Spectral Line Formation

A spectral line is the fingerprint of a quantized transition. When a bound electron jumps between two discrete levels of an atom, ion, or molecule, the photon it absorbs or emits carries exactly the energy difference, $h\nu = E_{\mathrm{upper}} - E_{\mathrm{lower}}$ . Because the levels are sharp, the interaction is concentrated in a narrow band of frequencies — a dark absorption line cut into a bright continuum, or a bright emission line against a dark background. The wavelength names the species: the pattern of lines is a barcode that lets us read the chemistry of a star we will never touch. But line formation is more than identification. Three physical strands must be braided together to predict a line. Statistical mechanics — the Boltzmann excitation law and the Saha ionization equation — fixes how many atoms sit in the lower level of the transition. Quantum mechanics — the oscillator strength or Einstein coefficients — fixes how eagerly each atom jumps. And radiative transfer carries the result outward: the line absorption coefficient enters the transfer equation like the continuum opacity but with a sharply peaked frequency dependence, so the emergent line reflects the run of temperature and source function through the atmosphere. Finally, the line is never infinitely sharp. Natural, Doppler, and pressure broadening each stamp the profile with a characteristic shape, and the curve of growth converts the measured equivalent width into a column density. This is why a single dark notch becomes a quantitative measurement of temperature, pressure, composition, velocity, and magnetic field — the engine of stellar and interstellar spectroscopy.

Key takeaways

Spectral lines arise from bound-bound transitions at frequencies fixed by the level spacing, $\Delta E = h\nu$ ; the wavelength identifies the atom, ion, or molecule.
Line strength is the product of three factors — ionization stage (Saha), excitation fraction (Boltzmann), and intrinsic transition probability (oscillator strength) — so a line can be weak even when the element is abundant.
Absorption versus emission is decided by the source function relative to the background: extra line opacity pushes the $\tau_\nu = 1$ surface into higher, cooler layers where $S_\nu$ is smaller, darkening the line.
The profile records the environment: a Gaussian Doppler core measures temperature, Lorentzian pressure wings measure density, and their convolution is the Voigt profile.
The curve of growth links the equivalent width $W_\lambda$ to the column density $N_\ell$ : linear when weak, flat when saturated, square-root when damping-dominated.
Key idea: A spectral line is radiative transfer with a sharply frequency-peaked opacity — its position identifies the atom, its strength counts absorbers through Boltzmann and Saha, and its shape records the velocity and pressure environment through the profile function.

The core insight

A spectral line is radiative transfer with a sharply frequency-peaked opacity: its position identifies the atom, its strength counts absorbers via Boltzmann and Saha, and its shape records the velocity and pressure environment through the profile function.

Where this lives today

Line formation is the daily engine of stellar astrophysics, and its modern incarnation is industrial in scale. Massive spectroscopic surveys — APOGEE, GALAH, and Gaia's RVS spectrograph — measure equivalent widths and full profiles for millions of stars, then feed them to automated synthesis codes that extract temperatures, gravities, microturbulence, and abundances of dozens of elements per star. This is the backbone of Galactic archaeology, which reconstructs the Milky Way's assembly history from the chemical fingerprints of its stars. At the precision frontier, spectrographs such as ESPRESSO on the VLT track Doppler shifts of thousands of lines to below $10$ cm/s, hunting for the reflex wobble of Earth-mass exoplanets — a measurement that lives or dies on how line profiles are distorted by stellar surface activity. And where classical analysis assumed LTE, current work solves the coupled statistical-equilibrium and transfer equations in full 3D model atmospheres, because the last decimal of an abundance now bears on questions from the primordial lithium problem to the ages of the oldest stars.

The physical picture

Picture continuum photons streaming upward through a stellar atmosphere. At most wavelengths, the gas is fairly transparent, and the photons we see were launched from deep, hot layers. But at the exact wavelength of a transition the gas is suddenly opaque: those photons escape only from the thin, cool upper layers, and cooler gas radiates less, so that wavelength looks dark. An absorption line is nothing more than a window into a higher, cooler part of the atmosphere — cooler because temperature falls outward through a stellar photosphere; where it instead rises outward, as in a chromosphere, the same extra opacity produces emission lines — and the first figure makes "higher opacity means a higher, colder escape surface" a single picture.

