OrbitalHub

The two brightest objects a wide-field optical survey can pick out at the cosmic dawn are not always what they appear to be. In a paper posted on 18 August 2026, a team led by Daming Yang of Leiden University reports that two sources originally classified as faint quasars in the Subaru High-z Exploration of Low-Luminosity Quasars (SHELLQs) catalogue are instead extremely UV-luminous galaxies, and that the only stellar populations able to reproduce their spectra contain stars heavier than about 225 solar masses each Yang et al. 2026 preprint on arXiv. The finding lands in a regime where galaxy and quasar brightness functions overlap, and the result means that source classifications, and the inferred demographics of both galaxies and quasars in this crossover regime, need revisiting.

The confusion zone sits at the extreme bright end of the ultraviolet galaxy luminosity function during the epoch of reionisation. Galaxies and quasars both produce blue ultraviolet continuum, and at absolute magnitudes around M_UV ≈ -23.5, the faintest quasars and the brightest starbursts are visually similar in a ground-based photometric catalogue. SHELLQs, a long-running campaign using the 8.2-metre Subaru telescope and the Hyper Suprime-Cam wide-field imager, has spent a decade picking out candidate high-redshift quasars from the Subaru Strategic Program imaging and confirming them spectroscopically Matsuoka et al. 2020 SHELLQs IX paper. Its sixteenth data release added 69 confirmed quasars at redshifts between 5.8 and 7.0 Matsuoka et al. 2021 SHELLQs XVI paper. Two sources in that wider sample, J1450-0144 at redshift 6.627 and J1429-0104 at redshift 6.796, looked like faint quasars in the discovery spectra from GTC/OSIRIS and Subaru/FOCAS, and they were catalogued as such.

JWST changed the picture. Yang and colleagues obtained near-infrared spectroscopy with NIRSpec using the G140H and G235H gratings, which together cover rest-frame wavelengths from about 1200 to 4000 angstroms in these high-redshift targets. The new spectra reveal three spectral signatures that the original discovery spectra could not resolve: narrow nebular emission lines including the [O II] doublet at 3727/3730 angstroms and [Ne III] at 3870/3969 angstroms, which are produced by interstellar gas ionised by hot stars rather than by an accreting black hole; broad He II 1640 emission with rest-frame equivalent widths of 8.8 plus or minus 1.2 angstroms in J1450 and 3.7 plus or minus 1.1 angstroms in J1429, a feature too strong for any AGN-powered continuum at this redshift; and prominent P Cygni profiles in the high-ionisation resonance lines N V 1240, Si IV 1400, and C IV 1550, the spectroscopic fingerprints of fast stellar winds driven by the hottest, most luminous stars.

The follow-up ALMA Band-6 observations add the second piece of evidence. ALMA detected luminous [C II] 158 micrometre emission in both systems, with line luminosities of about 0.8 times 10 to the 9 solar luminosities for J1450 and 4.1 times 10 to the 9 solar luminosities for J1429. J1429 additionally shows bright dust continuum, and the [C II] and dust peaks are offset by about 5.4 kiloparsecs from the rest-frame UV emission seen by NIRSpec and Subaru/HSC. The offset is a structural clue: it shows that J1429 is not a single point-like accretion disk masquerading as a galaxy, but a spatially resolved system with cold gas and dust that have decoupled from the hot stellar light, which is exactly the geometry of an extreme starburst rather than a quasar.

The interpretation hinges on a population-synthesis fit. The authors tested two classes of models against the new spectra. The first, BPASS (Binary Population and Spectral Synthesis) with standard upper-mass cutoffs around 100 solar masses, fails to simultaneously reproduce the broad He II 1640 emission and the strong wind profiles in N V, Si IV, and C IV. The second class, BPASS extended with a dedicated prescription for very massive stars above 100 solar masses, succeeds. Under the VMS models, the equivalent-width diagnostics imply an upper-mass cutoff above 225 solar masses for J1429, while J1450 lies beyond even the 475-solar-mass endpoint of the model grid. The star-formation durations are short: 2 to 4 million years for J1450, with a broader allowed range for J1429. Stellar masses come out at log(M_star/M_sun) between about 9.2 and 9.9, with star-formation rates of roughly 300 to 540 solar masses per year.

The reference spectrum in the model grid comes from R136, the central cluster of the Tarantula Nebula in the Large Magellanic Cloud Crowther et al. 2016 R136 massive-star census. HST/STIS spectroscopy of R136 has established that the cluster hosts individual stars well above the 150-solar-mass conventional upper limit. When the team plots the JWST/NIRSpec spectra of J1450 and J1429 against the stacked spectrum of the seven most massive stars in R136, initial masses between about 100 and 300 solar masses, the wind profiles and He II emission match almost feature for feature. The reionisation-era galaxies look, at high-ionisation wavelengths, like a cosmological-scale version of a local starburst cluster.

The result has direct implications for the demographics of the early Universe. Bright-end UV luminosity function measurements during reionisation have historically assumed that anything more luminous than a threshold around M_UV = -23 is dominated by quasars; the faintest “quasar” bin is then used to constrain the AGN duty cycle at early times. If a non-trivial fraction of that bin is instead powered by VMS-driven starburst galaxies, the inferred quasar space density drops, the inferred galaxy space density rises, and the contribution of early galaxies to reionising the intergalactic medium grows. The paper notes explicitly that the inferred demographics of both galaxies and quasars in the crossover regime need revisiting.

The story is also a methodological lesson in spectroscopic re-classification. SHELLQs selected these sources on the basis of photometric dropout signatures and broad-line spectral markers visible at low signal-to-noise in 8-metre-class spectra. JWST/NIRSpec at higher resolution and sensitivity shows that the broad lines were real stellar wind profiles, not the broad permitted lines of an accretion disk, and that the narrow nebular lines, which were below the discovery spectra’s noise floor, are decisive for reclassifying the sources as galaxies rather than quasars. Crossover-regime classification at cosmic dawn is now an empirical problem rather than a photometric one, and ALMA’s structural imaging (the 5.4-kpc UV/[C II] offset in J1429) provides the second confirming axis.

The paper is a preprint, resubmitted to Nature Astronomy on 20 July 2026 according to the arXiv metadata, so the model grids and equivalent-width fits will be re-examined through peer review. The next observational test is straightforward: more JWST/NIRSpec spectra of SHELLQs candidates in this brightness range, paired with ALMA [C II] imaging to look for the same offset geometry. If even a third of the M_UV around -23.5 bin turns out to be VMS-driven starbursts rather than faint quasars, the early-Universe census shifts noticeably, and R136 stops being a peculiar local object and starts being a representative sample of how the first massive stellar populations lit up their galaxies.