WorksPhysics and Astronomy
Dust Survival in Lensed Quiescent Galaxies at Depends on How They Quench
Abstract
The interstellar medium of massive quiescent galaxies holds key clues to how they quench, yet remain poorly understood. Although dust is expected to be depleted rapidly (100 Myr) after quenching, some quiescent galaxies host substantial dust reservoirs, raising the question of whether dust survives quenching or is replenished afterward. We present ALMA Band 6 continuum observations (250m) of eight strongly-lensed, spectroscopically-selected quiescent galaxies at z1-2.5 from the COOL-LAMPS and REQUIEM surveys, plus five archival REQUIEM systems. The sample spans more than two orders of magnitude in dust-to-stellar-mass ratio ( to ), and two galaxies host extended, resolved dust. Comparing with time since quenching, roughly half the sample, preferentially at higher redshift, is consistent with rapid quenching and harsh dust destruction, while lower-redshift, more slowly quenched systems are consistent with milder destruction, possibly reflecting an evolving dominant quenching mode. The two most dust-rich galaxies exceed even the mildest model destruction tracks, with J1201-3952 retaining the highest 1.8 Gyr after quenching. If these systems are truly quiescent, their dust masses approach maximum stellar yields, pointing to ISM grain regrowth and/or TP-AGB dust production. However due to our single-band continuum data, is uncertain (0.3 dex systematic uncertainty from the unknown dust temperature) and we cannot rule out obscured star formation in these two galaxies. Our results suggest that dust survival depends on how galaxies quench: rapid, harsh quenching efficiently destroys dust, while slower, milder quenching can leave dust reservoirs that persist or are replenished long after star formation ceases.
dex systematic uncertainty from the unknown dust temperature) and we cannot rule out obscured star formation in these two galaxies. Our results suggest that dust survival depends on how galaxies quench: rapid, harsh quenching efficiently destroys dust, while slower, milder quenching can leave dust reservoirs that persist or are replenished long after star formation ceases.
Introduction
In the current picture of massive galaxy evolution, a large fraction of massive galaxies exhibit low star formation rates (SFRs; e.g., Muzzin et al. [89]; Ilbert et al. [54]; Tomczak et al. [136]; Davidzon et al. [29]), compact, bulge-dominated morphologies (e.g., Daddi et al. [28]; van Dokkum et al. [140]; van der Wel et al. [139]), and red colors from old stellar populations (e.g., Kriek et al. [66]; Williams et al. [148]; Whitaker et al. [143, 144]; Muzzin et al. [90]). These quiescent galaxies are thought to evolve from star-forming disk galaxies, but the physical processes through which star formation shuts down and the origin of their compact morphologies is unclear (for recent comprehensive reviews, see Whitaker & Bezanson [142]; Bluck [9]). Proposed quenching mechanisms must either heat, remove, consume, or expel reservoirs of cold molecular gas, which serve as the fuel for star formation, either over rapid () or slower () timescales (e.g., Carnall et al. [17]; Park et al. [100]; Edward et al. [41]). Therefore, the joint analysis of star formation histories (SFH) and cold dust and gas content in quiescent galaxies can reveal how, when, and where quenching happens.
Several studies have sought to study molecular gas in quiescent and post-starburst galaxies, primarily utilizing CO spectroscopy (Hayashi et al. [51]; Spilker et al. [124]; Woodrum et al. [149]; Wu et al. 2023; Zanella et al. [152]; D’Onofrio et al. [36]; Setton et al. [118]; Scholtz et al. [113]). However, a challenge of this approach is the long integration times required to detect CO, particularly in systems that are expected to be intrinsically gas-poor if quenching has occurred through gas depletion. Consequently, CO has been detected in only a handful of quiescent galaxies, limiting robust constraints on the molecular gas content of the population. Beyond total gas content, spatially-resolved gas measurements can identify specific quenching mechanisms and elucidate how stellar mass assembles across different regions of a galaxy. For example, centrally depleted gas reservoirs could indicate inside-out quenching driven by AGN feedback or morphological quenching, consistent with the suppressed central star formation and older central stellar populations seen in resolved studies of massive galaxies at cosmic noon (Tacchella et al. [131]; Nelson et al. [94]; Tacchella et al. [132]; Akhshik et al. [5]). Similarly, gas displaced from the stellar body or swept up in outflows would point to environmental stripping or ejective feedback (Spilker et al. [125]; Belli et al. [7]; D’Eugenio et al. [31]). However, such resolved CO observations are prohibitively expensive for most quiescent galaxies at .
A less expensive tracer of cold gas is dust continuum at , which is closely correlated with CO luminosity in star-forming galaxies (Scoville et al. [115, 116]). While dust continuum has been used to estimate molecular gas mass in local quiescent galaxies (Gomez et al. [49]; Rowlands et al. [108]; Lianou et al. [73]; Michalowski et al. [85]; Lesniewska et al. [71]; Michalowski et al. [86]), studies at higher redshift often result in non-detections (Caliendo et al. [16]; Kalita et al. [61]; Whitaker et al. [145]; Suzuki et al. [130]; Chworowsky et al. [23]; Umehatа et al. 2025; Siegel et al. [121]; Spilker et al. [126]; Chang et al. [20]), indicating that little cold dust is present. There are several individual detections of dust continuum in quiescent galaxies; some of these systems are strongly gravitationally lensed, enabling observations of less massive dust reservoirs (Whitaker et al. [145]; Morishita et al. [88]), or they probe the dustiest subset of the quiescent population (Donevski et al. [34]; Lee et al. [67]). Alternatively, stacking has been used to measure dust masses below the detection limit for statistical samples of quiescent galaxies. Early studies showed that stacking recovered higher dust masses than individual measurements (Gobat et al. [48]; Magdis et al. [79]; Blánquez-Sesé et al. [8]), whereas recent mass-complete ALMA stacks found lower average values consistent with individual measurements (Adscheid et al. 2025).
This tension between stacking and individual measurements of dust mass may partly reflect methodological biases. Stacks can be boosted by the dustiest sources (Lorenzon et al. [74]), by source confusion in low-resolution far-IR maps (Adscheid et al. 2025), or by including lower-mass galaxies with higher fractions (Blánquez-Sesé et al. [8]), while individual detections remain limited to a handful of objects. Dust mass estimates from both approaches are further uncertain due to the assumed dust temperature (Cochrane et al. [24]). However, the wide range of measured dust masses may also reflect genuine diversity in the ISM of quiescent galaxies. Supporting this point, simulations predict gas-to-dust ratios spanning up to four orders of magnitude (Whitaker et al. [146]; Lorenzon et al. [74]), and observations find unusually high and varied gas-to-dust ratios in some systems [126, 75].
Dust is injected by supernovae and the winds of evolved stars and is subsequently reprocessed in the ISM, where grains grow in dense gas and are shattered into smaller grains in diffuse gas [109, 93, 99]. Once star formation ceases, these sources fade, with Type II supernovae ending within Myr, dust production by asymptotic giant branch (AGB) stars declining over the following several hundred Myr to Gyr as progressively lower-mass stars reach this phase, and grain growth stalling as cold, dense gas is consumed, heated, or expelled. Destruction, by contrast, can be rapid, through thermal sputtering in hot gas, supernova shocks, astration, and outflows (Li et al. 2019), with sputtering alone destroying grains within - Myr in hot halos [40, 59, 58, 52]. Quiescent galaxies are therefore naively expected to lose their dust shortly after quenching. Yet observations of quiescent galaxies with substantial dust reservoirs [51, 48, 8, 67] suggest that removal may be prolonged in some quiescent systems, or offset by replenishment through grain growth or asymptotic giant branch (AGB) production [34, 74, 35].
Beyond the timescale of dust destruction, a second confounding factor is the diversity of SFHs among quiescent galaxies, which encode a wide range of quenching timescales and times since quenching. While only of quiescent galaxies at are predicted to be recently quenched, this number rises to at [e.g. 142 and references therein]. If dust destruction proceeds more slowly than quenching, recently quenched galaxies may retain dust formed before quenching, or sustain enough residual star formation to continue producing it. Thermally pulsing asymptotic giant branch (TP-AGB) stars add a further channel: they are expected to be prominent in quiescent galaxies with intermediate-age stellar populations (- Gyr), particularly at , and could produce dust long after quenching, which would trace the intermediate-age stars rather than the dense core (Lu et al. 2025).
The details of dust destruction after star formation has been disrupted remain poorly understood. Donevski et al. [35] predict that the fate of the dust depends most strongly on the local ISM conditions, ranging from harsh conditions, with or without AGN feedback, that rapidly destroy dust, to milder conditions that prolong its survival. In some cases, particularly in low- to intermediate-mass quiescent galaxies, grain regrowth keeps the dust mass roughly constant [35, 74]. This heterogeneity may make dust continuum an unreliable tracer of molecular gas in quiescent galaxies [146, 126]. However, because dust content is tied to both quenching conditions and stellar population age, the dust continuum is itself a powerful probe of quenching, complementary to molecular gas studies.
In this paper, we present new ALMA Band 6 continuum observations of a sample of eight strongly lensed, spectroscopically-selected quiescent galaxies from the COOL-LAMPS [62] and REQUIEM surveys, and analyze their cold dust properties alongside SFHs derived from spectro-photometric spectral energy distribution (SED) fitting. We combine this sample with archival ALMA observations of five REQUIEM galaxies [145, 5]. Lensing magnification boosts both sensitivity, making it possible to detect, or place stringent limits on, the faint dust reservoirs of individual quiescent galaxies at [e.g. 145] that are otherwise only accessible through stacking. Where dust is detected, lensing also stretches the emission across the sky, so even our low-resolution (1″) data can begin to distinguish extended from compact dust, and this sets up higher-resolution follow-up to map the dust relative to the stars.
The paper is structured as follows. § 2 describes the sample, existing optical data, and the reduction of the ALMA dust continuum and ground-based spectroscopy. § 3 presents our method for determining lensing magnification, measuring dust masses, fitting SEDs, and deriving quenching timescales from SFHs. § 4 relates the dust properties of the sample to redshift, specific star formation rate (sSFR), stellar age, and SFH. In § 5, we examine the relationship between time since quenching and dust fraction, what it implies about changing modes of dust destruction across the redshift range of the sample, and whether the largest dust masses arise from stellar dust production (AGB stars and Type II supernovae) or grain growth in the ISM, as well as whether the dust-rich systems could host obscured star formation. We summarize our key findings in § 6.
