A star seven times the Sun’s mass took seven years to swallow its smaller companion, ending in an eruption on Oct. 17, 2025, that astronomers in Andromeda’s outskirts had tracked from the beginning.
The event, cataloged as AT 2025abao, is the first time the slow inspiral of one star inside another’s shared outer atmosphere has been observed directly rather than reconstructed after the fact. A study led by Viraj Karambelkar of Columbia, published in Nature with participation from the Instituto de Astrofísica de Canarias, lays out the full sequence, according to Gomera Verde.
Until now, astronomers had worked the other way around. As the sources put it, they had been doing reverse engineering on the corpses of events that happened hundreds or millions of years ago. With this system, the data came in while the stars were still dying.
The payoff is a measurement other researchers have wanted for decades: how efficient the “common envelope” phase actually is — the process that produces stars of neutron, black holes, and gravitational waves.
Two stars, seven orbits, one survivor
The system was an ordinary binary for millennia: a giant of roughly seven solar masses orbiting with a companion of about 0.7 solar masses. Archive records going back to 1989 show the giant held steady for nearly three decades.

In 2018, the smaller star touched the giant’s outer layers. Friction did the rest. Over the next seven years — about seven orbits, the NASA note stated — the companion spiraled inward, shedding orbital energy while the giant swelled to 2.5 times its original size, per the models built by the research team.
For most of that descent, the process was strangely gentle. More than 90 percent of the energy generated by the friction was transported to the surface and radiated to space in the form of light, without ejecting the star’s atmosphere. Then, in the final 300 days, the surface temperature dropped and the star expanded rapidly — the sign that equilibrium had broken and the companion was in free fall toward the helium core.
The eruption that broke the models
What came next was violent. In October 2025 the system flared in the style of a luminous red nova. In 25 days, its optical brightness multiplied a hundredfold and its surface temperature jumped from 3,500 to 8,000 kelvin, with a shock wave expanding at about 330 kilometers per second.
The energy accounting changed too. During the final plunge, the companion released a enormous amount of orbital energy. The eruption expelled between 0.1 and 0.2 solar masses of material.
Because the data confirmed that the stars fused, the team concluded both stars fused into a single object. Spectra also showed the ejecta slamming into a pre-existing disk of gas and dust, carving a bipolar structure resembling those seen in many planetary nebulas.
The survivor got a name: Bao, possibly rejuvenated by the merger. What Bao is today — a hot star, a stripped giant, something in between — remains an open question. And the merger itself still needs future observations to confirm in a conclusive way that the two stars fused.
How the alert chain came together
The catch began with a statistical oddity, not a flash. In February 2023, the team spotted a variable star in data from the retired NEOWISE mission that matched nothing known; before 2018 its brightness had been constant.

From there, the instruments stacked up. In September 2024, the Wide-field Infrared Transient Explorer (WINTER) showed the object was not only brightening but expelling dust. In July 2025, the SPHEREx satellite caught it dimming in visible light and brightening in infrared during a routine survey. On Oct. 17, 2025, the Zwicky Transient Facility caught the eruption itself.
The James Webb Space Telescope then closed the case. It observed Bao on Dec. 24, 2025 — just 47 days after the eruption peaked, which Karambelkar called the earliest follow-up ever done for an event of this kind.
“Se trata del seguimiento más temprano jamás realizado de un evento de este tipo.”
Karambelkar, of Columbia, said the data allowed the team to characterize the system’s state before, during and immediately after the merger.
“lo cual es, sencillamente, asombroso.”
The dates, laid out, show how long the patient watching lasted:
- 1989 — first archive data show the giant stable.
- 2018 — the companion enters the giant’s outer layers; brightness begins rising.
- February 2023 — NEOWISE data reveal an unclassifiable variable star.
- September 2024 — WINTER detects dust ejection.
- July 2025 — SPHEREx sees the object shift from visible to infrared brightness.
- Oct. 17, 2025 — Zwicky Transient Facility catches the eruption.
- Dec. 24, 2025 — Webb observes Bao, 47 days after peak.
- Nov. 18, 2026 — the Nature paper publishes.
Why the timing matters for what comes next
The discovery landed at an exceptional moment.
The stakes reach beyond one system. Common envelope evolution is the process that produces neutron stars, black holes, and gravitational waves. Measuring how the process actually behaves, rather than modeling it backward, calibrates predictions for the populations LISA and similar missions will detect.
Spanish institutions had a hand in the work: the Instituto de Astrofísica de Canarias, the Instituto de Ciencias del Cosmos de la Universidad de Barcelona, and the Instituto de Estudios Espaciales de Cataluña all participated. The Canary Island telescopes at Roque de los Muchachos, including the GTC and the NOT, are fundamental for following transient events of this kind.
The study, “An inspiral of a star inside a common envelope with its giant companion,” appears in Nature, with backing detailed in NASA’s announcement of the findings.