The 21 cm Line of Neutral Hydrogen

The 21 cm line is the softest whisper in the radio sky and one of the most quantitative tools in astrophysics. It comes from the ground state of the most abundant atom in the Universe: neutral atomic hydrogen, whose electron and proton each carry a spin magnetic moment. Those two tiny magnets can sit parallel or antiparallel, splitting the $1s\,^2S_{1/2}$ level into a triplet ( $F=1$ ) a hair above a singlet ( $F=0$ ). The gap is absurdly small — $\Delta E=5.87\times10^{-6}\,\mathrm{eV}$ , only $0.068\,\mathrm{K}$ — and the spontaneous spin-flip is one of the slowest transitions known, so slow that an atom radiates on average once every eleven million years. Two accidents of scale turn this feeble transition into a workhorse. First, the gap sits so far below any interstellar temperature that the populations freeze at their statistical-weight ratio of three-to-one, making the line brightness a nearly temperature-independent gauge of how many atoms lie along a sightline. Second, the Galaxy holds billions of solar masses of neutral hydrogen, so the sum of all those rare flips is a bright, ubiquitous line at $1420.4\,\mathrm{MHz}$ that dust cannot hide. Predicted by van de Hulst in 1944 and detected by Ewen and Purcell in 1951, it opened the neutral Universe: Galactic rotation curves, the cold and warm phases of the ISM, the dark-matter halos of spiral galaxies, and — redshifted to metre wavelengths — the cosmic dark ages and the first stars.

Key takeaways

The 21 cm line is the magnetic-dipole hyperfine transition $F=1\to F=0$ in the ground state of neutral atomic hydrogen, at $\nu_{10}=1420.405752\,\mathrm{MHz}$ ( $\lambda_{10}=21.106\,\mathrm{cm}$ ).
It is extraordinarily forbidden — $A_{10}=2.869\times10^{-15}\,\mathrm{s^{-1}}$ , lifetime $\approx1.1\times10^{7}\,\mathrm{yr}$ — yet the cosmic abundance of H I makes it easy to detect.
Excitation is set by the spin temperature through $n_1/n_0=(g_1/g_0)\,e^{-h\nu_{10}/k_B T_s}\approx3$ ; since $\Delta E/k_B=0.068\,\mathrm{K}$ lies far below any ISM temperature, the populations freeze at the $3{:}1$ ratio.
In the usually thin diffuse ISM the integrated brightness reads out the column directly: $N_{\mathrm{HI}}=1.823\times10^{18}\int T_B\,\mathrm{d}v\ \mathrm{cm^{-2}}$ ( $v$ in $\mathrm{km\,s^{-1}}$ ).
Because the line is narrow, Doppler shifts map onto line-of-sight velocity, making 21 cm the premier probe of Galactic rotation, spiral structure, and the two-phase neutral medium.
Redshifted against the CMB, with $\delta T_b\propto(1-T_{\mathrm{CMB}}/T_s)$ , it is the leading tomographic probe of the dark ages, cosmic dawn, and reionization.
Key idea: the 21 cm line is a magnetic-dipole hyperfine transition whose feeble emissivity is compensated by the enormous mass of Galactic H I; its narrowness preserves kinematics, its usual optical thinness makes brightness temperature a linear column-density gauge, and its excitation (the spin temperature) links Galactic ISM physics to high-redshift cosmology.

Where this lives today

The 21 cm line is simultaneously mature Galactic infrastructure and the cosmological frontier. All-sky emission surveys — LAB, GALFA-HI, and the arcminute-resolution HI4PI — give the standard hydrogen column map used to correct extragalactic observations for foreground gas and calibrate dust. Interferometers such as THINGS and WHISP trace H I rotation curves far beyond the optical disks of nearby galaxies, where the flat $\Theta(R)$ remains the cleanest galaxy-scale evidence for dark-matter halos and the linewidth feeds Tully–Fisher. At the frontier, redshifted 21 cm is the leading tomographic probe of the dark ages, cosmic dawn, and reionization: global-signal experiments (EDGES, SARAS) chase the all-sky trough while interferometers (HERA, LOFAR, the MWA, ultimately the SKA) map the fluctuating brightness field. H I intensity mapping (CHIME) uses the line at lower redshift to measure baryon acoustic oscillations.