Data
Our sample consists of 13 strongly lensed quiescent galaxies at , including seven from the Chicago Optically-selected strong Lenses – Located At the Margins of Public Surveys (COOL-LAMPS) collaboration [62] and six from the REsolving QUIEscent Magnified (REQUIEM) Galaxies Survey (Whitaker et al. [145]; Akhshik et al. [5]). We present new dust continuum imaging for 8 of the 13 galaxies (§ 2.3, see Figure 1); 7 of these galaxies are from COOL-LAMPS. The COOL-LAMPS and REQUIEM samples differ in their imaging depth and resolution (§ 2.1 and § 2.2) and in their rest-frame wavelength coverage (§ 3.2.3). While different tools are used to model their stellar populations and derive star formation histories (SFHs), the underlying assumptions are similar (§ 3.2.3). We describe each below and summarize the differences in § 3.2.3 and Table 1.

Figure 1. RGB images for the sample of eight lensed, spectroscopically-selected quiescent galaxies from the COOL-LAMPS and REQUIEM surveys. The optical imaging is with the exceptions of (i) J1201-3952, where we use H-band data, (iii) J1323-0343, where we use F140W data (Zhuang et al. [154]), and (iii) J1522-2525, for which we use HST imaging (F110W, F125W, and F140W). We show 3, 4, and contours from the 1.3mm dust continuum. Arrows point to the lensed arcs; in the case of J1323-0343, there is a nearby source which is associated with the arc.
| ID | [mJy] | Age [Gyr] | sSFR [yr] | ||||
| J1323-0343 | 1.016 | ||||||
| J2359-0208 | 1.121 | ||||||
| J1201-1443 | 1.152 | ||||||
| J0112-2053 | 1.176 | ||||||
| J1509-3308 | 1.105 | ||||||
| J1553-0142 | 1.285 | ||||||
| J1116-1443 | 1.483 | ||||||
| J1522-2535 | 2.450 |
Table 1. Sample summary for strong lens quiescent galaxies where we add new ALMA dust continuum imaging (§ 2.3). J1522-2535 is from the REQUIEM survey; stellar mass, sSFR, and age for J1522-2535 come from Newman et al. [95].
COOL-LAMPS
Sample Selection
COOL-LAMPS candidates were identified through visual inspection of image cutouts from the DESI Legacy Imaging Surveys DR9 (DECaLS; Dey et al. [32]) and DELVE (Drlica-Wagner et al. [37]), centered on red-sequence galaxies and galaxy clusters. Members of the collaboration ranked candidates by the likelihood of being a strong lens, prioritizing systems where background arcs lie at the margins of color-color and color-magnitude spaces (i.e., bright and red objects in co-adds-based photometric catalogs), and the highest-ranked systems were targeted for spectroscopic follow-up with Magellan (§ 2.1.3). From this parent sample we include the seven systems for which follow-up ALMA dust continuum imaging was acquired and whose lensed source shows a spectrum consistent with a quiescent stellar population (e.g., strong Balmer/4000 Å break and absorption features, with weak or absent emission lines). A detailed study of one of these systems, COOL J1323+0343, is presented in Sukay et al. [127], with the full sample (data and stellar population synthesis modeling) to be presented in Khullar et al. (in prep). Figure 1 shows RGB composite images of these seven sources (in addition to J1522-2535, to be presented in § 2.2).
Optical and Near-Infrared Imaging and Modeling
Optical imaging was obtained from DECaLS DR9 (Dey et al. [32]). For a subset of targets, we obtained follow-up imaging with the FourStar Infrared Camera (FOURSTAR; Persson et al. [102]) on the Magellan/Baade telescope. These data were reduced to astrometrically and photometrically calibrated stacks using a custom pipeline built via IRAF (Tody [134], Tody [135]) and PHOTPIPE (Rest et al. [105], Garg et al. [46], Miknaitis et al. [87]). The imaging reduction and photometric analysis broadly follow Sukay et al. [127], with COOL J1323+0343 serving as a representative example of the pipeline. The full details for the sample will appear in Khullar et al. (in prep). Typical seeing was 0.7″ in the optical and 1″ in the near-IR, and all targets were resolved due to their strongly lensed nature. There is significant contamination in the arcs due to flux from foreground/interloper cluster member galaxies that form the gravitational lens. Therefore, we measure photometry via GALFIT modeling (Peng et al. [101]). We fit the lens galaxy and nearby contaminants with Sérsic profiles and model the arc as multiple Sérsic components in across each available image. The model components belonging to the primary arc are merged into a single nominal “banana” aperture whose outer boundary traces the modeled arc (see also, Khullar et al. [62]; Sukay et al. [127]; Akhshik et al. [5]). The same aperture is effectively applied to all bands, and fluxes are measured by adding the inferred flux from all GALFIT model components 9in lens-subtracted images. The background is estimated by placing multiple apertures in blank sky with the same “banana” aperture as the science target.
Ground-Based Spectroscopy
Spectra of the lensed sources were obtained with the Low Dispersion Survey Spectrograph 3 (LDSS3) on the Magellan/Clay telescope between 2021-2024, using the 1″ x 4′ longslits and the VPH-Red Grism disperser, with total integration times of 1–2 hrs (except COOL J1323+0343, where the spectrum was observed with NOT/ALFOSC; Sukay et al. [127]). Slits were aligned along the brightest arc segment where possible. We reduced the data with a custom IRAF pipeline that performed bias subtraction, flat-fielding, wavelength calibration, sky subtraction, and optimal extraction of the 1D spectrum; we used imaging for relative and absolute flux calibration as a parameter in spectral energy distribution fitting (§ 2.1.2). Spectroscopic redshifts were measured via Prospector, with typical uncertainties of . The wavelength coverage of rest-frame 3500-4300 Å provides coverage of the [OII] emission line, and Balmer absorption features like Hγ, Hδ, Hε, H7, H8, and the Ca K&H doublet absorption features.
REQUIEM
The REsolving QUIEscent Magnified (REQUIEM) galaxies survey targets eight strongly lensed quiescent galaxies at with demagnified stellar masses of (Akhshik et al. [5]; Newman et al. [96])1. We include 7 of these galaxies in our sample, including new dust continuum imaging for J1522-2535 (Figure 1; see § 2.3); we exclude SDSSJ0851 (Akhshik et al. [3]) as the ALMA dust continuum observations were not executed. The REQUIEM-2D data consist of deep (5–15 orbit) HST/WFC3 G141 grism spectroscopy (HST-GO-15663) together with archival HST WFC3 and ACS imaging spanning F336W–F160W and Spitzer/IRAC 3.6 and 4.5 µm photometry, all reduced with Grizli (Brammer [12]). Contaminating neighbors are modeled with GALFIT in F160W and subtracted, and fluxes are measured in “banana”-shaped object apertures that follow the arc’s axis of symmetry (Wuyts et al. [151]; Akhshik et al. [5]), an approach analogous to the GALFIT-based arc apertures used for COOL-LAMPS (§ 2.1.2).
ALMA Dust Continuum Imaging
The ALMA continuum observations used in this study were taken as part of program 2023.1.00430.S and combined with the archival programs 2018.1.00276.S and 2019.1.0027.S presented in Whitaker et al. [145]. The dust continuum observations in this program are centered at 226 GHz in Band 6, corresponding to an observed wavelength of 1.3 mm (∼600 µm at ) covering the 7.5 GHz bandwidth in the Time Division Mode (TDM). In order to make a global measurement of dust mass, lower resolution data is optimal. Given the compact nature of dust in quiescent galaxies from previous studies (e.g. Whitaker et al. [145]), the antenna configuration was chosen to obtain low spatial resolution dust continuum ranging from 1″–1.5″. The baselines range from 15.1 m to 313.7 m with 43–45 antennae. On source time was 6.8 hours.
The ALMA data are reduced and calibrated using CASA (McMullin et al. [83]) version 6.5.4.9. We use pipeline-calibrated visibility data for the continuum imaging, generated with the CASA task tclean. The pipeline cleaning routine is performed down to noise level with Briggs weighting and a robust parameter set to . With this weighting scheme, the median circularized beam in the dataset has a radius of 1.12″ (ranging from 1.04″ to 1.28″) and the typical RMS noise is 20 µJy/beam. For targets that are undetected in the pipeline imaging, we test whether cleaning in interactive mode with natural weighting () and –-tapering to 2″ (a physical scale of 16.5 kpc at , or ∼3 kpc assuming ) recovers signal in the case where the dust morphology is extended. We find no extended emission in these sources and therefore default to the pipeline imaging scheme as it offers better sensitivity.
Out of the eight targets observed in program 2023.1.00430.S (Figure 1), we identify three targets with S/N dust continuum detection: J1201-3952, J1553-0142, J2359-0208. We fit J1553-0142 with a Gaussian using CARTA and find that the FWHM is smaller than the beam size; therefore we conclude that this source is unresolved, whereas the dust emission in J1201-3952 and J2359-0208 is extended. J1201-3952 has a sufficiently high S/N () to be flagged by CASA as a self-calibration candidate.
Following Equation 7.6 of the ALMA Cycle 12 Technical Handbook, the maximum recoverable scale ranges from 10″ to 10.8″. Dust emission more extended than this maximum recoverable scale is not recovered by these observations; since the major axes of J1201-3952 and J2359-0208 are 8″ and 13″, we expect that we should recover most of the flux density from these sources. However, for J1201-3952 and J2359-0208, we test cleaning in interactive mode with natural weighting () and –-tapering to 2″ (a physical scale of 16.5 kpc at , or ∼3 kpc assuming ) and 3″ (a physical scale of 25 kpc at , or ∼5 kpc assuming ) to determine whether we are resolving out any flux. In the case of J2359-0208, we find that the flux density after using natural weighting increases by and natural weighting with --tapering to 2″ decreases the flux density 17%; therefore, we conclude that natural weighting is sufficient to recover extended emission. For J1201-3952, natural weighting and --tapering to 2″ increases the flux density ∼15%.
Methods and Analysis
Gravitational Lensing Magnification
As with the optical photometry and spectroscopy, the lens modeling of the COOL-LAMPS systems from ground-based imaging is described in full in Khullar et al. (in prep.). We briefly summarize the procedure here. Lens models are constructed with the parametric lens modeling algorithm Lenstool (Jullo et al. [60]), which identifies the best-fit and samples the posterior via Markov Chain Monte Carlo (MCMC). The mass distribution of each lens is parameterized as dual pseudo-isothermal ellipsoid halos (dPIE, Elíasdóttir et al. [42]), constrained by multiple-image systems identified in the imaging where available (otherwise in imaging).
