JWST’s PEARLS: dust attenuation and gravitational lensing in the backlit-galaxy system VV 191 (2024)

1 Introduction

Dust is a key ingredient in both observed and physical properties of galaxies. It obscures and scatters stellar light, and enables star formation by acting as a catalyst for molecular gas formation. It is the ingredient in the ISM that most dramatically changes how we perceive a galaxy’s starlight, and the site of much of the astrochemistry. On broad scales, correction for the effects of dust is an important part of many techniques for estimating astronomical distances.

The cumulative effects of grains are often traced via attenuation (combined extinction and scattering) from theultraviolet to near-infrared regimes, and thermal emission in the far-infrared. Interpretation of theUV-optical-IR (UVOIR) attenuation signature is limited by the need to understand the local radiation field and three-dimensional mix of stars and dust, both of which are features of the many routines designed to model the observed spectral energy distributions (SEDs) of a few individual galaxies in great detail (e.g., Holwerda et al., 2012; De Looze et al., 2012; Gentile et al., 2015; Verstappen et al., 2013; Hughes et al., 2014; Allaert et al., 2015; De Geyter et al., 2014, 2015; Mosenkov et al., 2016, 2018) or populations with preset priors (e.g. Driver et al., 2018; Bellstedt et al., 2020, 2021; Thorne et al., 2021; Zou et al., 2022).In contrast to the extinction-based techniques for probing dust effects, the use of far-IR emissionis independent of the need for any background source, but becomes progressively lesssensitive for colder dust temperatures, requires knowledge of the emissivity behavior ofgrains much smaller than the wavelengths in question, and is generally observed at large angular diffraction limits.

Star-by-star observations within the Milky Way, and recentlyelsewhere in the Local Group, have been used to measure wavelength dependence of extinction(absorption plus scattering). A widely-used parametrization is theCCM curve (Cardelli et al., 1989), which is characterized by the ratio of total to selective extinction in the $V$ band, $R=A_{V}/E_{B-V}$¹¹1where $A_{V}$ is the extinction in magnitudes at $V$ and $E_{B-V}$ is the color excess - difference in extinctions - between $B$ and $V$ bands, the strength of the 2175-Å bump, and UV slope. In the near-infrared bands, a power-lawform in wavelength has been found useful at least from 0.5–5$\mu$m, $A_{\lambda}\propto\lambda^{-\beta}$ with estimates of $\beta$ ranging from 2.07 ((Wang & Chen, 2019) to 1.65 (using the summary data in Table 3 ofZhang & Yuan 2022). It has long been clear that there is not a universal extinction law, since variations are seen between Galactic environments as well as between our part of the Milky Way and, for example, the Magellanic Clouds.

In the context of almost all observations of distant galaxies, the extinction law (reflecting the grain properties)is inextricably mixed with geometric factors involving the relatave distributions of emitting sources and grains (Calzetti 2001, Chevallard et al. 2013), so weare properly deriving the attenuation behavior which combines both factors. Many studies have derived attenuation laws from the colorbehavior of galaxy samples and radiative-transfer models, or frommodels for the spectral energy distributions (SEDs) of galaxies compared to their observed properties, sometimes with a luminosity constraint including far-infrared emission (e.g., Salim et al. 2018). The Calzetti et al. (1994) attenuationlaw, derived for the inner regions of starburst galaxies, has proven to be of surprisingly wide applicability, a reasonable fit to the behavior of whole galaxies as well as the outer regions of some star-forming galaxies. However, its applicability into the infrared is limited because, as defined, this relation goes to zero at 2.2 $\mu$m.

Generically, attenuation laws for galaxies or large parts of galaxies will be flatter (have a smaller wavelength dependence) than the extinction laws applying to individual stars, due to the mixing of regions of differing extinction within the area measured (and possibly significant effects of scattering into the beam). This is true for the Calzetti et al. (1994) form versus the Cardelli et al. (1989) curve, and for the range ofnear-IR indices $\beta=1.0-1.4$ found for integrated galaxy light by Salim et al. (2018) compared to the steeper values $\beta\approx 1.8$ found for single stars as noted above. Our HST and JWST data allow a newapproach to this issue in the $0.6-4.4$ $\mu$m range.

In earlier work, we have explored the use of backlit (occulting) galaxy pairs to improve our understandingof attenuation in galaxies. In these cases, a background galaxy backlights alarge part of a foreground system, providing spatially continuous light (ideally froma smooth E or S0 galaxy) so that all the dust structure can be detected. While limitedto a small (random) subset of galaxies, this technique preserves the angular resolution ofUVOIR imaging, avoids the bias of small background objects toward preferential detection in more transparent spatial “windows (Holwerda et al., 2007), and works equally well independent of grain temperature (specifically even for the coldest grains, whose emissivity can be very low for far-IR and submillimeter detection).

This technique has been applied to a range of spirals, with the sample size growing as survey data improved the potential sample size.Small samples in ground-based images were considered by Keel (1983),White & Keel (1992), White et al. (2000),and using HST data by Elmegreen et al. (2001),Keel & White (2001a), Keel & White (2001b), and Holwerda et al. (2009).This technique was extended into the ultraviolet, mostly using GALEX data, byKeel et al. (2014), with more detailed treatment of the uncertainties introducedby imperfect symmetry in the galaxies. Addition of spectroscopic information can improve the separation of contributions from the two galaxies (Domingue et al., 2000) and provide a finer wavelength sampling of the attenuation (Holwerda et al., 2013). When both far-infrared (FIR) and backlighting measurements are possible, Domingue et al. (1999) found broad consistency in the dust masses derived in these complementary ways. These studies have provided estimates ofattenuation and dust masses, hints of fractal dust structure in well-resolved spiral arms,and has shown that the Calzetti et al. (1994) effective attenuation law applies to dust acrossmany spiral disks as well as in the star-forming regions for which it was initially derived.Comparison with simulations indicates that this behavior with wavelength, quite different from thatcharacteristic of extinction measured star-by-star in nearby galaxies, results from the mix ofoptical depths within each resolution element. Keel & White (2001a) showed, using HST data,that the derived attenuation law becomes steeper as the resolution element becomes smaller,across size ranges 50-200 pc, fitting with this idea.A handful of galaxies are known to host dust lanes or patches so far from their centers that the grains must be very cold ($<15$ K) (Holwerda et al., 2009), which is in the right sense to account for a low-temperature component of the overall emission of galaxiesincluding Herschel data (e.g., Bourne et al., 2012; Ciesla etal., 2012; Smith etal., 2012; Cortese etal., 2012, 2014; Clark etal., 2015; Beeston etal., 2018).