Figure

Why an absorption line is dark. At most wavelengths, photons escape from deep hot layers where τ_ν = 1 sits near T ≈ 8000 K, giving a bright continuum. At line center, the extra opacity lifts the τ_ν = 1 surface into high cool layers near T ≈ 5000 K, so the emergent intensity I_ν(0) ≈ S_ν(τ_ν = 1) drops and that wavelength looks dark.

The strength of a line is a product of three main factors: how many atoms are in the right ionization stage (Saha), how many of those sit in the right lower level (Boltzmann), and how eagerly each atom makes the jump (the oscillator strength). Shrink any one factor and the line is weak, however abundant the element, which is why the hydrogen that dominates most stellar atmospheres still produces only feeble Balmer lines in a cool K dwarf and very weak ones in a sufficiently hot O star. The shape of a line is a stopwatch and a barometer at once: thermal motions smear it into a Gaussian core whose width measures temperature, while collisions in dense gas broaden the Lorentzian wings, measuring pressure — but to turn that into a number, you must know where the line sits on the curve of growth, the second figure.

Figure

The curve of growth: how the equivalent width W_λ grows with the column density of absorbers N_ℓ. Weak lines grow linearly (slope 1, W_λ ∝ N_ℓ), saturated lines flatten (W_λ ∝ √ln N_ℓ) as the core goes black, and strong lines resume growth on the damping wings (slope 1/2, W_λ ∝ √N_ℓ Γ).

The core equations

The starting point is the Bohr frequency condition: a bound-bound transition emits or absorbs a photon whose energy is exactly the level separation. This is what makes the line a barcode — the wavelength depends only on the atom, not on temperature or density.

h\nu = E_{\mathrm{upper}} - E_{\mathrm{lower}} \qquad (1)

where: $h$ — Planck constant, $6.626\times10^{-34}\ \mathrm{J\,s}$ ; $\nu$ — photon frequency of the transition; $E_{\mathrm{upper,lower}}$ — energies of the upper and lower bound levels. Reading: the wavelength $\lambda = c/\nu$ is a fixed property of the species, so a line center identifies the atom or ion. How many atoms populate a given level is set by the Boltzmann law: in equilibrium, the ratio of two-level populations is the ratio of statistical weights times the Boltzmann factor of the energy gap. This is the excitation half of "how many absorbers are available."

\frac{n_b}{n_a} = \frac{g_b}{g_a}\,e^{-(E_b-E_a)/k_BT} \qquad (2)

where: $n_a,n_b$ — number densities in the lower level $a$ and upper level $b$ ; $g_a,g_b$ — statistical weights (degeneracies); $k_B$ — Boltzmann constant, $1.381\times10^{-23}\ \mathrm{J\,K^{-1}}$ ; $T$ — excitation temperature. Reading: higher levels become significantly populated when $k_BT$ approaches the excitation energy, which is why high-excitation lines need hot gas. The Saha equation is the ionization counterpart: it gives the ratio of adjacent ionization stages, and unlike Boltzmann, it depends on electron density, because recombination needs an electron to find the ion. The $T^{3/2}$ factor is the phase-space volume available to the freed electron.

\frac{N_{i+1}}{N_i}\,n_e = \frac{2\,Z_{i+1}}{Z_i}\left(\frac{2\pi m_e k_B T}{h^2}\right)^{3/2} e^{-\chi_i/k_BT} \qquad (3)

where: $N_i,N_{i+1}$ — number densities in ionization stages $i$ and $i+1$ ; $n_e$ — electron number density; $Z_i$ — partition function of stage $i$ ; $m_e$ — electron mass, $9.109\times10^{-31}\ \mathrm{kg}$ ; $\chi_i$ — ionization energy. Reading: raise $T$ and the gas ionizes; raise the electron pressure and recombination pushes it back down. These populations feed the line absorption coefficient, which enters the transfer equation like continuum opacity but concentrated by the normalized profile $\phi(\nu)$ . Its integrated strength is fixed by fundamental constants times the dimensionless oscillator strength.

