Discovery of a Damped Lyα Absorber Originating in a Spectacular Interacting Dwarf Galaxy Pair at z = 0.026

We refer the reader to Chen et al. (2020) for a description of the absorption-line observations of J2339–5523 from the Cosmic Origins Spectrograph (COS) on HST (λ = 1100–1800 Å, spectral resolution FWHM ≈ 20 km s⁻¹) and the Magellan Inamori Kyocera Echelle (MIKE) spectrograph on Magellan-Clay (λ = 3200–9200 Å, FWHM ≈ 8 km s⁻¹). We use Voigt profile fitting to determine the velocity (v), Doppler parameter (b), and column density (N) of the atomic and ionic transitions associated with the DLA using a grid search as described in Section 3 of Boettcher et al. (2021). The COS wavelength calibration is sufficiently uncertain that we allow small variations within δ v =±5kms⁻¹ for low and intermediate ions in COS with respect to Ca ii λ λ3934, 3969 in MIKE. We construct a marginalized probability density function for each parameter by calculating a likelihood of the form ${\mathscr{L}}\propto {{\rm{e}}}^{-{\chi }^{2}/2}$ at every point in parameter space, normalizing the total likelihood to unity, and integrating over the marginalized parameters. We adopt as the best-fit model the parameters corresponding to a value of 50% from the cumulative distribution function. The reported uncertainties correspond to the 68% confidence interval, and the upper limits are reported as the 95% one-sided confidence interval. We additionally perform a curve-of-growth (COG) analysis for the low and intermediate ions to corroborate the characteristic Doppler parameter determined by the Voigt profile fitting.

We observed the field centered at J2339–5523 using the MeerKAT-64 array on 2020 June 14 and 22. Of the 64 antennas, 59 and 60 antennas participated in the first and second observing run, respectively. We used the 32K mode of the SKA Reconfigurable Application Board correlator to split the total bandwidth of 856 MHz centered at 1283.9869 MHz into 32,768 frequency channels. The resultant frequency resolution is 26.123 kHz, or 5.7 km s⁻¹, at the redshifted H i 21 cm line frequency of the DLA. Objects PKS 1939–638 and PKS 0408–658 were observed for flux density, delay, and bandpass calibrations, and the compact radio source J2329–4730 was observed for complex gain calibration. The total on-source time on J2339–5523 was 112 minutes.

The MeerKAT data were processed using the Automated Radio Telescope Imaging Pipeline; we refer the reader to Gupta et al. (2021) for details. Here we focus on the Stokes I radio continuum and spectral line properties near the H i 21 cm line frequency corresponding to z_DLA (∼1384.4 MHz). The spatial resolution represented by the synthesized beam of the continuum image and image cube obtained using robust = 0 weighting is 84 × 66 (position angle = 138), with a spatial pixel size of 20. The cube has been deconvolved, or cleaned, using the CASA task tclean down to five times the single channel rms of 0.45 mJy beam⁻¹. The radio emission associated with J2339–5523 is compact with a deconvolved source size of <06.

We used the H i Source Finding Application (SOFIA v2.0; Serra et al. 2015; Westmeier et al. 2021) for the H i 21 cm analysis of the spectral line cube. We set SOFIA to subtract residual continuum subtraction errors from the image cube by fitting a polynomial of order 1. We used the smooth+clip (S+C) algorithm with a combination of spatial kernels of zero, 3, 6, and 9 pixels and spectral kernels of zero, 3, 7, and 15 channels to detect voxels containing H i 21 cm emission in the image cube. We set the threshold of the S+C finder to 3.5σ and used the reliability filter to reject unreliable detections. The output three-dimensional mask was then used to further clean the image cube down to the single channel rms. This deep-cleaned cube was then passed through SOFIA with the abovementioned setup to generate the final H i moment maps and the integrated spectrum. For the moment 1 and 2 maps, we used only those pixels where the signal was detected at 1.5 times the local rms. In the moment zero map, the mean value of pixels with a signal-to-noise ratio (S/N) in the range 2.5σ–3.5σ corresponds to an H i column density of N(H i) = 3.2 × 10²⁰ cm⁻². This represents the average 3σ column density sensitivity across the map (see de Blok et al. 2018 for details of this sensitivity estimate).

