The Andromeda giant swelled to about 860 times the Sun’s radius, then flared 100-fold in 25 days; a Nature study reads this as a 0.7-solar-mass companion spiralling inside it and probably merging with its core.

Peer reviewed · Observational (one event, many telescopes) · Inspiral and merger are model-based interpretations, and a rival study argues the pair survived

A team led by Viraj Karambelkar of Columbia University reports in Nature that AT 2025abao, a giant about 2.5 million light-years away in Andromeda, brightened slowly from 2018 and erupted in late 2025. They read the slow glow as friction from a smaller star dragging through the giant’s outer gas, and the eruption as its plunge into the core. That matters because this “common envelope” stage helps build close binary stars. The merger itself is inferred, and another team’s models say both stars survived.

What you need to know

  • Finding: A giant of about 7 solar masses brightened gradually for about seven years and then erupted as a luminous red nova, an outburst linked to two stars merging. The team estimates its companion at about 0.7 solar masses.
  • Method: Archival images back to 1989, years of survey data from ZTF, WINTER and NEOWISE, two chance SPHEREx scans, ground-based spectra and a James Webb Space Telescope observation 47 days after the optical peak, compared with energy and stellar models.
  • Source: Karambelkar and 44 co-authors, “An inspiral of a star inside a common envelope with its giant companion”, Nature 659, 628–633, published 18 November 2026 (accepted version on arXiv).
  • Why it matters: Theory has long said companions spiral inward through a shared envelope of gas. The authors call this the first direct observation of that drag releasing energy, and they use it to measure how efficiently the energy can throw gas off.
  • Limitation: It is one event, and the companion is never seen directly. Its mass, the inspiral and the merger all come from fitting models to the light, and dust now hides whatever remains.

Evidence at a glance

Evidence summary for the AT 2025abao common-envelope study
ItemAssessment
Evidence statusPeer reviewed (Nature). Observational: one transient watched across ultraviolet, visible and infrared light. The inspiral, the companion’s mass and the merger are interpretations from models.
Primary sourceKarambelkar, V., De, K., Temmink, K. D., Metzger, B. D. et al., “An inspiral of a star inside a common envelope with its giant companion”, Nature 659, 628–633, 18 November 2026. Journal page paywalled; accepted manuscript read in full on arXiv (2611.23470).
Evidence strengthStrong for what was seen, because decades of data from many independent instruments trace the brightening, fading and eruption. Moderate for the slow inspiral, because the authors’ drag estimate matches the extra light but rests on assumed star sizes and masses. Limited for the merger, because the authors call it a suggestion still to be tested, and a separate accepted study models a surviving pair.
Principal limitationThe companion has never been seen. A different model by Zhuo Chen and colleagues fits the same event with a 1.4-solar-mass companion that survives in a tight, stretched orbit.
Coverage at a glance1 independent newsroom (Science News) as of 19 November 2026, 09:45 UTC, found by SciQuest web searches for “AT 2025abao” and for Andromeda star merger news. Counted separately: the NASA release and its Phys.org reprint. Excluded: rewrite and aggregator sites. Coverage is not evidence that a finding is correct.

Sources: Karambelkar et al. 2026 (full accepted manuscript); Chen et al., arXiv:2610.05769 (accepted by ApJ Letters); NASA release of 18 November 2026; SciQuest coverage search at the time stated.

Is this a big deal?

Yes, as an observation. Astronomers have caught stars merging before, but those cases involved smaller, more compact stars and long, slow warm-ups that hid what was happening inside. Here the main star is a bloated giant of the kind used in the textbook theory of how close binaries form, and it was watched for years before, during and after the outburst. The final step, a merger rather than a surviving pair, is the least settled part, and it is the part the press release states most firmly.

What scientists found

The star sits in the outskirts of the Andromeda galaxy, also called M31. Archival photographic plates from 1989 and 1993 and later survey images show it holding steady until 2018. The team’s model of that quiet star gives about 7 solar masses, a surface near 3,650 K, about 18,000 times the Sun’s light and a radius of about 350 times the Sun’s. It was an early asymptotic giant branch star: a star that has used up the hydrogen in its core and swollen to enormous size.

