JADES Ultrared Flattened Objects: Morphologies and Spatial Gradients in Color and Stellar Populations

One of the more challenging populations of galaxies to characterize over the past few decades has been the optically dark/faint or "Hubble Space Telescope (HST)-dark" galaxies, where strong dust attenuation causes a lack of optical/UV emission making them difficult or impossible to detect with facilities such as the HST (e.g., Barger et al. 1998, 2022; Hughes et al. 1998; Coppin et al. 2006; Elbaz et al. 2011; Williams et al. 2019; Umehata et al. 2020; Gómez-Guijarro et al. 2022; Manning et al. 2022; Xiao et al. 2023). Surveys in the far-infrared (FIR) through millimeter wavelengths with instruments such as Spitzer/IRAC, Herschel, the James Clerk Maxwell Telescope (JCMT/SCUBA), and the Atacama Large Millimeter/submillimeter Array (ALMA) have gradually revealed this population to be primarily concentrated around z ≳ 2 (e.g., Barger et al. 1998; Hughes et al. 1998; Coppin et al. 2006; Elbaz et al. 2011; Wang et al. 2012, 2016, 2019; Tadaki et al. 2017; Franco et al. 2018). These and other subgroups of optically faint and/or IR/submillimeter-detected galaxies fall into the broad category of dusty star-forming galaxies (DSFGs), whose discovery, classification, and properties are reviewed in Casey et al. (2014). In addition to possessing large quantities of dust, measurements have revealed moderate-to-high stellar masses (M_* ≳ 10⁹ M_⊙) and high star formation rates (SFRs; e.g., Elbaz et al. 2011; Wang et al. 2019) with a small subset undergoing intense bursts of star formation with substantially higher SFRs of ≳1000/M_⊙ yr⁻¹ (e.g., Walter et al. 2012; Riechers et al. 2013; Marrone et al. 2018; Ma et al. 2019). This obscured galaxy population appears to be an important, and possibly dominant contributor to the cosmic SFR density out to z ∼ 4 (e.g., Barger et al. 2012; Madau & Dickinson 2014; Zavala et al. 2021; Barrufet et al. 2023).

Morphological studies of sources with detections at both optical/UV and submillimeter wavelengths have revealed that the DSFG population is often much more compact at longer wavelengths than at shorter wavelengths (e.g., Barro et al. 2016; Hodge et al. 2016; Oteo et al. 2016; Rujopakarn et al. 2016; Fujimoto et al. 2017; Tadaki et al. 2017; Gullberg et al. 2019; Nelson et al. 2019). However, this is not always the case, as Sun et al. (2021) report two spatially extended DSFGs with submillimeter sizes exceeding the IR sizes from Spitzer/IRAC data. These two wavelength regimes, rest-frame optical and FIR, are commonly taken to represent the distribution of existing stellar mass and star formation, respectively, suggesting that star formation is primarily centrally concentrated in these objects. This observation can be explained if these galaxies are in the process of building their central stellar bulges (e.g., Tadaki et al. 2017, 2020; Nelson et al. 2019), a strong indicator of future cessation of star formation (e.g., Franx et al. 2008; Bell et al. 2012; Lang et al. 2014; Whitaker et al. 2017). Lending further empirical evidence to support this physical interpretation, many DSFGs at higher redshifts (z ∼ 3) have FIR sizes that are consistent with the optical sizes of quiescent galaxies at lower redshifts (z ∼ 2). The leading hypothesis is thus that the DSFG population is the plausible progenitor of massive quenched systems at z ∼ 2 (e.g., Toft et al. 2014; Fujimoto et al. 2017; Tadaki et al. 2017; Suess et al. 2021), which in turn are likely the progenitors of the most massive and quenched galaxies observed in the local Universe. Given the relative importance of optically faint galaxies (OFGs) in the context of both the cosmic SFR budget and our understanding of massive galaxy evolution, it is imperative to have a more complete census of their morphologies and stellar populations (e.g., Gottumukkala et al. 2024; Williams et al. 2024).

Prior to the launch of the James Webb Space Telescope (JWST; Gardner et al. 2023), spatially resolved studies of the optically faint population have primarily been facilitated by interferometers such as the Submillimeter Array (SMA) or ALMA (e.g., Chen et al. 2014; Walter et al. 2016; Hsu et al. 2017; Tadaki et al. 2017; Cowie et al. 2018; Franco et al. 2018; Gómez-Guijarro et al. 2022). Due to their limited field of view and lower sensitivity, studies have therefore been limited to bright, lensed, or previously known galaxies. Thus, while interferometers can achieve the necessary spatial resolution to characterize galaxies at higher redshifts, their small field of view makes it challenging to observe or discover large samples of galaxies. Conversely, infrared telescopes like Spitzer or Herschel can observe many galaxies, but their spatial resolution is too poor for morphological studies of galaxies.

All of this has changed with the successful launch and commissioning of JWST, whose NIRCam instrument's large field of view ($\sim 10\ {\mathrm{arcmin}}^{2}$) and wavelength coverage redder than HST have already enabled numerous studies of the optically faint population out to 4.4 μm (Rieke et al. 2023a). Populations of these OFGs have been studied in the Cosmic Evolution Early Release Science (CEERS) survey (e.g., Barrufet et al. 2023; Gómez-Guijarro et al. 2023) as well as in observations of the SMACS0723 cluster (Rodighiero et al. 2023) finding that these galaxies are generally at redshifts 2 < z ≲ 7 with high stellar masses (M_* > 10¹⁰ M_⊙) and high dust attenuations (A_V > 2). Breaking down the JWST-observed OFGs observed in CEERS, Pérez-González et al. (2023) find that the majority of these sources (71%) are DSFGs at 2 < z < 6. However, to better understand how these galaxies fit into our overall picture of galaxy growth, we need to study their structures to better infer how these galaxies have assembled. Gómez-Guijarro et al. (2023) study the stellar masses and morphologies of DSFGs at 3 < z < 7.5 and find a subset of highly attenuated (defined as A_V > 1) galaxies whose main difference from their total sample is their ∼30% smaller effective size.

