An Ancient Descendant of the First Galaxies

TL;DR

The Sleeper is a remarkable relic galaxy at $z=5.95$ with a strong Balmer break $D_B\approx2.9$, and a stellar mass of $M_\ast\approx10^{9.14} M_\odot$. Non-parametric SFH modeling reveals a rapid early assembly by $z\gtrsim14$ forming a $M_\ast\approx10^{8.7} M_\odot$ progenitor, followed by a multi-hundred-Myr lull that yields a suppressed current star formation rate $\mathrm{SFR}_{10}\approx0.58^{+0.86}_{-0.37}$ M$_\odot$ yr$^{-1}$. This combination of a strong Balmer break in a relatively low-mass galaxy and its rarity in simulations supports an evolutionary pathway where luminous early galaxies fade into inefficient dwarfs, instead of always growing into the most massive descendants. The Sleeper resides in a dense, Balmer-break–rich environment, suggesting that such relics may trace ancient overdensities in the early universe and that burstiness driven by feedback can dominate the early mass assembly history.

Abstract

JWST has revealed unexpectedly bright galaxies in the first 500 Myr after the Big Bang. Their overabundance suggests that they are preferentially observed during burst phases, where their star formation rates increase dramatically. In cosmological simulations, such bursts transition into short ($\approx 40$ Myr) periods without star formation or naps. Using JWST/NIRCam medium-band observations, we report the discovery of the galaxy CANUCS-A370-2228423 ($z = 5.95 \pm 0.06$, $\log(M_\ast/M_{\odot}) = 9.14 \pm 0.09$), dubbed The Sleeper. Its star formation history indicates rapid assembly in the first 300 Myr ($z \gtrsim 14$), where it formed a $\log(M_\ast/M_{\odot}) = 8.7^{+0.3}_{-0.4}\ M_{\odot}$ progenitor, comparable in stellar mass to the few spectroscopically confirmed galaxies at those redshifts. Unexpectedly, this is followed by several hundred million years of suppressed star formation, in stark contrast to nappers. This results in a remarkably strong hydrogen Balmer break, exceeding that of any galaxy observed within the first billion years by a factor of $\approx 3$. Furthermore, Sleeper-like systems are overabundant in the observed survey volume compared to theory, as the probability of finding such galaxies in simulations is $< 0.2\%$. The discovery of The Sleeper therefore disrupts the current narrative that all luminous galaxies in the first few hundred million years grow into massive descendants. Instead it presents an alternative evolutionary pathway in which these unusually luminous galaxies fade into inefficient dwarfs after an early starburst, revealing greater diversity in the first stages of galaxy evolution.

Figures (9)