\alpha_\nu^{\mathrm{line}} = \frac{\pi e^2}{m_e c}\, f_{\ell u}\, n_\ell\,\phi(\nu) \qquad (4)

where: $\alpha_\nu^{\mathrm{line}}$ — line opacity per unit length (CGS); $e,m_e,c$ — electron charge, mass, and light speed; $f_{\ell u}$ — oscillator strength of the lower-to-upper transition; $n_\ell$ — lower-level number density; $\phi(\nu)$ — normalized profile with $\int\phi\,\mathrm{d}\nu = 1$ . Reading: the constant $\pi e^2/m_e c$ is the frequency-integrated classical cross-section — the cross-section itself is this constant times $f_{\ell u}\,\phi(\nu)$ — and $f_{\ell u}$ carries the quantum mechanics. The emergent intensity follows from the formal solution of radiative transfer. Expanded near the surface (Eddington-Barbier), it says the intensity we see equals the source function at optical depth unity along the ray — the single fact behind absorption lines.

I_\nu(0) = I_\nu(\tau_\nu)e^{-\tau_\nu} + \int_0^{\tau_\nu} S_\nu(t)\,e^{-t}\,\mathrm{d}t \;\;\Longrightarrow\;\; I_\nu(0) \approx S_\nu(\tau_\nu = 1) \qquad (5)

where: $I_\nu(0)$ — emergent specific intensity; $\tau_\nu$ — optical depth, $\mathrm{d}\tau_\nu = \alpha_\nu\,\mathrm{d}s$ ; $S_\nu = j_\nu/\alpha_\nu$ — source function. Reading: at line center the extra opacity brings $\tau_\nu = 1$ to a higher, cooler layer, so $S_\nu \approx B_\nu(T)$ is smaller and the line goes dark. The profile itself is the convolution of a Gaussian Doppler core with a Lorentzian damping wing — the Voigt profile. Its two width parameters are the temperature (through the Doppler width) and the pressure-plus-lifetime damping (through $\Gamma$ ).

\phi(\nu) = \frac{1}{\Delta\nu_D\sqrt{\pi}}\,H(a,u),\qquad a=\frac{\Gamma}{4\pi\Delta\nu_D},\quad u=\frac{\nu-\nu_0}{\Delta\nu_D} \qquad (6)

where: $H(a,u)$ — Voigt (Hjerting) function; $\Delta\nu_D = \nu_0\,v_{\mathrm{th}}/c$ — Doppler width with $v_{\mathrm{th}}=\sqrt{2k_BT/m}$ ; $\Gamma$ — total damping (natural + collisional); $\nu_0$ — line center. Reading: small $a$ means a nearly Gaussian, thermally broadened line; large $a$ means strong Lorentzian wings. Because thermal motion is rarely the whole story, the Doppler width in practice adds a non-thermal microturbulent term in quadrature — thermal motion dominates for light atoms, turbulence can win for heavy ones.

\Delta\lambda_D = \frac{\lambda_0}{c}\sqrt{\frac{2k_BT}{m} + \xi^2} \qquad (7)

where: $\lambda_0$ — line-center wavelength; $m$ — mass of the absorbing atom; $\xi$ — microturbulent velocity; $T$ — kinetic temperature. Reading: the thermal term scales as $1/\sqrt{m}$ , so an iron line is far narrower than a hydrogen line at the same temperature. Finally, the observable that ties everything to abundance is the equivalent width — the width of a perfectly black rectangle that removes the same total flux as the line. It is resolution-independent, measurable even when the profile is unresolved.

W_\lambda = \int \frac{F_c - F_\lambda}{F_c}\,\mathrm{d}\lambda \qquad (8)

where: $W_\lambda$ — equivalent width; $F_c$ — continuum flux interpolated across the line; $F_\lambda$ — observed flux in the line. Reading: the curve of growth maps $W_\lambda$ to the column density $N_\ell$ — the number the whole enterprise is built to extract.