Given the compact OFGs reported in Gómez-Guijarro et al. (2023) and expectations that optically faint populations, in general, might be compact, the discovery of OFGs detected in CEERS with apparent sizes greater than 025 has come as a surprise (Nelson et al. 2023). This visually striking galaxy population, dubbed ultrared flattened objects (UFOs), are noticeably more elongated and redder than the general population in both their inner and outer regions. Although the expectation may have been that these heavily obscured galaxies should be more compact than optically bright galaxies leading to a larger dust column density (as in, e.g., Gómez-Guijarro et al. 2023), the UFOs are not more compact than other galaxies at the same mass. Given the low axis ratios combined with inferred Sérsic indices close to unity, Nelson et al. (2023) suggest that it is most likely for the UFO population to be disk galaxies. Interestingly, in a spatially resolved study of a single previously known and spectroscopically confirmed z = 2.38 OFG in the A2744 field, Kokorev et al. (2023) were able to characterize the OFG as being a massive ($\mathrm{log}({M}_{* }/{M}_{\odot })\approx 11.3$), highly star-forming (200 M_⊙ yr⁻¹), dusty, and edge-on spiral galaxy with a nearly uniform A_V ∼ 4 across its disk. This galaxy has the same general characteristics as the UFO population described in Nelson et al. (2023) and appears consistent with having a disk morphology based on the visual identification of spiral structure in F277W imaging. Although there is some evidence that UFOs are highly inclined disks, low Sérsic indices and a projected axis-ratio distribution skewed toward low values can also be indicative of an intrinsically prolate population. Existing samples so far have been too small to perform axis-ratio modeling in order to statistically determine their intrinsic shapes, but here with deeper imaging over a wider area, we can acquire large enough samples to perform statistically robust axis-ratio modeling.

Looking at the observed color gradients (which can reveal spatial variations in stellar population properties, thus reflecting galaxy evolutionary and assembly histories) of the UFO population, Nelson et al. (2023) find that they are consistent with previously observed negative color gradients in galaxies with colors becoming bluer as you move out from the center (e.g., Tortora et al. 2010; Wuyts et al. 2010; Guo et al. 2011; Szomoru et al. 2013; Chan et al. 2016; Mosleh et al. 2017; Suess et al. 2019a, 2019b; Miller et al. 2023). However, the UFOs differ in that they are still red in their outskirts albeit not as red as in their inner regions. Nelson et al. (2023) suggest that this could be driven by large quantities of dust in these objects. This dust would need to be distributed throughout most of a galaxy to account for the red colors throughout unless there were other spatial gradients in the UFO stellar populations (e.g., younger dust-obscured stars in the outskirts and older dust-obscured stars centrally). Resolved stellar population modeling is needed to determine why these objects are so red out to large radii.

In order to better understand the intrinsic shapes of UFOs and to address what drives their observed colors, we present the identification and properties of 112 OFGs found in the JWST Advanced Deep Extragalactic Survey (JADES) observations of both Great Observatories Origins Deep Survey (GOODS) fields. Of the 112 OFGs, we identify 56 as being UFOs and, in this paper, we seek to understand the morphologies as well as the stellar population and dust content of these enigmatic objects leveraging the ≳5× larger sample along with deeper imaging. We describe our data and analysis methods in Section 2 and present the results of the integrated stellar populations in Section 3. In Section 4, we discuss the observed color gradients in the context of the stellar population and dust content of their inner and outer regions. We discuss the inferred projected shapes of the UFOs and show the constraints that can be made on their intrinsic shapes in Section 5. We discuss the implications of our results in Section 6 and we summarize our conclusions in Section 7.

Throughout the paper, we assume the WMAP9 ΛCDM cosmology with Ω_m = 0.2865, Ω_Λ = 0.7135, and H₀ =69.32 km s⁻¹ Mpc⁻¹ (Hinshaw et al. 2013). All magnitudes in this paper are expressed in the AB system (Oke 1974).

2.1. Observations

2.2. Data Reduction and Photometry

2.3. Sample Selection

For this work, our goal is to leverage the larger area of the JADES survey to generate a larger sample of UFOs in an attempt to address their nature (e.g., are they disks?). To this end, we select the galaxies in JADES that are brightest at the reddest wavelengths of our coverage (4.4 μm) and faint at HST wavelengths (<1.6 μm). From this sample of OFGs, we take those with semimajor axis effective (half-light) sizes greater than 025 as our sample of UFOs as illustrated in Figure 1.

Originally, we adopted the OFG selection criteria in Nelson et al. (2023) requiring AB magnitudes of F444W < 24.5, F115W > 27, and F150W > 25.5. However, during our analysis, we noted that some of the selected UFOs showed blue excesses either in the form of bluer clumps, excess blue emission on the outskirts, or SEDs more indicative of bluer stellar populations. These bluer outliers had inferred A_V values much lower (≲1) than their redder companions, and thus to select a more intrinsically red parent sample from which to find UFOs, we revised the OFG selection criteria of Nelson et al. (2023) to include the same F444W brightness cut, but with the addition of the following color cuts: F150W − F200W > 0.75 and F200W − F444W > 2.0. The selection change results in a much more robust sample of thoroughly red objects whose optical faintness is largely driven by dust attenuation (see Section 3).

Zoom In Zoom Out Reset image size

With these criteria, we identify 112 total galaxies with a range of apparent sizes. We fit for their morphological parameters (described in Section 2.5) allowing us to identify UFOs among our sample of OFGs. This selection identifies 56 UFOs whose stellar populations, morphologies, and comparisons with our broader optically faint sample we discuss below. In Figure 2, we show F200W, F277W, and F444W color images of the UFOs and in Figure 3, we show, for a subsample of UFOs, how HST color images (F606W, F125W, F160W) compare with the corresponding JWST image clearly illustrating the ability of JWST to reveal galaxies that were invisible or faint at HST wavelengths.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Additionally, from the F444W < 24.5 mag parent sample, we identify those that are bluer than our optically faint selected sample in both colors and analyze them as well. This sample is further limited to the same redshift and stellar mass range as the whole optically faint sample (described in Section 3). We fit the stellar populations and morphologies of this sample in the same way as for the OFGs in order to more broadly compare the two populations. Unless otherwise stated, figures will refer to the blue sample of F444W-selected galaxies as the F444W-parent sample and the smaller (non-UFO) optically faint galaxies as OFGs. Our three samples are shown in Figure 1 where we plot their F200W − F444W colors versus their semimajor axis effective size illustrating both our size cuts and one of our color cuts.