“Prime Extragalactic Areas for Reionization and Lensing Science” (PEARLS) is a JWSTGTO program (PI: R.Windhorst) designed to provide the community with medium-deep imaging in up to eight near-infrared filters (Windhorst etal., 2023). As one of the main program goals is to trace the evolution and assembly history of galaxies over cosmic time, the impact of extinction and scattering by interstellar dust grains on the observed tracers of such evolution poses a potential source of bias and uncertainty.To assess this issue, one of the PEARLS targets is a pair of galaxies with a fortuitous chance alignment on the sky: a face-on spiral galaxy partially backlit by an elliptical galaxy. In such a configuration, the properties ofthe dust in the foreground spiral galaxy can be measured in an absolute sense, and to very low dust column densities, and in both arm and interarm regions. This target, VV 191, serves as a GTO pilot program to demonstrate the power of JWSTfor future GO surveys that target a larger and more diverse sample of galaxies. In addition, unknown to us when selecting the target and clearly revealed by the JWST data, one of the components of the galaxy system observed here strongly lenses a distant background galaxy, illustrating multiple configurations and outcomes of overlapping galaxy images at once.

Previous work on dust in backlit galaxies has concentrated on grand-design spirals, often2-armed systems,where application of symmetry to model the foreground galaxy is manageable. VV 191 projects theouter disk against such a bright part of the background galaxy that we can complement earlier work by studying the disk attenuation of a multiple-armed spiral, where the dust lanes have flocculent structure. VV 191 wasinitially recognized as an occulting, rather than merging, system during the Galaxy Zoo survey for suchpairs (Keel et al., 2013). It nearly matches the schematic ideal for analysis of such a pair,with a face-on spiral roughly half-projected against a background E or S0 galaxy (itselfnearly circular in projection), and the spiral nearly face-on (photometric axial ratio within 0.15 of circular at all radii). The backlighting near the edge of the disk is so strong that uncertainties in correcting for the spiral’s own light barely contribute to the error budget in retrieving the attenuation. The analysistechnique and uncertainty analysis are described in detail in section 2.6.

We present here a combination of HST and JWST imaging studies for VV 191, showing the power of carrying this technique into the near-infrared (NIR), using a galaxy pair with nearly ideal geometry for this analysis. Adding the near-IR bands was expected to provide more leverage for measuring the reddening slope, and give accurate attenuation measurements in regions that might be effectively saturated at much shorter wavelengths. As it happened, the F090W and F150W JWST data also provide higher spatial resolution of dust structures.

2 Observations and data analysis

3 Spatial structure of dust attenuation

Important results came from the HST F606W snapshot image alone, informingthe selection of this pair for JWST observations. Most notable, the networkof dust lanes (both those tracing the spiral pattern and weaker lanes cross-cutting in a “‘dust web”)has a sharp outer boundary, so pronounced that it appears in the azimuthally averagedtransmission profile (Bradford 2016). Thisappears in HST and JWST data (Figure 3). These profiles also showdips in transmission where dust arms cross the sector of usable signal-to-noise. For reference, the region analyzed for the profiles in Fig. 3 is outlined in Fig. 4. The region more than 21.5″from the spiral core, with no evidence of dust attenuation, has a mean derived transmission within 0.4% of unity in all five bands from F606W to F444W, furnishing a check on the accuracy and applicability of the galaxy models. However, there are ripplesin the “zero-attenuation” level, including unphysical values greater than unity, at levels up to 1.5% and closely matchingin bands from F150W to F444W. This amplitude is greater than plausible flat-fielding uncertainties in NIRCam⁴⁴4B. Sunnquist, private communication.. We suspect these reflect mismatches between galaxy structure and our elliptically-symmetric models; this is an aspect of the low-level systematics which we attempt to deal with in fitting the reddening behavior in Section 4. The fact that they are consistent between filters and confined in radius gives reason to think that the systematic errors are similarly confined in radius within the foreground disk so will be approximately constant across each of the regions we fit.

We illustrate these features across our observed wavelength range with the two-dimensional transmission maps from both HST and JWST data in Figure 5. These illustrate the concentration of dust signatures to a system of lanes, some of which cut nearly perpendicular to the pitch angle of the overall spiral pattern (in what we term the dust web, some parts of which fit with spiral spurs of features as more often described). The shaping of these structures so far out in the disk appears to include processesbeyond the differential rotation which shapes so much of spiral patterns.

As a broad summary of the results, Figure 6 shows the cumulative distributions of transmission in broad radial bins. This is relevant to, for example, priors on color measurements of supernovae in galaxies of this type, and this technique offers the possibility of extension to a broader range of galaxy types and impact parameters.Figure 6 provides a complementary view to the averaged radial profiles in Figure 3, also useful in assessing the probability of a line of sight having a particular level of transmission all the way through the disk. This compares the cumulative (fractional) distribution of area with attenuation greater than a given level (transmission less than that value) for the three filters with the best dust measurements. This covers the well-measured range from 14.5-21 kpc, divided into inner and outer halves as well as showing the entire range.

4 The broadband attenuation law in VV191b

The slope of the reddening-attenuation relation (reddening law) combines information on the grain populations and spatial distribution, when observed at the resolutions available for galaxies (in contrast to sight lines to individual stars, where scattering and blending of regions with different column densities are negligible).We compare the distributions of points in the galaxy to the reddening trajectories predicted for various ratios of total-to-selective extinction $R=A_{V}/E_{B-V}$ or power-law index $\beta$.For a range of $A_{V}$ values, we calculate the fraction of the flux from an elliptical galaxy emerging in each of our passbands, incorporating photon weighting (Hogg, 2022) within the filter bands by folding in the SDSS spectrum of the core of VV 191a, the 1.5-µmspectrum of a typical luminous elliptical from François et al. (2019), andin the near-UV F336 band, a composite of data for NGC 4472, incorporating averaged IUE spectra (Norgaard-Nielsen & Kjaergaard 1981, Oke et al. 1981, Burstein et al. 1988) plus a Kitt Peak large-aperture spectrum extending near the atmospheric cutoff, all resampled to the redshift of VV 191a. Both photon weighting and finite passband widths mean that progressively more flux is detected at increasing values of $A_{V}$ than would be predicted from the central wavelength alone.

The linear resolution of these data ranges from 60–180 pc. Compared to the fractal-like distribution in the backlit spiral in AM0500$-$620 (Keel & White, 2001a), these values are in the regime where we might expect significant “graying” of the attenuation law as regions of different optical depth are mixed within each resolution element.

To the extent that a large part of the elliptical galaxy is unobscured so our model reflects its true brightness profile, this technique measures absolute attenuation, sensitive to gray attenuation as well as reddening.

