If the record for the largest star comes from a distance nobody has measured, how much is that record worth?
Welcome to FreeAstroScience. We went back to the 2012 paper behind the famous 2,150-solar-radii figure and found that the paper never publishes a radius at all.
Ask which is the largest star in the universe and almost every list answers Stephenson 2-18, a dust-wrapped red supergiant in Scutum credited with a radius 2,150 times the Sun’s. The arithmetic producing that figure is sound. Its input is not. Radius scales in direct proportion to assumed distance, and the distance to Stephenson 2-18 has never been measured. It was adopted, from a cluster the star may not even belong to.
Where 2,150 actually comes from
Everything traces to one table. Fok and colleagues modeled the infrared spectral energy distributions of red supergiants across six massive clusters using the DUSTY radiative-transfer code, and published in The Astrophysical Journal on 5 November 2012. Their Table 8 lists, for the source they label St2-18, an effective temperature of 3,200 K and a bolometric luminosity of log L = 5.64, which is 436,500 Suns. Read along the row and you get dust temperature, optical depth at 2.2 microns, mass-loss rate and luminosity. No radius column exists anywhere on the page.
What the world quotes as a measurement is what those two numbers produce when somebody pushes them through the Stefan-Boltzmann relation:
R / RSun = √(L / LSun) × (TSun / T)2
Here R is the star’s radius, L its bolometric luminosity, T its effective temperature, and the Sun subscript marks our own star’s value of each quantity. Said without symbols: size follows the square root of brightness divided by the square of surface temperature. That single relation is why the giants and supergiants sit in their own bands on the chart, as our guide to reading the Hertzsprung-Russell diagram lays out. Feed in 436,500 Suns, 3,200 K for Stephenson 2-18 and 5,772 K for our own star, and out drops 2,149.6. We ran it ourselves before writing this. The four digits match, so the algebra is not where the trouble lives.
Hold onto what 2,150 solar radii would mean physically. That radius comes to 1.50 billion kilometers, or 10.0 astronomical units. Saturn orbits at 9.58. Light needs 2 hours and 46 minutes to cross it from limb to limb, against 4.6 seconds for the Sun.
The distance is doing all the work
Luminosity is not something an instrument reads off a dial. A telescope records flux, the energy landing on each square meter at Earth every second, and turning flux into luminosity requires knowing how far the light traveled:
L = 4πd2F
Distance is d, measured flux is F, and 4πd2 gives the area of the sphere that light has spread itself across by the time any of it reaches us. Luminosity climbs with the square of distance. Radius climbs with the square root of luminosity. Put the two together and the squares cancel, leaving radius proportional to distance, one for one. Move Stephenson 2-18 ten percent farther out and it becomes ten percent wider without anyone observing anything new.
Which distance, then? Fok and colleagues adopted 5.5 kiloparsecs for the Stephenson 2 cluster. In the same table where that choice sits, their own kinematic distances, calculated from two different models of Galactic rotation, come out at 5.2 and 5.1 kiloparsecs. Run 5.1 through the chain and the radius drops to 1,993 solar radii, under the two-thousand line that makes the whole thing a headline. We cannot tell from the text why the rounder, bigger number was carried forward, and the paper never says.
Ben Davies, Don Figer and Rolf-Peter Kudritzki, who found the cluster’s core of 26 associated red supergiants in 2007, placed it at 5.8 kiloparsecs with an asymmetric error bar of +1.91 and −0.76. Nobody disputes that range. It is simply enormous, and every kiloparsec of it lands straight on the radius.
Table 1 — Stephenson 2-18’s radius recomputed at each published distance for the Stephenson 2 cluster (Davies et al. 2007; Fok et al. 2012)
| Cluster distance | Where the value comes from | Implied radius |
|---|---|---|
| 5.04 kpc | Davies et al., lower bound | 1,970 Suns |
| 5.1 kpc | Fok et al., Reid rotation model | 1,993 Suns |
| 5.5 kpc | Fok et al., adopted value | 2,150 Suns |
| 5.8 kpc | Davies et al., best estimate | 2,267 Suns |
| 7.71 kpc | Davies et al., upper bound | 3,013 Suns |
One paper, one flux, and a radius that swings by a factor of 1.5 across the range the cluster’s own discovery paper permits. Every row above rests on a legitimate published number.
Membership is the shakiest link
Ignacio Negueruela and his group went looking for red supergiants around the cluster rather than inside it. Cross-matching the DENIS, USNO-B1 and 2MASS catalogs produced more than 600 candidates, and they took intermediate-resolution spectra of roughly 250 with the William Herschel Telescope on La Palma.
The velocities refuse to line up
Their 2012 survey turned up about 35 red supergiants whose radial velocities match Stephenson 2 members, plus about 40 more with velocities that rule out any physical association. Several populations of massive stars lie stacked along this one sight line at different distances, and the cluster is a knot inside a far larger structure rather than an island. Sitting in the right direction on the sky makes a star a neighbor on a photograph, and nothing more.
A year later the same group took multi-epoch spectra around the calcium triplet for more than 30 of those supergiants. Most clustered tightly at spectral types M0 to M2, while the brightest objects in the near infrared stood apart, showing very late classes and heavy mass loss. Stephenson 2-18 is one of the outliers.
What that leaves us holding
Take away the cluster assumption and there is no distance to Stephenson 2-18 at all. None.
Siebert and colleagues hit the same wall in August 2025, working with ALMA on a different Stephenson 2 supergiant. They gave the problem an appendix of its own, argued that their star probably does belong to the cluster, and then closed on this: “a more accurate distance measurement is needed to confirm this.” A research team, writing in Astronomy & Astrophysics last summer, telling readers plainly that distances in this cluster remain unconfirmed.
