THE RADIAL AND AZIMUTHAL PROFILES OF Mg ii ABSORPTION AROUND 0.5 < z < 0.9 zCOSMOS GALAXIES OF DIFFERENT COLORS, MASSES, AND ENVIRONMENTS^*

The zCOSMOS survey (Lilly et al. 2007) is a survey in the 2 deg² COSMOS field (Scoville et al. 2007). The zCOSMOS survey is carried out with the VIMOS spectrograph on the ESO UT3 8 m Very Large Telescope. The survey itself is divided into two major components:

• 1.  
  zCOSMOS-bright, which consists of spectra from approximately 20,000 flux limited I_AB ⩽ 22.5 galaxies over the full 2 deg² COSMOS field. At this flux limit, the majority of the observed galaxies have redshifts in the range of 0 < z < 1.4. Observations in zCOSMOS-bright were obtained with the medium resolution grism using 1 arcsec slits, yielding a spectral resolution of R ∼ 600 at 2.5 Å pixel⁻¹. The average accuracy of individual redshifts has been demonstrated to be 110 km s⁻¹ (Lilly et al. 2009). Stellar masses for these galaxies have been estimated by spectral energy distribution fitting, using the Hyperzmass code, a modified version of the photo-z code Hyperz (Bolzonella et al. 2000). We refer the reader to Bolzonella et al. (2010) for a detailed technical description of this mass estimation. The absolute magnitudes used in this study were computed as in Zucca et al. (2009).
• 2.  
  zCOSMOS-deep, targets the central 1 deg² of the field and consists of approximately 10,000 spectra of B < 25.25 color-selected galaxies which are expected to predominantly lie in the redshift range of 1.4 < z < 3 (Lilly et al. 2007; S. J. Lilly et al. 2011, in preparation). The spectra were obtained using the low resolution blue grism with a resolution R ∼ 200 at 5.3 Å pixel⁻¹. The spectra cover the wavelength range from 3500 Å to 7000 Å. For each of the objects observed spectroscopically, photometric redshifts are available from the excellent COSMOS photometry using a number of codes (e.g., Ilbert et al. 2009). The photo-z accuracy for these objects for which reliable spectroscopic redshifts are available is σ_Δz ∼ 0.0381(1 + z). Further details about these two parts of zCOSMOS are given in Lilly et al. (2007, 2009) and S. J. Lilly et al. (2011, in preparation).

The first step in the analysis is to identify pairs of galaxies consisting of one foreground galaxy from zCOSMOS-bright, selected to lie at 0.5 < z_spec < 0.9, and one background galaxy from zCOSMOS-deep whose sight line passes within a specified projected radius (or impact parameter) b from the foreground galaxy. The foreground redshift range of 0.5 < z_spec < 0.9 is chosen so that the Mg ii 2799 absorption at the foreground redshift lies within the spectral range of the zCOSMOS-deep spectra. For the foreground galaxies we only use galaxies with high-confidence redshift measurements, i.e., those with redshift Confidence Classes of 4.x, 3.x, 2.5, and 1.5. As shown in Lilly et al. (2009) these are 99% reliable. Based on the photo-z estimates of the other galaxies, this secure sample is, in this redshift range, about 95% complete.

The background sample is selected from the zCOSMOS-deep survey. Because we do not need to know the exact redshift of the background object, provided only that we know it truly lies well behind the foreground one, we can use the spectra of background galaxies even if a confident redshift has not been secured. We therefore select the background galaxies using their photometric redshifts to have z_phot > 1.0.

Figure 1 shows the redshift distributions of the foreground galaxies and the background galaxies, both of the full samples and of those galaxies that are used in this analysis because they form suitably close projected pairs. The total number of foreground galaxies between 0.5 < z_spec < 0.9 is 7110. Of these, 3908 have at least one background galaxy projected within b ⩽ 200 kpc. These background galaxies comprise 5237 of the total potential background sample of 8529 galaxies. As expected, the samples that can be used for this study are a random set of the parent samples (Figure 1).