We measure the slope of the attenuation (reddening) law between each two of the three filters with the best attenuation measurements (F606W, F090W, F150W), matching the PSF core to the F606W data, by varying the slope between the filters (CCM parameter $R$ or power-law index $\beta$) and fitting to the curve formed by the transmissions predicted for the two bands as the total extinction is varied (for bookkeeping and comparison, we use $A_{V}$, following the prescription of Cardelli et al. (1989) since the Calzetti et al. (1994) form is not defined in our redder filter bands).Fitting attenuation laws to these clouds of points poses several specificchallenges. The predicted transmission varying a given parameter ($R$ or $\beta$) foreach filter pair is a one-parameter family anchored at both (1,1) and (0,0), as shown in Fig. 7. In the regime of small attenuation (transmission $\approx 1$), even smallsystematic errors in the data values (Section 3) will exercise undue influence on the fitting results. Such small systematics would be multiplicative if due to a model shortfall for the elliptical galaxy or additive if due to structure in the spiral; they are always small in these data, less than 4%, so we cannot distinguish and treat them as multiplicative for convenience.We explored several approaches to deriving these parameters, recognizing thatstill more detailed modeling may be warranted, going beyond the scope of this analysis and to be explored in future work.Working in a low-attenuation regime into the near-IR bands, the fitting isless well constrained than in, for example, the HST observations of NGC 3314reported by Keel & White (2001b), where dust lanes with larger attenuations weretargeted including shorter wavelengths and only values relative to the diffuse surroundingdust were retrieved, rather than the absolute attenuations we seek in this work.

These data are not accurate enough to discriminate between CCM and power-law formsfrom 0.6-1.5 $\mu$m; there are values of each parameter which give virtually identicalattenuation behavior in each filter pair over the attenuation range we observe although the value of $A_{V}$ between equivalent parts of the two curves differs significantly at large attenuations).

Respecting the scatter of pixel values, which may be large compared to the changes we seek to measure, cases of asymmetric scatter in one or both filters, and the two-dimensional nature of the scatter since it may be comparable in each filter of a fitting pair, we experimented withseveral ways to fit reddening slopes to data from each pairof filters with significant dust measures over large regions (F606W, F090W, F150W). Ordinary $\chi^{2}$ minimization gives results very sensitive to small systematic offsets in the zero-attenuation level, since each fitting function by itself gives a one-parameter family anchored at both ends. We first tried fit to the curves bisecting the lowest-transmission sections of each point cloud, with a heuristic cut to reduce the influence of systematics at high transmissions, and also used binned versions of each point cloud to attempt fits following the ridgelines as defined by contour plots. Our final fits for the reddening behavior combine features of these approaches, with explicit allowance for systematics in galaxy removal. As well as the reddening slope (separately for $R$ and $\beta$, we fit for a multiplicative factor near unity (in practice $\pm 3\%$) foreach filter as nuisance parameters. In some regions, there may be an additional component of fine-scale structure in the foreground galaxy (that is, surface-brightness fluctuations). Inspection shows that actual noise and real structure (due to the distribution of luminous stars) contribute about equally in the long-wavelength F356W and F44W bands in the inner zones we fit, so the color of the brighter and dimmer spots depends on the stellar population. This would be manifested in the two-filter data by a component near unity transmission with a different slope than the larger-scale reddening behavior; we try to avoid this effect by fitting onlypoints with transmission $<0.98$ in the bluer passband of each pair.

Our fits start by summing the pixel values of transmission in a filter pair into a grid, normally summed in bin of 0.01 in each transmission value. Passing a $3\times 3$-pixelLaplacian gradient filter over such a smoothed array allows numerical retrieval of theridgeline of each distribution, where the attenuation parameter and systematic scalingcan be fit in a way analogous to $\chi^{2}$ minimization, minimizing the 2-dimensional deviations of each point on the gradient-derived ridgelines from the transmission curve, although the expected scatter is nowfar removed from the individual pixel uncertainties. This approach therefore incorporates the 2-dimensional nature of the uncertainties for each pixel, and gives a quantitative way of assessing the accuracy of the best-fit parameters. We present the resultsfrom this gradient-fit approach in inTables 2 and 3 for $R$ and $\beta$ respectively.

These fits have been done for three regions spanning the arms as defined by dust lanes and two “interarm” regions (Figure 8), which together largely fill the region of highest S/N for attenuation. The fitting procedure is illustrated in Figure 7. We list the best-fit values for each region andfilter pair using each approach: values of $R$ in Table 2 andpower-law index $\beta$ in Table 3. The 2-band fits use pixel masks produced independently for each filter (and independent of the pixel masks used in fitting the galaxy models), to reject stellar associations in VV 191b and background galaxies. When matching resolutions, the masked regions are propagated to include all pixels whose value would be affected by more than 0.05% by a masked region in either filter band after the convolution, well below the intensity uncertainties in transmission values. The point clouds, ridgelines, and fitted 2-color transmission curves foreach pair of the three filters with widespread dust detection and each measured region are shown in Fig. 9. This illustrates how similar the best-fit $R$ and $\beta$ curves are for each filter pair. The behavior for $R<1$, formally found for the F090-F150 filter pair, becomes highly nonlinear (and poorly determines from existing data), so we list $R=1$ as a minimum value and show this in Fig. 9.

Across all regions and filter pairs, the value of $R=A_{V}/E_{B-V)}$ is very low, 1.94 from F606W-F090W and even smaller for filter pairs involving F150W (Table 2).This change with wavelength shows that the common fitting functions do not capture the broadband behavior in this situation (so a global fit across the well-measured bands from 0.6-1.5$\mu$m is not useful at this point).

We carried out analogous fitting for the index of a power-law attenuation law, where $A_{\lambda}=A_{V}(\lambda/\lambda_{V})^{-\beta}$, with results listed in Table 3. As with the CCM form, for each zone and filter pair, we did fits to the ridgeline of the point clouds as derived from the Laplacian transform of the gridded points, and as consistency tests, compare eyeball fitsboth to the point cloud and to the ridgeline of the smoothed distribution of data points. We include weighted means for the entire data set and separately for the arm and interarm zones, derived by analysis of all points in each category rather than by averaging the zone values. For adequate signal-to-noise ratio in tracing the gradients, some zones (indicated with asterisks) had their gridded point distribution smoothed by a Gaussian of $\sigma=0.03$ in each transmission value rather than the default 0.02. Comparison of zones where both Gaussian values give useful results indicate that the scatter introduced by this choice is slightly larger than the listed uncertainty of the fit. Very broadly, regions where regions of different attenuation are blended withineach resolution element will show flatter measured attenuation (large $R$, or $\beta$ closer to zero). The fitted values show trends of steeper slopes (more positive values of $\beta$) at longer wavelengths, smaller projected radii within the galaxy, and in interarm regions compared to the spiral arms.

Our fitting procedure gives useful bounds on the parameter values best fitting each data set. However, in most cases theattenuation tracks in these diagrams predicted by either CCM or power-law forms do not completely describe the data. Typically,the data exhibit more curvature in the reddening tracks than the fits - that is, more reddening than predicted at small attention and less than predicted at large attenuation. This is broadly consistent with larger-attenuation pixels including areas with different extinction within each resolution element, but the situation invites further, more detailed modeling.

