Probing star formation across cosmic time with absorption line systems
Brice Ménard, Vivienne Wild, Daniel Nestor, Anna Quider, Stefano Zibetti
aa r X i v : . [ a s t r o - ph . C O ] D ec Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed October 30, 2018 (MN L A TEX style file v2.2)
Probing star formation across cosmic time with absorptionline systems
Brice M´enard , Vivienne Wild , Daniel Nestor , , Anna Quider , Stefano Zibetti Canadian Institute for Theoretical Astrophysics Institut d’Astrophysique de Paris, C.N.R.S. Institute of Astronomy, University of Cambridge University of California in Los Angeles Max Planck Institut f¨ur Astronomie
Draft, October 30, 2018
ABSTRACT
We present an empirical connection between cold gas in galactic halos and star forma-tion. Using a sample of more than 8,500 Mg ii absorbers from SDSS quasar spectra,we report the detection of a 15 σ correlation between the rest equivalent width W of Mg ii absorbers and the associated [O ii ] luminosity, an estimator of star formationrate.This correlation has interesting implications: using only observable quantities weshow that Mg ii absorbers trace a substantial fraction of the global [O ii ] luminositydensity and recover the overall star formation history of the Universe derived fromclassical emission estimators up to z ∼
2. We then show that the distribution functionof Mg ii rest equivalent widths, d N/ d W inherits both its shape and amplitude fromthe [O ii ] luminosity function Φ( L ). These distributions can be naturally connected,without any free parameter.Our results imply a high covering factor of cold gas around star forming galaxies: C & .
5, favoring outflows as the mechanism responsible for Mg ii absorption. Wethen argue that intervening Mg ii absorbers and blue-shifted Mg ii absorption seen inthe spectra of star forming galaxies are essentially the same systems, implying thatthe observed outflowing gas can reach radii of ∼
50 kpc. These results not only shedlight on the nature of Mg ii absorbers but also provide us with a new probe of starformation, in absorption, i.e. in a way which does not suffer from dust extinction andwith a redshift-independent sensitivity. As shown in this analysis, such a tool canbe applied in a noise-dominated regime, i.e. using a dataset for which emission linesare not detected in individual objects. This is of particular interest for high redshiftstudies. Key words: absorbers: Mg ii – star formation rate – quasars – outflows In the past twenty years, we have seen enormous advances inour understanding of galaxy evolution. Gravitational lensinghas allowed us to interpret the properties of the host darkmatter halos, population synthesis models can characterizethe emission properties of galaxies over 3 orders of magni-tude in stellar mass, and the star formation history of theUniverse has been explored up to high redshift, revealing apeak at z ∼
2. Along with this, the importance of feedbackprocesses has emerged and one of the central questions whichnow needs to be addressed is how galaxies accrete, processand return gas into the inter-galactic medium (IGM). Absorption line spectroscopy provides us with a pow-erful tool to address this question, to explore in detail thedistribution of gas around galaxies and to study its inter-play with star formation. It gives us access to a wide arrayof elements and allows us to detect low-density gas that isorders of magnitude below the detection threshold of mostother techniques.Historically, the study of QSO absorber-galaxy relation-ships began with Mg ii absorption systems. This choice wasmainly driven by practical constraints: among the dominantions in H i gas only Mg ii could be studied from the groundat low redshifts ( z < ii absorber-galaxy connection by Bergeron (1986), a c (cid:13) B. M´enard et al. considerable amount of work has focused on revealing thenature and origin of these systems. Mg ii is known to be atracer of galactic halos, routinely seen up to about 50 kpcaround galaxies. However, despite significant observationalefforts in the last two decades, fundamental questions re-garding the physical nature of the absorbing gas remainunanswered. Various models have been proposed in the lit-erature: outflows (Bond et al. 2001), infalling gas (Tinker& Chen 2008), extended disks (Steidel et al. 2002), etc. butno consensus has been reached. Observationally, no com-pelling correlations between absorber properties and thoseof their associated galaxies have been reported. In particu-lar, the physical process(es) governing the observed equiva-lent widths W , the basic observable quantity, and the originof its distribution function are not yet understood.In this paper we present the detection of a correla-tion between the observed rest equivalent width of Mg ii absorbers and the associated [O ii ] luminosity and show thatthis relation has important implications. First, Mg ii gas ap-pears to trace a substantial fraction of the [O ii ] luminositydensity of the Universe. These absorber systems can there-fore be used as a new probe for star formation over cosmolog-ical times. We show that the observed redshift dependencereproduces the decline in cosmic star formation rate since z ∼ W and L OII and show that these twoquantities are related.This study was motivated by a number of recent re-sults, indicating a possible link between the presence ofstrong Mg ii absorbers and the star formation rate of theirassociated galaxies: Zibetti et al. (2007) analyzed the col-ors of stacked images of quasars with strong Mg ii absorbersand reported that the best-fit spectral energy distributions(SED) require stronger emission lines for stronger Mg ii ab-sorbers. Wild et al. (2007) created composite spectra of asubset of Mg ii -selected absorbers and was able to detecttheir mean [O ii ] emission. Rubin et al. (2009) studied theproperties of a Mg ii absorber-galaxy association and showedthat the absorbing gas found at about 15 kpc from the galaxyis consistent with having been ejected during the last burstof star formation of this galaxy. Finally, Nestor et al. (2010,in prep) studied the properties of two ultra-strong MgII ab-sorbers ( W = 5 . . ii absorption. Such an analysis does not provideany information on the distance between the absorbing gasand the galaxy, and the connection to intervening absorbersystems is not straightforward. However, we will show thatour new results suggest a strong link between these two trac-ers of gas in and around galaxies.The outline of this paper is as follows: after briefly re-viewing useful properties of Mg ii absorbers and [O ii ] emis-sion as a tracer of star formation, we describe the spectro-scopic data analysis in Section 3. We present the correlationbetween [O ii ] luminosity surface density and Mg ii equiva- lent width in Section 4. Section 5 presents the connectionbetween Φ( L OII ) and the distribution of Mg ii rest equiv-alent widths. Finally, we discuss the implications of theseresults in Section6 and summarize our work in Section 7.Throughout this paper we use the cosmology: Ω M =0 .