Derivation AThe weak-line linear curve of growth

The goal is to prove the cornerstone of quantitative spectroscopy: for an unsaturated line the equivalent width is directly proportional to the column density of absorbers, $W_\lambda \approx (\pi e^2/m_e c^2)\,\lambda_0^2\, f_{\ell u}\, N_\ell$ . We start from the emergent flux for a slab of pure line absorption over a continuum, take the optically thin limit so the exponential linearizes, and integrate over the line, using the profile's normalization to collapse the frequency dependence into a single fundamental-constant coefficient.

Step 1. Write the fractional line depth as $(F_c - F_\lambda)/F_c = 1 - e^{-\tau_\lambda} \approx \tau_\lambda$ . Why: for a weak line $\tau_\lambda \ll 1$ , so the residual depth is simply the optical depth — this is the approximation that makes the whole thing linear.

Step 2. Convert opacity to a per-column form: $\tau_\lambda = (\pi e^2/m_e c)\, f_{\ell u}\, N_\ell\, \phi(\nu)$ (evaluated at $\nu = c/\lambda$ ), where $N_\ell = \int n_\ell\,\mathrm{d}s$ is the column density along the line of sight. Why: the equivalent width depends only on how many absorbers lie along the ray, not on how they are distributed in depth.

Step 3. Integrate over the line, $W_\lambda = \int \tau_\lambda\,\mathrm{d}\lambda = (\pi e^2/m_e c)\, f_{\ell u}\, N_\ell \int \phi(\nu)\,\mathrm{d}\lambda$ , ready to collapse with the profile normalization $\int \phi\,\mathrm{d}\nu = 1$ . Why: the entire shape of the profile drops out because it is normalized — the equivalent width does not care whether the line is Doppler- or pressure-broadened, only how much total absorption it carries.

Step 4. Change variables with $\nu = c/\lambda$ , so $\mathrm{d}\nu = -(c/\lambda_0^2)\,\mathrm{d}\lambda$ , so $\int \phi(\nu)\,\mathrm{d}\lambda = (\lambda_0^2/c)\int \phi(\nu)\,\mathrm{d}\nu = \lambda_0^2/c$ at line center. Collecting terms gives $W_\lambda = (\pi e^2/m_e c^2)\,\lambda_0^2\, f_{\ell u}\, N_\ell$ . Why: the equivalent width is exactly linear in $N_\ell$ and $f_{\ell u}$ — this is why weak lines are the gold-standard abundance diagnostic and why the "gf-value" is tabulated for every useful line. *The coefficient $\pi e^2/m_e c^2 = 8.85\times10^{-13}$ cm (CGS); a dimensional check confirms that $W_\lambda$ carries units of length, as expected for an equivalent width.*

Go deeper · Derivation B · 1 minBoltzmann and Saha combine to set the lower-level population

Derivation BBoltzmann and Saha combine to set the lower-level population

The goal is to show why line strength peaks at a characteristic temperature rather than climbing forever with heat, by writing the fraction of an element sitting in the lower level of a transition as the product of an ionization fraction (Saha) and an excitation fraction (Boltzmann). This single product explains the entire OBAFGKM sequence — it is the payoff that turned stellar classification into a temperature scale.

Step 1. The excitation fraction from Boltzmann is $n_\ell/N_i = (g_\ell/Z_i(T))\,e^{-E_\ell/k_BT}$ , the fraction of ionization stage $i$ that sits in the specific lower level $\ell$ . Why: only atoms already in level $\ell$ can absorb the line, and that fraction rises steeply with temperature as $k_BT$ approaches $E_\ell$ .

Step 2. The ionization fraction from Saha is $N_i/N_{\mathrm{elem}} = (1 + N_{i+1}/N_i + \dots)^{-1}$ , obtained by summing the Saha ratios over all stages and normalizing. Why: as temperature climbs, atoms are stripped out of stage $i$ into $i+1$ , so this fraction eventually collapses even as excitation is still rising.

Step 3. Multiply the two: $n_\ell/N_{\mathrm{elem}} = (N_i/N_{\mathrm{elem}})\times(g_\ell/Z_i)\,e^{-E_\ell/k_BT}$ . Raising $T$ pushes the Boltzmann factor up but the Saha factor down, so the product peaks at an intermediate temperature. Why: two opposing exponentials in temperature can produce a maximum — the line is strongest where excitation has turned on but ionization has not yet emptied the stage.