The mean (all-region) values of $R$ (and $\beta$) in Tables 2 and 3 imply extinctions in these filter bands of $A_{F606W}/A_{V}=0.83$, $A_{F090W}/A_{V}=0.30$, and $A_{F150W}/A_{V}=0.047$. Slope fits to the longer-wavelength bands versus F606W give limits(combining all 5 measured zones) $A_{F356W}/A_{F606W}<0.035$ and$A_{F444W}/A_{F606W}<0.021$, which imply nonrestrictive limits onthe power-law index $\beta>0.4$ and $\beta>0.7$ respectively. Implicitly, the lack of relation between flux in these bands and attenuation measured at shorter wavelength shows the nondetection of emission from hot dust or PAH molecules in these (cold) dust lanes.

Because the red colorof the background galaxy limits the ultraviolet S/N so severely, we use only a subsetof 3510 pixels in fitting region 1 for F336W encompassing the Y-shaped dust lane projected closest to the nucleus of VV 191a, and constructed the 2D grid for fittingthe ridgeline of the F336W-F606W attenuation distribution using bin sizes of 0.05 in each filter. As shown in Figure 10, the extinction ratio between F336W and F606W bandsis best described by $R=1.7\pm 0.4$ whether one fits the ridgeline, a running mean when sorted by F606W transmission, or by bisecting the cloud of individual pixel values. However, the gradient measures of the estimated ridgeline value show evidence of a further decrease at larger attenuations, more characteristic of a screen than the more likelymix of extinction values. The low values of $R$ in some of ourfits may need to be understood as indicative, because thebehavior becomes so nonlinear for such low $R$ and there is little data on reddening behavior which is this strong.

The range of reddening behavior in galaxies as functions of galaxy type, metallicity, local environment, and averaging scale-size is surely broad and not yet understood. The most comparable previous results are from backlit galaxies observed with HST at linear resolutions $\approx 100$ pc. For disk dust in fairly luminous spiral disks, some systems show Milky Way-like reddening. The broad dusty arm in AM1316$-$241 (Keel & White, 2001a) gives $R_{V}=3.4\pm 0.2$, and multiple arm segments in NGC 3314 (Keel & White, 2001b) yield $R_{V}=3.5\pm 0.3$ with no radial trend. However, the spiral arms in AM0500$-$620 (Keel & White, 2001a) have $R_{V}=2.5\pm 0.4$, while the foreground system in NGC 1275 has a very different, almost grey extinction law ($R>4$, formally much greater, from Keel & White 2001a). This galaxy is moving at high velocity through the intracluster medium of the Perseus cluster, so its grains may have been exposed to alteration by the hot gas. Comparison of spectral slopes in star-forming regions led Calzetti et al. (1994) to propose a reddening law which has seen wide application, giving consistent results not only for the starburst regions where it was first derived, but for the outer parts of spirals in the UV (Keel et al., 2014). Further work (Calzetti et al., 2000) found a value $R_{V}=4.05$ for starburst regions in anchoring the parameters of this relation, which certainly responds to both the grain properties and the relative distributions of stars and dust in these intense star-forming regions.

These kinds of measurements, summing over substantial regions, unavoidably mix regions of various optical depths, so this mix affects the results along with the (astrophysically more interesting) properties of the grain populations. We see the effects of this mixing by fitting power-law reddening slopes $\beta$ after block-averaging the transmission maps by factors 2, 4, and 8, propagating the masked regions in each case cross the 5 zones shown in Figure 8). In the mean, $\beta$, measured between F606 and F090W, decreases by 0.41 between 2 and 8-pixel averagingsteps, and decreases by a mean of 0.36 between F606W and F150W. Decreasing $\beta$ moves in the direction of grayer behavior, as expected for mixing regions of different attenuation. As the number of points decreases in this averaging, our ridgeline procedure ceases to be useful, so the block-averaged data were fitted using minimization about the predicted curves for the points themselves. Thus these are strictly comparable to each other but not necessarily to the ridgeline fits used for the full-resolution maps.

The extinction behavior of the grains themselves is best derived when we can study their effects star by star, along essentially a single line of sight. This is how the classical Milky Way extinction curve was determined (albeit from a more limited part of the Galaxy than one might like), and has been done in the Magellanic Clouds and M31 (Clayton et al., 2015) as well; the reddening law in the SMC in particular becomes much steeper than in the Milky Way going into the ultraviolet. Farther afield, supernovae (especially when observed polarimetrically) offer ways to isolate the effects of the grain properties, often implying $R\approx 2$ (Wang & Wheeler, 2008; Patat et al., 2015). There is clearly wide scope to improve our understanding of how the properties and distribution of grains vary within and among galaxies. Our results for VV 191b, probing the outskirts of disk dust and extending to wavelengths where this behavior has been poorly known, show stronger reddening (steeper attenuation law with wavelength) than otherwise seen in galaxies, even with a possible role for mixtures of attenuation below the resolution limit flattening the observed behavior. This might hint at a very different population of grains, more weighted to smaller grains than in our neighborhood.

5 Gravitational Lensing

Our first inspection of the NIRCam images revealed an object 2.8″from the nucleus of the elliptical galaxy VV 191a, with a bright core and outskirts forming an arc nearly centered on the foreground galaxy. It is much redder than the elliptical galaxy, ruling out any sort of disrupted dwarf satellite and suggesting a gravitationally lensed image of a background galaxy. In this case VV 191 is now seen to show both dust backlighting and lensing (retroactively earning its place in a project whose overall acronym includes “Lensing”).

This object is shown in Figure 11 after subtracting the model for the foreground elliptical as described in section 2.6. It is so red that only its core was obvious in our HST F606W exposure. Galaxy-galaxy lens arcs often show two images, and indeed a similarly red, slightly resolved object appears on the opposite side of the foreground nucleus from the arc, at projected radius 1.03″. We quantify the colors using aperture photometry (with radius 0.5″) for the brightest part of the arc and for the foreground nucleus, flux integrated within a segment of an annulus for the entire arc, and integrated Gaussian fits for the counterimage. As we used data from an early version of the calibration pipeline while the photometric calibration was still in flux, we confirmed our understanding of the flux scale using the core of the elliptical galaxy VV 191a and the consistent SEDs of luminous elliptical galaxies (Table 1).The possible counterimage is so small that we measure its count rates (WFC3) or signal in mJy (NIRCam) by fitting two-dimensional Gaussian profiles using IRAF imexamine. These results are shown in Table 5, with ratios shown in magnitudes as well for quick interpretation. The listed errors include fluctuations in the residual background after model subtraction, which often exceed those from photon statistics.

We estimate photometric redshifts for the lensed components based on the magnitudes in Table 5, using the EAZY code (Brammer et al., 2008).The arc has $z_{phot}=0.90\pm 0.07$ (60% probability) with secondary solutions$z_{phot}=0.86\pm 0.07$ (26%) and $z_{phot}=1.17\pm 0.08$ (14%). Our initial analysis derived much larger values; the updated NIRCam zero points,reflecting greater sensitivity in the longer-wavelength bands and especially F444W thanpredicted, give a blue spectral slope between F356W and F444W which matches the longside of the generic 1.6$\mu$m peak in galaxy SEDs. The 4000-Å break is nowseen to fall between F606W and F090W bands. The filter selection, driven by the dustanalysis, leaves a large gap between F150W and F356W which drives the large remaininguncertainty in photometric redshifts. The counterimage has colors consistent withthe same redshift, albeit with larger uncertainties. This modest redshiftfits with the small angular sizes of the lensed components, since the derivedredshift range is near the maximum in angular-size distance; at $z=0.9$ theangular scale in the consensus cosmology is already within 10% of its peak value at $z=1.6$.

