THE DEPENDENCE OF GALACTIC OUTFLOWS ON THE PROPERTIES AND ORIENTATION OF zCOSMOS GALAXIES AT z ∼ 1^*

The galaxies used in this study are selected from the zCOSMOS survey. The zCOSMOS survey (Lilly et al. 2007) is a spectroscopic survey carried out in the COSMOS field (Scoville et al. 2007). This study utilizes the brighter part of the zCOSMOS survey known as zCOSMOS-bright (Lilly et al. 2009).

The zCOSMOS-bright survey was carried out with the VIMOS spectrograph on the ESO UT3 8 m VLT. The final sample used here consists of approximately 20,000 flux limited I_AB ⩽ 22.5 galaxies covering the full two square degree COSMOS field (Lilly et al. in preparation). At this flux limit, the observed redshift range is 0 < z < 1.5. zCOSMOS-bright objects were observed with the MR grism using 1 arcsec slits, yielding a spectral resolution of R ∼ 600 at 2.5 Å pixel⁻¹. The spectra cover the wavelength range from 5550 Å to 9650 Å. From repeat measurements the average accuracy of individual redshifts has been demonstrated to be 110 km s⁻¹.

We first identify the galaxies with secure spectroscopic redshifts, within 1.0 ⩽ z_sp ⩽ 1.5. The redshift range is chosen so that the wavelength coverage of the zCOSMOS-bright spectra will cover both the Mg ii 2796 Å, 2803 Å absorption doublet as well as the 3727 Å [O ii] emission line. This redshift selection yields a sample with 525 galaxies with a median redshift of z_median ∼ 1.1. We utilize the [O ii] emission lines to accurately measure the systemic redshifts of the galaxies. About 8% of the galaxies do not have measurable [O ii] emission lines and they are removed from this study. This yields a parent sample of 486 galaxies. Figure 1 shows the redshift distribution of these 486 galaxies (black line). The galaxies have redshift confidence classes of 4.x, 3.x, 2.5, and 9.5. As shown in Lilly et al. (2009) galaxies with these confidence classes are 99% reliable in terms of their redshifts. It should be noted that this sample is not mass complete.

Zoom In Zoom Out Reset image size

Each galaxy is associated with a stellar mass and absolute magnitude as described in the previous section. We divide our sample into blue star-forming and red passive galaxies with their rest frame (U − B) color, which is a weak function of mass. The line dividing red and blue galaxies is given by

$$\begin{equation} (U-B)_{{\rm rest}} = 1.05 + 0.075\; \log _{10} \left(\frac{M_{*}}{10^{10} {M_{\odot } }} \right) -0.18z; \end{equation} \tag{ 1 }$$

where M_* is the stellar mass of the galaxy in question and z is the redshift of the galaxy. The color–mass division between blue and red galaxies is shown in Figure 2. The red points are for red galaxies and blue points are for blue galaxies, respectively. After division, we are left with 70 red galaxies and 416 blue galaxies, respectively. The redshift distribution of the red and blue galaxies are shown in Figure 1. The red and blue galaxies essentially have the same N(z). A two-sample KS test cannot rule out the null hypothesis that they are drawn from the same parent distribution at 5% significance.

Zoom In Zoom Out Reset image size

Stellar masses (and the SFR) for these galaxies have been estimated by spectral energy distribution (SED) fitting at the known spectroscopic redshifts of the galaxies, using the Hyperzmass code, a modified version of the photo-z code Hyperz (Bolzonella et al. 2010). 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). We use the ZEST morphological classification (Scarlata et al. 2007) to estimate the morphologies of the galaxies. This classification is based on the HST/ACS F814W images in the COSMOS field. We compute the flux-averaged star formation rate surface density ($\rm {\Sigma _{SFR} = SFR / 2 \pi R_{1/2}^{2}}$) of the galaxies by dividing their star formation rates with the area enclosed within their half light radii (R_1/2). To study the dependence of galactic outflow on the apparent inclination of the host galaxies, we select only disk dominated galaxies. We identify disk galaxies classified as ZEST type 2, exclude the bulge dominated systems, and divide the sample of galaxies into three bins of inclination angles (i), the face-on galaxies (0° ⩽ i ⩽ 40°), the intermediately inclined galaxies (40° ⩽ i ⩽ 55°), and the edge-on galaxies (55° ⩽ i ⩽ 90°), respectively. This yields a total of 218 disk galaxies. Furthermore, we probe the outflows in the galaxies classified in ZEST to have irregular morphologies, i.e., galaxies with ZEST type 3 which gives a sample of 174 galaxies. The various samples used are shown later in Table 1.

