Comprehensive Study of a z = 2.35 DLA Galaxy: Mass, Metallicity, Age, Morphology and SFR from HST and VLT
Jens-Kristian Krogager, Johan P. U. Fynbo, Cedric Ledoux, Lise Christensen, Anna Gallazzi, Peter Laursen, Palle Møller, Pasquier Noterdaeme, Celine Peroux, Max Pettini, Marianne Vestergaard
MMon. Not. R. Astron. Soc. , 1–12 (2002) Printed 29 October 2018 (MN L A TEX style file v2.2)
Comprehensive Study of a z = 2 . DLA Galaxy:Mass, Metallicity, Age, Morphology and SFR from HST and VLT (cid:63)
Jens-Kristian Krogager, , † Johan P. U. Fynbo , C´edric Ledoux , Lise Christensen ,Anna Gallazzi , , Peter Laursen , Palle Møller , Pasquier Noterdaeme ,C´eline P´eroux , Max Pettini , Marianne Vestergaard , Dark Cosmology Centre, Niels Bohr Institute, Copenhagen University, Juliane Maries Vej 30, 2100 Copenhagen Ø, Denmark European Southern Observatory, Alonso de C´ordova 3107, Vitacura, Casilla 19001, Santiago 19, Chile INAF – Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, 50125 Firenze, Italy European Southern Observatory, Karl-Schwarzschild-Strasse 2, D-85748 Garching bei M¨unchen, Germany UPMC-CNRS, UMR7095, Institut d’Astrophysique de Paris, F-75014 Paris, France Aix Marseille Universit´e, CNRS, LAM (Laboratoire d’Astrophysique de Marseille) UMR 7326, 13388, Marseille, France Institute of Astronomy, Kavli Institute for Cosmology, Madingley Road Cambridge, CB3 0HA Steward Observatory, University of Arizona, 933 N Cherry Avenue, Tucson, AZ 85721, USA
Accepted 30 May 2013. Received ; in original form
ABSTRACT
We present a detailed study of the emission from a z = 2 . galaxy that causes dampedLyman- α absorption in the spectrum of the background QSO, SDSS J 2222 − HST imaging, and of spectroscopy from VLT/X-Shooter of thestrong emission lines: Ly α , [O II ], [O III ], [N II ], H α and H β . We compare the metallicity fromthe absorption lines in the QSO spectrum with the oxygen abundance inferred from the strong-line methods (R and N2). The two emission-line methods yield consistent results: [O/H] = − . ± . . Based on the absorption lines in the QSO spectrum a metallicity of − . ± . is inferred at an impact parameter of 6.3 kpc from the centre of the galaxy with a columndensity of hydrogen of log( N H I / cm − ) = 20 . ± . . The star formation rates of the galaxyfrom the UV continuum and H α line can be reconciled assuming an amount of reddening of E ( B − V ) = 0 . ± . , giving an inferred SFR of ± (cid:12) yr − (Chabrier IMF).From the HST imaging, the galaxy associated with the absorption is found to be a compact( r e =1.12 kpc) object with a disc-like, elongated (axis ratio 0.17) structure indicating that thegalaxy is seen close to edge-on. Moreover, the absorbing gas is located almost perpendicularlyabove the disc of the galaxy suggesting that the gas causing the absorption is not co-rotatingwith the disc. We investigate the stellar and dynamical masses from SED-fitting and emission-line widths, respectively, and find consistent results of × M (cid:12) . We suggest that the galaxyis a young proto -disc with evidence for a galactic outflow of enriched gas. This galaxy hints athow star-forming galaxies may be linked to the elusive population of damped Ly α absorbers. Key words: galaxies: formation – galaxies: high-redshift – galaxies: ISM – quasars: absorp-tion lines – quasars: individual: SDSS J 22:22:56.1 − (cid:63) Based on observations collected at the European Organisation for As-tronomical Research in the Southern Hemisphere, Chile, under program087.A-0085(A). Based on observations made with the NASA/ESA HubbleSpace Telescope, obtained at the Space Telescope Science Institute, whichis operated by the Association of Universities for Research in Astronomy,Inc., under NASA contract NAS 5-26555. These observations are associ-ated with program 12553. † E-mail: [email protected]
Mapping the structure, properties and chemical enrichment ofgalaxies over cosmic history is a major goal of contemporary as-trophysics. In the local Universe, important advances can be madeby studying the age-metallicity relation of stars in the Solar neigh-bourhood (e.g., Holmberg et al. 2007; Caffau et al. 2011) or in localgroup dwarf galaxies (Frebel et al. 2007, 2010). At redshifts higherthan z ≈ , a powerful method for studying chemical evolutionis spectroscopy of Damped Ly α Absorbers (DLAs) detected eithertowards background QSOs (see Wolfe et al. 2005, and references c (cid:13) a r X i v : . [ a s t r o - ph . C O ] S e p J.-K. Krogager et al. therein) or against Gamma-ray Burst (GRB) afterglow light (Fynboet al. 2006; Savaglio 2006; Prochaska et al. 2007). In those studies,abundances are determined from the gas phase as probed by the H I and metal absorption lines detected against the light of the back-ground QSOs and GRB afterglows, respectively.Mainly at lower redshifts, but recently also at z (cid:38) , H II -region abundances are determined using the relative strengths ofstrong emission lines (e.g., Kewley & Ellison 2008; Shapley 2011,and references therein). Currently, there have only been very fewcases where both methods have been applied to the same object.The first example is the case of the DLA towards SBS 1543+593(Bowen et al. 2005) where the two methods yielded consistent re-sults. Since then a handful of other cases have been studied (seethe compilation in P´eroux et al. 2012). As seen in Fig. 8 of P´erouxet al. (2012), the absorption-line measurements generally probe re-gions at larger galactocentric distances than the emission-line basedmeasurements and on average indicate lower metallicities than theresults based on emission-lines, possibly reflecting the early setupof metallicity gradients (O’Rourke et al. 2011). However, it is im-portant to stress that different recipes to derive oxygen abundancesusing strong emission-line fluxes reveal very inconsistent results(Kewley & Ellison 2008). It is therefore of interest to expand thesample of sources where strong-line based abundances can be in-dependently tested using other methods (see also Pettini 2006; Ku-dritzki et al. 2012). Also, the use of different elements, e.g., Fe,Zn or Si, to infer absorption metallicities may introduce systematicoffsets when comparing galaxies within a heterogeneous sample.Combining the two complementary methods of studying themetal enrichment in galaxies provides important hints to under-standing how galaxies turn their gas into stars, as the absorptionlines directly probe the cold gas, and the properties of the starforming region can be probed directly from the emission lines, andsometimes also from the continuum. However, linking the absorp-tion characteristics of DLAs to the emission characteristics of thegalaxies causing the absorption has been a great challenge at highredshift, due to the faint nature of DLA galaxies and due to theirproximity to a very bright QSO. So far, only a few DLA systemswith confirmed emission counterparts have been established at red-shifts around and higher than two (Krogager et al. 2012).Here we present an analysis of the DLA at z = 2 . with log( N H I / cm − ) = 20 . ± . seen in the spectrum of the z = 2 . QSO 2222 − [ O II ] , [ O III ] , and Balmer emission lines. Fur-thermore, we combine the spectroscopic data with imaging fromWide Field Camera 3 (WFC3) onboard the Hubble Space Telescope ( HST ). The high resolution images from
HST makes it possible todetect the continuum emission from the galaxy directly, allowing usto characterize the properties and structure of the absorbing galaxy.This enables us to start bridging the gap between the population ofDLA galaxies and star-forming galaxies.The paper is organized as follows. In Sect. 2, we presentthe spectroscopic data from X-Shooter and the imaging data from
HST /WFC3. In Sect. 3, we derive the oxygen abundance from neb-ular emission lines, derive the gas phase metal abundances, char-acterize the morphology from the imaging data, and finally weobtain fluxes of the galaxy used for spectral energy distribution(SED) fitting. In Sect. 4, we compare the absorption and emissionproperties, and discuss and outline the implications of our work.Throughout this paper, we assume a standard Λ CDM cosmologywith H = 71 km s − Mpc − , Ω Λ = 0 . and Ω M = 0 . . QSO 2222 − × α in the UVB.The final 2D-spectra were then spatially aligned and addedby error-weighting, and we correct for slit-loss by calculating howmuch light gets dispersed outside the slit at the given seeing in eachexposure before co-adding all the observations. We have not cor-rected for telluric absorption, since those regions are not crucialfor our analysis. The effective seeing in the final flux and wave-length calibrated 2D-spectrum is 0 . (cid:48)(cid:48) . (cid:48)(cid:48)
8, as the averageseeing in UVB was 0 . (cid:48)(cid:48)
77 (measured at 480 nm).In Figure 1 the spectrum blue-ward of the QSO Ly α emissionline is shown, demonstrating the quality of the combined spectrum;note the significant flux below the Lyman limit at z = 2 . . HST / WFC3 imaging
The field was observed with the Wide Field Camera 3 on 2011 Nov10 (with the UVIS detector in the F606W filter) and on 2012 Sep 14and 15 (with the IR detector in the F105W and F160W filters). Theroll-angle of the telescope was set such that the galaxy counterpartof the DLA fell between the diffraction spikes of the Point SpreadFunction (PSF). The two observations with the IR detector weretaken using the
WFC3-IR-DITHER-BOX-MIN pattern providing an c (cid:13)000
WFC3-IR-DITHER-BOX-MIN pattern providing an c (cid:13)000 , 1–12 tellar population of a z = 2 . DLA galaxy Figure 1.