We observed the field of J2339–5523 on 2020 November 22, December 5, and December 9 using VLT-MUSE. The total exposure time on source was 9390 s. We reduced the data using the standard ESO MUSE pipeline (Weilbacher et al. 2014) and the custom software CubExtractor from S. Cantalupo (Cantalupo et al. 2019). We subtracted the quasar light using a high-resolution spectral differential imaging method (e.g., Haffert et al. 2019). The wavelength coverage is λ = 4700–9350 Å, and the spectral resolution at the blue (red) end is R ≈ 1610 (R ≈ 3570). This corresponds to an FWHM of ≈185 km s⁻¹ (85 km s⁻¹), with approximately two wavelength bins sampling the spectral resolution element. Multiple pointings mapped a region of $1\buildrel{\,\prime}\over{.} 3\times 1\buildrel{\,\prime}\over{.} 4$ around the QSO. We spatially smoothed the MUSE data cube with a boxcar kernel of 3 × 3 pixels, where the kernel size was chosen to approximate the seeing disk at the time of the observations (≈06, or 0.3 kpc at the redshift of the DLA).

We construct continuum-subtracted narrowband images at the wavelength of all emission lines at the redshift of the DLA. We subtract the continuum by fitting a linear function locally around the masked lines. In parts of both dwarf galaxies, Balmer absorption from the stellar continuum modestly impacts the observed Hα and Hβ emission. To correct for this, we use the stellar population synthesis code bagpipes ²³ (Carnall et al. 2018) to fit a tau model to the stellar continuum of a bright, star-forming region in each galaxy assuming a spatially uniform star formation history; the equivalent widths of the Balmer lines in the best-fit models are in the range 2.5–3.0 Å. We then scale the models to the continuum level of any spaxel for which the continuum light is detected at ≥3σ at the wavelength of Hα or Hβ. To allow for a possible velocity offset between the stars and the gas, we adopt a broad window of ±500 km s⁻¹ (±180 km s⁻¹) when constructing the narrowband images of the Balmer lines for spaxels with (without) this correction applied. We derive an extinction correction using the extinction law of Cardelli et al. (1989) from the reddening, E(B − V), calculated from the Balmer decrement following Calzetti et al. (1994) for every spaxel for which Hβ and Hα are detected at the ≥3σ level.

Finally, we fit the six strongest emission lines (Hα, Hβ, [O iii] λ λ4960, 5008, and [S ii] λ λ6718, 6732) with a single Gaussian profile using the IDL software KUBEVIZ. ²⁴ This yields a velocity centroid and velocity dispersion, corrected for the instrumental resolution, for the ionized gas in each spaxel with statistically significant emission.

In Table 1, we list the best-fit column densities and Doppler parameters for all atomic and ionic species associated with the DLA, and we show a representative sample of the absorption lines and their best-fit models in Figure 1. The H i column density determined from the Lyα line is log[N(H i)/cm⁻²] =20.60 ± 0.05. The uncertainty is dominated by the continuum fitting, as determined by repeated measurements of N(H i) from independent, manual fits to the local continuum prior to the Lyα fitting.