This method is based on the decomposition method described in Weiner et al. (2009). We first remove the systemic ISM absorption component from the coadded spectra and then proceed to identify the strength of the outflowing gas in the coadd. The outflowing gas lying between the observer and the light source (the galaxy) will absorb photons at blueshifted negative velocities with respect to the systemic redshift of the galaxy. This outflowing gas could have some velocity dispersion (generally much lower than the thermal broadening of individual clouds) and even complex kinematic structures. However, due to the low resolution of the zCOSMOS spectra, such kinematic complexities can be neglected for the present purposes. The model describing the final observed absorption line is

$$\begin{equation} F_{{\rm obs}}(\lambda) = C(\lambda) \; F_{{\rm em}}(\lambda) \; [ 1 - A_{{\rm sym}}(\lambda) ] \; [1 - A_{{\rm flow}}(\lambda)] \end{equation} \tag{ 2 }$$

$$\begin{equation} A_{{\rm sym}}(\lambda) = A_{2796} \; G(v,\lambda _{2796},\sigma) + A_{2803} \; G(v,\lambda _{2803},\sigma), \end{equation} \tag{ 3 }$$

where F_obs(λ) is the observed flux density from the coadded spectra, C(λ) is the fit to the local continuum of the coadded spectra, F_em(λ) describes any emission above the continuum level and for simplicity is set to unity. [1 − A_sym(λ)] describes the systemic absorption component and [1 − A_flow(λ)] describes the absorption profile due to outflowing (blueshifted) gas. The symmetric absorption component is taken to be the sum of two Gaussians centered at the rest frame wavelength of each component of the Mg ii λλ2796 Å, 2803 Å doublet.

The decomposition of the symmetric and the outflowing component can be a complex problem due to the doublet structure of the absorption profile. We, therefore, infer the symmetric component by fitting a Gaussian on the red side of the 2803 Å line (within 0 km s⁻¹ ⩽ v ⩽ 1500 km s⁻¹). Since the Mg ii doublet is partially blended in our coadd, assuming symmetry, we impose the Gaussian fitted to the 2803 line onto the 2796 line as well. We keep the depth of these two Gaussians the same (A₂₇₉₆ = A₂₈₀₃), as the individual Mg ii lines are likely to be saturated. If these absorption lines are not completely saturated, then it may happen that the 2796 Å line is slightly deeper than the 2803 Å line. In that case, we will underestimate the systemic component in that line.

The results of this approach are illustrated in Figure 5. The black line is the observed spectrum, the red dashed double Gaussian is the symmetric component [1 − A_sym(λ)], and the blue line is the detected outflowing component [1 − A_flow(λ)].

Zoom In Zoom Out Reset image size

The total outflow equivalent width is the equivalent width of the outflowing component measured between −1500 km s⁻¹ ⩽ v ⩽ 0 km s⁻¹  with respect to the 2803 Å line and, therefore, include both the components of the doublet. It should, therefore, be noted that the quoted outflowing equivalent width throughout the paper is the total equivalent width of the outflowing components from both the 2796 Å and the 2803 Å lines. The mean outflow velocity (〈v_flow〉) is the mean velocity of the outflowing gas blueward of the zero velocity of the 2803 Å line up to a velocity adaptively chosen to be the point at which the absorption is absent or minimized and never greater than −768 km s⁻¹. We also calculate the mean outflow velocity with respect to the 2796 Å line, and the measured outflow velocities from both the Mg ii 2796 and 2803 lines are consistent within the uncertainties. Throughout the paper, we quote the outflow velocity with respect to the 2803 Å line.

