Free Access
Issue
A&A
Volume 590, June 2016
Article Number A68
Number of page(s) 29
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/201527038
Published online 12 May 2016

© ESO, 2016

1. Introduction

Galaxies at low and high redshift are surrounded by massive gaseous halos. These halos are believed to be built up and fed by large-scale gas circulation processes, such as accretion from the intergalactic medium (IGM), galactic winds, and outflows powered by star formation and active galactic nuclei, as well as minor and major mergers that result from the on-going hierarchical formation and evolution of galaxies. Such gaseous material in extended galaxy halos is nowadays often referred to as the circumgalactic medium (CGM). We here define the CGM as diffuse neutral and ionized gas that is located within the virial radius of a galaxy, but outside of its (main) stellar body.

Observational and theoretical studies imply that the CGM is a key component in the ongoing process of galaxy formation and evolution. Because the gas-consumption timescales of late-type galaxies are short compared to their lifetimes, they must gain gaseous material from outside (e.g., from the IGM and/or from merger processes) through the CGM to sustain their ongoing star formation. The manner in which the gas makes its way from the IGM/CGM to the disks of late-type spirals is not well understood, however. The classical picture of accretion (e.g., Rees & Ostriker 1977; White & Rees 1978; Birnboim & Dekel 2003) and the “hot-” and “cold-mode” scenarios (Kere et al. 2005) most likely are oversimplified, because the underlying physics that describe the large-scale flows of multiphase gas from the outer to the inner regions of a dynamically evolving galaxy is highly complicated (e.g., Mo & Miralda-Escude 1996; Maller & Bullock 2004). Most likely, only state-of-the art hydrodynamical simulations that cover the necessary physics and that have the necessary spatial resolution potentially provide realistic measures for the amount of gas that is reaching the disk of late-type galaxies and for the timescales that it requires to do so (e.g., van de Voort et al. 2011).

Observations that aim at studying the properties of the CGM are thus extremely important to improve our understanding of gas-accretion processes and provide constraints for numerical models. What observations can provide is a characterization of the distribution of the different gas phases in the CGM, the total mass that it contains (under reasonable assumptions), and the connection between the properties of the CGM and the properties of their host galaxies (e.g., morphology, luminosity, star-formation rate, etc.). The gas densities in the CGM are low, however, so that emission measurements of circumgalactic gas (e.g., in the X-ray regime) typically are limited to the very inner halo regions of nearby galaxies using current instruments (Anderson & Bregman 2011; Dai et al. 2012; Bogdán et al. 2013). Fortunately, absorption-line measurements that use background active galactic nuclei (AGNs; throughout the following we use the abbreviation QSO to refer to the various classes of AGNs collectively) can access physical tracers at the relevant densities with extremely high sensitivity. For such absorption-line measurements the ultraviolet (UV) range is particularly important, as it contains a large number of diagnostic transitions from low, intermediate and high ions of heavy elements and the Lyman series of neutral hydrogen. Consequently, UV absorption-line observations with past and present space-based UV spectrographs such as the Far-Ultraviolet Spectroscopic Explorer (FUSE), the Space Telescope Imaging Spectrograph (STIS) and the Cosmic Origins Spectrograph (COS; the latter two instruments installed on the Hubble Space Telescope; HST) provide a wealth of information on the physical and chemical properties of the different gas phases in the CGM of the Milky Way and other galaxies (e.g., Wakker et al. 1999, 2001; Sembach et al. 2003; Richter et al. 2001, 2009, 2011; Tripp et al. 2003; Fox et al. 2005, 2010; Collins et al. 2009; Wakker & Savage 2009; Shull et al. 2009; Prochaska et al. 2011; Tumlinson et al. 2011; Lehner et al. 2013; Keeney et al. 2013; Werk et al. 2013; Stocke et al. 2014; Liang & Chen 2014).

In two previous papers, we have studied the amount and distribution of neutral gas (the so-called high-velocity clouds) in the halos of low redshift galaxies via optical absorption spectroscopy and their contribution to the neutral-gas accretion rate in the local Universe (Richter et al. 2011; Richter 2012). We here continue our long-term strategy to study the circumgalactic medium in the local Universe with an absorption-line survey that aims at characterizing the extent and mass of diffuse ionized gas around low-redshift galaxies using archival UV absorption-line data from HST/COS.