3, Ω Λ = 0 . H = 70 km/s/Mpc. ii absorbers Strong Mg ii absorbers, classically defined with W λ > . . . L . L ⋆ (Bergeron & Boiss´e 1991; Stei-del, Dickinson, & Persson 1994; Nestor et al. 2007) and arefound at impact parameters ranging from a few to morethan 50 kpc (Zibetti et al. 2007; Steidel et al. 1997; Tinker& Chen 2008). In the range W λ & ii absorp-tion lines are in general saturated (Nestor et al. 2005) andno column density information can be directly extractedfrom the observed equivalent width, which can instead beused as an estimate of the overall gas velocity dispersion.High-resolution spectroscopic observations reveal that theabsorption often originates from several velocity components(Churchill, Kacprzak, & Steidel 2005). Empirically, it hasbeen found that in this regime the velocity dispersion of thegas follows ∆ v ≃
120 ( W / ˚A) km/s (see Ellison 2006, Fig.3). A fundamental quantity describing the statistical prop-erties of these absorbers is the distribution of Mg ii restequivalent width per unit redshift, d N / d W d z . This quan-tity has been accurately characterized using the Sloan Digi-tal Sky Survey (SDSS) by Nestor et al. (2005). These authorsfound that it is well described by an exponential distribu-tion:d N d W = N ⋆ W ⋆ e − W /W ⋆ (1)with the maximum likelihood values N ⋆ = 1 . ± .
13 (1 + z ) . ± . and W ⋆ = 0 . ± .
03 (1 + z ) . ± . . This resultis valid down to a rest equivalent width limit of ∼ . ii ] emission as a tracer of star formation Over the past 20 years many techniques have been exploredto estimate the star formation rates (SFRs) of galaxies and,all together, have established the steep rise in the SFR den-sity of the Universe from the present epoch to z ∼ α λ & M ⊙ ) and therefore provides anear-instantaneous ( τ .
10 Myr) measure of the SFR with c (cid:13) , 000–000 robing star formation with absorption line systems Figure 1.
Left: redshift distributions of detected Mg ii absorbers (red) and background quasars with z < ii and [O ii ] lines can be detected. Right: observed distribution of MgII- λ W ) departs from an exponential distribution. minimal dependence on the physical conditions of the ion-ized gas (Kennicutt 1998). Such observations are howeverlimited to z < . ii ] λ ii ] emission doublet is the shortest-wavelengthstrong emission line from low-density, photoionized galacticnebulae. The nebular lines effectively re-emit the integratedstellar luminosity short-ward of the Lyman limit. Only starswith masses M >
10 M ⊙ and lifetimes τ <
20 Myr con-tribute significantly to the integrated ionizing flux, so theemission lines provide a nearly instantaneous measure of theSFR, independent of the previous star formation history.[O ii ] luminosity depends explicitly on the chemical abun-dance and excitation state of the ionized gas, and suffers alarger amount of dust extinction than H α . Therefore, unlikethe Balmer recombination lines, [O ii ] is not directly pro-portional to the SFR and must be calibrated either empir-ically (Kennicutt 1992; Kewley et al. 2004) or theoretically(Barbaro & Poggianti 1997; Charlot & Longhetti 2001). Onaverage, it has been found thatSFR [M ⊙ / yr] ≃ A × − L OII [erg / s] . (2)The value of A is of order unity, with the precise value de-pending on the broad band luminosity and dust content ofthe galaxy, etc. (Moustakas, Kennicutt, & Tremonti 2006). Strong Mg ii absorbers can be detected in SDSS quasar spec-tra over the redshift range 0 . . z . .
2. The spectroscopic observations make use of 3 ′′ fibers receiving photons origi-nating from the entire path to the quasars (convolved withthe seeing of the observations). The detection of a strongMg ii absorber indicates the presence of a galaxy close tothe line-of-sight. If this galaxy is starforming, its narrowemission lines may be identified on top of the quasar spec-tral energy distribution. While the direct detection of theselines is possible only for a small number of absorber systemsat low redshift (see Noterdaeme, Srianand, & Mohan 2009),a statistical approach enables the measurement of the meanluminosity of such lines well below the noise level of individ-ual spectra and allow us to study the relationship betweenemission and absorption lines. The analysis makes use of a sample of Mg ii systems com-piled by Quider et al. (2010) using the method presented inNestor et al. (2005) and extended to the SDSS DR4 dataset(Adelman-McCarthy et al. 2006). In this section we brieflysummarize the main steps involved in the absorption linedetection procedure. We refer the reader to Nestor et al.(2005) for more detail.For each quasar spectrum in the SDSS DR4 database,Quider et al. (2010) fit a pseudo-continua in an iter-ative fashion using a combination of cubic-splines andGaussians for both emission and absorption features. Thecontinuum-normalized SDSS QSO spectra were searchedfor Mg ii λλ , W . All candidates are interactivelychecked for false detections, a satisfactory continuum fit,blends with absorption lines from other systems, and spe-cial cases. The identification of Mg II doublets required thedetection of the 2796 line and at least one additional line, c (cid:13)000
2. The spectroscopic observations make use of 3 ′′ fibers receiving photons origi-nating from the entire path to the quasars (convolved withthe seeing of the observations). The detection of a strongMg ii absorber indicates the presence of a galaxy close tothe line-of-sight. If this galaxy is starforming, its narrowemission lines may be identified on top of the quasar spec-tral energy distribution. While the direct detection of theselines is possible only for a small number of absorber systemsat low redshift (see Noterdaeme, Srianand, & Mohan 2009),a statistical approach enables the measurement of the meanluminosity of such lines well below the noise level of individ-ual spectra and allow us to study the relationship betweenemission and absorption lines. The analysis makes use of a sample of Mg ii systems com-piled by Quider et al. (2010) using the method presented inNestor et al. (2005) and extended to the SDSS DR4 dataset(Adelman-McCarthy et al. 2006). In this section we brieflysummarize the main steps involved in the absorption linedetection procedure. We refer the reader to Nestor et al.(2005) for more detail.For each quasar spectrum in the SDSS DR4 database,Quider et al. (2010) fit a pseudo-continua in an iter-ative fashion using a combination of cubic-splines andGaussians for both emission and absorption features. Thecontinuum-normalized SDSS QSO spectra were searchedfor Mg ii λλ , W . All candidates are interactivelychecked for false detections, a satisfactory continuum fit,blends with absorption lines from other systems, and spe-cial cases. The identification of Mg II doublets required thedetection of the 2796 line and at least one additional line, c (cid:13)000 , 000–000 B. M´enard et al.
Figure 2.
