Quasars probing intermediate redshift star-forming galaxies
aa r X i v : . [ a s t r o - ph . C O ] D ec Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 6 September 2018 (MN L A TEX style file v2.2)
Quasars probing intermediate redshift star-forming galaxies
P. Noterdaeme, R. Srianand and V. Mohan ⋆ Inter-University Centre for Astronomy and Astrophysics, Post Bag 4, Ganeshkhind, 411 007 Pune, India
ABSTRACT
We present a sample of 46 [O iii ]-emitting galaxies at z < . ii ] and H β emission lines from most of thesegalaxies in the SDSS spectra. We study both the emission and absorption properties of a sub-sample of 17 galaxies in the redshift range z = . ii lines are covered by theSDSS spectra. The measured lower-limits on the star-formation rates of these galaxies are inthe range 0.2-20 M ⊙ yr − . The emission line luminosities and (O / H) metallicities from R23measured in this sample are similar to what is found in normal galaxies at these redshifts.Thus, this constitutes a unique sample of intermediate redshift star-forming galaxies wherewe can study the QSO absorber - galaxy connection. Strong Mg ii ( W λ > ∼ i absorption lines are detected in the QSO spectra at the redshift of most of these galax-ies. Strong Fe ii ( W λ > α systems. We investigate various possible relations between theMg ii rest equivalent widths and the emission line properties. We find a possible (2 σ ) correla-tion between the emission-line metallicity of the galaxies and the Mg ii rest equivalent widthof the absorbers (log(O / H) + = . W λ + . iii ]-selected Mg ii systems represent only aminor fraction of the strong Mg ii absorbers. We find this cannot be attributed to biases relatedeither to the spectral signal-to-noise ratio or to the brightness of the QSOs. We measure theaverage observed fluxes (collected into the SDSS fibre) of the [O ii ] and [O iii ] lines associatedto Mg ii -selected systems through stacking technique. We find that the average lumiosities ofemission lines are higher for systems with larger W λ . The stacked luminosities are foundto be below the typical detection limit in individual spectra, indicating that faint galaxies cancontribute appreciably to the observed population of strong Mg ii absorbers at intermediateredshifts. We also present long-slit spectroscopic observations of SDSS J113108 + z ≥ . Key words: galaxies: abundances – galaxies: ISM – galaxies: fundamental parameters –quasar: absorption lines – quasar:individual: SDSS J113108 + The study of intervening absorption lines seen in the spectra ofbright distant objects is one of the most sensitive and powerfulprobe for understanding the early evolution of galaxies. Indeedat low and intermediate redshifts ( z < ∼ ii -selected systems. These absorbers are found to be statisti-cally associated with relatively bright field galaxies seen within afew tens of kpc to the QSO line of sight (Bergeron & Boiss´e 1991; ⋆ E-mail:[pasquiern, anand, vmohan]@iucaa.ernet.in
Steidel 1995). These studies established that Mg ii absorbers pro-vide an unbiased way to detect normal galaxies at di ff erent red-shifts. However, the success rate of detecting Mg ii absorption in thespectrum of QSOs that have known foreground galaxies with red-shift measurements is much less than one (Bechtold & Ellingson1992; Bowen et al. 1995; Tripp & Bowen 2005). These studies sug-gest that the gaseous halos around galaxies may be less uniformlypopulated than what was thought before (see Kacprzak et al. 2008).Also it is not necessary that the galaxies responsible for Mg ii ab-sorption are always bright L ⋆ galaxies.Integral field spectroscopy seems to be a promising techniquefor the study of galaxies associated to quasar absorption line sys- c (cid:13) P. Noterdaeme, R. Srianand and V. Mohan tems and allowed Bouch´e et al. (2007) to detect H α emission as-sociated to 14 strong z ∼ ii absorbers (with impact param-eters in the range 1-40 kpc), indicating large star formation rates(1-20 M ⊙ yr − ). Another possibility is to directly search for galaxylight from Mg ii absorbers in special cases where the quasar flux atshort wavelengths is switched-o ff by a higher redshift Lyman-limitsystem (Christensen et al. 2009). Till now, such observations haveonly resulted in stringent upper-limits on the broad-band luminos-ity of the related galaxies.It has also been proposed that other classes of QSO ab-sorbers, such as those characterised by strong Ca ii absorptionlines, could select the most metal-rich gas (see, e.g. Wild et al.2006; Nestor et al. 2008), hence probing more central parts of high-redshift galaxies. Wild et al. (2007) have statistically detected [O ii ]emission associated to strong Mg ii - and Ca ii -selected absorbersby stacking SDSS quasar spectra. However, only a few direct de-tections of emission lines from absorbing galaxies have been re-ported so far. Zych et al. (2007) presented direct imaging and long-slit spectroscopic observations of five quasars with strong Ca ii sys-tems at z < .
5. They detected [O ii ], [O iii ], H α and H β emissionlines at the redshift of the absorbers. The luminosity of the corre-sponding galaxies is high, L ≃ L ⋆ , with star-formation rates in therange 0.3-30 M ⊙ yr − .When galaxies are detected with some projected separation tothe QSO sight-line it is not obvious whether one is detecting thehalo gas associated with the galaxy or one is probing the correla-tion length of metals in the IGM with respect to the bright galaxies.The contribution of possible faint galaxies closer to the line of sight(i.e. within the point spread function of the QSO) that remain unde-tected is also not well explored. Therefore, even after two decadesof intense research activity to establish the Mg ii absorber-galaxyrelationship, there are still open questions in this field that need tobe answered.Here, we present direct detections of emission lines from in-tervening star-forming galaxies (with impact parameters ≤
10 kpcand redshifts in the range 0.1 ≤ z ≤ z ≥ .
4, the SDSS spectra allow us to search for Mg ii absorp-tion originating from these galaxies. Our approach is very di ff er-ent from all previous ones in the sense that we do not make anypre-selection of galaxies based on QSO absorption lines. On thecontrary, we take advantage of SDSS fibre spectra, by first search-ing for galaxy emission lines on top of background quasars spectra,then looking for the associated absorption lines. Throughout the pa-per, we adopt a Λ CDM cosmology with Ω m = . Ω Λ = . H =
70 km s − Mpc − (e.g. Spergel et al. 2003). We are mostly interested in detecting normal star forming galaxiesclose to the line of sight of background QSOs. We focus on theredshift range z = . ii absorption lines, fall in the spectral range coveredin the SDSS spectrum. In particular, [O iii ] doublet lines are usefulbecause the [O iii ] λ iii ] λ iii ] emitters that can be detected fromthe SDSS QSO spectra, describe our automatic routine to detectgalaxies and discuss the bias due to QSO luminosity. Here we try to get a rough estimate of the number of galaxieswith di ff erent [O iii ] luminosities, at 0 . ≤ z gal ≤ . iii ] λ L ⋆ [OIII] galaxy(log L ⋆ [OIII] (erg s − ) = z = . F ⋆ ( z = . ∼ − erg s − cm − Å − , assuminga line width FWHM ≃ L ⋆ [OIII] refers to the Schechter parameter of the [O iii ]-luminosity function.As we aim here at detecting low luminosity galaxies, we shouldreach a detection limit of the order of 10 − erg s − cm − Å − .If we assume the dominant noise for the detection of an emis-sion line is the photon noise from the quasar continuum (i.e. weignore the background noise) then the signal-to-noise ratio in thepeak of the emission line, SNR l , can be written as:SNR l = N l √ ( N c + N l ) , (1)where N l and N c are the counts in the line and in the continuumrespectively. The signal-to-noise ratio of the continuum, SNR c , isequal to N c / √ N c and the count ratio is equal to the flux ratio( N l / N c = F l / F c ). Therefore, we can writeSNR l = SNR c F l / F c √ + F l / F c . (2)The detectability of the emission line will not only depend on the[O iii ] line flux ( F l ) but also on the quasar flux ( F c ). For a givenemission line flux, the minimum signal-to-noise ratio required todetect the [O iii ] emission will depend upon the magnitude of thequasar. That is, the 3 σ detection of the [O iii ] λ L = . L ⋆ [OIII] galaxy towards a i =
19 quasar will re-quire a spectrum with signal-to-noise ratio ≥
10. In turn, detectingthe same galaxy towards a i =
17 quasar will require a signal-to-noise ratio higher than about 45.Next, we estimate the probability of a given line of sight topass very close or through a galaxy with line luminosity greaterthan a limiting luminosity L . Integrating the [O iii ] luminosityfunction over the luminosities gives the number density of galaxies: n L > L = Z ∞ L Φ ( L ) dL = Φ ⋆ Γ α + , L L ⋆ [OIII] , (3)where Φ ⋆ , α and L ⋆ [OIII] are the Schechter parameters of the lumi-nosity function, taken from Table 5 of Ly et al. (2007) and Γ ( a , b )is the incomplete gamma function. We can estimate the number ofgalaxies with an impact parameter less than the SDSS fibre radius(1 . ′′ , i.e. r ∼
10 kpc at z ∼ . N L > L = π r n L > L , (4)For L = . L ⋆ [OIII] , this gives us a number density per unit distance N L > . L ⋆ ∼ − × − Mpc − . That is, for a line of sight to a quasarprobing z ∼ . − .