This paper is organized as follows. A short discussion on the importance of doubly ionized silicon as a sensitive tracer for circumgalactic gas is presented in Sect. 2. In Sect. 3 we describe the HST/COS data acquisition, the COS data reduction, the spectral analysis method, and the galaxy data origin. In Sect. 4 we discuss the observed absorber properties, such as their incident rate, their redshift distribution, the distribution of equivalent widths and column densities. In Sect. 5 we model the expected absorption cross section of circumgalactic gas using the local galaxy luminosity function. Section 6 deals with the observed absorber-galaxy connection in our data sample. The ionization conditions and the cosmological mass density of the Si iii absorbers are considered in Sect. 7. We discuss our results in Sect. 8 and provide a summary of our study in Sect. 9.

2. On the importance of Si III as a tracer of circumgalactic gas

A particularly powerful transition in the UV to detect ionized interstellar, circumgalactic, and intergalactic gas in a wide range of physical conditions is that of doubly ionized silicon (Si iii) at 1206.500 Å (hereafter referred to as Si iiiλ1206). This is because of the very high oscillator strength of Si iiiλ1206 of f = 1.63 (Morton 2003) and the relatively large cosmic abundance of silicon (log (Si/H) = −4.49; Asplund et al. 2009). Doubly ionized silicon is a Mg-like ion with a closed-subshell 3s2 valence-electron structure. The ionization-energy boundaries of the Si iii ion are 16.35 and 33.49 eV (Morton 2003). As a result, detectable Si iii arises from both diffuse photoionized as well as collisionally ionized gas at moderate temperatures (T< 105 K; see Shull et al. 2009).

To demonstrate the diagnostic power of the Si iiiλ1206 transition for the study of the CGM we show in Fig. 1 the expected fractional abundance of the Si ions Si iiSi viii in gas at z = 0 that is exposed to the local UV background. For the UV background we use a modified version of the model by Haardt & Madau (2012), in which we re-scale the photoionization rate to a value of log Γ = −13.6 to compensate for the apparent discrepancies between the Haardt & Madau model and recent observational results (see Kollmeier et al. 2014; Wakker et al. 2015; Shull et al. 2015). The calculations (Fechner, in prep.) are based on ionization models performed with the ionization code Cloudy (v13.03; Ferland et al. 2013). As can be seen, doubly ionized Si represents the dominant observable ionization state of Si at densities and temperatures (log nH = −2.0 to −3.5, T< 105 K) that are typical for multiphase circumgalactic gas structures that are embedded in hot coronal gas (see Cen 2013).

In terms of quasar absorption-line systems, Si iiiλ1206 absorption in the low-redshift Universe is expected to be detectable (if not blended by other spectral features) in basically all damped Lyman α absorbers (DLAs; log N(H i)≥ 20.3), sub-damped Lyman α absorbers (sub-DLAs; 19.0 ≤ log N(H i)< 20.3), Lyman-limit systems (LLS; 17.2 ≤ log N(H i)< 19.0), because these are metal-enriched, multiphase absorbers with high gas columns. The majority of Si iii-selected absorbers (throughout the following referred to simply as Si iii absorbers), however, are expected to arise in ionized metal systems at lower H i column densities (log N(H i)< 17.2), where the minimum H i column density to detect Si iii with HST/COS in solar-metallicity gas at moderate gas densities (nH ~ 10-3 cm-3) can be as low as log N(H i)= 14 (see Sect. 7.1).

Low ions with lines in the observable UV, such as O i, C ii, and Si ii, are only present in mostly neutral or slightly ionized gas (i.e., at high N(H i)), while high ions such as C iv and O vi predominantly trace highly-ionized gas at gas densities that are typically below nH ~ 10-3 cm-3. A Si iii-selected absorption-line survey, such as presented in this paper, therefore is expected to be particularly sensitive to detect metal-enriched gas in the inner and outer halos of galaxies and to characterize its spatial distribution and physical properties.

thumbnail Fig. 1

Ionization fractions of different Si ionization states for different gas densities and temperatures, as calculated from a Cloudy ionization model (including photoionization and collisional ionization) of circumgalactic gas at z = 0 using the local UV background. The upper panel shows a model for log T = 4.2, the lower panel a model for log T = 4.4. The Si iii ion represents the dominant ionization state in the density range that is characteristic for circumgalactic gas structures that are embedded in hot coronal gas (log nH = −2.0 to −3.5).