We reproduced the image configuration, including the counterimage, with a two-component model for the lens:a singular isothermal ellipsoid (SIE) and an external shear. An initial optimization used the image centroids and flux ratio of the arc and counterimage in F444W as constraints. Then with those outputs as initial conditions, we performed a second optimization (using a particle swarm optimization algorithm in both instances) over the F444W band image (after subtraction of the foreground elliptical galaxy, see Figure 11), including a model for the source light.Since there is a $\approx 0.5$ mag color gradient between the core and outer region within the lensed arc, we modeled the source light with a point source and two extended components (disk and bulge) using two Sersic profiles. These profiles are constrained to be concentric during the optimization.The best fit model parameters values obtained after the second optimization are listed in Table 4. The strong lensing modeling was done using the lenstronomy⁵⁵5https://github.com/sibirrer/lenstronomy software package (Birrer & Amara 2018, Birrer et al. 2021).

This model is able to reproduce both the image positions and the flux ratio in the band F444W (see Table 5).The result is shown in Figure 12, along with the lens geometry which requires the source to be near the cusp of the tangential caustic.

Tracing the color of the arc as a function of angle around the lensing nucleus shows a color gradient, with the core 0.44 magnitude redder between F090W and F444W than the surrounding $20^{\circ}$ away. This might reflect a reddened star-forming core or AGN, where the outer regions of the source are where the outer regions of the source and the central redder component of the source are blended in the counterimage. This color gradient further motivates our two-component model for the lensed source.

The model optimization yields a reconstructed source with a bulge half-light radius of $R_{hl}=0.09"$.The physical size depends on the angular-diameter distance, and for the arc best fit photo-z source redshift $R_{hl}(z=0.9)\approx 0.75$ kpc. Similarly, the effective velocity dispersion for the isothermal lens model is $\sigma_{SIE}(z=0.9)\approx 250$ km/s which is very closeto the measured velocity dispersion of $\simeq 270\pm 7km/s$ dispersion from the SDSS for VV 191a, described above, though for a detailed comparison a modest correction for the light profile is required (note that this quantity is not the same as the measured dynamical value Treu & Koopmans 2002) because the measured dispersion in the fixed-diameter fiber of SDSS spectroscopy is inherently intensity weighted.The apparent physical size of the counterimage in the lens plane is $\approx 90$ pc.

We can take advantage of the symmetry of the lens and the low redshift of the lensing galaxy VV 191a ($z=0.0513$ corresponding to an angular diameter distance of only 208Mpc) to obtain a precise mass projected within the Einstein radius of 2.01″. This is model independent, related only to fundamental constants and the distances involved.The poorly-known redshift of the lensed galaxy does not affect this calculation as the lens distance $d_{L}$is so small that the distance ratio $d_{LS}/d_{S}$ is almost unity, for example for the photometricredshift estimates, the ratio is negligibly different for $z=0.8$ and $z=1.2$ with values of $0.922$ and $0.942$ respectively.This results in a projected mass within theEinstein radius of $M(<2.01")\approx 1.1\times 10^{11}M_{\odot}$ and $1\arcsec=1.008kpc$ at the redshift of VV191a.We can compare this to the projected light from our profile fit for VV 191a summed within the same $2.01\arcsec$ radius.

To estimate the mass-to-light ratio we convert the HST F606W (which is close to SDSS $r$) luminosity to the $g$ band using the observation that $(g-r)$ colors of low-z luminous elliptical galaxies are always close to 0.80, and use $(g-B)$ transformations from Jordi et al. 2006 for elliptical galaxies. Applying the Galactic foreground extinction estimate from Schlafly & Finkbeiner (2011) for the F606W data, we obtain $L_{B}(<2.01")=1.45\times 10^{10}L_{\odot,B}$, which yields a central mass-to-light ratio of $M/L_{B}=7.6M_{\odot}/L_{\odot,B}$.This value can be compared to the higher-redshift elliptical galaxy lens mass-to-light ratio in the B-band from Treu & Koopmans (2002) of $M/L_{B}=8\pm 1M_{\odot}/L_{\odot}$, as well as the results from Gerhard et al. (2001) for local E/S0 galaxies, which have sample mean values near 7.2 from both dynamical models and individual metallicities plus Salpeter IMF. Because the lens is comparatively nearby, these values depend only weakly on the distance to the background source. For example, were the source as distant as $z=3$, $M/L_{B}$ would decrease by 8%.

The image configuration is quite similar to that seen in the strongly magnified systems IRAS F10214+4724 (Eisenhardt et al., 1996) and 3C 220.3 (Haas et al., 2014). In exploring mass and lensing models, we note the SDSS DR15 velocity dispersion $272\pm 7$ km s^-1 with the BOSS fiber aperture of diameter 2″.

An additional faint red galaxy appears at right in Figure 11, 5.25″from the foreground nucleus. This object is elongated tangential to the foreground galaxy (best seen in F444W), in a region where the arc model predicts that weak lensing should stretch the image by about a factor 1.6 without generating a counterimage, so this is a second plausibly lensed background system. It is also very red; the F356W and F444W magnitudes are more reliable because the F090W and F150W images are confused with a blue star image $0.18$″away (which is too faint to be a major contaminant in the longer bands). Uncertainty due to deblending star and galaxy fluxes is indicated by colons for these passbands in Table 5. EAZY gives a primary photometric redshift $z=2.06\pm 0.10$ (55% likelihood) with a secondary (34%) peak at $z=1.7\pm 0.1$.

A similar EAZY fit gives photometric redshift estimate$z=0.73\pm 0.18$ for the diffuse component outside the western end of the lensed arc, allowing but not requiring its source to bephysically associated with that of the arc.

Galaxy-galaxy gravitational lensing was a later discovery than lensing by groups and clusters, since the Einstein radii are smaller and the light of the lensed source may be blended with that from the lens galaxy. However, the more detailed work needed to find such lenses has paid off in tracing the mass distributions of (especially early-type) galaxies over a substantial redshift range, with detection techniques ranging from high-redshift AGN spectra in low-redshift galaxies (Huchra et al., 1985), through multiple redshift systems in SDSS spectra (the SLACS survey, Bolton et al. 2004, 2005, 2006) and searches for lens structures in HST images (a recent summary of the many advances in finding such lens systems is given by O’Donnell et al. 2022). The difficulty of recognizing the lensed components in VV191 from even HST data in the optical, together with their prominence in the NIRCam images, suggests that large numbers of galaxy-galaxy lenses will appear (even serendipitously) in JWST galaxy studies. Such samples will improve our understanding of the mass distribution in the lens galaxies over cosmic time, in favorable cases also telling us about the masses of supermassive black holes. In VV 191a, the lack of a detectable third image (not seen either in the data we show here or the HST near-UV data) is consistent with the SMBH masses typically expected at this luminosity, as is the mass model.