The errors on the equivalent width as well as the outflow velocities are estimated using a bootstrap approach. For each set of galaxy spectra to be coadded, a thousand coadded spectra are generated by randomly selecting spectra from that sample. For each of these coadds, the outflow equivalent width and the mean outflow velocities are estimated. The width of the distribution of the measured equivalent widths and the outflow velocities in these thousand coadds are taken as representative of the error in the original measurements and should also account for sample variance, continuum uncertainties, and the error in fitting the decomposition model.

This method assumes that the outflowing profile described in Equation (2) is only due to absorption from outflowing gas. Any redshifted emission in the 2803 Å line will change the amplitude and the width of the Gaussian fitted to the symmetric component. Emission to the red side of the 2796 Å line will effectively cause us to underestimate the equivalent width of the outflowing gas. It is very hard to account for the emission filling because the strength of Mg ii emission lines will depend on the galaxy mass as well as the dust content of the galaxy (see Kornei et al. 2013 and Martin et al. 2012 for a more detailed description of Mg ii emission lines). We are also assuming that there is a negligible amount of gas falling into the galaxy. Infalling gas can also contribute to the red wing of the 2803 Å absorption. However, recent observations of individual galaxy spectra have shown that the covering fraction of infalling gas is quite low, approximately ∼6 % (Rubin et al. 2012; Martin et al. 2012). Moreover, the equivalent widths of the inflowing component is always lower than that of the outflowing component, and hence it can only be observed when there is little or no outflow. Because of these considerations, we shall neglect the inflowing component in the reminder of this study.

It should be noted that this method works optimally for high S/N spectra, as for such cases the symmetric component can be modeled accurately and hence gives a better estimate of the outflowing gas.

To estimate the equivalent width of the outflowing gas, we also use a second method similar to the method used by Rubin et al. (2010). This method is more accurate for low S/N coadds as compared to the decomposition method. We make an assumption that the Mg ii doublet is saturated near the systemic velocities, in our spectra, such that the systemic components of 2796 Å and 2803 Å lines have the same depths at v = 0 km s⁻¹. This is a reasonable assumption because the Mg ii absorption line becomes saturated at column densities ∼10¹⁴ cm⁻², which correspond to relatively low hydrogen column densities >10¹⁹ cm⁻² at solar abundance (Rubin et al. 2010). The contribution from the ISM and the stellar atmosphere, which likely makes the dominant contribution to the systemic absorption, typically have column densities exceeding these values.

We measure the outflow equivalent width as follows

$$\begin{equation} W_{{\rm out}} \; =\; 2 \;\left(W_{2796\,\mathring{\rm{A}} ,\;{\rm blue}} \; - W_{2803\,\mathring{\rm{A}} ,\;{\rm red}} \right) \end{equation} \tag{ 4 }$$

where

$$\begin{equation} W_{2796\,\mathring{\rm{A}} ,\;{\rm blue}} \; =\; \sum \limits _{\lambda =2785\,\mathring{\rm{A}} }^{2796\,\mathring{\rm{A}}} \left(1 - \frac{F_{{\rm obs}}(\lambda)}{C(\lambda)} \right) \Delta \lambda \end{equation} \tag{ 5 }$$

$$\begin{equation} W_{2803\,\mathring{\rm{A}} ,\;{\rm red}} \; =\; \sum \limits _{\lambda =2803\,\mathring{\rm{A}}}^{2815\,\mathring{\rm{A}} } \left(1 - \frac{F_{\rm obs}(\lambda)}{C(\lambda)} \right) \Delta \lambda \end{equation} \tag{ 6 }$$

and F_obs(λ) is the observed coadded spectra and C(λ) is the local continuum around the absorption feature. W_{2803 Å, red} is the amount of absorption due to gas that is not outflowing, this is subtracted from W_{2796 Å, blue} to avoid overestimating the outflow absorption strength due to inclusion of the systemic component of absorption. This difference is multiplied by two to give the total equivalent width of the outflowing component in the doublet W_out. This method is more reliable for low S/N spectra because, as opposed to the decomposition method, this method does not directly rely on modeling the systemic component.