Median flux in erg/s/cm /˚A from composite spectrain the region of the [O ii ] emission line, for five bins of Mg ii restequivalent width. The red line is the best fit emission doublet.The spectra have been offset for display purposes. the most convenient being the 2803 doublet partner. A 5 σ significance level was required for all λ σ significance level for the corresponding λ − α emission were accepted. From this catalogue weselect systems where both Mg ii and [O ii ] λλ . < W < . . < z < . observed distributionof absorber rest equivalent width W . As described above,the intrinsic distribution of rest equivalent widths follows anexponential distribution down to W = 0 . ii ] luminosities In general, the [O ii ] emission lines from galaxies with z & . ii ] emission of samples ofMg ii absorbers by creating composite spectra in the ab-sorber rest frame. In this procedure masks are applied toabsorption lines expected at the redshift of all known Mg ii absorber systems as well as sky emission lines. The contin-uum is defined with an iterative running median of sizesranging from 500 to 15 pixels. Narrow features caused byunidentified systems with z = z abs cannot be masked witha reliable uniformity, and are thus treated as an additionalnoise contribution. Each continuum-subtracted spectrum isconverted into units of luminosity surface density (see be-low) and samples of spectra are combined to create averageand median composites. In general the results from thesetwo estimators are found to be similar. Since the formerones are usually noisier we will focus only on median valuesbelow.In each composite spectrum the [O ii ] λλ < λ rest < ii ]line are shown in Fig. 2 for samples of Mg ii absorbers withdifferent rest equivalent widths. We note that no shift in thecentroid of the [O ii ] lines with respect to those of the Mg ii lines is detected. These results indicate a strong correlationbetween [O ii ] emission and Mg ii absorption rest equivalentwidth. This correlation will be analyzed quantitatively be-low. The available sample of absorbers for which both Mg ii and[O ii ] lines are simultaneously accessible spans a substantialredshift interval: from 0.4 to 1.3. In this range, the physicalarea corresponding to a given angular aperture varies by afactor of about 2.5. This is illustrated in Fig. 7 in the Ap-pendix in which we show the physical size covered by the3 ′′ diameter SDSS fibre aperture as a function of redshift.In order to combine coherently, or compare, signals fromdifferent redshifts, we estimate the [O ii ] luminosity surfacedensity, Σ L OII . For each Mg ii absorber, we thus convert thespectrum from flux units into luminosity units, at the red-shift of the absorber. We then divide by the surface area ofthe SDSS fibre at this redshift, to obtain luminosity surfacedensity:Σ L OII ( z ) = L OII ( z )Ω f D A , (3)where Ω f = πθ f and θ f is the angular radius of the SDSSfibre and D A ( z ) is the angular diameter distance at the red-shift of the absorber. c (cid:13) , 000–000 robing star formation with absorption line systems Figure 3.
Left: mean [O ii ] luminosity surface density, Σ L OII (averaged within the 3-arcsec fibre) as a function of Mg ii rest equivalentwidth W . The solid line is the best fit power-law. Right: mean [O ii ] luminosity surface density measured in three redshift bins, as afunction of W . The dashed lines show the best 3-parameter fit to the entire sample. ii ABSORPTION AS A TRACER OFSTAR FORMATION
In this section we use Mg ii absorbers as tracers and showthey can be used as a probe of star formation. This is doneindependently of any relation between Mg ii absorbers andgalaxies. ii - [O ii ] correlation Using a logarithmic binning in Mg ii rest equivalent width, W MgII0 , we create composite spectra in the absorber rest-frame and estimate Σ L OII , the median [O ii ] luminosity sur-face density. The observed relation between these two pa-rameters is presented in Fig. 3 with red data points. Mg ii absorbers with equivalent widths 0 . < W < ii ] emission, spanning more thana factor 30 in luminosity surface density. The observed re-lation, averaged over the redshift range 0 . < z < .
3, issimply described by h Σ L OII ( W ) i = A „ W « α (4)with A = (1 . ± . × erg s − kpc and α = 1 . ± .
11. This fit is shown in Figure 3 with the solid line.This result shows that, without any prior knowledge ofthe underlying galaxy distribution, Mg ii absorbers can beused to trace the distribution of [O ii ] emission, i.e. star for-mation. This property is interesting since the detectabilityof absorber systems does not depend on redshift and is notsignificantly affected by dust extinction. They can therefore provide us with a powerful probe of star formation over cos-mological times. This will be discussed in more detail below. In order to investigate the redshift dependence of the aboverelation, we measure the mean [O ii ] luminosity surface den-sity for three redshift intervals: 0 . < z < . < z < . < z < .
3. The results are shown in the right panel ofFig. 3 where each color corresponds to a given range of Mg ii rest equivalent widths, resulting in 12 independent datapoints. Remarkably, they appear to be distributed regularlyin the L OII − z plane. This property further demonstratesthat the relation between Mg ii absorption properties andobserved [O ii ] luminosity surface density holds over a sig-nificant cosmological time span. This figure illustrates thatMg ii absorbers are found, on average, in regions having apredictable value of [O ii ] luminosity surface density. Thiscorrespondence is valid over a large range in both redshiftand [O ii ] luminosity.A global fit to all the data points in the redshift- W plane gives: h Σ L OII ( W , z ) i = A „ W « α (1 + z ) β (5)with A = (7 . ± . × erg s − kpc , α = 1 . ± . β = 1 . ± .
4. The results of this global fit are shown inthe right panel of Fig. 3 as dotted lines. Using this simpleformula, we are able to fit 12 data points with 3 parametersresulting in a reduced χ of 1 .
3. The redshift dependence ofthis relation is remarkably weak. The mean [O ii ] luminos-ity surface density changes by less than 50% over a unity c (cid:13) , 000–000 B. M´enard et al.
Figure 4.