Open with DEXTER

3. Observations, data handling, and analysis method

3.1. COS spectra selection and data reduction

For our study we make use of archival HST/COS data that were retrieved from the HST Science Archive at the Canadian Astronomy Data Centre (CADC). Because we aim at studying intervening Si iii λ1206 absorption in the IGM at z ≤ 0.1 together with corresponding H i Ly αλ1215 absorption we are primarily interested in the wavelength range between 1208 and 1338 Å. This range is covered by the COS G130M grating which operates between λ = 1150−1450 Å, providing a spectral resolution of R ≈ 15 000−20 000 (corresponding to an instrumental FWHM of 15−20 km s-1, while the native pixel size is 2 km s-1; Green et al. 2012; Debes et al. 2016). Using the CADC web interface we searched for all publicly available COS data from all types of extragalactic point sources such as the various types of AGN and galaxies that were observed with the COS G130M grating. By the end of February 2014 we found (and downloaded) G130M data sets for 552 extragalactic sightlines. Whenever G160M data (covering the range λ = 1405−1775 Å  including the important C iv doublet at 1548.2,1550.8 Å) were available for these sightlines, we retrieved them as well.

For the further data reduction we used the raw COS spectra of the individual science exposures. The individual exposures were processed with the CALCOS pipeline (v2.17.3) to produce the standard COS x1d fits files. For the coaddition of the individual spectra we then used a custom-written code that aligns the individual exposures in wavelength space in a fully automated fashion. The code calculates for each exposure a pixel/wavelength calibration based on the line flanks (for spectra with S/N> 5) or line centers (for spectra with S/N ≤ 5) of various interstellar anchor lines that are distributed over the wavelength range of the G130M and G160M spectral ranges. The heliocentric velocity positions of the anchor lines were calibrated for each sightline using H i 21 cm data from the Leiden-Argentine-Bonn (LAB) survey (Kalberla et al. 2005). The individual spectra then were uniformly rebinned and coadded pixel-by-pixel (using the count rate in each pixel), where pixels with known artifacts were flagged accordingly. Errors were calculated in the coadded spectra by weighting by inverse variance. In this way, we obtained for each sightline a calibrated, co-added G130M spectrum (and G160M spectrum, if available). For each sightline we checked the quality of the data reduction by a visual inspection of the final spectrum.

As it turns out, many of the data sets have very low signal-to-noise ratios (S/N) and/or sample background sources with very low redshifts. These spectra are thus not suited to investigating intervening metal absorption at z ≤ 0.1. We selected only those spectral data that have a minimum S/N per resolution element of four in the wavelength range between 1208 and 1338 Å  and for which the redshift of the background source is zQSO> 0.03. This selection reduces the total sample to 303 lines of sight (LOS). In Table A.1 in the Appendix we present a complete list of all 303 QSO sightlines in our COS sample including QSO names and coordinates.

3.2. Redshift-path coverage, absorber identification, and spectral analysis

The next step in our analysis was to characterize the available redshift path to detect intervening Si iii+H i Ly α absorption at z ≤ 0.1 in each spectrum. We generally exclude the velocity range between | vhelio | = 0−500 km s-1 where absorption by local disk gas and by intermediate- and high-velocity clouds in the Milky Way halo is found (see, e.g., Wakker & van Woerden 1998; Richter 2006; Shull et al. 2009; Lehner et al. 2012; Putman et al. 2012; Herenz et al. 2013). To sort out regions that may be associated with the background AGN we further ignore the spectral range that lies within 5000 km s-1 of zQSO. Finally, we identify and flag along each sightline those spectral regions that are heavily blended by other intervening absorbers (in particular by higher-redshift LLS and DLAs). As a result, we obtain a total absorption redshift path Δzabs< 0.1 for each sightline that is available to identify intervening Si iii+H i Ly α absorption (see Table A.1, last column).