The final 1D spectrum from the UVB arm of the z = 2 . Q 2222 − z = 2 . is clearly seen and the damped Lyman-alpha absorption line is identified in the middle of the figure by the vertical dashed line. The Lyman-alphaemission line at the QSO redshift is seen in the right part of the figure. optimal 4-point sampling of the PSF. The UVIS observation wastaken using the WFC3-UVIS-DITHER-BOX pattern.The field was observed using a 4-point sub-pixel dither pat-tern, which allowed us to regain more spatial information. We haveused the software package multidrizzle to align and com-bine the images. By shifting and combining the images taken withsub-pixel offsets one achieves a better sampling of the PSF, whichin the case of the NIR observations is quite crucial as the PSF ispoorly sampled in the native 0 . (cid:48)(cid:48)
13 px − images. Furthermore, thedrizzle-algorithm allows us to reduce the pixel size when combin-ing the images providing sharper and better sampled images. Forthe combination in this work we have set the parameter pixfrac to . and used a final pixel scale of 0 . (cid:48)(cid:48)
06 px − for NIR and0 . (cid:48)(cid:48)
024 px − for UVIS. For a detailed description of the parame-ters in the software we refer the user to the multidrizzle usermanual. In order to detect the faint emission lines from the foregroundDamped Lyman- α Absorbing galaxy we needed to subtract theQSO continuum. We did this by modelling the spectral trace andthe spectral point spread function (SPSF) as a function of wave-length. The trace position and amplitude were fitted as a function ofwavelength in a small (84 ˚A) region around each emission line fromthe galaxy, e.g., [ O II ] , [ O III ] , and [ N II ] . In the same region aroundthe line we estimated the SPSF by averaging the observed spatialprofile along the spectral direction. The modelled 2D-spectrum ofthe QSO was then subtracted from the observed spectrum, and a1D-spectrum for each line was extracted with the optimal extrac-tion algorithm of Horne (1986). All extracted emission lines areshown in Figure 2. multidrizzle is a product of the Space Telescope Science Institute,which is operated by AURA for NASA. We were able to detect emission from Ly α , H α , H β , the [ O II ] λλ , doublet, the two [ O III ] lines at ˚A and ˚A, and [ N II ] λ . The line profiles for all the extractedlines were fitted with Gaussian profiles to measure the flux in theline. The parameters of the Gaussian were allowed to vary in all fitsexcept for [ O II ] and [ N II ] . The line-width and redshift of the two [ O II ] components were tied in the fit, because of a sky line that fallsright between the two components. For the very faint [ N II ] -line, were-binned the spectrum by a factor of two and fixed the redshift andline-width to the values of the nearby H α line. The errors on thefluxes were estimated by varying the line profile within the errorof each fit parameter 1,000 times. The uncertainty was then deter-mined from the 1 σ width of the resulting distribution of fluxes. Thefluxes and errors are listed in Table 1.All the extracted lines with Gaussian line fits are shown inFig. 2 along with the best fitting Gaussian profile. The grey shadedareas indicate telluric emission or absorption features. These re-gions were included in the fits to give the optimal estimate of un-certainties on the fit parameters. The Ly α line is shown in Fig. 3.Due to the asymmetric shape and the complex nature of the res-onant line, we determine the line flux simply by integrating theobserved line profile as opposed to fitting the line. The ratio of the H α and H β line fluxes provides us with informa-tion about the dust extinction in the system as we know what thisratio should be intrinsically, given the physical conditions of theemitting region. Assuming case B recombination with an electrontemperature of T e = 10 K and density of n e = 10 cm − , whichare standard assumptions in the literature for star forming regions,the line ratio has a value of . . Requiring that our measured lineratio, after reddening, and the intrinsic ratio be the same, we esti-mate the reddening: E ( B − V ) = 2 . k (H β ) − k (H α ) log (cid:18) (H α/ H β ) obs (H α/ H β ) (cid:19) , c (cid:13) , 1–12 J.-K. Krogager et al.
Table 1.
Measured emission line fluxesTransition Wavelength (1)
Flux (2)
FWHM (3) z Ly α ± [ O II ] ± ± β ± ± [ O III ] ± ± [ O III ] ± ± α ± ± [ N II ] ± (1) Transition rest frame wavelength in ˚ A . (2) Flux in units of − erg s − cm − , before reddening correction. (3) Line width at FWHM in units of km s − corrected for the instrumentalresolution of 45 km s − . Figure 2.
Emission lines extracted after the subtraction of the QSO contin-uum in the 2D spectrum. Each panel shows the 2D spectrum where the hor-izontal residuals from the QSO subtraction is seen ( top ) and the extracted1D spectrum of the line ( bottom ). The wavelengths are given in µ m andall flux-density units are − erg s − cm − ˚ A − . The red line in eachpanel indicates the best fitting Gaussian to the line profile. The grey filledareas indicate skylines or telluric absorption features. Figure 3.