The physical picture

Picture the proton as a tiny bar magnet and the electron's spin as another. When their moments align parallel the atom has slightly more energy than when antiparallel — it "wants" to flip. The flip is agonizingly slow because it cannot proceed as an electric-dipole transition (which powers ordinary bright lines); it must use the far weaker magnetic-dipole coupling, hence the eleven-million-year lifetime. The first figure lays the two spin configurations against the energy ladder and hangs every governing number on it.

Figure

The hyperfine splitting of the hydrogen ground state: parallel spins give the F = 1 triplet, antiparallel spins the F = 0 singlet, separated by ΔE = 5.87×10⁻⁶ eV. The spin-flip emits a 21 cm photon at ν₁₀ = 1420.406 MHz with rate A₁₀ = 2.87×10⁻¹⁵ s⁻¹, a mean lifetime near 11 million years

The line's power rests on the same two accidents that make it faint. Because $\Delta E/k_B=0.068\,\mathrm{K}$ lies far below any interstellar temperature, the Boltzmann factor is essentially unity, the populations lock to $3{:}1$ , and the emissivity becomes a nearly temperature-blind counter of atoms. Because the transition is intrinsically narrow, the observed profile is a velocity histogram — a warm envelope in a broad hump, a cold cloud in a narrow spike — the two phases the second figure records, whose area through one constant is the hydrogen column.

Figure

A synthetic Galactic 21 cm emission profile: a narrow cold-neutral-medium spike (≈ 1.8 km/s at 70 K) on a broad warm-neutral-medium hump (≈ 19 km/s at 8000 K). The area fixes the column through N_HI = 1.823×10¹⁸ ∫ T_B dv; here ∫ T_B dv ≈ 819 K km/s gives N_HI ≈ 1.5×10²¹ cm⁻²

Whether the line appears in emission or absorption is a thermodynamic question: it probes the contrast between the gas spin temperature and the radiation behind it. Against the faint Galactic sky, warm gas ( $T_s\sim8000\,\mathrm{K}$ ) glows in emission while cold gas ( $T_s\sim70\,\mathrm{K}$ ) across a bright continuum source appears in absorption. Run the same physics against the CMB and you get 21 cm cosmology: intergalactic hydrogen with $T_s<T_{\mathrm{CMB}}$ absorbs a dip at cosmic dawn, then emits once the first sources heat it past $T_{\mathrm{CMB}}$ .

The core equations

The first three relations record the atomic data. The transition frequency is a laboratory constant known to a dozen digits.

h\nu_{10}=\Delta E_{\mathrm{HF}},\qquad \nu_{10}=1420.405751768\,\mathrm{MHz},\qquad \lambda_{10}=\frac{c}{\nu_{10}}=21.106\,\mathrm{cm} \qquad (1)

where: $\nu_{10}$ — rest frequency of the $F=1\to0$ transition; $\Delta E_{\mathrm{HF}}$ — hyperfine splitting, $5.874\times10^{-6}\,\mathrm{eV}$ ; $\lambda_{10}$ — rest wavelength. Expressed as a temperature, that gap is the most consequential number here.

\frac{\Delta E_{\mathrm{HF}}}{k_B}=\frac{h\nu_{10}}{k_B}=0.0682\,\mathrm{K} \qquad (2)

where: $k_B$ — Boltzmann constant. Reading it: since $T_s\gg0.07\,\mathrm{K}$ always, both levels stay near their statistical-weight ratio. The spontaneous decay rate is the smallest of any astrophysically important line; its inverse is the wait before an atom flips.

A_{10}=2.869\times10^{-15}\,\mathrm{s^{-1}},\qquad t_{\mathrm{rad}}=A_{10}^{-1}\approx1.1\times10^{7}\,\mathrm{yr} \qquad (3)

where: $A_{10}$ — Einstein coefficient (magnetic-dipole); $t_{\mathrm{rad}}$ — mean radiative lifetime of the $F=1$ level. The excitation is packaged into the spin temperature, defined so the populations obey a Boltzmann law — bookkeeping, not a real temperature.