2.4. Stellar Populations

We utilize the photometric redshift code, EAzY (Brammer et al. 2008), to infer the redshifts of our sample of galaxies. The templates and photometry used to constrain the redshifts follow the methods outlined in Hainline et al. (2024). For the rest of the stellar population parameters, we use the Prospector SED fitting code (Johnson et al. 2021) fixing the redshift to the value inferred from EAzY. We verified that allowing the redshift to remain free in our Prospector fits yields redshifts consistent with EAzY as well as consistent stellar population parameter inferences as the fixed-redshift results. To vastly decrease the time needed to perform each SED fit, we make use of a neural net emulator, parrot, which has been trained to infer the photometry for some underlying stellar population model (see Mathews et al. 2023). This emulator gives consistent results with standard Prospector, but with a factor of ≈10³ increase in speed.

We adopt the Prospector-β model described in detail in Wang et al. (2023) adjusted to keep the redshift fixed. In brief, this is the same model as the Prospector-α model (Leja et al. 2017) but makes use of a joint prior on the redshift, stellar mass, and stellar metallicity disfavoring high-mass, high-z solutions while still allowing a nonnegligible probability of obtaining such solutions (thus making it possible to discover massive galaxies at high redshifts). The stellar populations are modeled using the MILES spectral templates (Sánchez-Blázquez et al. 2006; Falcón-Barroso et al. 2011) with the MIST stellar isochrones (Choi et al. 2016) as implemented in FSPS (Conroy et al. 2009, 2010). We adopt the Chabrier initial mass function (Chabrier 2003) and model the star formation history by fitting for the mass formed in seven logarithmically spaced time bins using a continuity prior (Leja et al. 2019a), which weights against sharp changes in the SFR between adjacent time bins. In practice, this prior is a Student's t-distribution for the log ratio of the SFR in adjacent time bins, $\mathrm{log}\ ({\mathrm{SFR}}_{n}/{\mathrm{SFR}}_{n-1})$, with a width of σ = 0.3, and degrees of freedom, ν = 2.

To accurately characterize the red and likely dusty nature of these galaxies, we adopt the two-component dust model of Charlot & Fall (2000), which characterizes the dust in galaxies with a diffuse dust component attenuating all stars equally and a separate birth-cloud component providing additional dust attenuation for stars younger than 10 Myr. Each component can have extinctions up to A_V ∼ 4.3 and we also allow for a varying dust attenuation curve following the prescription in Noll et al. (2009). We additionally test two alternative dust models in which we: (1) remove the young star component, and (2) fix the slope of the attenuation curve to the Calzetti et al. (2000) value. These two variations give consistent results with our fiducial dust model. The effects of dust emission are included using the dust emission templates from Draine & Li (2007) with three free parameters that control the shape of the IR SED (see Leja et al. 2017 for details). Finally, the possibility of an active galactic nucleus (AGN) powering hot dust emission is included (Leja et al. 2018) utilizing the AGN templates of Nenkova et al. (2008a, 2008b) and nebular line emission is included following the Cloudy (Ferland et al. 1998, 2017) FSPS implementation described in Byler et al. (2017). We note that certain aspects of this model (i.e., FIR dust and AGN emission) are not strictly necessary, but are included as default parameters in the emulator we use. It is possible to fix certain parameters to narrow the explored parameter space, but doing so has a negligible impact on our results.

We fit each object using the photometry as described in Section 2.2 and values are reported as the median of the posteriors and 1σ uncertainties as the 84th–50th and 50th–16th interquartile ranges. To approximately account for systematic errors or issues with early JWST photometric calibrations, we enforce an error floor of 5%. We show the results of our fits in Figure 4 for a representative group of galaxies in our sample. Each fit shows the observed SED, the model SED, and a spectrum generated in FSPS using the inferred model parameters.

Zoom In Zoom Out Reset image size

2.5. Morphologies

In Figure 5, we show how our galaxies are distributed in stellar mass and redshift as well as how they are situated within the "star-forming main sequence" (SFMS) of Leja et al. (2022) for z = 2.5 galaxies. The UFOs tend to occupy the same region of both the z–M_* and SFR–M_* planes as their optically faint parent sample suggesting that the primary distinguishing factor is their structure. These objects tend to be clustered between 2 < z < 4, with moderate-to-high stellar masses ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 10$) and SFRs that tend to lie on or above the SFMS of z = 2.5 galaxies from Leja et al. (2022) except at the massive end where there is a noticeable drop to below the SFMS at a mass of ∼10¹⁰ M_⊙. We discuss this drop more in Section 6. The median values and percentiles (16th and 84th) for these three parameters ($z,\mathrm{log}({M}_{* }/{M}_{\odot }),\mathrm{log}(\mathrm{SFR}/({M}_{\odot }\,{\mathrm{yr}}^{-1})$) are ${2.42}_{-0.50}^{+0.58}$, ${10.04}_{-0.33}^{+0.63}$, and ${1.58}_{-0.37}^{+0.50}$ for the UFOs and ${2.78}_{-0.64}^{+0.85}$, ${10.29}_{-0.39}^{+0.50}$, and ${1.48}_{-0.57}^{+0.48}$ for the smaller OFGs. We show these values along with the values for other populations in Table 1.

Zoom In Zoom Out Reset image size

Table 1. Integrated Stellar Population Parameters

Parameter	UFOs	OFGs	F444W-parent	Herschel Ultrared
z	${2.42}_{-0.50}^{+0.58}$	${2.78}_{-0.64}^{+0.85}$	${2.19}_{-0.49}^{+0.68}$	${3.24}_{-0.51}^{+1.10}$
$\mathrm{log}({M}_{* }/{M}_{\odot })$	${10.04}_{-0.33}^{+0.63}$	${10.29}_{-0.39}^{+0.50}$	${9.81}_{-0.21}^{+0.63}$	${11.57}_{-0.42}^{+0.18}$
$\mathrm{log}(\mathrm{SFR}/({M}_{\odot }\,{\mathrm{yr}}^{-1}))$	${1.58}_{-0.37}^{+0.50}$	${1.48}_{-0.57}^{+0.48}$	${0.94}_{-1.27}^{+0.51}$	${2.86}_{-0.47}^{+0.23}$
AV	${2.76}_{-0.43}^{+0.47}$	${2.61}_{-0.65}^{+0.52}$	${0.87}_{-0.40}^{+0.62}$	⋯
MWA [Gyr]	${0.97}_{-0.51}^{+1.43}$	${0.90}_{-0.51}^{+1.50}$	${1.09}_{-0.35}^{+1.88}$	⋯

Note. The reported quantity is the median and percentiles (16th and 84th) of the distribution for that parameter.

