The cosmos is full of old stories written in light. Some galaxies blaze with newborn stars, others go quiet, their stellar nurseries shuttered as if by nightfall. For years, astronomers chasing the quiet giants—the first massive quiescent galaxies—taced the universe with a question mark: when did these giants stop forming stars, and how did they get so big so fast? The James Webb Space Telescope (JWST) has rewritten that chapter. A new program called DeepDive goes beyond snapshots, pulling out deep, detailed spectra of some of the universe’s earliest massive dead-ends in star formation, around redshifts z ~ 3–4. The result isn’t a single answer but a richer, more nuanced story of how and when these behemoths settled down. The work is led by K. Ito and F. Valentino at the Cosmic Dawn Center (DAWN) in Copenhagen, a collaboration spanning institutions such as the University of Copenhagen and DTU Space, among others. It’s a testament to how a deep dive can illuminate the handwriting of cosmic history more clearly than a postcard glimpse.
What DeepDive adds to the literature is not merely a larger catalog of galaxies, but a way to read the timelines of their lives. The team collected very deep (1–3 hours) JWST/NIRSpec spectra for ten primary, massive quiescent galaxies at z ~ 3–4, targeting the rest-frame optical window around the Balmer break and, crucially, extending into the Hα and [N II] region. They didn’t stop there. They also combed the JWST Archive for similar spectra of quiescent galaxies, building a statistical context with a much larger sample: about 140 archival sources spanning 1 < z < 5, with wide coverage in stellar mass. Together, the DeepDive target set and the archival companions comprise roughly 150 quiescent galaxies, offering a rare chance to study how these systems evolve across a broad swath of cosmic time and mass. This is the first robust census of such galaxies that leverages medium-resolution spectroscopy to characterize both their stellar populations and their ionized gas in detail.
In a sense, DeepDive is archaeology on cosmic scales. It digs into the fossil records of star formation, metallicity, and the dynamics of stars that formed billions of years ago. It also asks a critical, practical question that underpins how we interpret the oldest galaxies: how reliable are our photometric selections when the universe’s look-back time stretches to the edge of detectability? The study uses three classical selection routes—strength of the 4000 Å break (Dn4000), rest-frame UVJ colors, and specific star formation rate (sSFR)—and finds a remarkable ~90% overlap among these methods for the quiescent population, with around 70% of archival sources meeting all three. That convergence is itself a note of confidence in how we carve out quiescent galaxies from the distant fog. And the public data release is a quiet revolution in how open science accelerates progress in a field where data quality is as decisive as theory.
A program built for the deep dive
The DeepDive program is not a single telescope night but a coordinated campaign that combines several Jules Verne-like moves: targeted deep spectroscopy, careful flux calibration, and an archival backbone that anchors the findings in a broader cosmic timeline. The DeepDive team used JWST/NIRSpec with the G235M/F170LP setup to cover wavelengths around the Balmer break for targets at z > 3.4, achieving signal-to-noise ratios on the stellar continuum that let them detect faint spectral features in the rest-frame optical. Their ten primary, massive galaxies sit at redshifts between 3.4 and 4, with stellar masses around 10^10.8 to 10^11.5 solar masses. The long exposure times—ranging from about 2,000 to 10,500 seconds per mask—were tailored to ensure a fair comparison of S/N across different targets, a crucial step when you’re chasing subtle absorption lines and narrow emission features in these distant systems.
One of the technical triumphs of the work lies in how they handled the wavelength coverage. By extending the nominal G235M/F170LP range with a novel reduction pipeline, they could access the Hα+[N II] complex in all their main targets. That matters a lot: Hα is a direct tracer of star formation when ionized by young, massive stars. Detecting Hα at z ~ 4 is not just a technical feat; it provides a clean empirical anchor for the current star formation rate (SFR) that can be compared with inferences from the stellar continuum and spectral energy distribution (SED). In five of the ten main targets, they detected Hα at S/N ≥ 3, and in two galaxies, the Hα line shows a broad component, a signature often associated with active galactic nuclei (AGN). The group quantified instantaneous SFRs of roughly 0–5 solar masses per year for these galaxies, consistent with a quiescent status given their stellar masses, though with the caveat that some Hα flux could be AGN-driven in those particular cases.
To place these newly observed galaxies in a wider context, DeepDive also performed a comprehensive archival search. They pulled together 140 quiescent galaxies with grating spectra from the DAWN JWST Archive, extending the redshift coverage to 1 < z < 5 and spanning more than an order of magnitude in mass. This archival slice is not a mere appendix; it provides a statistical backbone that lets the authors examine how the high-redshift DeepDive targets compare with lower-redshift quiescent galaxies, how quenching signatures evolve, and how consistent UVJ, Dn4000, and sSFR criteria are across a broad cosmic timeline. All photometric and spectroscopic data from both the DeepDive program and the archival compilation are being released publicly, a move that invites the global community to test, refine, and reuse these measurements for years to come.