Left:
Relative contribution of [O ii ] luminosity surface density, Σ L OII as a function of Mg ii rest equivalent width W . Thesolid line is the best fit given by Eq. 9. Right: the [O ii ] luminosity volume density from Mg ii absorbers (with W > . ii absorbers (d N/ d z ) is measured. The other data points show direct [O ii ] luminosity density measurements fromthe literature. Note that no dust correction has been applied to any of these [O ii ] luminosity measurements, allowing direct comparisonbetween samples. Mg ii absorbers appear to trace a substantial fraction of the global [O ii ] luminosity density in the Universe. redshift range, corresponding to a timescale of about 5 Gyr.Understanding the origin of this dependence would be ofparticular interest and would allow us to directly use Mg ii absorbers as a probe of star formation without the need ofdetecting [O ii ] emission. The SDSS fibers used to observe quasars can be consideredas a set of cones integrating light emitted along the line-of-sight. Each quasar spectrum therefore probes a given volumeand the set of (high-redshift) quasar spectra can be con-sidered as a random sampling of the low-redshift Universe.Within this volume, we estimate the [O ii ] luminosity den-sity, L OII , traced by Mg ii absorbers. Note that we are onlyinterested in estimating a density and therefore do not needto include any emission contribution originating outside thefibers. We then use L OII to estimate the star formation ratedensity probed by these systems as a function of redshiftand compare it to the total value in the Universe.The co-moving [O ii ] luminosity density probed by Mg ii absorbers is given by L OII ( z ) = LV C = 1(1 + z ) d LD A ( z ) dΩ d N d r (6)where L is the [O ii ] luminosity associated with absorbersobserved within the co-moving volume V C , which is accessi-ble along the pathlength towards the background quasars (with and without absorbers). In the second expression, D A ( z ) (1 + z ) dΩ is the co-moving surface of the aperture,d N/ d r is the mean number of absorbers expected withina pathlength d r and d L is the [O ii ] luminosity observedwithin the corresponding volume element. Introducing thecosmology-dependent function E ( z ) = d z d r z ) = H c p Ω M (1 + z ) + Ω Λ (1 + z ) (7)we can simply express the luminosity density probed by apopulation of Mg ii absorbers as a function of observablequantities L OII ( z ) = Σ L OII ( z ) d N d z E ( z ) , (8)where d N/ d z is the line-of-sight incidence of absorber sys-tems and Σ L OII the mean luminosity surface density associ-ated with them. Eq. 8 indicates that, for a given cosmology, the [O ii ] luminosity density traced by Mg ii absorbers can bederived without any assumption . In particular it does notdepend on the spatial extent of the gas and the relationbetween absorbers and galaxies. A similar formalism wasused by Wolfe, Gawiser, & Prochaska (2003) and Wild etal. (2007) to estimate the contribution of damped Lyman- α systems and CaII absorption systems to the overall starformation rate.In order to account for the varying completeness of theabsorber sample with W , we estimate Eq. 8 by integratingover the intrinsic rest equivalent width distribution c (cid:13) , 000–000 robing star formation with absorption line systems L OII ( z ) = E ( z ) Z d W d N d W d z Σ L OII ( W , z ) (9)where d N / d W d z is the completeness-corrected distribu-tion of rest equivalent widths given by Nestor et al. (2005),see Eq. 1. The relative contribution of the integrand is shownin the left panel of Fig. 4 for the sample spanning the entireredshift range. It shows that most of the [O ii ] luminositysurface density, and therefore most of the star formation, isassociated with systems having W ∼ ii absorbers (M´enard et al. 2008). We will return to this pointin the discussion. We note that estimating the integral inEq. 9 from the lowest rest equivalent width available in thisanalysis, i.e. W = 0 . L OII using Eq. 9 for the three redshiftbins defined above and the results are given in Table 1. Theresults are presented in the right panel of Figure 4. The or-ange data points show direct measurements for our sampleof Mg ii absorbers with W > . ii absorbers, i.e. including the con-tribution from weaker systems. For comparison, we show acompilation of measurements of the total [O ii ] luminositydensity obtained from narrow band filter and emission linesurveys. Note that no dust correction has been applied toany of these datasets, allowing direct comparisons betweenobserved quantities. This figure shows that, at z ∼ theamount of [O ii ] luminosity density traced by Mg ii absorbersis a substantial fraction of the total density in the Universe estimated from direct emission measurements. Given thenon-negligible scatter in the reported values of L OII fromphotometric and spectroscopic surveys at around z = 1, itis not yet possible to precisely quantify the fraction of [O ii ]luminosity traced by Mg ii absorbers. Nevertheless our re-sults show that at least half of the [O ii ] luminosity densityis traced by strong Mg ii absorbers. This implies that a highfraction of star-forming galaxies are associated with strongMg ii absorbers and vice-versa. In other words, the presenceof Mg ii absorbers around a galaxy depends on its (recent)star formation rate.To further illustrate this connection, we explore the red-shift dependence of L OII . As shown above, this quantity isgiven by the product of two observables: d N / d W d z , whichhas been accurately measured up to z = 2 . L OII ( W , z ) which is parametrized in Eq. 5 up to z = 1 . L OII up to z = 2 .
2, we use an extrapola-tion of Eq. 5, a relation which is weakly redshift dependent.The result is shown in Fig. 4 with the solid line. The over-all redshift dependence appears to be in agreement with thetrend given by direct [O ii ] emission surveys. We now show how the global star formation probed by Mg ii systems compares to the overall star formation history inthe Universe estimated by various techniques. As mentionedabove, [O ii ] luminosity can be used as an estimator of star redshift L OII for W > . L OII for all W interval [10 erg s − Mpc − ] [10 erg s − Mpc − ]0 . < z < .
67 1 . ± .
13 1 . ± . . < z < .
99 1 . ± .
17 1 . ± . . < z < .
30 1 . ± .
24 2 . ± . Table 1. [O ii ] luminosity density traced by Mg ii absorbers as afunction of redshift. No dust corrections have been applied. Figure 5.
Star formation rate density as a function of redshift.The gray data points are from the Hopkins (2004) compilationand use various types of estimators. The orange and red datapoints show the SFR density traced by W > . ii absorbers. The red curve shows the SFR density estimate usingd N/ d z measured up to z = 2 . ii ]-Mg ii relation. formation rate, however the scaling between these two quan-tities is not straightforward: it is based on empirical rela-tions, calibrated at various redshifts and subject to dust ex-tinction corrections (Kennicutt 1998; Moustakas, Kennicutt,& Tremonti 2006). Here we choose to apply the average scal-ing coefficient used in a recent analysis by Zhu et al. (2009)for galaxies with 0 . < z < .
45 (see their Table 2). Theseauthors made use of the empirical correlation derived byMoustakas et al. (2006) between the absolute B-band mag-nitude, and the L[O ii ]/SFR ratio. This calibration statisti-cally accounts for the gross systematic effects of reddening,metallicity, and excitation, all of which correlate with opti-cal luminosity, and has been shown to work reasonably wellfor star-forming galaxies at 0 . < z < . . < z < .
45, we get˙ ρ ⋆ = 4 . − × L OII M ⊙ yr − Mpc − (10) c (cid:13) , 000–000 B. M´enard et al. where L OII is in erg s − . With the limitations of such ascaling in mind, we convert our [O ii ] luminosity density es-timate into a star formation rate density and present theresults in Figure 5. The gray data points show the estimatessummarized in the Hopkins (2004) compilation . We showthe estimated star formation rate density from Mg ii ab-sorbers with W > . ii ] - Mg ii scal-ing given in Eq. 5 and d N/ d z measurements available up to z = 2 .