\frac{n_1}{n_0}=\frac{g_1}{g_0}\,e^{-h\nu_{10}/k_B T_s}=3\,e^{-0.0682\,\mathrm{K}/T_s}\approx3 \qquad (4)

where: $n_1,n_0$ — densities in the $F=1$ and $F=0$ levels; $g_1=3,\,g_0=1$ — statistical weights; $T_s$ — spin temperature. Reading it: three-quarters of H I atoms sit in the upper state, nearly independent of $T_s$ . The absorption coefficient carries the crucial $1/T_s$ dependence, born from the near-cancellation of absorption by stimulated emission in the Rayleigh–Jeans limit; the visible $h\nu_{10}/k_B T_s$ factor shows where it comes from.

\kappa_\nu=\frac{3c^2}{32\pi}\,\frac{A_{10}}{\nu_{10}^2}\,\frac{h\nu_{10}}{k_B T_s}\,n_{\mathrm{HI}}\,\phi(\nu)=\frac{3\,h c^2 A_{10}}{32\pi\,\nu_{10}\,k_B T_s}\,n_{\mathrm{HI}}\,\phi(\nu) \qquad (5)

where: $\kappa_\nu$ — absorption coefficient per unit length; $n_{\mathrm{HI}}$ — neutral hydrogen number density; $\phi(\nu)$ — normalized profile, $\int\phi\,\mathrm{d}\nu=1$ . Reading it: opacity falls as $1/T_s$ because hotter gas has more stimulated emission cancelling absorption. Integrating opacity along the path and profile collapses the atomic constants into one famous number; the thin limit then removes even the spin temperature.

N_{\mathrm{HI}}=1.823\times10^{18}\,\frac{T_s}{\mathrm{K}}\!\int\tau\,\mathrm{d}\Big(\frac{v}{\mathrm{km\,s^{-1}}}\Big)\ \xrightarrow{\ \tau\ll1\ }\ 1.823\times10^{18}\!\int T_B(v)\,\mathrm{d}v \qquad (6)

where: $\tau$ — optical depth along the sightline; $N_{\mathrm{HI}}$ — H I column in $\mathrm{cm^{-2}}$ ; $T_B(v)$ — brightness in K, $v$ in $\mathrm{km\,s^{-1}}$ . Reading it: in the thin limit the area under the profile alone counts the atoms, no temperature needed. The radiative-transfer solution ties emission and absorption together and fixes the sign of the signal.

T_B(v)=(T_s-T_{\mathrm{bg}})\left(1-e^{-\tau_\nu(v)}\right) \qquad (7)

where: $T_{\mathrm{bg}}$ — background brightness (Galactic continuum, or the CMB). Emission when $T_s>T_{\mathrm{bg}}$ , absorption when $T_s<T_{\mathrm{bg}}$ . The intrinsic width of one component is thermal, set by the Maxwellian speeds of hydrogen; turbulence broadens it further.

\left(\frac{\Delta v_{\rm FWHM}}{\mathrm{km\,s^{-1}}}\right)_{\rm therm}=\sqrt{\frac{8\ln 2\,k_B T_k}{m_{\rm H}}}\Big/10^{5}=2.14\,\sqrt{\frac{T_k}{100\,\mathrm{K}}} \qquad (8)

where: $T_k$ — kinetic temperature; $m_{\rm H}$ — hydrogen mass; the $10^5$ converts $\mathrm{cm\,s^{-1}}$ to $\mathrm{km\,s^{-1}}$ . Reading it: cold gas gives lines a couple of $\mathrm{km\,s^{-1}}$ wide, warm gas tens — the widths separating the phases above. Finally, the spin temperature is set by three competing couplings — CMB, collisions, and Lyman- $\alpha$ scattering — as a coupling-weighted harmonic mean.

T_s^{-1}=\frac{T_{\mathrm{CMB}}^{-1}+x_c\,T_k^{-1}+x_\alpha\,T_\alpha^{-1}}{1+x_c+x_\alpha} \qquad (9)

where: $x_c$ — collisional coupling; $x_\alpha$ — Lyman- $\alpha$ (Wouthuysen–Field) coupling; $T_\alpha\approx T_k$ — colour temperature of the Lyman- $\alpha$ field. Reading it: when collisions or Lyman- $\alpha$ dominate, $T_s\to T_k$ and the gas decouples from the CMB.

Derivation AThe optically-thin column-density constant

The goal is to earn the single most-used number in H I astronomy — the coefficient $1.823\times10^{18}$ that turns an integrated 21 cm brightness temperature into a hydrogen column — and to see why the hard-to-measure spin temperature drops out in the optically-thin limit. We integrate the absorption coefficient along the sightline and across the profile, insert the atomic constants, then take the thin limit so $T_s$ cancels.