To estimate the mean outflow velocity in this second method, we consider the mean absorption weighted outflow velocity

$$\begin{equation} \langle v_{{\rm out}} \rangle \; =\; \frac{W_{{\rm all}}}{W_{{\rm out}}} \langle v_{{\rm all}}\rangle \end{equation} \tag{ 7 }$$

since we can assume that the mean velocity of the systemic component (〈v_sym〉) is zero. W_all is the total equivalent width of the Mg ii 2796, 2803 doublet and 〈v_all〉 is the mean absorption weighted velocity of the observed absorption line F_obs(λ) with respect to its systemic velocity.

The errors of measurement in this method are estimated similar to the decomposition method. Out of the sample of galaxies under study, randomly selected spectra are coadded one thousand times. For each coadd equivalent widths and velocities are measured as described above. The width of the distribution of the measured equivalent widths and the velocities in these thousand coadds are taken as representative of the error in the original measurements.

Both methods described above assume that there is no redshifted absorption from the outflowing gas. However, we are dealing with finite spectral resolution, which effectively puts some of the absorption from the outflowing gas into the redshifted absorption. This implies that we will overestimate the systemic component and as a result the outflow equivalent widths will be underestimated as the spectral resolution is degraded.

To estimate the typical magnitude of this effect, we set up a simple empirical model, similar to the model described by Equations (2) and (3). We set up a systemic absorption component described as [1 − A_sym(λ)] centered on the Mg ii 2796, 2803 absorption doublet. The depths of the two Gaussians are ∼40% of the continuum (we assume that the systemic component is saturated and hence the line ratio is 1:1) and σ_sym = 100 km s⁻¹ is chosen. The outflowing component [1 − A_flow(λ)] can be of any arbitrary shape, but we assume that it can also be described by two Gaussian profiles that are blueshifted from the systemic velocity by an arbitrary velocity offset Δv, which is varied. The Gaussians describing the outflowing components are truncated at zero velocity. The depth of the outflowing absorption profile is also varied and the width of the outflowing component varied to be Δv/2 ⩽ σ_out ⩽ Δv. We again assume that there is no emission and get an ideal absorption profile $\mathcal {F}(\lambda)$. To account for finite spectral resolution $\mathcal {F}(\lambda)$ is convolved with a Gaussian window function W to give the observed flux F(λ) as follows

$$\begin{equation} F(\lambda)= \left(W * \mathcal {F}\right)(\lambda) = \int _{-\infty }^{\infty } W(\tau) \mathcal {F}(\lambda - \tau)\, d\tau. \end{equation} \tag{ 8 }$$

Here, W is a Gaussian with an FWHM that is changed to mimic the effect of varying resolution. The equivalent width of the outflow component is then estimated from F(λ) using both the decomposition and the boxcar method. Figure 6 shows the observed equivalent width of the outflow component for a set of models with a fixed systemic component and varying outflow component chosen to span an outflow velocity range of −326 km s⁻¹ ⩽ 〈v_out〉 ⩽ −195 km s⁻¹. For a model with an intrinsic outflow equivalent width of ∼2.7 Å, the observed outflow equivalent width is ∼1.55 Å for an instrument with an FWHM of 500 km s⁻¹. This correction needs to be applied to the measurements done with both the decomposition and the boxcar method to estimate the true outflow equivalent width corrected for resolution effects.