The Ly α emission line extracted from the DLA trough in the X-shooter spectrum. The top panel shows the observed 2D spectrum and thepanel below shows the extracted 1D spectrum, and 1 σ uncertainty (greyshaded area). The dotted vertical line shows the systemic line centre. Thecommon asymmetric line shape of Ly α is seen in this case, where the blue-shifted part of the line is heavily absorbed. On top of the observed spectrumthe best fitting model of the emission profile is shown (red line) along withthe 68% confidence interval (red shaded area), see Sect. 3.7 for details. where k denotes the reddening law evaluated at the given wave-length and (H α/ H β ) indicates the intrinsic line ratio. We use theextinction law from Calzetti et al. (2000) to quantify the extinctionwhile adopting a R V value of 4.05. From the measured ratio in ourspectrum of (H α / H β ) obs = 3 . ± . we obtain the corre-sponding extinction of E ( B − V ) = 0.05 ± β line in the case of the Balmer line ratio. However,the effect in the K - and H -band, where we extract the Balmer lines,is minor ( ∼ %) compared to the uncertainty introduced by the skylines, especially at the position of the H β -line. We have inferred the metallicity of the system using three indepen-dent methods; two measures of the emission-line metallicity usingthe strong line ratios R and N2, respectively, and one measure ofthe absorption-line metallicity from Voigt-profile fitting. In the nextsections we present each determination in detail. calibration We determined the metallicity
12 + log (O/H) by use of thestrong line diagnostic, R (Pagel et al. 1979), defined as the ra-tio ( [ O II ] λλ , O III ] λ O III ] λ ) / H β .One complication of using this diagnostic is that the metallic-ity is double-valued for a given value of R (the so-called up-per and lower branches). However, due to our relatively highvalue of R we are in the region of the diagram where the twobranches are close to one another. The calibration by Kobulnicky c (cid:13)000
12 + log (O/H) by use of thestrong line diagnostic, R (Pagel et al. 1979), defined as the ra-tio ( [ O II ] λλ , O III ] λ O III ] λ ) / H β .One complication of using this diagnostic is that the metallic-ity is double-valued for a given value of R (the so-called up-per and lower branches). However, due to our relatively highvalue of R we are in the region of the diagram where the twobranches are close to one another. The calibration by Kobulnicky c (cid:13)000 , 1–12 tellar population of a z = 2 . DLA galaxy log( R ) + l o g ( O / H ) Figure 4.
Plot of the R strong line ratio using the calibration by Kobul-nicky & Kewley (2004) (two solid lines) with model uncertainty of . dexindicated by the light grey shaded area around the curves. The orangefilled area between the two solid lines indicates the constraint from ourmeasurement of R . The dashed line indicates the inferred metallicity of
12 + log(O / H) = 8 . and the red shaded area around it indicates the σ uncertainty of . dex. The dot-dashed line shows the Solar oxygenabundance of
12 + log(O / H) (cid:12) = 8 . (Asplund et al. 2009). & Kewley (2004) that we use depends on the ionization param-eter, q , which can be determined by using the line ratio O =[ O III ] λ / [ O II ] λ , but requires prior knowledge of themetallicity. Given the two measured quantities, R and O , wecan solve for both metallicity and q at the same time by iteratingthe two expressions given an initial guess of metallicity.From our spectrum we measure log(R ) = 1 . ± . andlog(O ) = 0 . ± . . These line ratios are corrected for redden-ing as indicated by the Balmer decrement. However, in Sect. 3.8 weinfer E ( B − V ) = 0 . ± . , which is the value that we haveadopted for the correction here. We perform the iterative calcula-tion of metallicity 500 times while varying the two line-ratios (R and O ) within their respective errors. The metallicity and uncer-tainty is then determined as the median and standard deviation ofthe resulting distribution This in turn yields an inferred metallic-ity of
12 + log(O / H) = 8 . ± . , see Fig. 4. We furthermorenote that the result does not depend on the initial guess, and thatthe model uncertainty introduced by the error on the ionization pa-rameter, q , is negligible. The uncertainty of . dex includes thecontribution from the scatter in the R calibration which we con-servatively set to . dex. Our detection of [ N II ] enables us to infer the metallicity indepen-dently from the N2 index ( [ N II ] / H α ) as this line ratio has beenfound to correlate with metallicity (Pettini & Pagel 2004). The cal-ibration from these authors give
12 + log(O / H) = 8 .
90 + 0 . · log(N2) with a scatter of ± α and [ N II ] means that we do not need to worry aboutreddening effects in the N2 ratio. The measured line fluxes yield aline ratio of N2 = 0 . ± . , giving an inferred metallicity of
12 + log(O / H) = 8 . ± . .P´eroux et al. (2012) infer an upper limit of
12 + log(O / H) < . from their non-detection of [N II ]. Our measurement is slightlyhigher, but still consistent with theirs within the uncertainties.Since the two measures of the emission metallicity, R andN2, are independent we can combine both measures to obtain amore precise estimate of
12 + log(O / H) = 8 . ± . . Adoptingthe Solar abundance presented by Asplund et al. (2009) this givesus [O / H] = − . ± . . In the following section we present detailed measurements of theabsorption-line abundances detected in the DLA system. Fynboet al. (2010) present an analysis of the metal absorption-line sys-tem of this DLA using a three-component Voigt-profile fit. Theyderive fairly high metallicities from [ Zn/H ] = − . ± . , [ Si/H ] = − . ± . , and [ S/H ] = − . ± . , and see evi-dence for dust depletion from [ Mn/H ] = − . ± . , [ Ni/H ] = − . ± . , and [ Fe/H ] = − . ± . . Since our data havemuch better resolution (recall values in Section 2.1 compared to R = 4700 , , in the Fynbo et al. study for the UVB, VISand NIR arms, respectively) and much higher Signal-to-Noise Ra-tio (SNR ∼ ∼
50) we can alleviate theeffect of hidden saturation, thereby further constraining the metalabundances.In order to properly decompose the absorption profiles in dif-ferent velocity components we have selected unsaturated, but well-defined low-ionization lines to infer the metallicity of the absorb-ing gas (see Fig. 5). Furthermore, we have access to a few high-ionization lines: C IV λλ VI λλ VI λ II λ FitLyman in MIDAS (Fontana & Ballester 1995) while tying thebroadening parameter, b , and the redshift for each velocity compo-nent for all lines. The results for the individual components of thefit are summarized in Table 2. We derive the total column densitiesfrom the fit by adding the column densities for all components ofeach species. These and the derived metallicities are listed in Ta-ble 3. For the Ti II λλ σ ) of [ Ti/H ] (cid:54) − . , as the line is shallow and the continuum level inthis part of the spectrum is uncertain. The abundance of Zn is prob-ably overestimated due to blending with the weak Mg I λ I λ I to be ∼ IV whereas we used two components for themore noisy O VI and S VI lines. The results from the fit are listed inthe bottom part of Table 2. Our possible detection of S VI is quotedas an upper limit, and the first component of C IV is quoted as alower limit since the line is saturated. The velocity of the maincomponent of C IV is fully consistent with the systemic velocityfrom the emission lines within the uncertainty of 9 km s − . Theoxygen and sulfur lines have slightly larger offsets relative to theemission redshift of +20 km s − and +55 km s − , respectively. c (cid:13) , 1–12 J.-K. Krogager et al.
Table 2.
