Neutral Hydrogen 21-cm Line and Galactic Kinematics

Every hydrogen atom in the cold interstellar medium is a tiny magnet holding two spins — a proton and an electron — and once every eleven million years or so, entirely on its own, one of them flips. The atom sheds the sliver of magnetic energy that separated the parallel from the antiparallel configuration as a single photon of wavelength $21.106\,\mathrm{cm}$ , frequency $\nu_0 = 1420.4\,\mathrm{MHz}$ . Per atom this is the most forbidden event in common astrophysics; across the $\sim 10^9\,M_\odot$ of neutral atomic hydrogen (HI) that fills the Milky Way, it is the brightest spectral line in the radio sky. This article is where the microscopic quantum mechanics of a hyperfine transition becomes a survey instrument for an entire Galaxy. Two properties make the line extraordinary. At $21\,\mathrm{cm}$ interstellar dust is transparent, so the emission reaches us from HI on the far side of the disk — regions optical astronomy can never see through kiloparsecs of obscuration. And the line is narrow with a rest frequency known to nine figures, so every Doppler shift is a clean velocity. Fold those velocities through the Galaxy's differential rotation and the line delivers the rotation curve $\Theta(R)$ : the tangent-point method reads it straight off the extreme velocity in the inner disk, with no distance at all. The curves come out flat — one of the founding pieces of evidence for dark matter. The same line, redshifted, is the great hope of early-universe cosmology. Whether HI at cosmic dawn glows against the microwave background or bites a shadow into it is decided by one number — the spin temperature $T_s$ , set by collisions, radiation, and the Wouthuysen–Field coupling to the first starlight. One transition, three eras of astrophysics.

Key takeaways

The 21-cm line is the magnetic-dipole spin-flip between the hyperfine levels of the hydrogen $1s$ ground state ( $F=1 \to F=0$ ), rest frequency $\nu_0 = 1420.405751\,\mathrm{MHz}$ , wavelength $\lambda_0 = 21.106\,\mathrm{cm}$ .
Its rate is tiny, $A_{10} \approx 2.869\times10^{-15}\,\mathrm{s^{-1}}$ (an $\sim 11$ Myr lifetime), yet Galactic sightlines hold $10^{20}$ – $10^{22}$ atoms cm $^{-2}$ , so the collective line is bright and dust-transparent.
In the thin limit the velocity-integrated brightness gives the column directly, $N_{\mathrm{HI}} = 1.823\times10^{18}\int T_B(v)\,\mathrm{d}v$ , independent of $T_s$ .
The tangent-point method extracts $\Theta(R) = v_{r,\max} + \Theta_0\sin\ell$ at $R = R_0\sin\ell$ from Doppler shifts alone, with no distance measurement.
Observed HI rotation curves stay flat ( $\Theta \approx 220\,\mathrm{km\,s^{-1}}$ ) far beyond the stellar disk, implying $M(<R)\propto R$ and a massive dark halo.
The spin temperature $T_s$ sets the excitation; its contrast $(T_s-T_\gamma)/T_s$ decides emission versus absorption and drives the cosmological signal.
Key idea: the 21-cm hyperfine line is a dust-transparent tracer of neutral hydrogen whose column density maps the atomic gas and whose Doppler shift, folded through a differential-rotation model, delivers the Galactic rotation curve.

The core insight

The 21-cm hyperfine line is a dust-transparent tracer of neutral hydrogen whose column density gives the atomic gas distribution and whose Doppler shift, folded through a differential-rotation model, delivers the Galactic rotation curve.

Where this lives today

The 21-cm line is the backbone of modern Galactic and extragalactic surveys. The Green Bank Telescope, the Arecibo legacy surveys, and the all-sky HI4PI compilation map Milky Way hydrogen in exquisite velocity detail, feeding models of the disk, its warp, and the high-velocity clouds raining onto it. In external galaxies, THINGS and the WALLABY survey on ASKAP trace HI rotation curves out to radii where no starlight remains, and those curves stay flat — the most direct dynamical weighing of dark-matter halos; a galaxy's integrated profile width also powers the Tully–Fisher distance ladder. At the frontier, LOFAR, MWA, HERA, and the Square Kilometre Array chase the redshifted signal from reionization and cosmic dawn, while EDGES and SARAS pursue the sky-averaged absorption trough — a claimed detection whose anomalous depth, if real, would demand exotic gas cooling or an unexpected radio background. One transition, predicted in a blacked-out wartime city, now maps a galaxy, weighs its dark matter, and listens for the first stars.