Using the apertures defined from the optical imaging (§ 2.1), we determine the number and spatial extent of the lensed images that make up each primary arc, and use these to compute an effective magnification, , for the region of the source contributing to our photometry and spectroscopy. We define as the harmonic mean of the magnification over the pixels within the aperture (as is standard practice; e.g. see Sharon et al. [120]). By drawing 1000 posterior chains from the Lenstool output, we report the 16th, 50th, and 84th percentiles of the resulting distribution in Table 1; we note here that ground imaging-based lens models are severely limited in the ability to constrain multiple images at various redshifts and spectral energy distributions (SEDs), and hence the magnification uncertainties are likely biased and/or the estimated median value is biased.
Strong lensing is wavelength-independent, so quantities derived from spectral shape such as dust attenuation and stellar age are roughly independent of magnification, as is the specific star formation rate, in which the magnification cancels out. Absolute quantities such as stellar mass and star formation rate are scaled by . To propagate the uncertainties in both the stellar population fits and the lens models, we construct the delensed distributions by Monte Carlo sampling. First we make 1000 draws from the stellar mass and star formation rate posteriors, and then, independently, we draw from the magnification distribution. Each of these physical quantities is divided by the drawn (considering the magnification boost to be a multiplicative factor). The resulting delensed stellar masses and star formation rates are reported in Table 1. We note that the stellar mass of J1553-0142 may be overestimated due to uncertainty in the lensing model.
For the REQUIEM targets, we adopt the magnifications compiled in Table 4 of Akhshik et al. [5] and the references therein (estimated using the same process as COOL-LAMPS), with one exception. For MRG-M0138, we replace the pre-JWST value of [95, 107] with , derived from recent lens models that incorporate new JWST imaging and VLT/MUSE spectroscopic constraints [1, 43, 129].
Modeling Stellar Populations
COOL-LAMPS
We derive physical properties (e.g., stellar mass, SFR, age) and non-parametric SFHs for the COOL-LAMPS galaxies by jointly fitting the ground-based spectroscopy and the multi-band photometry (, or the subset available for each target) measured within the apertures described in § 2.1.2, using the Bayesian stellar population synthesis (SPS) framework prospector [69, 57]. Stellar populations are modeled with FSPS [26, 25] using MIST isochrones [22] and the MILES spectral library [112], assuming a Chabrier [19] initial mass function. Nebular line and continuum emission are included using the CLOUDY-based models embedded in FSPS [15, 14].
We adopt a non-parametric SFH with 10 fixed lookback-time bins and the continuity prior of Leja et al. [68], which is placed on the ratio of star formation rates in adjacent bins and favors smooth transitions between them. The first two bins span 0–30 Myr and 30–100 Myr, the last bin spans –, where is the age of the Universe at the observed redshift, and the remaining seven bins are spaced evenly in logarithmic time between 100 Myr and . Dust attenuation follows the Kriek & Conroy [65] law, with the diffuse optical depth at 5500 Å (, or equivalently ) and the power-law slope offset (, dust index) as free parameters; the strength of the 2175 Å bump is tied to [97, 65]. We adopt uniform priors of and [69, 68]. Stellar metallicity is a free parameter, with flat priors in log-space (based on the limits of the MILES spectral library within FSPS).
In our joint spectro-photometric fit, we utilise a fifth-order Chebyshev polynomial for relative and absolute flux-calibration vector, marginalizing over systematic wavelength-dependent differences between the spectroscopic and photometric fluxes. We also fit for the redshift within to allow for uncertainties in the wavelength calibration of the ground-based spectra. The posterior is sampled with dynesty [123]. ALMA observations are not included in the fits, so no energy-balance assumptions enter our fiducial SPS model.
Because magnification enters only as a normalization of the spectrophotometric data, the inferred SFH shape is independent of the lens model, while extensive quantities are corrected using the magnifications from § 3.1 (post-fitting). We primarily rely on ratios (e.g., specific SFR, dust-to-stellar mass ratio), for which the uncertain effect of magnification is minimized. Further details of the SFH modeling and its caveats are given in Sukay et al. [127] and Khullar et al. (in prep.).
REQUIEM
For the REQUIEM galaxies we adopt the SFHs of Akhshik et al. [5], derived with the requiem2d code [3], which jointly fits the multi-band HST photometry and the 2D G141 grism spectra in their native 2D space, with a multiplicative polynomial fit alongside the stellar population model. While Akhshik et al. [5] focus on spatially resolved stellar populations, we use only their global SFHs (summed over all spatial bins), which additionally incorporate the unresolved Spitzer/IRAC constraints, to match the integrated COOL-LAMPS measurements. As for COOL-LAMPS, the models use FSPS templates with a Chabrier [19] IMF and a non-parametric SFH with a continuity prior, here with 20 lookback-time bins. Dust attenuation follows the same Noll et al. [97]/Kriek & Conroy [65] parameterization, but with the birth-cloud (), diffuse (), and slope () parameters all free. Metallicity and priors are set by first fitting the photometry with Prospector-α [69], dividing the resulting posteriors into 16 template boxes, and refitting the box weights simultaneously with the SFH, such that the Prospector posteriors act as weakly informative priors. The posterior is sampled with the No-U-Turn Sampler [53] in pymc3 [111].
Comparison of Modeling Assumptions
Both subsamples are modeled with FSPS stellar populations, a Chabrier [19] IMF, non-parametric SFHs with a continuity prior, and a Kriek & Conroy [65] dust law. The main differences are the SFH time resolution (10 bins for COOL-LAMPS versus 20 for REQUIEM); the treatment of young-star attenuation (fixed at twice the diffuse attenuation versus a free birth-cloud component); the metallicity prior (a Gallazzi et al. [45] mass–metallicity prior versus a weakly informative prior derived from photometry-only fits); the sampler (dynesty versus NUTS); and the spectroscopic constraint (ground-based spectra versus HST/G141 grism spectra). Since non-parametric SFH constraints are prior-sensitive, particularly in the oldest bins [68], we compare the subsamples using derived quantities that are less sensitive to bin choice, namely the quenching timescale and the time elapsed since quenching, defined following Whitaker & Bezanson [142]. We note that Akhshik et al. [5] report stellar masses 0.1–0.5 dex higher than earlier parametric fits to the same galaxies [95, 80], an offset attributed both to the non-parametric SFHs and to their aperture photometry; the COOL-LAMPS masses should be similarly subject to the first of these effects.
Deriving Quenching Timescales
Based on the SFHs we obtain from SED fitting (§ 3.2), we derive a quenching timescale and the time since quenching for each galaxy as follows. We smooth the 10 bin SFH by taking the midpoints of each bin and interpolating. We then determine the sSFR at each time step by dividing the SFR by the stellar mass at that time. Integrating the SFH yields the total mass formed, whereas the sSFR should be normalized by the mass that survives, after mass loss from stellar evolution. We therefore correct the integrated mass formed downward by 0.2 dex, corresponding to a surviving mass fraction of ∼0.6 for a Chabrier [19] IMF [78]. Although the surviving mass fraction evolves with the age of the stellar population, it declines most rapidly within the first ∼1 Gyr after star formation and changes only slowly thereafter. Because and occur after the bulk of the stellar mass has formed, a constant correction is adequate for these timescales. It would be less reliable for any timescale derived from the sSFR history at earlier times, when the population is young and the surviving fraction is higher. To determine when a galaxy is quenched, we adopt a criteria of , where is the age of the Universe [98]. When the sSFR drops below this value, it has quenched; we define this time as . We further define when the sSFR is at 0.6 dex above this cutoff to be the time when the galaxy begins to transition off of the star-forming main sequence (SFMS), or . The time since quenching is then , or the difference between the time at observation and the time at quenching. Figure 2 demonstrates how we derive these times. We also test using the offset from the SFMS measured in Leja et al. [70] as our quenching threshold, and find that it is generally in close agreement to the cutoff at Gyr. However, at earlier times there is a sharp

Figure 2. The median smoothed SFH of J1116-1443 is shown (pink line), as well as the SFHs from individual chains and the original 10 bin SFH (black dashed line). The quenching threshold is defined by (red line), where is the age of the Universe; the transition time is defined as 0.6 dex above this relation (gray line). We note where , and are located according to these criteria.
drop in the SFMS likely due to extrapolation; therefore we adopt the criteria. To derive timescale errors, we draw 1000 SFH chains from the posterior distribution and calculate and for each chain. We adopt the median, 16th, and 84th percentile when reporting in § 5.
Measuring Dust Mass
We estimate dust mass assuming a modified black-body (e.g. Greve et al. [50]) with dust emissivity of 1.8 and a dust mass opacity coefficient [38, 27]. Following Magdis et al. [79], who measure dust temperatures of stacked quiescent galaxy samples at , we adopt a dust temperature of for COOL-LAMPS galaxies at and for J1522-2535 at . For the three detected sources, J1201-3952, J1553-0142, and J2359-0208, we measure the dust continuum flux within apertures derived from modeling optical data with GALFIT.
We propagate the uncertainty in the dust-to-stellar mass fraction () by incorporating errors from SED fitting in determining the stellar mass, in addition to the noise in the dust continuum data used to calculate the dust mass. While the lens models are largely based on ground-based data alone, and therefore relatively uncertain [120], we mainly consider relative fractions of physical parameters that largely mitigate lensing uncertainties (e.g., , sSFR). However, our uncertainty analysis does not account for a potentially significant source of systematic uncertainty in our measurement of dust mass, which is dust temperature. If we instead assume dust temperatures range between 15–35K, this could shift the dust-to-stellar mass fraction by dex. Robust constraints on dust temperature require additional data sampling the peak of the far-IR, which are challenging to acquire at cosmic noon for these quiescent systems. Ongoing work performing a joint optical to far-IR modeling of the MRG-M0138 strong lensing system assuming energy balance will start to make progress in this area (Medrano et al. in prep). To reflect the uncertainty in dust temperature, we add a systematic error bar to all plots that include dust mass.