6, the probability of a L ≥ . L ⋆ [OIII] galaxy be-ing at an impact parameter less than the SDSS fibre radius will beabout 5 − × − . We need to consider this as a very conservativeupper limit as while estimating this number we have not consid-ered (1) any bias due to the luminosity of the background QSO(see Sect. 2.3), (2) the emission line attenuation due to dust internalto the galaxy, (3) the fibre losses (i.e. the fact that only a fractionof [O iii ] flux may go through the fibre) and (4) the colour selectionof QSOs missing dusty sightlines (Noterdaeme et al. 2009a), as ex-pected from galaxies at very low impact parameters. From these c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies simple-minded calculations, one can already see that a huge num-ber of quasar spectra ( > ) with adequate signal-to-noise ratioswill be required to detect a handful star-forming galaxies with animpact parameter ≤
10 kpc along the line of sight to distant QSOs.The availability of a large number of high quality QSO spectra inthe SDSS database makes it a realistic possibility to build a sam-ple of such star forming galaxies to bridge the connection betweenstrong Mg ii systems and star-forming galaxies. iii ] lines in SDSS quasar spectra We employ a correlation analysis to select emission line galaxiesclose to the line of sight to 98 978 quasars from the Sloan Digi-tal Sky Survey II, Data Release 7, without any prior knowledge ofthe absorption properties of the galaxies. As a first step we iter-atively fit the quasar continuum by applying Savitsky-Golay fil-tering and removing deviant pixels. We then cross-correlate thecontinuum-subtracted quasar spectra with a template profile of[O iii ] λλ α forest (in particularfor high redshift QSOs) and below λ = z gal < . C > . iii ] λ iii ] λ σ level. Notethat because of the galactic template used, wide emission lines aris-ing from AGNs are not picked-up by our procedure. Each candi-date was then inspected visually for the presence of other emis-sion lines in particular H β , H γ and [O ii ] λλ ≤ z gal ≤ ii absorption in theSDSS spectrum itself. These objects form the main sample of thispresent study.In addition, we performed an automatic search for the[O ii ] λλ , associated to Mg ii systemsto detect higher redshifts systems, for which telluric lines make anysearch based on [O iii ] a di ffi cult task. Mg ii systems were found byan automatic procedure based on correlation analysis, similar tothat used by Noterdaeme et al. (2009b) to search for DLAs. Thisprovided us with two additional galaxies, at z = .
669 and 0.788towards SDSS J120908 + + iii ]-selected galax-ies at 0 . < z gal < . ii λλ ff ect of QSO luminosity In Fig. 1 we present the fibre magnitudes and spectral signal-to-noise ratios in i -band of the whole quasar sample searched for[O iii ] emission lines and compare to that of quasars with detectedintervening emission lines. It is somehow surprising to see thatthe detections do not occupy any preferred region of the SNR-magnitude diagram. This is better seen in the upper (resp. right)panel where we compare the di ff erential and cumulative distribu-tions of the magnitudes (resp. signal-to-noise ratios) of QSOs with The two lines are always blended at the SDSS resolution
Table 1.
Sample of galaxies with [O iii ] emission around QSO sightlineQSO plate,MJD,fiber z QSO z gal a selectionJ023914-070557 0455,51909,562 0.715 0.342 [O iii ]J080216 + iii ]J080808 + [O iii ]J081154 + [O iii ]J082057 + iii ]J085113 + iii ]J091417 + [O iii ]J092913 + [O iii ]J094041 + [O iii ]J094335-004322 0266,51630,125 0.271 0.099 [O iii ]J094759 + iii ]J095228 + [O iii ]J101246 + iii ]J103309 + iii ]J104223 + [O iii ]J104257 + iii ]J110224 + iii ]J111343 + iii ]J112146 + iii ]J113002 + iii ]J113108 + [O iii ]J114340 + iii ]J120538 + [O iii ]J120908 + Mg ii + [O ii ]J121510 + [O iii ]J122752 + [O iii ]J125339 + [O iii ]J131804 + iii ]J132542 + [O iii ]J133733 + iii ]J132918 + iii ]J140103-005030 0301,51942,052 0.927 0.357 [O iii ]J140159 + iii ]J142421 + [O iii ]J143458 + iii ]J144412 + iii ]J145240 + iii ]J150140 + iii ]J154542 + [O iii ]J160521 + iii ]J161016 + iii ]J161607 + iii ]J161728 + Mg ii + [O ii ]J165508 + [O iii ]J165632 + [O iii ]J235621 + iii ] a The redshifts of galaxies studied in this paper, z gal ≥ .
4, are marked inbold face. and without a foreground [O iii ]-emitting galaxy. Indeed, a double-side Kolmogorov-Smirnov test shows that the probability of themagnitudes of the quasar with [O iii ]-selected intervening galax-ies to arise from the same parent population as the whole quasarsample is high (P KS = . KS = . / N ratio required forthe 1 σ detection of an emission line with peak intensity F l = − erg s − cm − Å − (or equivalently the 3 σ detection of F l = × − erg s − cm − Å − ) as a function of the i -band magnitudeof the background quasar. The spectral S / N ratios and i -band mag-nitudes of the quasars roughly follow this relation. While detec-tions are naturally made easy towards faint quasars, the steeply in-creasing S / N ratio of SDSS spectra with the quasar brightness also c (cid:13)000
4, are marked inbold face. and without a foreground [O iii ]-emitting galaxy. Indeed, a double-side Kolmogorov-Smirnov test shows that the probability of themagnitudes of the quasar with [O iii ]-selected intervening galax-ies to arise from the same parent population as the whole quasarsample is high (P KS = . KS = . / N ratio required forthe 1 σ detection of an emission line with peak intensity F l = − erg s − cm − Å − (or equivalently the 3 σ detection of F l = × − erg s − cm − Å − ) as a function of the i -band magnitudeof the background quasar. The spectral S / N ratios and i -band mag-nitudes of the quasars roughly follow this relation. While detec-tions are naturally made easy towards faint quasars, the steeply in-creasing S / N ratio of SDSS spectra with the quasar brightness also c (cid:13)000 , 000–000 P. Noterdaeme, R. Srianand and V. Mohan
14 16 18 20 22I (mag)020406080 S N R (I) [OIII], z<0.4[OIII], z>0.4MgII+[OII], z>0.4 . . . Figure 1. i -band signal-to-noise ratio and magnitude of the searchedquasars. Contours are drawn in the region with highest density of pointsfor presentation purpose. The upper and right panels show, respectively themagnitude and the signal-to-noise ratio distributions for the whole quasarsample (unfilled histogram) and for quasars selected upon the presence ofintervening [O iii ] emission line (filled histogram). The filled histogramshave been scaled for presentation purpose only. Dotted (resp. solid) curveson the top and right panels represent the normalised cumulative distribu-tions for the whole quasar sample (resp. quasars with [O iii ]-selected inter-vening galaxies). The three curves in the main panel represent the minimumsignal-to-noise required for the 1 σ detection of a emission line with peakflux F l = / × − (dotted), 10 − (dashed) and 5 × − (dashed-dotted) erg s − cm − Å − . allows for detections towards relatively bright quasars. However,the signal-to-noise ratios reached by the SDSS for the very brightquasars ( i <
17) are still not su ffi ciently high to detect interveningemission lines with F l ∼ − erg s − cm − Å − . As can be seen inthe top panel of Fig. 1, quasars with i <
17 are rare anyway and donot contribute much to the statistics.
