Astronomers From Gauhati University and the Indian Institute of Astrophysics Report From Gaia Data That the Star Cluster UBC 63 Is Two Clusters, 21 and 562 Million Years Old, Moving Apart After a Close Flyby About 6 Million Years Ago

A group of stars about 4,892 light years from Earth, listed in star catalogues as a single cluster named UBC 63, is two clusters, according to a team led from India. The Press Information Bureau (PIB) announced the result on 5 November 2026. The younger cluster, UBC 63A, is about 21 million years old. The older, UBC 63B, is about 562 million years old, and the two passed within about 7 parsecs (about 23 light years) of each other roughly 6 million years ago.

Astronomers from Gauhati University in Guwahati and the Indian Institute of Astrophysics (IIA) in Bengaluru led the work. The IIA is an autonomous institute of the Department of Science and Technology (DST), was founded by Vainu Bappu, runs the Kodaikanal Solar Observatory, and was one of the institutes behind the ultraviolet telescope on AstroSat. The first author, Samrat Biswas, is a DST INSPIRE Senior Research Fellow at Gauhati University. His co-authors include Biman J. Medhi and A. H. Sheikh of Gauhati University, Sukanta Deb of Cotton University, H. S. Das of Assam University, Maheswar Gopinathan and Ram Sagar of the IIA, and colleagues in Italy, China and Japan. The paper appeared in The Astronomical Journal (volume 172, number 3) on 13 August 2026, open access, after a preprint went on the arXiv server on 20 July.

What the team found

UBC 63 is an open cluster, which the PIB release describes as a poorly populated group of stars held loosely together by gravity. Britannica defines an open cluster as a group of young stars bound by mutual gravitation. The PIB says UBC 63 had been largely overlooked, with earlier papers giving only preliminary numbers that treated it as one system. The Biswas paper says earlier studies put its age between 21.3 and 60.3 million years.

The Indian team split it into 246 likely member stars: 98 in UBC 63A and 148 in UBC 63B. The paper gives the ages as 21 million years (plus or minus 4 million) and 562 million years (plus or minus 43 million). The gap of 541 million years, plus or minus 43 million, is too large for the two to have formed together. The paper’s tables give UBC 63A a mass of about 294 Suns (plus or minus 58) and UBC 63B about 435 (plus or minus 87).

The PIB release credits machine learning and dynamical simulations run on high-performance computers. In the paper, the machine-learning step is a statistical tool called a Gaussian mixture model, and the simulations are orbit calculations plus a direct N-body run, in which the gravitational pull between the stars is followed step by step.

How two clusters hid inside one

The first clue was in the colours. When the team plotted the 246 stars by brightness and colour, they fell along two separate tracks, where a single cluster would give one. The authors compare the pattern to a known old pair, NGC 1605a and NGC 1605b. That plot had not yet been corrected for uneven dust between the stars and Earth, so the team repeated it star by star with dust corrections from published extinction values and fitted theoretical stellar-evolution tracks (MIST models) to each group.

The second check used motion. A Gaussian mixture model run on each star’s position on the sky, parallax and two proper-motion values (explained below) tested one to five groups, and the statistical score had a clear minimum at two. A second, parallel run with stricter membership cut-offs gave consistent results, the paper says. UBC 63A overlaps almost completely with the stars earlier catalogues had assigned to UBC 63. UBC 63B is the new component, with a slightly different parallax and proper motion.

The third check came from the stars themselves. UBC 63B contains two red clump stars, evolved stars that burn helium in their cores, and a probable blue straggler. The authors say the red clump stars independently support an older age for UBC 63B. Across 100 simulated re-runs that varied the measurements within their errors, the two stars had mean membership probabilities of 0.62 and 0.59 for UBC 63B, above the 0.5 threshold the team used. Radial velocities, measured from the Doppler shift in starlight along the line of sight, exist for 9 stars in UBC 63A and 18 in UBC 63B. They differ by 3.46 plus or minus 1.70 kilometres per second.

What Gaia supplies

The measurements come from Gaia, the European Space Agency (ESA) mission that observed from 27 July 2014 to 15 January 2025. ESA says it made more than three trillion observations of two billion stars and other objects. Gaia Data Release 3, published on 13 June 2022 and the release the Indian team used, includes full position, parallax and proper-motion solutions for around 1.46 billion sources, according to the Gaia archive.

ESA explains the two quantities that matter here. Parallax is the small apparent shift of a nearby star against the distant background as the observer’s position changes, and it gives the star’s distance. Proper motion is the star’s real change of position across the sky. Because parallax repeats every year and proper motion is steady, a few years of monitoring let Gaia scientists separate the two for each star. Combined with radial velocity, they give each star’s motion through space.

Those measurements are why the pair could be pulled apart. In the paper’s table, UBC 63A is at about 1,502 parsecs and UBC 63B at about 1,527 parsecs, and their proper motions differ slightly. A census by Emily Hunt and Sabine Reffert that searched Gaia DR3 recovered 7,167 clusters, of which 2,387 were candidate new objects.

The flyby, run backwards and forwards

To find out how two clusters of different ages came to be neighbours, the team ran their orbits through the Milky Way’s gravity, back in time and forward. Both are on near-circular paths, with eccentricities of about 0.08 and 0.07, and have similar angular momentum. When UBC 63A formed 21 million years ago, the clusters were about 60 parsecs apart, a figure the PIB and the paper both give (the paper’s uncertainty is plus or minus 29). The paper adds that this is comparable to the size of a giant molecular cloud, hinting that both may have come from the same cloud complex.

The closest approach was about 7 parsecs (plus or minus 2), some 6 million years ago. Today the 3D separation is 26 parsecs (plus or minus 8), which the PIB converts to about 85 light years. The relative speed is 3.60 plus or minus 1.80 kilometres per second, well above the 0.51 kilometres per second the pair would need to stay bound. In 10,000 random draws within the uncertainties, the paper finds a probability of about 98 per cent that the clusters are unbound. They are moving apart, and the paper’s forward calculation puts them about 491 parsecs apart (plus or minus 213) after another 100 million years.

The authors also report low tidal-factor values and candidate member stars lying beyond each cluster’s Jacobi radius, the limit set by tidal forces. They read these as signs of a brief, strong tidal interaction during the flyby, and they say a detailed study of those stray stars is needed to separate the flyby’s effect from the Milky Way’s.

Why a cluster pair matters

The PIB release says encounters between star clusters are relatively uncommon, so finding a pair soon after one gives a view of how gravity reshapes clusters. The paper cites earlier estimates that binary clusters make up 8 to 12 per cent of open clusters and lists two other flyby pairs found recently, ASCC 71 with ESO 064-65 and NGC 2129 with UBC 437. The authors call UBC 63 a rare, real-time laboratory for studying tidal effects between clusters near the Sun.

How firmly it is established

The PIB release says the analysis confirmed two distinct populations. The paper’s abstract describes UBC 63 as “a compelling candidate for a double cluster undergoing an unbound flyby interaction”, and its conclusion says the pair is best interpreted as an unbound, tidally interacting flyby system. It also lists its limits. The orbit calculation treats each cluster as a point mass, the backward N-body run starts from a simplified set-up, and the first colour plot lacked the dust correction. The N-body run gives separations of about 7.4 parsecs at closest approach, 25 today and 452 in 100 million years, which the authors say agree within error limits with the orbit calculation.