From 2018 to early 2025 its visible light rose by two to three times, in two separate episodes, each building over roughly 1,000 to 1,500 days, while its surface temperature hardly moved from about 3,800 K. Then, for about 300 days, the visible light halved while the infrared kept climbing by five times or more. In July 2025 NASA’s SPHEREx telescope happened to scan the star and caught it at about 860 times the Sun’s radius and some 100,000 times its light.

The eruption began on 19 September 2025 in Zwicky Transient Facility images. In 25 days the star grew about 100 times brighter and its surface heated to about 8,000 K. Observer Koichi Itagaki first noticed the outburst on 17 October. Visible light peaked on 7 November, and the total output kept rising through a plateau of about 80 days, reaching about 1.5 million times the Sun’s light. The spectra match a luminous red nova, ending with titanium oxide forming in the cooling gas.

The team’s reading of the whole sequence: a companion of about 0.7 solar masses, a little lighter than the Sun, spent the seven years spiralling through the giant’s outer layers, then plunged to the helium core within fewer than ten orbits and most likely merged with it. That blast flung out about 10% of the giant’s envelope at roughly ten times the speed needed to escape the star.

How they found it

The star drew attention before it erupted. Karambelkar’s group spotted its slow change in infrared surveys and reported it in April 2025, months before the outburst, in a paper later published in The Astrophysical Journal. That paper weighed two ideas: a companion beginning a common-envelope phase, or a burst of nuclear burning known as a thermal pulse.

Once the eruption began, the team collected spectra from October 2025 to February 2026 and asked for time on the James Webb Space Telescope. Webb’s mid-infrared instrument observed on 24 December 2025, 47 days after the optical peak, while SPHEREx made a second pass. Karambelkar said in NASA’s release that “Stellar mergers are very, very difficult to observe in real time” because they are hard to predict.

Seven years of warning: how the giant star behind AT 2025abao changed Measured changes in Andromeda, with the inspiral and merger readings proposed by Karambelkar et al. (Nature, 2026) Feb 2023: NEOWISE flags it Sep 2024: WINTER sees brightening Jul 2025: SPHEREx, radius ~860 Suns 19 Sep 2025: eruption begins 17 Oct: first spotted 7 Nov: optical peak 24 Dec: Webb Steady Slow brightening (about 7 years) Optical fade Eruption // 20172018202020222024 Dec 2024May 2025Oct 2025Feb 2026 Visible light rises 2 to 3 times, in two episodes; surface stays near 3,800 K. Proposed: a ~0.7-solar-mass companion drags through the giant’s outer layers. Fewer than 10 orbits. Quiet in archival images from 1989 to 2018. ~300 days: visible light halves, infrared climbs 5 times or more. Eruption: ~100 times brighter in 25 days, ~8,000 K, then an ~80-day plateau. Proposed: the companion plunges into the core; about 10% of the envelope is thrown off. How big was the giant? Its radius if it sat where the Sun is (1 AU = Earth–Sun distance) Earth’s orbit, 1 AU Mars, 1.52 AU Jupiter’s orbit, 5.20 AU Sun’s centre Before (quiet star): ~350 Sun radii = 1.63 AU, past Mars’ orbit July 2025 (SPHEREx): ~860 Sun radii = 4.0 AU, inside Jupiter’s orbit Sources: Karambelkar et al., Nature 659, 628–633 (2026), arXiv:2611.23470; NASA Webb release, 18 Nov 2026; NASA fact sheets (Sun radius 695,700 km; 1 AU = 149.6 million km). Time axis is zoomed after the // break. Radius conversions are SciQuest calculations. SciQuest original graphic.
The measured changes in AT 2025abao from 2017 to early 2026, with the time axis zoomed after the break, and the star’s radius before and during its swelling compared with planetary orbits if it sat where the Sun is. The inspiral and merger labels are the authors’ interpretation, not direct observations. Data from Karambelkar et al. 2026, Nature; radius conversions use NASA fact-sheet values. SciQuest original graphic.