The [O ii ] λλ β are the other stronglines that are expected in the wavelength range covered by theSDSS spectrum for the redshift range of our interest. We detect[O ii ] λλ z = .
445 galaxyalong the line of sight towards J081154 + / N in the wavelength range of [O ii ]emission is poor. The H β line is detected in all but two cases. Herewe perform a detailed analysis of H β , [O iii ] and [O ii ] emissionlines through Gaussian fitting. As a first step we subtract the contin-uum emission that includes the continuum light from both the QSOand the galaxy. The unabsorbed continuum (including the quasarbroad emission lines but excluding intervening emission lines) isdetermined accurately by manually adjusting a spline function tothe observed spectrum. Then we simultaneously fit the detectedemission lines using an IDL code based on MPFIT (Markwardt2009), which performs χ -minimisation by Levenberg-Marquardttechnique. We use a single redshift for all emission lines. The [O ii ] λλ , iii ] λλ iii ], [O ii ] and H β lines are then measured from thefitted Gaussian parameters. In the following, we always refer to[O iii ] flux (or luminosity) as the sum of fluxes (or luminosity) of[O iii ] λ iii ] λ ii ], we usethe sum of [O ii ] λ ii ] λ σ upper limits. Using the measured redshifts and fluxes we estimate the emissionline luminosities for the cosmological parameters noted above. Lu-minosities for all the three lines are also summarised in Table 2.Note that we do not apply any correction for the dust reddeningor the fact that the fibre need not sample the whole galaxy. Thismeans that the quoted luminosities should be treated as lower lim-its. In Fig. 3, we compare the distribution of the measured [O iii ]luminosities of our galaxies and the [O iii ] luminosity functions at z = . − . L ⋆ [OIII] galaxy. As expected we mostly de-tect galaxies with luminosities in the range 0.1-3 L ⋆ [OIII] , with a me-dian [O iii ] luminosity L [OIII] ∼ L ⋆ [OIII] . The galaxy along the lineof sight towards J113108 + L [OIII] ∼ L ⋆ [OIII] . We studythis system in detail using our observations with IUCAA Girawaliobservatory (IGO) in Section 6.The sharp decrease seen in the number of galaxies detectedat the low luminosity end (log L [OIII] <
41) is a consequence ofour detectability limit, as can be seen from the departure fromthe dotted curve in Fig. 3, while the decrease at high luminosities(log L [OIII] >
42) is a natural consequence of the decrease in thenumber density of very luminous galaxies. As there is a factor 30spread in the luminosity, this set of galaxies provides a good sam-ple for various followup studies such as measuring the cross-sectionand filling factor of Mg ii absorbers at low impact parameters (i.e ≤
10 kpc).
In this section, we aim at deriving the star-formation rates of in-tervening galaxies using emission line luminosities. It is knownthat the relationship between the observed luminosity of a line andstar formation rate depends on dust extinction and metallicity [seefor example Argence & Lamareille (2009)]. In the redshift range ofgalaxies that we focus on in this paper, the H α line redshifts into thenear-IR wavelengths. Thus we can not use the Balmer decrement toget the dust extinction. Also, as the continuum of the galaxy is com-bined with that of the QSO, we can not use the SED fitting to getthe estimates for reddening.Using more than 100 000 star-forming galaxies from SDSS,Argence & Lamareille (2009) have provided fitting formulae (theirEqs. 23 and 24) that use uncorrected [O ii ] and [H β ] luminosities c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies Fe ii , Mn ii Mg ii Mg i [O ii ] H β [O iii ] J . + . −20246 //// J . + . −20246 //// J . + . −101234 //// J . + . −202468 //// J . + . −5051015 //// J . + . −20246 //// J . + . −50510152025 //// J . + . −1001020304050 //// J . + . −202468 //// J . + . //// Rest wavelength (Å) Rest wavelength (Å) ← ′′ → Figure 2.
Absorption (left) and emission (middle) lines from the intervening galaxies. Note that in the left panels (absorption), the spectrum is normalised bydividing the observed spectrum by the QSO continuum, while in the middle panels (emission), the QSO continuum is subtracted from the observed spectrumand the flux-scale is in units of 10 − erg s − Å − . Best fitted absorption and emission lines are overplotted. The SDSS images of the QSOs are shown in theright panels. The black circle represents the position of the 3 ′′ -diameter SDSS fibre. North is top and East is left.c (cid:13) , 000–000 P. Noterdaeme, R. Srianand and V. Mohan Fe ii , Mn ii Mg ii Mg i [O ii ] H β [O iii ] J . + . −202468 //// J . + . −2024681012 //// J . + . −5051015 //// J . + . −20246810 //// J . + . −202468 //// J . + . −505101520 //// J . + . −5051015 //// J . + . −20246810 //// J . + . //// Rest wavelength (Å) Rest wavelength (Å) ←
10 arcsec → Figure 2. continued to get the star formation rates. The typical quoted dispersion in theSFR is about 0.23 dex. As noted by Argence & Lamareille (2009),the SFR derived using their Eq. 23 is weakly sensitive to the varia-tions in dust attenuation. However, their Eq. 24 provides a SFR es-timate that is weakly sensitive to the variations in metallicity. Thusfor each galaxy we consider the SFR estimates based on the above-mentioned two equations to provide the realistic range in SFR (lastcolumn in Table 2). In this table we also give SFR based only onthe [O ii ] and [H β ] calibrators. However, these estimates should betreated as lower limits as the SDSS fibres may collect only part ofthe emission from the galaxies. It is clear from Table 2 that the star formation rate in the galax-ies in our sample varies between 0.2 and 20 M ⊙ yr − . Roughly halfof the galaxies in the sample have a SFR less than 2 M ⊙ yr − . Notethat in the case of J132542 + ff erent calibrators. We cautionthat the [O ii ] and H β emission lines in this system are close to thedetection limit and using their ratio may lead to a significant over-estimation of the SFR and metallicities (see next sub-section). It isinteresting to note that all the 5 Ca ii absorption selected galaxiesstudied by Zych et al. (2007) have a SFR similar to that of galaxiesin our sample. c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies Table 2.
