Significant H I and Metal Differences around the z = 0.83 Lens Galaxy Towards the Doubly Lensed Quasar SBS 0909+532
DDraft version March 1, 2021
Typeset using L A TEX twocolumn style in AASTeX63
Significant H I and Metal Differences around the z = 0.83 Lens Galaxy Towards the Doubly LensedQuasar SBS 0909+532 Frances H. Cashman ,
1, 2
Varsha P. Kulkarni , and Sebastian Lopez Department of Physics & Astronomy, University of South Carolina, Columbia, SC 29208, USA Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA Departamento de Astronom´ıa, Universidad de Chile, Casilla 36-D, Santiago, Chile (Received June 20, 2020; Revised November 27, 2020; Accepted December 9, 2020; Published January 28, 2021)
ABSTRACTWe report a large difference in neutral hydrogen (H I ) and metal column densities between the twosight lines probing opposite sides of the lensing galaxy at z lens = 0.83 toward the doubly lensed quasarSBS 0909+532. Using archival HST -STIS and
Keck
HIRES spectra of the lensed quasar images, wemeasure log N H I = 18.77 ± − toward the brighter image ( A ) at an impact parameter of r A = 3.15 kpc and log N H I = 20.38 ± − toward the fainter image ( B ) at an impact parameterof r B = 5.74 kpc. This difference by a factor of ∼
41 is the highest difference between sight lines for alens galaxy in which H I has been measured, suggesting patchiness and/or anisotropy on these scales.We estimate an average Fe abundance gradient between the sight lines to be ≥ +0.35 dex kpc − .The N Fe II / N Mg II ratios for the individual components detected in the
Keck
HIRES spectra havesupersolar values for all components in sight line A and for 11 out of 18 components in sight line B ,suggesting that Type Ia supernovae may have contributed to the chemical enrichment of the galaxy’senvironment. Additionally, these observations provide complementary information to detections ofcold gas in early-type galaxies and the tension between these and some models of cloud survival. Keywords: galaxies: abundances – galaxies: elliptical – quasars: absorption lines INTRODUCTIONTraditional quasar absorption line studies probe a sin-gle sight line through a galaxy; however, it is difficult tolink the properties of an absorber to the galaxy host andspeculate about the galaxy’s properties based on a sin-gle sight line. The use of gravitationally lensed quasars(GLQs) to probe foreground galaxies improves on thesingle sight line method, as one has multiple sight linesto characterize the absorption regions of the galaxy. Us-ing multiple sight lines has the advantage of studyingvariations in gas, dust, and structure to offer a uniquetransverse study of a galaxy. Locally, multiple sight lineshave been successfully implemented to probe the inter-stellar medium (ISM) of the Milky Way (MW) and other
Corresponding author: Frances [email protected] nearby galaxies. Lauroesch et al. (2000) and Andrews etal. (2001) revealed turbulence-driven astronomical unit-scale variations in cold neutral gas structures traced bylow column density Na I absorption lines along closelyspaced stellar sight lines. Similarly, closely spaced sightlines toward GLQs can distinguish small-scale structurein the ISM of a lens galaxy. Additionally, any absorbersthat exist between the lens and the background quasarare magnified by lensing, potentially revealing parsec-to kiloparsec-scale structure depending on the locationof the absorber.Absorption line studies of lenses are not as common asnon-lens absorbers since most lenses lie at a redshift z (cid:46) I column density, which requires UV spectroscopy(as does any absorption line system with z < a r X i v : . [ a s t r o - ph . GA ] F e b Cashman et al. mits not only the determination of galaxy mass but alsoabundance gradients within the galaxy. This is note-worthy, as some lenses have shown positive or invertedgradients, i.e., gradients opposite to what is seen in theMW and other nearby galaxies, suggesting central di-lution from mergers or infall from metal-poor gas ( ∼− − − in the MW, Friel et al. 2002;Luck & Lambert 2011; Cheng et al. 2012; − − in M101, Kennicutt et al. 2003; − ± − in M33, Rosolowsky & Simon 2008; − ± − in nearby isolated spirals, Rupke etal. 2010)Previous lens galaxy imaging surveys suggest that themajority of lens galaxies are passively evolving normalearly-type galaxies (e.g., Keeton et al. 1998; Zahedy etal. 2016). Keeton et al. (1998) describe that lens galax-ies are a biased sample, typically very massive, as mas-sive galaxies are more likely to lens background objects.This mass bias favors early-type galaxies, with late-typespirals expected to compose only 10 −
20% of all grav-itational lenses. It has also been reported that approx-imately a third of nearby early-type ellipticals are notgas-poor but contain large amounts of H I gas, despitebeing quiescent (e.g., recent 21 cm surveys by Grossi etal. 2009; Oosterloo et al. 2010; Serra et al. 2012; Younget al. 2014). Additionally, there are QSO absorption linestudies of luminous red galaxies in which Mg II λλ (cid:38)
100 kpc (e.g., Gau-thier et al. 2010, 2009; Huang et al. 2016), suggestingthat these halos are chemically enriched. This detectionof enriched cool gas within passive galaxies raises ques-tions as to what processes and mechanisms exist withinthe galaxy that keep the gas from cooling further andforming stars.The mass bias in lensing studies can be advanta-geous, as studying variations within these lensing galax-ies along sight lines to the multiple images can con-tribute significantly to what we know about the ISMof passively evolving elliptical galaxies. In a QSO ab-sorption line study along multiple sight lines to threelensing galaxies, Zahedy et al. (2016) and Zahedy etal. (2017) reported that while the gas content variedamongst the lenses and within sight lines of the samelenses, supersolar [Fe/Mg] relative abundance patternswere observed in all sight lines that also had detectionsof cool gas. The high [Fe/Mg] ratios suggest a significantcontribution from Type Ia supernovae (SNe Ia) to thechemical enrichment history of the inner ISM of theselenses. These observations support current theories thatthe presence of mature stellar populations could preventfurther star formation from occurring in the reservoirsof chemically enriched cool gas due to a combination of injected energy from SNe Ia and winds from asymptoticgiant branch stars (e.g., Conroy et al. 2015).We used public archive
HST -STIS UV absorptionspectra and
Keck
HIRES optical spectra to study theunexplored difference in the chemical composition andkinematic structure of the lens galaxy at z = 0 .
83 to-ward the double GLQ SBS 0909+532. We report signif-icant H I and metal column density differences at pro-jected distances from the lens galaxy’s center of r A =3.15 and r B = 5.74 kpc on opposite sides of the lens inthe inner ISM. We see much heavier H I and metal ab-sorption in image B than in image A . We discuss theobserved N Fe II / N Mg II relative abundance pattern inSBS 0909+532 in comparison to the lenses Q1017 − − − − II absorption at a redshift of z abs = 0.611 along both linesof sight in § A and B . In section 4, we show the results of ourmeasurements. In section 5 we compare our results tosimilar studies from the literature. We adopt the cos-mology H = 70 km s − Mpc − , Ω M = 0.3, and Ω Λ =0.7 throughout this paper. OBSERVATIONS AND DATA REDUCTION2.1.
SBS 0909+532 Background
SBS 0909+532 is a well-studied doubly imaged grav-itationally lensed system identified by Kochanek et al.(1997), who initially observed two images at z QSO =1.377 separated by 1 . (cid:48)(cid:48)
1. Oscoz et al. (1997) were ableto confirm the redshift of the source quasar, as well asthe redshift of the reported Mg II absorption seen at z = 0 .