Zoom In Zoom Out Reset image size

Table 1.  z = 0.026 DLA Properties

Column Densities for Limits and Ions Fit with b as a Free Parameter a
Ion	b (km s−1)	log[N/cm−2]	Comments
H i	⋯	20.60 ± 0.05	b unconstrained due to Lyα damping wings; N independent of b
C i	⋯	<13.4	b fixed to 10 km s−1 based on Ca ii
C ii*	⋯	<13.3	b fixed to 10 km s−1 based on Ca ii
C iv	${23}_{-8}^{+12}$	13.45 ± 0.08
N i	12 ± 2	${14.46}_{-0.04}^{+0.05}$
N v	⋯	<12.9	b fixed to 25 km s−1 based on C iv; comparable N found for b ≲ 100 km s−1
Si iv	⋯	<13.0	b fixed to 25 km s−1 based on C iv; both 1393 and 1402 Å transitions contaminated
P ii	⋯	${12.9}_{-0.2}^{+0.1}$	Probability density function flat in b for 10 km s−1 ≤ b ≤ 20 km s−1
S ii	${10}_{-1}^{+2}$	${14.77}_{-0.05}^{+0.07}$
S iii	⋯	14.4 ± 0.1	S iii λ1190 blended with Si ii λ1190; S iii b value not well constrained
Ca ii	${10.3}_{-0.7}^{+0.8}$	11.9 ± 0.02
Fe ii	${14}_{-3}^{+4}$	14.55 ± 0.05
Column Densities for b = 10, 20 km s−1 b
	log[N/cm−2]	log[N/cm−2]
	(b = 10 km s−1)	(b = 20 km s−1)
C ii	17.6	15.1	C ii λ1334 is saturated
N ii	14.7	14.2	Only low-S/N N ii λ1083 transition available
O i	17.7	15.5	O i λ1302 is saturated
Al ii	14.9	13.3	Al ii λ1670 is saturated
Si ii	15.9	14.5	Five saturated lines available
Si iii	15.6	13.8	Si iii λ1206 is saturated

Notes.

^a We impose the prior 10 km s⁻¹ ≤ b ≤ 20 km s⁻¹ in the Voigt profile fitting of neutral, singly, and doubly ionized species based on the COG analysis. ^b We report the best-fit column density corresponding to b = 10 and 20 km s⁻¹ for ions with only saturated and/or low-S/N transitions available; the b values are chosen based on the COG analysis.

Download table as:  ASCIITypeset image

The Ca ii λ λ3934, 3969 lines are the only associated absorption features detected in the MIKE data and indicate a single velocity component at z = 0.02604, which we adopt as the velocity zero-point in the following analysis. The Ca ii lines have a Doppler parameter of b ≈ 10 km s⁻¹, which we corroborate using a COG analysis for low and intermediate ions in COS that span a wide range of oscillator strengths. The best constraints come from six Fe ii and six N i transitions; these measurements suggest that 10 km s⁻¹ ≤ b ≤ 20 km s⁻¹, and we impose this prior in the Voigt profile fitting of neutral, singly, and doubly ionized species. Only saturated or low-S/N lines are available for O i, C ii, N ii, Al ii, Si ii, and Si iii, and we report a range of column densities corresponding to 10 km s⁻¹ ≤ b ≤ 20 km s⁻¹ for these species. There is some indication of moderate ionization state gas associated with the DLA from the detection of C iv (the Si iv transitions are contaminated by interlopers). The C iv Doppler parameter exceeds that of the low ions by a factor of 2, hinting that it may arise in a boundary layer between the neutral absorber and a hot ambient medium (e.g., Fox et al. 2005).

In panel (c) of Figure 1, we show the chemical abundances, [X/H], for a range of elements ordered from most volatile to most refractory. Our best leverage on the metallicity of the DLA comes from S, which is only mildly depleted in a dusty medium (e.g., De Cia et al. 2016) and suggests Z ≈ 0.1Z_⊙. In contrast, we find [Fe/H] = −1.60 ± 0.07 and [Ca/H] =−3.00 ± 0.05, suggesting dust depletion of the most refractory elements. The decrement in [Fe/H] compared to [S/H] is consistent with the Galactic halo depletion pattern of Savage & Sembach (1996). Due to the low second ionization energy of Ca (11.9 eV), it is possible that we underestimate [Ca/H] by assuming N(Ca) = N(Ca ii). However, the presence of dust depletion is clear from the underabundance of Fe alone. Applying the method of De Cia et al. (2021), we use the well-constrained S and Fe abundances to determine a total, dust-corrected metallicity of [X/H] = −0.8 ± 0.1 using the depletion patterns of De Cia et al. (2016). This is consistent with being drawn from the metallicity distribution of known DLAs at low redshift (Lehner et al. 2019).