For the determination of the Si iii number density (Sect. 4.2) we further need to consider the detection limit for Si iii absorption along each sightline, which depends on the S/N in the relevant part of the spectrum where intervening Si iii absorption is expected to occur. The minimum column density, Nlim, that can be detected at 3σ significance from an unresolved absorption line with a laboratory wavelength λ0 and an oscillator strength f in a spectrum with a resolving power R and a given S/N per resolution element is given by (e.g., Richter et al. 2001; Tumlinson et al. 2002): (1)For the only available Si iii transition in the UV range we have λ0 = 1206.500 Å  and f = 1.62732 (Morton 2003), so that for a S/N of 4 per resolution element the formal 3σ detection limit for R = 15 000 is log Nlim(Si iii)= 12.46. More than half of the considered COS spectra have a S/N per resolution element of 12, so that for these LOS log Nlim(Si iii)≤ 12.0.

thumbnail Fig. 2

Five examples for velocity profiles of Si iii-selected absorption line systems in our HST/COS survey with decreasing Si iii λ1206 absorption strengths (gray-shaded area; from left to right). Shown are the velocity profiles of H i, Si iii, Si ii, Si iv, C ii, and C iv (from top to bottom) plotted against the absorber’s restframe velocity. The names of the background QSOs and the absorption redshifts are indicated above each panel. A complete set of velocity plots for all detected Si iii absorbers discussed in this paper is available in the Appendix.

Open with DEXTER

For the identification of intervening Si iii absorbers at z ≤ 0.1 in our COS data sample we used the following strategy. In a first step, we let an automated line-finder algorithm identify absorption features whose wavelengths would correspond to combined Si iii/H i Ly α absorption and created a candidate list of possible Si iii absorbers. In a second step, we inspected each individual spectrum by eye and created a different (independent) candidate list. For each of the candidates from both lists velocity plots (including all relevant low and high ions in the available wavelength range) were generated and the absorber candidates were analyzed in detail. To meet our selection criteria of a bona fide Si iii absorber, we require the significant detection of Si iii and H i together with the detection of at least one additional metal ion (C iv, C ii, Si iv, Si ii) to avoid mis-indentifications and false detections due to absorption features of other intervening absorbers at higher redshifts. Si iii traces both diffuse ionized gas (as also traced by high ions such as C iv and Si iv) as well as denser, partly neutral gas (as also traced by singly-ionized species such as Si ii and Mg ii; see Fig. 1), so that the simultaneous detection of Si iii with either one of the above listed low and high ions is expected.

We also identified a number of absorber candidates, that possibly show absorption in H i and Si iii but lack absorption from any other ion, so that a unambiguous identification cannot be given.

For each intervening Si iii absorber we checked for associated absorption in other metal ions (in particular Si ii, Si iv, C ii, C iv, N v, Fe ii). We then measured equivalent widths (and their limits) for the strongest lines of the three different ionization states of silicon (Si ii λ1260, Si iii λ1206, Si iv λ1393), C ii λ1334, and C iv λ1548, as well as for H i Ly α by a direct pixel integration. Because of the limited spectral resolution and S/N of the COS data and the resulting lack of information about the intrinsic component structure in the absorbers we refrain from performing line fits for the analysis of the metal absorption in the systems. We instead use the apparent-optical depth method (AOD method; Savage & Sembach 1991) to derive column densities for unsaturated absorption in the above-listed lines and lower limits for lines that appear to be saturated. We assume that saturation becomes important for lines that have an absorption depth >0.3 at the resolution of COS. Only for a small number of absorbers do we use Voigt-profile fitting to estimate the total H i column density from the damping wings of the H i Ly α absorption. The AOD method and Voigt-profile fitting are implemented in the custom-written span software package that is based on ESO-MIDAS and its data-reduction libraries (Fontana & Ballester 1995). Laboratory wavelengths and oscillator strengths for all ions were adopted from the compilation by Morton (2003).

3.3. Galaxy data

To statistically investigate the relation between Si iii absorbers and low-redshift galaxies we searched for publicly available galaxy data in the SIMBAD data archive1. Using SIMBAD we generated a list of galaxies with known redshifts located at z ≤ 0.1 and within 2deg of each sightline. A redshift of z = 0.1 corresponds to a proper distance of 420.9 Mpc at z = 0.1 for a standard ΛCDM cosmology with ΩΛ = 0.72,Ωm = 0.28 and H0 = 69.7 km s-1 Mpc-1 (Hinshaw et al. 20132). In this way, we obtained redshifts and coordinates for 64 280 galaxies distributed around the 303 COS sightlines. For 40 907 of these galaxies we have additional information on their B-band magnitudes. The distribution of mB for these systems is shown in Fig. 3. The distribution breaks down at mB = 20.0, with only two percent of the galaxies having mB> 20.0. For z = 0.1 and the above given cosmology, mB = 20.0 corresponds to an absolute magnitude of MB ≈ −18.1. If we consider the g-band SDSS galaxy-luminosity function (Montero-Dorta & Prada 2009) as reference, this value translates into a lower luminosity cutoff of L ≈ 0.5 L.