2. As can be seen, both the overall amplitude and theredshift dependence are in good agreement with other starformation rate estimators. The global star formation probedby Mg ii absorbers can therefore be used as an independentprobe of the star formation history in the Universe.Interestingly, extrapolating the comparison between theSFR density traced by Mg ii absorbers and observations ofthe global SFR density to higher redshifts suggests that theproduct Σ L OII × d N/ d z (see Eq. 8) should reach a max-imum value at around redshift two and then decrease athigher z . Since Σ L OII ( z, W ) appears to be weakly redshiftdependent, the expected decrease in this product should bedriven by a decrease in d N/ d z beyond redshift two. This canbe tested observationally using near-infrared spectroscopy ofan ensemble of quasars, with instruments such as X-shooter(D’Odorico et al. 2006) or the upcoming FIRE spectrographSimcoe et al. (2008). ii ] LUMINOSITY FUNCTION ANDTHE DISTRIBUTION OF Mg ii RESTEQUIVALENT WIDTHS
In the previous section we made use of Mg ii absorbers astracers and showed that, independently of their relations togalaxies, they can be used as a probe of star formation inthe Universe. The connection between Mg ii absorbers andstar formation can be explored one step further. We nowshow that the distribution of Mg ii rest equivalent widthsd N/ d W , and the [O ii ] luminosity function, i.e. a measureof the probability distribution function (p.d.f.) of star forma-tion rate in the Universe, are connected, both in amplitudeand shape. Our goal here is only to reveal the existence ofthis relation and explore its implications. A more detailedcomparison and study of possible deviations is left for futurework. Moreover, as explained below, the scatter in the re-ported values of the [O ii ] luminosity function (see Fig. 5 and6) currently limits the level to which we can meaningfullyrefer to the amplitude of the [O ii ] luminosity function.Given the existence of a relation between the mean [O ii ]luminosity surface density and W we can express, in a sta-tistical sense, the [O ii ] luminosity function Φ( L OII ) in termsof W (Mg ii ). We can write:Φ(log L ) = d N d log L d V C The more recent compilation by Hopkins et al. (2006) focuses onstar formation measurements derived from rest-frame UV obser-vations rather than emission-line based star formation estimates.It is therefore less appropriate for our comparison. = d N d W d z d W d log L E ( z ) C σ (11)where L represents the mean [O ii ] luminosity per galaxytraced by a strong Mg ii absorber, σ is the cross-section forabsorption and C is the covering factor for Mg ii absorp-tion within this cross-section, which can be expressed as theproduct of the fraction of galaxies giving rise to Mg ii ab-sorption and the spatial cross-section taking into account theangular variations and clumpiness of the gas: C = C f C Ω .The function E ( z ) carries the dependence on cosmology, asdefined in Eq. 7.In order to show the connection between the distribu-tion of Mg ii rest equivalent widths and the [O ii ] luminos-ity function, we use a simplifying assumption: we considerthat the cold gas distribution around star forming galaxiesis isothermal, as motivated by Chelouche et al. (2008). Inthis case, the 2-dimensional density distribution goes like ρ ∝ b − and the probability of observing a system at animpact parameter b , ρ ( b ) b d b , is constant up to a maxi-mum impact parameter b max . The uniform cross-section isthen simply given by σ = π b . The fraction of galaxy lightfalling within the fiber of projected radius r f is, on average,given by ( r f /b max ) and the mean [O ii ] luminosity per Mg ii galaxy is L OII , gal = Σ L OII σ , where Σ L OII is the mean [O ii ]luminosity surface density derived above. This allows us toexpress the luminosity function as:Φ(log L ) = d N d W d z d W d log(Σ L OII ) E ( z ) C σ = d N d W d z E ( z ) α C σ , (12)with α = d log Σ L OII / d W . Using this approximation, wenow compute the [O ii ] luminosity function expected fromMg ii absorbers, as a function of W . The first two terms ofthe above equation are observable quantities, described byEq. 1 and 5. Various observations of Mg ii absorbers-galaxyassociations indicate that the gas extends up to projectedradii of about 50 kpc (Steidel et al. 1995, Zibetti et al. 2007,Chen & Tinker 2008). We will therefore use b max = 50 kpcto characterize the cross-section in our estimate of Eq. 12.We compute Eq. 12 at z = 1 .
2, a redshift at whichseveral direct measurements of the [O ii ] luminosity func-tion are available (see below). We first use C = 1, implyingthat all star forming galaxies are surrounded by Mg ii gas.The resulting [O ii ] luminosity function is shown in the leftpanel of Figure 6 with the solid red line. For comparison,we show several estimates of the [O ii ] luminosity functionat the same redshift from narrow-band imaging surveys of[O ii ] emitting galaxies in the HST COSMOS 2 square de-gree field using the Suprime-Cam on the Subaru Telescope(Takahashi et al. 2007) as well as from Ly et al. (2007). Theoverall agreement, both in shape and amplitude, is striking.The Mg ii rest equivalent width distribution appears to berelated to the [O ii ] luminosity function, and therefore thep.d.f. of star formation rate. The connection between Mg ii absorption and star formation is not only seen as a functionof redshift, as shown through estimates of the [O ii ] luminos-ity density L OII , but appears to also operate as a functionof W .Comparing the red curve to the set of blue curves c (cid:13) , 000–000 robing star formation with absorption line systems Figure 6.
The connection between the [O ii ] luminosity function, Φ( L ), and the distribution of Mg ii rest equivalent width, dN / d W . Left: Φ( L ) derived from the distribution of Mg ii absorbers (solid red) using the observed scaling relation between [O ii ] emission andMg ii absorption, dN / d W estimated at z ≃ . ii ] luminosity functions are shown for comparison. No dust corrections are applied allowing direct comparisons. Right: Mg ii rest equivalent width distribution at z ≃ . N / d W d z based on the observed [O ii ] luminosity functions. reveals a number of interesting properties: first of all wecan see that the above relation maps Mg ii absorbers with1 . W . ii ] line luminosities ranging from 10 to about 5 × erg/s, i.e. the range corresponding to theexponential decline of the [O ii ] luminosity function. Second,the match in amplitude shows that the incidence of strongMg ii absorbers can be explained by the number density of[O ii ] bright galaxies and finally the shape of the two dis-tributions are in good agreement, further indicating thatthese emission and absorption processes are related. Thesetwo quantities ultimately allow us to probe the same phe-nomenon: star formation.In Eq. 12, the terms d N / d W d z , α and E ( z ) are al-ready constrained by direct observations and a set of cosmo-logical parameters. Some constraints on the spatial extent ofthe cross-section σ and the covering factor C already existbut are less robust. Keeping in mind that the scatter in thereported values of Φ( L ) only allows us to compare distribu-tions within a factor ∼
2, we now comment on these twoquantities. Changing the value of the covering factor C inEquations 11 and 12 results in shifting the curves vertically.By definition, the amount of [O ii ] luminosity density tracedby Mg ii absorbers must be lower than the total value, esti-mated from direct measurements. Depending on the chosendataset, Figure 6 shows that the amplitude of the [O ii ] lu-minosity function implied by Mg ii absorbers is either com-parable to the measured Φ( L ) or lower by about a factortwo. This puts a strong constraint on the covering factor C = C f C Ω ≃ . − . (13) Therefore, the covering factor for Mg ii absorption aroundstar forming galaxies (with L OII & erg / s) is expectedto be high, possibly close to unity. Using a different value ofthe cross-section σ has several effects: it changes both therelation between L and W , and the amplitude of the pre-dicted Φ( L ), shifting the prediction diagonally and changingits curvature. We find that only values of 20 . b max .