The physical picture

Picture the proton and electron of ground-state hydrogen as two bar magnets. Parallel spins cost a little energy; when the electron flips to anti-align, the atom drops to its true ground state and emits the difference as one photon. That energy is only $5.87\,\mathrm{micro\text{-}eV}$ — millionths of the scale of ordinary spectral lines — because it comes from magnetic coupling between spins, not the electric binding of electron to proton. Hence a radio photon, not an optical one, and an equivalent temperature $T_\star = h\nu_0/k_B$ of a mere $0.068\,\mathrm{K}$ : at every interstellar temperature the gap is trivially crossed and roughly three-quarters of the atoms sit in the upper level, ready to flip. The first figure is the ledger of that transition.

Figure

Hyperfine splitting of the hydrogen 1s ground state: the parallel-spin triplet F = 1 with statistical weight g = 3 lies above the antiparallel singlet F = 0 with g = 1 by E₁₀ = hν₀ = 5.87 μeV, so the spin-flip emits a photon at ν₀ = 1420.4 MHz, λ₀ = 21.106 cm, with equivalent temperature T* = hν₀/k_B = 0.068 K

Now let the emitting atoms rotate with the Galaxy. Each parcel broadcasts at a Doppler-shifted frequency, so a sightline through the differentially rotating disk returns not a spike but a spread of frequencies — a profile. The extreme velocity in that profile is special: it comes from the tangent point, where the orbit runs parallel to our line of sight and the full orbital speed projects onto it. Reading that extreme velocity across longitudes traces the circular-speed curve of the inner Galaxy — and it refuses to fall the Keplerian way. The second figure builds the geometry and its flat-curve payoff side by side.

Figure

The tangent-point construction and the flat rotation curve it yields: a sightline from the Sun at longitude ℓ reaches its smallest radius R_min = R₀ sin ℓ at the tangent point T, where the projected velocity peaks at v_r,max = Θ(R_min) − Θ₀ sin ℓ; mapping this across longitudes gives Θ(R) ≈ 220 km s⁻¹, flat far beyond the Sun at R₀ = 8.2 kpc

Finally, whether the line glows or shadows is a contest of temperatures. If the gas's spin temperature exceeds the brightness temperature behind it, the line emits; if colder, it carves an absorption trough. In the Galaxy the background is faint and HI emits; in the early universe the background is the cosmic microwave background (CMB), and the same inequality — now $(T_s-T_\gamma)/T_s$ — decides whether cosmic dawn is a whisper of emission or, as models predict, a deep absorption shadow.

The core equations

The transition energy is fixed by the hyperfine interaction between the proton and electron magnetic moments — a fantastically small energy from which everything downstream follows: the radio wavelength, the near-equal level populations, the reach to the coldest gas.

E_{10} = h\nu_0, \qquad \nu_0 = 1420.405751\,\mathrm{MHz}, \qquad \lambda_0 = \frac{c}{\nu_0} = 21.106\,\mathrm{cm} \qquad (1)

where: $E_{10}$ — gap between the $F=1$ and $F=0$ levels, $\approx 5.87\times10^{-6}\,\mathrm{eV}$ ; $h$ — Planck constant; $\nu_0$ — rest frequency, known to nine figures; $\lambda_0$ — rest wavelength. Reading it: the gap sits six orders of magnitude below optical line energies — hence a radio photon, and levels that stay nearly equally populated at all interstellar temperatures. The rate at which an isolated excited atom flips is the Einstein $A$ coefficient for a magnetic-dipole transition — a minuscule value that makes the line simultaneously "forbidden" per atom and ubiquitous in bulk.

A_{10} = \frac{64\pi^4}{3}\,\frac{\nu_0^3}{h c^3}\,\mu_B^2 \approx 2.869\times10^{-15}\,\mathrm{s^{-1}} \qquad (2)

where: $A_{10}$ — spontaneous-emission coefficient for the spin-flip; $\mu_B$ — Bohr magneton, the scale of the electron magnetic moment; $c$ — speed of light. Reading it: the mean radiative lifetime is $1/A_{10}\approx11\,\mathrm{Myr}$ — longer than the interval between atomic collisions in most of the medium, so collisions, not spontaneous decay, usually control the excitation. The relative populations of the two levels define the spin temperature — not the gas's thermodynamic temperature but the temperature that would produce the observed level ratio in a Boltzmann distribution.