For each sightline, we then calculated the projected impact parameters, ρ, of the galaxies to the LOS. There are 11 127 galaxies within ρ = 1 Mpc located in the cylindrical volume around the 303 sightlines. This number implies a mean galaxy density of φ ≈ 0.03 Mpc-3 in our total galaxy sample, which is about seven times the space density of L galaxies in the SDSS g-band (Montero-Dorta & Prada 2009).

At this point it is important to mention that the galaxy catalog created in this way is highly inhomogeneous because the data stems from different galaxy surveys and pointed observations. As a result, the individual values for φ for each sightline vary considerably. In terms of luminosities, only a few sightlines have galaxy data that are sufficiently deep to detect faint (L ≤ 0.05 L) galaxies. In general, our galaxy data is incomplete for L< 0.5L. This aspect needs to be carefully taken into account for the interpretation of the observed absorber/galaxy relation. For many sightlines, we are missing low-luminosity galaxies that may be close to the LOS and/or responsible for the observed metal absorption (see discussion in Sect. 6). We do not consider any other morphological parameters of the selected galaxies (Hubble type, star-formation rate, etc.), because such information is available to us only for a small fraction of the galaxies in our sample.

4. Absorber properties

4.1. Si III detection rate, equivalent widths and column densities

Using the above outlined strategy, we identify 69 intervening Si iii absorbers along the 303 selected QSO sightlines. The statistical and physical properties of these 69 intervening absorbers are discussed in this section.

In Table A.2 in the Appendix we list the measured equivalent widths and derived column densities for the various ions detected in these systems. These tables also contain information on two additional strong metal absorbers towards SDSSJ141542.90+163413.8 and UKS0242-724 that are detected in Si ii, Si iv, and other ions, but not in Si iii, because there are no useful data in the relevant spectral region where Si iii is expected. In Fig. 2 we show five examples for velocity profiles of intervening Si iii absorbers with different Si iii absorption strengths. The complete set of velocity profiles for all 69 confirmed Si iii absorbers is provided in Figs. A.1A.7 in the Appendix of this paper. Following the criteria defined in the previous section we further identify 20 Si iii candidate absorbers along the 303 lines of sight. Although these candidate systems are not considered in the further analysis, we list the QSO names and (possible) absorption redshifts of these systems in Table A.3 in the Appendix.

thumbnail Fig. 3

Distribution of B-band magnitudes for ~41 000 galaxies in our sample.

Open with DEXTER

The redshift distribution of the 69 absorbers is shown in Fig. 4, left panel. The absorption redshifts lie between z = 0.00014 and z = 0.09789. While for z = 0.01−0.10 the absorbers are uniformly distributed over the surveyed redshift range, the distribution shows a peak in the first redshift bin at z< 0.01. This peak is related to an overdensity of Si iii absorbers in the broader Virgo-Cluster environment that is traced along several sightlines in our QSO sample. In the middle panel of Fig. 4 we show the distribution of Si iii λ1206 equivalent widths of the 69 absorbers. The distribution peaks at relatively low equivalent widths (Wr ≤ 100 mÅ); 65 percent of the absorbers have Wr ≤ 200 mÅ, while most of the remaining 35 percent are spread over a large range in Wr between 200 and 700 mÅ. There are two systems that have very large equivalent widths of Wr> 700 mÅ, belonging to strong absorption systems towards SDSSJ140732.25+550725.6 and PG 0832+251 (see Appendix).