Zoom In Zoom Out Reset image size

Clearly a major problem will occur if a "background" galaxy actually lies at the same redshift as the foreground galaxy, since then strong interstellar absorption in the former will be mistaken as absorption from the intergalactic medium around the latter. In the current study, this should be unlikely: most galaxies in the background sample lie at z_photo > 1.4, reflecting the use of the well-established gzK and ugr color selection criteria in zCOSMOS-deep. Random scattering to redshifts 0.5 < z < 0.9 is unlikely. Catastrophic failures in the photo-z will generally involve degeneracies between z > 2 and z < 0.5. The loss of high-redshift galaxies with spuriously low photometric redshifts will be of no consequence except to reduce the size of the sample. The inclusion of low-redshift galaxies that are assigned a spuriously high photo-z will merely dilute the signal from the foreground objects, since these sight lines do not in fact penetrate through the foreground system, but will not introduce a false signal. Using zCOSMOS-deep objects having secured spectroscopic redshifts (i.e., zCOSMOS-deep galaxies with spectroscopic redshift Confidence Classes 4.x and 3.x), we estimate that the fraction of objects in this latter category to be about 3%. We have not removed these known interlopers, but have checked that none of them lie at the same redshift as the foreground galaxy.

We associate to each foreground galaxy a stellar mass and corresponding absolute magnitude as described in the previous section. We divide our foreground galaxies into blue star-forming and red passive galaxies on the basis of their rest-frame (u − B) color, which is a weak function of mass. The line dividing the blue and red galaxies is similar but not identical to that used in (Peng et al. 2010):

$$\begin{equation} ({u-B})_{\rm AB}= 0.98 + 0.075\, \log \left({\frac{M}{10^{10} \, {M_{\odot }}}} \right) -0.18{z}, \end{equation} \tag{ 1 }$$

where M is the mass of the galaxy in question, for simplicity we use the average redshift z ∼ 0.7 of the sample as a whole. The color–mass division between blue and red galaxies is shown in Figure 2. The left-hand panel shows this division for non-group galaxies and the right-hand panel shows for the group galaxies.

Zoom In Zoom Out Reset image size

For environmental information, we use the group catalog of C. Knobel et al. (2011, in preparation) to separate the foreground sample into group and non-group galaxies. Almost 25% of the initial sample of 7110 galaxies can be assigned to a group. We refer the reader to Knobel et al. (2009) for a detailed description of the group-finding algorithm and details of the zCOSMOS group catalog.

To study the azimuthal dependence on Mg ii absorption line strength, we use the Zurich Estimator of Structural Types (ZEST) morphological classification (Scarlata et al. 2007) for each potential foreground galaxy. This is based on the HST/ACS F814W images of the COSMOS field. We first isolate those foreground galaxies that have a disk-dominated morphology, i.e., ZEST Type 2 excluding bulge-dominated systems, and further require that the galaxy has 0 < b/a < 0.65, corresponding to inclination angles of 50 < i < 90. This yields a sample of 595 foreground systems probed by 1134 background spectra within b < 200 kpc.

The spectra of all background galaxies whose sight lines pass within a given region of a given set of foreground galaxies are then co-added as follows. The spectrum of each individual background galaxy is first shifted in wavelength to the rest frame of the foreground galaxy in question. This means that absorption at the redshift of the foreground galaxy will appear at the correct "rest wavelength," i.e., 2799 Å in the shifted spectrum and will thereby allow co-addition of the spectra of background galaxies associated with foreground galaxies at different wavelengths. The continuum is then normalized by fitting the continuum with a running median filter of width 59 Å and then dividing the spectrum by it. To facilitate co-adding the spectra of different galaxies, this normalized spectrum is re-sampled onto a uniform grid in wavelength. The co-added spectra are produced by taking the median of the normalized re-sampled spectra. The median is chosen to reduce the sensitivity to absorption or emission features in the background galaxies, and to any other artifacts such as sky residuals, even though the effects of these should be random because of the range of redshifts of both foreground and background sources. We checked that other summation techniques, including straight averaging and average-sigma-clipping, produce very similar co-added spectra, but a marginally less clean continuum. The use of the median, plus the extended nature of the background source, means that the absorption in the co-added spectra represents a "typical" sight line rather than a true average including rare high-absorption sight lines. Three examples of the resulting co-added spectra are shown in Figure 3.