A census across cosmic time
With the two datasets in hand, the authors map what quiescence looks like across 1 < z < 5. The archival set, which peaks around z ~ 2, sits in mass and SFR territory consistent with earlier findings: substantial numbers of massive galaxies that have already quenched long before the present day. The DeepDive main targets, by contrast, inhabit a higher redshift corridor, z ~ 3.4–4, and, at fixed redshift, skew toward higher stellar masses (log M*/M⊙ about 10.8 to 11.5). In other words, they are among the early population of truly massive galaxies that stop forming stars while the universe is still a few billion years old or younger. The stellar mass–SFR relation shows most of the DeepDive galaxies well below the star formation main sequence at their epoch, reinforcing their quiescent status, while a handful of archival galaxies at z ~ 2 live closer to, or on, the main sequence, illustrating diversity in the quenching journey.
Beyond the individual galaxies, the team asks how well photometric redshifts track the actual, spectroscopically measured redshifts. The comparison shows excellent agreement at z < 3.5, with a normalized median absolute deviation (NMAD) around 0.022 and a small bias near −0.01. However, at the highest redshifts, the photometric redshifts tend to underestimate z, with a bias around −0.17 for 3.5 < z < 5. The takeaway: for the most distant quiescent galaxies, spectroscopy matters—photometric estimates can slip by, and having spectra makes for a more reliable census of the early quiescent population. That reliability is essential when you’re trying to quantify how quickly galaxies quench and how their stellar populations age in the first few billion years after the Big Bang.
Jumping to the color-and-age plane, the DeepDive and archival quiescent galaxies largely populate the UVJ diagram as expected, but with a twist at high redshift: many bright, massive galaxies in the DeepDive sample sit toward bluer UVJ colors and lower Dn4000 values. That indicates younger stellar populations in these early quiescent systems, precisely the signal you’d expect if you’re catching galaxies closer to their quenching epoch. The authors go further and quantify a relation between Dn4000 and a diagonal UVJ-based coordinate they define, S_Q. The emergent pattern is intuitive: as you move toward younger, recently quenched systems, Dn4000 declines and the rest-frame colors trend bluer; older quiescent systems display stronger 4000 Å breaks and redder colors. The analysis even reveals faint metal absorption features in the stacked high-S/N spectra of the z ~ 3 subsample, suggesting that metal enrichment had already begun even as the stellar populations remained relatively young. This paints a picture of rapid, diverse quenching: some galaxies snuff out star formation quickly, while others still bear the fingerprints of their metal-rich early histories.
What DeepDive reveals about stars and metals
If you want a mental image, imagine listening to a chorus where some singers have already paused, others still warming up. The stacked spectra in the DeepDive sample betray this mix. When the authors group galaxies by Dn4000, the lower-Dn4000 bin—the more recently quenched—shows stronger nebular emission lines, including [O II], [O III], and Hα, hinting at residual activity that could be powered by a residual star formation or by AGN processes. In contrast, the higher-Dn4000 bin—the older cohort—shows stronger absorption features tied to older stellar populations, such as Ca II H+K and Mg, and overall redder continua. The faint metal lines, including Fe4383 and Mgb, become detectable in the higher-Dn4000 stacks, offering a rare peek into the metallicity and alpha-element enrichment of massive galaxies when the universe was still forming its first billions of years of history. The results are consistent with a scenario where early, rapid star formation injects alpha elements (like magnesium) into the interstellar medium on short timescales, followed by a slower iron enrichment as Type Ia supernovae contribute later on.
In a nice demonstration of how far spectroscopy has come, the team measures not just the ages but also the chemical fingerprints of these systems. By comparing the observed line strengths with modern stellar population models (including BPASS, which incorporates binary stars and varied metallicities), they illustrate how Dn4000 tracks stellar age while the metal indices encode metallicity and abundance patterns. Even with a simplifying assumption of single-burst populations for illustration, the trends line up with expectations: younger, recently quenched galaxies show signs of ongoing or recent activity in their emission lines, while older quiescent galaxies carry the stronger absorption features shaped by longer-ago star formation histories. The broad implication is that DeepDive is unlocking the metallurgical history of the early universe as seen in massive quiescent galaxies, offering a rare window into how elements were synthesized and distributed in the first few gigayears of cosmic time.