To explain the slow brightening, the team compared two energy sources. Gas leaking from the pair, the usual explanation in earlier red novae, falls short here by about 100 times in the authors’ estimate, because the giant’s weak gravity gives escaping gas little energy. Drag on a 0.7-solar-mass companion grazing the giant’s surface, worked out with a standard formula for a body moving through gas, gives about 3 × 1038 erg per second. That matches the extra light the star gave off, which was about 1038 erg per second.

The closest everyday comparison is a satellite in a low orbit. Thin air at the top of Earth’s atmosphere drags on it, its orbit shrinks over years, and the final fall is fast. (SciQuest covered one such case, the re-entry of NASA’s Swift observatory, which also observed this eruption.) The comparison has limits: here the “air” is the giant’s own outer gas, and the companion is itself a star whose lost orbital energy heats and puffs up the giant.

Show the maths: how many orbits fit in seven years?

A radius of 350 Suns is 350 × 695,700 km, about 243 million km or 1.63 AU, where 1 AU (149.6 million km) is the Earth–Sun distance. That is farther out than Mars’ orbit at 1.52 AU. Kepler’s third law, with period in years, distance in AU and mass in solar masses, gives P = √(a³ ÷ M) = √(1.63³ ÷ 7.7) ≈ 0.75 years, or about 270 days, for a companion skimming the surface. The paper gives about 280 days. Seven years is about 2,560 days, so 2,560 ÷ 280 ≈ 9 orbits, which is why the authors say fewer than ten. The 2.5 million light-year distance comes from the paper’s 0.77 megaparsecs (1 megaparsec ≈ 3.26 million light-years), so the light seen in 2025 left the star about 2.5 million years ago.

Why it matters to you

Nothing here changes life on Earth. The Sun has no companion star and will not do this. The interest is in where the universe’s close pairs come from. Common-envelope episodes are thought to produce tight binaries that later become pairs of white dwarfs, sources of the gravitational waves that the planned Laser Interferometer Space Antenna is designed to detect, and the systems behind some supernovae.

Population models of those objects rely on one number, an efficiency usually written α, for how much of the lost orbital energy goes into throwing off the shared envelope. From this single event the authors bound it between about 0.01 and 0.4, in line with values inferred from surviving binaries. They also find that most of the energy in the plunge went into flinging a small amount of gas very fast, which the standard formula does not capture.

What the evidence does not show

It does not show the companion. Its 0.7-solar-mass estimate comes from two model comparisons, the dust around the quiet star and the mass shed during the slow brightening, which agree on a mass ratio of about 0.1 but carry wide error bars.

It does not prove a merger. The paper says the authors “suggest” one, based on the energy of the ejected gas, the small amount of envelope lost and the eruption’s likeness to known mergers. They list tests that could still rule out a tight surviving pair: how the dust shell expands, whether the remnant stays unusually bright for decades, and pulsations in a surviving envelope. NASA’s release goes further and says Webb’s data “proved the stars merged,” which is firmer than the paper.

It also does not settle the masses. The star’s size, mass and brightness come from fitting models to its light, and the efficiency range rests on a computer model of the giant’s internal structure.

How this fits what we already knew

Luminous red novae have been tied to stellar mergers since V1309 Scorpii, a contact binary in our own galaxy whose merger was traced in detail. AT 2025abao is the fourth such event found in Andromeda, according to a study by Andrea Reguitti and colleagues in Astronomy & Astrophysics, which measured its peak at magnitude 15.1 in the g band.

The difference is the main star. Known red nova progenitors were stars still burning hydrogen in their cores or just past that stage, with tightly bound outer layers, and their precursors brightened by several magnitudes over hundreds of orbits. AT 2025abao’s giant has a loose, churning envelope and brightened by only about 1 magnitude over fewer than ten orbits. For background on how stars reach this swollen stage, see the stages of a star’s life cycle.