\frac{n_1}{n_0} = \frac{g_1}{g_0}\,e^{-E_{10}/k_B T_s} = 3\,e^{-T_\star/T_s}, \qquad T_\star \equiv \frac{h\nu_0}{k_B} = 0.0682\,\mathrm{K} \qquad (3)

where: $n_1, n_0$ — densities in the upper ( $F=1$ ) and lower ( $F=0$ ) levels; $g_1=3,\ g_0=1$ — statistical weights $2F+1$ ; $T_s$ — spin temperature; $T_\star$ — equivalent temperature of the transition. Reading it: since every relevant $T_s\gg T_\star$ , the exponential is $\approx1$ and $n_1/n_0\approx3$ — three-quarters of the atoms sit in the upper state regardless of conditions. Emission from an optically thin cloud integrates directly into a column density with a universal coefficient — the workhorse formula of HI astronomy: a brightness-temperature profile in, atoms per square centimetre out.

N_{\mathrm{HI}} = 1.823\times10^{18}\,\mathrm{cm^{-2}} \int T_B(v)\,\mathrm{d}v \qquad (\text{thin, } v \text{ in } \mathrm{km\,s^{-1}}) \qquad (4)

where: $N_{\mathrm{HI}}$ — HI column density; $T_B(v)$ — brightness temperature versus velocity (K); $\mathrm{d}v$ — velocity element. Reading it: $T_s$ has cancelled — in the thin limit the line measures how many atoms lie along the sightline, not how they are excited. The kinematic content is the projection of differential rotation onto the line of sight, which in the Galactic plane reduces to a single clean expression in longitude.

v_r(\ell) = R_0\left[\frac{\Theta(R)}{R} - \frac{\Theta_0}{R_0}\right]\sin\ell \qquad (b=0) \qquad (5)

where: $v_r$ — velocity relative to the Local Standard of Rest; $\ell$ — Galactic longitude; $R_0\approx8.2\,\mathrm{kpc}$ — Sun's radius; $\Theta_0\approx220\,\mathrm{km\,s^{-1}}$ — circular speed at the Sun; $\Theta(R)$ — circular speed at $R$ . Reading it: only the difference of angular speeds is observable, so $v_r\to0$ toward the centre and anticentre and kinematic distances fail there. In the first quadrant the profile's extreme velocity picks out the tangent point, giving the rotation curve with no distance measurement.

v_{r,\max}(\ell) = \Theta(R_{\min}) - \Theta_0\sin\ell, \qquad R_{\min} = R_0\sin\ell \qquad (6)

where: $v_{r,\max}$ — maximum radial velocity on an inner sightline; $R_{\min}$ — tangent-point radius. Reading it: each longitude samples one radius $R_0\sin\ell$ , so scanning $\ell$ from $0$ to $90^\circ$ sweeps the curve from centre to solar circle. The line-centre optical depth carries the spin temperature back in — absorption weighs the difference of populations, so unlike emission it is sensitive to excitation.

\tau_\nu = \frac{3 c^2 A_{10}}{32\pi\,\nu_0^2}\,\frac{h}{k_B T_s}\,\frac{N_{\mathrm{HI}}}{\Delta v} \qquad (7)

where: $\tau_\nu$ — line-centre optical depth; $\Delta v$ — velocity width; $T_s$ — spin temperature. Reading it: cold gas is far more opaque than warm gas of the same column, so absorption against background sources preferentially reveals the cold neutral medium. Redshifted to cosmic dawn, the line's contrast against the CMB is the target of a generation of experiments — its sign and amplitude set by the same spin-temperature logic, now competing against the microwave background.

\delta T_b \approx 27\,x_{\mathrm{HI}}(1+\delta)\left(\frac{T_s - T_\gamma}{T_s}\right)\sqrt{\frac{1+z}{10}}\ \mathrm{mK} \qquad (8)

where: $\delta T_b$ — differential brightness against the CMB; $x_{\mathrm{HI}}$ — neutral fraction; $\delta$ — gas overdensity; $T_\gamma$ — CMB temperature at $z$ ; $z$ — redshift. Reading it: $T_s < T_\gamma$ gives absorption, $T_s > T_\gamma$ emission — the whole cosmological signal is a story about which way that inequality runs.