Step 4. Apply to hydrogen: the $n=2$ level that absorbs Balmer photons is barely populated below $\sim 8000$ K, but above $\sim 11000$ K hydrogen is largely ionized. The window between is the A-star regime near $9500$ K, exactly where Balmer lines peak. Why: this quantitative peak, first computed in the 1920s, is the direct microphysical origin of the spectral sequence. *The $n=2$ Boltzmann fraction rises by a factor $\approx 2700$ from 6000 K to 10000 K, directly explaining the Balmer-strength maximum in A-type stars.*

Go deeper · Derivation C · 1 minNatural line width from the finite lifetime

Derivation CNatural line width from the finite lifetime

The goal is to derive the Lorentzian natural profile and its width from a damped-oscillator argument, and to show that the same Lorentzian form governs pressure broadening. An excited level with finite lifetime $\Delta t$ cannot have a perfectly sharp energy, by the energy-time uncertainty relation, and this intrinsic fuzziness sets the minimum width any line can have — collisions, which shorten the effective lifetime, simply add to $\Gamma$ .

Step 1. Model the emitted field as a damped oscillation, $E(t) \propto e^{-i2\pi\nu_0 t}\,e^{-\Gamma t/2}$ , decaying with damping constant $\Gamma$ equal to the inverse lifetime. Why: an excited atom radiates a finite wavetrain, not an eternal sine wave, and a finite wavetrain cannot be monochromatic.

Step 2. Fourier transform to frequency, $\tilde{E}(\nu) \propto [\,i2\pi(\nu-\nu_0) + \Gamma/2\,]^{-1}$ . Why: the transform of a decaying exponential is a complex Lorentzian amplitude — the finite duration spreads the energy over a band of frequencies.

Step 3. Square to get the power spectrum, $|\tilde{E}(\nu)|^2 \propto [\,(2\pi)^2(\nu-\nu_0)^2 + (\Gamma/2)^2\,]^{-1}$ , which normalizes to the Lorentzian $\phi(\nu) = (\Gamma/4\pi^2)/[(\nu-\nu_0)^2 + (\Gamma/4\pi)^2]$ . Why: this is the intrinsic natural profile — sharply peaked at $\nu_0$ with slowly falling wings that never quite vanish.

Step 4. Read off the width, $\Delta\nu_{\mathrm{FWHM}} = \Gamma/2\pi \sim 1/(2\pi\,\Delta t)$ , consistent with $\Delta E\,\Delta t \sim \hbar$ . Collisions interrupt the wavetrain and add a collisional rate to $\Gamma$ , giving pressure broadening with the identical Lorentzian shape. Why: natural and pressure broadening are the same mathematics with different clocks — spontaneous decay versus collision rate — which is why both live in the damping wings. *For Hα the natural width $\approx 2\times10^{-4}$ Å is roughly $10^3$ times narrower than the thermal Doppler width, consistent with the energy-time uncertainty relation $\Delta E\,\Delta t \sim \hbar$ .*

Limiting cases

The character of a line depends on which regime of the curve of growth and which broadening mechanism dominates:

Case	Result	Meaning
Weak (linear) lines	$W_\lambda \propto N_\ell$	Unsaturated: equivalent width grows linearly with the column density of absorbers — the best abundance diagnostics.
Saturated (flat) lines	$W_\lambda \propto \sqrt{\ln N_\ell}$	Core is fully black; adding absorbers barely widens it. Equivalent width nearly independent of $N_\ell$ — abundances poorly constrained.
Damping (square-root) lines	$W_\lambda \propto \sqrt{N_\ell\,\Gamma}$	Strong lines with pressure-broadened Lorentzian wings; growth resumes as the wings soak up flux.
LTE source function	$S_\nu = B_\nu(T)$	When collisions dominate, the source function is Planckian and Kirchhoff's laws hold exactly.
Light vs heavy atom	$\Delta\lambda_D \propto 1/\sqrt{m}$	Hydrogen lines are broad and thermal; iron lines are narrow, exposing microturbulence and rotation.

