21 cm Cosmology

The line that maps the invisible universe

For its first few hundred million years, the universe was dark. Roughly 380,000 years after the Big Bang the cosmos cooled enough for protons and electrons to combine into neutral hydrogen — the recombination era that released the cosmic microwave background. After that flash, the universe held no stars, no galaxies, no light sources at all: just a vast, expanding fog of neutral hydrogen and helium. This stretch, lasting until the first stars switched on, is the cosmic dark ages. Optical and infrared telescopes — even the James Webb Space Telescope — see essentially nothing there, because there was nothing yet emitting visible light.

But that hydrogen fog was not silent. Every neutral hydrogen atom is a tiny radio transmitter. Its single electron and single proton each carry an intrinsic magnetic moment, and the two can be aligned (spins parallel) or anti-aligned (spins anti-parallel). The parallel state has very slightly more energy. When an atom drops from the parallel to the anti-parallel state, it emits one photon carrying that energy difference — about 5.9 microelectronvolts, corresponding to a wavelength of 21.106 cm and a frequency of 1420.405751 MHz. This is the 21 cm line, produced by the spin-flip (hyperfine) transition.

The transition is fantastically forbidden by the usual rules of atomic physics: a lone atom in the parallel state takes, on average, about 11 million years to flip. No laboratory could ever detect emission that slow. But interstellar and intergalactic space hold so many hydrogen atoms — on the order of 10⁸⁰ in the observable universe — that at any instant a huge number are flipping. The aggregate glow is bright enough that it has been the workhorse of radio astronomy since its first detection in 1951, used to map the spiral arms and rotation of the Milky Way. 21 cm cosmology pushes the same line to its ultimate use: probing the gas of the entire early universe.

Reading cosmic time off the dial

The trick that turns one spectral line into a map of cosmic history is redshift. As space expands, a 21 cm photon emitted long ago stretches in wavelength on its way to us. A photon emitted at redshift z arrives at an observed frequency of 1420.4 MHz divided by (1 + z). The further back in time the emission, the lower its observed frequency. This makes the radio dial a direct timeline of the cosmos:

Where the redshifted 21 cm signal lands for each epoch

Epoch	Redshift z	Age of universe	Observed frequency
Recombination / CMB	~1100	~380,000 yr	~1.3 MHz
Dark ages	~30-200	~2-100 Myr	~7-46 MHz
Cosmic dawn (first stars)	~15-30	~100-270 Myr	~46-89 MHz
Epoch of reionization	~6-12	~0.4-0.95 Gyr	~109-203 MHz
Post-reionization (intergalactic + galaxies)	~0-6	~0.95-13.8 Gyr	~203-1420 MHz

Because each observed frequency corresponds to a unique distance and look-back time, tuning a radio telescope across the band is equivalent to slicing the universe into shells at different ages. Add the two angular dimensions on the sky, and the result is genuine three-dimensional tomography — a data cube where the third axis is cosmic time. No other observable gives a continuous 3D map of so much of the universe's volume, which is why 21 cm is sometimes called the richest untapped reservoir of cosmological information.

Emission, absorption, and the spin temperature

Whether the gas shows up as a bright signal, a dark one, or nothing at all depends on the spin temperature, T_S — a number that encodes the ratio of atoms in the upper versus lower hyperfine state. The 21 cm signal is always measured against the cosmic microwave background, which acts as a backlight. The comparison gives the differential brightness temperature:

• If T_S > T_CMB, the gas emits — a bright (positive) signal.
• If T_S < T_CMB, the gas absorbs the CMB — a dark (negative) signal, seen as an absorption trough.
• If T_S = T_CMB, the line vanishes and the gas is invisible at 21 cm.

Three processes fight over T_S. CMB photons drive it toward the CMB temperature. Atomic collisions drive it toward the kinetic gas temperature (important when the gas was dense early on). And — crucially — once the first stars light up, their Lyman-alpha photons couple T_S to the cold gas via the Wouthuysen-Field effect. This is the chain that makes 21 cm a clock: as the universe evolves, the signal flips between absorption and emission, and the timing of those flips reveals when collisions, first starlight, and finally X-ray heating took over.