Stars are not supposed to get this big
Evolutionary models put a ceiling on how far a red supergiant envelope can inflate before it turns unstable and gets thrown off, the same runaway swelling that will one day drive our own Sun through its red giant phase on a far smaller scale. The number usually quoted for that ceiling is around 1,500 solar radii. Stephenson 2-18 sits 43 percent above it.
Honesty about that 1,500 is owed here. We could not trace it to a single primary calculation in anything we read for this piece, and it circulates as a rule of thumb drawn from evolutionary models rather than as a measured limit. Reporting it that way beats dressing it up with a citation it does not have.
Run the logic backward and the mismatch becomes a statement about distance. For the same temperature and the same measured flux to yield 1,500 solar radii, Stephenson 2-18 would have to sit at 3.84 kiloparsecs, about 12,500 light-years, which undercuts every published estimate, the low end of the Davies range included. Three exits present themselves, and no observation currently picks between them. Stephenson 2-18 may lie much closer than anyone assumes. Perhaps it was never a supergiant. Or the ceiling drawn by the models sits in the wrong place.
We are leaving the pulsation-instability calculations out. They set that ceiling, they deserve an article of their own, and sketching them here would not alter the shape of the problem: an observed number and a predicted number disagreeing by nearly half.
The measurement that would settle it
A few approaches could put a real number on this distance, and one of them is already halfway built.
- Maser astrometry. Stephenson 2-18 emits SiO and water masers, and very-long-baseline interferometry can pin a maser spot down to tens of microarcseconds, which is a parallax arriving by a side door.
- Resolve the disk directly.
- A verdict on membership. Negueruela’s team has sorted dozens of neighboring supergiants by radial velocity, and settling this one star either way decides whether the cluster distance may be applied to it.
That middle item is less far-fetched than it sounds. On 21 November 2024, Keiichi Ohnaka and five co-authors published GRAVITY observations from ESO’s Very Large Telescope Interferometer that resolved WOH G64, a dust-enshrouded red supergiant in the Large Magellanic Cloud, producing the first interferometric image of such a star outside the Milky Way. WOH G64 lies roughly nine times farther away than Stephenson 2 is believed to be. Dust rather than distance is the obstacle in Scutum: the cluster carries an average infrared extinction of 1.40 magnitudes in Fok’s own table, and the visual extinction along that sight line is heavier still.
Our verdict on the largest star in the universe
Stephenson 2-18 is a genuine, heavily obscured red supergiant, and about that much we are confident. Calling it the largest star in the universe without qualification is where the reporting outruns the evidence. Blame does not belong to Fok and colleagues, who published a temperature and a luminosity and drew no such conclusion themselves. The claim got assembled afterward, out of their table plus a distance they had already flagged as provisional.
So our position is this: the honest ranking today makes Stephenson 2-18 the largest candidate, not the largest star, and that difference is not pedantry. UY Scuti wore this crown until a Gaia parallax cut its radius from roughly 1,700 solar radii to about 909, a reversal we covered in our profile of the star that lost the title. One parallax did that. A maser parallax could do the same in Scutum.
We would carry four things away from all this.
- Distance drives everything.
- Fok et al. (2012) published a temperature and a luminosity for this star, and no radius at all.
- Radius tracks distance one for one, and the published cluster distance spans 5.04 to 7.71 kiloparsecs.
- Membership in Stephenson 2 has never been confirmed for Stephenson 2-18.
Keep asking what a number is made of
A record is a poor way to learn astronomy and an excellent way to get somebody started. The habit worth keeping is the one that walks a headline figure back to the measurement underneath, then asks what got assumed along the route. Do that here and the leaderboard stops mattering, because a 43 percent standoff between theory and observation opens a better question than any ranking closes. We will revise this page the day anyone publishes a parallax for Stephenson 2-18. Until then, ask what your numbers are made of. Gerd Dani, for FreeAstroScience.
Sources
- Fok, T. K. T., Nakashima, J., Yung, B. H. K., Hsia, C.-H., & Deguchi, S. (2012). Maser Observations of Westerlund 1 and Comprehensive Considerations on Maser Properties of Red Supergiants Associated with Massive Clusters. The Astrophysical Journal, 760(1), 65. Published 5 November 2012. https://arxiv.org/abs/1209.6427
- Davies, B., Figer, D. F., Kudritzki, R.-P., et al. (2007). A Massive Cluster of Red Supergiants at the Base of the Scutum-Crux Arm. The Astrophysical Journal, 671, 781. https://arxiv.org/abs/0708.0821
- Negueruela, I., et al. (2012). Red supergiants around the obscured open cluster Stephenson 2. Astronomy & Astrophysics, 547, A15. https://arxiv.org/abs/1208.3282
- Negueruela, I., et al. (2013). The population of M-type supergiants in the starburst cluster Stephenson 2. Astronomy & Astrophysics. https://arxiv.org/abs/1303.1837
- Siebert, M. A., De Beck, E., Quintana-Lacaci, G., & Vlemmings, W. H. T. (2025). Stephenson 2 DFK 52: Discovery of an exotic red supergiant in the massive stellar cluster RSGC2. Astronomy & Astrophysics, 700, L11. Published 6 August 2025. https://doi.org/10.1051/0004-6361/202555975
- Ohnaka, K., Hofmann, K.-H., Weigelt, G., van Loon, J. Th., Schertl, D., & Goldman, S. R. (2024). Imaging the innermost circumstellar environment of the red supergiant WOH G64 in the Large Magellanic Cloud. Astronomy & Astrophysics, 691, L15. Published 21 November 2024. https://doi.org/10.1051/0004-6361/202451820