Step 1. Integrate equation (5) along the path. Since $N_{\mathrm{HI}}=\int n_{\mathrm{HI}}\,\mathrm{d}s$ and $\int\phi\,\mathrm{d}\nu=1$ , $\int\tau\,\mathrm{d}\nu=(3\,h c^2 A_{10})/(32\pi\,\nu_{10}\,k_B)\times(N_{\mathrm{HI}}/T_s)$ . Why: pulling the density integral out of the profile separates "how many atoms" from "how they are spread in velocity."

Step 2. Convert to velocity with $\mathrm{d}\nu=(\nu_{10}/c)\,\mathrm{d}v$ , gaining a factor $\nu_{10}/c$ : $\int\tau\,\mathrm{d}v=(3\,h c^3 A_{10})/(32\pi\,\nu_{10}^2\,k_B)\times(N_{\mathrm{HI}}/T_s)$ . Why: observers record spectra in $\mathrm{km\,s^{-1}}$ , and the extra $\nu_{10}$ turns the $\nu^{-1}$ opacity into the correct $\nu^{-2}$ scaling.

Step 3. Invert and insert numbers: $N_{\mathrm{HI}}=(32\pi\,\nu_{10}^2\,k_B)/(3\,h c^3 A_{10})\times T_s\int\tau\,\mathrm{d}v$ ; with $A_{10}=2.869\times10^{-15}\,\mathrm{s^{-1}}$ and $\nu_{10}=1.42041\times10^{9}\,\mathrm{Hz}$ ( $\mathrm{d}v$ in $\mathrm{km\,s^{-1}}$ ), the prefactor is $1.823\times10^{18}\,\mathrm{cm^{-2}\,K^{-1}\,(km\,s^{-1})^{-1}}$ . Why: a pure combination of fundamental and atomic constants, so the number is universal.

Step 4. Take the thin limit: for $\tau\ll1$ , $T_B=T_s(1-e^{-\tau})\approx T_s\tau$ , so $T_s\int\tau\,\mathrm{d}v=\int T_B\,\mathrm{d}v$ and $T_s$ cancels: $N_{\mathrm{HI}}=1.823\times10^{18}\int T_B\,\mathrm{d}v$ . Why: this cancellation is what makes 21 cm quantitative — you count the atoms without knowing the excitation, as long as the gas is not too thick. *Why the coefficient is $1.823\times10^{18}$ : inserting $A_{10}$ and $\nu_{10}$ into the prefactor $(32\pi\,\nu_{10}^2\,k_B)/(3\,h c^3 A_{10})$ gives $1.8224\times10^{18}$ , rounding to the standard value. Note also that the absorption coefficient carries $\nu_{10}^{-1}$ , not $\nu_{10}^{-2}$ ; the extra $\nu_{10}$ in the denominator comes from the $h\nu_{10}/k_B T_s$ stimulated-emission factor, which is what makes the dimensional accounting consistent.*

Go deeper · Derivation B · 1 minThe spin temperature as a weighted mean

Derivation BThe spin temperature as a weighted mean

The goal is to show how three processes — radiative coupling to the CMB, atomic collisions, and resonant scattering of Lyman- $\alpha$ photons — each drive the hyperfine ratio toward their own reservoir temperature, and how the outcome is a coupling-weighted average that decides whether the line is visible against the CMB. We balance the rates in statistical equilibrium, apply detailed balance, and linearize using $h\nu_{10}/k_B T\ll1$ .

Step 1. Write the equilibrium balance over every spin-flip process: $n_1(C_{10}+P_{10}+A_{10}+B_{10}I)=n_0(C_{01}+P_{01}+B_{01}I)$ , with $C$ collisional, $P$ Lyman- $\alpha$ -induced, and $B\,I$ radiative rates from the CMB intensity $I$ . Why: in the diffuse ISM and IGM the populations are set by competing rates, not one bath.

Step 2. Apply detailed balance: $C_{01}/C_{10}=3\,e^{-h\nu_{10}/k_B T_k}$ , $P_{01}/P_{10}=3\,e^{-h\nu_{10}/k_B T_\alpha}$ , radiative tied to $T_{\mathrm{CMB}}$ . Why: each process alone thermalizes the ratio to its own reservoir — $T_k$ , $T_\alpha$ , or $T_{\mathrm{CMB}}$ — so the answer interpolates.