83. The lens redshift was later spectroscopicallyconfirmed to be z = 0.8302 ± z = 1 . ± . r eff = 1 . (cid:48)(cid:48) ± . (cid:48)(cid:48)
90) and low surfacebrightness. Lubin et al. (2000) estimated the mass in-side the Einstein ring to be 1.42 × M (cid:12) h − . Thecore of the lens is observable in I -band imaging as a sig-nificant residual after the quasar images are subtracted,but it does not appear in the V band. The local envi-ronment of the lensing galaxy consists of three nearbygalaxies within ∼
200 kpc h − , two of which are within ∼
100 kpc h − . ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 A and B imagesinferred from photometry, as well as low-resolution spec-troscopy, of SBS 0909+532 shows an extinction curveproduced by dust in the lens galaxy (Motta et al. 2002),the first optical extinction curve measurement at cosmo-logical distances that matches the quality seen locally.We note that differential extinction has also been de-tected in another galaxy at z = 0.93 (Wucknitz et al.2003). Both Motta et al. (2002) and Mediavilla et al.(2005) suggest that there may be a unique link betweenthe activity of a galaxy and the strength of the 2175˚A feature since passive, normal early-type galaxies aresubjected to fewer shocks and processing by radiation,unlike the environments of starburst galaxies and activegalactic nuclei.What has been missing from the portrait of SBS0909+532 is a description of the variation in metallicityand kinematic structure between the two sight line im-ages. As many lens galaxies lie at z <
1, the hydrogenLyman series lines fall in the UV portion of the electro-magnetic spectrum and are therefore only accessible tostudy with UV spectrographs on board space-based tele-scopes. Therefore, measurements of H I exist for only ahandful of lenses, and this, in turn, means that metal-licity measurements of lens galaxies are also scarce. Inthis paper, we use archival HST and
Keck spectroscopymeasurements to provide a deeper probe into the envi-ronment of early-type lens galaxies by providing multi-ple metallicity measurements at different impact param-eters through the ISM of the lens galaxy. Only four othermeasurements of H I exist for lens galaxies (Q1017 − AB , Q1355 − AB , Kulkarni et al. 2019; Q0047 − AB , Zahedy et al. 2017; Q0512 − AB , Lopez et al.2005; all lenses with redshift 0 . < z < . − AB , Q1355 − AB ,Q0512 − AB ) show a positive average metallicitygradient (or inverse gradient).A summary of the observational and spectroscopic pa-rameters for the QSO SBS 0909+532 and the absorptionline systems detected along the line of sight are shownin Tables 1 and 2. The impact parameters r , i.e., theprojected positions of the lensed images A and B rel-ative to the lens galaxy, are visually marked in Figure1. Image A is the brighter image and has a projecteddistance of 3.15 kpc from the lens galaxy. Image B isfainter and is at a projected distance of 5.74 kpc fromthe lens galaxy.2.2. Keck HIRES Observations and Data Reduction
Spatially resolved spectra of SBS 0909+532 A and B were downloaded from the Keck
Observatory archive .These data were obtained with the HIRES spectrograph(single CCD setup) on the night of 1998 December 18as part of program C140H (PI: W. Sargent) on the KeckI telescope. Five 2700 s exposures were taken of im-age A and 12 3000 s exposures were taken of image B .These data were reduced using the MAKEE package de-veloped by T. Barlow. The 2-pixel projected slit usedwas ∼ (cid:48)(cid:48) wide as projected on the sky, which yielded aspectral resolving power R ∼ linetools (Prochaskaet al. 2017) by fitting a spline function to the contin-uum of each quasar image. The wavelength coveragewas ∼ − Hubble Space Telescope STIS Observations andData Reduction
SBS 0909+532 was observed with
HST -STIS (SpaceTelescope Imaging Spectrograph) on 2003 March 7 inboth the optical and the UV using the CCD detector andthe G430L and G230LB gratings with STIS during Cycle11 as part of program GO-9380 (PI: E. Mediavilla). Thedata were initially presented in Mediavilla et al. (2005)and were used to determine an optical-FUV differentialextinction curve of the dust in the lens galaxy. We referthe reader to that paper for more specific details on theobservation.The H I column density is needed to measure thegas metallicity. The lens galaxy at z ∼ α line being redshifted to λ ∼ − IRAF /STSDAS X1D (Tody 1993) task to extract the one-dimensional spec-tra. Background subtraction, charge transfer efficiencycorrection, conversion to heliocentric wavelengths, andabsolute flux calibration were also performed during the X1D task. The average wavelength dispersion was 1.37˚A pixel − , the average resolution was ∼
500 km s − , and ∼ tb/makee/index.html IRAF is distributed by the National Optical Astronomy Observa-tory (operated by the Association of Universities for Research inAstronomy Inc.) under cooperative agreement with the NationalScience Foundation
Cashman et al.
AB G
Figure 1.
Left : HST image of SBS 0909+532 AB in the H band from the CASTLeS Survey (Kochanek et al. 1999). The lensgalaxy is clearly visible in between the lensed images. Leh´ar et al. (2000) report an absorption-corrected optical flux ratio inthe H band of A/B = 1.12.
Right : Chandra
Observations of SBS 0909+532 AB from Dai & Kochanek (2009). Table 1.
Basic Properties of the QSO SBS 0909+532 AB and the Lens Galaxy z QSO z lens Mag A , Mag B a ∆ θ AB (arcsec) b r A (kpc) c r B (kpc) d l A,B (kpc) e ± ± I ) 1.17 3.154 5.744 8.9 a I -band magnitude of each lensed quasar image (Leh´ar et al. 2000) b Angular separation between lensed quasar images c Impact parameter of image A from the lens galaxy, i.e., the projected distance between sight line A and the galaxy center d Impact parameter of image B from the lens galaxy, i.e., the projected distance between sight line B and the galaxy center e Transverse separation between the GLQ sight lines at the lens redshift
Table 2.