Zoom In Zoom Out Reset image size

The spectral resolution of the zCOSMOS-deep spectra is very low (R ∼ 200), so the two components of the Mg ii λλ 2796, 2803 doublet cannot be separated. The wavelength FWHM of almost 30 Å is much larger than the observed-frame 12 Å separation of the doublet at the redshifts of interest. We therefore fit a single Gaussian profile around the Mg ii rest-frame wavelength and measure the (rest-frame) absorption EW, integrating across both components of the doublet. In order to limit the number of free parameters in the fit, we determine the shape of this Gaussian (central wavelength and dispersion) from a fit to the best absorption spectrum, which is obtained within 40 kpc of all foreground galaxies (shown in Figure 3, top spectra). This Gaussian profile is then used for all other spectra, with only the depth as a free parameter. We have checked that other integration schemes, e.g., simply summing the flux deficit through the region of the line, give indistinguishable results within the uncertainties.

The errors on the EW measurements are determined using a bootstrap approach. For each set of background spectra, a thousand co-added spectra are generated from random selections of the sample. The width of the distribution of measured EWs in these thousand spectra are taken as representative of the error in the original measurement, and should account for sample variance, continuum uncertainties, and profile fitting errors.

We first investigate the variation of Mg ii line strength with impact parameter as a function of the color and mass of the foreground galaxy. The foreground sample is divided into four sub-samples in terms of their stellar masses and rest-frame colors: a high-mass blue sample is defined with log₁₀(M_stellar) ⩾ 9.88 M_☉, a low-mass blue sample comprises galaxies with log₁₀(M_stellar) < 9.88 M_☉. A high-mass red sample is defined with log₁₀(M_stellar) ⩾ 10.68 M_☉ and a low-mass red sample comprises galaxies with log₁₀(M_stellar) < 10.68 M_☉, as shown in Figure 2. These mass ranges are chosen so that each sample has approximately the same number of background spectra and so that the high-mass blue sample and the low-mass red sample have more or less the same mean stellar mass, thereby enabling a direct comparison between blue and red galaxies at the same mass. For each sample, a co-added spectrum is produced for impact parameters b < 50 kpc, 50 kpc < b < 65 kpc, and 65 kpc < b < 80 kpc.

The absorption EWs measured in the 12 co-added spectra are plotted in Figure 4 and displayed in Table 1. The error bars on mass give the 1σ spread of the foreground galaxy mass in that bin. It can clearly be seen that blue foreground galaxies are associated with stronger Mg ii absorption relative to red galaxies, especially at small impact parameters. To study this in more detail, we also create a single relatively narrow intermediate bin centered in mass on the overlap region (i.e., 10.1 M_☉ ⩾ log₁₀(M_stellar) ⩾ 10.6 M_☉, shaded region in Figure 4). This is shown in the panel to the right. In this overlap region of mass, the blue galaxies have almost eight times larger absorption at b < 50 kpc. There is also a significant mass dependence within the blue sample, again most clearly seen at smaller impact parameters. This may be present for the red sample but the statistical significance is very low.

Zoom In Zoom Out Reset image size

These results are consistent with the findings of Zibetti et al. (2007), who used the orthogonal approach of stacking images from Sloan Digital Sky Survey (SDSS) to show that stronger absorption systems in quasar spectra were associated with bluer, star-forming, galaxies. The current analysis provides a direct proof of the dependence of Mg ii absorption on the colors of the associated galaxies.

This clear demonstration of a color dependence is consistent with the scenario that Mg ii absorption within 65 kpc of galaxies is associated with material entrained in outflows driven by star formation. This is also indicated by analysis of blueshifted Mg ii absorption profiles in the spectra of star-forming galaxies (Weiner et al. 2009; Rubin et al. 2010). However, this evidence is also consistent with the idea that in-falling gas triggers star formation activity in blue galaxies. We show later in the paper, however, that the azimuthal dependence of the absorption strongly favors the first hypothesis.