Step 3. Solve for the effective ratio, $e^{-h\nu_{10}/k_B T_s}=\big(\sum_i R_i\,e^{-h\nu_{10}/k_B T_i}\big)/\sum_i R_i$ with $R_i$ each process rate. Why: $T_s$ is by definition the temperature reproducing the rate-averaged population ratio.

Step 4. Linearize: since $h\nu_{10}/k_B T_i\ll1$ , expand to first order and identify $x_c\propto$ collisional rate, $x_\alpha\propto$ Lyman- $\alpha$ rate, giving $T_s^{-1}=(T_{\mathrm{CMB}}^{-1}+x_c T_k^{-1}+x_\alpha T_\alpha^{-1})/(1+x_c+x_\alpha)$ . Why: $T_s$ is a harmonic mean weighted by coupling strength, so strong collisions or Lyman- $\alpha$ pumping pull it to $T_k$ and let the gas absorb the CMB. *Why the limits make sense: when $x_c\gg1$ or $x_\alpha\gg1$ the weighted mean gives $T_s=T_k$ , and when both couplings vanish it gives $T_s=T_{\mathrm{CMB}}$ — the two extremes that bracket all astrophysical situations.*

Go deeper · Derivation C · 1 minKinematic distance from Galactic rotation

Derivation CKinematic distance from Galactic rotation

The goal is to turn a measured 21 cm Doppler velocity into a location in the Galaxy, which is what lets the narrow, dust-penetrating line map spiral arms and the rotation curve. We project the differential rotation of a flat, axisymmetric disk onto the sightline at Galactic longitude $\ell$ , then specialize to the tangent point where the geometry gives the circular speed directly.

Step 1. A cloud at radius $R$ orbits at speed $\Theta(R)$ ; the observed radial velocity subtracts the Sun's projected motion, $v_r=\Theta(R)\cos\alpha-\Theta_0\sin\ell$ , with $\Theta_0$ the Sun's speed at $R_0$ and $\alpha$ the angle between cloud velocity and sightline. Why: only the line-of-sight component Doppler-shifts the line, and both cloud and observer orbit.

Step 2. Eliminate $\alpha$ with the sine rule of the Sun–centre–cloud triangle, $R\cos\alpha=R_0\sin\ell$ . Why: $\ell$ is measurable but $\alpha$ is not.

Step 3. Combine to the rotation law $v_r=R_0\sin\ell\,\big(\Theta(R)/R-\Theta_0/R_0\big)$ . Why: a velocity at known longitude constrains $R$ (with a near/far ambiguity for inner sightlines).

Step 4. At the tangent point the velocity is extremal: $v_{r,\max}=\Theta(R_{\min})-\Theta_0\sin\ell$ , $R_{\min}=R_0\sin\ell$ . Why: there $\cos\alpha=1$ , so the peak velocity along an inner sightline yields $\Theta(R)$ directly — the method that first showed the rotation curve stays flat beyond the visible disk.

Limiting cases

Three knobs — spin temperature, optical depth, and background radiation — govern the line's appearance. The four regimes below cover essentially every use, from Galactic surveys to cosmology.

Case	Result	Meaning
Optically thin emission	$\tau_\nu\ll1\ \Rightarrow\ T_B\approx T_s\tau_\nu\propto N_{\mathrm{HI}}$	The usual diffuse-ISM regime: brightness is independent of $T_s$ and linear in column density — the basis of $N_{\mathrm{HI}}=1.823\times10^{18}\int T_B\,\mathrm{d}v$ .
Optically thick emission	$\tau_\nu\gg1\ \Rightarrow\ T_B\to T_s$	Saturated cold H I; the profile flattens at the spin (≈ kinetic) temperature and the thin formula underestimates $N_{\mathrm{HI}}$ .
Absorption against a background	$T_B=(T_s-T_{\mathrm{bg}})(1-e^{-\tau_\nu})$	If $T_s<T_{\mathrm{bg}}$ the line is seen in absorption against a bright continuum source — how $T_s$ of the cold neutral medium is measured.
CMB-coupled IGM (cosmology)	$\delta T_b\propto\big(1-T_{\mathrm{CMB}}/T_s\big)$	Differential brightness against the CMB vanishes at $T_s=T_{\mathrm{CMB}}$ and is negative (absorption) when $T_s<T_{\mathrm{CMB}}$ — the cosmic-dawn trough.