Results
A Wide Range of Dust to Stellar Mass Ratios
In Figure 3, we show the ratio of dust mass to stellar mass, , as a function of redshift, measuring dust mass based on the calculations described in § 3.4 and stellar mass from SED fitting as described in § 3.2. As both dust mass and stellar mass are measured in the image plane, the magnification should approximately cancel in . However, differential magnification could bias if the dust and stars have different intrinsic distributions within the galaxies, since regions closer to the critical curve are more strongly magnified (e.g. Serjeant [117]). The optical and near-IR imaging is seeing-limited (–”) and the ALMA beam is ”. Hence, the stellar and dust emission are each measured over broadened profiles that average over different parts of the magnification map; at this moment, we cannot directly verify that the two components share the same effective magnifications, or are strictly co-spatial. We compare our sample with other studies of quiescent galaxies, with results from stacking from to ([48, 79, 8]), individual measurements at (Lorenzon et al. [74]) and (Spilker et al. [126]), and ([51, 16, 145, 130, 67, 137, 121, 126, 20]). For reference, we also include the Tacconi et al. [133] scaling relation, assuming a typical molecular gas-to-dust mass ratio, , of 100 and for the SFMS, as well as 0.9 dex below the SFMS (the threshold we use to determine time at quenching, see § 3.3), with

Figure 3. The dust fraction () of our sample of quiescent galaxies spans over two orders of magnitude, including objects that are among the dustiest and least dusty quiescent galaxies discovered. Points are color coded by sSFR, with tracks from the Tacconi et al. [133] scaling relation assuming a gas-to-dust ratio (GDR) of 100 and at the star-forming main sequence (solid line), 0.9 dex below the main sequence with a GDR=100 (dashed line), and 0.9 dex below the main sequence with a GDR=1000 (dotted line). We also include results from the literature ([51, 48, 79, 16, 61, 145, 130, 8, 67, 137, 74, 126, 20]). We add a systematic error bar (bottom right corner) representing the uncertainty in dust mass introduced by allowing dust temperature to vary from 15–35K.
a of 100 and 1000 to account for the wide variety of ratios observed in quiescent galaxies ([146, 126, 75]). When assuming , most individual measurements of from the literature that fall dex below the standard Tacconi et al. [133] relation are upper limits, while stacked results and a handful of individual detections are roughly consistent with the relation extrapolated to the defined quenching threshold. Two of the systems with 1.3mm detections in our sample (J1201-3952 and J2359-0208) have higher values than previously measured in lensed quiescent samples at higher redshifts ([145]) and are more comparable with both the stacked results ([79, 48, 8]) and the brightest individual dust detections ([51, 130, 67]) – however, these earlier studies rely on photometrically-selected quiescent galaxies. The dust masses of these two objects are consistent with the scaling relation between dust mass and redshift found for ALMA-detected star-forming galaxies ([33, 64]). J1201-3952 in particular may have the highest dust fraction yet measured in a spectroscopically-selected quiescent galaxy, offset 0.3 dex below the SFMS Tacconi et al. [133] relation. Only the source ALMA.14 Hayashi et al. [51] has a higher dust fraction, but as discussed by Lee et al. [67], its classification as a quiescent galaxy is based solely on optical/near-infrared photometry, which may significantly underestimate its true SFR; thus its classification as a quiescent galaxy is uncertain.
On the other hand, the remainder of the sample is fairly dust poor. J1553-0142 has the second lowest value of any quiescent galaxy with a solid dust detection, after MRG-M0138 from the lensed REQUIEM sample (Whitaker et al. [145]). Moreover, J1323-0343 and J1116-01443 are the two most stringent upper limits on , with values of order . In summary, our sample shows a wide diversity in ; additionally, aside from this study – there are only two other existing dust measurements in individual quiescent galaxies at (Hayashi et al. [51]; Spilker et al. [126]), so our work provides an 4.5 increase in the number of such systems.
Dust Properties with sSFR and Age
We plot dust fraction against specific star formation rate (sSFR) in Figure 4, revealing a mild positive association between and sSFR. We measure the Kendell correlation coefficient using the python package pymccorrelation (Privon et al. [103]), treating upper limits as left-censored points (Isobe et al. [56]), and find with a p-value of 0.08. There is significant scatter in sSFR – at , the dynamic range of values extends over two orders of magnitude. The large scatter in dust fraction may reflect differing timescales in dust destruction, complicated by varying times since quenching across the sample (see § 5.1). Two of the galaxies with dust detections at (J1201-3952, J2359-0208) have , similar to previous studies at , while the sSFR of J1553-0142 is lower (). The galaxies that are undetected tend to have lower values, with for three out of the four cases.

Figure 4. There is a mild association (Kendell ) between specific star formation rate (sSFR) and dust fraction (), and scatter over two order of magnitudes in dust fraction. We add a systematic error bar (bottom right corner) representing the uncertainty in dust mass introduced by allowing dust temperature to vary from 15-35K.
In Figure 5, we plot the mass-weighted stellar age as a function of redshift. Stellar ages taken from the literature are mass-weighted ([61, 130, 67, 121]), with the exception of REQUIEM galaxies, which are derived. from the time at which 50% of stellar mass has formed () [5]. Although it is difficult to surmise a trend given the large number of upper limits, appears to generally decrease with increasing stellar population age. We also calculate a Kendell correlation for age and dust fraction, and find with a -value of 0.6, indicating a positive association between age and dust fraction with low statistical significance. The lower redshift galaxies in our sample tend to have older ages (Gyr) relative to higher redshift galaxies (Gyr). This largely reflects that higher redshift galaxies have simply had less time to grow their stellar populations and quench, with slower quenching systems only emerging at later times – high redshift sources are generally caught closer to the epoch of quenching where residual dust is more likely (in the case where quiescent galaxies are the descendants of dusty star-forming galaxies).

Figure 5. Dust fraction () plotted as a function of age; we do not find a statistically significant association between these two properties. Points are color coded by redshift; lower redshift galaxies are older. We add a systematic error bar (bottom right corner) representing the uncertainty in dust mass introduced by allowing dust temperature to vary from 15-35K.
Star Formation Histories of COOL-LAMPS and REQUIEM Galaxies
To explore how the dust properties of our sample relate to their SFHs, we reconstruct how the star formation in these galaxies grew and subsequently shut down over cosmic time. In Figure 6, we plot the SFHs normalized by the total mass formed and compare to those derived for the higher redshift REQUIEM galaxies, with stellar mass formed calculated by integrating over the SFH. The details of the SED fitting, from which we derive the SFHs, can be found in § 3.2.

Figure 6. Left: The star formation histories normalized by total mass formed over the entire SFH; the SFRs of several COOL-LAMPS galaxies drop sharply within 2 Gyrs of formation. Two of the galaxies with dust detections (solid lines) show more extended or recent star formation relative to galaxies that are undetected in dust continuum (dashed lines). SFHs are color coded by dust-to-stellar mass fraction; for galaxies with non-detections in dust continuum, the dust fractions are upper limits. Right: The star formation histories normalized by total mass formed for REQUIEM galaxies exhibit a range of shapes, with most undergoing quenching about 2 Gyr after the Big Bang (at ).
For the COOL-LAMPS sample at (left, Figure 6), we find that two of the galaxies with dust detections (2359-0208 and J1201-3952) have more extended SFHs on average relative to dust-undetected galaxies and J1553-0142. The latter have steeper drops in SFR when the age of the Universe was less than 2 Gyr old. Unlike the rest of the sample, which quench star formation early without rejuvenation, J2359-0208 appears to have experienced a second burst of star formation starting when the Universe was 3 Gyr, with very recent and extreme quenching occurring over the last Myr. We note that the fit for J2359-0208 is uncertain due to a noisy spectrum therefore, the shape of the SFH is more uncertain relative to the rest of our sample. The uncertainties are folded into our measurements of age, sSFR, and stellar mass.
Figure 6 (right) shows that the majority of the REQUIEM galaxies (at ) cross the quenching threshold when the Universe was Gyr, close to the epoch of observation, with no evidence of residual cold dust remaining, though for a range of quenching timescales [5]. The most extended quenching timescale (MRG-M0138), more gradual than even the lower redshift COOL-LAMPS galaxies, has a tightly constrained dust detection at , with the dust residing only in the core (Whitaker et al. [145]). Similarly, the other REQUIEM galaxy with a dust detection, MRG-M2129, shows centrally-concentrated dust at a more moderate value, with the overall (s)SFR similarly elevated relative to the other galaxies in the sample. This stands in contrast to the COOL-LAMPS galaxies J2359-0208 and J1201-3952, where the cold dust is unambiguously extended and more closely follows the stellar light profile, which we discuss in more detail in § 5. The destruction is irrevocable in the case where it is coupled with similarly harsh ISM conditions, though can extend over several gigayears for more mild ISM conditions (Donevski et al. [35]).
The REQUIEM and COOL-LAMPS samples seem to collectively offer us a glimpse into the life cycle of dust post-quenching and an opportunity to empirically constrain dust destruction conditions and associated timescales. Figure 7 (left) shows the full sample, with the measurements of dust fraction relative to the time that has elapsed since crossing the quenching threshold for each of these systems (derived according to the methodology described in § 3.3). As one can also gather from Figure 6, the REQUIEM galaxies have all recently shut down star formation within the last Gyr, and often only a few hundred Myrs before the epoch of observation. Conversely, the COOL-LAMPS sample spans a wide range of time since quenching and a similarly wide range of dust fractions. While this sample is notably biased due to the nature of strong-lens system discovery, the combined set of REQUIEM and COOL-LAMPS galaxies broadens the parameter space we can explore across redshift, SFH shapes, and dust content, yielding some general trends that we discuss next.

Figure 7. Left: Lower redshift COOL-LAMPS galaxies quenched further in the past, and have dust fractions consistent with mild dust destruction or dust regrowth. Higher redshift REQUIEM galaxies are more recently quenched and are more consistent with harsh dust destruction. We add a systematic error bar (bottom right corner) representing the uncertainty in dust mass introduced by allowing dust temperature to vary from 15-35K. Right: The star formation histories for COOL-LAMPS and REQUIEM galaxies, grouped by where they fall on the time since quenching vs dust fraction plot. SFHs with dust regrowth/mild dust destruction are shown at the top, mild dust destruction in the middle, and harsh dust destruction on the bottom.
Discussion
It has been predicted that dust should be destroyed over rapid timescales after quenching (Li et al. 2019), although recent models suggest that this is highly dependent on the density and energetics of the local ISM, with dust regrowth possible in cold, dense, and less energetically disturbed environments (Lorenzon et al. [74]; Donevski et al. [35]). In Figure 7, we compare our observed dust fractions as a function of time since quenching to the model predictions of Donevski et al. [35]. The solid lines represent dust destruction trajectories when AGN feedback is included, whereas the dotted lines do not include AGN feedback. Turning on AGN feedback is associated with increased dust destruction (decreasing ) with increasing severity as more time elapses post-quenching. In effect, the AGN energetics likely keep the gas within the extended dark matter halos hot for prolonged periods, with thermal sputtering from hot electrons destroying the dust (Whitaker et al. [146]). This destruction is irrevocable in the case where it is coupled with similarly harsh ISM conditions, though can extend over several gigayears for more mild ISM conditions (Donevski et al. [35]).