Absorption Line Systems Observedalong the Lines of Sight to SBS 0909+532 z W A (˚A) a W B (˚A) b l A,B (kpc) c ± ± ± ± a Equivalent width of the Mg II λ A b Equivalent width of the Mg II λ B c Transverse separation between the GLQ sight lines atthe absorber redshift the S/N ratio per pixel is ∼
30 in image A and ∼
16 inimage B in the regions of interest. As the flux levelsbetween the exposures were consistent, the three expo-sures of A were combined with equal weighting. Duringthe combination of exposures, gaps in spectral coveragedue to bad pixels were recovered by replacing the pixelwith the average flux value from the other two frames.The same procedure was performed for image B . The combined exposures of A and B were then continuumnormalized using linetools by fitting a spline functionin featureless spectral regions. ABSORPTION LINE AND COLUMN DENSITYMEASUREMENTSWhere possible, two methods were used to measurecolumn densities: the apparent optical depth (AOD)method (Savage & Sembach 1996) using the program
SPECP and Voigt profile fitting using the program VPFIT version 11.1 (Carswell & Webb 2014). The atomic datautilized by
VPFIT and
SPECP were adopted from the com-pilations of Cashman et al. (2017) and Morton (2003)(see Table 3 notes a - d for atomic information on thespecific transitions included in the fits).3.1. STIS H I Measurements
The H I Ly α and Ly β lines were used for estimatingthe column density along the sight line to the brighterimage A . Unfortunately, all shorter wavelength Lyman SPECP was developed by D. Welty and J. Lauroesch. ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 I Lyman series line thatcould be used for estimating the column density alongthe sight line to the fainter image B was the H I Ly α line.Although the column density in sight line B is estimatedwithout these higher order Lyman series lines (possiblywithin orders of magnitude), we determine a reasonablyrobust value for log N H I . Of course, we note that ob-servations of SBS 0909+532 AB at higher resolution inthe far-UV could permit measurements of higher orderLyman series lines up to the H I Lyman limit. The H I column densities for A and B were measured based onsingle component fits to the Ly α and/or Ly β lines using VPFIT . The methods to obtain the H I column densitiesare described below.3.1.1. SBS 0909+532 A
We find weak H I absorption near z lens = 0.83 in thespectrum of SBS 0909+532 A . As can be seen in the leftpanels of Figure 2 and in Figure 3, the H I Ly α featurespans a velocity range from ∼ −
600 to +600 km s − ,and although it is broad and saturated, at this resolu-tion ( v FWHM = 333 km s − ) it does not display dampingwings. Since both Ly α and a partially blended Ly β weredetected, they were fit together to constrain the H I col-umn density. We chose a conservative estimate for the b -value of the single component to be v FWHM /10 (i.e., b = 33 km s − ), based on Legendre-Gauss quadraturerepresenting the integration over the instrumental pro-file. The resulting log N H I from fitting both lines with
VPFIT is 18.77 ± − and the profile shows goodagreement with the data (see the left panels of Figure2). Repeating the fit with a b -value as small as 12 kms − results in log N H I = 18.83 ± − , which iswithin the margin of error of our log N H I measurementbased on the instrument profile. Although such a small b -value cannot be justified given the instrument profile,we include this result to show the robustness of the es-timate corresponding to a conservative b -value based onthe data resolution. We performed an additional fit fora b -value as large as 45 km s − resulting in log N H =18.60 ± − . This result is consistent with ourconservative measurement at about the 1 σ level. Thesefits can be seen in the left panels of Figure 2.3.1.2. SBS 0909+532 B
There is significantly more H I absorption observed inthe fainter sight line B than in sight line A . Given thelow resolution of the G230LB data and the broad Ly α feature (see Figures 2 and 3) both the z and b -value forthe single H I component was fixed to the redshift and b -value of the dominant metal line component observedin the high-resolution HIRES spectra, and only log N H I was allowed to vary using
VPFIT .Executing a χ minimization analysis to determinedamped and sub-damped H I column densities and un-certainties has historically given rise to unrealisticallylow uncertainties (see Prochaska et al. 2003b). There-fore, we created a series of Ly α profiles varying in stepsof ± B , it was determined that a ± N H I = 20.38 cm − gavethe most consistency between the fitted profile and theflux within a 2 σ buffer. Furthermore, to increase confi-dence in the adopted log N H I and its uncertainty, giventhe lower signal-to-noise ratio and apparently dampednature of the absorption seen along the sight line to im-age B , we utilized a technique (originally described byRao & Turnshek 2000) to examine the possibility thatadditional uncertainty may have also arisen from sub-jective continuum placement. Our original “most likely”normalized continuum was shifted above and below byan offset per pixel from the 1 σ error array of the fluxdata. Each offset spectrum was then renormalized anda Voigt profile was fitted again to the H I Ly α feature.The +1 σ continuum resulted in a column density of log N H I = 20 . ± .
20 cm − . The − σ continuum resultedin a column density of log N H I = 20 . ± .
25 cm − . Asthe mean value of these high and low log N H I values of20 . ± .
33 cm − is within the range of the value deter-mined from the initial fit, we chose to retain the initiallog N H I value of 20.38 cm − and the conservative ± − wide, as shown in the multiple componentsof the Mg II and Fe II lines detected in the correspond-ing HIRES data (see Figures 5 and 6). Other poten-tially existing UV metal lines at the redshift of the lensare shown in Figure 4, however, obtaining higher reso-lution UV spectra is necessary to confirm the identityof these metal lines and to measure the metal columndensities. Observations of other metal ions at high res-olution would permit more accurate determinations ofrelative ionization fraction corrections using ion ratiosbesides the only ratio available to us (Mg II /Mg I ). Mea-surements of the undepleted elements S and O wouldprovide more robust determinations of the metallicity, Cashman et al. and ion ratios such as Si II /S II or Si II /O I would allowfor dust depletion determinations.3.2. Keck HIRES Metal Line Measurements
Tables 3 and 4 list individual component column den-sity measurements for the metal line transitions detectedin the limited wavelength range of the available HIRESspectra. For sight line A , all velocity components seenin Mg II were also seen in Fe II . Mg I and Mn II werenot detected in image A ; therefore, we calculated the 3 σ upper limit to the column density from the 3 σ observedframe equivalent width upper limit for Mg I λ II transition at λ RDGEN (Carswell et al.2014) to select the velocity ranges of the metal lines andto initially mark and estimate component column densi-ties. The redshifts of the stronger components were se-lected by examining the weaker lines, and the redshiftsof the weaker components were selected by examiningthe stronger lines. For ions where multiple lines weredetected, they were fit together to constrain the ioniccolumn densities. The redshifts and b -values were alsotied together for ions of similar ionization stage. Prelim-inary initial guesses of the redshift, b -value, and columndensity were then fed into VPFIT to determine the finalresults for the individual components. These parameterswere allowed to vary and the program was permitted toadd, remove, and/or move the positions of the compo-nents until a Voigt profile fit to the data produced thelowest possible χ . In image A , the absorbing regionconsists of three very weak components that span a to-tal velocity range of ∼
66 km s − , whereas the absorbingregion in image B was fit with up to 21 componentsspanning a total velocity range of ∼
650 km s − , some ofwhich are saturated. All 21 components were detectedin Mg II , of which 18 could be detected in Fe II , 10 inMg I , and 2 in Mn II . The total column densities for theindividual ions are computed by adding all their con-stituent velocity components together. Voigt profile fitsto the spectra are shown in Figures 5 and 6. The totalcolumn densities, as well as the comparable AOD mea-surements (for the detected lines), can be seen in Table5. RESULTSThe redshifts, b -values, and column densities of eachsight line’s H I and metal absorption components aresummarized in Tables 3 and 4. The total column den-sities, calculated abundances, and overall average abun-dance gradients are summarized in Table 5. Followingcommon practice, the abundances of each element X aredefined as [X/H] = log ( N X / N H I ) − log (X/H) (cid:12) . Ele- ment abundances in the solar photosphere from Asplundet al. (2009) were adopted.4.1. Variations in the H I Column Density between theSight Lines
For the z lens = 0.83 absorber toward SBS 0909+532 AB , the H I column density is significantly higher alongsight line B , by a factor of 1.61 dex ( ∼
41 times higher),indicating that the neutral gas is not distributed homo-geneously around the lens. This asymmetry suggeststhat the sight lines are not probing a spatially coherentregion. This difference is interesting for two reasons.First, the separation between the sight lines is small,1.11 (cid:48)(cid:48) or 8.9 kpc at the redshift of the lens. This meansthat structural differences exist within this normal el-liptical galaxy on scales less than 8.9 kpc. Second, theimpact parameters ( r ) for sight lines A and B are r A =3.15 kpc and r B = 5.74 kpc; thus sight line B is located ∼ A . Thus, more H I exists furtherfrom the center of the galaxy, which could suggest thatthe region probed by sight line A at ∼ AB was observed on 2006 December 17 with the AdvancedCCD Imaging Spectrometer on board the Chandra
X-ray Observatory by Dai & Kochanek (2009). Similarlyto what we observe, Dai & Kochanek (2009) also sawmuch more absorption in image B than in image A .They measured log ∆ N H , B − A = 20.74 +0 . − . dex betweenthe sight lines. This difference is consistent with our re-ported H I column density measurement for sight line B within the margin of error. We consider the impact ofionization effects on our measurements for sight line A in § Element Abundances and Abundance Gradients
Similarly, we see significantly more metal absorptionin sight line B than in A (see Figure 7). All Mg II andFe II velocity components seen in sight line A are alsoseen in sight line B . However, sight line B shows severalvelocity components that are not seen in sight line A .The metallicities were calculated for each sight lineand are shown in Table 5. The average abun-dance gradient is calculated from the difference inFe abundances measured in the lensed images andthe difference in the impact parameters as mea-sured from the center of the lensing galaxy, i.e.,∆[Fe/H]/∆ r = ([Fe/H] B − [Fe/H] A )/( r B − r A ). Thiscalculation shows how the abundance between the im-ages would change per unit distance if the lens wereconsidered to be uniform. Although Fe depletes read-ily onto dust grains, we use [Fe/H] to characterize the ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 Figure 2.