Common trapThe spin temperature is the kinetic temperature, and the line is always optically thin

Both halves are false, and believing them is the classic 21 cm error. The spin temperature equals the kinetic temperature only when a coupling (collisions in dense gas, Lyman- $\alpha$ pumping at cosmic dawn) overpowers the CMB; in the warm neutral medium or the general IGM the two can differ by orders of magnitude — and the cosmological signal exploits exactly that difference. The line is likewise only usually thin: along cold, high-column sightlines the optical depth exceeds unity, self-absorption saturates the profile at $T_s$ , and the naive estimator underestimates the column, sometimes badly. A last trap is reading the tiny $\Delta E/k_B=0.068\,\mathrm{K}$ as an excitation barrier: it is the opposite, a gap so small both levels stay populated at the $3{:}1$ ratio.

Worked examples

1 · A hydrogen column from an emission profile (fully worked)

The sightline in the figure above has $\int T_B\,\mathrm{d}v=819\,\mathrm{K\,km\,s^{-1}}$ . Find the H I column, decide whether the thin approximation holds for each phase, and estimate the mass surface density.

Step 1. Apply the thin estimator to the integrated area: $N_{\mathrm{HI}}=1.823\times10^{18}\times819=\mathbf{1.49\times10^{21}\,cm^{-2}}$ — typical for a few hundred parsecs through the disk.

Step 2. Test thinness phase by phase. The warm hump peaks at $T_B\approx14\,\mathrm{K}$ ; with $T_s\approx8000\,\mathrm{K}$ , $\tau\approx T_B/T_s\approx1.8\times10^{-3}$ — emphatically thin, so the estimator is exact for the WNM. The cold spike peaks at $T_B\approx58\,\mathrm{K}$ ; with $T_s\approx70\,\mathrm{K}$ , $1-e^{-\tau}=58/70=0.83$ , hence $\tau=-\ln(0.17)\approx\mathbf{1.8}$ — the CNM is not thin.

Step 3. Since $\tau_{\rm CNM}\approx1.8$ , the thin estimator undercounts the cold gas by $\tau/(1-e^{-\tau})\approx2.2$ at line centre; recovering the true CNM column needs paired emission-and-absorption toward a background source.

Step 4. Convert to mass surface density with $\Sigma_{\rm HI}=1.4\,m_{\rm H}N_{\mathrm{HI}}$ (helium correction): $1.4\times1.67\times10^{-24}\times1.49\times10^{21}\approx\mathbf{17\,M_\odot\,pc^{-2}}$ , an ordinary disk value. Result: $N_{\mathrm{HI}}\approx1.5\times10^{21}\,\mathrm{cm^{-2}}$ ; the WNM is thin but the CNM ( $\tau\approx1.8$ ) is undercounted by $\sim2\times$ .

2 · The cosmic-dawn absorption trough (fully worked)

At $z=17$ , $T_{\mathrm{CMB}}=2.725(1+z)=49\,\mathrm{K}$ . If Lyman- $\alpha$ pumping couples the spin temperature to the adiabatically-cooled gas at $T_k\approx6\,\mathrm{K}$ , estimate the differential 21 cm brightness against the CMB and compare with EDGES.

Step 1. Use the standard global-signal expression $\delta T_b\approx27\,x_{\mathrm{HI}}(1+\delta)\Big[\frac{0.15}{\Omega_m h^2}\frac{1+z}{10}\Big]^{1/2}\Big(\frac{\Omega_b h^2}{0.023}\Big)\Big(1-\frac{T_{\mathrm{CMB}}}{T_s}\Big)\,\mathrm{mK}$ , with $x_{\mathrm{HI}}=1,\ \delta=0$ , $\Omega_m h^2=0.143$ , $\Omega_b h^2=0.0224$ .

Step 2. The prefactor: bracket $\sqrt{(0.15/0.143)(1.8)}=1.37$ and $\Omega_b h^2/0.023=0.974$ give amplitude $27\times1.37\times0.974=\mathbf{36\,mK}$ .

Step 3. With strong Lyman- $\alpha$ coupling, $T_s\to T_k=6\,\mathrm{K}$ : $1-T_{\mathrm{CMB}}/T_s=1-49/6=-7.2$ — the gas is far colder than the CMB, so the sign is negative (absorption).