Si iii column densities for the 69 absorbers, as derived from the AOD method (see previous section), are shown in the right panel of Fig. 4. The gray-shaded area (which adds to the green-shaded area) indicates lower limits of N(Si iii) for absorbers where the Si iii λ1206 line shows evidence for saturation. The decline of the observed distribution at low column densities reflects both the inhomogeneous S/N in the spectra as well as the column density distribution that is intrinsic to the absorber population. From the estimate of Nlim(Si iii) for each sightline (see Eq. (1)) follows that more than 90 percent of all sightlines are sensitive to detect Si iii absorbers with log N(Si iii)< 12.2, but only four such systems are found in our data (fractional abundance 4/69 ≈ 0.04). Therefore, intervening Si iii absorbers with log N(Si iii)< 12.2 are rare and may even represent a population that is distinct from the absorbers with higher column densities (as is discussed later). At the high-column density end, the distribution breaks down at log N(Si iii)= 14. Even if some of the saturated absorbers (gray-shaded range) would have logarithmic Si iii column densities > 14, their number would be small compared to absorbers with log N(Si iii)< 14. Our conclusion is that the characteristic column density range for intervening Si iii absorbers is log N(Si iii)= 12.2−14.0.

thumbnail Fig. 4

Left panel: redshift distribution of the 69 intervening Si iii absorbers in our sample. Middle panel: distribution of Si iii λ1206 restframe equivalent widths of the 69 absorbers, obtained from the pixel integration over the Si iii velocity profiles. Right panel: distribution of logarithmic Si iii column densities for the 69 absorbers, derived using the AOD method. The gray-shaded area indicates lower limits for log N(Si iii) for absorbers that have saturated Si iii λ1206 lines.

Open with DEXTER

4.2. Number density of Si III absorbers

For the determination of the number density of intervening Si iii absorbers per unit redshifz, , we need to consider in detail the completeness of our absorber survey and the selection bias in our QSO sample.

As discussed above, log Nlim(Si iii)≥ 12.2 represents the column density range that is characteristic for intervening Si iii absorbers. 63 Si iii absorbers in our sample have column densities log Nlim(Si iii)≥ 12.2 and 280 out of the 303 sightlines are sensitive to this level, covering a total redshift path of Δzabs = 21.95. The resulting number density thus is . The given errors represent statistical errors calculated from Poisson statistics. For log Nlim(Si iii)= 11.8 we derive (31 absorbers, Δzabs,tot = 7.8), while for log Nlim(Si iii)= 12.6 the number density is slightly smaller (, 49 absorbers, Δzabs,tot = 23.6). If we consider only low-column density absorbers with log N(Si iii)= 11.8−12.2 we obtain .

One important effect that influences the observed frequency of intervening absorption systems is the selection bias in the QSO sample that we are using. Most of the bright QSOs in our COS sample have been previously studied using instruments such as HST/STIS, FUSE, and others. While the original motivation to select these background sources certainly was their intrinsic brightness, the re-observation of these sources with COS and the chosen integration time (e.g., the achieved S/N), possibly was motivated by targeting particularly interesting intervening absorbers to perform a detailed study of these systems. Therefore, one of the reasons for limiting the absorber sample to redshifts z ≤ 0.1 was to avoid the inclusion of targeted observations of stronger intervening metal absorbers at z> 0.12 that can be observed in both H i Ly α as well as Ly β.

We have scanned the various original COS proposals that outline the motivation for observing the LOS along which we detect intervening Si iii at z ≤ 0.1. As it turns out, our QSO sample covers most of the sightlines selected for the COS-Dwarfs survey (Bordoloi et al. 2014), a targeted survey of sightlines passing through the virial radius of z ≤ 0.1 dwarf galaxies to study the CGM of these systems. For an un-biased estimate of (Si iii) all these sightlines need to be excluded. All in all, we identify 40 QSO sightlines in our sample that have been specifically observed to study circumgalactic gas in the vicinity of known low-redshift galaxies. By removing these sightlines from our sample we derive the following bias-corrected number densities of Si iii absorbers for the different limiting Si iii column densities: for log Nlim(Si iii)= 11.8, for log Nlim(Si iii)= 12.2, and for log Nlim(Si iii)= 12.6. For the range log N(Si iii)= 11.8−12.2 we again obtain .

It is evident that these numbers are not substantially different from those derived from the biased sample. This is, however, not surprising because it is known that the UV absorption cross section of the warm CGM around dwarf galaxies appears to be small when compared to the warm CGM of more massive galaxies (e.g., Bordoloi et al. 2014; Liang & Chen 2014; see also Sect. 8.3). In other words: only a few additional Si iii absorbers are (in a statistical sense) added to our absorber sample when including the pre-selected Bordoloi et al. sightlines, but because of the overall large size of our QSO sample their influence on is small.