Star formation rates of the intervening galaxies a QSO z gal F λ (10 − erg s − cm − ) L (10 erg s − ) b SFR (M ⊙ yr − )[O ii ] [O iii ] H β [O ii ] [O iii ] H β [O ii ] c H β c [O ii ] + H β d J080808 + + ≤ ≤ . + ≤ ≤ . + + + + + + + e + + + + + + + e ≤ . ≤ . + + a Due to fibre losses, luminosities and star formation rates should be considered as lower limits. b Luminosities are notcorrected for dust-extinction. c Updated SFR calibrations for [O ii ] (Kewley et al. 2004) and H α (Kennicutt 1998) were taken from Argence & Lamareille (2009), assuming an intrinsicBalmer decrement of 2.85: H α = β . d Self-consistent two-lines calibration ([O ii ] + H β ) from Argence & Lamareille (2009). e (Mg ii + [O ii ])-selectedgalaxies.
40 41 42 43log L [OIII] (erg s −1 )−6−5−4−3−2−1 l og φ ( M p c − d e x − ) N u m b e r Figure 3.
The distribution of [O iii ] luminosities for the intervening galaxies(light histogram: [O iii ]-selected, dark histogram: (Mg ii + [O ii ])-selected) iscompared to the [O iii ] luminosity functions at z = . z = .
64 (lower curve), to be read on the left axis. The vertical dashed linemarks the position of L ⋆ [OIII] = . × erg s − (Hippelein et al. 2003).The dotted curve represent the [O iii ] luminosity function in linear space andarbitrary scaling to illustrate the completeness of the luminosity distributionof the galaxies presented in this paper. We use the R23 ratio, ([O ii ] + [O iii ]) / H β ), as calibrated byKobulnicky et al. (1999) to measure the metallicities. This calibra-tion provides two solutions for most values of R23, a low and a highmetallicity estimates. These are generally referred to as “lower” and “upper” branches of R23. The use of additional line ratios is nec-essary to break the degeneracy (see Kewley & Ellison 2008). Un-fortunately, the corresponding lines are not covered by the SDSSspectra. Therefore, we provide the lower and upper estimates of themetallicity derived using the R23 ratio only. The oxygen metallic-ity estimated using uncorrected fluxes are given in Table 3. Thecolumns R23 l and R23 u refer to the lower and upper branch ofR23. It is known that the value of (O / H) estimated with and with-out dust corrections are consistent with each another within 0.1 dex(Moustakas & Kennicutt 2006; Lamareille et al. 2006).Mouhcine et al. (2006) pointed out that, for the intermedi-ate redshift galaxies where the degeneracy is lifted using otherline indicators, the (O / H) is found to be predominantly close tothe value obtained for the upper branch. We find that the (O / H)range in our sample obtained using R23 u compares well withthat measured in intermediate redshift field galaxies and clustergalaxies (See Kobulnicky & Phillips 2003; Kobulnicky et al. 2003;Lilly et al. 2003; Mouhcine et al. 2006). This confirms that ourgalaxy selection is not heavily biased towards high or low metal-licity galaxies. However, we wish to point-out that the (O / H) de-termination based on R23 u for the Ca ii and DLA-selected galaxies(Zych et al. 2007) are slightly higher than the values we find for thegalaxies in our sample. The emission line analysis presented in the previous section clearlysuggests that the distribution of physical properties of the emissionline galaxies in our sample are consistent with that found for fieldgalaxies at similar redshift range. Thus, we have an unbiased, al-beit small sample of star-forming galaxies where we will be able toprobe the nature of absorption lines they produce in the spectra ofbackground QSOs that are at an impact parameter ≤
10 kpc.In this section, we analyse the absorption lines produced in the c (cid:13) , 000–000 P. Noterdaeme, R. Srianand and V. Mohan
Table 3.
Emission-line metallicitiesQSO z gal log(O / H) + l R23 u J080808 + + + + + + + + + + † + + + + + + + † + + / H) + = . ± . † (Mg ii + [O ii ])-selected galaxies. QSO spectra by the emitting galaxies. For this, each quasar spec-trum is normalised by dividingthe observed spectrum by the quasarcontinuum. In principle, the galaxy continuum emission should besubtracted prior to normalisation of the spectrum. However, con-tinuum emission from the galaxies is expected to be very smallcompared to that of the QSO and should have a negligible e ff ecton the measurement of absorption line parameters. The equiva-lent widths of metal absorption lines are then obtained by simul-taneous Gauss-profile fitting, using a single absorption redshift (i.e z abs ) to describe all detected absorption lines but allowing for itto be di ff erent from z gal obtained from emission lines. The nor-malised spectra and gaussian fits to the absorption lines are alsoshown in Fig. 2. The rest equivalent width of Fe ii λλ , ii λλ , i λ z gal ) and the relative ve-locity shift between the centroids of the absorption and emissionlines ( ∆ v / c = ( z abs − z gal ) / (1 + z gal )). ii and Fe ii absorption lines We have chosen the redshift range such that the available SDSSspectrum of each QSO covers the expected wavelength range ofMg ii absorption from the galaxy. In the case of the z gal = . + z em = . ii absorption is completely absorbed by ahigher redshift Lyman limit system. Similarly in the case of the z gal = .
565 galaxy towards J122752 + ii absorption is blended with the Lyman- α forest absorptionfrom high redshift. In the remaining 15 galaxies that are selectedmainly through [O iii ] emission we detect Mg ii absorption with restequivalent widths greater than 1 Å in 13 cases (See Fig. 2). This im-plies a detection rate of Mg ii absorption with W λ ≥ ∼ W λ ≥ ∼
60% detection rate.The mean Mg ii equivalent width of our [O iii ]-selected galaxysample is h W λ i ∼ . h W λ i ∼ . r (MgII λ N u m b e r r (MgII λ C u m u l . fr ac ti on Figure 4.
Distribution of Mg ii λ ii sample (with z abs = . − .
7, unfilled histogram) and that ofthe [O iii ]-selected galaxies, (light grey histogram). The values for the two(Mg ii + [O ii ])-selected galaxies are represented in dark grey. The two distri-butions are represented with di ff erent scales for presentation purpose only(SDSS left, galaxy sample right). The dotted line represents the parametri-sation by Nestor et al. (2005), scaled to match the number of systems with W λ ≥ W λ = ii from SDSS: dashed; Mg ii from our galaxysample: solid). the two [O ii ]-selected galaxies). In Fig. 4 we present the distribu-tion of Mg ii λ . < z < . ii equiva-lent widths with the exponential parametrization by Nestor et al.(2005), we find that our sample of Mg ii -selected absorbers is prob-ably complete down to W λ = iii ]-selected Mg ii absorbers are selected without any a priori information on their ab-sorbing properties, we can compare the two distributions for equiv-alenth widths above this value. It is clear from the figure that theMg ii equivalent widths in our [O iii ]-selected sample are predomi-nantly distributed towards higher equivalent widths. A double-sideKolmogorov-Smirnov test gives a probability 1 . × − that the twodistributions (for W λ ≥ ii absorption lines associated to line-emittinggalaxies are clearly characterised by larger equivalent widths thanthe Mg ii -selected absorbers. Below, we address the question ofwhat fraction of strong Mg ii absorption systems in the same red-shift range are detected in our emission line search.In our [O iii ]-emission selected sample, we cover the restwavelength range of Fe ii λλ ii absorption lines in all the cases when W λ ≥ W λ ≥ i absorption is also detected.These systems satisfy the criteria defined by Rao et al. (2006) onthe equivalent widths of Fe ii , Mg ii and Mg i to select DampedLyman- α systems (see also Rao & Turnshek 2000). We thereforeexpect that more than half of the systems in our sample are bona-fide DLAs with log N (H i ) ≥ .
3. From Fig. 3 of Steidel (1995)it is clear that galaxies associated with DLAs have low impact pa-rameters (i.e ≤
14 kpc) (see also Rao et al. 2003; Chen & Lanzetta2003). c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies Table 4.