Remembering that our EW measurements are averaged over the annuli surrounding the foreground galaxies, it is noticeable how strong this average absorption is around the most massive blue galaxies, even allowing for the fact that we compute integrated EW across both components of the doublet. In quasar spectra (which individually probe a much smaller region) an EW of greater than 0.3 Å is usually regarded as a "strong" system. This level of absorption (i.e., integrated EW > 0.6 Å) is attained across an area of more than 8000 kpc² around the blue galaxies with masses around 2 × 10¹⁰ M_☉.

To study further the radial profile around galaxies, we divide the sample more finely in b but lump together all blue galaxies and all red galaxies. We also distinguish between galaxies that are identified as lying in groups (as described in Section 2) and those not in groups. The resulting radial profiles are shown in Figure 5 and also displayed in Table 2.

Zoom In Zoom Out Reset image size

The non-group galaxies, in the left-hand panel, show a profile that drops steeply with radius beyond about 70 kpc. To ease comparison with previous work, we parameterize the radial profiles in terms of a singular isothermal sphere (SIS) as in Tinker & Chen (2008). We show later in Section 3.3 that the Mg ii absorption is certainly not spherically distributed, and thus our choice has no physical meaning per se, and is used simply as a convenient parameterization. In a future paper we will explore more physically motivated models for the radial profile.

In the isothermal sphere distribution, the distribution of Mg ii gas is characterized as

$$\begin{equation} \rm {EW_{r}^{iso} = \left\lbrace \begin{array}{@{}l@{\quad }l@{}}\dsty\frac{{\rm EW}_{0}}{\sqrt{(b^{2}/a_{h}^{2} + 1)}} \arctan {\sqrt{\dsty\frac{R_{{\rm gas}}^{2} - b^{2}}{b^{2} + a_{h}^{2}}}}, & \mbox{if }b \le R_{{\rm gas}} \\ \\ 0, & \mbox{otherwise. } \end{array}\right. } \end{equation} \tag{ 2 }$$

Here the core radius a_h is defined to be a_h = 0.2 R_gas and does not affect EW^iso_r at large b. We fit this model to the EW profiles of the non-group galaxies. We find R_gas = 115.2 ± 2.3 kpc (physical) and EW₀ = 0.7 ± 0.12 for "all" galaxies, R_gas = 107.6 ± 1.3 kpc (physical) and EW₀ = 1.1 ± 0.14 for blue galaxies, and R_gas = 118.1 ± 5. kpc (physical) and EW₀ = 0.3 ± 0.11 for red galaxies. For completeness, the results for all galaxies (group plus non-group) is very similar R_gas = 114.7 ± 1.5 kpc (physical) and EW₀ = 0.71 ± 0.12. It is noticeable that the R_gas for the blue and the red galaxies are very similar despite large changes in EW₀.

Although the methodologies are different, in the sense that we measure a spatially averaged absorption, our own analysis clearly confirms the trend between impact parameter (b) and EW for galaxies, for example as observed in Chen et al. (2010a) with quasar absorption lines in z < 0.5 SDSS galaxies. However, Chen et al. (2010a) did not find any color dependence for non-group z < 0.5 galaxies. As shown in Figure 4, our co-added spectra (at <z > ∼0.7) show a strong dependence on the color of the host galaxy for b ⩽ 70 kpc. The reason for this discrepancy is unclear to us. It might reflect the higher redshifts, and thus stronger star formation activity, of blue galaxies in our own sample, but the redshift difference is small enough to make this interpretation unlikely in our view.

The right-hand panel in Figure 5 shows the radial profile around galaxies, which are selected to lie in groups. This is somewhat artificial, essentially because the absorption from each line of sight will contribute to several different b bins, corresponding to the different members of the group. This will automatically produce a flatter distribution of Mg ii with impact parameter. We include it in this paper only for comparison with previous work (e.g., Chen et al. 2010a). With this caveat in mind, we see a much flatter absorption profile that extends beyond 140 kpc, substantially larger than seen for isolated, non-group galaxies. It should be noted that the dependence on color that we have observed for non-group galaxies is also evident in group galaxies at small radii b < 70 kpc but is much weaker at larger radii.