60 kpcdo not over-predict the observed luminosity density at cer-tain values of L OII . The match between these two sets ofcurves is possible for only a restricted range of impact pa-rameters, which encloses the mean value reported by directobservations of Mg ii -galaxy associations.We now look at the relation in Eq. 11 from the oppositedirection, i.e. we use it as a model for the distribution ofMg ii rest equivalent width. Using C = 1, as discussed above,we can writed N d W d z = Φ(log L ) α σ E ( z ) . (14)This relation provides a model explaining both the ampli-tude, shape and redshift-dependence of the Mg ii rest equiva-lent width distribution, with no free parameter. The inferreddistribution of Mg ii rest equivalent widths is shown in theright panel of Figure 6 with blue lines using the three mea-surements of the [O ii ] luminosity function introduced above.For comparison, we show the measurement of d N / d W d z from SDSS obtained by Nestor et al. (in prep.) for systemscentered at z = 1 .
2, with a half width of δz = 0 .
2. This Fig-ure allows us to illustrate the accuracy with which the latterdistribution is measured with current datasets. The estimate c (cid:13) , 000–000 B. M´enard et al. of d N/ d W d z originates from the analysis of about 45,000SDSS quasar spectra spread over a large fraction of the sky.It therefore does not suffer from cosmic variance.While we have focused on the similarities between theshape and amplitude of Φ(log L ) and d N / d W d z , we nowcomment on possible departures between these quantities.As surveys become larger and selection effects better un-derstood, a more robust estimate of the [O ii ] luminosityfunction will allow us to explore possible deviations fromthe predictions given by Eq. 11. As discussed above, the ra-tio between the predicted and observed Φ( L ) will provideus with an estimate of the covering factor for absorptionaround star forming galaxies. Deviations in shape could alsoindicate variations of the cross-section as a function of [O ii ]luminosity and/or Mg ii rest equivalent width. As mentionedabove, the overall distribution of Mg ii W can be describedby a Schechter function, with W ⋆ ≃ . z = 1. Observa-tional constraints exist down to about 0.1 ˚A. If the corre-spondence between W and L OII holds in this regime, thiswould provide us with a mean of probing the [O ii ] luminos-ity function down to values significantly fainter than currentemission derived estimates at these redshifts.In summary, we have shown that the distribution ofMg ii rest equivalent widths and the [O ii ] luminosity func-tion track each other, as a function of both redshift and W , suggesting that these two observables probe the samephenomenon: star formation. This connection allows us topresent the first non-parametric model for the distributionof Mg ii rest equivalent widths, d N/ d W d z , which appearsto inherit its amplitude, shape and redshift dependence fromthe [O ii ] luminosity function or equivalently the p.d.f. of starformation in the Universe. As mentioned earlier, the nature of strong Mg ii absorbershas been a matter of debate for more than two decades. Oneway to address this question has been to look for observa-tional correlations between absorber and galaxy properties.However, until recently no compelling detection had beenreported. This analysis has allowed us to show that Mg ii absorbers trace a substantial fraction of the overall star for-mation rate of the Universe. This property actually shedslight on why previous attempts to detect correlations be-tween W (MgII) and galaxy secular properties have eitherfailed, or reported rather weak trends (e.g. Kacprzak et al.2007). As shown by Brinchmann et al. (2004) at z < .
2, therelation between SFR and galactic stellar mass (measuredover five decades) is relatively shallow: SFR ∝ M . ∗ . At highredshifts, observational constraints on the SFR-stellar massrelation are limited to massive galaxies (Noeske et al. 2007,and Daddi et al., 2007) but indicate a large scatter in SFRat a given mass. Therefore selecting galaxies with respectto their SFR, or equivalently W , will result in a samplewith a wide range in stellar mass, broad band luminosityand colour. Correlations between absorber rest equivalentwidths and galaxy secular parameters such as broad band luminosity or mass are thus expected to be weak, in agree-ments with results from Steidel et al. (1997); Zibetti et al.(2007); M´enard & Chelouche (2009); Tinker & Chen (2008)and Gauthier, Chen, & Tinker (2009). ii absorbers The nature of Mg ii absorbers has been long been debatedand so far no consensus has been reached regarding the phys-ical process governing the observed equivalent widths. In thiswork we have shown the connection between the presence ofMg ii absorbers in galactic halos and nearby star formation.Gas around a galaxy can be related to star formation in twodifferent ways: it can be infalling and feeding an episode ofstar formation or it can be outflowing, due to star forma-tion feedback processes such as supernova winds or radiationpressure.Enhancements of star formation are expected to be trig-gered on a timescale of order the dynamical time ( τ d = R/v c ∼ yr) following gas accretion. The time requiredfor these systems to travel a distance r of 20 kpc is∆ t = 60 Myr „ r
20 kpc « „ v
300 km s − « − , (15)where the value of the characteristic velocity is motivatedby observations of blue-shifted Mg ii absorption in the spec-tra of star-forming galaxies (Weiner et al. 2009) or usingthe observed Mg ii rest equivalent width as a proxy for theprojected velocity: ∆ v ≃
120 ( W / ˚A) km/s (see section 2.1).Such velocities are however comparable to the circular ve-locity of L ⋆ galaxies and can also be reached by infallingmaterial. The above relation indicates that the timescale forthe gas to travel is comparable to the timescale of a starformation episode. Note that we would not expect any cor-relation between W and instantaneous star formation rateif τ d ≪ ∆ t .The high covering factor around star forming galaxiesimplied in our analysis allows us to shed light on the di-rection of motion of the gas. Although there is currentlyno direct observational evidence for infalling cold gas ontogalaxies, results from numerical hydrodynamic simulationssuggest that gas accretion leading to efficient star forma-tion mostly occurs through cold flows, i.e. dense filamentarystructures penetrating through the diffuse gaseous halos ofgalaxies (Keres et al. 2005; Dekel et al. 2009). Such modelssuggest that these elongated structures give rise to coveringfactors of order 10-20%. In Section 5 we showed that, in or-der not to over-predict the observed [O ii ] luminosity func-tion, the covering factor for Mg ii absorption around starforming galaxies is necessarily high, with C = C f C Ω & . ii absorbers and starformation, and the recent results by Weiner et al. (2009) c (cid:13) , 000–000 robing star formation with absorption line systems who showed that blue-shifted Mg ii absorption, i.e. outflow-ing gas, is ubiquitous in star-forming galaxies at z ∼ B -band luminosity, and a factor 30in stellar mass. Their blue-shifted Mg ii systems appear toshare all the properties of the intervening systems used inour study: rest equivalent widths, redshift range, high cov-ering factor and especially the correlation with star forma-tion. Occam’s razor (c. 1320), which tells us that entitiesshould not be multiplied unnecessary, leads us to postulatethat the nature of the Mg ii gas is the same in both types ofanalyses. In other words, blue-shifted Mg ii absorbers seen inself absorption and intervening Mg ii absorbers might be thesame gas clouds. This implies that the blue-shifted absorp-tion seen in star forming galaxies (e.g. Heckman, Armus, &Miley 1990; Martin 1999; Weiner et al. 2009 may well extendto radii up to tens of kpc. Our analysis has shown that the bulk of star formation istraced by Mg ii absorbers. This connection provides us witha new tool to explore star formation in the Universe, overcosmological times. Here we discuss its potential and high-light differences with existing techniques. • Absorption measurements offer a number of advan-tages: the detectability of absorption lines does not dependon redshift (while galaxy fluxes drop rapidly at high red-shifts) and perhaps more importantly, the observed proper-ties of absorption lines are not sensitive to dust extinction.Mg ii absorbers can therefore provide us with a powerful andcomplementary tracer of star formation. • Absorber systems allow us to apply emission line studiesin a noise-dominated regime. Only a handful of low-redshiftMg ii absorbers allow for the direct detection of [O ii ] emis-sion lines in individual systems. For the vast majority, nodirect detection is possible in SDSS quasar spectra. How-ever, being able to average over thousands of lines-of-sightsallows us to estimate the mean [O ii ] luminosity of the corre-sponding population, down to emission levels comparable towhat is currently obtained with some of the deepest narrowband surveys. • This technique does not suffer from contamination frommisidentified emission lines, a severe limitation for narrow-band imaging surveys of [O ii ] emission (see Takahashi et al.2007; Ly et al. 2007). • In optical spectra Mg ii absorbers are detected fromthe ground over the range 0 . < z < .