6 Summary and Conclusions

Combining HST and JWST imaging, we have used thebacklighting of the face-on spiral galaxy VV 191b by a luminous elliptical to trace the structure and reddening law of dust in its outer disk. The unusually favorable geometry of this system allows us to map the dust attenuation with spatial resolution $\approx 60$ pc, spanning radii 12-21 kpc (much of this range, beyond the stellar spiral arms), and with high-quality mapping from 0.6-1.5$\mu$m.

The distinct dust lanes end sharply at radius 21 kpc (1.7 Petrosian radii), and exhibit not only the usual pattern following the pitch angle of the bright spiral arms but additional lanes cutting at a variety of angles to this.

We fit curves for various values of total-to-selective extinction $R$ to the PSF-matched attenuation values measured pixel-by-pixel in multiple filters from 0.6-1.5µm, to estimate the slope of the effective reddening law. Most regions can be described by $R\approx 2.0$ using a CCM reddening law, or with power-law index $\beta\approx 2.0$, with longer wavelengths showingsmaller $R$ and more positive $\beta$ (less-gray extinction). Limits to the attenuation at longer wavelengths 3.5-4.4$\mu$m are 0.02-0.035 $A_{V}$. Although even sub-percent systematics in modeling the galaxy brightness profiles are important for small attenuation values, we see some evidence that small-attenuation regions have larger $R$ than regions with more attenuation, which could be a sign that clumpy structure affects strong-attenuation areas more strongly. The most striking feature of these measurements is how strongly reddening the behavior is, even in the likely presence of mixing effects combining region of different attenuation in each resolution element, which we in fact detect by measuring our transmission maps after various degrees of averaging. Broadly, this might suggest a population strongly dominated by small grains, such as has been suggested in the outer disk of the Milky Way by Zasowski et al. (2009) based on stellar colors in the near-infrared. It may be important in shaping the grain population the the dust features we sample in VV 191b are as far as 7.5 kpc from regions of ongoing star formation as traced in the UV or by hot dust.

This investigation of dust far from star-forming regions may be relevant to theextinction and reddening in cosmologically-interesting SN Ia, which may explode anywhere ina galaxy (in contrast to the concentration of core-collapse supernova near theirformation regions). Supernovae probe extinction much more precisely than ourspatially-averaged measurements, approaching the point-source probes provided by ordinary stars.Thorp & Mandel (2022) use a hierarchical Bayesian approach to address the possibility ofunphysical ranges of $R$ retrieved in supernova samples due to observational error,finding $R=2.2-2.6$ for type Ia supernova in the Carnegie Supernova project. Theycontrast this with $R=1.9$ from fits to individual SN from Johansson et al. (2021). Allthe values are significantly lower than the frequently-used Milky Way (local)mean $R\approx 3.5$, in the same sense as our results for the outer disk of VV 191b.

The high quality of the NIRCam data from 0.9 to 1.5µm means that the NIR is now the most sensitive regime to trace backlit dust in galaxies, even with its greater attenuation at shorter wavelengths. These observations detect dust lanes farther out than the spiral pattern is seen in starlight even with the deep IR images. F150W is surprisingly valuable for its leverage in fitting the slope of the effective reddening law (F070W+F150W would be a fast survey pair with NIRCam, with longer wavelengths observed at no extra cost in observing time). The Galaxy Zoo overlapping-galaxy catalog alone (Keel et al., 2013) includes a number of deeper overlap pairs within the distance $z=0.05$ of this one, which would be at least as well resolved and allow construction of composite attenuation profiles which could be valuable in constructing priors for supernova photometry into the near-IR, and improving our understanding of the energy budget in disk galaxies. Complementing such efforts, these optical/NIR attenuation maps could provide a comparison set for data tracing the PAH emission bands farther into the infrared, and far-IR or submillimeter tracers of the overall dust emission, to refine our understanding of how the components of the grain population change in various environments.

The JWST images further revealed a galaxy-galaxy gravitational lensing instance through the elliptical galaxy VV 191a, with a background system at $z\approx 0.9$ appearing as a $90^{\circ}$ arc and counterimage. An additional distant background galaxy appears weakly lensed with modeled axial ratio 1.6. All these can be reproduced simultaneously with a lens modeling including a singular isothermal sphere and shear (likely associated with the adjacent spiral galaxy). The small redshift of the lens allows a determination of the mass/light ratio within its Einstein radius almost independent of the source redshift, $M/L_{B}=7.6$ in solar units.

VV 191 vividly demonstrates a facet of backlit-galaxy dust analysis which becomes important only at significant redshifts: large redshift differences increase the role of gravitational lensing in distorting the background sources, so using galaxy symmetry in modeling is more effective for small redshift offsets between the two galaxies.

We dedicate this paper to Prof. Hendrik van de Hulst, who during his long and very productive career at Sterrewacht Leiden had a mission to understand cosmic dust.This project is based on observations made with the NASA/ESA Hubble Space Telescope and the NASA/ESA/CSA James Webb Space Telescope, obtained from the Mikulski Archive for Space Telescopes, which is a collaboration between the Space Telescope Science Institute (STScI/NASA), the Space Telescope European Coordinating Facility(ST-ECF/ESA), and the Canadian Astronomy Data Centre (CADC/NRC/CSA). The IUE spectra were retrieved from theINES service of LAEFF in Spain.RAJ, RAW, and SHC acknowledge support from NASA JWSTInterdisciplinaryScientist grants NAG5-12460, NNX14AN10G and 80NSSC18K0200 from GSFC.CNAW acknowledges funding from the JWST/NIRCam contract NASS-0215 to theUniversity of Arizona. We acknowledge support from the ERC Advanced Investigator Grant EPOCHS (788113), as well as a studentship from STFC. AZ acknowledges support by Grant No. 2020750 from the United States-Israel Binational Science Foundation (BSF) and Grant No. 2109066 from the United States National Science Foundation (NSF), and by the Ministry of Science & Technology, Israel.This research was supported by the Australian Research Council Centre of Excellencefor All Sky Astrophysics in 3 Dimensions (ASTRO 3D), through project number CE170100013.We thank our Program Coordinator, Tony Roman, for his expert help scheduling this program.We thank the Galaxy Zoo volunteers for finding so many of these backlit-galaxy systems (particularly Galaxy Zoo users kodemunkey andGoniners for first pointing out this pair).The data presented in this paper were obtained from the Mikulski Archive for SpaceTelescopes (MAST) at the Space Telescope Science Institute. The specific observations analyzed can beaccessed via https://doi.org/10.17909/78wg-t688 (catalog 10.17909/78wg-t688).