A more physically motivated approach would be to determine the Mg ii absorption profile around a particular location in the group, e.g., either the geometric center defined in C. Knobel et al. (2011, in preparation) as the geometric mean of the positions in the sky of all the group members, or the location of the most massive member of the group. The results are shown in Figure 6. The left panel is for group centers and the right panel is for the most massive galaxies in the groups.

Zoom In Zoom Out Reset image size

In order to try to understand the profiles in the right-hand panel of Figure 5 and in Figure 6, we construct a simple model in which the absorption profile around each group galaxy is given by that which we have measured around non-group galaxies—i.e., the model assumes that the absorption of the group can be represented by a simple superposition of the absorption of individual members. Knowing the locations of each individual members within each group, we can predict, for the ensemble of group members, the average profile that would be expected if all group members exhibited the same average profile as non-group members, as described by the SIS profiles in the left-hand panel of Figure 5. On average, only 2/3 of the I_AB ⩽ 22.5 galaxies were observed spectroscopically in zCOSMOS-bright. C. Knobel et al. (2011, in preparation) have developed a scheme to include in the groups, the missing members for which only a photo-z is available. In the model we include along with the spectroscopic members, the photometric members with a probability of membership p > 0.7. We first ignore the color of the other group members, but then incorporate this information also into the model.

If, within some bin in impact parameter, we have a sample of n group members, which are spread among some number of groups, and if the nth galaxy in this sample has m fellow members of its particular group, then the final EW that is expected for this impact parameter is given by

$$\begin{equation} {\rm EW}_{{\rm total}} = \frac{1}{n} \sum _{n}\Big(\!\sum _{m} {\rm EW}_m(b_m)\Big), \end{equation} \tag{ 3 }$$

where EW_m(b_m) is the EW appropriate for the SIS sphere of the galaxy type for galaxy m (i.e., blue or red) at the impact parameter b_m that the background galaxy has to this mth member galaxy. In summing the EWs, we assume that the velocities associated with each galaxy are large enough that there are no saturation effects. Typical velocity dispersions within the groups in question are typically 200 km s⁻¹ < σ < 600 km s⁻¹ (Knobel et al. 2009).

The results of this exercise are shown in Figure 7 and in Figure 6. Figure 7 re-plots the average group member profile of Figure 5 but no longer differentiating in color of the foreground galaxies. Both in Figure 7 and in Figure 6, the dotted lines show what would be expected from the SIS model around a single galaxy, i.e., ignoring other group members entirely, the dark shaded area includes the effects of other group members using the superposition approach, while the light shaded area also incorporates the information on the colors of those other members. Both of the latter shaded areas provide a better representation of the extended profile observed in group galaxies. This simple superposition model provides a reasonable representation of the absorption profiles around group galaxies. Interestingly, such a model would predict that the absorption profile at b  >  100 kpc should be independent of the color of the member galaxy, since at these radii it would be dominated by other group members. This may indeed be seen in the right-hand panel of Figure 5.

Zoom In Zoom Out Reset image size

This superposition model reproduces both the extended distribution and the magnitude of the Mg ii absorption profile very well using the geometrical group centers. However, while using the most massive galaxies, the result is less satisfactory. The differences between the two panels of Figure 6 arise from differences in both the models and in the observational data points, especially at large impact parameters. The difference in the models arises because the group centers are, by construction, more or less equidistant from the group members, maximizing the number which are close enough to contribute to the expected EW. Within the data, there are about 2.5 times more group members within 100 kpc of the geometrical center as there are within 100 kpc of the most massive galaxies. The differences in the observational points are not of large significance compared with the statistical error bars. The offsets between the geometrical centers and the most massive galaxies in each group extend up to 140 kpc, producing substantial scrambling of the data in impact parameter between the two approaches. Therefore, the error bars between the plots should be largely independent.