Ionic column densities for individual absorption components oflow-ionization lines (top) and high-ionization lines (bottom). The errorsquoted below only include the formal errors from
FitLyman .Component log( N/ cm − ) b km s − Si II λ v = − km s − ± ± v = 11 km s − ± ± v = 73 km s − ± ± v = 118 km s − ± ± v = 174 km s − ± ± II λ v = − km s − ± ± v = 11 km s − ± ± v = 73 km s − ± ± v = 118 km s − ± ± v = 174 km s − ± ± II λλλ v = − km s − ± ± v = 11 km s − ± ± v = 73 km s − ± ± v = 118 km s − ± ± v = 174 km s − ± ± II λλλ v = − km s − ± ± v = 11 km s − ± ± v = 73 km s − ± ± v = 118 km s − ± ± v = 174 km s − ± ± II λλλ v = − km s − ± ± v = 11 km s − ± ± v = 73 km s − ± ± v = 118 km s − ± ± v = 174 km s − ± ± II λ v = − km s − ± ± v = 11 km s − ± ± v = 73 km s − ± ± v = 118 km s − ± ± v = 174 km s − ± ± IV λλ v = 5 km s − > ± v = 134 km s − ± ± v = 197 km s − ± ± VI λλ v = 55 km s − ± ± v = 242 km s − ± ± VI λ v = 20 km s − (cid:54) ± ± v = 210 km s − (cid:54) ± ± The velocity, v , indicated at each component shows the relativevelocity with respect to the emission redshift z = 2 . . Lower limit due to saturation. Upper limit due to possible blending.
Figure 5.
Results of Voigt-profile fitting to the metal absorption lines. Thezero point of the velocity scale is fixed to the redshift from the emissionlines, z em = 2 . with an uncertainty of 9 km s − . The figure shows thenormalized spectra around each fitted line. The part of the spectrum plottedin light grey without error-bars in the panels of Mn II λ II λ II λ VI λλ (cid:13)000
Results of Voigt-profile fitting to the metal absorption lines. Thezero point of the velocity scale is fixed to the redshift from the emissionlines, z em = 2 . with an uncertainty of 9 km s − . The figure shows thenormalized spectra around each fitted line. The part of the spectrum plottedin light grey without error-bars in the panels of Mn II λ II λ II λ VI λλ (cid:13)000 , 1–12 tellar population of a z = 2 . DLA galaxy Table 3.
Total column densities and metallicities for the low-ionization linesin the z = 2 . absorbing system.Element log( N X / cm − ) [X/H] log( N (cid:12) / cm − ) H 20.65 ± ± − ± ± − ± ± − ± ± − ± ± − ± ± − ± Our data exhibit a number of dips at the expected position of H lines at z abs = 2 . , which suggest a detection of molecular hy-drogen at the level of log( N H / cm − ) ∼ . However, assert-ing the presence of H and deriving accurate column densities orlimits requires much higher spectral resolution data than what wehave available here (see, e.g., Ledoux et al. 2003; Noterdaeme et al.2008). This is needed in order to ( i ) deblend the possible H linesfrom the Lyman- α forest, ( ii ) estimate the true continuum of theQSO from unabsorbed spectral regions, and ( iii ) resolve the ve-locity structure of the system. Fynbo et al. (2011) reported thedetection of H lines at a resolving power of R = 6400 in the z abs = 2 . DLA towards SDSS J 0918 + log N ( H ) ∼ − ). The present tentative detection of H lines towards SDSS J 2222 − R (cid:38) , ). HST images
We started out by modelling the Point Spread Function in each im-age using the software T
INY T IM to create sub-sampled PSFs sim-ulated for each of the individual frames at each position in the four-point dither pattern. We preferred this approach over using stellarsources as the T INY T IM PSFs are more sensitive to the outer re-gions of the PSF, where the stellar sources are dominated by noisein the sky background.After resampling the modelled PSFs to the native pixel size andconvolving with the filter-specific Charge Diffusion Kernel, these”raw” PSF images were drizzled together in the same way as theactual data to replicate the effect of multidrizzle on the PSFshape. We were not able to simulate the PSFs for the infrared im-ages properly because the images were mildly saturated. We there-fore chose to use stellar PSFs generated by median combination ofstars in the field. Since there are only a handful isolated stars in thefield of view with signal-to-noise ratios similar to that of the QSO,we used a few stars with slightly lower SNR in order to reduce thebackground noise in the PSF.We used the software G
ALFIT (Peng et al. 2002) to subtractthe QSO and to characterize the absorbing galaxy. Using the mod-elled PSF for the F606W image we first subtracted the QSO bysimply modelling it as a point source while modelling a constantsky background. We then located the nearby galaxy in the residu-als. Hereafter, we re-did the fit, this time simultaneously fitting theQSO, the background, and the galaxy using a S´ersic surface bright- ness profile: Σ( r ) = Σ e exp (cid:34) − κ (cid:32)(cid:18) rr e (cid:19) /n − (cid:33)(cid:35) , where Σ e is the surface brightness at the effective radius, r e , de-fined as the radius enclosing half the flux. The S´ersic index, n ,determines the concentration of the profile, with high n profileshaving steeper inner slopes and larger extended wings. The oppo-site is the case for low n . The parameter κ is linked to n to en-sure that half of the light is enclosed within the effective radius.The software G ALFIT uses a 2D S´ersic profile allowing for ellip-tical isophotes. The output from fitting the S´ersic profile is givenin terms of the total flux from the integrated profile, the effectivesemi-major axis, a e , the S´ersic index, n , the axis ratio of semi-major and -minor axes, b/a , and the angle of a e with respect tothe image-axes. We fitted the galaxy allowing all parameters tovary freely. This resulted in the following best-fitting parameters: mag AB = 24 . ± . , n = 0 . ± . , a e = 5 . ± . px, b/a = 0 . ± . , and PA = − . ± . o in the F606Wimage.Due to the broader PSF in the NIR images the QSO light isspatially overlapping with the galaxy. We therefore fixed the struc-tural parameters in the S´ersic fit of the galaxy to those of the well-constrained F606W fit. We note that simply locking all structuralparameters of the S´ersic profile in the NIR images might not givethe most accurate description of the galaxy, as parameters such assize and n depend on wavelength and thus yield different resultswhen analysed in different wavelength band-passes (Kelvin et al.2012). However, in order to obtain the most reliable fluxes and toget the fit to converge we had to keep the variables fixed. In orderto estimate how robust our obtained fluxes are with respect to theparameters that were held fixed, we varied the S´ersic parameterswithin their errors (as given by the fit to F606W) and re-did the fitfor each new set of profile parameters. The uncertainty on the fluxwas very minor (0.03 dex) compared to the large uncertainty causedby the PSF subtraction ( ∼ . (cid:48)(cid:48)
74 corresponding to a pro-jected distance of 6.3 kpc at z = 2 . , and the angle between themajor axis of the galaxy and the line connecting the QSO to thecentral region of the galaxy is ± o . α emission modeling We observe a typical, double-peaked Ly α emission line, with astrong component redwards of line centre and a less prominent bluecomponent (Fig. 3). The large difference between the two peaks in-dicates that the Ly α photons escape through an expanding medium,and the purpose of this section is to investigate what constraints wecan put on the outflow velocity, V out . Qualitatively we can see thatthe red wing extends to several hundreds of km s − . However, inorder to get a more accurate description of the emission profile we c (cid:13) , 1–12 J.-K. Krogager et al.
Table 4.
Results from G
ALFIT analysisFILTER mag [ AB ] a e / kpc n b/a F606W 24.29 ± . ± .
06 0 . ± .
15 0 . ± . F105W 24.51 ± . ∗ . ∗ . ∗ F160W 23.53 ± . ∗ . ∗ . ∗∗ Parameters that were fixed in the corresponding fit.
Figure 6.