Derivation ALine-of-sight velocity of differentially rotating gas

The goal is to turn the picture of a rotating disk into the observable of equation (5): the radial velocity a 21-cm profile records. We assume every parcel moves on a circular orbit at the local circular speed $\Theta(R)$ , work in the Galactic plane ( $b=0$ ), and measure velocities relative to the Local Standard of Rest so the Sun's motion is already subtracted. The task is to project both the parcel's and the Sun's velocity onto the line of sight and difference them — a spectrograph sees only relative approach or recession.

Step 1. Place the Galactic centre at the origin, the Sun at radius $R_0$ , and a parcel at radius $R$ along a sightline of longitude $\ell$ . The parcel moves at speed $\Theta(R)$ tangent to its orbit; the Sun at $\Theta_0$ tangent to the solar circle. Why: circular orbits are the minimal axisymmetric model, and they make a one-dimensional velocity encode a two-dimensional position.

Step 2. Project each velocity onto the line of sight. With $\alpha$ the angle between the parcel's orbital velocity and the sightline, the observed radial velocity is the difference of projections,

v_r = \Theta(R)\cos\alpha - \Theta_0\sin\ell.

Why: only the along-sightline component shifts the frequency; the transverse part is invisible, so the Doppler signal is purely this difference.

Step 3. Eliminate $\alpha$ with geometry. In the Sun–centre–parcel triangle, the law of sines (equivalently, dropping a perpendicular from the centre to the sightline) gives

R\cos\alpha = R_0\sin\ell.

Why: this converts the unmeasurable orbit angle into the measurable longitude, and it is the same relation that later defines the tangent point.

Step 4. Substitute $\cos\alpha = R_0\sin\ell / R$ and factor:

v_r = \Theta(R)\frac{R_0\sin\ell}{R} - \Theta_0\sin\ell = R_0\left[\frac{\Theta(R)}{R} - \frac{\Theta_0}{R_0}\right]\sin\ell.

Why: the result depends only on how angular speed $\Theta/R$ runs with radius — differential rotation, not rotation itself, is what we observe. *Verified: reduces to $v_r=0$ at $\ell = 0,\,180^\circ$ and at $R=R_0$ ; matches Binney & Tremaine §3.2 and Carroll & Ostlie §24.3, assuming circular orbits and axisymmetry.*

Go deeper · Derivation B · 1 minThe tangent-point rotation curve of the inner Galaxy

Derivation BThe tangent-point rotation curve of the inner Galaxy

The goal is to extract the circular-speed curve $\Theta(R)$ for $R<R_0$ using nothing but the extreme velocity in each 21-cm profile — no photometric distance, no standard candle, no parallax. The idea rests on one observation: along an inner sightline ( $0<\ell<90^\circ$ ), gas at different distances lies at different radii, and because $\Theta/R$ falls outward the radial velocity peaks at one special radius. Pin down that radius and the extreme velocity hands us the curve point by point.

Step 1. Start from projection formula (5), $v_r = R_0[\Theta(R)/R - \Theta_0/R_0]\sin\ell$ , and hold $\ell$ fixed. Looking outward, $R$ first decreases to a minimum, then increases. Why: the sightline is a chord of the solar circle, dipping inward to a closest approach before receding — that closest approach is where the physics happens.

Step 2. Identify the tangent point, the minimum radius reached:

R_{\min} = R_0\sin\ell,

the foot of the perpendicular from the centre to the sightline, where the line of sight is exactly tangent to the orbit. Why: there the orbital velocity is entirely along the line of sight, so its full magnitude is Doppler-visible — nowhere else projects as much.

Step 3. Because $\Theta(R)/R$ is largest at the smallest $R$ , the velocity peaks at the tangent point. Evaluate (5) there, using $R_0\sin\ell / R_{\min} = 1$ :

v_{r,\max} = R_0\left[\frac{\Theta(R_{\min})}{R_{\min}} - \frac{\Theta_0}{R_0}\right]\sin\ell = \Theta(R_{\min}) - \Theta_0\sin\ell.

Why: the bracket collapses because $R_{\min}$ is tuned so $R_0\sin\ell$ and $R_{\min}$ cancel, leaving the bare circular speed.

Step 4. Solve for the circular speed:

\Theta(R_{\min}) = v_{r,\max}(\ell) + \Theta_0\sin\ell, \qquad R_{\min} = R_0\sin\ell.