Download table as:  ASCIITypeset image

Figure 5 also compares the optically faint populations in the context of the bluer F444W-selected parent sample and an FIR-selected sample from Ma et al. (2019). The FIR-selected galaxies are much more extreme than the UFOs with stellar masses predominately higher than M_* ∼ 10¹¹ M_⊙ and SFRs > 100M_⊙ yr⁻¹, and z > 3, although we note that there is likely a systematic offset in their measured SFRs toward higher values compared to those we infer here owing to different methodologies used (e.g., Leja et al. 2019b). These galaxies are examples of extremely red objects with large A_V that were selected to have rising Herschel/SPIRE flux densities (S₅₀₀ > S₃₅₀ > S₂₅₀). Compared to the bluer F444W-selected galaxies, the OFGs tend to have slightly higher redshifts and stellar masses than the OFGs. The SFMS in Leja et al. (2022) is fit to the 1.6 μm selected 3D-HST galaxies, so the fact that the majority of our sample lies on and around this line means that these are not extreme objects but rather a normal subset of the massive galaxy population that was previously missed. Thus, the population of UFOs studied here are not extreme objects in regard to the stellar mass, SFR, or redshift.

We also show the position of the three F444W-selected samples in the mass-weighted age (MWA)−A_V plane and the stellar mass–A_V plane in Figure 6 with the MWA and A_V (for the diffuse dust component) representing two likely drivers of optical faintness. Most of the OFGs have large A_V (${2.76}_{-0.43}^{+0.47}$ for UFOs and ${2.61}_{-0.65}^{+0.52}$ for the smaller OFGs) and a young MWA (${0.97}_{-0.51}^{+1.43}$ for UFOs and ${0.90}_{-0.51}^{+1.50}$ for OFGs). The right panel additionally shows the inferred A_V in the inner and outer regions as discussed in Section 4, demonstrating the large A_V values in both the inner and outer regions of the UFOs. Compared to the bluer F444W-selected galaxies, the UFOs have much higher dust attenuation, but similar MWAs suggest that their red colors—and the reason they were previously missed—is owing to dust attenuation and redshift as opposed to old stellar ages. These results are discussed further in Section 6.

Zoom In Zoom Out Reset image size

To further quantify the differences between these populations (UFOs, smaller OFGs, and F444w-parent) in these 2D parameter spaces, we perform 2D Kolmogorov–Smirnov (KS) tests on these distributions, using the public code ndtest, ²⁵ based on the algorithm described in Peacock (1983) and Fasano & Franceschini (1987). First, in comparing the F444W-parent sample with either optically faint population, we find very small p-values, consistent with zero, showing that the OFGs are a distinct population in terms of stellar mass, SFR, dust attenuation, and age. Comparing the UFOs and the smaller OFGs, we find a large p-value (0.53) for the MWA–A_V distribution, a moderate p-value (0.14) for the log(SFR)–$\mathrm{log}(M/{M}_{\odot })$, and small p-values (<0.06) for the z–$\mathrm{log}(M/{M}_{\odot })$ and A_V –$\mathrm{log}(M/{M}_{\odot })$ distributions. This suggests that UFOs are most similar to their smaller counterparts in their ages, dust attenuation, and SFRs, while the stellar mass is where UFOs differ most from the smaller OFGs.

Looking at the images of our galaxies, it is apparent that these galaxies are consistently red throughout their extended light profiles. To make this visual assessment more quantitative, we calculate the F200W – F444W colors for the UFOs in the inner regions (03 diameter circular aperture) and outer regions (06–03 diameter apertures) of the PSF-matched images. We additionally calculate the integrated colors for the smaller optically faint and bluer F444W-selected galaxies. We show these colors in Figure 7 with the UFO inner and outer colors shown with filled and unfilled red circles, respectively. As can be seen, the UFOs are strikingly red in both their inner and outer regions with both region colors much redder than the bulk of F444W-selected galaxies. These are the same trends that were observed in the sample of UFOs first reported in Nelson et al. (2023) who saw that their 12 galaxies had red colors throughout the inner and outer regions that were generally much redder than the parent population. This is a unique population in that the inner and outer regions are both truly red as opposed to many galaxies that are observed to have red centers and blue outskirts. However, the color gradient is still negative in both cases with the UFOs being less red in the outskirts than they are in the centers.

Zoom In Zoom Out Reset image size

To further elucidate the persistent redness shown in Figure 7, we also infer the UFO stellar populations within the inner 03 diameter (2.42 kpc at z = 2.5) and the outer 06–03 annulus. This approximate spatially resolved analysis serves to place constraints on what is driving the red colors observed throughout the extent of these objects. A representative best-fit spectrum from each region is shown in the lower panel of Figure 8. To generate these region-specific spectra, we first generate a rest-frame spectrum in FSPS for each UFO/region taken from the maximum of the parameter posteriors for that object fit. Then, for each region, we stack the individual rest-frame spectra by taking a median at each wavelength/flux value to generate a "median" spectrum for each region. It is clear that the two spectra are most similar at long wavelengths and least similar at shorter/bluer wavelengths with more blue flux in the outer regions consistent with the slight color gradients observed in UFOs that we showed in Figure 7.

Zoom In Zoom Out Reset image size

Turning to what drives these slight color gradients, in the upper panels of Figure 8, we show the comparison between the inner and outer SED fitting results for the following parameters: stellar mass (M_*), visible attenuation from dust (A_V ), MWA, and specific star formation rate (sSFR). All comparisons are shown as the ratio or difference of the inner parameter inference to the outer parameter inference as a function of the stellar mass inferred from the integrated fits (Section 3). The red dashed line in each panel reflects the median ratio or difference between the inner and outer parameters, the gray dashed line is where the two regions have the same value, and the red shaded region shows the ∼1σ spread in the ratio/difference. This information is additionally summarized in Table 2, where we show the median and percentiles of the inner and outer stellar population parameters.

Table 2. Inner and Outer Stellar Population Parameters

Parameter	Inner	Outer
$\mathrm{log}({M}_{* }/{M}_{\odot })$	${9.86}_{-0.42}^{+0.48}$	${10.11}_{-0.47}^{+0.42}$
$\mathrm{log}(\mathrm{sSFR}/{\mathrm{yr}}^{-1})$	$-{8.28}_{-0.70}^{+0.55}$	$-{8.64}_{-0.78}^{+0.58}$
AV	${4.26}_{-0.70}^{+1.46}$	${2.29}_{-0.57}^{+0.32}$
MWA [Gyr]	${0.94}_{-0.54}^{+0.49}$	${1.08}_{-0.57}^{+0.44}$

Note. The reported quantity is the median and percentiles (16th and 84th) of the distribution for that parameter.