Plots of the detected H I line(s) in the z = 0 .
83 lens galaxy in the
HST
STIS spectrum of SBS 0909+532 A and B . Ineach panel, the normalized data are shown in black, the dashed red line shows the continuum level, and the blue curve near thebottom shows the the 1 σ error in the normalized flux. The solid green curve in each panel indicates the theoretical Voigt profilefit. The dashed green curve above and below the fitted profile shows the uncertainty in log N H I . The black vertical dashed lineindicates the position of the H I component that was used in the fit (see Tables 3 and 4). Left panels: The Voigt profile fitsfor H I Ly α and Ly β corresponding to log N H I = 18.77 ± − with b = 33 km s − in the HST
STIS spectrum of SBS0909+532 A . The cyan and purple curves show the effects of different b -values on the fitted Voigt profiles profiles resulting from b = 12 km s − and b = 45 km s − respectively. Right panel: The Voigt profile fit for H I Ly α corresponding to log N H I = 20.38 ± − in the HST
STIS spectrum of SBS 0909+532 B . The solid colored vertical lines (magenta: Si II λ I λλ III λ z = 0 . I feature. We were unable to detect H I Ly β toward sight line SBS 0909+532 B because of extremenoise in the region. Table 3.
Line parameters in the z lens = 0 .
83 galaxy toward SBS 0909+532 A z b eff log N H I a log N Mg I b log N Mg II c log N Mn II d log N Fe II e (km s − ) (cm − ) (cm − ) (cm − ) (cm − ) (cm − )0.830129 ± ± ≤ ± ≤ ± ± ± ± ± ± ± ± ± ± ± a H I λ f -value = 0.4164, Pal´chikov (1998) b Mg I λ f = 1.71, Froese Fischer et al. (2006); There is a non-detection of Mg I λ σ upper limit to thecolumn density from the 3 σ observed frame equivalent width upper limit. c Mg II λλ f = 0.613, f = 0.306, Froese Fischer et al. (2006) d Mn II λλ f = 0.358, f = 0.279, f = 0.196, Den Hartog et al. (2011); There is a non-detection of Mn II λλ σ upper limit to the column density from the 3 σ observed frame equivalent width upperlimit for the strongest transition at λ e Fe II λλ f = 0.239, f = 0.320, f = 0.0313, Bergeson et al. (1996); Fe II λ f = 0.0717, Fuhr& Wiese (2006) average abundance gradient because measurements of[Fe/H] exist for other lenses for comparison. Table 5shows these results. The average abundance gradient∆[Fe/H]/∆ r ≥ +0.20 dex kpc − between the two sightlines, where the lower limit results from the lower limit of [Fe/H] B from the many saturated components at 150km s − (cid:46) v (cid:46) −
150 km s − . This positive gradientis much higher than the range of metallicity gradientsobserved in the MW and nearby galaxies ( ∼ − − − in the MW, Friel et al. 2002; Luck & Cashman et al.
Table 4.
Line parameters in the z lens = 0 .
83 galaxy toward SBS 0909+532 B z b eff log N H I log N Mg I log N Mg II log N Mn II log N Fe II (km s − ) (cm − ) (cm − ) (cm − ) (cm − ) (cm − )0.827682 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± > > ± ± ± > > ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± > ± ± ± ± > ± ± ± ± > ± ± ± ± > > ± ± ± > > ± > ± ± > > ± > ± ± ± ± ± ± ± ± ± ± Figure 3.
Velocity overplot of H I Ly α in the sight linetoward SBS 0909+532 A (gray+blue) with the same featureobserved in the sight line toward SBS 0909+532 B (pur-ple+pink). Lambert 2011; Cheng et al. 2012; − − inM101, Kennicutt et al. 2003; − ± − in M33, Rosolowsky & Simon 2008; − ± − in nearby isolated spirals, Rupke et al. 2010).4.3. Ionization Effects
The metallicities calculated for each sight line madethe assumption that the ion stages H I , Fe II , and Mg II can represent the total column density of that elementin the Lyman limit system (LLS; sight line A ) and thedamped Ly α absorber (DLA; sight line B ). The lower N H I detected in sight line A could be an indication thatthe environment is highly ionized; therefore, we investi-gate to what extent our results may be affected by ion-ization of the absorbing gas. In the case of the DLA insight line B , H I is expected to be self-shielding againstphotons capable of ionizing it ( hν > II and Fe II are indeed the dominantionization stages since Mg I and Fe I can be ionizedby even photons that cannot ionize H I , and in princi-ple some of the photons ionizing H I (those with hν > II and Fe II . Unfortunately, no higher ions have confirmeddetections in either sight line.We ran a suite of CLOUDY photoionization models us-ing version 17.01 (Ferland et al. 2017) to determine theextent of ionization effects in both sight lines. We usedthe approximation that the absorption regions are plane-parallel slabs and included the cosmic microwave back-ground at the redshift of the absorber and the extra-galactic UV background from Khaire & Srianand (2019)at the redshift of the lens (KS18 in
CLOUDY ) as the ra-diation fields. Additionally, we include the cosmic raybackground from Indriolo et al. (2007) since cosmic raysnot only heat ionized gas but also heat neutral gas andcreate secondary ionizations. The neutral hydrogen col-umn densities were fixed to the estimated values mea-sured from the STIS spectra listed in Tables 3 and 4.The gas metallicities were fixed to the values obtained ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 Figure 4.
Velocity plots of potentially present metal lines with respect to z = 0.83 in STIS image B (right two panels), withthe same lines for A (left two panels) included for comparison. There are no data at the red end of the panel for Al II λ I or Al II transitions may not be present inimage A . Table 5.
Total Column Densities, Metallicities, and Gradients in the z lens = 0 .