Absorption line measurementsQSO z gal ∆ v E(B-V) a Rest equivalent widths (Å)(km s − ) Fe ii λ ii λ ii λ ii λ i λ + + + b + + d . . . . . . 2.1(0.6) 2.1(0.5) 0.6(0.3)J095228 + + < < < + d < + + c + + < < + + < . < . < + < + d + c + d . . . . . . 2.2(0.3) 2.2(0.3) 0.6(0.3)J165632 + a Negative values for E(B-V) are the consequence of intrinsic quasar shape variations. b A Lyman-limit system is present at z abs ∼ . λ < c (Mg ii + [O ii ])-selected galaxies. From Fig. 1 of Argence & Lamareille (2009), we can see that theaverage dust optical depth in star forming galaxies in their SDSSsample is ∼ .
2. This corresponds to an A V of 1.3 and E(B-V)of 0.42 for the assumed R V = ii ] emission by stackingthe spectra of quasars with strong Ca ii absorbers, which have beenproved to contain on an average larger amounts of dust than H i -selected DLAs (Wild et al. 2006; Nestor et al. 2008). While dustyabsorbers are good candidates to search for the host galaxy emis-sion lines, it is very interesting to verify whether the reciprocalis also true (i.e. whether the absorbers associated to star-forminggalaxies within an impact parameter of 10 kpc are also dusty).We aim here at deriving the selective reddening E(B-V) ofthe background QSO produced by the absorbing galaxy. We usethe same procedure as described in Srianand et al. (2008a) andNoterdaeme et al. (2009a). In short, we fit the observed spectrumwith a SDSS quasar composite spectrum (Vanden Berk et al. 2001),reddened by an extinction law shifted to the redshift of the in-tervening galaxy. We use the SMC extinction curve given byGordon et al. (2003), which has been shown to reproduce well theaverage reddening due to Mg ii absorbers (e.g. Khare et al. 2005;M´enard et al. 2005; Wild et al. 2006; York et al. 2006, see howeverSrianand et al. 2008a for individual cases). Other extinction laws(LMC, MW) provide similar results as they are very similar in therest wavelength range of the absorbers covered by the SDSS spec-tra. QSO-to-QSO intrinsic shape variations are actually the mainsource of uncertainties. The distribution of E(B-V) is shown onFig. 5. As can be seen from this figure, the range in intrinsic QSOUV slopes introduces a scatter of about 0.02 mag in the distributionof measured E(B-V).We found three systems (towards J091417 + −0.1 0.0 0.1 0.2 0.3 0.4 0.5E(B−V)012345 fr e qu e n c y Figure 5.
Distribution of E(B-V) derived by fitting the SED of the back-ground quasar. The light grey histogram represents [O iii ]-selected galax-ies while the E(B-V) values for the two (Mg ii + [O ii ])-selected galaxies arerepresented in dark grey. The dashed line represents the median value forthe 19 galaxies. The Gaussian illustrates the ∼ J095228 + + ≥ .
15. Forthe first system, the reddening of the quasar is derived using alimited wavelength-range because of the presence of a Lyman-limit system. The quasar also has a high redshift ( z QSO = . + ii absorptionassociated with the [O iii ] emitting galaxy at z gal = .
419 there isa strong Mg ii system at z abs = .
977 with rest equivalent width c (cid:13) , 000–000 P. Noterdaeme, R. Srianand and V. Mohan of the Mg ii doublets 2.6 and 2.3 Å respectively. This systemalso has very strong Fe ii lines. The QSO SED is reasonably wellreproduced assuming the reddening is produced in this system.However, the best fitted curve under-predicts the flux in the blueend of the spectrum. Thus, the E(B-V) value for the galaxy at z gal = .
419 should be considered as an upper limit. In the case ofJ125339 + ii H and K lines are seenin absorption at the redshift of the QSO. Thus the reddening couldbe mainly due to the QSO host galaxy.As can be seen Fig. 5, the distribution is concentrated arounda median value E(B-V) = < < ii ab-sorbers (Wild et al. 2006) and dusty 21-cm and CO absorbersat intermediate redshifts (Srianand et al. 2008a; Noterdaeme et al.2009a). As expected the measured E(B-V) along the QSO line ofsight is much less than that measured for the SDSS galaxies usingemission line ratios (Argence & Lamareille 2009). Ledoux et al. (2006) have established a correlation between thevelocity width of low ionisation lines and the metallicity of highredshift DLAs. The slope of this relationship is shown to be con-sistent with the mass-metallicity relation found in local galaxies(Tremonti et al. 2004). In this sub-section we explore various pos-sible correlations between star formation indicators and metallicityindicators from the emission line fluxes and Mg ii equivalent width.We make the assumption that the Mg ii equivalent width reflects thenumber of components and the velocity spread between them andis not due to line saturation. This assumption allow us to use W λ as an indicator of the velocity spread along the QSO line of sight(see e.g. Nestor et al. 2003; Ellison 2006).In the three upper panels of Fig 6 we plot the luminosities of[O ii ], H β and the derived star formation rates as a function of therest equivalent width of the Mg ii λ ii ], H β luminosity and W λ .We also do not find any strong correlation between the SFR given inthe last column of Table 2 and W λ . However, the largest equiv-alent width systems are also associated to the largest luminosities.Except from the system towards J132542 + ii ] and H β emis-sion line fluxes are very low in this galaxy, it is well possible thatthe SFR and metallicities, which depend on the line ratios, are sig-nificantly over-estimated (see Sect. 3.2).In the bottom panel of Fig. 6, we plot the upper-branch (R23 u )estimate of (O / H) as a function of W λ . As pointed-out before,in the absence of additional constraints we end up with two de-generate metallicity measurements using R23. However, in thefive line-emitting galaxies studied by Zych et al. (2007) – withsimilar properties as those presented here – the upper-branch ofR23 is preferred. A trend for increasing emission-line metallici-ties with increasing equivalent width can be seen. The correlation(log(O / H) + = . W λ + .
27) is significant at the 2 σ level.However, to draw a firm conclusion on the velocity metallicity cor-relation we need to remove the degeneracy in the (O / H) estimationand get the velocity spread in the absorbing gas using high reso-lution spectroscopy. Interestingly, a similar trend is also observed L H β ( e r g s − ) L [ O II] ( e r g s − ) SF R ( M O • y r - ) r (MgII λ l og ( O / H ) + ( R u ) Figure 6.
From top to bottom: [O ii ] luminosities, H β luminosities, SFRsand emission-line metallicities vs the Mg ii equivalent width of the associ-ated absorbing gas. The SFRs are that from the two lines calibration (lastcolumn of Table 2). The dotted line on the bottom panel represents the lin-ear regression fit to the metallicity - equivalent width correlation. Grey cir-cles represent [O iii ]-selected galaxies with black circles the (Mg ii + [O ii ])-selected galaxies. between the Mg ii equivalent width and the gas-phase metallicitymeasured along the quasar line of sight (e.g. Nestor et al. 2003;Murphy et al. 2007).The velocity shifts ( ∆ v ) between the emission and the absorp-tion lines are quite small (see Table 4), at most about 100 km s − .These are consistent with the expected circular velocities of typicalgalaxies suggesting the absorbing gas is bounded to the emissionline galaxy. We note that the ∆ v we measure here are lower thanthat measured with respect to luminous galaxies at larger impactparameters (100-200 km s − for impact parameters ∼ − kpcSteidel et al. 2002). In Fig. 7 we plot ∆ v against W λ . We do notfind any trend between the two quantities. Although the sampleis too small to conclude, this may be explained by the variety of c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies r (MgII λ ∆ v ( k m s − ) Figure 7.