Zoom In Zoom Out Reset image size

The Mg ii absorption doublet from the coadd of all the 486 galaxies was shown in Figure 5. Before correcting for resolution, the outflow equivalent width measured from the decomposition method is W_flow = 1.42 Å and the outflow equivalent width measured from the boxcar method is W_out = 1.6 Å. Since the resolution of zCOSMOS spectra corresponds to an FWHM of ∼500 km s⁻¹, the resolution corrected outflow equivalent widths are W_flow = 2.52 ± 0.25 Å and W_out = 2.7 ± 0.23 Å, respectively. The outflow velocities inferred by both the methods are 〈v_flow〉 = − 293 ± 32 km s⁻¹ and 〈v_out〉 = − 270 ± 63 km s⁻¹, respectively. 〈v_out〉 is estimated using Equation (7). It should be noted that W_all and 〈v_all〉 are not affected by the degradation in resolution and hence are robustly estimated from the observed absorption feature without any corrections. As a result, the use of Equation (7) to deduce 〈v_out〉 only requires the correction of W_out. These outflow velocities are not taking into account, emission filling because of redshifted backscattered Mg ii emission. We account for this effect by applying an average correction of 67 km s⁻¹ to all the outflow velocity estimates. This correction gives emission corrected 〈v_flow〉 = − 226 ± 32 km s⁻¹ and 〈v_out〉 = − 203 ± 63 km s⁻¹, respectively. The average correction for emission filling is obtained by comparing the Mg ii and Fe ii outflow velocities at higher redshifts as described below. Throughout the paper, all of the outflow velocities quoted were found after correcting for an average emission filling.

In this section, we measure outflow velocities using the Fe ii 2586, 2600 lines and check that these velocities are indeed consistent with that measured with the blueshifted Mg ii doublet. Fe ii lines have the added value that they will suffer less from redshifted emission filling, allowing a robust characterization of the systemic component (Prochaska et al. 2011). However, they are too blue to be observed in our spectra at all redshifts. The Fe ii 2586, 2600 doublet can only be analyzed with our sample at z > 1.15. We select a subsample of 207 galaxies with z > 1.15, which are coadded and used to analyze the Fe ii doublet. We continuum normalize and coadd the spectra as before. Figure 7 shows the coadded Fe ii 2586, 2600 lines as well as the corresponding Mg ii 2796, 2803 doublet, respectively. The Mg ii doublet is offset vertically for presentation. We model the systemic components of the Fe ii 2586, 2600 lines by fitting a single Gaussian on the red side of both lines independently (Figure 7, red lines). The blue lines show the estimated outflowing component for both the transitions. The Fe ii 2586, 2600 outflow equivalent widths are almost the same for both lines (W₂₅₈₆ = 0.85 ± 0.30 Å, W₂₆₀₀ = 0.98 ± 0.29 Å), indicating that the outflowing gas is optically thick in the coadd. We compute for the same coadded spectra the Mg ii outflow equivalent width, W_flow = 2.1 ± 0.36 Å. We obtain the mean outflow velocity for each species as before and find for the Fe ii 2586 line 〈V_flow〉(2586) = −223 ± 62 km s⁻¹ and for the Fe ii 2600 line 〈V_flow〉(2600) = −259 ± 75 km s⁻¹. For the Mg ii transition, we estimate the mean outflow velocity 〈V_flow〉(2803) = −290 ± 47 km s⁻¹ and 〈V_flow〉(2796) = −285 ± 53 km s⁻¹. Hence, the measured mean outflow velocities are consistent within the uncertainties for both Mg ii and Fe ii doublets. However, the inferred mean Fe ii outflow velocities are slightly lower than the mean outflow velocities inferred from the Mg ii absorption. This is owing to the fact that emission filling in the Mg ii doublet will systematically create bias, leading us to underestimate the systemic absorption components. The difference between 〈V_flow〉(2803) and 〈V_flow〉(2586) is 67 km s⁻¹. However, the general trends seen regarding dependence of galactic outflows with galaxy properties will remain the same for both the absorption species. We compensate the effect of emission filling by applying an average correction of 67 km s⁻¹ to all Mg ii traced outflow velocity estimates. Throughout this study, all of the outflow velocities are offset by 67 km s⁻¹ to compensate the effect of backscattered redshifted emission filling.