Download table as:  ASCIITypeset image

The largest radial differences are observed between the inner and outer A_V , where it is clear that the majority of UFOs have higher central A_V than outer A_V although both regions have high A_V (Figure 6). Taken at face value, this implies that dust concentrations are higher in the center as has been found before in studies of other galaxy populations (e.g., Nelson et al. 2016; Whitaker et al. 2017; Tacchella et al. 2018). On the other hand, the gradients in the other parameters, MWA and sSFR, are not quite as strong, especially for the MWA which has a median ratio of ∼1, but with a significant scatter about the median (typical spread of ∼0.7). Here, the slight color gradients of our UFOs appear to be primarily caused by stronger central dust attenuation consistent with what was found in Miller et al. (2023) for HST-selected star-forming galaxies (SFGs).

5.1. Sizes and Sérsic Indices

Inferred morphological parameters are shown in Figure 9 and the median/percentiles of these parameters are shown in Table 3. In the leftmost panel of Figure 9, we show how the UFOs and their optically faint and F444W-selected parent populations are situated in the size–mass plane. For comparison, we show the z = 2 size–mass relations from Suess et al. (2019a) for SFGs and quiescent galaxies. Our sample of UFOs lies on or above the SFG size–mass relation, showing that these are truly extended objects, both physically and apparently. Further, the full sample of OFGs is fairly evenly distributed in the size–mass plane relative to the parent sample. This is surprising. While we expected OFGs to have compact sizes to drive their high values of dust attenuation, this does not appear to be the case. The distribution of the sizes of the optically faint population is similar to that of the parent population.

Zoom In Zoom Out Reset image size

Table 3. Morphological Parameter Summary

Parameter	UFOs	OFGs
$\mathrm{log}({r}_{e}/\mathrm{kpc})$	${0.43}_{-0.09}^{+0.17}$	${0.04}_{-0.30}^{+0.19}$
Axis-ratio (q)	${0.39}_{-0.17}^{+0.20}$	${0.60}_{-0.17}^{+0.21}$
Sérsic Index (n)	${1.12}_{-0.39}^{+1.53}$	${1.72}_{-0.68}^{+1.99}$

Note. The reported quantity is the median and percentiles (16th and 84th) of the distribution for that parameter.

Download table as:  ASCIITypeset image

The middle two panels of Figure 9 show the distribution of Sérsic indices versus the stellar mass and effective size for the UFOs in the context of all OFGs and the F444W-selected parent sample. The majority of the UFOs and the OFG population, in general, have n < 2, suggesting they have a minimal structural contribution from a bulge. Contrary to our physical expectation that OFGs would be centrally concentrated to drive their high dust attenuations, the bulk of these objects have Sérsic indices consistent with surface brightness profiles that are only slightly steeper than exponential. Interestingly, among F444W-selected massive galaxies ($\mathrm{log}({M}_{* }/{M}_{\odot })\gt 10$) with low Sérsic indices (n < 2), and r_e > 2 kpc, 69% are optically faint UFOs. Thus, neither compact sizes nor high Sérsic indices can explain the high values of dust attenuation in these galaxies.

5.2. 3D Shapes from Axis-ratio Distribution Modeling

Here we explore the intrinsic shapes of the UFOs by modeling the distribution of their projected axis ratios (e.g., Chang et al. 2013; van der Wel et al. 2014). In general, the shape of a galaxy can be approximated as an ellipsoid characterized by three axis lengths, A ≳ B ≳ C, which can be used to define the ellipticity (E) and triaxiality (T) parameters. Within the family of ellipsoids, there are three broad categories: the oblate ellipsoid (A ∼ B > C, disky), the prolate ellipsoid (A > B ∼ C, elongated), or the spheroid (A ∼ B ∼ C). The projected axis ratios of these shapes can be inferred from how each ellipsoid type would appear in projection from a set of randomly drawn viewing angles with distinct shapes having unique projected axis-ratio distributions (e.g., Chang et al. 2013; van der Wel et al. 2014) characterized by values of E and T and thus values of A, B, and C. For example, a q distribution peaking between 0.8 and 1.0 would be predominately comprised of spheroids as they can only have large observed axis ratios. A disk population would have a roughly flat q distribution reflecting that disks can be observed edge-on, face-on, and everything in between, while only observing inclined disks would result in an axis-ratio distribution without any higher q objects. Prolate populations have axis-ratio distributions that peak at shorter values (q ∼ 0.4) with a drop at q ∼ 0.2 and a tail toward higher axis ratios.

In Figure 10, we plot the observed axis-ratio distributions of the full F444W-selected parent sample, all OFGs, and the size-selected subset (the UFOs). We find that the UFOs typically have q < 0.5 and the smaller OFGs typically have q ≥ 0.5. Taken at face value, this suggests that the UFOs are largely prolate or disk-shaped and the more compact OFGs are more spheroid-dominated. We additionally perform a KS test on the UFO and OFG axis-ratio distributions to constrain the probability that these samples were drawn from the same underlying population. We find a p-value of ∼10⁻⁶ showing that, at the very least, the UFOs and OFGs represent different populations in terms of their axis ratios, though we note that owing to their sizes being closer to the scale of the PSF, the shapes of the compact OFGs are much less certain. Figure 11 shows the observed UFO axis-ratio distribution compared to various mock populations. Our sample of UFOs has many more objects with q ∼ 0.1–0.2 than the sample of Nelson et al. (2023) and as such neither inclined thick-disk or prolate populations by themselves adequately match the observed UFO q distribution. The tail of very low axis ratios can only be achieved by having a population of intrinsically inclined thin disks capable of reaching such low q values. However, as can be seen in the right panel of Figure 11, an inclined thin-disk population by itself overpredicts these low q objects and underpredicts those with higher q. For this reason, we show combinations of prolate and disk-like populations on the right. With the exception of the prolate+thin disk (no inclination restriction), which overpredicts higher q objects, the various combinations all seem to fairly well match the observed distribution making it challenging to uniquely assign an intrinsic shape to the UFOs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

For these reasons, we turn to a modeling scheme that directly fits the observed axis-ratio distributions given some assumptions about the true underlying distribution. We use the BEAST (S. H. Price et al. 2024, in preparation) axis-ratio modeling code to perform nested sampling with dynesty (Speagle 2020) to infer the intrinsic shapes of the UFOs. Here, we briefly summarize the functionality of this software. The ellipticity (E = 1 − C²) and triaxiality ($T=\left[1-{B}^{2}\right]/\left[1-{C}^{2}\right]$) can be used to define specific subtypes of ellipsoids characterized by ranges of values of E and T reflecting the underlying relationships between A, B, and C. In practice, one can combine an observed q distribution and, assuming the galaxies come from a population that can be described by a distribution of E and T, perform Markov Chain Monte Carlo or nested sampling of the posterior to determine the most likely values for the parameters of the E and T distributions, and thus on the intrinsic shapes (i.e., oblate/disky or prolate) given these assumptions.