The velocity o ff set between the absorption and emission lineredshifts is plotted against the Mg ii rest equivalent width. Symbols are asper Fig. 6 galaxy morphologies found to be associated with low-redshift Mg ii systems (Le Brun et al. 1997). ii ABSORBERS
Based on our automatic search for Mg ii absorption in SDSS-DR7 we find 2319, 1807, 494 and 118 systems respectively with W λ ≤ ≥ . ≤ z ≤ .
7. It is clear from the previous discussion that whilemost of the emission line galaxies produce strong Mg ii absorp-tion not all the strong Mg ii absorption systems are detected in our[O iii ] emission searches. Indeed, the number of strong Mg ii sys-tems (with W λ ≥ iii ]-emitting galaxiesdetected within the SDSS fibre is about a hundred times less thanthe total number of strong Mg ii -absorbers. This could be due to (i)poor S / N of the spectra and / or the relative brightness between theQSO and the galaxy, (ii) low star formation rate in the underlyinggalaxies or (iii) the impact parameter of the emission line regionsbeing larger than ∼
10 kpc (or angular separations more than 1.5”).To explore the e ff ect of spectral signal-to-noise ratio andgalaxy-QSO contrast further, we plot the i -band magnitude vs theS / N for all QSOs with intervening Mg ii systems with W λ ≥ . ≤ z ≤ . i -band magnitudes in a narrow range18 to 19.5 mag. The spectral signal to noise ratio is constant withina factor 2 in this magnitude range. To the eye there seems to be atendency for the [O iii ]-selected galaxies to prefer slightly fainter i -band magnitude (see the upper histogram). However, KS tests donot indicate the di ff erences between two populations to be statisti-cally significant. Thus it seems that there is no clear indication thatwe could have missed strong emission from most of the Mg ii sys-tems mainly because of the poor signal to noise. It is also clear fromFig. 1 that, in the i -magnitude range 18 to 20 mag, the spectral S / Nachieved in all the QSO spectra are good enough to detect emissionlines with peak flux in excess of 10 − erg s − cm − Å − . Thus thelack of direct detection of emission lines from the strong Mg ii sys-tems is consistent with either their fluxes being small or the impact S N R ( i ) MgII−selected[OIII]−selected . . . Figure 8. i -band signal-to-noise ratios and magnitudes for quasars withstrong ( W λ > ii -systems (black dots) and inter-vening [O iii ] emission lines (red squares). The top and right panels showthe distributions and cumulative distributions of magnitudes and signal-to-noise, respectively. Red filled histograms (solid lines) stand for quasars withintervening [O iii ]-emission lines while blue unfilled histograms (dashedline) stand for quasars selected for strong intervening Mg ii absorption lines. parameters being large enough so that the emitting regions are notfalling inside the fibre.In order to investigate the average [O ii ] and [O iii ] emissionwithin 1.5” of associated intervening Mg ii absorption systems, webuild several composite spectra corresponding to di ff erent ranges ofMg ii equivalent widths. The quasar spectra featuring Mg ii absorp-tion lines in the range z abs = . − . ii ] and [O iii ] emission lines from intervening Mg ii absorptionsystems with rest equivalent width W λ in the ranges ≤
1, 1-2,2-3 and > ii equivalent widths. This canalso be seen from Table 5 where we give the average luminosityof [O ii ] and [O iii ] emission lines for di ff erent sub-samples definedusing W λ . We note that the average [O ii ] flux for W λ ∼ ii -selectedDLAs (1 < W λ / W λ < W λ > .
6; see Rao et al.2006).The average [O iii ] luminosity found by stacking Mg ii sys-tems with W λ > L ⋆ [OIII] /
10 and close to the low-est luminosity we directly measure in the [O iii ]-selected galax-ies (in z gal = .
444 towards J091417 + i = ii absorber towardsJ224630.62 + ii λ c (cid:13) , 000–000 P. Noterdaeme, R. Srianand and V. Mohan F λ ( − e r g s − c m − Å − ) [OII] 4950 5000 5050 5100 [OIII]Wavelength (Å) Figure 9. [O ii ] (left) and [O iii ] (right) emission lines from stacking QSOspectra with intervening Mg ii absorbers at 0 . < z abs < . ff erentMg ii λ W λ <
1; 1 ≤ W λ <
2; 2 ≤ W λ < W λ ≥ Table 5.
Luminosities of the emission lines in the stacked spectrumline Average luminosity a for systems withW <
1Å 1 < W < < W <
3Å W > ii ] 0.5(0.2) 1.4(0.1) 3.2(0.2) 5.1(0.6)[O iii ] 2.5(0.3) 3.5(0.3) 5.1(0.4) 10.0(0.9) a in units of 10 erg s − . However a galaxy is clearly seen within the area covered by thefibre (See Fig. 5 of Zych et al. 2007) and emission lines (with inte-grated [O iii ] flux of 8 × − erg s − cm − i.e. lower than our detec-tion limit with SDSS) are detected in the VLT / FORS spectra. Wealso note that the [O iii ] luminosity of this galaxy is similar to thestacked luminosity obtained using systems with 2 < W λ < z ∼ . ii absorbers at 0 . < z < . α emitting re-gions of Mg ii -selected galaxies have similar sizes. This means thatthe galaxies contributing to the stacked emission lines should haveimpact parameters less than about 20 kpc. Therefore, it is most un-likely that a bright galaxy at high impact parameter (several tensof kpc) will contribute to the average detection of emission linesin the stacked spectrum. The stacking method alone does not pro-vide the higher moments of the L [OIII] -distribution. However, withadditional constraints on the galaxy sizes and the small number ofdirect detections, our results are consistent with at least part of thestrong Mg ii absorbers arising from low L [OIII] luminosity galaxiesat low-impact parameters, as seen in the case of J224630 + iii ] and [O ii ] emissionlines are detected in the stacked spectrum even when we consideronly low equivalent widths. In particular there are roughly 4 timesmore systems with 1 ≤ W λ ≤ ≤ W λ ≤ iii ] and [O ii ] lumi-nosities are less only by a factor 2. Even though our Mg ii -selectedsystems with low rest equivalent widths ( W λ < / N spectra, it is interestingto see that we detect [O ii ] emission at 3 σ level and that [O iii ] isdetected at > σ . This means that galaxies with low impact param-eters can also produce low equivalent width absorption lines. This is consistent with the fact that we do not detect Mg ii absorptionwith W λ ≥ iii ]-selected galaxies.In summary, star-forming galaxies with low [O iii ] luminosi-ties seem to provide an important contribution to the population ofMg ii absorption selected galaxies. + In this section, we study the galaxy at z gal = .
563 towardsJ113108 + z > . ii absorptionlines. An extended galaxy (SDSS J113108.31 + ′′ from the quasar image and extendedtowards the quasar image (see Fig. 10). Moreover, the photomet-ric redshift provided by SDSS, z = . ± .