Step 4. Combine: $\delta T_b\approx36\times(-7.2)\approx\mathbf{-260\,mK}$ . Standard models with weaker coupling and early X-ray heating predict a shallower $\sim-200\,\mathrm{mK}$ ; EDGES (2018) reported $\sim-500\,\mathrm{mK}$ at $z\approx17$ — twice any prediction, which if real demands extra cooling or an excess radio background. Result: a cosmic-dawn absorption signal of order $-200$ to $-260\,\mathrm{mK}$ , the observable dark-ages experiments chase.

Physics of the portrait

The 21 cm line is a rare case where theory came first: in wartime occupied Holland a young Dutch student worked out that neutral hydrogen ought to have one usable radio line — and told astronomers where to point.

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

Hendrik van de Hulst

1918–2000 · radio astronomy of neutral hydrogen

Hendrik Christoffel van de Hulst was a graduate student in the German-occupied Netherlands when, at a 1944 colloquium organized by Jan Oort, he asked whether interstellar hydrogen might produce a detectable radio line. Reasoning from the hyperfine structure of the ground state, he predicted a spin-flip transition near 21 cm and — crucially — argued that although the transition probability was minuscule, the sheer quantity of Galactic hydrogen would make it observable. It took seven years and post-war radio technology to confirm him: Ewen and Purcell detected the line at Harvard in March 1951, Muller, Oort, and Christiansen confirming within weeks. Van de Hulst went on to co-develop the theory of interstellar dust and help found the Dutch radio-astronomy program that built Westerbork. His single wartime insight opened the neutral Universe, and it remains the backbone of Galactic H I astronomy and the target frequency, redshifted, of the largest radio telescopes ever built.

History

1944

— van de Hulst predicts the 21 cm hyperfine line of interstellar hydrogen

1951

— Ewen and Purcell detect the line at Harvard; Muller, Oort, and Christiansen confirm within weeks

1952

— the first 21 cm surveys begin tracing Galactic spiral structure

1958

— Field formulates the spin-temperature theory of collisional, radiative, and Lyman-α coupling

2018

— EDGES reports an anomalously deep 21 cm absorption trough from cosmic dawn near z ≈ 17

Going further

Texts. Draine, Physics of the Interstellar and Intergalactic Medium, Ch. 8 (hyperfine transition, spin temperature, column-density constant) · Carroll & Ostlie, An Introduction to Modern Astrophysics, 2nd ed., §§12.1–12.2 · Binney & Tremaine, Galactic Dynamics, 2nd ed., Ch. 1 and §9.1 (H I kinematics) · Mo, van den Bosch & White, Galaxy Formation and Evolution, Ch. 11 · Field (1958), Proc. IRE 46, 240 (spin-temperature theory) · Furlanetto, Oh & Briggs (2006), Phys. Rep. 433, 181 (the 21 cm transition). The atomic data used throughout — $\nu_{10}=1420.405751768\,\mathrm{MHz}$ , $\lambda_{10}=21.106\,\mathrm{cm}$ , $\Delta E=5.874\times10^{-6}\,\mathrm{eV}=0.0682\,\mathrm{K}$ , $A_{10}=2.869\times10^{-15}\,\mathrm{s^{-1}}$ giving $t_{\mathrm{rad}}\approx1.1\times10^{7}\,\mathrm{yr}$ — are standard values. Two points of care: the absorption coefficient carries $\nu_{10}^{-1}$ , not $\nu_{10}^{-2}$ (the compensating $\nu_{10}$ comes from the $h\nu_{10}/k_B T_s$ stimulated-emission factor, making the dimensions consistent); and the thermal FWHM coefficient is $2.14\,\sqrt{T_k/100\,\mathrm{K}}\,\mathrm{km\,s^{-1}}$ , not $0.21$ (which would be ten times too small).

Prerequisites. Phases of the Interstellar Medium · Interstellar Gas Heating and Cooling Processes · Thermal Instability and the Two-Phase Medium. Leads to. CO as a Molecular Gas Tracer and the X-Factor · Turbulence in the Interstellar Medium · Molecular Clouds and Giant Molecular Clouds.

Ask the tutor: Ask me why the spin temperature drops out of the thin-limit column density, or how the Wouthuysen–Field effect turns on the cosmic-dawn absorption trough.