Changing Modes of Dust Destruction with Redshift
We find that the higher redshift galaxies from the REQUIEM sample have quenched more recently, closer to the epoch of observation, and show dust fractions that align with harsh dust destruction, likely driven at least in part by AGN feedback. This is consistent with recent JWST observations that reveal neutral outflows from AGN in recently quenched systems at (Belli et al. [7]; Davies et al. [30]; Bugiani et al. [13]; Luo et al. [77]; Zhu et al. [153]), indicating the prevalence of AGN feedback at earlier times. For galaxies in this sample with dust detections, the dust is compact and centrally located, which may be the result of recent (in the last several hundred Myr) star formation in the core. While MRG-M2129 has a moderate dust fraction, its SFH is consistent with recent, rapid quenching. We therefore place it in harsh-destruction category, as it is observed shortly after quenching, before its residual core dust has had time to be destroyed. There is evidence that MRG-M2129 hosts an AGN (Newman et al. [95]; Man et al. [80]; Morishita et al. [88]), supporting that AGN feedback could play a role in the process of dust destruction.
By contrast, the lower redshift COOL-LAMPS galaxies are largely consistent with a mild dust destruction scenario at face value. That said, four COOL-LAMPS systems (J0112-2053, J1116-1443, J1323-0343, and J1509-3308) have stringent upper limits on that do not rule out more harsh ISM conditions, similar to the REQUIEM sample. These sources also show similar SFHs, with early formation and rapid quenching, observed Gyr after quenching. By contrast, the two COOL-LAMPS systems with extended dust emission are more unambiguously overlapping with the model tracks for mild ISM conditions and dust regrowth. The difference in the implied dust destruction timescales may reflect the changing ISM conditions and a shift in the dominant physical mechanisms driving quenching (see Whitaker & Bezanson [142], for a review). At , more intense feedback leads to more efficient dust destruction, while these processes become slower by .
Grain Regrowth and Dust Production after Quenching
In the cases of J1201-3952 and J2359-0208, their dust fractions exceed even the most optimistic dust destruction scenario with no AGN feedback. J2359-0208 has a SFH consistent with a post-starburst classification, with a relatively high sSFR that is near the quenching boundary (e.g. the sSFR is ∼0.9 dex below the SFMS), similar to MRG-M2129, although J2359-0208 also experienced a late rejuvenation followed by rapid quenching close to the epoch of observation (Figure 6). While both do not formally qualify as quiescent according to the criteria from 3.3, and as such we assign both a time since quenching of zero, their dust morphologies show a striking difference: MRG-M2129 retains compact dust in its core ([145, 88]), consistent with the harsh, AGN-driven destruction discussed in § 5.1, whereas J2359-0208 shows extended dust emission. J2359-0208 may appear consistent with dust growth because rejuvenation and stellar dust production combine, with fresh dust from the recent burst adding to the grain regrowth and/or TP-AGB dust production by the intermediate-age stellar population formed during rejuvenation. For the latter case, dust from TP-AGB stars should trace the spatial distribution of the stellar population formed ∼0.2–2 Gyr before observation, here the rejuvenation episode, rather than the overall stellar light. Residual dust that has not yet had enough time to deplete, on the other hand, would primarily be expected to reside in the core where higher gas densities shield grains and slow the destruction timescale ([106]). Distinguishing the dust distributions robustly will require higher-resolution continuum imaging than the present data provide, together with resolved stellar-population maps.
The dust excess in J1201-3952 and J2359-0208 relative to even the mildest destruction tracks may also point to grain growth in the ISM that is more efficient than in the fiducial models, for example if metals re-accrete onto grains on timescales shorter than ∼100 Myr ([35]). Rejuvenation, as observed in a minority of quiescent systems (e.g., [4, 21]) and predicted by simulations ([104, 63, 44]), provides a complementary channel, supplying fresh dust both directly and by enabling grain growth ([74]). However, both interpretations rely on the assumption that the far-IR emission is not powered by ongoing star formation.
Are Dust-Rich Systems Truly Quiescent?
An obvious concern is whether these dust-rich systems are truly quiescent, or whether they host dust-obscured star formation that is invisible in the rest-frame optical spectra used to derive their SFHs. Strong Balmer absorption and weak emission lines do not uniquely identify a quenched galaxy, as an obscured starburst can sit alongside, or be spatially offset from, an optically post-starburst stellar population (Schreiber et al. [114]). For example, among local E+A galaxies with molecular gas measurements, a subset (7 of 26) host obscured starbursts detectable only in the far-IR, some with starburst-like SFRs (Baron et al. [6]), although infrared luminosities alone can substantially overestimate SFRs in galaxies below the main sequence ([138, 147]). In the FIREbox simulations, bursty, interaction-driven SFHs produce gas-rich post-starburst “impostors” that are still forming stars at rates typical of star-forming galaxies (Cenci et al. [18]).
However, a substantial dust reservoir, as we see in J1201-3952 and J2359-0208, does not by itself imply star formation. Local post-starbursts host compact, turbulent molecular reservoirs that form stars roughly an order of magnitude less efficiently than starbursts of similar gas surface density (Smercina et al. [122]). Furthermore, at , the IR SEDs of most post-starbursts can be reproduced by dust heated by their evolved stellar populations without buried star formation (Setton et al. [118]), although spatially resolved Pa shows that the most CO-luminous systems are still forming stars behind optically thick dust (Setton et al. [119]). Our ALMA Band 6 continuum samples the Rayleigh–Jeans tail and therefore constrains dust mass rather than the infrared luminosity; it cannot by itself rule out obscured star formation, since inferred from a single long-wavelength point depends steeply on the assumed dust temperature.
Two lines of evidence nonetheless argue against a hidden starburst. First, the dust fractions of J1201-3952 and J2359-0208 remain 0.3-0.6 dex below those of main-sequence galaxies at similar stellar mass and redshift, when assuming a value of of 100 (Figure 3), whereas an obscured starburst sufficient to place them on the main sequence would typically be accompanied by a dust reservoir comparable to those of star-forming galaxies. Second, the dust emission in both systems is extended and broadly co-spatial with the stars, rather than arising from a compact nuclear or spatially offset clump as seen in misclassified “quiescent” galaxies (Schreiber et al. [114]). We note, however, that at the resolution of our ALMA data (-), a compact obscured component contributing up to 39% and 28% of the continuum flux could remain unresolved within the extended emission for J1201-3952 and J2359-0208 respectively (see also Sun et al. [128]), so this argument constrains only an obscured starburst that dominates the dust emission.
We therefore cannot rule out modest obscured star formation in these two systems. If present, it would place them in a still-quenching or rejuvenating phase
rather than true quiescence, consistent with the rejuvenation already seen in the SFH of J2359-0208. The dust-poor systems, however, are unaffected, as their ultra-deep continuum limits themselves make significant obscured star formation unlikely. Multi-band ALMA continuum spanning the Rayleigh–Jeans tail and approaching the far-IR peak would jointly constrain and the emissivity index , removing the dominant uncertainty in converting the far-IR emission to either a dust mass or an obscured SFR (Cochrane et al. [24]). Independently, rest-frame near-IR hydrogen recombination lines (e.g., Pa, Pa), which are far less sensitive to attenuation (e.g., Setton et al. [119]), and mid-IR PAH emission would directly test for obscured star formation, though PAHs in quiescent systems can also be excited by evolved stellar populations.
Dust Yields from AGB stars and SN
Taking the quiescent interpretation at face value, we quantify the dust yield from AGB stars and Type II supernovae (SN) required to account for the observed dust masses, following Michalowski et al. [84] and Lee et al. [67]. We assume a Salpeter IMF (Salpeter [110]) and that stellar mass ranges from for Type II SNe and for AGB stars. Ventura et al. [141] and Nanni et al. [91] report maximum dust yield for AGB stars 0.25 , although estimated dust yields vary significantly with estimated initial mass and metallicity; we adopt a maximum dust yield of 0.03 per AGB star. We adopt a maximum dust yield for Type II SN of 0.8 from SN1987A (dust mass measurements vary for this SN from 0.23–0.8 , see Indebetouw et al. [55]; Dwek & Arendt [39]; Matsuura et al. [82]), although once sufficient time has passed for the reverse shock to destroy dust grains, it is predicted the dust yield from Type II SN should drop to (Bocchio et al. [11]). Figure 8 shows the dust yield per star assuming that all of the dust mass comes from either Type II SN or AGB stars. In most cases, the dust mass that we observe could be explained entirely by either Type II SN or AGB stars. However, the two dust-rich sources, J2359-0208 and J1201-3952, approach or exceed the theoretical limits when attributing the dust to one mechanism alone. In particular, J1201-3952 falls on the limit for AGB stars and sits higher than models that predict a dust yield of for Type II SN when including dust destruction from SN shocks. These dust yield estimates therefore suggest efficient post-quenching grain regrowth is possible in 2 out of the 11 systems (). This is lower than the of simulated quiescent galaxies that maintain roughly constant dust content for Gyr after quenching in SIMBA (Lorenzon et al. [74]), but is consistent with their prediction that such systems are more common at lower stellar masses than those investigated in this paper (see also Donevski et al. [35]).

Figure 8. The dust yield per star for either AGB stars or Type II SN to account for the total dust mass approaches or exceeds the theoretical limit for several quiescent galaxies. Results from this study are outlined in black.
The dust yield calculation treats each production channel as integrated over the full stellar population, but the rate of AGB dust injection at the epoch of observation depends on the shape of the SFH. Stars of – reach the AGB between Myr and – Gyr after they form, and the thermally pulsing AGB phase dominates the near-IR light of stellar populations aged – Gyr (Maraston [81]); since these same stars are efficient dust producers (e.g., Nanni et al. [91, 92]), AGB dust injection should likewise be strongest at intermediate ages. The SFHs in Figure 6 show a qualitative trend in this direction. The dust-poor COOL-LAMPS systems formed nearly all of their stellar mass in a single burst within the first Gyr, and are observed several gigayears later, when the stars now reaching the AGB fall near or below the threshold for significant dust production. By contrast, the two dust-rich COOL-LAMPS systems formed their stars over an extended period of several gigayears, with low-level star formation continuing close to the epoch of observation (through a late rejuvenation in J2359-0208, and through a prolonged decline along the quiescence threshold in J1201-3952). Both therefore host stellar populations that are still feeding the TP-AGB phase. The most dust-rich REQUIEM system likewise formed a large fraction of its mass within the last Gyr before observation. In the REQUIEM sample, however, recent star formation and a short time since quenching coincide, so residual dust and fresh AGB dust cannot be separated. The COOL-LAMPS systems break this degeneracy, with J1201-3952 quenched Gyr before observation yet has the highest in the sample, suggesting that its extended SFH, rather than a recent quenching event, sustains its dust reservoir. Quantifying this trend requires SFHs with sub-gigayear precision at these ages, which can be tested further with future rest-frame near- to mid-IR spectroscopy.