07, is close to thatof the detected emission and absorption lines. This is therefore avery good case to study the connection between the galaxy and theabsorber.In order to obtain an accurate measurement of the location ofthe emission region, we performed long-slit spectroscopy with the2 m telescope of the IUCAA Girawali Observatory. Observationswere carried out on March 20, 2009 using IUCAA Faint ObjectSpectrograph and Camera (IFOSC). A 2 ′′ slit and GR7 grism cov-ering the wavelength range between 3900 Å to 6800 Å were used.The detector used is a LN2 cooled thinned 2k ×
2k CCD camera.Each pixel on the CCD covers 0.34 ′′ of sky which corresponds to1.4 Å using the above grism. Three exposures of 2700 s each weretaken at slit position 10 deg from North (P1) and two exposuresof 2700 s each were taken at slit position 92 deg from North (P2).The slit positions have been indicated in Fig. 10. For flat fieldingHalogen lamp flats were used. Helium and Neon lamps were usedsimultaneously for getting comparison spectrum. The IRAF routine response was used for making a normalised flat. The doslit pack-age has been used to extract and calibrate the 1D spectrum. Twodimensional analysis was performed using the method described inVivek et al. (2009).Fig. 11 represents the two-dimensional spectra obtained withthe two slit orientations as illustrated on Fig. 10. We removed thequasar trace by fitting a Gaussian in the spatial direction, whoseamplitude is allowed to vary smoothly (2nd-order polynomial) withthe wavelength. We also left the position of the Gaussian in thespatial direction to vary linearly with the wavelength, to take intoaccount the possible misalignment between the quasar trace and thepixels of the CCD.The [O ii ] emission can be seen as a bright spot in both spec-tra. Interestingly, despite of aligning the slit with the galaxy seenat 3.5 ′′ from the quasar (see Fig. 10), there is no emission lineat the corresponding position (marked by a ’X’) in the 2D spec-trum. On the contrary, while the second slit angle was chosen toavoid including the galaxy, the [O ii ] emission is still detected. Thisdemonstrates the galaxy seen on the SDSS image is not responsi-ble for the detected emission lines. Interestingly, this galaxy doesnot produce any other detectable absorption line system in the QSOspectra over the wavelength range covered by the SDSS and IGOspectra. We obtained the centroid of the [O ii ] emission by fittingthe line with a two-dimensional Gaussian function. From triangu-lating the positions with the two slits, we are able to put a goodconstraint on the position of the emitting region. We measure animpact parameter D = . ′′ , which corresponds to 7 kpc at the red-shift of the galaxy. The centroid of the emission region is shown as c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies a r c s ec −6−4−20246 −6−4−20246 0 50 100 150Pixels −6−4−20246 0 50 100 150Pixels −6−4−20246 Dispersion axis (pixels)
Figure 11.
The background-subtracted 2D spectra of the quasar SDSS J113108 + z gal = .
56. Top: Total (quasar + galaxy) spectraobtained with two slit orientations (left: P1, 10 degrees from North; right: P2, 92 degrees from North, see Fig. 10). Bottom: Same spectra after removing thequasar trace. The centre of the quasar trace is represented by the horizontal dashed lines, and its FWHM by the horizontal dotted lines. The ellipses representthe FWHM of a 2D-Gaussian fitted over the galaxy emission line. The ’X’ on the bottom left panel represent the expected position of a [O ii ] emission linelocated 3.5 ′′ south from the quasar, i.e., at the centroid of the galaxy resolved by the SDSS. The data has been smoothed 2 × −10 −5 0 5 10arcsec−10−50510 a r c s ec P1 P2
Figure 10.
Layout of the slits (solid lines) and SDSS fibre (circle) on thequasar SDSS J113108 + σ constraints on the centroid of the intervening emission line galaxy, asobtained from each slit orientation. The cross marks the centroid of thegalaxy. a white cross on Fig. 10. We see the integrated [O ii ] flux measuredin IGO and SDSS spectra are consistent with one another suggest-ing that the whole line-emitting region is within the SDSS fibre. Taking advantage of the available ∼
100 000 fibre spectra of quasarsin the Sloan Digital Sky Survey-II, DR7, we build a unique sampleof 46 star-forming galaxies at z < . iii ] and / or [O ii ]) emission lines seen on top of the backgroundquasar spectra. We show the the detectability of [O iii ] lines is notbiased by the luminosity of the background quasars. We study both the emission and absorption properties of asub-sample of 17 galaxies at z ≥ . ii lines are covered by the SDSS spectra. The detectionsshow that we are probing a unbiased population of low luminosity[O iii ]-emitting galaxies at small impact parameters ( < ∼
10 kpc; i.e.the SDSS fibre radius) from the quasar lines of sight. We find thattypical properties (metallicity, star-formation rates, kinematics) ofthese galaxies are similar to that of normal star-forming galaxies atthese redshifts. The low E(B-V) we measure along the quasar linesof sight indicates that the absorption lines arise from regions rela-tively free of dust. This implies that quasar absorbers selected uponthe presence of cold gas and dust features (Srianand et al. 2008a,b;Noterdaeme et al. 2009a) might still be the best way to probe theinterstellar medium in the densest regions of normal galaxies in thedistant Universe.We find that the equivalent widths of Mg ii absorption linesarising from the [O iii ]-selected galaxies are skewed towards higherequivalent widths than the overall population of Mg ii absorbers.However, the [O iii ]-selected Mg ii absorbers represent only a smallfraction of the overall Mg ii population. From stacking the spec-tra of quasars featuring strong Mg ii absorbers, we detect the [O ii ]and [O iii ] emission lines. The average line fluxes are below ourtypical detection limit in individual spectrum. This suggests thatat least part of the strong ( W λ > ii absorption sys-tems arise from low luminosity galaxies at small impact parame-ters. Also strong Mg ii systems have been detected at higher ratearound clusters (Lopez et al. 2008). In such cases, the halo sizes ofMg ii systems are inferred to be less than 10 h − kpc (Padilla et al.2009).The absorption properties of the galaxies indicate that at leasthalf of the emission line galaxies in our sample, if not all, con-tain su ffi cient neutral gas to produce Damped Lyman- α absorption(Rao et al. 2006) as well as 21-cm absorption Gupta et al. (2009)along the quasar line of sight. Unfortunately, most of the QSOs inour sample do not have su ffi cient radio fluxes to carry-out 21-cmsearches.SDSS spectra allowed us to explore the possible connectionsbetween various parameters of the galaxies (such as metallicity,dust content and kinematics) derived from the absorbing gas andthat derived from emission lines in a limited redshift range. Nev-ertheless, our representative sample of 46 galaxies presented in c (cid:13) , 000–000 P. Noterdaeme, R. Srianand and V. Mohan
