nu\,\mathrm{d}s$ , and it is dimensionless. You can raise the optical depth either by making the gas more opaque or simply by looking through more of it — a tenuous but enormous cloud can have a large $\tau_\nu$ , and a dense but thin skin can have a small one. Opacity is the rate; optical depth is the accumulated total.

Why measure τ inward from the observer, giving a minus sign?

It is a convention chosen so that $\tau_\nu$ increases as you go deeper into a star, which matches how atmosphere models are built — layer by layer from the surface down. Because the photon travels outward, toward decreasing $\tau_\nu$ , the differential $\mathrm{d}\tau_\nu=-\alpha_\nu\,\mathrm{d}s$ picks up a minus sign relative to the direction of travel. Nothing physical depends on the choice; you could measure $\tau_\nu$ along the ray instead and flip the signs consistently. The one thing you must not do is mix the conventions halfway through a calculation, which turns the attenuation $e^{-\tau_\nu}$ into a spurious exponential growth.

How can the diffuse Sun have a sharp visible edge?

Because visibility is set by optical depth, not by any change of phase. As your sightline slides from the disc centre toward the limb, it grazes the atmosphere at an ever-shallower angle, so it reaches $\tau_\nu=2/3$ while still high in the cool, thin outer gas — and beyond that surface the gas is opaque to you. The transition from "we see in" to "we cannot" happens over a scale height that is tiny compared with the solar radius, so the edge looks razor-sharp even though the density falls off smoothly. The same effect, seen in projection, is limb darkening: the limb shows you cooler, higher layers than the disc centre, so it looks fainter.

What is the source function, and why does it appear everywhere here?

The source function $S_\nu=j_\nu/\alpha_\nu$ is the ratio of how much a parcel of gas emits to how much it absorbs. It is the value the intensity is always being driven toward: wherever $I_\nu=S_\nu$ , the transfer equation says the intensity stops changing. In local thermodynamic equilibrium, emission and absorption are locked together by Kirchhoff's law and $S_\nu$ becomes the Planck function $B_\nu(T)$ , so an optically thick medium simply glows as a blackbody at its local temperature. When scattering dominates instead, $S_\nu$ is set by the ambient radiation field rather than the local temperature, and the problem becomes genuinely non-local — which is where the full theory of radiative transfer begins.

Does a large optical depth mean no light gets out at all?

Not at all — it means no light gets through from behind, but the medium radiates plenty of its own. A thick slab erases whatever was behind it ( $e^{-\tau_\nu}\to0$ ) yet emerges at $I_\nu\to S_\nu$ , the blackbody intensity of its own surface layer. A star is the example: enormously optically thick, so we see nothing of its interior, yet its photosphere pours out a near-blackbody spectrum. Large $\tau_\nu$ hides the inside and replaces it with a surface; it does not make the object dark.

How does the same optical depth idea reach from stars to cosmology?

The physics is identical; only the interaction changes. In a star the removal is bound-bound and bound-free absorption plus scattering by atoms; in the reionized intergalactic medium it is Thomson scattering of CMB photons off free electrons. In both cases you integrate an interaction rate (an absorption coefficient, or $n_e\sigma_T$ ) along the line of sight to get a dimensionless $\tau$ , and in both cases $e^{-\tau}$ is the survival probability. The cosmological twist is only bookkeeping — the path element is written in redshift, $\mathrm{d}s=c\,\mathrm{d}z/[(1+z)H(z)]$ — but the meaning of $\tau\approx0.054$ is exactly the meaning of $\tau=0.054$ for a fog: about $5\%$ of the photons scattered on the way to you.