Why is the transition called "forbidden" if we detect it everywhere?

"Forbidden" is spectroscopic jargon for "not allowed as an electric-dipole transition," not "impossible." The spin-flip cannot use the strong electric-dipole channel of ordinary bright lines, so it falls back on the far weaker magnetic-dipole coupling, giving the tiny rate $A_{10}=2.869\times10^{-15}\,\mathrm{s^{-1}}$ and an eleven-million-year lifetime per atom. Numbers rescue it: one sightline holds $\sim10^{21}$ hydrogen atoms per square centimetre, and the sum of all those rare flips is a bright line. Forbidden lines are in fact prized because they arise in gas too thin to collisionally de-excite the upper level before it radiates.

What exactly is the spin temperature, and is it a real temperature?

The spin temperature $T_s$ is defined as whatever temperature makes the hyperfine populations obey $n_1/n_0=3\,e^{-h\nu_{10}/k_B T_s}$ — a bookkeeping device, not necessarily the temperature of anything. In dense gas, frequent collisions drive $T_s$ to the kinetic temperature $T_k$ , so there it is a real thermometer. In the rarefied IGM it is instead pulled between $T_{\mathrm{CMB}}$ , $T_k$ , and the Lyman- $\alpha$ colour temperature, as a coupling-weighted harmonic mean. All 21 cm cosmology hinges on this: the signal exists only when some process pushes $T_s$ away from $T_{\mathrm{CMB}}$ .

How can 21 cm give a column density without knowing the temperature?

Only in the optically-thin limit, where the cancellation is exact. Opacity scales as $N_{\mathrm{HI}}/T_s$ (hotter gas has more stimulated emission cancelling absorption), while thin-limit brightness is $T_B\approx T_s\tau$ . Multiply and $T_s$ cancels: $\int T_B\,\mathrm{d}v=T_s\int\tau\,\mathrm{d}v\propto N_{\mathrm{HI}}$ , giving $N_{\mathrm{HI}}=1.823\times10^{18}\int T_B\,\mathrm{d}v$ . The catch is "thin": along cold, high-column sightlines the optical depth exceeds one and this estimator undercounts the gas — so cold clouds are weighed with paired emission-and-absorption instead.

Why is the line so narrow, and why does that matter?

The intrinsic width is thermal: hydrogen atoms move at Maxwellian speeds, Doppler-broadening the line by $\Delta v_{\rm FWHM}=2.14\sqrt{T_k/100\,\mathrm{K}}\,\mathrm{km\,s^{-1}}$ — under $2\,\mathrm{km\,s^{-1}}$ for cold gas, about $20$ for warm gas. That is far narrower than the velocity spread from Galactic rotation (hundreds of $\mathrm{km\,s^{-1}}$ ), so the profile is essentially a velocity histogram. This makes 21 cm a kinematic tool: each component pins a cloud's line-of-sight velocity, which a rotation model turns into a distance, revealing spiral arms, the disk warp, and high-velocity clouds that dust would hide.

What is the Wouthuysen–Field effect and why does cosmic dawn need it?

In the dark ages, intergalactic hydrogen sits at $T_s\approx T_{\mathrm{CMB}}$ because collisions in the thin, expanding gas are too rare to couple $T_s$ to the colder kinetic temperature — so there is no signal. The Wouthuysen–Field effect is the rescue: Lyman- $\alpha$ photons from the first stars scatter resonantly off hydrogen, and each scattering can flip the hyperfine spin, coupling $T_s$ to the Lyman- $\alpha$ colour temperature ( $=T_k$ ). Once it switches on, $T_s$ drops below $T_{\mathrm{CMB}}$ and the gas absorbs the microwave background — the cosmic-dawn trough that EDGES and HERA are built to detect.

How does 21 cm reveal dark matter in galaxies?

Neutral hydrogen disks extend well beyond the stellar disk, so 21 cm interferometry traces the rotation curve $\Theta(R)$ to radii where starlight has faded to nothing. If only luminous matter were present, orbital speeds should fall as $\Theta\propto R^{-1/2}$ (Keplerian) past the light; instead H I rotation curves stay flat to the last point, requiring an extended halo of unseen mass — dark matter. Because the line penetrates dust and reaches large radii, it gives the cleanest galaxy-scale dynamical evidence for dark halos, and underpins the Tully–Fisher relation linking H I linewidth to luminosity.