83 galaxy
SBS 0909+532 A SBS 0909+532 BIon log N AOD log N fit [X/H] a log N AOD log N fit [X/H] a ∆[X/H]/∆ r b Mg i ≤ c ... ... 13.04 ± ± ii ± ± − ± ± ≥ ≥ − ii ≤ d ... ≤ − ± ± − ± ii ± ± − ± ± ≥ ≥ − ≥ +0.20 a For sight line B , [Mg/H] is calculated from the total sum of log N Mg II and log N Mg I . For sight line A , [Mg/H] is calculated from log N Mg II onlydue to the non-detection of Mg I . b Average abundance gradient ∆[Fe/H]/∆ r = ([Fe/H] B − [Fe/H] A )/( r B − r A ) in dex kpc − ; does not include ionization corrections. c Non-detection of Mg I in the lensed image of A , column density in cm − is 3 σ upper limit to the column density that was calculated based on the3 σ observed frame equivalent width upper limit assuming a linear curve of growth. d Non-detection of Mn II in the lensed image of A , column density in cm − σ upper limit to the column density from the 3 σ observed frameequivalent width upper limit for the strongest transition at λ for Fe from the HIRES spectra (Z A , LLS ∼ (cid:12) andZ B , DLA (cid:38) (cid:12) for sight line A and sight line B re-spectively).The constraints on the number density log n H and theionization parameter log U were estimated by comparingthe value of the only observed column density ratio ofadjacent ions available, N Mg II /N Mg I , to the calculatedmodel ratio for a range of hydrogen number densitiesfrom 10 − to 10 cm − (see Figure 8). We acknowl-edge that this is a rather broad approach to determinethe number density, and thus the ionization parame-ter, in these systems. For sight line B , there are 10components with both Mg I and Mg II detected in theHIRES data. The observed log ( N Mg II /N Mg I ) values for these components range from ∼ I profiles athigher resolution based on higher-order Lyman serieslines; thus, we proceed by considering the ratio of thetotal of all components for Mg I and Mg II .The model for sight line A estimates log n H ≤ − . − and log U ≥ − .
56. Furthermore, it predicts log N H II = 20.78 cm − with total log N H = 20.78 cm − ,indicating that this region is dominated by ionized hy-drogen. For sight line B , log n H ≤ − and log U ≥ − .
71. For this cooler sight line, the model esti-mates log N H II = 18.29 cm − and total log N H = 20.380 Cashman et al.
Figure 5.
Voigt profile fits for the metal lines in the z ∼ .
83 lens galaxy in the Keck HIRES spectrum of SBS 0909+532 A ,ordered by ion and then by decreasing oscillator strength. In each panel, the normalized data are shown in black, the solidgreen curve indicates the theoretical Voigt profile fit to the absorption features, and the dashed red line shows the continuumlevel. The 1 σ error values in the normalized flux are represented by the blue curves near the bottom of each panel. The verticaldotted lines indicate the positions of the components that were used in the fit. As these lines show weaker absorption, thenormalized flux scales are shown starting at 0.5 and the 1 σ error arrays are offset by 0.5, so that they can be viewed in the samepanels. Note that Mg I and Mn II were not detected in image A , but Mg I λ II transition at λ B , as shown in Figure 6. cm − (essentially equal to log N H I ) and thus confirmsthat neutral hydrogen is the dominant stage. The modelalso predicts an H column density of 17.57 cm − . Wethen estimate the Fe abundance of the gas by correctingthe observed column density ratio by the predicted rel-ative ionization fraction for the estimated log n H , i.e.,log (Fe/H) = log ( N Fe II /N H I ) − log ( f Fe + /f H ). If therelative ionization fraction f Fe + /f H is ∼
1, then thegas metallicity can be approximated directly from log( N Fe II /N H I ), where then [Fe/H] ≈ log ( N Fe II /N H I ) − log (Fe/H) (cid:12) .In sight line B , we obtain a relative ionization fractioncorrection between Fe + and H of ( f Fe + /f H ) ≈ N Fe II /N H I )and obtain [Fe/H] ≥ − .
26. Dividing the model N H I ( ≈ model log N H ) by model n H = 3.6 cm − , the DLA regionis estimated to be ∼
21 pc along the line of sight. In sight line A , we obtain a relative ionization fractioncorrection log ( f Fe + /f H ) = 0.38. After correcting log( N Fe II /N H I ) by this amount, we adopt a metallicity of[Fe/H] = − .
16. This corrected metallicity results in anionization-corrected lower limit of ∆[Fe/H]/∆ r ≥ +0 . − . Dividing the model N H by the model n H ,the LLS absorbing region is estimated to be ∼
76 kpcalong the line of sight.We note that these models are based on H I columndensities derived from low resolution spectra as well asmetal column densities of refractory elements. Higherresolution UV spectra would not only provide more ro-bust H I column densities, but could also potentiallyprovide additional adjacent ion ratios to better constrainlog n H and the relative ionization fraction corrections.However, even with a model based on log N H I measure-ments from low resolution spectra, the results are con- ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 Figure 6.
Voigt profile fits for the metal lines in the z ∼ .
83 lens galaxy in the
Keck
HIRES spectrum of SBS 0909+532 B ,ordered by ion and then by decreasing oscillator strength. In each panel, the normalized data are shown in black, the solidgreen curve indicates the theoretical Voigt profile fit to the absorption features, and the dashed red line shows the continuumlevel. The 1 σ error values in the normalized flux are represented by the blue curves near the bottom of each panel. The verticaldotted lines indicate the positions of the components that were used in the fit. Shaded regions indicate absorption unrelated tothe presented line. sistent with the photoionization studies from Bergeron& Stasi´nska 1986, which suggest that Mg II absorptiondominates in regions of cool, photoionized gas of T ∼ K. We also note that ionization modeling results aresensitive to the atomic parameters used, such as dielec-tronic recombination coefficients. Improvements to theaccuracies of these parameters are essential for improv-ing the reliability of ionization modeling calculations,such as those presented here.4.4.
The Transverse Separation and Mass of the LensGalaxy
The transverse separation between the GLQ imageswas calculated for the absorber at z abs = 0.611 and thelens at z lens = 0.8302. For an absorber with a redshiftgreater than or equal to the lens redshift, the transverseseparation l AB between the sight lines is calculated as l AB = D aq (1 + z l )∆ θ AB D l D lq (1 + z a ) , (1) where D l , D lq , and D aq are the angular diameter dis-tances between the observer and the lens, between thelens and the quasar, and between the absorber and thequasar, respectively, and ∆ θ AB is the angular separa-tion between quasar images A and B . In the case whenthe absorber is the lens itself, i.e., z l = z a and D aq = D lq , and Equation 1 simplifies to l AB = ∆ θ AB D l .The angular diameter distances in Equation 1 are cal-culated using D = cH (1 + z ) (cid:90) z z [Ω Λ + Ω m (1 + z ) ] − / dz, (2)following Hogg (1999).In addition to the unique transverse study of the lensand other absorbers that GLQs provide, analysis of thelensed images provides an opportunity to determine themass and the mass distribution of the lens. The massdistribution for an early-type galaxy is presumed tobe that of a singular isothermal sphere (SIS) given by2 Cashman et al.
Figure 7.