We observe two dwarf galaxies at the redshift of the DLA and employ a deep galaxy redshift survey to characterize their environment. Chen et al. (2020) described the redshift survey conducted in all CUBS fields with LDSS-3C and IMACS on Magellan and VLT-MUSE. The spectroscopic component of this survey targets galaxies fainter than 18th magnitude within $\theta \lesssim 10^{\prime}$, so we primarily use photometric redshifts from the Dark Energy Survey (DES; Abbott et al. 2018). At the redshift of the DLA, a blue galaxy with L = 0.01L_* has an apparent r-band magnitude of m_r = 19.2 assuming M_r,* = −21.1 (Cool et al. 2012), where M_r,* is the characteristic rest-frame absolute r-band magnitude at the break of the luminosity function. We estimate the uncertainty on these photometric redshifts using 16 galaxies with m_r < 19 with robust spectroscopic redshifts within $\theta =11^{\prime}$. For galaxies at z < 0.3, we find dz/(1 + z) <0.02. We do not find any galaxies with ∣z_phot − z_DLA∣ < 0.02. There are two galaxies with ∣z_phot − z_DLA∣ = 0.02–0.04. The first is at z_phot = 0.055 and d = 95 kpc; this system has m_r = 18.4. The second is at z_phot = 0.064 and d = 255 kpc and has m_r = 17.4. Thus, aside from the two dwarf galaxies, there are no galaxies more luminous than ≈0.05L_* within d ≈ 350 kpc.

As shown in Figure 2 and Table 2, the two dwarf galaxies are separated by 33 kpc and within a projected velocity of Δv_g ≲ 10 km s⁻¹ of the DLA. We refer to these galaxies as G1 and G2, where G1 is the galaxy at a smaller impact parameter, d, with respect to the QSO. At d = 6.2 kpc, G1 is projected in close enough proximity to the QSO that the 3σ optical continuum light overlaps with the wing of the QSO point-spread function (PSF) on the southwest side of the galaxy. Galaxy G2 is found at the edge of the MUSE footprint at d = 32.5 kpc, and ≈15% of the area within the 3σ r-band contour is found outside of the IFU coverage.

Zoom In Zoom Out Reset image size

We estimate the stellar mass of both galaxies to be log (M_star/M_⊙) = 8.5 ± 0.2 from extinction-corrected, r-band photometry from DES using the M_r − M_star relation derived from NASA Sloan Atlas galaxies by Liang & Chen (2014). We find comparable mass estimates from stellar population synthesis models using bagpipes. We estimate the halo mass of the galaxies to be log (M_h/M_⊙) = 10.8 ± 0.2 from the stellar mass–halo mass relation for field dwarfs determined by Read et al. (2017); this is consistent with the models of Behroozi et al. (2019). Following Maller & Bullock (2004), this yields a virial radius of ${R}_{\mathrm{vir}}\approx {100}_{-10}^{+20}$ kpc. Here R_vir is defined based on the overdensity condition of Bryan & Norman (1998) at the redshift of the dwarfs. The projected separation of the dwarfs is thus approximately one-third of their virial radii. As shown in Figure 2, the optical continuum morphology of the faint outskirts of both dwarfs shows disturbances; these irregularities are most prominent on the sides of the galaxies closest to their companion, including a faint stellar stream protruding from G1 that points toward the QSO. The extinction-corrected Hα luminosity of G1, L(Hα) ≈ 8 × 10³⁹ erg s⁻¹, suggests a star formation rate (SFR) of ≈0.04 M_⊙ yr⁻¹ following the relation of Calzetti (2008). About two-thirds of the star formation is found in the bright, star-forming knot labeled G1R1 in Figure 2. Galaxy G2 is ≈40% dimmer than G1 in Hα and thus has SFR(Hα) ≈0.03 M_⊙ yr⁻¹ (a small fraction of the Hα flux may be unaccounted for due to the proximity of the galaxy to the edge of the detector).

We detect spatially extended H i 21 cm emission that reveals a neutral gas envelope surrounding G1 and G2 on a scale of >40 kpc, including an apparent tidal bridge between them. We display the H i column density contours in Figure 2. The bridge has a width of ∼20'', or 10 kpc, at the level of 5 × 10¹⁹ cm⁻² and is thus resolved by MeerKAT's synthesized beam (FWHM = 4 kpc × 3 kpc). The total integrated line flux of the system is 0.44 ± 0.06 Jy km s⁻¹, which corresponds to a total H i gas mass of M(H i) ≈1.3 × 10⁹ M_⊙. If we bisect the gas distribution at the midpoint between the galaxies, the two dwarfs are associated with nearly the same H i mass.