Data images ( left ) and PSF-subtracted residuals ( right ) fromG
ALFIT in all three filters. All images are aligned North up and East left.The galaxy causing the absorption is clearly visible in the F606W imageeven before PSF subtraction. The same position is shown by the red circlein the IR images. have constructed a semi-realistic model of the system and run Ly α radiative transfer (RT) through it, to find the best-fitting spectrum.The RT is conducted using the code M O C A L A TA (Laursen et al.2009a,b), while the galaxy model is similar to the one described in(Laursen et al. 2013). In short, the galaxy is modelled as a sphereof multiphase gas, with warm, neutral clouds floating in a hot, pre-dominantly ionized medium. A similar procedure was used in No-terdaeme et al. (2012), although here we employ a more rigorousapproach. A high number of parameters dictate the outcome of such asimulation. Luckily, our observations offer excellent constraints onmany of these: The effective radius, r e = 1 . kpc, used to modelthe size of the emitting region, the metallicity Z (cid:39) . Z (cid:12) (weassume that the amount of metals that condense to dust is simi-lar to the local Universe (Zafar & Watson 2013)), and an averageof 130 km s − from the measured emission-line widths are usedas a proxy for the intrinsic Ly α line width. From the widths ofthe low-ionization absorption lines, we use a velocity dispersionof the clouds of 115 km s − . Finally, we infer an intrinsic equiv-alent width of 150 ˚A, in accordance with the F606W magnitude.For details on the rest of the parameters, e.g., cloud size distribu-tion, temperatures and densities of the two phases, etc., see Laursenet al. (2013).We set r gal = 10 kpc, but note that the exact size of the sys-tem is not important, rather the total column density (cid:104) N H I (cid:105) from thecentre and out, averaged over all directions, determines the shape ofthe spectrum. This leaves us with two unknown parameters, (cid:104) N H I (cid:105) and V out , where (cid:104) N H I (cid:105) is dominated by the number of clouds ( N cl ).We first run a rough fit to the spectrum, providing us with informa-tion about the initial conditions for the system: N cl = 10 and V out = 150 km s − , and consequently run a grid of 11 ×
11 modelswith N cl ∈ [10 . , . ] and V out ∈ [100 , km s − .Instead of doing a regular χ minimization of the pixel-wisedifference between model and spectrum, we compare the observedand the simulated spectra using the following four observables: Thepeak separation in ˚A; the width of the red peak in ˚A; the ratioof the integrated flux in the two peaks; and the ratio of the peakheights. The best fit is defined as the model which minimizes allof the four above mentioned criteria simultaneously, given the con-straints on metallicity, emission line velocity dispersion, Ly-alphaflux and structure of the emitting region. From the best fit, we findan outflow velocity of V out = 160 +20 − km s − . Here, the confi-dence intervals are given by the range of models, for which all fourestimators lie within the 68% confidence interval of those observedin our spectrum. In the best-fitting model, the average column den-sity and number of clouds intercepted by a sightline towards theQSO at a distance of 6.3 kpc are log( N H I / cm − ) = 20 . +0 . − . and n c = 2 ± , respectively. This is somewhat lower although notinconsistent with the measured value of . ± . and the factthat five absorption profiles were used in Sect. 3.5.The escape fraction of Ly α photons ( f esc ∼ %) is higherthan what is found when comparing the total Ly α -to-H α ratio, andwe find that a SFR of only 6.0 M (cid:12) yr − is needed to match theobserved spectrum. However, the galaxy was modelled as a spher-ically symmetric system, and the galaxy appears disc-like from theG ALFIT analysis in Sec. 3.6. For such a disc-like system, the escapefraction will be significantly lower when observed edge-on, whichindeed seems to be the case here.
The F606W flux of the galaxy as measured from the HST data cor-responds to a rest frame wavelength of 1775 ˚A at the redshift ofthe galaxy. Based on the AB magnitude obtained from our G AL - FIT analysis (see Table 4) we get a luminosity after correcting forGalactic extinction. We use the emission redshift, z em = 2 . ,to compute the luminosity distance. This gives a luminosity of L ν = F ν π d L (1 + z ) − = 9 . × erg s − Hz − . Weuse the extinction correction factor for the F606W band, ∆mag = c (cid:13) , 1–12 tellar population of a z = 2 . DLA galaxy − . mag, from the NASA/IPAC Extragalactic Database (NED).This luminosity corresponds to a star formation rate of SFR UV =13 . ± . M (cid:12) yr − using the relation from Kennicutt (1998).For the H α emission line we take the observed line flux (seetable 1), which corresponds to a luminosity of L H α = 2 . ± . × erg s − . Converting the luminosity into SFR againusing Kennicutt (1998) gives SFR H α = 18 . ± . (cid:12) yr − .The discrepancy between the two inferred star formation ratesindicates some degree of dust extinction. To quantify the amount ofextinction we assume that the two measures should yield the samevalue, when corrected for dust reddening. The correction factor tothe SFR can be expressed as: SFR int = SFR obs · . · E ( B − V ) k ( λ ) . Requiring that the two SFRs from UV and H α be equal we arriveat the following expression: E ( B − V ) = 2 . k (UV) − k (H α ) · log (cid:18) SFR H α SFR UV (cid:19) , where k (UV) and k (H α ) denote the Calzetti et al. (2000) extinc-tion curve evaluated at the UV rest frame wavelength, 1755 ˚A,and at the rest frame wavelength of H α , respectively. From thisrelation we derive a colour excess of E ( B − V ) = 0 . ± . consistent with the previously mentioned measure from the Balmerdecrement (sect. 3.3). The extinction corrected star formation rateis SFR = 22 . ± . (cid:12) yr − with an assumed Salpeter initialmass function (IMF). If we convert this to the Chabrier IMF we get SFR ch = 12 . ± . (cid:12) yr − . This measurement agrees very wellwith the results of Fynbo et al. (2010) and P´eroux et al. (2012) whofind SFRs of SFR > M (cid:12) yr − and SFR = 17 . ± . M (cid:12) yr − ,respectively.We note that the two proxies for star formation do not tracestar formation on the same physical time-scales, and therefore donot necessarily have to yield the same measured quantity. The H α -line primarily traces the ongoing star formation responsible for ion-izing the H II -regions, whereas the UV continuum traces previousstar formation as well, linked to the O and B stars already in place.However, the young, UV-bright stars may also contribute to pho-toionization of the gas, and the quantification of the difference de-pends on assumptions about IMF, star formation history, and thedistribution of dust. The detailed, exact modelling of all these fac-tors is beyond the scope of our simple extinction estimate, we havetherefore chosen to neglect this effect in the above analysis.The star formation rate can also be determined directly fromthe absorption lines associated with the DLA if the C II ∗ absorptionline is available (the method is described in Wolfe et al. 2003). Un-fortunately the line is blended, so we cannot put any firm constrainton the star formation using this method. We fit the three broad-band photometric points from the
HST imag-ing mentioned in Sect. 3.6 to obtain estimates of stellar mass, ageand star formation rate. But first, we apply a correction for Galacticextinction available from the NED online database . To simplifythe fit we subtract the emission-line fluxes from the broad-bandfluxes and fit the continuum only, instead of fitting both continuum http://ned.ipac.caltech.edu/ Wavelength / Å A B m a g n i t u d e F606W F105W F160W
Figure 7.