Velocity overplots of the detected metal lines in the sight line to the lensed image of SBS 0909+532 A (in purple),with the same metal lines observed in the sight line to the lensed image B (in gold) to illustrate the high degree of differencein absorption between the two lines of sight. The three vertical gray lines at ∼
21, 39, and 64 km s − are the locations of thethree components in the metal lines observed in the sight line to lensed image A . These three components lie entirely withinthe width of the two components near 28 and 60 km s − in lensed image B . ρ ∝ r − (e.g., Koopmans et al. 2009), in which the lensmatter behaves as an ideal gas in thermal and hydro-static equilibrium confined by a spherically symmetricgravitational potential. The velocity dispersion of anSIS of a galaxy lensing a quasar that produces the ob-served lensed image separation ∆ θ is σ SIS = c D q ∆ θ πD lq , (3)where D q and D lq are the same angular diameter dis-tances between the observer and the quasar and the lensand the quasar that were calculated from Equation 2.We estimate a velocity dispersion of 258 km s − for SBS0909+532. This value is comparable to the velocity dis-persion obtained from Oscoz et al. (1997) of 272 km s − .We estimate the mass of the lens galaxy from the as-trometry of the lensed images relative to the lens itselfusing M = − c ∆ θ AG ∆ θ BG D q D l GD lq , (4)where ∆ θ AG and ∆ θ BG (with opposite signs) are theangular separations of lensed images A and B from thelens (e.g., Schneider et al. 1992). Given the angular sep-arations and our calculated angular diameter distances,we estimate log ( M/M (cid:12) ) = 11.3. Our value for the massof the lens of SBS 0909+532 is in agreement with theestimate of log (
M/M (cid:12) ) = 11.31 by Lubin et al. (2000), which they based on the galaxy surface brightness profilewithin the Einstein radius. DISCUSSIONMeasurements of N H I , and therefore measurementsof metallicities, have been performed for only four otherlenses (Q1017 − AB , Q1355 − AB , Kulkarni etal. 2019; HE 0047 − AB , Zahedy et al. 2017; HE0512 − AB , Lopez et al. 2005) along multiple sightlines through the lens galaxy. Thus, the measurementsfor the lens at z = 0.83 toward the two sight lines towardSBS 0909+532 AB add important information to thissmall sample.5.1. H I Absorption in Lenses
Large differences in H I and metal column density areobserved at small impact parameters on either side ofthe galaxy. Sight line A , with an impact parameter r A = 3.15 kpc from the lensing galaxy, shows significantlyless neutral hydrogen absorption than sight line B at animpact parameter r B = 5.74 kpc from the galaxy. Thedifference between log N H I , A = 18.77 ± − andlog N H I , B = 20.38 ± − of 1.61 dex shows thatthe column density of neutral hydrogen drops by a factorof 41 over a transverse distance of 8.9 kpc. To comparethe physical extent of the H I absorption and the scaleover which it varies, we compute the fractional differencein log N H I measured at the redshift of the lens along ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 Figure 8.
Estimated number densities for the two sight linestoward SBS 0909+532 AB using CLOUDY version 17.01.
Top :Comparison of model log ( N Mg II /N Mg I ) over a range of log n H values and the observed lower limit of log ( N Mg II /N Mg I ) ≥ N H I = 18.77 ± − ) observedin the sight line to the lensed image SBS 0909+532 A . Weestimate log n H ≤ − − and log U ≥ − . Bottom :Comparison of model log ( N Mg II /N Mg I ) over a range of log n H values and the observed log ( N Mg II /N Mg I ) ≥ N H I = 20.38 ± − ) at z = 0 .
83 in the sightline to the lensed image of SBS 0909+532 B . We estimate log n H ≤ − and log U ≥ − both lines of sight, (log N H I , X − log N H I , Y )/log N H I , X ,where sight line X has stronger H I absorption out ofthe two, and compare this difference to other lenses inthe left panel of Figure 9. In fact, SBS 0909+532 AB shows the largest fractional difference in N H I (a differ-ence of 0.98 between log N H I , B and log N H I , A ) for thesmall sample of lenses for which measurements of N H I exist. Three other lenses (at z lens = 0.408, 0.933, and1.085 toward quasars HE 0047 − − − N H I ≤ − N H I .While such a large difference in H I absorption mayseem surprising for this small sample of lenses, itis consistent with the N H I differences seen for otherlenses (e.g. Q1355 − z lens , and even HE 0047 − > N H I for quasar absorption line systems(LLS, sub-DLAs, and DLAs) along the lines of sightto gravitationally lensed quasars, but they are not thelenses themselves (see Table 26 in Kulkarni et al. 2019for details on these absorption systems). However, thesenon-lens absorbers are likely probing absorption regionsbelonging to a variety of unknown host galaxies, and itis entirely possible that these lines of sight intersect thegalaxy on the same side with larger and similar impactparameters and thus may be expected to show strongerspatial correlations (i.e., smaller fractional variations).The large fractional variations seen in some non-lens ab-sorbers could perhaps be explained if those absorbersare associated with less well-mixed gas, e.g., in massivegalaxies with more extended star-formation histories. Inany case, it is interesting to note that the large differ-ence in H I (1.61 dex) observed between the sight lines ofSBS 0909+532 is the highest observed so far amongst thesmall sample of multiply imaged early-type galaxies. Wealso note that the large range of possible fractional dif-ferences underscores the need for higher resolution UVspectra of these GLQs.As mentioned in § z = 0.83 in their effort to study the evolutionof the dust-to-gas ratio. They also reported heavier N H absorption in image B than in image A with ∆log N H B − A = 20.74 +0 . − . , which is consistent with our adopted valueof log N H I , B = 20.38 ± − within the margin oferror. Hydrogen seen in elliptical galaxies comes in mul-tiple forms, predominately as X-ray-emitting hot gas,perhaps from SNe and stellar winds (e.g., Loewensteinet al. 1998, Mathews & Brighenti 2003, Pipino & Mat-teucci 2011). A large presence of hot hydrogen sup-ports the idea that mature stellar populations could bewhat prevents reservoirs of chemically enriched cool gasfrom collapsing into furthering star formation in ellip-tical galaxies, due to a combination of injected energyfrom SNe Ia and winds from asymptotic giant branchstars. SBS 0909+532 also happens to reside in a groupenvironment, with three nearby galaxies, two of whichare within ∼
100 kpc, that likely contribute tidal effects(Leh´ar et al. 2000). Thus, we cannot rule out that pastinteractions between group members could have heatedor tidally stripped cool gas from the inner regions ofSBS 0909+532. As there is a large difference in boththe presence and amount of neutral gas and ions seen4
Cashman et al. between images A and B given the small impact param-eters on either side of the galaxy, it is possible that thegalaxy’s evolutionary history includes a mixed merger(e.g., a wet-dry merger).5.2. Metal Absorption in Lenses
The lack of coherence between the two sight lines sep-arated by 8.9 kpc is clearly seen in the metal absorptionlines of the HIRES high resolution spectra in Figures 5and 6. We calculated the fractional difference in Fe II for all lenses in the small sample that also have H I measurements. These calculations are displayed in theright panel of Figure 9. SBS 0909+532 AB and HE1104 − AB (Zahedy et al. 2016) show the highestfractional differences in Fe II between lens sight lines inthe sample of lenses. In fact, if the galaxy along theline of sight to Q1355 − z = 0.70 is not the lens,then SBS 0909+532 and HE 1104 − II above ∼ [Fe/Mg] Abundance Ratios As mentioned in §