Zoom In Zoom Out Reset image size

We first investigate the variation of the outflow equivalent width and outflow velocity with the rest frame color and stellar mass of the host galaxies. The galaxy sample is divided into red and blue galaxies as described before. To study the mass dependence, we divide the blue galaxies into three stellar mass bins. The low-mass blue galaxy sample consists of galaxies with M_* ⩽ 10¹⁰M_☉, the intermediate-mass blue galaxy sample consists of galaxies with stellar masses 10¹⁰M_☉ < M_* ⩽ 10^10.45M_☉ and the high mass blue galaxy sample is defined as galaxies with M_* > 10^10.45M_☉

The high mass blue galaxy sample has a mean stellar mass 〈M_*〉 = 10^{10.71 ± .22}M_☉, which is reasonably close to the mean mass of the red galaxy sample 〈M_*〉 = 10^{11 ± .24}M_☉ to compare the dependence of outflowing properties with color. Figure 8 shows the variation of outflow equivalent width as a function of rest frame color and stellar mass. It can be seen that both the decomposition method (filled circles), and the boxcar method (open diamonds) give similar outflow equivalent widths within the uncertainties. Among the blue galaxies, the high-mass blue galaxies exhibit higher outflow equivalent widths as compared to the low mass blue galaxy samples. At similar stellar mass ranges, the blue galaxies exhibit significantly higher outflow equivalent widths as compared to the red galaxies. In the sample of red galaxies, very little outflow absorption is observed. This color and mass dependence in the equivalent width of the outflowing component hints at a dependence of the outflowing gas on the star formation rate of the host galaxy. This trend is also similar to the trend observed in Bordoloi et al. (2011) for impact parameters within 50 kpc of 0.5 ⩽ z ⩽ 0.9 galaxies, seen in the spectra of background galaxies.

Zoom In Zoom Out Reset image size

The gray points are data points taken from the literature, the gray open circles are taken from Weiner et al. (2009) and the gray open diamonds are taken from Rubin et al. (2010). The study done in Weiner et al. (2009) used 1406 spectra of blue star forming galaxies at z ∼ 1.4 from the DEEP2 galaxy survey. They estimated the outflow equivalent width of the 2796 Å line only and we have multiplied that value by two to compare with our estimates of the total outflow equivalent width of the doublet. The study done in Rubin et al. (2010) used a sample of 468 galaxies in the TKRS survey in the GOODS-N field at 0.7 ⩽ z ⩽ 1.5 with a median redshift of 0.9. Both of these studies have a higher spectral resolution than the zCOSMOS bright survey. Although these studies were done in different redshift ranges, their general trends are consistent with our finding of a general increase of outflow equivalent width with stellar mass. We perform a Pearson linear rank correlation test, which rules out the null hypothesis that there is no correlation between stellar mass and outflow equivalent width among blue galaxies at a 95% confidence level. Our own analysis highlights the importance of incorporating color in interpreting the trends with mass and/or luminosity.

Figure 9 shows the outflow velocity estimates done with both the decomposition and the boxcar method as a function of stellar mass. For the red galaxies, only an upper limit for outflow velocity is measured. The gray points are again taken from Weiner et al. (2009). We find that, within the error bars, there is no significant trend in outflow velocity as a function of stellar mass, although the low mass blue sample gives a lower estimate of outflow velocity in the boxcar method as compared to the decomposition method. In general, we find a typical range of outflow velocity ranging approximately from −150 to −200 km s⁻¹ among these galaxies.