Why: every longitude yields one clean $(R,\Theta)$ pair, so a longitude scan from $0$ to $90^\circ$ maps the entire inner rotation curve — the classic HI result that first revealed the curve does not fall. *Verified: consistent with Binney & Tremaine §3.2 and Carroll & Ostlie §24.3; assumes circular orbits, axisymmetry, and declining angular speed, and fails in the barred region within a few kpc of the centre.*

Go deeper · Derivation C · 1 minColumn density from the optically thin emission profile

Derivation CColumn density from the optically thin emission profile

The goal is to prove equation (4) — that the velocity-integrated brightness temperature of an HI line gives the column density with the universal coefficient $1.823\times10^{18}\,\mathrm{cm^{-2}\,(K\,km\,s^{-1})^{-1}}$ , independent of how the gas is excited. This is the most-used formula in HI astronomy, and its power lies in the cancellation of the spin temperature. We treat a thin isothermal slab with no background and follow the radiative transfer to an integral over the line.

Step 1. Solve the transfer equation for a uniform slab with zero background. The emergent excess brightness temperature is

T_B = T_s\,(1 - e^{-\tau_\nu}),

the source function $T_s$ attenuated by the slab's own opacity. Why: with nothing behind it, all the brightness is self-emission, so $T_s$ sets the ceiling the line can reach.

Step 2. Take the thin limit $\tau_\nu \ll 1$ , expanding $1-e^{-\tau_\nu}\approx\tau_\nu$ :

T_B \approx T_s\,\tau_\nu.

Why: since $\tau_\nu$ carries a factor $1/T_s$ from (7), the spin temperature is about to disappear entirely.

Step 3. Insert the optical depth (7), written per unit velocity so $N_{\mathrm{HI}}/\Delta v \to n_{\mathrm{HI}}\,\mathrm{d}s/\mathrm{d}v$ . The $T_s$ factors cancel:

T_s\,\tau_\nu = \frac{3 c^2 A_{10}}{32\pi\,\nu_0^2}\,\frac{h}{k_B}\,\frac{n_{\mathrm{HI}}\,\mathrm{d}s}{\mathrm{d}v}.

Why: emission counts atoms while opacity divides by excitation — the two dependences are reciprocal, leaving a pure atom count.

Step 4. Integrate over velocity; $\mathrm{d}s$ accumulates into the full column $N_{\mathrm{HI}}=\int n_{\mathrm{HI}}\,\mathrm{d}s$ :

\int T_B\,\mathrm{d}v = \frac{3 c^2 A_{10}\,h}{32\pi\,\nu_0^2 k_B}\,N_{\mathrm{HI}}.

Substituting the constants (with $v$ in $\mathrm{km\,s^{-1}}$ ) gives $1/(1.823\times10^{18})$ , hence $N_{\mathrm{HI}} = 1.823\times10^{18}\int T_B\,\mathrm{d}v$ . Why: the prefactor is a pure combination of atomic constants — the same number in every HI survey pipeline. *Verified: coefficient matches Draine ch. 8 and standard radio-astronomy references; valid only for $\tau_\nu\lesssim 0.5$ , above which self-absorption underestimates $N_{\mathrm{HI}}$ .*

Limiting cases

The line's behaviour splits cleanly according to opacity and to what controls the spin temperature — each regime a different instrument reading of the same transition.

Case	Result	Meaning
Optically thin emission	$T_B \approx T_s\,\tau_\nu \propto N_{\mathrm{HI}}$	For $\tau_\nu\ll1$ brightness is proportional to column and independent of $T_s$ ; the profile directly counts atoms.
Optically thick / self-absorbed	$T_B \to T_s\,(1-e^{-\tau_\nu}) \to T_s$	For $\tau_\nu\gg1$ brightness saturates at $T_s$ ; the line stops measuring total column and HI mass is underestimated.
Collision-dominated (dense gas)	$T_s \to T_K$	Frequent collisions thermalize the levels, driving $T_s$ to the kinetic temperature.
Radiation-dominated (rarefied gas)	$T_s \to T_\gamma$	With few collisions the levels pin to the ambient radiation (CMB); the contrast vanishes and the line fades.
Cosmic dawn with Lyman-α coupling	$T_s \to T_K < T_\gamma$	Wouthuysen–Field coupling ties $T_s$ to the cold gas via scattered Lyα photons, driving $\delta T_b<0$ — the absorption trough.