Conclusion
In this paper, we combine new ALMA Band 6 dust continuum observations of eight strongly lensed, spectroscopically-selected quiescent galaxies at – from the COOL-LAMPS and REQUIEM surveys with five archival REQUIEM measurements (Whitaker et al. [145]; Akhshik et al. [5]). This sample is the largest set of individually targeted quiescent galaxies at cosmic noon with both deep dust continuum constraints and spectroscopically derived SFHs. Our key findings are:
The dust-to-stellar mass ratio spans more than two orders of magnitude, from to , including some of the most stringent upper limits to date and J1201-3952, which may have the highest yet measured in a spectroscopically-selected quiescent galaxy. Our sample increases the number of individual dust measurements in quiescent galaxies at by a factor of 4.5.
Compared with the models of Donevski et al. [35], the recently quenched REQUIEM galaxies and the early-forming, rapidly quenched COOL-LAMPS galaxies are consistent with harsh dust destruction, likely driven by AGN feedback, whereas the more slowly quenched COOL-LAMPS systems with extended dust are consistent with mild destruction or dust regrowth. This suggests that dust survival depends on how galaxies quench, and possibly that the dominant quenching mode evolves between and .
The two most dust-rich galaxies, J1201-3952 and J2359-0208, exceed even the mildest model destruction tracks, and their dust masses approach or exceed the maximum yields from AGB stars or Type II supernovae alone. Their extended SFHs, including a late rejuvenation in J2359-0208, point to efficient ISM grain growth and/or ongoing TP-AGB dust production by intermediate-age stellar populations.
These interpretations carry two main caveats: the unknown dust temperature introduces a dex systematic uncertainty in , and our single-band, continuum cannot exclude modest obscured star formation in the two dust-rich systems.
While calibrations of dust continuum as a molecular gas tracer remain uncertain in quiescent galaxies [126], this study illustrates the value of dust continuum for studying quenching and feedback. The dust content of quiescent galaxies at cosmic noon varies by more than two orders of magnitude, consistent with a picture in which the mode of quenching (slow versus rapid) and the strength of feedback (mild versus harsh) have lasting consequences for their dust. Three observational steps would test this picture most directly. First, the multi-band and higher-resolution ALMA continuum described in § 5 would both constrain dust masses and separate compact residual dust from extended, stellar-produced dust. Second, JWST rest-frame mid-IR spectroscopy would directly trace circumstellar dust around AGB stars, testing the contribution of late-stage stellar dust production (e.g., [35, 119]). Third, CO observations would test whether dust traces the remaining molecular gas, given evidence that in quiescent galaxies can be unusually high ([126]).
Strong lensing currently makes these measurements feasible for individual galaxies, bringing intrinsically faint dust continuum within reach of ALMA and beginning to resolve the dust relative to the stars, the key discriminant between residual core dust and dust produced by the stellar populations. Looking further ahead, the recently selected PRobe far-Infrared Mission for Astrophysics (PRIMA; [47]) will sample the far-IR peak of cold dust emission for statistical samples of quiescent galaxies, making it possible to map how dust survival depends on quenching pathway across the population rather than galaxy by galaxy.
Acknowledgements
The University of Massachusetts Amherst acknowledges that it was founded and built on the unceded homelands of the Pocumtuc Nation on the land of the Norrwutuck community. We recognize and honor the ongoing presence and stewardship of Indigenous peoples. As a Land Grant institution, UMass Amherst also occupies land obtained through the Morrill Act of 1862, which expropriated Indigenous territories to support the development of public universities. We recognize this historical context and the responsibilities it entails.
We also wish to acknowledge the many individuals whose labor sustains the institutions that make this research possible, including the janitorial, facility maintenance, technical, and administrative support staff at the University of Massachusetts, Amherst, whose contributions are often unrecognized but essential to the functioning of the scientific enterprise.
SB thanks Samantha Berek for helpful discussion. SB and KEW acknowledge support from the National Science Foundation under CAREER Award No. AST-2236773. GK would like to thank the Baum Grant and Fellowship at the University of Washington for support during this work, as well as the ALMA Ambassador Program (administered by NAASC and NRAO). GK acknowledges the International Space Science Institute (ISSI), Bern, for their hospitality, financial support and collaboration during the time of writing this manuscript, as part of their Visiting Scientist program. GK would also like to thank the DiRAC Institute in the Department of Astronomy at the University of Washington. The DiRAC Institute is supported through generous gifts from the Charles and Lisa Simonyi Fund for Arts and Sciences, Janet and Lloyd Frink, and the Washington Research Foundation. This research was supported in part by the University of Pittsburgh Center for Research Computing, RRI:DCR_022735, through the resources provided. Specifically, this work used the H2P/MPI cluster, which is supported by NSF award number OAC-2117681. JAD acknowledges funding from the Dunlap Institute, funded through an endowment established by the David Dunlap family and the University of Toronto.
A subset of spectra in the COOL-LAMPS program are observations made with the Nordic Optical Telescope, owned in collaboration by the University of Turku and Aarhus University, and operated jointly by Aarhus University, the University of Turku and the University of Oslo, representing Denmark, Finland and Norway, the University of Iceland and Stockholm University at the Observatorio del Roque de los Muchachos, La Palma, Spain, of the Instituto de Astrofisica de Canarias. The NOT data were obtained under program ID P61-031. The data presented here were obtained [in part] with ALFOSC, which is provided by the Instituto de Astrofisica de Andalucia (IAA) under a joint agreement with the University of Copenhagen and NOT.
References
- [1]Acebron, A., Bergamini, P., Rosati, P., et al. 2025, A&A, 699, doi: 10.1051/0004-6361/202554468DOI
- [3]Akhshik, M., Whitaker, K. E., Brammer, G., et al. 2020, ApJ, 900, 184, doi: 10.3847/1538-4357/abac62DOI
- [4]Akhshik, M., Whitaker, K. E., Leja, J., et al. 2021, ApJL, 907, L8, doi: 10.3847/2041-8213/abd416DOI
- [5]Akhshik, M., Whitaker, K. E., Leja, J., et al. 2023, ApJ, 943, 179, doi: 10.3847/1538-4357/aca677DOI
- [6]Baron, D., Netzer, H., French, K. D., et al. 2023, MNRAS, 524, 2741, doi: 10.1093/mnras/stad1792DOI
- [7]Belli, S., Park, M., Davies, R. L., et al. 2024, Nature, 630, 54, doi: 10.1038/s41586-024-07412-1DOI
- [8]Blánquez-Sesé, D., Gómez-Guijarro, C., Magdis, G. E., et al. 2023, A&A, 674, A166, doi: 10.1051/0004-6361/202345977
- [9]Bluck, A. F. L. 2026a, arXiv e-prints, arXiv:2608.07221, doi: 10.48550/arXiv.2608.07221arxiv.org/abs/2608.07221
- [11]Bocchio, M., Marassi, S., Schneider, R., et al. 2016, A&A, 587, A157, doi: 10.1051/0004-6361/201527432DOI
- [12]Brammer, G. 2019, Grizli: Grism redshift and line analysis software, Astrophysics Source Code Library, record ascl:1905.001
- [13]Bugiani, L., Belli, S., Park, M., et al. 2025, ApJ, 981, 25, doi: 10.3847/1538-4357/adaeafDOI
- [14]Byler, N. 2018, cloudyFSPS, 1.0.0, Zenodo, doi: 10.5281/zenodo.1156412. https://doi.org/10.5281/zenodo.1156412