2. Infrared spec-troscopy can extend this redshift path up to the reioniza-tion epoch. For example, the upcoming near-infrared spec-trograph FIRE (Simcoe et al. 2008) will allow the detec-tion of Mg ii absorption and [O ii ] emission up to z ∼ with the same tracer ,and provide a new set of constraints on the decline of theSFR density at high redshift. A vast amount of Mg ii data has been accumulated overthe years. Its analysis has mainly focused on attempting toreveal the nature of these systems. We can now use thesedatasets to obtain new insights into star formation, overcosmological times. About 16,000 Mg ii absorbers have beendetected in SDSS quasar spectra Quider et al. (2010). Theanalysis of the entire sample (DR7) will double this number.The details of Mg ii absorbers (number of components, ve-locity widths, etc.) accessible through high-resolution spec-troscopy can be used to put constraints on star formationprocesses. Understanding this type of information might al-low us to constrain the recent star formation history ofgalaxies. Such a technique will be complementary to spectralenergy distribution analyses.More work needs to be done in order to understandthe origin of the empirical [O ii ]-Mg ii scaling relation andespecially its redshift dependence. If the redshift depen-dence of Σ L OII ( W ) can be explained, Mg ii absorbers canthen, alone, be used to trace star formation in the Uni-verse. Finally, comparing SFR( z ) from emission and absorp-tion techniques might shed light on the redshift dependenceof the dust extinction affecting emission based estimators.With appropriate wavelength coverages, similar absorption-emission scaling relations can be explore, for example be-tween Mg ii absorption and H- α emission. It will be inter-esting to extend the current analysis to other absorptiontransitions (FeII, CIV, etc.) and test their connection to theprocess of star formation. We present an empirical connection between cold gas ingalactic halos and star formation. Using a sample of morethan 8,500 Mg ii absorbers from SDSS quasar spectra, wereport the detection of a 15 σ correlation between the restequivalent width W of Mg ii absorbers and the associated[O ii ] luminosity, an estimator of star formation rate. Thiscorrelation allows us to show that(i) Mg ii absorbers trace a substantial fraction of the [O ii ]luminosity density in the Universe, i.e. global star formation.The SFR density probed by Mg ii absorbers recovers theoverall star formation history of the Universe derived fromclassical emission estimators up to z ∼ ii rest equivalentwidths, d N/ d W inherits both its shape and amplitude fromthe [O ii ] luminosity function Φ( L ). These distributions canbe naturally connected, without any free parameter.(iii) the covering factor for cold gas around star forminggalaxies appears to be high: C & .
5. The determination ofthis value is currently limited by the scatter in the reportedestimates of [O ii ] luminosity densities.These results not only shed light on the nature of Mg ii ab-sorbers but also provide us with a new probe of star forma-tion, in absorption, i.e. in a way which does not suffer fromdust extinction and with a redshift-independent sensitivity.As shown in this analysis, such a tool can be applied in anoise-dominated regime, i.e. using a dataset for which emis- c (cid:13) , 000–000 B. M´enard et al. sion lines are not detected in individual objects. This is ofparticular interest for high redshift studies.We argue that outflows appear as a favored explanationregarding the nature of Mg ii absorbers and that blue-shiftedMg ii absorption seen in the spectra of star forming galax-ies could be the same systems, implying that the observedoutflowing gas can reach radii of ∼
50 kpc.
ACKNOWLEDGEMENTS
We thank David Turnshek and Sandhya Rao for their cru-cial role the compilation of the Mg ii absorber catalog basedon SDSS data. We are grateful to the authors of the Ly etal. (2007) for having provided us with the values of theirobserved [O ii APPENDIX: FIBER APERTURE CORRECTIONReferences
Adelman-McCarthy J. K., et al., 2006, ApJS, 162, 38Bahcall J. N., Spitzer L. J., 1969, ApJ, 156, L63Barbaro G., Poggianti B. M., 1997, A&A, 324, 490Bergeron J., 1986, A&A, 155, L8Bergeron J., Boiss´e P., 1991, A&A, 243, 344Bohlin R. C., Savage B. D., Drake J. F., 1978, ApJ, 224,132Boissier S., Boselli A., Buat V., Donas J., Milliard B., 2004,A&A, 424, 465Bond N. A., Churchill C. W., Charlton J. C., Vogt S. S.,2001, ApJ, 562, 641Bouch´e N., Murphy M. T., P´eroux C., Davies R., Eisen-hauer F., F¨orster Schreiber N. M., Tacconi L., 2007, ApJ,669, L5Bowen D. V., Blades J. C., Pettini M., 1995, ApJ, 448, 662Bowen D. V., Blades J. C., Pettini M., 1996, ApJ, 464, 141Bruzual G., Charlot S., 2003, MNRAS, 344, 1000
Figure 7.