Common trapThe 21-cm line is just Lyman-α from cold hydrogen

No. The 21-cm line is a magnetic-dipole spin-flip within the $1s$ ground state — the electron never changes orbital; only the relative orientation of the proton and electron spins does. Lyman- $\alpha$ ( $121.6\,\mathrm{nm}$ ) is an electric-dipole transition between the $n=2$ and $n=1$ orbitals, a million times more energetic. The two even play opposite roles: Lyα photons from the first stars are the pump (via Wouthuysen–Field coupling) that sets the spin temperature governing the 21-cm line. A subtler trap: the "temperature" in the level ratio is the spin temperature $T_s$ , which in general equals neither the kinetic nor the radiation temperature — $T_s=T_K$ holds only where collisions dominate, $T_s=T_\gamma$ only where they are negligible.

Worked examples

1 · Reading the rotation curve off a tangent-point velocity (fully worked)

Along the sightline at Galactic longitude $\ell = 30^\circ$ the 21-cm profile cuts off at a maximum radial velocity $v_{r,\max} = 110\,\mathrm{km\,s^{-1}}$ . With $R_0 = 8.2\,\mathrm{kpc}$ and $\Theta_0 = 220\,\mathrm{km\,s^{-1}}$ , find the tangent-point radius and the circular speed there. Then predict the tangent-point velocity at $\ell = 60^\circ$ if the curve is flat.

Step 1. Geometry fixes the tangent-point radius: $R_{\min} = R_0\sin\ell = 8.2\times0.5 = \mathbf{4.1\,kpc}$ — halfway out to the Sun.

Step 2. Equation (6) solved for the circular speed: $\Theta(R_{\min}) = v_{r,\max} + \Theta_0\sin\ell = 110 + 220\times0.5 = \mathbf{220\,km\,s^{-1}}$ . The speed at $4.1\,\mathrm{kpc}$ equals the speed at the Sun — flat between them.

Step 3. Predict $\ell=60^\circ$ : $R_{\min} = 8.2\times0.866 = 7.1\,\mathrm{kpc}$ , and a flat curve gives $v_{r,\max} = 220 - 220\times0.866 = \mathbf{29.5\,km\,s^{-1}}$ . The tangent-point velocity shrinks at large $\ell$ even though the gas moves just as fast, because there the tangent point approaches the Sun and the differential velocity vanishes. A flat curve predicts $110\,\mathrm{km\,s^{-1}}$ at $\ell=30^\circ$ but only $\sim 30\,\mathrm{km\,s^{-1}}$ at $\ell=60^\circ$ .

2 · Column density and gas mass from a line profile (fully worked)

An HI cloud shows a roughly Gaussian 21-cm line with peak brightness temperature $T_{B,\mathrm{peak}} = 80\,\mathrm{K}$ and full width at half maximum $\Delta v_{\mathrm{FWHM}} = 25\,\mathrm{km\,s^{-1}}$ , and it is optically thin. Find the HI column density and the corresponding gas surface density.

Step 1. Integrate the Gaussian profile: area $\int T_B\,\mathrm{d}v = T_{B,\mathrm{peak}}\,\Delta v\,\sqrt{\pi/(4\ln 2)} = 80\times25\times1.0645 = \mathbf{2129\,K\,km\,s^{-1}}$ .

Step 2. Apply equation (4): $N_{\mathrm{HI}} = 1.823\times10^{18}\times2129 = \mathbf{3.9\times10^{21}\,cm^{-2}}$ — a typical cloud column, and with $\tau_\nu$ below $\sim0.5$ the thin-limit formula is safe.

Step 3. As a surface mass density: $\Sigma = N_{\mathrm{HI}}\,m_{\mathrm{H}} = 3.9\times10^{21}\times1.67\times10^{-24} = 6.5\times10^{-3}\,\mathrm{g\,cm^{-2}} \approx \mathbf{31\,M_\odot\,pc^{-2}}$ — right in the range of a spiral disk's atomic gas layer. A single 80 K, 25 km s⁻¹ line encodes $\sim4\times10^{21}$ atoms cm⁻² and $\sim30\,M_\odot\,\mathrm{pc^{-2}}$ of neutral gas.