As shown in Figure 2, the QSO sight line intersects the outer gaseous envelope of G1. In the vicinity of the QSO, H i emission from several voxels in the image cube contributes to the H i 21 cm column density. The emission is detected in individual spectral channels at the 2σ–3σ level. Within the MeerKAT synthesized beam, the characteristic column density at the location of the peak flux density of the QSO is log[N(H i)/cm⁻²] ≈20.6. Despite the possible impact of beam dilution in a clumpy medium, this is remarkably consistent with the column density determined from the HST-COS absorption-line spectroscopy.

The QSO J2339–5523 is unresolved (size <300 pc) in the MeerKAT image. We visually examined the spectral line image cube and confirmed that no negative features representing H i 21 cm absorption at the >2σ level (∫τ d v = 0.028 km s⁻¹) are present toward the QSO. The strength of the H i 21 cm absorption depends on the H i column density, as well as the spin temperature, T_s , and the covering factor of the absorbing gas, f_c . If the radio-emitting region of the QSO is more spatially extended than the absorbing gas, then f_c < 1. The H i column density toward the UV-emitting region of the QSO is well constrained through the HST-COS observations of the DLA. If the UV- and radio-emitting regions of the QSO are spatially aligned, we can use the measured N(H i) from COS to constrain the ratio of the spin temperature to the covering factor. Using a 5σ upper limit on H i 21 cm absorption in the unsmoothed MeerKAT spectrum with a spectral rms of 0.44 mJy beam⁻¹ channel⁻¹, we estimate a lower limit of T_s /f_c > 1880 K assuming an FWHM of 17 km s⁻¹ (b = 10 km s⁻¹) for the H i absorption line based on the metal lines detected in the COS and MIKE spectra.

In the top panels of Figure 3, we show the line-of-sight velocity and velocity dispersion of the neutral gas traced by H i 21 cm emission. The velocity shear across the neutral gas distribution is ≈±20 km s⁻¹. The characteristic velocity dispersion within the bodies of G1 and G2 is 10–15 km s⁻¹, and this value is a factor of 2 higher in parts of the extended medium, suggesting possible dynamical disturbance. Due to the relatively small number of MeerKAT beams sampling the gas distribution, we cannot draw strong conclusions about whether the bridge gas arises from G1 or G2. It is also not clear if the gas distribution in either galaxy is disklike; there is no definitive evidence for organized rotation, and it is common for the gas content in galaxies of this mass to have significant dispersion support (e.g., El-Badry et al. 2018).

Zoom In Zoom Out Reset image size

We analyze the warm ionized medium (WIM) in both dwarf galaxies with the goal of comparing the observable properties of the interstellar medium (ISM) with those of the damped absorber to better characterize the origin of the neutral gas envelope. Among the emission lines covered by MUSE are Hα, Hβ, [O iii] λ λ4960, 5008, [N ii] λ λ6549, 6585, [S ii] λ λ6718, 6732, and [S iii] λ6313, 9071, all of which are detected in at least one star-forming region in one or both galaxies.

3.4.1. WIM Mass and Morphology

As shown in Figure 2, there is no evidence of spatially extended optical line emission on scales larger than the stellar components of G1 and G2. The 3σ Hα surface brightness detection threshold in the vicinity of G1 (G2) is 2.6 × 10⁻¹⁸ erg s⁻¹ cm⁻² arcsec⁻² (3.8 × 10⁻¹⁸ erg s⁻¹ cm⁻² arcsec⁻²). Notably, the WIM does not overlap with the quasar sight line at the detection threshold of the data. The line-emitting feature in closest projected proximity to the quasar sight line (d ≈ 2 kpc) is an isolated cloud separated from the main body of G1 in the 3σ Hα surface brightness contour (see Figure 2). We used an optimal extraction method implemented in CubExtractor to search for Hα and [N ii] emission on top of the QSO, but the noise properties of the region affected by the bright QSO PSF prohibit a sensitive constraint on the Hα surface brightness at this location.