The BEAST modeling assumes all galaxies in the sample are drawn from a single population with Gaussian distributions of E and T, characterized by four free parameters: centers μ_E and μ_T and sigmas, σ_E and σ_T . In our default model, we consider galaxies that are randomly observed from any viewing angle. However, we also employ a secondary model that restricts the inclination to be between 50° and 90°, and a third that additionally enforces oblateness (i.e., diskiness) by restricting the triaxiality parameter <0.33, in addition to restricting the inclination. These secondary models encode the physical expectation that these objects are inclined disks. Each model fits the q distribution in 10 bins. For each parameter, we assume uniform priors from zero to unity except for the third model in which we restrict the range of T to be between zero and 0.33. In what follows, we define four unique regions of the intrinsic axis-ratio space (C versus B), which represent distinct shapes and can be seen in the middle panel of Figure 12. These are spheroidal (green/upper right), prolate (blue/lower left), thin disk (red/bottom of the circular region), and thick disk (purple/top of the circular region). The results of fitting these three models are summarized in Figure 12 where we show the best-fit (maximum a posteriori) model q distribution, random draws from the distribution of the best-fit model in the intrinsic axis-ratio space, and the fractions of each shape that can be attributed to that single-population model.

Zoom In Zoom Out Reset image size

From the top panels in Figure 12, we can see that each different version of the model (free, inclined, inclined and oblate) does a reasonably good job of reproducing the observed UFO axis-ratio distribution. However, the model with no triaxiality or inclination constraints achieves this agreement in a markedly different way than the constrained models. In particular, when there are no constraints on the inclination or triaxiality of the population, the model prefers a population that is dominated by intrinsically prolate objects (∼90%) with only a ∼10% contribution from disks (mostly thin) and essentially zero spheroidal objects. This tells us that the UFO q distribution can be generated from a population that is dominated by prolate ellipsoids with a small contribution from thin disks to reproduce the lowest-q UFOs.

In the other two models, the story is very different. These disk assumption models can also produce axis-ratio distributions that are fairly consistent with the data showing that disk-based distributions are not inconsistent with the observations. In the middle panel of the bottom row, we can see that the restricted inclination model still predicts that prolate objects are the dominant contributor to the observed q distribution, but only slightly with a ∼40% contribution from thick and thin disks. Thus, even if we assume that we are only observing galaxies closer to edge-on, the axis-ratio modeling still statistically prefers a contribution from prolate objects. Moving to the third panel of the bottom row, we can see that if we restrict the triaxiality to be less than 0.33 as well as restricting the inclination, then the model predicts the dominant contribution coming from thick and thin disks (∼60%) with a contribution from prolate ellipsoids of at most ∼40%. This shows that even in the most disk-friendly modeling scenario, there is evidence for a contribution from the prolate set of objects. That some optically faint objects might be prolate is a surprising finding as it is not straightforward how a prolate object viewed along its long axis could have dust column densities large enough to render the galaxy optically faint. We can restrict the triaxiality even more, which will put more objects into the disk category, but this leads to much worse agreement between observed and model q distributions and requires a ∼20% spheroid contribution.

We note here a few modeling caveats that should be mentioned before we continue to discuss the broader implications of these results. First, we assume that the intrinsic shapes of the UFOs can be accurately modeled as ellipsoids, which is a good approximation but will clearly break down for certain objects. For example, ongoing galaxy mergers, galaxies with clumpy features, or galaxies with spiral arms and/or dust lanes could deviate somewhat from the ellipsoidal shape assumption. In fact, in some of our UFOs (Figure 2), we see evidence of clumps (29312), spiral features (187160), or multiple components/possible mergers (e.g., 167032 and 172813). These nonsmooth components could impact the ensemble axis-ratio distribution leading to higher or lower inferences on the fractions of one or more shape subcategories. Additionally, if the UFOs are not a unified population of intrinsic shapes that can be characterized by a single distribution, then our modeling assumptions start to break down and the inferred parameters do not necessarily tell us about the true breakdown of intrinsic shapes.

This paper presents an analysis of the stellar populations and structural properties of an unexpected population of elongated, OFGs from the JADES survey. Here, we discuss why these results are surprising in terms of existing studies of the intrinsic shapes of galaxies, what drives the optical faintness of these objects, and their subsequent evolution.

Probably the most surprising aspect of this population of objects is their structure: they have fairly large radii, nearly exponential light distributions, and are highly elongated. Beginning with their sizes, the leftmost panel of Figure 9 shows that a population of OFGs exists (which we call UFOs) whose sizes are larger than average SFGs at their masses and redshifts. This is surprising because at a fixed dust mass, galaxies with a dust distribution that is more concentrated will have higher dust column densities and hence larger values of dust attenuation (e.g., Nelson et al. 2014). These higher dust column densities could come from compact sizes or concentrated distributions (i.e., high Sérsic indices). As such, we may expect OFGs to preferentially have small sizes and/or high Sérsic indices. Thus, while the naive physical expectation for the structures of OFGs would be to have compact sizes to drive their high values of dust attenuation, this does not appear to be the case with the distributions of the optically faint and bluer F444W-selected parent samples being similar. In particular, among objects with stellar masses greater than 10¹⁰ M_⊙ and sizes >2 kpc, 39% are optically faint while among objects with sizes less than 2 kpc, 22% are optically faint. Optical faintness does not appear to correlate with size. This is surprising given the findings of Gómez-Guijarro et al. (2023) who find that among SFGs at 3 < z < 7.5, those with A_V > 1 have ∼30% smaller sizes. However, recent simulations find that orientation (i.e., viewing angle), and not physical characteristics, is the main determinate of whether a galaxy will be optically faint or not, with galaxies only appearing optically faint in some fraction (depending on the object) of the orientations (Cochrane et al. 2024).