Physical radius mapped by the SDSS spectroscopicfiber, as a function of redshift
Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345,245Chelouche D., M´enard B., Bowen D. V., Gnat O., 2008,ApJ, 683, 55Charlot S., Longhetti M., 2001, MNRAS, 323, 887Churchill C. W., Kacprzak G. G., Steidel C. C., 2005,pgqa.conf, 24Dav´e, R., & Oppenheimer, B. D. 2007, MNRAS, 374, 427Dekel A., et al., 2009, Natur, 457, 451D’Odorico S., et al., 2006, SPIE, 6269,Ellison, S. L., Churchill, C. W., Rix, S. A., & Pettini, M.2004, ApJ, 615, 118Ellison S. L., 2006, MNRAS, 368, 335Ellison S. L., et al., 2006, MNRAS, 372, L38Gallagher J. S., Hunter D. A., Bushouse H., 1989, AJ, 97,700Gauthier J.-R., Chen H.-W., Tinker J. L., 2009, ApJ, 702,50Guillemin, P., & Bergeron, J. 1997, A&A, 328, 499Heckman T. M., Armus L., Miley G. K., 1990, ApJS, 74,833Hopkins A. M., et al., 2003, ApJ, 599, 971Hopkins A. M., 2004, ApJ, 615, 209Hopkins A. M., Beacom J. F., 2006, ApJ, 651, 142Issa M. R., MacLaren I., Wolfendale A. W., 1990, A&A,236, 237Jenkins E. B., Savage B. D., Spitzer L., Jr., 1986, ApJ, 301,355Kacprzak G. G., Churchill C. W., Steidel C. C., MurphyM. T., Evans J. L., 2007, ApJ, 662, 909Kennicutt R. C., Jr., 1992, ApJ, 388, 310Kennicutt R. C., Jr., 1998, ApJ, 498, 541Keres D., Katz N., Weinberg D. H., Dav´e R., 2005, MN-RAS, 363, 2Kewley L. J., Geller M. J., Jansen R. A., 2004, AJ, 127,2002P. Khare, D. G. York, D. Vanden Berk, V. P. Kulkarni,et al., Proceedings of the IAU Colloquium 199 ”Prob-ing Galaxies through Quasar Absorption Lines”, astro-ph/0504532 c (cid:13) , 000–000 robing star formation with absorption line systems Ly C., et al., 2007, ApJ, 657, 738Martin C. L., 1999, ApJ, 513, 156M´enard, B. and P´eroux, C., A&A 410, 43M´enard, B., Nestor, D., Turnshek, D., Quider, A.,Richards, G., Chelouche, D., & Rao, S. 2008, MNRAS,385, 1053M´enard B., Chelouche D., 2009, MNRAS, 393, 808Mo H. J., Miralda-Escude J., 1996, ApJ, 469, 589Mouhcine M., Lewis I., Jones B., Lamareille F., MaddoxS. J., Contini T., 2005, MNRAS, 362, 1143Moustakas J., Kennicutt R. C., Jr., Tremonti C. A., 2006,ApJ, 642, 775Murphy M. T., Curran S. J., Webb J. K., M´enager H., ZychB. J., 2007, MNRAS, 376, 673Nestor 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, 637Nestor D. B., Turnshek D. A., Rao S. M., Quider A. M.,2007, ApJ, 658, 185Noterdaeme P., Srianand R., Mohan V., 2009, arXiv,arXiv:0912.0736P´eroux C., Deharveng J.-M., Le Brun V., Cristiani S., 2004,MNRAS, 352, 1291P´eroux C., Meiring J. D., Kulkarni V. P., Ferlet R., KhareP., Lauroesch J. T., Vladilo G., York D. G., 2006, MN-RAS, 372, 369Petitjean, P., & Bergeron, J. 1990, A&A, 231, 309Prochter G. E., Prochaska J. X., Burles S. M., 2006, ApJ,639, 766Quider, A, Nestor, D, Turnshek, D, Rao, S, Monier, E &Weyant, A., 2010, in preparationRao, S. M., Turnshek, D. A., & Nestor, D. B. 2006, ApJ,636, 610Rubin K. H. R., Prochaska J. X., Koo D. C., Phillips A. C.,Weiner B. J., 2009, arXiv, arXiv:0907.0231Savage B. D., et al., 2000, ApJS, 129, 563Simcoe R. A., et al., 2008, SPIE, 7014,Snow T. P., Rachford B. L., Figoski L., 2002, ApJ, 573, 662Steidel C. C., Kollmeier J. A., Shapley A. E., ChurchillC. W., Dickinson M., Pettini M., 2002, ApJ, 570, 526Steidel C. C., Dickinson M., Persson S. E., 1994, ApJ, 437,L75Steidel, C. C., Dickinson, M., Meyer, D. M., Adelberger,K. L., & Sembach, K. R. 1997, ApJ, 480, 568Takahashi M. I., et al., 2007, ApJS, 172, 456Tinker J. L., Chen H.-W., 2008, ApJ, 679, 1218Tremonti C. A., et al., 2004, ApJ, 613, 898Tremonti C. A., Moustakas J., Diamond-Stanic A. M.,2007, ApJ, 663, L77Turnshek D. A., Rao S. M., Nestor D. B., Belfort-MihalyiM., Quider A., 2005, astro, arXiv:astro-ph/0506701Veilleux S., Cecil G., Bland-Hawthorn J., 2005, ARAA, 43,769Vladilo G., Centuri´on M., Levshakov S. A., P´eroux C.,Khare P., Kulkarni V. P., York D. G., 2006, A&A, 454,151Vladilo G., Prochaska J. X., Wolfe A. M., 2008, A&A, 478,701Wang, J., Hall, P. B., Ge, J., Li, A., & Schneider, D. P.2004, ApJ, 609, 589Weiner B. J., et al., 2009, ApJ, 692, 187 Welty D. E., Frisch P. C., Sonneborn G., York D. G., 1999,ApJ, 512, 636Wild, V., Hewett, P. C., & Pettini, M. 2006, MNRAS, 367,211Wild, V., Hewett, P. C., & Pettini, M. 2007, MNRAS, 374,292Wolfe A. M., Gawiser E., Prochaska J. X., 2005, ARA&A,43, 861Wolfe A. M., Gawiser E., Prochaska J. X., 2003, ApJ, 593,235York D. G., Dopita M., Green R., Bechtold J., 1986, ApJ,311, 610York, D. G., et al. 2006, MNRAS, 367, 945Zibetti, S., M´enard, B., Nestor, D., & Turnshek, D. 2005,ApJl, 631, L105Zibetti, S., M´enard, B., Nestor, D. B., Quider, A. M., Rao,S. M., & Turnshek, D. A. 2007, ApJ, 658, 161Zuo L., Beaver E. A., Burbidge E. M., Cohen R. D.,Junkkarinen V. T., Lyons R. W., 1997, ApJ, 477, 568 c (cid:13)000