- [15]Byler, N., Dalcanton, J. J., Conroy, C., & Johnson, B. D. 2017, ApJ, 840, 44, doi: 10.3847/1538-4357/aa6c66DOI
- [16]Caliendo, J. N., Whitaker, K. E., Akhshik, M., et al. 2021, ApJL, 910, L7, doi: 10.3847/2041-8213/abe132DOI
- [17]Carnall, A. C., McLure, R. J., Dunlop, J. S., & Davé, R. 2018, MNRAS, 480, 4379, doi: 10.1093/mnras/sty2169DOI
- [18]Cenci, E., Feldmann, R., Wellons, S., et al. 2026, MNRAS, 551, stag1492, doi: 10.1093/mnras/stag1492DOI
- [19]Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392DOI
- [20]Chang, W., Wilson, G., Forrest, B., et al. 2026, arXiv e-prints, arXiv:2601.22844, doi: 10.48550/arXiv.2601.22844DOI
- [21]Chauke, P., van der Wel, A., Pacifici, C., et al. 2019, ApJ, 877, 48, doi: 10.3847/1538-4357/ab164dDOI
- [22]Choi, J., Dotter, A., Conroy, C., et al. 2016, ApJ, 823, 102, doi: 10.3847/0004-637X/823/2/102DOI
- [23]Chworowsky, K., Finkelstein, S. L., Boylan-Kolchin, M., et al. 2024, AJ, 168, 113, doi: 10.3847/1538-3881/ad57c1DOI
- [24]Cochrane, R. K., Hayward, C. C., & Anglés-Alcázar, D. 2022, ApJL, 939, L27, doi: 10.3847/2041-8213/ac951dDOI
- [25]Conroy, C., & Gunn, J. E. 2010, FSPS: Flexible Stellar Population Synthesis, Astrophysics Source Code Library, record ascl:1010.043. http://ascl.net/1010.043
- [26]Conroy, C., Gunn, J. E., & White, M. 2009, ApJ, 699, 486, doi: 10.1088/0004-637X/699/1/486DOI
- [27]da Cunha, E., Hodge, J. A., Casey, C. M., et al. 2021, ApJ, 919, 30, doi: 10.3847/1538-4357/ac0ae0DOI
- [28]Daddi, E., Renzini, A., Pirzkal, N., et al. 2005, ApJ, 626, 680, doi: 10.1086/430104DOI
- [29]Davidzon, I., Ilbert, O., Laigle, C., et al. 2017, A&A, 605, A70, doi: 10.1051/0004-6361/201730419DOI
- [30]Davies, R. L., Belli, S., Park, M., et al. 2024, MNRAS, 528, 4976, doi: 10.1093/mnras/stae327DOI
- [31]D’Eugenio, F., Pérez-González, P. G., Maiolino, R., et al. 2024, Nature Astronomy, 8, 1443, doi: 10.1038/s41550-024-02345-1DOI
- [32]Dey, A., Schlegel, D. J., Lang, D., et al. 2019, AJ, 157, 168, doi: 10.3847/1538-3881/ab089dDOI
- [33]Donevski, D., Lapi, A., Małek, K., et al. 2020, A&A, 644, A144, doi: 10.1051/0004-6361/202038405DOI
- [34]Donevski, D., Damjanov, I., Nanni, A., et al. 2023, A&A, 678, A35, doi: 10.1051/0004-6361/2023246066
- [35]Donevski, D., Nanni, A., Whitaker, K. E., et al. 2026, arXiv e-prints, arXiv:2608.27571, doi: 10.48550/arXiv.2608.27571DOI
- [36]D’Onofrio, V. R., Spilker, J. S., Bezanson, R., et al. 2025, ApJ, 990, 166, doi: 10.3847/1538-4357/adf5bdDOI
- [37]Drlica-Wagner, A., Carlin, J. L., Nidever, D. L., et al. 2021, ApJS, 256, 2, doi: 10.3847/1538-4365/ac079d
- [38]Dunne, L., Eales, S., Edmunds, M., et al. 2000, MNRAS, 315, 115, doi: 10.1046/j.1365-8711.2000.03386.xDOI
- [39]Dwek, E., & Arendt, R. G. 2015, ApJ, 810, 75, doi: 10.1088/0004-637X/810/1/75DOI
- [40]Dwek, E., & Scalo, J. M. 1980, ApJ, 239, 193, doi: 10.1086/158100DOI
- [41]Edward, A. H., Antwi-Danso, J., Muzzin, A., et al. 2026, arXiv e-prints, arXiv:2605.27555, doi: 10.48550/arXiv.2605.27555DOI
- [42]Elíasdóttir, Á., Limousin, M., Richard, J., et al. 2007, arXiv e-prints, arXiv:0710.5636. https://arxiv.org/abs/0710.5636
- [43]Ertl, S., Suyu, S. H., et al. 2025, A&A, 702, doi: 10.1051/0004-6361/202554477DOI
- [44]Fortuné, S., Remus, R.-S., Kimmig, L. C., Burkert, A., & Dolag, K. 2025, A&A, 704, A185, doi: 10.1051/0004-6361/202554805DOI
- [45]Gallazzi, A., Charlot, S., Brinchmann, J., White, S. D. M., & Tremonti, C. A. 2005, MNRAS, 362, 41, doi: 10.1111/j.1365-2966.2005.09321.xDOI
- [46]Garg, A., Stubbs, C. W., Challis, P., et al. 2007, AJ, 133, 403, doi: 10.1086/510118DOI
- [47]Glenn, J., Meixner, M., Bradford, C. M., et al. 2025, Journal of Astronomical Telescopes, Instruments, and Systems, 11, 031628, doi: 10.1117/1.JATIS.11.3.031628DOI
- [48]Gobat, R., Daddi, E., Magdis, G., et al. 2018, Nature Astronomy, 2, 239, doi: 10.1038/s41550-017-0352-5DOI
- [49]Gomez, H. L., Baes, M., Cortese, L., et al. 2010, A&A, 518, L45, doi: 10.1051/0004-6361/201014530DOI
- [50]Greve, T. R., Vieira, J. D., Weiß, A., et al. 2012, ApJ, 756, 101, doi: 10.1088/0004-637X/756/1/101DOI
- [51]Hayashi, M., Tadaki, K.-i., Kodama, T., et al. 2018, ApJ, 856, 118, doi: 10.3847/1538-4357/aab3e7DOI
- [52]Hirashita, H., & Nozawa, T. 2017, Planet. Space Sci., 149, 45, doi: 10.1016/j.pss.2017.01.009DOI
- [53]Homan, M. D., & Gelman, A. 2014, J. Mach. Learn. Res., 15, 1593–1623
- [54]Ilbert, O., McCracken, H. J., Le Fèvre, O., et al. 2013, A&A, 556, A55, doi: 10.1051/0004-6361/201321100DOI
- [55]Indebetouw, R., Matsuura, M., Dwek, E., et al. 2014, ApJL, 782, L2, doi: 10.1088/2041-8205/782/1/L2DOI
- [56]Isobe, T., Feigelson, E. D., & Nelson, P. I. 1986, ApJ, 306, 490, doi: 10.1086/164359DOI
- [57]Johnson, B. D., Leja, J., Conroy, C., & Speagle, J. S. 2021, ApJS, 254, 22, doi: 10.3847/1538-4365/abef67DOI
- [58]Jones, A. P., Tielens, A. G. G. M., & Hollenbach, D. J. 1996, ApJ, 469, 740, doi: 10.1086/177823DOI
- [59]Jones, A. P., Tielens, A. G. G. M., Hollenbach, D. J., & McKee, C. F. 1994, ApJ, 433, 797, doi: 10.1086/174689DOI
- [60]Jullo, E., Kneib, J. P., Limousin, M., et al. 2007, New Journal of Physics, 9, 447, doi: 10.1088/1367-2630/9/12/447DOI
- [61]Kalita, B. S., Daddi, E., D’Eugenio, C., et al. 2021, ApJL, 917, L17, doi: 10.3847/2041-8213/ac16dcDOI
- [62]Khullar, G., Gozman, K., Lin, J. J., et al. 2021, ApJ, 906, 107, doi: 10.3847/1538-4357/abcb86DOI
- [63]Kimmig, L. C., Remus, R.-S., Seidel, B., et al. 2025, ApJ, 979, 15, doi: 10.3847/1538-4357/ad9472DOI
- [64]Kokorev, V. I., Magdis, G. E., Davidzon, I., et al. 2021, ApJ, 921, 40, doi: 10.3847/1538-4357/ac18ceDOI
- [65]Kriek, M., & Conroy, C. 2013, ApJL, 775, L16, doi: 10.1088/2041-8205/775/1/L16DOI
- [66]Kriek, M., van Dokkum, P. G., Franx, M., et al. 2006, ApJ, 645, 44, doi: 10.1086/504103DOI
- [67]Lee, M. M., Steidel, C. C., Brammer, G., et al. 2024, MNRAS, 527, 9529, doi: 10.1093/mnras/stad3718DOI
- [68]Leja, J., Carnall, A. C., Johnson, B. D., Conroy, C., & Speagle, J. S. 2019, ApJ, 876, 3, doi: 10.3847/1538-4357/ab133cDOI
- [69]Leja, J., Johnson, B. D., Conroy, C., van Dokkum, P. G., & Byler, N. 2017, ApJ, 837, 170, doi: 10.3847/1538-4357/aa5ffeDOI
- [70]Leja, J., Speagle, J. S., Ting, Y.-S., et al. 2022, ApJ, 936, 165, doi: 10.3847/1538-4357/ac887dDOI
- [71]Leśniewska, A., Michalowski, M. J., Gall, C., et al. 2023, ApJ, 953, 27, doi: 10.3847/1538-4357/acdcfcDOI
- [73]Lianou, S., Xilouris, E., Madden, S. C., & Barmby, P. 2016, MNRAS, 461, 2856, doi: 10.1093/mnras/sty1467
- [74]Lorenzon, G., Donevski, D., Lisiecki, K., et al. 2025a, A&A, 693, A118, doi: 10.1051/0004-6361/202450393DOI
- [75]Lorenzon, G., Donevski, D., Man, A. W. S., et al. 2025b, ApJL, 995, L63, doi: 10.3847/2041-8213/ae226cDOI
- [77]Luo, Y., Rowlands, K., Alatalo, K., et al. 2026, ApJ, 1000, 24, doi: 10.3847/1538-4357/ae4559
- [78]Madau, P., & Dickinson, M. 2014, ARA&A, 52, 415, doi: 10.1146/annurev-astro-081811-125615DOI
- [79]Magdis, G. E., Gobat, R., Valentino, F., et al. 2021, A&A, 647, A33, doi: 10.1051/0004-6361/202039280DOI
- [80]Man, A. W. S., Zabl, J., Brammer, G. B., et al. 2021, ApJ, 919, 20, doi: 10.3847/1538-4357/ac0ae3DOI
- [81]Maraston, C. 2005, MNRAS, 362, 799, doi: 10.1111/j.1365-2966.2005.09270.xDOI
- [82]Matsuura, M., Dwek, E., Barlow, M. J., et al. 2015, ApJ, 800, 50, doi: 10.1088/0004-637X/800/1/50DOI
- [83]McMullin, J. P., Waters, B., Schiebel, D., Young, W., & Golap, K. 2007, in Astronomical Society of the Pacific Conference Series, Vol. 376, Astronomical Data Analysis Software and Systems XVI, ed. R. A. Shaw, F. Hill, & D. J. Bell, 127
- [84]Michalowski, M. J., Watson, D., & Hjorth, J. 2010, ApJ, 712, 942, doi: 10.1088/0004-637X/712/2/942DOI
- [85]Michalowski, M. J., Hjorth, J., Gall, C., et al. 2019, A&A, 632, A43, doi: 10.1051/0004-6361/201936055DOI