3 · The sign of the cosmic-dawn signal (quick estimate)

At $z=17$ the CMB temperature is $T_\gamma = 2.725(1+z) = 49\,\mathrm{K}$ ; adiabatically cooled gas sits near $T_K \approx 7\,\mathrm{K}$ . If Lyman-α coupling drives $T_s\to T_K$ in fully neutral, mean-density gas, then $(T_s-T_\gamma)/T_s = (7-49)/7 = -6$ and equation (8) with $x_{\mathrm{HI}}=1$ , $\delta=0$ , $\sqrt{1.8}=1.34$ gives $\delta T_b = 27\times(-6)\times1.34 \approx \mathbf{-0.2\,K}$ — a deep absorption trough, the feature 21-cm cosmology hunts for and (at a controversially deeper $\sim -0.5\,\mathrm{K}$ ) what EDGES claimed in 2018.

Physics of the portrait

The 21-cm line was theory before observation — a wartime prediction by a graduate student who reasoned that if hydrogen had a hyperfine transition, the Galaxy must be shining in it.

Origins · 1 minHendrik van de Hulst and how this was discovered

Hendrik van de Hulst

1918–2000 · radio astronomy and the 21-cm prediction

In occupied Leiden in 1944, with observatories shuttered, Jan Oort asked his student Hendrik van de Hulst a theorist's question: was there any spectral line by which radio waves could map the hydrogen that optical astronomy could not penetrate? Van de Hulst worked out that the hydrogen ground state must be hyperfine-split by the proton and electron spins, and that the forbidden magnetic-dipole transition would radiate near $21\,\mathrm{cm}$ . He doubted it could ever be detected — a flip every eleven million years per atom — but he had the line. Seven years later three groups found it within weeks: Ewen and Purcell at Harvard in March 1951 with a horn antenna bolted out a lab window, then Muller and Oort in the Netherlands, then Christiansen and Hindman in Australia. Van de Hulst went on to help lead the Dutch effort that produced the first 21-cm maps of spiral structure and the first HI rotation curve of the Milky Way — a spin-flip he doubted could be seen became the instrument that weighed the Galaxy and exposed its dark halo.

History

1944

— van de Hulst predicts the hydrogen hyperfine line near 21 cm

1951

— Ewen and Purcell detect it at Harvard; Muller, Oort, and the Australians confirm within weeks

1954

— the Leiden and Sydney groups publish the first 21-cm map of Galactic spiral structure

1978

— Bosma's HI rotation curves stay flat far beyond the stellar disk, hardening the dark-matter case

2018

— EDGES reports a claimed 21-cm absorption trough from cosmic dawn near z ≈ 17

Going further

Texts. Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §§24.3–24.4 (the interstellar medium, the 21-cm line, Galactic rotation) · Binney & Tremaine, Galactic Dynamics, 2nd ed., §§1.1, 3.2 (differential rotation, the tangent-point method) · Mo, van den Bosch & White, Galaxy Formation and Evolution, Ch. 11 (gas in galaxies, HI observables) · Draine, Physics of the Interstellar and Intergalactic Medium, Ch. 8 (HI physics, $A_{10}$ , spin temperature) · Dodelson, Modern Cosmology, 2nd ed. (spin temperature, Wouthuysen–Field coupling, the cosmological signal).

Provenance. A verifier should confirm: $\nu_0 = 1420.405751\,\mathrm{MHz}$ , $\lambda_0 = 21.106\,\mathrm{cm}$ ; $A_{10}\approx2.869\times10^{-15}\,\mathrm{s^{-1}}$ , lifetime $\sim1.1\times10^7\,\mathrm{yr}$ (Draine ch. 8); $T_\star = h\nu_0/k_B = 0.0682\,\mathrm{K}$ , $E_{10}\approx5.87\times10^{-6}\,\mathrm{eV}$ , $g_1/g_0 = 3$ ; the column coefficient $1.823\times10^{18}\,\mathrm{cm^{-2}\,(K\,km\,s^{-1})^{-1}}$ ; the tangent-point relation at $R_{\min}=R_0\sin\ell$ (Binney & Tremaine §§1.1, 3.2; Carroll & Ostlie §24.3), $R_0\approx8.2\,\mathrm{kpc}$ , $\Theta_0\approx220\,\mathrm{km\,s^{-1}}$ ; and $\delta T_b\approx27\,\mathrm{mK}$ (Dodelson; Furlanetto, Oh & Briggs 2006). Worked numbers ( $\Theta(4.1\,\mathrm{kpc})=220\,\mathrm{km\,s^{-1}}$ ; $N_{\mathrm{HI}}=3.9\times10^{21}\,\mathrm{cm^{-2}}$ , $\Sigma\approx31\,M_\odot\,\mathrm{pc^{-2}}$ ; $\delta T_b\approx-0.2\,\mathrm{K}$ at $z=17$ ) are reproduced in the verification script.