What other astronomers have found

SciQuest has not yet obtained comment from an astronomer outside the team, and Science News quoted only the lead author. The clearest outside view is in the literature. Zhuo Chen of Tsinghua University and colleagues, in a paper accepted by The Astrophysical Journal Letters and posted in October 2026, modelled the same event and reached a different ending. In their models a companion of about 1.4 solar masses drifts in through the giant’s wind for about eight years, then plunges and launches a shock, and the companion and the giant’s core survive in an eccentric orbit at least 400 days later. They note that their one-dimensional models lack some effects seen in three-dimensional simulations.

Reguitti’s team, writing before both studies, classed the eruption as a canonical luminous red nova and proposed that the extent of a common envelope explains why some red novae show two peaks and others a long plateau.

How other headlines framed it

  • Science News led on two stars merging in front of astronomers, using the nicknames Big Bao and Little Bao, and noted lower down that dust hides the remnant and that confirmation may take years.
  • NASA’s release said Webb data clinched the merger and called the giant a supergiant. The paper describes an asymptotic giant branch star and presents the merger as its suggested outcome.
  • Phys.org carried the NASA text under a headline saying Webb confirmed the merger, which is stronger than the paper.

Science News briefly raised the possibility that the stars survived in a tighter orbit before setting it aside. None of the three mentioned the published rival model by Chen’s team.

What happens next

The star went behind the Sun in mid-February 2026, so the next step is watching it as it returns. The authors want late-time infrared measurements of the dust shell, long-term monitoring of the line profiles and checks for pulsations. Each could separate a merged star from a tight surviving pair. Beyond this event, they expect the Vera C. Rubin Observatory and the Nancy Grace Roman Space Telescope to find similar cases in about 100 galaxies like the Milky Way, out to about 15 megaparsecs (roughly 50 million light-years). For a sense of how far Andromeda itself is, see how many light-years away the Andromeda galaxy is.

In one sentence

Astronomers watched a giant star in Andromeda glow for seven years and then erupt, and they read it as a smaller star spiralling through the giant’s outer gas and probably merging with its core, a final step that a rival model disputes and future observations can test.

Quick check: did it click?
  1. What powered the seven-year brightening, according to the Nature team? Friction, or drag, on a companion star moving through the giant’s outer layers, turning orbital energy into light.
  2. Has anyone seen the companion star? No. Its mass of about 0.7 solar masses is a model estimate, and another team estimates about 1.4.
  3. Is the merger confirmed? No. The paper suggests a merger and lists tests; a rival accepted study models a surviving pair.

Read next

Sources and disclosure

  1. Karambelkar, V., De, K., Temmink, K. D., Metzger, B. D., Blagorodnova, N. et al., An inspiral of a star inside a common envelope with its giant companion, Nature 659, 628–633, 18 November 2026. Accepted manuscript: arXiv:2611.23470.
  2. Karambelkar, V. R. et al., The slow brightening of WNTR23bzdiq/WTP19aalzlk: possible onset of common envelope evolution in an asymptotic giant branch star?, The Astrophysical Journal 993, 109 (2025).
  3. Reguitti, A. et al., AT 2025abao: the fourth luminous red nova in M 31, Astronomy & Astrophysics 709, A25 (2026), open access.
  4. Chen, Z., Cui, Y., Wu, T., Cai, Y. et al., Unravelling the nature of AT 2025abao: from precursor to final fate, arXiv:2610.05769 (accepted by The Astrophysical Journal Letters).
  5. NASA Webb Mission Team, NASA Data Clinches First-of-Its-Kind Stellar Merger, NASA, 18 November 2026.
  6. NASA Space Science Data Coordinated Archive, Sun Fact Sheet (radius 695,700 km) and Planetary Fact Sheet (Mars 228.0, Jupiter 778.5 million km from the Sun).
  7. Bartels, M., Two stars in the Andromeda galaxy merged before astronomers’ eyes, Science News, November 2026.

Disclosure: This article was drafted with AI assistance and checked against the sources above. The conversions to AU, the orbital period check, the orbit count and the light-year distance are SciQuest calculations from the cited values; the graphic is a SciQuest original drawn from values in the Nature paper and NASA fact sheets. The featured image is credited in its caption. Spotted an error? See our corrections policy.


Discover more from SciQuest

Subscribe to get the latest posts sent to your email.