Zoom In Zoom Out Reset image size

In this section, we investigate the correlation between the star formation rates and the outflow properties of the galaxies. The galaxy sample is subdivided at the 25th and 75th percentile values of star formation rates. The lowest star forming sample consists of galaxies with log₁₀(SFR/M_☉ yr⁻¹) ⩽ 1.0396. The intermediate star forming sample consists of galaxies with 1.0396 < log₁₀(SFR/M_☉ yr⁻¹) ⩽ 1.572. The highest star forming sample is made out of galaxies with log₁₀(SFR/M_☉ yr⁻¹) > 1.572.

Figure 10 shows the variation of outflow equivalent width as a function of the star formation rate of the galaxies. There is a general trend of steady increase in outflow equivalent width with increasing star formation rates. This trend is seen in both the decomposition method estimates and the boxcar estimates. These trends are consistent with the trends found in the literature (gray points) though the other studies were done at different redshift ranges.

Zoom In Zoom Out Reset image size

Figure 11 shows the variation of outflow velocity with varying star formation rates. We find that the mean velocity of the total Mg ii absorption (〈v_all〉) increases with the increase of SFR (see Table 1), the inferred outflow velocities also show a very weak trend of increasing with SFR. Owing to large uncertainties in our measurements, both estimates are broadly consistent with the observations of Weiner et al. (2009; gray circles), who found a weak correlation between increasing outflow velocity with increasing star formation rates. The outflow velocity estimates in Weiner et al. (2009) are not corrected for emission filling, which accounts for the small offset with our outflow velocity estimates. Combining our measurements with that from Weiner et al. (2009), we perform a Pearson linear rank correlation test and find a 2.7σ correlation between log SFR and equivalent width/outflow velocity.

Zoom In Zoom Out Reset image size

In this section, we test for dependence of outflow strength on the SFR surface density (Σ_SFR). We estimate a flux-averaged SFR surface density ($\Sigma _{{\rm SFR}} = {\rm SFR} / 2 \pi R_{1/2}^{2}$), for each object assuming that star formation is distributed like the (observed frame) i-band light (i.e., half the star forming regions of the host galaxies are within their half light radius (R_1/2)). The left panel of Figure 12 shows the distribution of log SFR and $\log {\pi R_{1/2}^{2}}$ for the sample used in this section. The red and cyan lines indicate the 25th and the 75th percentile values of log Σ_SFR (−1.04, −0.25), which are used to subdivide our sample.

Zoom In Zoom Out Reset image size

In Figure 12, the right panel shows the outflow equivalent width versus log Σ_SFR. The filled circles are measurements done with the decomposition method and the open diamonds are estimated with the boxcar method. The subsample with the lowest log Σ_SFR exhibits very little outflow equivalent width and as we probe the higher log Σ_SFR subsamples the measured outflow equivalent width increases.

It is interesting to discuss these results in the context of the canonical star formation rate surface density "threshold" (Σ_SFR = 0.1 M_☉ yr⁻¹ kpc⁻²). It has been suggested that it is the minimum Σ_SFR required for driving outflows in local starbursts (Heckman 2002). By chance, this threshold corresponds to the Σ_SFR between the first and the second bin in this study, and it is noticeable that the lowest bin has only an upper limit for outflowing equivalent width. The Pearson linear rank correlation test shows a 2.9σ correlation between log Σ_SFR and outflow equivalent width.

At higher redshifts, Rubin et al. (2010) and Steidel et al. (2010) have reported weaker correlations of outflow absorption strength with Σ_SFR. Recently, Kornei et al. (2012) reported a 3.1σ correlation of outflow velocity with Σ_SFR at these redshifts. At z ∼ 2, Newman et al. (2012) reported a correlation between outflow strength (quantified via the broad flux fraction of the Hα line) and Σ_SFR of star forming galaxies. They reported a Σ_SFR threshold of 1 M_☉ yr⁻¹ kpc⁻², above which M_* > 10¹⁰M_☉ galaxies might have stronger outflows as compared to the low mass galaxies.