References

• Allaert et al. (2015) Allaert, F., Gentile, G., Baes, M., et al.2015, A&A, 582, A18. doi:10.1051/0004-6361/201526667
• Astropy Collaboration (2013) Astropy Collaboration,Robitaille, T.P., Tollerud, E.J., Greenfield, P., etal.2013,A&A 558, A33
• Astropy Collaboration (2018) Astropy Collaboration,Price-Whelan, A.M., Sipőcz, B.M., Günther, H.M., etal.2018,AJ 156, 123
• Beeston etal. (2018)Beeston, R.A. etal. 2018, MNRAS, 479, 1077
• Bellstedt et al. (2020) Bellstedt, S., Robotham, A.S.G., Driver, S.P., et al.2020, MNRAS, 498, 5581. doi:10.1093/mnras/staa2620
• Bellstedt et al. (2021) Bellstedt, S., Robotham, A.S.G., Driver, S.P., et al.2021, MNRAS, 503, 3309. doi:10.1093/mnras/stab550
• Bertin & Arnouts (1996) Bertin, E., & Arnouts, S. 1996,A&AS 117, 393
• Birrer & Amara (2018)Birrer, S. & Amara, A. Lenstronomy: multi-purpose gravitational lens modelling software package. (2018)
• Birrer et al. (2021) Birrer, S., Shajib, A., Gilman, D., et al.2021, The Journal of Open Source Software, 6, 3283. doi:10.21105/joss.03283
• Bolton et al. (2004) Bolton, A.S., Burles, S., Schlegel, D.J., et al.2004, AJ, 127, 1860. doi:10.1086/382714
• Bolton et al. (2005) Bolton, A.S., Burles, S., Koopmans, L.V.E., et al.2005, ApJ, 624, L21. doi:10.1086/430440
• Bolton et al. (2006) Bolton, A.S., Burles, S., Koopmans, L.V.E., et al.2006, ApJ, 638, 703. doi:10.1086/498884
• Bourne et al. (2012) Bourne, N., Maddox, S.J., Dunne, L., et al.2012, MNRAS, 421, 3027. doi:10.1111/j.1365-2966.2012.20528.x
• Bradford (2016) Bradford, S. (2016) M.S. Thesis, Dust mapping of spiral galaxy disks from theHubble STARSMOG survey. [electronic resource]. [University of Alabama Libraries].https://ir.ua.edu/handle/123456789/2608 (Accessed: 5 July 2022).
• Brammer et al. (2008) Brammer, G.B., van Dokkum, P.G., & Coppi, P.2008, ApJ, 686, 1503. doi:10.1086/591786
• Brown et al. (2014) Brown, M.J.I., Moustakas, J., Smith, J.-D.T., et al.2014, ApJS, 212, 18. doi:10.1088/0067-0049/212/2/18
• Burstein et al. (1988) Burstein, D., Bertola, F., Buson, L.M., et al.1988, ApJ, 328, 440. doi:10.1086/166304
• Calzetti et al. (1994) Calzetti, D., Kinney, A.L., & Storchi-Bergmann, T.1994, ApJ, 429, 582. doi:10.1086/174346
• Calzetti et al. (2000) Calzetti, D., Armus, L., Bohlin, R.C., et al.2000, ApJ, 533, 682. doi:10.1086/308692
• Calzetti (2001) Calzetti, D.2001, PASP, 113, 1449. doi:10.1086/324269
• Cardelli et al. (1989) Cardelli, J.A., Clayton, G.C., & Mathis, J.S.1989, ApJ, 345, 245. doi:10.1086/167900
• Chevallard et al. (2013) Chevallard, J., Charlot, S., Wandelt, B., et al.2013, MNRAS, 432, 2061. doi:10.1093/mnras/stt523
• Ciesla etal. (2014)Ciesla, L. etal. 2014, A&A, 565, A128
• Ciesla etal. (2012)—. 2012, A&A, 543, A161
• Clark etal. (2015)Clark, C.J.R. etal. 2015, MNRAS, 452, 397
• Clayton et al. (2015) Clayton, G.C., Gordon, K.D., Bianchi, L.C., et al.2015, ApJ, 815, 14. doi:10.1088/0004-637X/815/1/14
• Cortese etal. (2012)Cortese, L. etal. 2012, A&A, 540, A52
• Cortese etal. (2014)—. 2014, MNRAS, 440, 942
• De Geyter et al. (2014) De Geyter, G., Baes, M., Camps, P., et al.2014, MNRAS, 441, 869. doi:10.1093/mnras/stu612
• De Geyter et al. (2015) De Geyter, G., Baes, M., De Looze, I., et al.2015, MNRAS, 451, 1728. doi:10.1093/mnras/stv1104
• De Looze et al. (2012) De Looze, I., Baes, M., Bendo, G.J., et al.2012, MNRAS, 427, 2797. doi:10.1111/j.1365-2966.2012.22045.x
• Domingue et al. (1999) Domingue, D.L., Keel, W.C., Ryder, S.D., et al.1999, AJ, 118, 1542. doi:10.1086/301062
• Domingue et al. (2000) Domingue, D.L., Keel, W.C., & White, R.E.2000, ApJ, 545, 171. doi:10.1086/317767
• Driver et al. (2018) Driver, S.P., Andrews, S.K., da Cunha, E., et al.2018, MNRAS, 475, 2891. doi:10.1093/mnras/stx2728
• Eisenhardt et al. (1996) Eisenhardt, P.R., Armus, L., Hogg, D.W., et al.1996, ApJ, 461, 72. doi:10.1086/177038
• Elmegreen et al. (2001) Elmegreen, D.M., Kaufman, M., Elmegreen, B.G., et al.2001, AJ, 121, 182. doi:10.1086/318036
• François et al. (2019) François, P., Morelli, L., Pizzella, A., et al.2019, A&A, 621, A60.
• Fruchter etal. (2010) Fruchter, A.S., Hack, W.,Dencheva, N., etal.2010, in the Proceedings of the “2010 STScICalibration Workshop”, eds. S.Deustua & C.Oliveira (Baltimore:STScI), pp.376–381.
• Gentile et al. (2015) Gentile, G., Tydtgat, C., Baes, M., et al.2015, A&A, 576, A57. doi:10.1051/0004-6361/201425279
• Gerhard et al. (2001) Gerhard, O., Kronawitter, A., Saglia, R.P., et al.2001, AJ, 121, 1936. doi:10.1086/319940
• Haas et al. (2014)Haas, M., Leipski, C., Barthel, P., et al. 2014, ApJ, 790, 46
• Hoffmanm etal. (2021) Hoffmann, S.L., Mack, J.,Avila, R.J., etal.2021, “The DrizzlePac Handbook (v2.0)”(Baltimore: STScI)