Worked examples

1 · Thermal Doppler width of Hα (fully worked)

Estimate the Doppler width of the hydrogen Hα line ( $\lambda_0 = 6562.8$ Å) in a solar-type photosphere at $T = 6000$ K, compare it to the natural width, and check whether a microturbulence of $\xi = 1$ km/s matters.

Step 1. Thermal speed of hydrogen: $v_{\mathrm{th}} = \sqrt{2k_BT/m_H} = \sqrt{2(1.381\times10^{-23})(6000)/(1.673\times10^{-27})} = \mathbf{9.95\ km/s}$ .

Step 2. Doppler width (the $1/e$ half-width): $\Delta\lambda_D = \lambda_0\,v_{\mathrm{th}}/c = (6562.8\ \text{Å})(9.95\times10^3/2.998\times10^8) = \mathbf{0.218\ Å}$ , so the full width at half maximum is $2\sqrt{\ln 2}\,\Delta\lambda_D = \mathbf{0.36\ Å}$ .

Step 3. Compare to the natural width. With $\Gamma \approx 10^8\ \mathrm{s^{-1}}$ for the Balmer upper level, $\Delta\lambda_{\mathrm{nat}} = \lambda_0^2\,\Gamma/(2\pi c) \approx \mathbf{2\times10^{-4}\ Å}$ — about $\mathbf{10^3}$ times narrower. In a stellar photosphere the line core is Doppler-dominated; the natural (and pressure) broadening lives only in the far wings.

Step 4. Add microturbulence: $\Delta\lambda_D = (\lambda_0/c)\sqrt{v_{\mathrm{th}}^2 + \xi^2}$ with $\xi = 1$ km/s gives $\mathbf{0.219\ Å}$ — negligible, because for light hydrogen the thermal speed dominates. For an iron line ( $m = 56\,m_H$ ) the thermal speed is $\sqrt{56}\approx 7.5$ times smaller, so the same turbulence would visibly broaden it. The Hα core is thermal, 0.22 Å wide, and a thousand times broader than its natural width.

2 · Why cool stars have weak Balmer lines (Boltzmann)

Compute the fraction of neutral hydrogen atoms sitting in the $n=2$ level (the lower level of the Balmer series) at $T = 6000$ K and at $T = 10000$ K, using $E_2 - E_1 = 10.2$ eV, $g_1 = 2$ , $g_2 = 8$ .

Step 1. Boltzmann ratio at 6000 K: $k_BT = (8.617\times10^{-5})(6000) = 0.517$ eV, so $n_2/n_1 = (8/2)\,e^{-10.2/0.517} = 4\,e^{-19.7} = \mathbf{1.1\times10^{-8}}$ .

Step 2. At 10000 K: $k_BT = 0.862$ eV, so $n_2/n_1 = 4\,e^{-10.2/0.862} = 4\,e^{-11.8} = \mathbf{2.9\times10^{-5}}$ .

Step 3. The $n=2$ population jumps by a factor $\approx \mathbf{2700}$ between these temperatures. Since Balmer absorption requires an atom already in $n=2$ , a 6000 K G star has almost none available, so its Balmer lines are weak despite hydrogen's abundance. Raising T from 6000 to 10000 K multiplies the Balmer absorbers by nearly 3000.

Step 4. Push higher and Saha takes over: above $\sim 11000$ K hydrogen ionizes, emptying the neutral $n=2$ level and weakening Balmer again. The two effects together put the peak near $9500$ K — the A stars. The competition between excitation (Boltzmann) and ionization (Saha) is why Balmer lines peak in A stars, not in the hottest ones.

3 · Column density from a measured equivalent width (curve of growth)

A weak, unsaturated Fe I line at $\lambda_0 = 5000$ Å with oscillator strength $f = 0.02$ is measured to have $W_\lambda = 10$ mÅ. Estimate the column density of iron atoms in the lower level.

Step 1. Invert the linear curve of growth: $N_\ell = W_\lambda / [(\pi e^2/m_e c^2)\,\lambda_0^2\, f]$ , with the constant $\pi e^2/m_e c^2 = 8.85\times10^{-13}$ cm (CGS).