In this section, we present the variation of Mg ii outflow kinematics with the apparent inclination of the associated disk galaxies. We select the galaxies to be disk-dominated, as described before. We divide the disk galaxies into three bins in inclination angles, with 0° ⩽ i ⩽ 40°, 40° < i ⩽ 55°, and 55° < i < 90°. The definition of the inclination angle is such that i = 90° represents an edge-on system and i = 0° represents a face-on system. Some examples of the typical galaxies selected are shown in Figure 13. We also investigate the distribution of mass, SFR, and Σ_SFR in each inclination bin and find that they are consistent with having the same parent distribution according to a two-sample KS test at a 10% significance level. The cumulative distributions of the stellar mass, SFR, and Σ_SFR of the galaxies in each inclination bin are shown in Figure 14.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 15 shows the variation of outflow equivalent width (left panel) with inclination and outflow velocity (right panel) with inclination. The error bars in inclination angles are the standard deviation of the inclination angles within that bin. The face-on systems show the strongest outflow absorption equivalent width and this decreases with increasing inclination. The edge-on systems show the weakest outflow equivalent width and only a 1σ upper limit was measured in both the decomposition (filled circles) and the boxcar (open diamonds) methods. The face-on systems also exhibit the highest outflow velocities and the observed outflow velocities decrease as we probe galaxies with higher inclinations. The Pearson linear rank correlation test shows a 3σ correlation between galaxy inclination and outflow equivalent width/outflow velocities.

Zoom In Zoom Out Reset image size

The galaxies with irregular morphologies, classified as ZEST type 3 have similar outflow equivalent widths as the face-on disk galaxies and similar outflow velocities (∼ − 200 km s⁻¹, red points Figure 15). Since no inclination can be assigned to the galaxies with irregular morphologies, their outflow estimates are essentially averaged over all possible orientations. However, it should be noted that the mass distribution of these galaxies is not the same as that of the disk galaxies. A two-sample KS test performed on the distribution of mass, SFR, and Σ_SFR of the irregular and disk galaxies rules out the null hypothesis that they are drawn from the same parent distribution at a 5% significance level. To the first order, they are shifted to lower mass by about 0.15 dex in mass and 0.3 dex is SFR.

The trend of variation of outflow velocity with galaxy inclination is similar to the trends observed for low-redshift samples. Chen et al. (2010) studied a sample of ∼150,000 SDSS galaxies and created stacks of galaxy spectra as a function of apparent disk inclination and studied the outflow properties of the Na D absorption doublet. They found a strong increase in outflow velocity as the galaxies became more and more face-on, which is consistent with the picture of galactic winds escaping along the disk rotation axis. Furthermore, Heckman et al. (2000) observed Na D absorption in 18 local starbursts and found that a higher fraction of the face-on galaxies exhibit outflows as compared to the edge-on systems. These low-redshift studies were feasible because of relatively bright targets and high resolution spectroscopy.

Recently, Weiner et al. (2009) examined 118 galaxies at z ∼ 1.4 with HST I-band imaging and did not find a correlation between apparent inclination and wind strength or outflow velocity. They inferred that this might be due to uncertainties in estimating axis ratios of irregular galaxies imaged in the rest-frame U band. Kornei et al. (2012), using a sample of 72 z ∼ 1 galaxies, divided their sample at i = 45° and found that the coadded spectra of the face-on systems exhibit higher outflow velocity as compared to the more edge-on systems.

Our results for inclination dependence of outflowing gas in the "down the barrel" spectra of disk galaxies are entirely consistent with the azimuthal dependence (around edge-on galaxies) of the distributed strong Mg ii absorbers, seen against the spectra of background galaxies in Bordoloi et al. (2011).

The new results tie directly to the picture that the azimuthal asymmetry of strong Mg ii absorption line systems around galaxies, observed in the spectra of background galaxies or quasars, are primarily due to cool Mg ii gas entrained in star-formation-driven bipolar winds, especially at low impact parameters (Bordoloi et al. 2011, 2014).