The success of this superposition model, especially for geometrical group centers is quite surprising as it suggests that the existence of Mg ii absorption halos around galaxies may not be significantly affected by the group environment. The fact that the specific star formation rates of star-forming galaxies do not appear to depend on environment, even though the fraction of galaxies that are star-forming does (Peng et al. 2010), suggests that the source of star-formation-driven winds is likely the same in the two environments. However, tidal effects have been invoked to enhance absorption for low-redshift groups (Kacprzak et al. 2010b). One can also imagine scenarios whereby different mechanisms may counteract each other, and the accuracy of the agreement anyway allows for some non-negligible differences.

With the current data set, we can probe the strength of Mg ii absorption lines around disk galaxies as a function of the azimuthal angle relative to the projected disk axis. As described in Section 2, we select a set of disk-dominated galaxies that lie within 40° of being edge-on. We then compute the azimuthal angle ϕ between the projected semi-minor axis of the disk, and the projected vector from the center of the foreground galaxy to the background galaxy. In other words, values of |ϕ| > 45° represent lines of sight that pass near to the plane of the disk, while values of |ϕ| < 45° are associated with lines of sight passing close to the symmetry (rotation) axis of the disk. If the Mg ii absorption around disk galaxies is associated with a bi-polar outflow along the disk axis, then the latter might be expected to show stronger absorption. Conversely, if the absorption was due to an extension of the disk itself, then the former would be expected to be stronger.

Figure 8 shows the radial Mg ii absorption profiles for these disk galaxies for two azimuthal bins split at |ϕ| = 45°. At small impact parameters, i.e., b < 50 kpc, the absorption along the disk axis is significantly stronger than in the plane of the disk. The difference disappears at larger radii because the polar quadrants have a much steeper radial decline. It is quite noticeable how flat the radial profile in the plane of the disk is. The error bars in impact parameter give the standard deviation of the distribution of impact parameters within that bin. The measurements are given in Table 3.

Zoom In Zoom Out Reset image size

This effect is also seen in Figure 9 (Table 4) which shows the azimuthal dependence for three radial bins. At small impact parameters, there is a strong azimuthal gradient in absorption strength from the poles down to the plane of the disk, nominally declining by a factor of about three. At the larger impact parameters, the effect vanishes, or may even possibly reverse (at low significance). The error bars in azimuthal angles are the standard errors on the mean of azimuthal angles within that bin.

Zoom In Zoom Out Reset image size

We take this as strong evidence that the close-in Mg ii absorption within 50 kpc is associated, at least statistically, with bipolar regions aligned with the axes of symmetry of the disks. This is almost certainly the signature of a bipolar outflow of material from the disk.

After the submission of our original manuscript, Kacprzak et al. (2011) have presented evidence from analysis of the inclinations of Mg ii selected galaxies that the distribution of Mg ii is co-planar with galaxy disks, especially for weak absorption systems. At first sight, this is in direct contrast to the evidence in the previous section that Mg ii absorption is strongest in bi-polar regions aligned with the poles of the disks. However, Kacprzak et al.'s (2011) result primarily refers to weaker Mg ii systems at relatively large impact parameters and this may account for some of the apparent differences.

Although we believe that the azimuthal effects explored above offer a much stronger and more robust diagnostic of the geometry of the Mg ii distribution than inclination effects, we present in this section, for completeness, the variation of Mg ii absorption with the inclination of the associated disk galaxies. As in the previous section, we select the foreground galaxies to be disk dominated, as defined in Section 2. We divide the disk galaxies into three bins in inclination angles, within 0° < i < 50°, 50° < i < 65°, and 65° < i < 90°.

The sense of the inclination angles is that i = 90° represents an edge-on system and i = 0° represents a face-on system. The sample is not sub-divided in terms of azimuthal angles, i.e., for any range of inclination the derived EW are averaged over all azimuthal angles. Figure 10 (Table 5) shows the variation of Mg ii absorption around disk galaxies as a function of inclination for three impact parameter bins. The error bars in inclination angles are the standard errors on the mean of the inclination angles within that bin. We find no significant trends of EW with the apparent inclination of disk-dominated galaxies.

Zoom In Zoom Out Reset image size

We plan to present a more comprehensive exploration of the dependence of EW on both inclination and azimuthal angles for a range of spatial distributions in a future paper. In this we will explore any differences between the two approaches described by Kacprzak et al. (2011) and here.