Best fitting template (in grey) to our three broad-band photomet-ric points (black, squares). The error-bars in y-direction indicate the 1 σ -uncertainty on the fluxes. In the bottom of the figure the transmission curvesfor each filter are shown, and the blue, round points indicate the model mag-nitudes calculated in each filter passband. The median age and mass of thefit is t = 98 Myr and M (cid:63) = 2 . · M (cid:12) , respectively. and emission lines simultaneously. The only filter which is influ-enced by strong emission lines is the F160W band, which containsflux from H β and the two [ O III ] lines. In order to subtract the emis-sion lines, we assume a flat continuum spectrum, since we only de-tect the emission lines, and subtract the integrated line fluxes fromthe total flux in the observed band, weighted by the filter curve. Wefind that the emission lines contribute 33% of the total flux, and weinfer a corrected F160W magnitude of m = 23 . ± . .Our fitting code uses the stellar population templates fromBruzual & Charlot (2003) convolved with a large Monte Carlo li-brary of star formation histories (exponential plus random bursts)assuming a Chabrier (2003) IMF. Dust is added following the two-component model of Charlot & Fall (2000), with the parametersbeing the total optical depth, τ v , and the fraction of dust contributedby the ISM, µ . We restrict the range of metallicities of the mod-els to be consistent with our measurement from the emission lines log( Z/Z (cid:12) ) = − . ± . . We then adopt a Bayesian approachby comparing the observed photometry to the one predicted by allthe models in the library, and we construct the probability densityfunctions of stellar mass, mean light-weighted stellar age, and starformation rate. Taking the mean and 16 th and 84 th percentiles ofthe PDF we obtain a stellar mass of M (cid:63) = 2 . +1 . − . × M (cid:12) ,age of the galaxy of t = 98 +113 − Myr, and a star formation rateof
SFR = 8 . +4 . − . M (cid:12) yr − . We also get an estimate of the dustextinction from the fit by comparing the intrinsic template with thebest fitting reddened template. From this we infer A V = 0 . +0 . − . ,which corresponds to E ( B − V ) = 0 . +0 . − . .The SFR from the SED fit agrees well with the one inferredin Sect. 3.8 within σ , and also the median value of the dust isconsistent with what we find in Sect. 3.3 and 3.8. The photometryand best fitting template are shown in Fig. 7. For details on the prior distribution of the SFH and dust parameters seeSalim et al. (2005)c (cid:13) , 1–12 J.-K. Krogager et al.
We have presented a measurement of the gas-phase metallicity inthe H II regions of the emission counterpart of the DLA towards thequasar Q 2222 − andN2 diagnostics expressed in Solar units is [O / H] = − . ± . .From the absorbing gas located 6.3 kpc away, we find a metallic-ity in the neutral ISM from sulfur of [S / H] = − . ± . , thusslightly lower than the metallicity inferred from the central emit-ting region. Sulfur is not depleted onto dust, and thereby traces theoverall metallicity very well. We find a consistent metallicity from [Si / H] = − . ± . and [Zn / H] = − . ± . (note that [Zn / H] is probably overestimated by 0.1 dex). Neither Zn nor Siare significantly depleted onto dust and hence provide a good mea-surement of the gas-phase metallicity (Meyer & Roth 1990; Pettiniet al. 1997). The observed metallicities of DLAs at these redshiftsrange from [M / H] ≈ − . to [M / H] ≈ − . with an averagemetallicity weighted by N H I of − . ± . (Rafelski et al. 2012).The DLA in our study is thus amongst the most metal-rich DLAsat this redshift.From the elements Mn, Fe and Ni we clearly see that dustdepletion is in fact at play, the metallicities are [Fe / H] = − . , [Ni / H] = − . , and [Mn / H] = − . , indicating that the refrac-tory elements are, to some degree, removed from the gas-phase,which is to be expected in a high-metallicity system as this partic-ular one (Ledoux et al. 2003). Also, we report a tentative detectionof molecular hydrogen, see Sect. 3.5.1.Bowen et al. (2005) find similar results regarding gas phasemetallicities for a low-redshift galaxy, where a quasar intersects thegalaxy 3 kpc from a star forming region. The authors find consistentmetallicities based on the emission region and the absorbing, neu-tral gas, respectively. For a compilation of systems with emissionand absorption based metallicities, see P´eroux et al. (2012).The presence of highly enriched material 6 kpc above (almostperpendicular to) the galactic plane of this galaxy with nearly thesame metallicity as the star forming regions within the galaxy in-dicates that metal-rich material has been expelled from the galaxyinto the halo. We see independent evidence for outflowing gas fromthe Ly α emission line with a velocity of 160 km s − , see Sect. 3.7for details. At this velocity we estimate that it would take of theorder of 40 Myr for this enriched material to reach a distance of6.3kpc from the galaxy plane. Given this relatively short time-scaleit seems reasonable that the two metallicities are similar, since nolarge amount of enrichment has had time to occur after the expul-sion of the outflowing gas, which is expected to mix with lowermetallicity gas further out, lowering the observed metallicity in theabsorbing gas. This scenario can be compared to the wind observedin the nearby galaxy M82, where neutral gas and molecules forma filamentary structure in the outflow extending to great distancesfrom the disc (Veilleux et al. 2009; Melioli et al. 2013).A recent study by Bregman et al. (2013) shows evidence fora similar scenario, where near-Solar metallicity gas of a nearby,edge-on disc galaxy has been detected 5 kpc above the disc witha neutral hydrogen column density of . × cm − . However,as these authors study a local galaxy a direct comparison to highredshift might not be fully valid. We can use our information about the size of the galaxy and thekinematics, as probed by the emission lines (see FWHM measures in Table 1), to get an estimate of the dynamical mass of the sys-tem. We follow the method described in Rhoads et al. (2013) toestimate the dynamical mass given the measured size and velocitydispersion: M dyn ≈ σ a e G sin ( i ) , where i denotes the inclination of the system with i = 90 o being edge-on. In order to estimate the velocity dispersion of thesystem we use the FWHM of the emission lines as a probe of theintegrated gas-kinematics of the system. We then take a weightedmean of all the measured line-widths, and correct the FWHM forthe instrumental resolution (45 km s − ). This gives a measure thevelocity dispersion of σ = 49 . ± . km s − , and we get the sizefrom the G ALFIT analysis in physical units: a e = 1 . ± . kpc(see Table 4).From the G ALFIT analysis we infer an axis ratio of the galaxyof b/a = 0 . . The system may be described as reasonably disc-like, given the elongated shape, and the fact that we see a value ofS´ersic n very close to 1. Using the results of Haynes & Giovanelli(1984) that disc galaxies on average have axis ratios around . − . , we conclude that the galaxy in our study is very close to edge-on, even when assuming the lowest intrinsic value of . . We thusadopt a value of i = 90 o and use the fitted half-light semi-majoraxis for our estimate of the dynamical mass of the system: M dyn ≈ . × M (cid:12) . This estimate should only be considered a roughapproximation (valid within a factor of ∼ ) as we have assumedthe system to be in virial equilibrium, which it may not be. From our SED fit to the broad-band imaging data, we obtain a stel-lar mass of M (cid:63) = 2 . × M (cid:12) . We can use this measurementto test the recently proposed evolving mass-metallicity relation forDLA systems (Ledoux et al. 2006, see also Møller et al. (2004,2013); Neeleman et al. (2013)). Using the relation in Møller et al.(2013), which relates emission metallicity and redshift to stellarmass, with our direct measurement of the emission metallicity, wefind a stellar mass of M (cid:63) = 6 +9 − × M (cid:12) . Though the scatterin their relation is substantial ( ∼ . dex), the agreement betweenthe relation and our best fit stellar mass from the SED fit is striking.Moreover, the mass is in very good agreement with the medianstellar mass of Lyman Break Galaxies (LBGs) at redshift z = 1 − (Erb et al. 2006). Hathi et al. (2013) find M (cid:63), LBG = 2 × M (cid:12) ,and further characterize the general LBG population in terms ofmedian age ( ∼