Prerequisites. Structure of the Milky Way Galaxy · Galactic Coordinate System and the Solar Neighborhood · Interstellar Medium of Galaxies. Leads to. Galactic Rotation Curves · Evidence for Dark Matter in Galaxies · Spiral Galaxies and Density Wave Theory · Galactic Dark Matter Halos and Density Profiles.

Ask the tutor: Ask me to walk through a tangent-point velocity into a rotation-curve point, or to show how the spin temperature decides whether cosmic dawn glows or shadows.

If each atom emits only once in 11 million years, how is the line bright?

Because interstellar space is an enormous integrator. The per-atom rate $A_{10}\approx2.9\times10^{-15}\,\mathrm{s^{-1}}$ is tiny, but a single sightline through the disk threads $10^{20}$ to $10^{22}$ atoms per square centimetre, and the Galaxy holds $\sim10^9\,M_\odot$ of HI. A vanishing rate times that many emitters is a strong collective signal. The rarity is even a feature: weakly coupled to the radiation field, the line rarely saturates, so over most of the sky it stays optically thin and its brightness faithfully reports the column density.

What is the difference between spin, kinetic, and radiation temperature?

They answer three questions. The kinetic temperature $T_K$ is the real thermodynamic temperature of the atoms' thermal motions. The radiation temperature $T_\gamma$ describes the ambient photon field at $21\,\mathrm{cm}$ (in the early universe, the CMB). The spin temperature $T_s$ is a bookkeeping temperature defined purely by the ratio of atoms in the upper versus lower level. $T_s$ is pulled toward $T_K$ by collisions, toward $T_\gamma$ by 21-cm photons, and toward $T_K$ by Lyman- $\alpha$ scattering (Wouthuysen–Field). Which pull wins depends on density and radiation, and the resulting $T_s$ governs whether the line emits or absorbs.

Why does a flat rotation curve imply dark matter?

For circular orbits, gravity versus centripetal acceleration gives $\Theta^2(R) = GM(<R)/R$ , so $M(<R) = \Theta^2 R/G$ . If $\Theta$ stays constant to large radius — as the 21-cm curves show, well beyond where starlight fades — then $M(<R)$ grows linearly with $R$ . Yet the luminous matter runs out, so its enclosed mass should level off and the curve fall Keplerian, $\Theta\propto R^{-1/2}$ . The persistent flatness means mass keeps accumulating where there is no light: an extended, roughly $\rho\propto R^{-2}$ dark halo. HI curves, reaching farther out than any stellar tracer, are the cleanest demonstration.

Why does the tangent-point method only work in the inner Galaxy?

Because only sightlines passing inside the solar circle have a tangent point. For $0<\ell<90^\circ$ the line of sight dips to a closest approach $R_{\min}=R_0\sin\ell<R_0$ , where the orbit runs tangent and the velocity peaks — giving $\Theta(R_{\min})$ directly. Toward the outer Galaxy ( $90^\circ<\ell<270^\circ$ ) there is no inner turning point; the gas is always farther out, its angular speed lower, and no extreme velocity to exploit. Mapping the outer curve therefore needs independent distances to the HI (associated stars or masers), which is why it was pinned down decades later.

What is the Wouthuysen–Field effect and why does it matter for cosmology?

It lets the first starlight control the 21-cm signal. A Lyman- $\alpha$ photon absorbed to $n=2$ and re-emitted to $n=1$ can flip the hyperfine spin state. When the early universe fills with Lyα photons from the first stars, this scattering couples the spin temperature to the cold gas, driving $T_s$ below $T_\gamma$ . That turns a silent sky into an absorption trough: without the coupling, $T_s$ stays pinned at $T_\gamma$ and $\delta T_b=0$ . The timing and depth of the trough thus encode when the first sources switched on — the reason EDGES, SARAS, and HERA chase it.

Can I trust a kinematic distance from a 21-cm velocity?

With care. A rotation model converts velocity to distance well over much of the disk, but has two failure modes. Toward the centre and anticentre ( $\ell\approx0^\circ,180^\circ$ ) the radial velocity vanishes at all distances, so the method is degenerate. And in the inner Galaxy a given velocity occurs at two points along the sightline — the near/far ambiguity, resolved only with extra data such as HI self-absorption. Streaming motions near arms and the bar break the circular-orbit assumption, so kinematic distances carry real systematics.