Step 2. Convert units: $W_\lambda = 10$ mÅ $= 1.0\times10^{-10}$ cm; $\lambda_0 = 5000$ Å $= 5.0\times10^{-5}$ cm, so $\lambda_0^2 = 2.5\times10^{-9}\ \mathrm{cm^2}$ .

Step 3. Evaluate: $N_\ell = (1.0\times10^{-10})/[(8.85\times10^{-13})(2.5\times10^{-9})(0.02)] = \mathbf{2.3\times10^{12}\ cm^{-2}}$ .

Step 4. Check the regime: the problem states the line is weak and unsaturated — equivalently its line-center optical depth is $\tau_0 \lesssim 1$ , which is the real test of linearity; the tiny ratio $W_\lambda/\lambda_0 = 2\times10^{-6}$ alone would not prove it, since the comparison that matters is $W_\lambda$ against the Doppler width, not $\lambda_0$ . With that assumption the linear formula is valid and the abundance is well constrained. The lower-level iron column is about 2×10¹² cm⁻²; a Saha–Boltzmann correction then converts it to a total iron abundance.

Physics of the portrait

The pieces of line-formation theory were all in place by 1925 — Bohr's levels, Boltzmann's excitation, Saha's ionization, the transfer equation — but it took one doctoral thesis to fit them together and read the composition of the stars.

Origins · 1 minCecilia Payne-Gaposchkin and how this was discovered

Cecilia Payne-Gaposchkin

1900–1979 · stellar spectroscopy and the chemical composition of stars

Cecilia Payne-Gaposchkin turned the abstract machinery of line formation into the first correct census of the universe. In her 1925 Harvard PhD thesis — often called the most brilliant ever written in astronomy — she combined Saha's newly published ionization equation with the Boltzmann excitation law and applied it to the vast Harvard collection of stellar spectra. The relative strengths of lines from different elements and ionization stages, she showed, are governed almost entirely by temperature through the Saha–Boltzmann product, not by composition: the OBAFGKM sequence is a temperature sequence. Pushing further, she found that hydrogen and helium overwhelmingly dominate stellar matter — hydrogen by roughly a million to one over the metals. The result so contradicted the prevailing belief in Earth-like stellar compositions that Henry Norris Russell persuaded her to call it "almost certainly not real"; within four years he confirmed she had been right. Her work established that a line's strength is a thermometer first and an abundance gauge only second — the lesson of the curve of growth — and it remains the foundation of all stellar abundance analysis.

History

1802

— Wollaston notes dark gaps in the solar spectrum

1814

— Fraunhofer maps and labels hundreds of solar absorption lines

1859

— Kirchhoff and Bunsen show that lines identify chemical elements

1920

— Saha derives the ionization equation linking temperature to line strength

1925

— Payne-Gaposchkin uses Saha and Boltzmann to prove stars are mostly hydrogen

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., Ch. 8–9 (Formation of Spectral Lines) · Gray, The Observation and Analysis of Stellar Photospheres, 3rd ed., Ch. 11–16 · Rutten, Radiative Transfer in Stellar Atmospheres (Utrecht lecture notes) · Rybicki & Lightman, Radiative Processes in Astrophysics, Ch. 1–2 · Mihalas, Stellar Atmospheres, 2nd ed. Key numerical benchmarks: the linear curve-of-growth coefficient $\pi e^2/m_e c^2 = 8.85\times10^{-13}$ cm (CGS); the Hα thermal Doppler width $0.22$ Å at 6000 K, roughly $10^3$ times broader than the natural width; the hydrogen $n=2$ Boltzmann fraction rising by a factor $\approx 2700$ from 6000 K to 10000 K; and the Balmer-strength maximum near 9500 K in A stars. Constants ( $h$ , $k_B$ , $m_e$ , $e$ ) are CODATA standard values.