125 Myr), median SFR ( ∼
15 M (cid:12) yr − ), and me-dian dust extinction ( E ( B − V ) ≈ z = 2 − (Fynbo et al. 2003; Rauch et al. 2008). Never-theless, it is important to remember that the DLA in question wasspecifically chosen to have high metallicity, and is of unusuallyhigh metallicity; it is an order of magnitude higher than the me-dian metallicity of DLAs at z = 2 − , [ M/H ] (cid:39) − . or 1/30 ofSolar (Noterdaeme et al. 2008; Rafelski et al. 2012). The proposedmass-metallicity relation for DLA galaxies (Ledoux et al. 2006), c (cid:13)000
15 M (cid:12) yr − ), and me-dian dust extinction ( E ( B − V ) ≈ z = 2 − (Fynbo et al. 2003; Rauch et al. 2008). Never-theless, it is important to remember that the DLA in question wasspecifically chosen to have high metallicity, and is of unusuallyhigh metallicity; it is an order of magnitude higher than the me-dian metallicity of DLAs at z = 2 − , [ M/H ] (cid:39) − . or 1/30 ofSolar (Noterdaeme et al. 2008; Rafelski et al. 2012). The proposedmass-metallicity relation for DLA galaxies (Ledoux et al. 2006), c (cid:13)000 , 1–12 tellar population of a z = 2 . DLA galaxy then supports the original suggestion by Fynbo, Møller, & Warren(1999) that most DLA galaxy counterparts are too faint to be iden-tified via their stellar or nebular emission.We are further able to infer the expected metallicity from thestellar mass and SFR using the Fundamental Metallicity Relationfrom the work of Mannucci et al. (2010). Given the expressionfrom these authors we find an oxygen abundance in the range from [O / H] = − . to [O / H] = − . , depending on the fitting func-tion assumed. Since the SFR of our target is slightly outside therange over which the relation is derived there will be uncertaintyrelated to the extrapolation. The metallicity is, however, still in per-fect agreement with our measurements. This indicates that the DLAin our study follows the same relation as other galaxies studied atboth lower and higher redshifts, strengthening the link between thisDLA galaxy and the general population of star-forming galaxies. The Tully-Fisher Relation
We now turn to look at how this galaxy is located on the stellar-mass Tully-Fisher relation (M (cid:63) -TFR) to test our assumption thatthe system is disc-like and relaxed. The Tully-Fisher relation, orig-inally stated in terms of luminosity and velocity (Tully & Fisher1977), can also be presented in terms of stellar mass (which cor-relates with luminosity) and velocity. The M (cid:63) -TFR was studied indetail by Kassin et al. (2007) who gave their best fit to the data as: log( S . ) = 1 . ± .
03 + 0 . ± . · log (cid:18) M (cid:63) M (cid:12) (cid:19) , where S . is defined by the authors as S K ≡ K V + σ g , with K = 0 . . Using the measured velocity dispersion as a proxy for S . (see a discussion of this in Rhoads et al. 2013) we find that theinferred stellar mass is M (cid:63) = 2 . +1 . − . × M (cid:12) . This is in verygood agreement with our previously mentioned mass estimates,including our rough estimate of the dynamical mass.The stellar (and dynamical) mass inferred from the emissionline widths is subject to uncertainties caused by the fact that we donot know the detailed structure of the velocity field. The emissionlines are most certainly influenced by turbulence in the gas, whichwe cannot quantify. This would overestimate our line-widths andthereby our TF-based stellar mass and dynamical mass estimates.Also, we required that the system be in virial equilibrium; however,we observe gas at large galactocentric radii, almost perpendicular tothe disc, with similar metallicity as the line emitting region, indicat-ing outflowing gas from the central parts. This has an impact on ourcalculation of the dynamical mass, and our ability to use the line-widths as a tracer of the ordered rotation of the system. Moreover, anon-negligible gas mass is expected in a young ( ∼ Myr), star-forming galaxy. This mass would not be accounted for in the stellarmass estimate from the SED fit, but would contribute to the dynam-ical mass, thus increasing the observed velocity dispersion. We es-timate the gas mass from the star formation density (see Kennicutt1998) using the half-light radius and the axis-ratio of the galaxy.We infer a gas mass of M gas = 1 +3 − . × M (cid:12) . This estimate isa very rough approximation given the large scatter ( ∼ . dex) inthe relation, and is therefore not to be trusted as a true value of theamount of gas. It does, however, indicate that a gas mass of roughlyhalf the stellar mass is present. We note that the inferred dynami-cal mass is consistent with our best-fit stellar mass within σ , anda significant gas mass is therefore not required in order to recon-cile the two mass estimates, and within the (large) uncertainties, all three mass estimates agree well. The compact nature of the galaxyalso means that the kinematics of the emission lines only probe theinnermost region, and are therefore mostly sensitive to the stellar mass concentrated in the central region, and not the gas in the outerparts.Studies of the TFR at higher redshifts find that the relation isoffset to lower stellar masses for a given velocity. We see a similarthough not statistically significant trend in our data. Cresci et al.(2009) find that LBGs at z ≈ are offset by 0.4 dex comparedto the local TFR, and they find that the relation has low scatter.Gnerucci et al. (2011) find similar results in their sample of z ≈ galaxies with an offset up to 1 dex, however, the scatter in theirdata is very large. This may indicate that the TFR has not yet beenestablished at these redshifts, and that the galaxies are influencedheavily by random motions. We have presented our analysis of a high-redshift galaxy selectedfrom its neutral hydrogen absorption seen in the spectrum of abackground quasar. We have presented the extracted emission linesfrom the galaxy counterpart of the absorption, and combined thesewith our detailed absorption-line study to probe the metallicitiesseen in the two phases. We find that the two metallicities are sim-ilar, but the absorbing gas has a slightly lower metallicity than theemitting gas. We use
HST imaging to constrain the stellar popula-tion; our data are consistent with the picture in which the galaxycausing the Ly α absorption is a young, small ( ∼ kpc), disc-likesystem, with a not fully ordered (proto-) disc structure. Moreover,we see evidence for a so-called galactic fountain , where enrichedgas gets blown out from the star-forming regions, forms the neutralhydrogen absorption that we see ∼ kpc above the galactic plane,and in the end may settle back onto the disc.This galaxy demonstrates exactly how star-forming galaxiesat high redshift may overlap with the population of the most metal-rich DLAs. However, the very faint nature of damped Lyman- α absorbing galaxies renders the majority of these almost impossibleto detect. The few exceptions, such as the case reported here, thusoffer a rare glimpse into this elusive galaxy population. ACKNOWLEDGEMENTS
We thank the anonymous referee for the a nice and constructive re-port. The Dark Cosmology Centre is funded by the DNRF. JPUFand PL acknowledge support form the ERC-StG grant EGGS-278202. JK acknowledges support from an ESO Studentship. CPhas beneted from support of the Agence Nationale de la Recherchewith reference ANR-08-BLAN-0316-01. This research has madeuse of the NASA/IPAC Extragalactic Database (NED) which isoperated by the Jet Propulsion Laboratory, California Institute ofTechnology, under contract with the National Aeronautics andSpace Administration. The RT simulations were conducted on thefacilities provided by the Danish Center for Scientific Computing.A.G. acknowledges support from the EU FP7/2007-2013 undergrant agreement n. 267251 AstroFIt.