The second surprise is their Sérsic indices: the bulk of the UFOs and the OFGs in general have n < 2 (see Figure 9). Previous multiwavelength studies looking at the morphologies of dusty galaxies in both optical/UV and submillimeter wavelengths generally find that these sources are more compact at longer observed wavelengths (e.g., Hodge et al. 2016; Tadaki et al. 2017; Nelson et al. 2019) suggesting that these galaxies may be in the process of building dense stellar bulges (e.g., Lang et al. 2014; Whitaker et al. 2017) with dust attenuation causing these bulges to be missed at shorter wavelengths. Bulge-dominated galaxies are typically characterized by high Sérsic indices reflecting more centrally concentrated light profiles, which is the opposite of what we see in the sample of OFGs studied in this paper. The UFOs and OFGs in our sample have light distributions more indicative of exponential profiles suggesting they are strongly disk-dominated (though prolate is also a possibility). The massive portion of the galaxies in our sample are approaching stellar masses at which they will likely quench yet most show little evidence of having or building bulges which appears to be a prerequisite for quenching (e.g., Lang et al. 2014; Whitaker et al. 2017). Although most of the UFOs have low Sérsic indices in F444W emission, it is possible that even the F444W light is attenuated in the central regions and a bulge component could appear in even redder wavelengths. MIRI or ALMA data would be needed in order to definitively rule out the presence of a highly obscured bulge component.

The third surprise is that most of the UFOs are elongated. These objects are selected based on their fluxes, colors, and sizes; there is no explicit selection for them to be elongated. The most obvious explanation perhaps for a highly reddened cohort of large galaxies to be elongated is that we are looking at an edge-on subset of a disk galaxy population. Another possible explanation for high dust attenuation is geometric: if galaxies are intrinsically disky, then the dust column density will be much larger when viewed edge-on than face-on (e.g., Wild et al. 2011; Patel et al. 2012). If the UFOs are in fact disk-dominated objects, viewing them edge-on could explain why they are optically faint and would not have appeared in previous studies.

While there is significant physical reason to believe the UFOs are intrinsically edge-on disks, axis-ratio modeling suggests a significant fraction of objects may be intrinsically prolate instead of oblate/disky. This type of statistical modeling was not possible in Nelson et al. (2023) given the small sample size; having the large sample provided by JADES is essential. In a framework assuming UFOs are a single population, axis-ratio modeling suggests that at least 30% of UFOs are intrinsically prolate. This is very surprising. In an intrinsically prolate population, a physical argument akin to the higher dust columns in inclined disk galaxies would suggest a preference for optical faintness when looking at the prolate object with the long axis along the line of sight. In this configuration, the object would appear small and circular (perhaps like some of the smaller OFGs in our sample) and it would physically make sense for these objects to have large dust columns. Viewed with the long axis perpendicular to the line of sight, it is harder to understand how dust attenuation could be strong enough to produce optical faintness.

To put these results in context, we compare them to previous studies of the intrinsic shapes of galaxies as a function of the redshift and stellar mass. Our stellar population modeling suggests that these objects have fairly high stellar masses; observations of SFGs with HST over the redshift range 1 < z < 3 have shown that high-mass (M_* ≥ 10¹⁰ M_⊙) SFGs are predominately disks (e.g., van der Wel et al. 2014; Zhang et al. 2019), especially at z < 2. However, at higher redshifts (z ≳ 2) and lower masses (M_* ≤ 10^9.5 M_⊙) the shapes of SFGs become dominated by prolate objects. In other words, the fraction of star-forming prolate galaxies increases toward higher redshifts and lower masses, while the opposite is true of star-forming disks.

For this reason, and given the fairly wide range of inferred UFO stellar masses, we consider the possibility that the UFOs are composed of a lower-mass population and a higher-mass population with distinct axis-ratio distributions. Figure 13 shows the axis-ratio distributions of UFOs and the parent population of F444W-selected galaxies divided by mass at M_* = 10^10.4 M_⊙. As expected, the axis-ratio distribution of the low-mass UFOs is more skewed toward lower q values than that of the high-mass UFOs, and axis-ratio modeling suggests nearly 100% are prolate. There are not enough galaxies in the high-mass bin to determine their intrinsic shapes from statistical modeling but their axis-ratio distribution is certainly not that of a purely prolate population.

Zoom In Zoom Out Reset image size

It has been shown in Vega-Ferrero et al. (2024), based on a machine-learning-based classification scheme for inferring galaxy morphologies, that approximately half of galaxies visually classified as disks from CEERS imaging are more consistent with prolate or spheroidal populations than being true disks. Thus, we may indeed be looking at some objects that visually appear to be disks, but in reality, are prolate. Further, it is worth noting that populations of massive but quenched prolate galaxies at z ∼ 0 have been identified in cosmological simulations (e.g., Ebrová & Łokas 2017; Thob et al. 2019) as well as observationally (e.g., Tsatsi et al. 2017). Ebrová & Łokas (2017) suggest that these prolate objects are either formed prolate (or become prolate early in their lifetimes) or formed through radially aligned mergers that induce the stellar shapes toward prolate ellipsoids. These types of observations could be consistent with our plausible finding of star-forming prolate intermediate-mass (M_* < 10^10.4 M_⊙) galaxies at 2 < z < 6, which could then be the progenitors of massive quenched galaxies at z ∼ 0. Although we can rule out that UFOs are not a bulge-dominated population in F444W light, definitively determining the intrinsic shapes of these objects will require kinematic measurements.

Turning back to why these objects are optically faint, in Figure 6 we show the distribution of OFGs and our F444W-selected parent sample in a plane of two likely culprits: age and dust attenuation, which both act to redden the SEDs of galaxies. The most significant effect is dust attenuation: most galaxies at 2 < z < 6 and F444W < 24.5 that have A_v > 2 are optically faint. There is a secondary trend with age in which the OFGs selected from JADES are on average slightly older than the parent population but this is decidedly subdominant to the dust attenuation. Further, as shown in Figure 9, 58% of galaxies in our F444W-selected parent sample with q < 0.3 are optically faint. On the other hand, both the size and Sérsic index do not appear to drive optical faintness (at least as defined in this paper). Thus, having a low projected axis ratio appears to be the best predictor of optical faintness among the standard morphological indicators–effective radius, Sérsic index, and axis ratio. That said, a very small fraction of the compact OFGs (i.e., non-UFOs) have q < 0.3 so inclination does not appear to provide an explanation for the optical faintness of these more compact objects.