Table 1 is ideally suited for several follow-up observations usingspace and ground-based telescopes. This should allow one to ex-plore various issues such as: the connection between the redden-ing along the QSO line of sight and the dust extinction in theline-emitting region; the comparison between the emission andabsorption line metallicities; the dependence of the properties ofthe absorbing gas on the galaxy morphology, kinematics and im-pact parameter, etc. Finally, we perform long-slit observations ofthe most luminous galaxy with Mg ii absorption in our sample(SDSS J113108 + ii ] emission de-tected in the SDSS spectrum is not detected in the extended brightgalaxy seen on the SDSS image. This once again suggests that oneshould be cautious in associating intervening absorption (or emis-sion) to bright galaxies seen in the field with photometric redshiftmeasurements only. ACKNOWLEDGEMENTS
We gratefully thank the anonymous referee for thorough readingof the paper and helpful comments and suggestions. We also thankP. Petitjean and P. Boiss´e for useful comments on the manuscript.PN acknowledges support from the french Ministry of Foreign andEuropean A ff airs. We wish to acknowledge the IUCAA / IGO sta ff for their support during our observations. We acknowledge the useof the Sloan Digital Sky Survey. Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Par-ticipating Institutions, the National Science Foundation, the U.S.Department of Energy, the National Aeronautics and Space Ad-ministration, the Japanese Monbukagakusho, the Max Planck So-ciety, and the Higher Education Funding Council for England. TheSDSS Web Site is . The SDSS is managedby the Astrophysical Research Consortium for the Participating In-stitutions. The Participating Institutions are the American Museumof Natural History, Astrophysical Institute Potsdam, University ofBasel, University of Cambridge, Case Western Reserve University,University of Chicago, Drexel University, Fermilab, the Institutefor Advanced Study, the Japan Participation Group, Johns HopkinsUniversity, the Joint Institute for Nuclear Astrophysics, the KavliInstitute for Particle Astrophysics and Cosmology, the Korean Sci-entist Group, the Chinese Academy of Sciences (LAMOST), LosAlamos National Laboratory, the Max-Planck-Institute for Astron-omy (MPIA), the Max-Planck-Institute for Astrophysics (MPA),New Mexico State University, Ohio State University, Universityof Pittsburgh, University of Portsmouth, Princeton University, theUnited States Naval Observatory, and the University of Washing-ton. REFERENCES
Argence, B. & Lamareille, F., 2009, A&A, 495, 759Asplund, M., Grevesse, N., Sauval, A. J., & Scott, P., 2009,ARA&A, 47, 481Bechtold, J. & Ellingson, E., 1992, ApJ, 396, 20Bergeron, J. & Boiss´e, P., 1991, A&A, 243, 344Bouch´e, N., Lehnert, M. D., Aguirre, A., P´eroux, C., & Bergeron,J., 2007, MNRAS, 378, 525Bowen, D. V., Blades, J. C., & Pettini, M., 1995, ApJ, 448, 634Chen, H. & Lanzetta, K. M., 2003, ApJ, 597, 706Christensen, L., Noterdaeme, P., Petitjean, P., Ledoux, C., &Fynbo, J. P. U., 2009, A&A, 505, 1007 Dahlen, T., Mobasher, B., Dickinson, M., Ferguson, H. C., Gi-avalisco, M., Kretchmer, C., & Ravindranath, S., 2007, ApJ, 654,172de Mello, D. F., Wadadekar, Y., Dahlen, T., Casertano, S., & Gard-ner, J. P., 2006, AJ, 131, 216Ellison, S. L., 2006, MNRAS, 368, 335Ellison, S. L., Hall, P. B., & Lira, P., 2005, AJ, 130, 1345Gordon, K. D., Clayton, G. C., Misselt, K. A., Landolt, A. U., &Wol ff , M. J., 2003, ApJ, 594, 279Gupta, N., Srianand, R., Petitjean, P., Noterdaeme, P., & Saikia,D. J., 2009, MNRAS, 398, 201Hippelein, H., Maier, C., Meisenheimer, K., et al., 2003, A&A,402, 65Kacprzak, G. G., Churchill, C. W., Steidel, C. C., & Murphy,M. T., 2008, AJ, 135, 922Kennicutt, Jr., R. C., 1998, ARA&A, 36, 189Kewley, L. J. & Ellison, S. L., 2008, ApJ, 681, 1183Kewley, L. J., Geller, M. J., & Jansen, R. A., 2004, AJ, 127, 2002Khare, P., York, D. G., vanden Berk, D., et al., 2005, in IAU Col-loq. 199: Probing Galaxies through Quasar Absorption Lines,P. Williams, C.-G. Shu, & B. Menard, ed., pp. 427–429Kisielius, R., Storey, P. J., Ferland, G. J., & Keenan, F. P., 2009,MNRAS, 397, 903Kobulnicky, H. A., Kennicutt, Jr., R. C., & Pizagno, J. L., 1999,ApJ, 514, 544Kobulnicky, H. A. & Phillips, A. C., 2003, ApJ, 599, 1031Kobulnicky, H. A., Willmer, C. N. A., Phillips, A. C., et al., 2003,ApJ, 599, 1006Lamareille, F., Contini, T., Brinchmann, J., Le Borgne, J.-F., Char-lot, S., & Richard, J., 2006, A&A, 448, 907Le Brun, V., Bergeron, J., Boisse, P., & Deharveng, J. M., 1997,A&A, 321, 733Ledoux, C., Petitjean, P., Fynbo, J. P. U., Møller, P., & Srianand,R., 2006, A&A, 457, 71Lilly, S. J., Carollo, C. M., & Stockton, A. N., 2003, ApJ, 597,730Lopez, S., Barrientos, L. F., Lira, P., et al., 2008, ApJ, 679, 1144Ly, C., Malkan, M. A., Kashikawa, N., et al., 2007, ApJ, 657, 738Markwardt, C. B., 2009, ArXiv e-prints 0902.2850M´enard, B., Zibetti, S., Nestor, D., & Turnshek, D., 2005, inIAU Colloq. 199: Probing Galaxies through Quasar AbsorptionLines, P. Williams, C.-G. Shu, & B. Menard, ed., pp. 86–91Mouhcine, M., Bamford, S. P., Arag´on-Salamanca, A., Nakamura,O., & Milvang-Jensen, B., 2006, MNRAS, 369, 891Moustakas, J. & Kennicutt, Jr., R. C., 2006, ApJ, 651, 155Murphy, M. T., Curran, S. J., Webb, J. K., M´enager, H., & Zych,B. J., 2007, MNRAS, 376, 673Murphy, M. T. & Liske, J., 2004, MNRAS, 354, L31Nestor, D. B., Pettini, M., Hewett, P. C., Rao, S., & Wild, V., 2008,MNRAS, 390, 1670Nestor, D. B., Rao, S. M., Turnshek, D. A., & Vanden Berk, D.,2003, ApJ, 595, L5Nestor, D. B., Turnshek, D. A., & Rao, S. M., 2005, ApJ, 628, 637Noterdaeme, P., Ledoux, C., Srianand, R., Petitjean, P., & Lopez,S., 2009a, A&A, 503, 765Noterdaeme, P., Petitjean, P., Ledoux, C., & Srianand, R., 2009b,A&A, 505, 1087Noterdaeme, P., Petitjean, P., Srianand, R., Ledoux, C., & Le Petit,F., 2007, A&A, 469, 425Padilla, N., Lacerna, I., Lopez, S., Barrientos, L. F., Lira, P., An-drews, H., & Tejos, N., 2009, MNRAS, 395, 1135 c (cid:13) , 000–000 uasars probing intermediate redshift star-forming galaxies Rao, S. M., Nestor, D. B., Turnshek, D. A., Lane, W. M., Monier,E. M., & Bergeron, J., 2003, ApJ, 595, 94Rao, S. M. & Turnshek, D. A., 2000, ApJS, 130, 1Rao, S. M., Turnshek, D. A., & Nestor, D. B., 2006, ApJ, 636, 610Spergel, D. N., Verde, L., Peiris, H. V., et al., 2003, ApJS, 148,175Srianand, R., Gupta, N., Petitjean, P., Noterdaeme, P., & Saikia,D. J., 2008a, MNRAS, 391, L69Srianand, R., Noterdaeme, P., Ledoux, C., & Petitjean, P., 2008b,A&A, 482, L39Steidel, C. C., 1995, in QSO Absorption Lines, Meylan, G., ed.,p. 139Steidel, C. C., Kollmeier, J. A., Shapley, A. E., Churchill, C. W.,Dickinson, M., & Pettini, M., 2002, ApJ, 570, 526Storey, P. J. & Zeippen, C. J., 2000, MNRAS, 312, 813Tremonti, C. A., Heckman, T. M., Kau ff mann, G., et al., 2004,ApJ, 613, 898Tripp, T. M. & Bowen, D. V., 2005, in IAU Colloq. 199: ProbingGalaxies through Quasar Absorption Lines, Williams, P., Shu,C.-G., & Menard, B., eds., pp. 5–23Vanden Berk, D. E., Richards, G. T., Bauer, A., et al., 2001, AJ,122, 549Vivek, M., Srianand, R., Noterdaeme, P., Mohan, V., & Kuri-akosde, V. C., 2009, MNRAS, L321 + Wild, V., Hewett, P. C., & Pettini, M., 2006, MNRAS, 367, 211—, 2007, MNRAS, 374, 292York, D. G., Khare, P., Vanden Berk, D., et al., 2006, MNRAS,367, 945Zych, B. J., Murphy, M. T., Pettini, M., Hewett, P. C., Ryan-Weber, E. V., & Ellison, S. L., 2007, MNRAS, 379, 1409 c (cid:13)000