Another striking result comes from the evidence for ionized gas in several quiescent galaxies. The stacked spectra across the Dn4000 bins reveal emission lines that can be powered by active galactic nuclei, residual star formation, or shocks. The pattern—stronger emission in the more recently quenched stacks—suggests that AGN activity or low-level star formation may persist in the immediate post-quenching phase for some of these galaxies. Adding to the intrigue, roughly half of the DeepDive main targets show signs of blue-shifted Na I D absorption, a classic tracer of neutral gas outflows. In other words, even as these galaxies seem largely dormant in star formation, they are not sitting completely still: they may be expelling gas or experiencing feedback processes that help keep star formation suppressed, a crucial piece of the baryon-cycle story in the early universe.
Why this matters for how we tell the story of galaxy formation
The DeepDive results illuminate the era when the universe was only a few billion years old and the cosmic census of massive galaxies was still taking shape. They reinforce the view that massive quiescent galaxies form very quickly, in episodes of intense star formation, and then quench on surprisingly short timescales. The presence of young stellar ages in z ~ 3–4 quiescents, together with clear signatures of alpha-enhancement in some stacked spectra, provides empirical leverage on the timescales of star formation and the relative timing of supernova contributions to chemical enrichment. In other words, these galaxies seem to have formed their stars fast, pumped out metals into their surroundings, and shut down rapidly enough that the universe’s light reveals a relatively pristine, young stellar population in some sources, while older members carry stronger metal fingerprints in their spectra.
There’s also a methodological payoff. JWST’s ability to access Hα at these redshifts is a game changer. Hα is one of the most direct tracers of current star formation, and measuring it in z ~ 4 galaxies provides a much more immediate measure of their ongoing activity than is possible from UV light alone, which can be muddied by dust and recent star formation histories. The finding that several DeepDive galaxies host broad Hα components—clear AGN signatures—reminds us that the quenching story is not just about starving stars but also about how massive black holes might regulate gas on galaxy-wide scales. And the outflow signatures seen in Na I D offer a tangible mechanism by which galaxies might keep a lid on future star formation, ejecting gas and heating their surroundings to thwart future stellar nurseries.
Another big plus is the statistical backbone built by combining the DeepDive sample with a broad archival population. This is the kind of dataset that allows theorists to test models of how galaxies assemble mass, regulate star formation, and evolve structurally over a wide redshift range. The work hints at a continuity in the quenching process across 1 < z < 5, with high-redshift quiescent galaxies appearing younger and bluer than their later-time cousins, yet still bearing the hallmarks of rapid formation and swift cessation. That continuity is essential for calibrating hydrodynamical simulations and semi-analytic models, which must reproduce not only the existence of these galaxies but also their spectral fingerprints, abundance, and outflow behavior across cosmic time.
What’s next and why it matters
The DeepDive team is careful to frame their paper as the opening move in a longer conversation. They’ve laid down a robust, publicly accessible data set that will feed a tide of follow-up work. Future papers will dive into star-formation histories with greater nuance, extract metallicities and abundance ratios like [Mg/Fe] to pin down formation timescales, and attempt to measure stellar velocity dispersions to constrain dynamical masses and test the mass–size–velocity dispersion relations that inform the so-called fundamental plane for massive galaxies out to high redshift. With JWST imaging (NIRCam) providing precise sizes and light profiles, researchers will be able to compare dynamical masses to stellar masses inferred from SED fitting, offering insights into the initial mass function and the mass assembly of these giants. And because several targets show AGN or AGN-like signatures, the work will feed into the broader conversation about how black hole growth and feedback influence the quenching of star formation in the early universe.
In practical terms, the study’s approach—combining deep, high-quality spectroscopy with a large archival chorus—offers a blueprint for how we transform snapshots into statistics. The authors estimate a population-level view of quiescent galaxies at z ~ 3–5 by stitching together a carefully curated sample that spans wide ranges in mass and redshift. This isn’t merely catalog-building; it’s about building the empirical scaffolding that will support the next generation of galaxy formation theories. And because the data will be publicly accessible, other researchers can apply new models, test alternative selection criteria, or search for subtle correlations that the current study might not have captured. This openness accelerates a field where every photon captured from the early universe matters as a data point in a grand, cosmic census.
Ultimately, the DeepDive project turns a whisper into a chorus. It shows that the first massive quiescent galaxies were already writing their own epics in the universe’s first few billion years, with rapid star formation, swift quenching, hidden AGN activity, and faint metallic signatures that only now, through JWST’s power, we can hear clearly. It’s a reminder that the story is not simply one of “how did galaxies stop forming stars?” but a richer tale of how, when, and under what conditions the most massive galaxies in the universe shut down and then settled into the quiet majesty of the present day. The lead authors—K. Ito and F. Valentino—along with their collaborators at the Cosmic Dawn Center and partner institutions, have given us a robust framework to read that history with more fidelity than ever before. The universe, it seems, had more to teach us about quiet giants than we realized, and DeepDive is the map that helps us listen more carefully to what they have to say.