Finally, we consider how this enigmatic galaxy population may evolve toward lower redshifts. With stellar masses of log(M_*) > 10 at z ∼ 2 − 3, UFOs are likely to evolve into quiescent galaxies with log(M_*) > 11 in the local universe (e.g., Behroozi et al. 2019). In fact, in Figure 5, we can see that a small fraction of our three samples have SFRs up to ∼3 dex below the z = 2.5 SFMS from Leja et al. (2022) suggesting that these galaxies have already ceased their star formation. This is a sharp transition to quiescence at a characteristic mass of log(M_*) ∼ 10.4 consistent with previous studies finding that quenching is efficient at a similar mass (e.g., Contini et al. 2020). With the typically small quantity of dust attenuation in quiescent galaxies in the local universe, the dust attenuation seen in these objects at z > 2 will need to almost completely disappear in the intervening time (e.g., Whitaker et al. 2021). Significant structural change is also likely as they need to transition from prolate or oblate morphologies to having significant dominant bulge components.

We have performed a morphological and stellar population analysis of a sample of 112 OFGs with a specific focus on the elongated subset of OFGs, dubbed UFOs. In particular, we explored how these galaxies are situated in certain fundamental galaxy scaling relations such as the size–mass plane and the star-forming main sequence as well as investigations into how the stellar populations vary between the inner and outer regions and finally an in-depth look at the morphological parameters of these objects such as their sizes, Sérsic indices, and axis ratios. Our main goal has been to place stronger constraints on the nature of the UFO galaxy population and to place the optically faint and UFO populations within the larger context of the massive, SFG population. Our main results are summarized below.

• 1.  
  We identify 112 optically faint objects in the JADES survey of which 56 are classified as UFOs with apparent sizes larger than 025. The UFOs are typically between 2 < z < 4 with a range of stellar masses (M_* ∼10⁹⁻¹¹ M_⊙), high SFRs, and high A_V . Compared to the mass- and redshift-matched sample of bluer F444W-selected galaxies in JADES, the UFOs tend to have higher stellar masses and SFRs but have much lower stellar masses than the Herschel-detected ultrared galaxies from Ma et al. (2019).
• 2.  
  UFOs have red colors throughout the extent of their bodies but still possess slight negative color gradients with outskirts that are less red than the interiors. These color gradients are likely driven by increased central dust concentration and not by the ages of the stellar populations.
• 3.  
  The observed UFO axis-ratio distribution in F444W is consistent with observing some combination of randomly oriented disks and randomly oriented prolate galaxies. We perform a detailed Bayesian modeling of the q distribution using the BEAST and find that the UFOs can not be uniquely said to consist of one single galaxy shape. It is possible that the UFOs consist of a lower-mass prolate population and a higher-mass oblate/disk population, but without a larger sample, it is hard to verify this.
• 4.  
  We find that the strongest predictors of optical faintness in the galaxy populations are A_V and q. That increasing A_V is a strong predictor of optical faintness is not surprising given that larger dust attenuation will naturally lead to less optical light escaping a galaxy. Morphologically, it is somewhat surprising that the strongest predictor for optical faintness is a low q as it was thought that small physical size would be the easiest way to drive large enough dust columns to make galaxies optically faint, but that is not what we see given the large sizes of UFOs.

UFOs are an unanticipated galaxy population that JWST has made easy to reveal and study. The existence of OFGs with extended structures was surprising owing to the expectation that only compact sizes could lead to large enough dust columns and as early JWST identified OFGs were found to be relatively compact (Gómez-Guijarro et al. 2023). The suggestion that a significant fraction of these objects are prolate instead of disks is an interesting possibility with implications for the z > 2 shapes of SFGs and how these shapes evolve to the present day. Future work will increase these samples to larger numbers and seek stronger constraints on the morphologies with kinematics from JWST spectra allowing us to come to a definitive conclusion on the intrinsic 3D shapes of UFOs.

J.L.G. gratefully acknowledges the support provided by NASA through grants 20-ASTRO20-0200 and HST-AR-16146 and support from the NRAO through grant SOSPADA-025. E.J.N. acknowledges support provided by NASA through grants HST-AR-16146 and JWST-GO-01895. The research of C.C.W. is supported by NOIRLab, which is managed by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation. S.H.P. acknowledges support through grant JWST-GO-2561. The Cosmic Dawn Center (DAWN) is funded by the Danish National Research Foundation under grant No. 140. A.J.B. acknowledges funding from the First Galaxies Advanced Grant from the European Research Council (ERC) under the European Union's Horizon 2020 research and innovation program (grant agreement No. 789056). W.B. acknowledges support by the Science and Technology Facilities Council (STFC), ERC Advanced Grant 695671 "QUENCH." K.B. acknowledges support from the Australian Research Council Centre of Excellence for All Sky Astrophysics in 3 Dimensions (ASTRO 3D), through project No. CE170100013. E.C.L. acknowledges the support of an STFC Webb Fellowship (ST/W001438/1). D.J.E. is supported as a Simons Investigator and by a JWST/NIRCam contract to the University of Arizona, NAS5-02015. R.H. acknowledges support from the Johns Hopkins University, Institute for Data Intensive Engineering and Science (IDIES). R.M. acknowledges support from the Science and Technology Facilities Council (STFC), by the ERC through Advanced Grant 695671 QUENCH, and by the UKRI Frontier Research grant RISEandFALL. R.M. also acknowledges funding from a research professorship from the Royal Society. B.D.J., G.R., M.R., and B.E.R. acknowledge support from a JWST/NIRCam contract to the University of Arizona NAS5-02015. This work utilized the Alpine high-performance computing resource at the University of Colorado Boulder. Alpine is jointly funded by the University of Colorado Boulder, the University of Colorado Anschutz, and Colorado State University.

Facility: JWST - James Webb Space Telescope (NIRCam).

Software: astropy (Astropy Collaboration et al. 2013, 2018, 2022), Dynesty (Speagle 2020), GALFIT (Peng et al. 2002, 2010), Lenstronomy (Birrer & Amara 2018), matplotlib (Hunter 2007), numpy (Harris et al. 2020), photutils (Bradley et al. 2020), Prospector (Johnson et al. 2021).