REFERENCES
Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P., 2009,ARA&A, 47, 481 c (cid:13) , 1–12 J.-K. Krogager et al.
Bowen, D. V., Jenkins, E. B., Pettini, M., & Tripp, T. M., 2005,ApJ, 635, 880Bregman, J. N., Miller, E. D., Seitzer, P., Cowley, C. R., & Miller,M. J., 2013, ApJ, 766, 57Bruzual, G. & Charlot, S., 2003, MNRAS, 344, 1000Caffau, E., Bonifacio, P., Franc¸ois, P., et al., 2011, A&A, 534, A4Calzetti, D., Armus, L., Bohlin, R. C., Kinney, A. L., Koornneef,J., & Storchi-Bergmann, T., 2000, ApJ, 533, 682Chabrier, G., 2003, Publications of the Astronomical Society ofthe Pacific, 115, 763Charlot, S. & Fall, S. M., 2000, ApJ, 539, 718Cresci, G., Hicks, E. K. S., Genzel, R., et al., 2009, ApJ, 697, 115Erb, D. K., Shapley, A. E., Pettini, M., Steidel, C. C., Reddy,N. A., & Adelberger, K. L., 2006, ApJ, 644, 813Fontana, A. & Ballester, P., 1995, The Messenger, 80, 37Frebel, A., Johnson, J. L., & Bromm, V., 2007, MNRAS, 380, L40Frebel, A., Simon, J. D., Geha, M., & Willman, B., 2010, ApJ,708, 560Fynbo, J. P. U., Laursen, P., Ledoux, C., et al., 2010, MNRAS,408, 2128Fynbo, J. P. U., Ledoux, C., Møller, P., Thomsen, B., & Burud, I.,2003, A&A, 407, 147Fynbo, J. P. U., Ledoux, C., Noterdaeme, P., et al., 2011, MNRAS,413, 2481Fynbo, J. P. U., Starling, R. L. C., Ledoux, C., et al., 2006, A&A,451, L47Fynbo, J. U., Møller, P., & Warren, S. J., 1999, MNRAS, 305, 849Gnerucci, A., Marconi, A., Cresci, G., et al., 2011, A&A, 528,A88Hathi, N. P., Cohen, S. H., Ryan, Jr., R. E., et al., 2013, ApJ, 765,88Haynes, M. P. & Giovanelli, R., 1984, AJ, 89, 758Holmberg, J., Nordstr¨om, B., & Andersen, J., 2007, A&A, 475,519Horne, K., 1986, PASP, 98, 609Kassin, S. A., Weiner, B. J., Faber, S. M., et al., 2007, ApJ, 660,L35Kelvin, L. S., Driver, S. P., Robotham, A. S. G., et al., 2012, MN-RAS, 421, 1007Kennicutt, Jr., R. C., 1998, ARA&A, 36, 189Kewley, L. J. & Ellison, S. L., 2008, ApJ, 681, 1183Kobulnicky, H. A. & Kewley, L. J., 2004, ApJ, 617, 240Krogager, J.-K., Fynbo, J. P. U., Møller, P., Ledoux, C., Noter-daeme, P., Christensen, L., Milvang-Jensen, B., & Sparre, M.,2012, MNRAS, 424, L1Kudritzki, R.-P., Urbaneja, M. A., Gazak, Z., Bresolin, F., Przy-billa, N., Gieren, W., & Pietrzy´nski, G., 2012, ApJ, 747, 15Laursen, P., Duval, F., & ¨Ostlin, G., 2013, ApJ, 766, 124Laursen, P., Razoumov, A. O., & Sommer-Larsen, J., 2009a, ApJ,696, 853Laursen, P., Sommer-Larsen, J., & Andersen, A. C., 2009b, ApJ,704, 1640Ledoux, C., Petitjean, P., Fynbo, J. P. U., Møller, P., & Srianand,R., 2006, A&A, 457, 71Ledoux, C., Petitjean, P., & Srianand, R., 2003, MNRAS, 346,209Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci,A., 2010, MNRAS, 408, 2115Melioli, C., de Gouveia Dal Pino, E. M., & Geraissate, F. G., 2013,MNRAS, 430, 3235Meyer, D. M. & Roth, K. C., 1990, ApJ, 363, 57Modigliani, A., Goldoni, P., Royer, F., et al., 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Se-ries, Vol. 7737, Society of Photo-Optical Instrumentation Engi-neers (SPIE) Conference SeriesMøller, P., Fynbo, J. P. U., & Fall, S. M., 2004, A&A, 422, L33Møller, P., Fynbo, J. P. U., Ledoux, C., & Nilsson, K. K., 2013,MNRAS, 430, 2680Neeleman, M., Wolfe, A. M., Prochaska, J. X., & Rafelski, M.,2013, ApJ, 769, 54Noterdaeme, P., Laursen, P., Petitjean, P., et al., 2012, A&A, 540,A63Noterdaeme, P., Ledoux, C., Petitjean, P., & Srianand, R., 2008,A&A, 481, 327O’Rourke, D. J. P., Shabala, S. S., & Alexander, P., 2011, MN-RAS, 418, 2145Pagel, B. E. J., Edmunds, M. G., Blackwell, D. E., Chun, M. S.,& Smith, G., 1979, MNRAS, 189, 95Peng, C. Y., Ho, L. C., Impey, C. D., & Rix, H.-W., 2002, AJ, 124,266P´eroux, C., Bouch´e, N., Kulkarni, V. P., York, D. G., & Vladilo,G., 2012, MNRAS, 419, 3060Pettini, M., 2006, in The Fabulous Destiny of Galaxies: BridgingPast and Present, Le Brun, V., Mazure, A., Arnouts, S., & Bur-garella, D., eds., p. 319Pettini, M., King, D. L., Smith, L. J., & Hunstead, R. W., 1997,ApJ, 478, 536Pettini, M. & Pagel, B. E. J., 2004, MNRAS, 348, L59Prochaska, J. X., Chen, H.-W., Dessauges-Zavadsky, M., &Bloom, J. S., 2007, ApJ, 666, 267Rafelski, M., Wolfe, A. M., Prochaska, J. X., Neeleman, M., &Mendez, A. J., 2012, ApJ, 755, 89Rauch, M., Haehnelt, M., Bunker, A., et al., 2008, ApJ, 681, 856Rhoads, J. E., Malhotra, S., Finkelstein, S. L., Fynbo, J. P. U.,McLinden, E. M., Richardson, M. L. A., & Tilvi, V. S., 2013,ArXiv e-prints, 1301.3140Salim, S., Charlot, S., Rich, R. M., et al., 2005, ApJ, 619, L39Savaglio, S., 2006, New Journal of Physics, 8, 195Shapley, A. E., 2011, ARA&A, 49, 525Tully, R. B. & Fisher, J. R., 1977, A&A, 54, 661Veilleux, S., Rupke, D. S. N., & Swaters, R., 2009, ApJ, 700, L149Vernet, J., Dekker, H., D’Odorico, S., et al., 2011, A&A, 536,A105Wolfe, A. M., Gawiser, E., & Prochaska, J. X., 2005, ARA&A,43, 861Wolfe, A. M., Prochaska, J. X., & Gawiser, E., 2003, ApJ, 593,215Zafar, T. & Watson, D., 2013, arXiv:1303.1141 c (cid:13)000