• Hogg (2022) Hogg, D.W.2022, arXiv:2206.00989
• Holwerda et al. (2007) Holwerda, B.W., Meyer, M., Regan, M., et al.2007, AJ, 134, 1655. doi:10.1086/521824
• Holwerda et al. (2009) Holwerda, B.W., Keel, W.C., Williams, B., et al.2009, AJ, 137, 3000. doi:10.1088/0004-6256/137/2/3000
• Holwerda et al. (2012) Holwerda, B.W., Bianchi, S., Böker, T., et al.2012, A&A, 541, L5. doi:10.1051/0004-6361/201118615
• Holwerda et al. (2013) Holwerda, B.W., Böker, T., Dalcanton, J.J., et al.2013, MNRAS, 433, 47. doi:10.1093/mnras/stt669
• Holwerda & Keel (2013) Holwerda, B.W. & Keel, W.C.2013, A&A, 556, A42. doi:10.1051/0004-6361/201321455
• Huchra et al. (1985) Huchra, J., Gorenstein, M., Kent, S., et al.1985, AJ, 90, 691. doi:10.1086/113777
• Hughes et al. (2014) Hughes, T.M., Baes, M., Fritz, J., et al.2014, A&A, 565, A4. doi:10.1051/0004-6361/201323245
• Hunter (2007) Hunter, J.D. 2007, MCS&E 9, 3
• Jedrzejewski (1987) Jedrzejewski, R.I.1987, MNRAS, 226, 747. doi:10.1093/mnras/226.4.747
• Johansson et al. (2021) Johansson, J., Cenko, S.B., Fox, O.D., et al.2021, ApJ, 923, 237. doi:10.3847/1538-4357/ac2f9e
• Jordi et al. (2006) Jordi, K., Grebel, E.K., & Ammon, K.2006, A&A, 460, 339. doi:10.1051/0004-6361:20066082
• Keel (1983) Keel, W.C.1983, AJ, 88, 1579. doi:10.1086/113448
• Keel et al. (2013) Keel, W.C., Manning, A.M., Holwerda, B.W., et al.2013, PASP, 125, 2. doi:10.1086/669233
• Keel et al. (2014) Keel, W.C., Manning, A.M., Holwerda, B.W., et al.2014, AJ, 147, 44. doi:10.1088/0004-6256/147/2/44
• Keel & White (2001a) Keel, W.C. & White, R.E.2001, AJ, 121, 1442. doi:10.1086/319386
• Keel & White (2001b) Keel, W.C. & White, R.E.2001, AJ, 122, 1369. doi:10.1086/322117
• Koekemoer etal. (2003) Koekemoer, A.M., Fruchter, A.S.,Hook, R.N., & Hack, W. 2003, in: ‘The 2002 HST Calibration Workshop:Hubble after the Installation of the ACS and the NICMOS CoolingSystem’, eds. S.Arribas, A.Koekemoer, & B.Whitmore (Baltimore:STScI), p. 337
• Mosenkov et al. (2016) Mosenkov, A.V., Allaert, F., Baes, M., et al.2016, A&A, 592, A71. doi:10.1051/0004-6361/201628676
• Mosenkov et al. (2018) Mosenkov, A.V., Allaert, F., Baes, M., et al.2018, A&A, 616, A120. doi:10.1051/0004-6361/201832899
• O’Donnell et al. (2022) O’Donnell, J.H., Wilkinson, R.D., Diehl, H.T., et al.2022, ApJS, 259, 27. doi:10.3847/1538-4365/ac470b
• Oke et al. (1981) Oke, J.B., Bertola, F., & Capaccioli, M.1981, ApJ, 243, 453. doi:10.1086/158612
• Norgaard-Nielsen & Kjaergaard (1981) Norgaard-Nielsen, H.U. & Kjaergaard, P.1981, A&A, 93, 290
• Patat et al. (2015) Patat, F., Taubenberger, S., Cox, N.L.J., et al.2015, A&A, 577, A53. doi:10.1051/0004-6361/201424507
• Rauscher et al. (2012) Rauscher, B.J., Arendt, R.G., Fixsen, D.J., et al.2012, Proc.SPIE, 8453, 84531F. doi:10.1117/12.926089
• Rigby et al. (2022) Rigby, J., Perrin, M., McElwain, M., et al.2022, arXiv:2207.05632
• Salim et al. (2018) Salim, S., Boquien, M., & Lee, J.C.2018, ApJ, 859, 11. doi:10.3847/1538-4357/aabf3c
• Schlafly & Finkbeiner (2011) Schlafly, E.F. & Finkbeiner, D.P.2011, ApJ, 737, 103. doi:10.1088/0004-637X/737/2/103
• Smith etal. (2012)Smith, D.J.B. etal. 2012, MNRAS, 427, 703
• Thorne et al. (2021) Thorne, J.E., Robotham, A.S.G., Davies, L.J.M., et al.2021, MNRAS, 505, 540. doi:10.1093/mnras/stab1294
• Thorp & Mandel (2022) Thorp, S. & Mandel, K.S.2022, MNRAS, 517, 2360. doi:10.1093/mnras/stac2714
• Treu & Koopmans (2002) Treu, T. & Koopmans, L.V.E.2002, ApJ, 575, 87. doi:10.1086/341216
• Verstappen et al. (2013) Verstappen, J., Fritz, J., Baes, M., et al.2013, A&A, 556, A54. doi:10.1051/0004-6361/201220733
• Vorontsov-Velyaminov (1959) Vorontsov-Velyaminov, B.A.1959,Atlas i Katalog Vzaimodeistvuyuschikh Galaktik I(Atlas and Catalog of Interacting Galaxies I), Shternberg Institute, Moscow State University (1959)
• Wang & Chen (2019) Wang, S. & Chen, X.2019, ApJ, 877, 116. doi:10.3847/1538-4357/ab1c61
• Wang & Wheeler (2008) Wang, L. & Wheeler, J.C.2008, ARA&A, 46, 433. doi:10.1146/annurev.astro.46.060407.145139
• White & Keel (1992) White, R.E. & Keel, W.C.1992, Nature, 359, 129. doi:10.1038/359129a0
• White et al. (2000) White, R.E., Keel, W.C., & Conselice, C.J.2000, ApJ, 542, 761. doi:10.1086/317011
• Windhorst etal. (2023) Windhorst, R.A., Cohen, S.H., Jansen, R.A., et al.2023, AJ, 165, 13. doi:10.3847/1538-3881/aca163
• Wright (2006) Wright, E.L.2006, PASP, 118, 1711. doi:10.1086/510102
• Zasowski et al. (2009) Zasowski, G., Majewski, S.R., Indebetouw, R., et al.2009, ApJ, 707, 510. doi:10.1088/0004-637X/707/1/510
• Zhang & Yuan (2022) Zhang, R. & Yuan, R. 2022, ApJSupp (in press), arXiv:2206.06432
• Zou et al. (2022) Zou, F., Brandt, W.N., Chen, C.-T., et al.2022, arXiv:2206.06432