Event Horizon Telescope

The orange on the Moon

To photograph the edge of a black hole you need a telescope that can pick out an object about the angular size of an orange sitting on the surface of the Moon. M87*, the supermassive black hole the Event Horizon Telescope imaged first, has a shadow only about 42 microarcseconds across — a microarcsecond being a millionth of an arcsecond, or 1/3,600,000,000 of a degree. The Hubble Space Telescope, magnificent as it is, resolves about 50,000 microarcseconds. You would need an optical telescope a thousand times sharper, or a radio dish the size of a continent.

You cannot build a dish the size of a continent. But you can fake one. If you scatter radio receivers across the planet, record the incoming wave at each, tag every sample with an atomic-clock timestamp, and later combine the recordings in a computer, the array behaves — for the purpose of angular resolution — like one telescope as wide as the farthest pair of dishes. That trick is interferometry, and the EHT is its most extreme expression: a telescope as wide as the Earth, pointed at the darkest object in the universe.

Why bigger means sharper: θ ≈ λ/D

Every telescope is limited by diffraction. Light entering an aperture of diameter D spreads into a blur whose angular width is

θ ≈ λ / D     (radians)

where λ is the observing wavelength. (For a single filled circular dish the precise Rayleigh factor is 1.22 λ/D; for an interferometer the relevant scale is set by the baseline length, so the order-of-magnitude estimate uses λ/D.) Two points closer together than θ merge into one. To see a tiny object you make D larger or λ smaller. The EHT does both, pushed to extremes. It observes at λ = 1.3 mm, where the plasma swirling around the black hole becomes transparent, and it builds a virtual D equal to the longest baseline between its stations — roughly the diameter of the Earth, 1.07 × 10⁷ m. Plugging in:

θ ≈ 1.3 × 10⁻³ m / 1.07 × 10⁷ m
  ≈ 1.2 × 10⁻¹⁰ rad
  ≈ 25 microarcseconds

That is the headline number. Twenty-five microarcseconds is about 2,000 times finer than Hubble, and it is no accident that it sits just below the predicted shadow size of M87* and Sagittarius A*. The targets were chosen precisely because they are the two black holes whose shadows are largest on our sky — the only two the EHT can currently resolve.

How interferometry forges an Earth-sized dish

A filled dish works by collecting a wavefront across its whole surface and bringing it to a focus, where the contributions interfere. An interferometer skips the surface and keeps only the edges. Two dishes separated by a baseline vector B each sample the same incoming wavefront, but a wave arriving from angle θ reaches one dish a little before the other. The extra path is B·ŝ, where ŝ points at the source. Correlating the two recorded signals measures the phase of that delay, which encodes one Fourier component of the sky brightness — one "spatial frequency" set by the baseline length projected onto the sky, in units of λ.

The longer the baseline, the higher the spatial frequency it samples, and the finer the detail it carries. A single baseline gives you one Fourier component. As the Earth rotates, each baseline sweeps through a range of orientations, filling in more of the Fourier ("u-v") plane. With eight stations you have 8 × 7 / 2 = 28 simultaneous baselines, and a full night of Earth rotation paints a sparse but informative track across the u-v plane. An imaging algorithm then inverts those measurements into a picture. Because the coverage is sparse, the inversion is under-determined — which is why the EHT runs several independent imaging pipelines and demands they agree before trusting a feature.

Atomic clocks and shipped hard drives

For the correlation to work, each station must timestamp its data so precisely that recordings made thousands of kilometres apart can be aligned to a fraction of the 1.3 mm wave period — about 4 picoseconds. That demands a hydrogen-maser atomic clock at every site, stable to roughly one part in 10¹⁵, so the clocks drift by less than the wave period over the hours of an observation.

The data volume is staggering. Each station digitises the radio field at roughly 64 gigabits per second across both polarisations and several frequency bands. Over a multi-night campaign that fills thousands of helium-sealed hard drives — of order 5 petabytes per observing run. There is no network on Earth that can move that fast, so the drives are packed in crates and physically flown to two correlator centres: the MIT Haystack Observatory in Massachusetts and the Max Planck Institute for Radio Astronomy in Bonn. The South Pole Telescope's drives famously had to wait for the Antarctic winter to end before a plane could collect them, delaying the first M87* result.

What the dark patch actually is

The EHT does not photograph the event horizon. It photographs the shadow: a central deficit of brightness ringed by a bright crescent. Light from the hot plasma behind and around the black hole is bent by gravity; rays that pass within a critical impact parameter are captured and never reach us, leaving a dark disk on the sky. Outside that boundary, gravitationally lensed emission piles up into a thin, bright photon ring.

For a non-spinning (Schwarzschild) black hole the shadow's angular radius corresponds to a critical impact parameter

b_crit = 3√3 · GM/c² ≈ 5.196 GM/c²
shadow diameter = 2 b_crit = 6√3 GM/c² ≈ 10.4 GM/c²
              ≈ 5.2 Schwarzschild radii (R_s = 2GM/c²)

So the shadow is about 2.6 times wider than the event horizon itself — gravitational lensing magnifies the silhouette. The photon ring lies just outside, near the unstable photon orbit at 3 GM/c² for Schwarzschild (the photon sphere). The crescent is brighter on one side because the plasma orbits at a sizeable fraction of light speed; relativistic Doppler beaming boosts the side approaching us. That asymmetry is how the EHT reads off the black hole's spin orientation.

The two targets, by the numbers

Property	M87*	Sagittarius A*
Mass	6.5 × 10⁹ M☉	4.15 × 10⁶ M☉
Distance	16.8 Mpc (≈ 55 Mly)	8.15 kpc (≈ 26,700 ly)
Shadow angular diameter	≈ 42 µas	≈ 52 µas
Schwarzschild radius R_s	≈ 1.9 × 10¹⁰ km (≈ 128 AU)	≈ 1.2 × 10⁷ km (≈ 0.08 AU)
ISCO orbital period	≈ days–weeks	≈ 4–30 minutes
Image released	10 April 2019	12 May 2022
Key difficulty	Faint, distant, but steady	Bright but rapidly variable

Notice the two shadows have almost the same angular size despite the masses differing by a factor of 1,570. M87* is about 2,000 times farther away than Sgr A*, and that distance almost exactly cancels its larger physical size. It is a cosmic coincidence that the largest black hole in our extragalactic neighbourhood and the one at our own Galactic Centre both happen to subtend roughly 50 microarcseconds — and that this is just within reach of an Earth-sized telescope.

Why 1.3 mm, and why not visible light?

Three constraints meet at 1.3 mm. First, resolution: since θ ∝ λ, you want the shortest wavelength you can use. Second, the source: the magnetised plasma around the black hole is optically thick at centimetre wavelengths — it self-absorbs and is smeared by interstellar scattering, especially toward the Galactic Centre — but it becomes transparent in the millimetre band, so the photon ring shines through. Third, the atmosphere: water vapour absorbs sub-millimetre radiation, but there is a usable window near 1.3 mm at high, dry sites like the Atacama plateau and Mauna Kea.

Visible light is hopeless here for two reasons. The plasma does not emit a clean image at optical wavelengths, and an optical interferometer would need its mirror paths matched to a fraction of an optical wavelength (~500 nm) across an Earth-sized baseline — mechanically impossible. Radio interferometry instead records the field digitally and aligns it in software after the fact, which is why the trick works at 1.3 mm but not in the visible. The EHT is now commissioning observations at 0.87 mm (345 GHz), which sharpens the resolution by roughly 30% and tightens the photon ring.

The array and the campaign

The 2017 campaign that produced the M87* and Sgr A* images linked eight facilities at six geographic sites: the Atacama Large Millimeter/submillimeter Array (ALMA) and the Atacama Pathfinder Experiment (APEX) in Chile; the Large Millimeter Telescope (LMT) in Mexico; the South Pole Telescope (SPT) in Antarctica; the IRAM 30 m telescope in Spain; the James Clerk Maxwell Telescope (JCMT) and Submillimeter Array (SMA) on Mauna Kea; and the Submillimeter Telescope (SMT) in Arizona. ALMA, itself an array of 66 dishes acting as one giant collecting area, is the most sensitive node and anchors the network.

Observing windows are brief and brutal. The team needs simultaneous clear, dry skies at sites on four continents, which restricts campaigns to about a week each spring. The black hole never sets at some stations and never rises at others, so any given baseline only operates for part of the night. The array has since grown to include Greenland, the NOEMA array in France, and the Kitt Peak telescope, improving u-v coverage for later campaigns.

Common misconceptions and edge cases

• "It's a photo of the event horizon." No. The dark region is the shadow — a lensed silhouette about 2.6× larger than the horizon. The horizon itself emits nothing and cannot be imaged; what you see is the absence of background light plus the bright photon ring.
• "The whole Earth is one giant dish, so it collects like one." Only for resolution, not sensitivity. The collecting area is just the sum of the real dishes, so the EHT is faint-limited; it can only see the brightest, largest-shadow targets. The Earth-sized aperture buys angular resolution, not light grasp.
• "They took a single snapshot." The image is reconstructed from sparse Fourier samples accumulated over hours, then run through multiple independent algorithms (CLEAN, regularised maximum likelihood, Bayesian methods) that must agree. A feature only survives if all pipelines reproduce it.
• "M87* is bigger on the sky because it's the bigger black hole." M87* is far more massive, but also vastly more distant; the two effects nearly cancel, leaving M87* and Sgr A* with comparable ~50 µas shadows.
• "The ring is the accretion disk." The bright ring is dominated by gravitationally lensed photon-ring emission, not a direct view of a flat disk. Light from behind the hole is bent around it, so the ring contains contributions from material we could never see in a straight line.
• "Sgr A* was harder because it's faint." It is actually bright and nearby; the difficulty is time variability. Its innermost gas orbits in minutes, so the source morphs during one night's observation — the image had to be averaged over thousands of plausible variable realisations.