The expected 21 cm signal through cosmic history

Phase	Dominant coupling	Signal vs CMB
Early dark ages (dense)	Collisions to cold gas	Absorption
Mid dark ages (rarefied)	CMB wins; TS → TCMB	No signal
Cosmic dawn	Lyman-α (Wouthuysen-Field) to cold gas	Deep absorption
X-ray heating	Gas heated above CMB	Emission
Reionization	Hydrogen ionized — no line	Signal disappears (bubbles)

Bubbles: tracing reionization

The most observable stretch is the epoch of reionization, roughly z ≈ 6-15 (about 0.3-1 billion years after the Big Bang). Ultraviolet light from the first stars and galaxies ionized the neutral hydrogen around them, and because ionized hydrogen has no electron to flip, it produces no 21 cm line. Each early galaxy therefore carves a growing bubble of silence into the glowing neutral fog. Over time the bubbles expand, overlap, and finally merge until the intergalactic medium is fully ionized — the state it remains in today.

A 21 cm tomographic map of this era is a movie of those bubbles growing. Their characteristic size (tens of millions of light-years across at the peak), how their pattern correlates with the underlying matter, and how quickly the universe transitions from mostly-neutral to mostly-ionized all encode the nature of the first light sources: were they faint dwarf galaxies, rare bright quasars, or something exotic? Independent constraints from CMB polarization (the optical depth to reionization measured by Planck) and from the Lyman-alpha forest in distant quasar spectra agree that reionization ended near z ≈ 6, but only 21 cm can show the process unfolding in space and time.

EDGES, foregrounds, and the hardest measurement in radio astronomy

In 2018 the EDGES experiment — a single, table-sized dipole antenna in the Western Australian desert — reported a sky-averaged absorption trough near 78 MHz, corresponding to z ≈ 17, about 180 million years after the Big Bang. If real, it is the fingerprint of cosmic dawn: the first starlight coupling the spin temperature to cold gas. But the trough was roughly twice as deep as standard cosmology predicts. Explanations ranged from the gas being colder than expected (perhaps via interactions between baryons and dark matter) to an unaccounted-for radio background. As of the mid-2020s the result is still unconfirmed: the SARAS 3 experiment reported no detection of the same profile, so the community treats EDGES as tentative pending independent verification.

The difficulty is brutal. The cosmological 21 cm signal is on the order of tens of millikelvin (or less). Sitting in front of it is Galactic synchrotron foreground emission that is 10,000 to 100,000 times brighter, plus terrestrial radio interference, ionospheric distortion, and instrument-calibration errors that mimic the signal. Two complementary strategies attack the problem:

Global versus interferometric 21 cm experiments

Approach	What it measures	Instruments
Global (sky-averaged)	Mean 21 cm signal vs frequency	EDGES, SARAS 3, REACH
Interferometric (power spectrum)	Statistical fluctuations	LOFAR, MWA, HERA
Interferometric (tomography)	3D images of bubbles	Square Kilometre Array (SKA)

The Square Kilometre Array, now under construction, is designed to be the first instrument capable of true imaging tomography of reionization rather than just a statistical detection. Combined with a possible future lunar far-side radio array — shielded from Earth's interference and ionosphere — 21 cm cosmology may eventually reach all the way back into the pristine dark ages at z ≈ 30-200, an era that holds clean cosmological information untouched by messy astrophysics.

Why 21 cm cosmology matters

• The only probe of the dark ages. No light sources existed, but neutral hydrogen still spoke at 21 cm.
• 3D mapping. Redshift turns frequency into distance, yielding tomographic data cubes across cosmic time.
• The first stars. The cosmic-dawn absorption signal marks when starlight first changed the gas.
• Reionization history. Growing ionized bubbles reveal what sources reionized the universe and when.
• Fundamental physics. Anomalies like EDGES could hint at dark-matter interactions or new radio backgrounds.
• Enormous volume. 21 cm can access far more of the observable universe than any galaxy survey.

Common misconceptions

• "21 cm comes from electron orbital transitions." No — it is the hyperfine spin-flip, not a jump between orbital energy levels.
• "We observe it at 21 cm." Only nearby gas; cosmological signals are redshifted to metre wavelengths (50-200 MHz).
• "It is always emission." The line can appear in emission or absorption depending on spin temperature versus the CMB.
• "Ionized gas glows at 21 cm too." Ionized hydrogen has no bound electron, so it shows up as silence — the bubbles.
• "EDGES confirmed cosmic dawn." The signal is tentative and not yet independently verified.
• "Foregrounds are a minor nuisance." They outshine the signal by up to 100,000 times — they are the central challenge.