1, Zahedy et al. (2016) and Za-hedy et al. (2017) investigated [Fe/Mg] ratios in threelens galaxies along sight lines where cool gas was de-tected to look for possible contributions to the chemi-cal enrichment history of the inner ISM of lenses fromSNe Ia. Mg II absorption traces cool, photoionized gasof T ∼ K (Bergeron & Stasi´nska 1986). Currenttheories suggest that a combination of injected energyfrom SNe Ia and winds from asymptotic giant branchstars from mature stellar populations could be respon-sible for quenching star formation in the reservoirs ofchemically enriched cool gas in passive galaxies. Za-hedy et al. (2017) reported that while the gas contentvaried amongst the lenses and within sight lines of thesame lenses, supersolar [Fe/Mg] relative abundance pat-terns were observed in all sight lines that also had de-tections of cool gas. As each SN Ia is estimated to con-tribute ∼ M (cid:12) of iron and (cid:46) M (cid:12) of magnesium(see Iwamoto et al. 1999), the high relative abundancethey observed suggests a significant contribution to thechemical enrichment of these lens galaxies from SNe Ia.Zahedy et al. (2017) also reported supersolar values ofobserved log N Fe II / N Mg II for most of the individualcomponents detected in each sight line toward one of thelens galaxies in their sample, using abundances adopted from Asplund et al. 2006 (log (Fe/Mg) (cid:12) > − II and Fe II are the respective dominant ionization stages and the ra-tio N Fe II / N Mg II should trace the total [Fe/Mg] abun-dance ratio. Thus, each cloud of cool gas shows thesupersolar relative abundance suggestive of a significantcontribution from SNe Ia. Another factor that can af-fect the [Fe/Mg] ratio is that Fe depletes more stronglyon dust grains than Mg. This would decrease [Fe/Mg],so the true [Fe/Mg] is even higher.The top panel of Figure 10 shows the observed log N Fe II / N Mg II values for the individual components re-solved in the HIRES spectra of SBS 0909+532 AB ver-sus their velocity offset. All three resolved componentsin sight line A show a much higher log N Fe II / N Mg II value than the typical solar abundance level of − I and Mg II column densities and photoionization modeling suggestthat this sight line is not probing cool gas. Most of thecomponents in sight line B are saturated and blended.Eleven out of 18 components detected in sight line B show an observed log N Fe II / N Mg II value higher thanthe solar value. Four of these components can be con-sidered unblended; however, they show subsolar log N Fe II / N Mg II values. These components are very weak,11.85 < log N Mg II (cm − ) < r e of the lens galaxyis 12.01 ± r A ≈ r e and r B ≈ r e . Thus, even though there is not aunanimous trend of supersolar relative abundance ratiosseen in the cloud components in sight line B at this lowimpact parameter, it does suggest the possibility thatthe chemical enrichment in this region of the lensinggalaxy could be from SNe Ia.Additionally, we plot the observed log N Fe II / N Mg II values for the individual components resolved in theMagE spectra of both Q1017 − AB and Q1355 − AB (see the middle and bottom panels of Figure 10,where both lenses are described in Kulkarni et al. 2019).Q1017 − AB is doubly imaged by a lens galaxy at z lens = 1.086 with sight lines separated by l AB = 6.9kpc and log N H I , A = 19.87 ± N H I , B =19.79 ± − A andtwo components in the sight line toward Q1017 − B .1355 − AB is doubly imaged by a lens galaxy at z lens = 0.48 and l AB = 7.3 kpc and log N H I , A = 18.81 ± N H I , B = 19.43 ± N Fe II / N Mg II if the lensgalaxy toward Q1355 − B is instead at redshift z lens ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 Figure 9.
Fractional difference in log N H I (left panel) and log N Fe II (right panel) for SBS 0909+532 AB , calculated as (log N X − log N Y )/log N X , where sight line X has stronger absorption of the two. The blue and green shapes are lenses for which H I and Fe II , respectively, have been measured (Q1017 − AB , Q1355 − AB , Kulkarni et al. 2019; HE 0047 − AB , Zahedyet al. 2017; HE 0512 − AB , Lopez et al. 2005; HE 1104 − AB , Zahedy et al. 2016, Fe II only). SBS 0909+532 AB ,represented with a magenta circle, shows the highest H I fractional difference between lens sight lines. Only four other lensesto date have measurement of H I in all sight lines and thus have computed H I fractional differences; however, five are shown inthe figure, as Q1355 − z lens = 0.48 and 0.70. Regardless, the H I fractional uncertaintyfor either lens redshift measurement is in the close range of 0.74 − II , shown in the rightpanel. The orange diamonds in both panels are quasar absorption line systems along the lines of sight to the gravitationallylensed quasars but are not the lenses themselves (H1413+117, Monier et al. 2009, H I only; Q0957+561AB, Churchill et al. 2003;Q1104 − I only; SDSS J1442+4055, Krogager et al. 2018; see Table 26in Kulkarni et al. 2019 for a summary of measurements for these particular absorption systems). The bars on the points showthe maximum and minimum fractional difference possible given the uncertainty in log N X . = 0.70, which is discussed in greater detail in Kulkarniet al. (2019). None of the components in the sight linestoward Q1355 − A or Q1355 − B at either red-shift show a supersolar [Fe/Mg] ratio.5.4. Cloud Survival in Early-type Galaxies
The origin of the observed neutral gas in SBS0909+532 is unknown. We consider the possibility thatthe absorbing gas may arise in the ISM of the lensgalaxy. The gas detected in sight line A shows very lowFe enrichment and if it were ISM dominated, we wouldexpect to see a higher Fe enrichment in the lower impactparameter to the lensed image A . For the cooler gas insight line B , we observe ∼
6% solar Fe enrichment, al-though the true metallicity may be higher since we arecurrently unable to account for depletion. Our detectionof a supersolar [Fe/Mg] relative abundance in 11 out of18 cloud components could suggest ongoing enrichmentin the ISM due to an aging stellar population. For thesesight lines positioned on opposite sides of the galaxy,we cannot know to what extent the low impact param-eter means a higher probability of hitting the disk, andtherefore we consider multiple origin scenarios.A recent study by Afruni et al. (2019) explores multi-ple interpretations of the origin and fate of cool circum-galactic gas clouds associated with massive early-type galaxies. Their favored interpretation is that the coolclouds originate as the result of filaments of low metal-licity intergalactic gas accreting into the galaxy halo,which then fall in through the halo. They conclude thatit is very unlikely that these clouds survive the entire in-fall and instead evaporate in the hot corona. Thus, theinternal regions of the halo are mostly devoid of cool gas,and early-type galaxies are quiescent even though largereservoirs of cool circumgalactic gas are present. On theother hand, Nelson et al. (2020) report the possibility ofharboring cool gas at small impact parameters based onhigh-resolution TNG50 cosmological magnetohydrody-namical simulations, which explore the origin and prop-erties of cold circumgalactic medium (CGM) gas aroundmassive early-type galaxies at z ∼ I and Mg II , concluding that these halos mayhost ∼ discrete absorbing cloudlets, approximately akiloparsec or smaller in size. In this scenario, it is notsurprising to find H I column densities as high as thatobserved in SBS 0909+532 B at an impact parameterof 5.74 kpc away from the galaxy center.However, if we consider the neutral gas observed insight line B within the context of it being of external6 Cashman et al.
Figure 10.