• Figure 1: Photometry and spectral energy distribution (SED) modeling of The Sleeper. Left: JWST/NIRCam color-composite image. The Sleeper is located at a projected distance of $\approx 3$ kpc from a more massive ($\log(M_\star/M_\odot) = 9.55 \pm 0.35$) companion (CANUCS-A370-2201097), both situated in an overdensity at $z = 5.87 - 6.14$ with 36 other galaxies showing Balmer breaks. Right: maximum likelihood DENSE BASIS fit ($\chi_\nu^2 = 1.14$) to the 29-band photometry from CANUCS/Technicolor. This comprises all the available JWST/NIRCam medium and broad bands, combined with two narrow bands and ancillary HST data. The F250M and F300M medium bands, as well as the low fluxes in F200W and F210M, robustly identify a strong flux excess caused by the redshifted hydrogen Balmer spectral break ($\lambda_{\mathrm{rest}} = 3645~$Å), which unambiguously indicates an evolved population when the observed SED is dominated by stars.
• Figure 2: Relationship between Balmer break strength ($\mathrm{D_B}$), age, and star formation history at $z \sim 5 - 10$. We show the break strength measured for a flux-limited CANUCS/Technicolor sample (shaded contours), spectroscopically-confirmed individual literature galaxies vikaeus2024witten2025strait2023looser2024 (gray filled points), and stacks rb2024 (black filled points). We exclude "little red dots," e.g., wang2024setton2024 because they may have a significant flux contribution from gas heated by AGN and not from stars, making the interpretation of their Balmer break strengths ambiguous. The solid lines show the evolution of maximally-old models: a stellar population with age equal to the Hubble time, $t_H(z)$, and models assuming various onsets of star formation after the Big Bang ($200,~300,~\mathrm{and}~ 400$ Myr). These models assume an instantaneous burst (timescale, $\tau = 3$ Myr) and 10% solar metallicity. The dashed line is for an extended burst model ($\tau = 400$ Myr at $z = 6$). Massive (log $M_\ast/M_\odot \geq 10$) quiescent galaxies at these redshifts degraaff2025_natureweibel2025 are included as open points. The extreme Balmer break strength of The Sleeper is best explained by a stellar population that formed at least 50% of its stars in the first 500 Myr of cosmic history.
• Figure 3: The Sleeper compared to spectroscopically-confirmed galaxies at $z > 10$robertson2023naidu2025castellano2024. Left: the median non-parametric star-formation history inferred using DENSE BASIS. The shading indicates the 16th and 84th percentiles of the posterior estimated as a function of cosmic time. The vertical lines show when the galaxy assembled 25%, 50%, and 75% of its stellar mass. The Sleeper earns its name from the extended period of suppressed star formation at $z \lesssim 10$ relative to the time-dependent main sequence traced by its stellar mass growth speagle2014. This results in a low current specific SFR (log $\text{sSFR}_{10} [ ~\textrm{yr}^{-1}] \approx -9.4$). Middle: the cumulative mass assembly history, accounting for mass loss due to stellar evolution. The Sleeper forms a log($M_\ast/M_\odot) = 8.70_{-0.35}^{+0.32}$ progenitor at $z \sim 14$ in a rapid ($\approx 300$ Myr) burst, consistent with the stellar masses of UV-bright $z > 10$ galaxies. Right: the subsequent period of suppressed star formation results in a $\sim 2$ mag decline in the UV luminosity from $z \sim 14$ ($0.4\mathcal{L^{*}}$) to $z \sim 6$ ($0.04\mathcal{L^{*}}$), consistent with predictions from analytical models of burstiness wyithe2014. This assumes $\mathcal{L^{*}} = -20.75_{-0.81}^{+0.64}$willott2024.
• Figure 4: Evolutionary pathways of the oldest galaxies observed in the first two billion years. We show the lookback time at which the galaxy formed 50% of its stellar mass as a function of observed redshift for spectroscopically-confirmed massive quiescent galaxies at $z \gtrsim 3$glazebrook2024degraaff2025_natureweibel2025carnall2024onoue2025sato2024 and UV-luminous galaxies at $z > 10$robertson2023naidu2025. Lighter symbols use the $t_{50}$ corresponding to a higher metallicity solution, where available. Shaded tracks show the ages of passively evolving simple stellar populations at $5 < z_\mathrm{form} \leq 25$. The star formation history and stellar mass of The Sleeper are broadly consistent with the measured ages of the first galaxies at $z \sim 14$. Given that the oldest ($t_{50} \gtrsim 500$ Myr) galaxies observed at these redshifts tend to be extremely massive ($\gtrsim 10^{10} ~ \mathrm{M_\odot}$), The Sleeper presents a unique opportunity to study a possible progenitor of the low-mass quiescent population at $z \lesssim 3$. Assuming passive evolution at its current star formation rate, The Sleeper predicts a population of ancient, low-mass galaxies at $z \sim 3$ which are missing from current samples likely due to observational biases cutler2024.
• Figure 5: The flux profile of The Sleeper . The left panel shows a $\sim 1\times1$$^{\prime\prime}$ cutout from the PSF-homogenized F277W image in logarithmic scaling, showing the 2D flux profile of its companions labelled B (CANUCS-A370-2217627, $z = 6.78 \pm 0.14$) and C (CANUCS-A370-2201097, $z = 6.00 \pm 0.03$). The catalog apertures are also shown. Right panel: 1D flux profile of The Sleeper determined using the ratio of its growth curves in the highest S/N filter (F277W) and the PSF reference (F444W), normalized at $r = 0.27$$^{\prime\prime}$. While the scaled Kron aperture, $r_\mathrm{tot} = 2.5\times r_\mathrm{circ}$, contains the total source flux, it also includes $\sim 10\%$ contamination from B and C. This would introduce a systematic of the same magnitude in estimating its photometric redshift and stellar population parameters. Therefore, we determine its total flux using the optimal aperture, $r = 0.25$$^{\prime\prime}$, scaled to $r=0.35$$^{\prime\prime}$ .