Builds on: The Equation of Radiative Transfer · Opacity and Absorption Coefficients · Emission and the Source Function · The Boltzmann Excitation Equation · The Saha Ionization Equation · Local Thermodynamic Equilibrium. Leads to: Line Broadening Mechanisms · Kirchhoff's Laws and Thermal Emission · Specific Intensity and Radiative Flux.

Ask the tutor: Ask me to walk through a curve-of-growth abundance measurement end to end, or to show exactly how the Saha–Boltzmann product makes Balmer lines peak in A stars.

Why does the wavelength of a line depend only on the atom and not on the star?

Because the wavelength is fixed by the Bohr condition $h\nu = E_{\mathrm{upper}} - E_{\mathrm{lower}}$ , and the energy levels are intrinsic properties of the atom or ion, set by quantum mechanics. Temperature, pressure, and density change how many atoms occupy each level and how broad the line is, but not where its center sits. That is what makes spectroscopy a chemical assay: a sodium D line at 5890 Å in a flame, the Sun, and a distant galaxy is the same transition, so identifying it identifies the element. Bulk motion Doppler-shifts the whole pattern, but the relative spacing is preserved, so the barcode stays recognizable.

What is the difference between the excitation temperature and the kinetic temperature?

The kinetic temperature describes the Maxwellian distribution of particle speeds and sets the thermal Doppler width. The excitation temperature is whatever value you must plug into the Boltzmann law to reproduce the observed level populations. In local thermodynamic equilibrium (LTE) these coincide, along with the ionization and radiation temperatures, and a single $T$ governs everything. In low-density gas — chromospheres, nebulae, the interstellar medium — collisions are too rare to enforce LTE, so populations depart from Boltzmann and the excitation temperature can differ sharply from the kinetic one. Recognizing when LTE fails is a central subtlety of modern abundance work.

How can I separate temperature from pressure using a single strong line?

Fit the full Voigt profile. The Gaussian core is set by Doppler broadening, whose width scales as $\sqrt{T/m}$ , so it delivers the temperature (once the atomic mass and microturbulence are accounted for). The Lorentzian wings are set by the damping constant $\Gamma$ , which grows with collision rate and hence with gas pressure. A strong line shows both features at once — a rounded thermal core flanked by extended pressure wings — so a single Voigt fit disentangles the two. This is why the wings of strong lines such as the Balmer or Na D lines are prized surface-gravity diagnostics: gravity sets photospheric pressure, which sets the wing damping.

Why is the equivalent width preferred over the line depth for measuring abundances?

Because the equivalent width is a conserved integral, immune to the one thing that varies most between instruments: spectral resolution. A spectrograph that cannot resolve the true profile smears a deep narrow line into a shallow broad one, changing the central depth dramatically — but the area removed from the continuum, and hence $W_\lambda$ , is unchanged. Since the curve of growth relates $W_\lambda$ (not depth) to column density, it is both the natural theoretical quantity and the robust observational one, requiring only a well-defined continuum.

What actually happens on the flat part of the curve of growth?

The line core has become optically thick — its center is essentially fully black, absorbing all the continuum flux it possibly can. Adding more absorbers cannot make it darker than black, so the equivalent width grows only through the slow spread of the Doppler wings, the feeble $W_\lambda \propto \sqrt{\ln N_\ell}$ dependence. A tenfold change in column density might then change $W_\lambda$ by only a few percent, so the abundance is poorly constrained and microturbulence errors dominate. Practitioners therefore avoid saturated lines, preferring weak lines on the linear part or, for the strongest transitions, the damping part where the pressure wings restore a usable $W_\lambda \propto \sqrt{N_\ell}$ growth.

Does line formation always assume local thermodynamic equilibrium?

No, and the assumption is increasingly abandoned where precision matters. LTE says the source function is Planckian, $S_\nu = B_\nu(T)$ , with Saha and Boltzmann holding at the local temperature — excellent deep in dense photospheres where collisions dominate. But in the tenuous upper layers where many strong lines form, radiative rates compete with collisional ones, the populations depart from equilibrium, and the source function is no longer Planckian. Solving this requires the coupled equations of statistical equilibrium and radiative transfer together — the non-LTE problem, whose corrections routinely shift derived abundances by tenths of a dex for elements like oxygen, sodium, and iron.