We estimate the warm ionized gas mass in the ISM of G1 and G2 as follows. For a clumping factor ${ \mathcal C }\equiv \langle {n}_{e}^{2}\rangle /\langle {n}_{e}{\rangle }^{2}$, the Hα surface brightness depends on the mean electron density, 〈n_e 〉, and the path length through the gas, L, according to

$$\begin{eqnarray}&&I({\rm{H}}\alpha )\approx 2\times {10}^{-15}\ \displaystyle \frac{{ \mathcal C }\langle {n}_{e}{\rangle }^{2}}{{\left(1+z\right)}^{4}}\ \displaystyle \frac{L}{\mathrm{kpc}}\ \mathrm{erg}\ {{\rm{s}}}^{-1}\ {\mathrm{cm}}^{-2}\ {\mathrm{arcsec}}^{-2}.\end{eqnarray} \tag{ 1 }$$

Here the electron density is in units of cm⁻³, and we assume an electron temperature T_e = 10⁴ K. If the ionized gas is found in a relatively thin disk (L ≈ 1 kpc) in G1, then M(H ii) ≈ 4–10 × 10⁷ M_⊙ for ${ \mathcal C }$ between 10 and 1. If, instead, L is closer to the projected size of the line-emitting region (L ≈ 7 kpc), then the ionized gas mass may be as high as 0.9–3 × 10⁸ M_⊙ for the same range of clumping factors. At approximately half of the Hα luminosity, the ionized gas mass estimates for G2 are roughly half of those for G1. Across these models, the electron density varies between 〈n_e 〉 ≈ 0.005 and 1 cm⁻³, which is consistent with the constraint from I([S ii] λ6718)/I([S ii] λ6732) that n_e ≲ 10² cm⁻³ (e.g., Osterbrock & Ferland 2006). Thus, the warm ionized phase of the ISM likely contributes between a few percent and 50% as much mass as the neutral phase.

3.4.2. WIM Kinematics

3.4.3. WIM Metallicity and Dust Content

For G1R1, we detect both the auroral [S iii] λ6313 and the nebular [S iii] λ9071 transitions, permitting a measurement of T_e and thus our most direct constraint on the gas-phase metallicity. In the top left panel of Figure 4, we show the aperture within which both lines are detected at the 3σ level, and we find I([S iii](λ9533 + λ9071))/I(S iii] λ6313) =29.0 ± 0.9 within this region (note that the nebular lines have a fixed ratio of I([S iii] λ9533)/I([S iii] λ9071) = 2.44). Following Osterbrock & Ferland (2006), this yields T_e ≈ 1.4 × 10⁴ K at the density constrained by the [S ii] doublet (n_e ≲ 10² cm⁻³). Adopting this value of T_e , we use the I([S ii] λ6718)/I(Hα) and I([S iii] λ9533)/I(Hα) line ratios to estimate S/H = S⁺/H + S⁺⁺/H ≈ 10%–20% of the solar value, consistent with the abundance determined for the DLA in Section 3.1.

Zoom In Zoom Out Reset image size

We additionally detect several strong-line metallicity indicators in multiple star-forming regions in both galaxies, including N₂ = I([N ii] λ6585)/I(Hα) (e.g., Pettini & Pagel 2004, hereafter PP04; see Figure 4). Following PP04, we find 12 + log(O/H) ≈8.0 (20% solar) from N₂ in G1R1 (note the 0.2 dex systematic uncertainty in the PP04 relation). We observe values up to 12 + log(O/H) = 8.3–8.4 (40%–50% solar) in lower surface brightness regions of this galaxy, but we regard this as an upper limit due to the likely presence of diffuse ionized gas that elevates the observed line ratios and thus biases the metallicity measurement (e.g., Zhang et al. 2017; Vale Asari et al. 2019). In the star-forming regions of G2, 12 + log(O/H) is generally ≈0.1 dex higher than in G1, suggesting slightly elevated metallicity in this galaxy. We corroborate a metallicity of a few tens of percent solar for G1 and G2 using Cloudy photoionization models to reproduce the observable emission-line ratios (Ferland et al. 2017). The dust extinction maps shown in the bottom panel of Figure 4 indicate a characteristic E(B − V) = 0.05 and 0.1 for G1 and G2, respectively. This is comparable to the characteristic E(B − V) ≈ 0.1 found for galaxies of this mass by Salim et al. (2018), suggesting that G1 and G2 are typical in their dust content.