- [86]Michalowski, M. J., Gall, C., Hjorth, J., et al. 2024, ApJ, 964, 129, doi: 10.3847/1538-4357/ad1b52DOI
- [87]Miknaitis, G., Pignata, G., Rest, A., et al. 2007, ApJ, 666, 674, doi: 10.1086/519986DOI
- [88]Morishita, T., Abdurro’uf, Hirashita, H., et al. 2022, ApJ, 938, 144, doi: 10.3847/1538-4357/ac9055DOI
- [89]Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013a, ApJ, 777, 18, doi: 10.1088/0004-637X/777/1/18DOI
- [90]Muzzin, A., Marchesini, D., Stefanon, M., et al. 2013b, ApJS, 206, 8, doi: 10.1088/0067-0049/206/1/8DOI
- [91]Nanni, A., Bressan, A., Marigo, P., & Girardi, L. 2013, MNRAS, 434, 2390, doi: 10.1093/mnras/stt1175DOI
- [92]Nanni, A., Marigo, P., Girardi, L., et al. 2018, MNRAS, 473, 5492, doi: 10.1093/mnras/stx2641DOI
- [93]Narayanan, D., Torrey, P., Parente, M., et al. 2026, The Open Journal of Astrophysics, 9, 72863, doi: 10.33332/001c.172863
- [94]Nelson, E. J., van Dokkum, P. G., Förster Schreiber, N. M., et al. 2016, ApJ, 828, 27, doi: 10.3847/0004-637X/828/1/27DOI
- [95]Newman, A. B., Belli, S., Ellis, R. S., & Patel, S. G. 2018, ApJ, 862, 125, doi: 10.3847/1538-4357/aacd4dDOI
- [96]Newman, A. B., Gu, M., Belli, S., et al. 2026, Science, 392, 1065, doi: 10.1126/science.adx5816DOI
- [97]Noll, S., Pierini, D., Cimatti, A., et al. 2009, A&A, 499, 69, doi: 10.1051/0004-6361/200811526DOI
- [98]Pacifici, C., Kassin, S. A., Weiner, B. J., et al. 2016, ApJ, 832, 79, doi: 10.3847/0004-637X/832/1/79DOI
- [99]Parente, M., Narayanan, D., Torrey, P., et al. 2026, arXiv e-prints, arXiv:2609.35533, doi: 10.48550/arXiv.2609.35533DOI
- [100]Park, M., Belli, S., Conroy, C., et al. 2024, ApJ, 976, 72, doi: 10.3847/1538-4357/ad7e15DOI
- [101]Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W. 2002, AJ, 124, 266, doi: 10.1086/340952DOI
- [102]Persson, S. E., Barkhouser, R., Birk, C., et al. 2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7014, Ground-based and Airborne Instrumentation for Astronomy II, 70142V
- [103]Privon, G. C., Ricci, C., Aalto, S., et al. 2020, ApJ, 893, 149, doi: 10.3847/1538-4357/ab8015DOI
- [104]Remus, R.-S., & Kimmig, L. C. 2025, ApJ, 982, 30, doi: 10.3847/1538-4357/ad8b4bDOI
- [105]Rest, A., Stubbs, C., Becker, A. C., et al. 2005, ApJ, 634, 1103, doi: 10.1086/497060DOI
- [106]Richie, H. M., & Schneider, E. E. 2026, ApJ, 996, 17, doi: 10.3847/1538-4357/ae12e8DOI
- [107]Rodney, S. A., Brammer, G. B., Pierel, J. D. R., et al. 2021, Nature Astronomy, 5, 1118, doi: 10.1038/s41550-021-01450-9DOI
- [108]Rowlands, K., Dunne, L., Maddox, S., et al. 2012, MNRAS, 419, 2545, doi: 10.1111/j.1365-2966.2011.19905.xDOI
- [109]Salim, S., & Narayanan, D. 2020, ARA&A, 58, 529, doi: 10.1146/annurev-astro-032620-021933DOI
- [110]Salpeter, E. E. 1955, ApJ, 121, 161, doi: 10.1086/145971DOI
- [111]Salvatier, J., Wiecki, T. V., & Fonnesbeck, C. 2016, PyMC3: Python probabilistic programming framework, Astrophysics Source Code Library, record ascl:1610.016. http://ascl.net/1610.016
- [112]Sánchez-Blázquez, P., Peletier, R. F., Jiménez-Vicente, J., et al. 2006, MNRAS, 371, 703, doi: 10.1111/j.1365-2966.2006.10699.x
- [113]Scholtz, J., D’Eugenio, F., Maiolino, R., et al. 2026, Nature Astronomy, 10, 431, doi: 10.1038/s41550-025-02751-zDOI
- [114]Schreiber, C., Labbé, I., Glazebrook, K., et al. 2018, A&A, 611, A22, doi: 10.1051/0004-6361/201731917DOI
- [115]Scoville, N., Aussel, H., Sheth, K., et al. 2014, ApJ, 783, 84, doi: 10.1088/0004-637X/783/2/84DOI
- [116]Scoville, N., Sheth, K., Aussel, H., et al. 2016, ApJ, 820, 83, doi: 10.3847/0004-637X/820/2/83DOI
- [117]Serjeant, S. 2012, MNRAS, 424, 2429, doi: 10.1111/j.1365-2966.2012.20761.xDOI
- [118]Setton, D. J., Spilker, J. S., Bezanson, R., et al. 2025, AJ, 170, 351, doi: 10.3847/1538-3881/ae1607DOI
- [119]Setton, D. J., Greene, J. E., Bezanson, R., et al. 2026, arXiv e-prints, arXiv:2609.31307, doi: 10.48550/arXiv.2609.31307DOI
- [120]Sharon, K., Bayliss, M. B., Dahle, H., et al. 2020, ApJS, 247, 12, doi: 10.3847/1538-4365/ab5f13DOI
- [121]Siegel, J. C., Setton, D. J., Greene, J. E., et al. 2025, ApJ, 985, 125, doi: 10.3847/1538-4357/adc7bc7
- [122]Smercina, A., Smith, J.-D. T., French, K. D., et al. 2022, ApJ, 929, 154, doi: 10.3847/1538-4357/ac5d5fDOI
- [123]Speagle, J. S. 2020, MNRAS, 493, 3132, doi: 10.1093/mnras/staa278DOI
- [124]Spilker, J., Bezanson, R., Barisic, I., et al. 2018, ApJ, 860, 103, doi: 10.3847/1538-4357/aac438DOI
- [125]Spilker, J. S., Sueß, K. A., Setton, D. J., et al. 2022, ApJL, 936, L11, doi: 10.3847/2041-8213/ac75eaDOI
- [126]Spilker, J. S., Whitaker, K. E., Narayanan, D., et al. 2025, ApJL, 993, L40, doi: 10.3847/2041-8213/ae14d6DOI
- [127]Sukay, E., Khullar, G., Gladders, M. D., et al. 2022, ApJ, 940, 42, doi: 10.3847/1538-4357/ac9974DOI
- [128]Sun, F., Egami, E., Fujimoto, S., et al. 2022, ApJ, 932, 77, doi: 10.3847/1538-4357/ac6e3fDOI
- [129]Suyu, S. H., Acebron, A., Grillo, C., et al. 2026, A&A, 708, doi: 10.1051/0004-6361/202557235DOI
- [130]Suzuki, T. L., Glazebrook, K., Schreiber, C., et al. 2022, ApJ, 936, 61, doi: 10.3847/1538-4357/ac7ce3DOI
- [131]Tacchella, S., Carollo, C. M., Renzini, A., et al. 2015, Science, 348, 314, doi: 10.1126/science.1261094DOI
- [132]Tacchella, S., Carollo, C. M., Förster Schreiber, N. M., et al. 2018, ApJ, 859, 56, doi: 10.3847/1538-4357/aabf8bDOI
- [133]Tacconi, L. J., Genzel, R., Saintonge, A., et al. 2018, ApJ, 853, 179, doi: 10.3847/1538-4357/aaa4b4DOI
- [134]Tody, D. 1986, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 627, Instrumentation in astronomy VI, ed. D. L. Crawford, 733
- [135]Tody, D. 1993, in Astronomical Society of the Pacific Conference Series, Vol. 52, Astronomical Data Analysis Software and Systems II, ed. R. J. Hanisch, R. J. V. Brissenden, & J. Barnes, 173
- [136]Tomczak, A. R., Quadri, R. F., Tran, K.-V. H., et al. 2014, ApJ, 783, 85, doi: 10.1088/0004-637X/783/2/85DOI
- [137]Umehata, H., Kubo, M., & Nakanishi, K. 2025, ApJL, 985, L8, doi: 10.3847/2041-8213/add1d4DOI
- [138]Utomo, D., Kriek, M., Labbé, I., Conroy, C., & Fumagalli, M. 2014, ApJL, 783, L30, doi: 10.1088/2041-8205/783/2/L30DOI
- [139]van der Wel, A., Franx, M., van Dokkum, P. G., et al. 2014, ApJ, 788, 28, doi: 10.1088/0004-637X/788/1/28DOI
- [140]van Dokkum, P. G., Franx, M., Kriek, M., et al. 2008, ApJL, 677, L5, doi: 10.1086/587874DOI
- [141]Ventura, P., di Criscienzo, M., Schneider, R., et al. 2012, MNRAS, 424, 2345, doi: 10.1111/j.1365-2966.2012.21403.xDOI
- [142]Whitaker, K. E., & Bezanson, R. 2026, arXiv e-prints, arXiv:2606.12156, doi: 10.48550/arXiv.2606.12156DOI
- [143]Whitaker, K. E., Labbé, I., van Dokkum, P. G., et al. 2011, ApJ, 735, 86, doi: 10.1088/0004-637X/735/2/86DOI
- [144]Whitaker, K. E., van Dokkum, P. G., Brammer, G., et al. 2013, ApJL, 770, L39, doi: 10.1088/2041-8205/770/2/L39DOI
- [145]Whitaker, K. E., Williams, C. C., Mowla, L., et al. 2021a, Nature, 597, 485, doi: 10.1038/s41586-021-03806-7DOI
- [146]Whitaker, K. E., Narayanan, D., Williams, C. C., et al. 2021b, ApJL, 922, L30, doi: 10.3847/2041-8213/ac399fDOI
- [147]Wild, V., Asari, N. V., Rowlands, K., et al. 2025, The Open Journal of Astrophysics, 8, 3, doi: 10.33232/001c.128125DOI
- [148]Williams, R. J., Quadri, R. F., Franx, M., van Dokkum, P., & Labbé, I. 2009, ApJ, 691, 1879, doi: 10.1088/0004-637X/691/2/1879DOI
- [149]Woodrum, C., Williams, C. C., Rieke, M., et al. 2022, ApJ, 940, 39, doi: 10.3847/1538-4357/ac9af7DOI
- [151]Wuyts, E., Barrientos, L. F., Gladders, M. D., et al. 2010, ApJ, 724, 1182, doi: 10.1088/0004-637X/724/2/1182DOI
- [152]Zanella, A., Valentino, F., Gallazzi, A., et al. 2023, MNRAS, 524, 923, doi: 10.1093/mnras/stad1821DOI
- [153]Zhu, P., Ito, K., Valentino, F., et al. 2026, A&A, 712, A169, doi: 10.1051/0004-6361/202659510DOI
- [154]Zhuang, Z., Leethochawalit, N., Kirby, E. N., et al. 2023, ApJ, 948, 132, doi: 10.3847/1538-4357/acc79bDOI