Observed column density ratio for log( N Fe II /N Mg II ) for individual components versus their veloc-ity offset from the center of the lens galaxy for SBS 0909+532 AB . The black dashed line is the solar ratio log (Fe/Mg) (cid:12) = − σ uncertainties for the calculated ratios. Additionally, we showthe same plot for both Q1017 − AB and Q1355 − AB , which were observed and reported on in Kulkarni et al.(2019). Top : SBS 0909+532 shows a supersolar value oflog ( N Fe II /N Mg II ) for all components in sight line A , butfor only 11 out of 18 components in the cooler sight line B . Middle : Q1017 − AB is doubly imaged by a lens galaxyat z lens = 1.0859 with sight lines separated by l AB = 6.9 kpc.Only one of the three resolved components in the MagE spec-trum shows a supersolar value for log ( N Fe II /N Mg II ) in sightline A . All three resolved components show a supersolar valuefor log ( N Fe II /N Mg II ) in sight line B . Bottom : Q1355 − AB is doubly imaged by a lens galaxy at z lens = 0.48 and l AB = 7.3 kpc. None of the components in the MagE spec-trum along either sight line show a supersolar value for log( N Fe II /N Mg II ). origin, as described in Afruni et al. (2019), then wecan hypothesize possible scenarios for this low metal-licity cloud’s survival. We do note that the abundanceis based on Fe ([Fe/H] B ≥ − . § B , which supports a possibility thatwe are seeing signatures of energy interjection.5.5. An Mg II absorber at z abs ≈ While examining the HIRES spectrum of SBS0909+532 AB for metal lines at the redshift of the lens,we found an Mg II absorption system in both sight lineswith the dominant Mg II component centered at a red-shift of z abs , A = 0.6116 and z abs , B = 0.6114. No othermetal lines were detected at these redshifts. The STISspectrum was examined for corresponding H I absorp-tion; however, none was seen due to extremely high noiseat the blue end. The Mg I and Mg II column densi-ties were measured using both Voigt profile fitting andthe AOD method where possible. The Voigt profile fitsare shown in Figure 11. The equivalent widths of theMg II lines are listed in Table 2 and the column densi-ties for the individual components and the total columndensities are listed in Table 6. While the Mg I andMg II absorption is stronger in sight line B comparedto sight line A , the Mg II /Mg I ratio is comparable inboth sight lines, with N Mg II / N Mg I ∼
53 in sight line A and N Mg II / N Mg I ∼
56 in sight line B . Although thissystem is unrelated to the lens galaxy, we report its ex-istence for those interested in lensed Mg II absorptionline systems. CONCLUSIONWe examined both
HST -STIS UV spectra and
Keck
HIRES optical spectra of the images of the doubly lensedquasar SBS 0909+532 at z QSO = 1.37 to study the spa-tial differences in neutral hydrogen and metal absorp-tion lines of the lens galaxy at z lens = 0.83. We detecta significant difference in H I , Mg II , and Fe II betweenthe lensed images separated by 8.9 kpc. We calculate alarge fractional difference ( ≥ r A = 3.15 kpc (0.26 r e ) and r B = 5.74 ignificant Differences in H I and Metals around the Lens Galaxy Towards SBS 0909+532 Table 6.
Results of Voigt Profile Fitting for Ions in the z = 0.611 Mg II Absorber along the Sight Line towardSBS 0909+532 A and B SBS 0909+532 A SBS 0909+532 Bz A b eff a log N Mg I b log N Mg II c z B b eff a log N Mg I b log N Mg II c ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± N X , fit ± ± N X , fit ± ± N X , AOD ... 12.59 ± N X , AOD ± ± a In km s − b Measurements determined from Mg I λ − . c Measurements determined from Mg II λλ − . kpc (0.48 r e ) from the lens galaxy ( r e ∼
12 kpc). Wemeasure log N H I , A = 18.77 ± − and log N H I , B = 20 . ± .
20 cm − in the STIS spectra and find thatthe lens for SBS 0909+532 AB shows the highest frac-tional difference in H I of lensing galaxies for which onlyfour other H I measurements currently exist (see the leftpanel of Figure 9). High ionization in the region probedby sight line A is likely responsible for these differencesin H I , as Dai & Kochanek (2009) also reported similardifferences in X-ray absorption. The iron abundance islow in both sight lines, [Fe/H] = − .
78 (uncorrected) inimage A and [Fe/H] ≥ − .
26 in image B , but could behigher possibly due to dust depletions for which we areunable to account, given the limited wavelength rangeof the optical spectra. Thus, we are unable to ascertainas to what extent dust depletion may be influencing thedifferences we observe. We performed ionization modelson both sight lines, with the model for sight line A sug-gesting that the region is highly ionized. The resultingrelative ionization fraction is f F e + /f H = 0.38, and weadopt [Fe/H] = − .
16 for sight line A . Ionization cor-rections were unnecessary for sight line B , as the modeldetermined that the relative ionization fraction is ∼ ≥ +0.35 dex kpc − based on a relative ionizationfraction corrected [Fe/H] and a sight line separation of8.9 kpc, showing that the metallicity could be increasingwith radius for this particular lens galaxy. However, weemphasize that higher resolution UV spectra of the indi-vidual images are essential to obtain even firmer valueson both log N H I for sight lines A and B and to measurecolumn densities of additional metals of higher ioniza-tion stages in order to explore the multiphase nature ofthe absorbing gas. These measurements would permita more robust determination of metallicities and ion-ization fraction corrections, which would be beneficial since our initial findings suggest that these significantdifferences are due to differences in ionization. We alsocompare our results with those from recent studies sug-gesting that SNe Ia make significant contributions tothe chemical enrichment of the environment of ellipticalgalaxies (e.g., Zahedy et al. 2016, Zahedy et al. 2017,and Mernier et al. 2017). We find that 11 of 18 cloudcomponents in the cooler sight line B show a supersolar N Fe II /N Mg II value, which supports the idea that energyinterjection from aging stars in elliptical galaxies mayprevent reservoirs of cool gas from collapsing to furtherstar formation. Finally, even though the origin of thecold gas detected along the line of sight to image B isunknown, we discuss its existence within the context ofsome models of cloud survival. We find that our observa-tions resemble the scenario described in the simulationsfrom Nelson et al. (2020), who predict a population ofsmall-scale ( < I and Mg II . In this frame-work, it would not be unexpected for a high H I columndensity cloud to survive at an impact parameter of 5.74kpc from the center of the lens galaxy.Robust measurements of both volatile and refractoryelements will enable determination of the differences indust depletions between the different sight lines. Thesedifferences, together with differential dust extinctioncurves and measurements of the 2175 ˚A bump (alreadyavailable for SBS 0909+532 AB from Motta et al. 2002)will allow a detailed look at differences in dust struc-ture and composition on kiloparsec scales. Increasingthe samples of UV and optical high-resolution spectrafor other gravitationally lensed quasar sight lines is alsoessential to understand how common the findings fromSBS 0909+532 AB are. Measurements of metallicities,relative abundances, dust depletions, and ionization pa-rameters along the lensed sight lines for a large sample8 Cashman et al.
Figure 11.
Voigt profile fits of the Mg II absorber detectedat z = 0 . A (top) and z = 0 . σ error values in the normalized flux are representedby the blue curves near the bottom of each panel. will improve understanding of the spatial scales of pro-cesses important for galaxy evolution, e.g., chemical en-richment and heating of the ISM by SNe Ia vs. SNe IIsupernovae and the processing of dust grains. ACKNOWLEDGMENTSWe thank an anonymous referee for constructivecomments that have helped to improve this paper.F.H.C. and V.P.K. gratefully acknowledge support fromNASA/ Space Telescope Science Institute (grant HST-GO-13801.001-A, PI Kulkarni) and from NASA grantNNX17AJ26G (PI Kulkarni). V.P.K. also gratefully ac-knowledges support from NASA grant 80NSSC20K0887and NSF grant AST/2009811. S.L. was funded byproject FONDECYT 1191232. Facilities:
HST(STIS), Keck(HIRES)
Software: