Two feeding black holes found 4,900 light-years apart in an early galaxy merger

Webb’s spectra split one dust-buried X-ray source into two black holes of roughly 69 million and 17 million Suns, about 1.5 kiloparsecs apart, though the claim that the pair is feeding faster than the Eddington limit rests on a single, unresolved X-ray spectrum.

Peer reviewed · Observations of one system · Masses and growth rate are estimates

Two feeding supermassive black holes sit about 4,900 light-years apart inside a merging galaxy seen some 1.3 billion years after the Big Bang. Hyewon Suh of the International Gemini Observatory/NSF NOIRLab and 17 co-authors report the system, LID-1166, in Nature Astronomy. Close pairs like this are expected in the young universe but rarely confirmed, because dust hides them. The catch: the masses are indirect estimates, and the headline growth rate depends on how obscured the X-ray source is assumed to be.

Evidence at a glance

Is this a big deal?

A careful detection with an interpretation still to test. Two black holes this close, this early, confirmed by two telescopes, is uncommon. The super-Eddington growth is the part to hold loosely.

What scientists found

LID-1166 began as a strong X-ray source in the Chandra COSMOS Legacy Survey with nothing visible at its position, even in Hubble’s deepest images: a black hole feeding behind thick dust.

JWST saw two compact sources of hydrogen-alpha light, each with broad lines, the signature of gas whirling near a black hole. Their redshifts, 4.502 and 4.499, put the companion moving about 164 km/s toward us relative to the primary.

From those broad lines the team estimates about 69 million solar masses for the primary and 17 million for the companion. This standard shortcut for weighing black holes has wide error bars: the primary could sit anywhere from 49 to 155 million Suns.

How they found it

Think of two headlights in fog. X-rays get through, but Chandra’s roughly 2-arcsecond view blurs both lamps into one. JWST, observing for 1.45 hours in April 2023, split them. The analogy breaks in one place: these lamps differ in brightness and line widths, part of how the team knows they are not one source seen twice.

The primary’s glare could fake a neighbour, so Suh’s team modelled and subtracted the telescope’s blur pattern for a point of light. The companion survived when they shifted the model, changed the aperture and swapped in a real calibration star.

On 16 March 2025, ALMA’s 25-minute map of ionised carbon showed two gas components at the same positions and velocities. The differing redshifts and line ratios rule out a gravitationally lensed double image, the authors argue.

What “rapid growth” means here

The Eddington limit is the output at which radiation pushing on infalling gas balances gravity. Exceeding it is one proposed way early black holes grew so big so fast.

Assuming the X-rays pass through Compton-thick gas (about 10²⁴ hydrogen atoms per square centimetre, dense enough to scatter X-rays), the team estimates a combined output of 10^46.6 erg/s, about 4 × 10³⁹ watts or 10 trillion Suns. That is 3.7 times the pair’s Eddington limit (range 2.4 to 8.9), or about 6.3 solar masses swallowed per year if 10% of infalling mass is radiated.

SciQuest’s check: at a steady 6.3 Suns a year, the pair’s 86 million solar masses would double in about 14 million years. The authors expect such growth to be episodic, so treat that as an upper bound.

Why it matters to you

Nothing here changes daily life. It speaks to how giant elliptical galaxies and their black holes were built; our galaxy’s own Sagittarius A* is a far smaller nearby example. Pairs like this may be ancestors of the binaries behind low-frequency gravitational waves, which pulsar-timing arrays search for.

What the evidence does not show

The 1.5 kpc gap is projected on the sky; the true distance could be larger. The obscured X-ray model improved the Chandra fit only modestly, and simpler models drop the Eddington ratio to 0.2 to 0.5. The team also notes that the two host galaxies’ starlight is not cleanly detected, and calls its dynamical mass an order-of-magnitude constraint.

How this fits what we already knew

The authors’ July preprint called LID-1166 the first close-separation dual AGN known beyond the local universe. The published abstract instead says such pairs remain rare. That is the safer reading: other JWST teams have reported candidates, including a z ≈ 5 galaxy with possible black holes 190 parsecs apart (Übler et al., preprint).

Using dynamical mass as a stand-in for bulge mass, the authors place the pair on the black-hole-to-host relation of today’s massive ellipticals, and propose that hidden, merger-driven feeding sets it up early.

Outside views

SciQuest has not interviewed an independent researcher. Live Science reported that Anna Trindade Falcão of NASA Goddard, who was not involved, said the companion’s survival across several subtraction tests supported it being real, but suggested a more detailed follow-up to rule out one black hole plus leftover glow. Independent JWST work suggests such pairs are not unusual: Perna et al. (2025) found multiple AGN in 3 to 5 of 16 JWST-observed systems, at 3 to 28 kpc.

How other headlines framed it

• Live Science called it the oldest, closest pair, reported from the preprint, and included an outside view.
• Science Times repeated the superlatives without naming the paper or giving masses.

SciQuest drops the “first” and “closest” framing, which the published abstract no longer uses.

What happens next

The authors call for systematic integral-field surveys with JWST and adaptive-optics spectrographs on large ground telescopes to find more hidden pairs.

In one sentence

JWST and ALMA have pulled two feeding black holes out of one dust-hidden galaxy merger 1.3 billion years after the Big Bang, and whether they are outgrowing the Eddington limit depends on X-ray modelling still open to test.

Read next

• What is the biggest black hole? Why many black-hole masses are estimates.
• JWST measures gas near Sagittarius A*: our own galaxy’s black hole, up close.
• Can a black hole destroy a galaxy? What feeding black holes do to their hosts.

Sources and disclosure

1. Suh H, Decarli R, Farina EP, Scharwächter J, Mezcua M, Lanzuisi G, Loiacono F, Lemaux BC, Yi SK, Marchesi S, Volonteri M, Hasinger G, Civano F, Gloudemans A, Choi E, Nam Y, Foord A, Onorato S. “Rapid growth in a dual active galactic nucleus during a gas-rich merger at z ≈ 4.5.” Nature Astronomy, published 30 October 2026. doi:10.1038/s41550-026-03008-z. Accepted manuscript: arXiv:2607.19491.
2. Data: JWST GO programme 1760, MAST doi:10.17909/pvsx-yv31; ALMA programme 2024.1.01025.S.
3. Übler H et al. “BlackTHUNDER: Evidence of three massive black holes in a z~5 galaxy.” Preprint, arXiv:2509.21575.
4. Perna M et al. “GA-NIFS: High number of dual active galactic nuclei at z~3.” Astronomy & Astrophysics 696, A59 (2025). arXiv:2310.03067.
5. Coverage read for framing: Live Science (13 August 2026), Science Times (14 August 2026).
6. ESO release eso2208, 12 May 2022: Sagittarius A* is about four million times the Sun’s mass (used for the Milky Way comparison).

Disclosure: This article was drafted with AI assistance and checked against the sources above, including the full accepted manuscript on arXiv. Calculations by SciQuest: cosmic age at z = 4.5 is about 1.34 billion years (lookback about 12.5 billion years) for flat ΛCDM with H₀ = 67.66 km/s/Mpc and Ω_m = 0.311 (Planck 2018), radiation neglected; 1.5 kpc × 3,262 light-years per kpc ≈ 4,900 light-years; 10^46.6 erg/s ≈ 4.0 × 10³⁹ W ≈ 1.0 × 10¹³ solar luminosities (3.828 × 10²⁶ W each); doubling time = 86 million ÷ 6.3 per year ≈ 14 million years. The graphic is a SciQuest original built from the paper’s reported values. Spotted an error? See our corrections policy.