In Fig. 5 we show the bias-corrected number density of intervening Si iii absorbers per unit redshift in our survey, , for the three different limiting Si iii column densities, log Nlim(Si iii)= 11.8,12.2 (dashed red line), and 12.6. In the following, we refer to these values when discussing the number densities of intervening Si iii absorbers.

4.3. Associated absorption from other ions

4.3.1. H I

As outlined in the introduction, the Si iii ion is, because of its ionization potential, a very sensitive tracer for both predominantly neutral as well as predominantly ionized gas in the halos of galaxies. As a consequence, the H i column density in the Si iii absorbing gas varies over several orders of magnitudes, ranging from log N(H i)~ 14 up to log N(H i)~ 22, depending on whether the sight line passes through a galaxy disk or a galaxy halo, and on the local ionization conditions. Detailed ionization models of the Si iii absorbers are presented in Sect. 7. In the left panel of Fig. 6 we show the distribution of H i Ly α equivalent widths in the Si iii-selected absorbers, which similarly span an extremely large range from Wr ≈ 250 mÅ  to Wr ≈ 5 Å. We generally refrain from estimating the H i column densities from the saturated H i Ly α lines because of the very large uncertainties that such an estimate would be afflicted with.

It is worth noting that our absorber sample contains four damped Ly α/sub-damped Ly α systems (DLAs/sub-DLAs) with log N(H i)> 19.2, as estimated from fitting the damping wings of the Ly α absorption in these systems. With a total redshift path of Δzabs,tot = 21.95, this absorber frequency corresponds to a number density of . Despite the low-number statistics, it is worth mentioning this value agrees well with the expected number density of sub-DLAs/DLAs as estimated from the H i mass function of z ≈ 0 galaxies (Zwaan et al. 2005).

4.3.2. Si II

thumbnail Fig. 5

Number densities, , of intervening Si iii absorbers for different limiting column densities.

Open with DEXTER

Detailed ionization models (see Sect. 7) indicate that regions with gas densities log nH ≥ −3.5 and sufficiently large neutral hydrogen column densities are expected to show both Si iii as well as Si ii absorption. Because Si ii and Mg ii have almost identical ionization potentials (Morton 2003) and silicon and magnesium have similar solar abundances (Asplund et al. 2009), the absorber statistics for Si ii can be directly compared with the well-established statistics on intervening Mg ii absorbers. The presence of Si ii absorption without associated Si iii is basically impossible for any realistic galactic or circumgalactic gas environment (see Sect. 7), so that the fact that our absorber sample is Si iii-selected is not expected to introduce a selection bias in our Si ii statistics.

Fourty of our Si iii absorbers show associated Si ii absorption in the strongest of the available Si ii lines at λ0 = 1260.42 Å. The equivalent-width distribution for Si ii λ1260 in these absorbers is shown in the middle panel of Fig. 6. It shows that the majority (80 percent) of the detected Si ii lines have equivalent widths in the λ1260 line of <300 mÅ. To compare the observed number density of Si ii-bearing absorbers with that of the prominent strong Mg ii systems, i.e., systems that have an equivalent width of Wr ≥ 300 mÅ in the Mg ii λ2976 line, we need to convert the observed Si ii λ1260 equivalent widths into Mg iiλ2976 equivalent widths. For this we use the Si ii/Mg ii conversion scheme presented in Herenz et al. (2013), after which an equivalent widths of Wr = 300 mÅ  in the Mg ii λ2976 line corresponds to an equivalent width of Wr = 140 mÅ in Si ii λ1260, assuming solar relative abundances of Mg and Si. In our absorber sample we have 22 systems that have Wr ≥ 140 mÅ in the Si iiλ1260 line, from which we indirectly infer a number density of strong Mg ii absorbers at z ≤ 0.1 of . This value is in good agreement with the expectations for (Mg iiλ2976) based on the redshift evolution of strong Mg ii absorbers in SDSS data (; Nestor et al. 2005; Prochter et al. 2006; Lundgren et al. 2009; Zhu & Menard 2013).

4.3.3. Si IV

There are 38 Si iii absorbers in our sample that show associated Si iv absorption (the two available Si iv transitions are located at λ0 = 1393.8 and 1402.8 Å, the former being the stronger of the two transitions; Morton 2003). The distribution of Si iv λ1393 equivalent widths of the 38 absorbers detected in Si iv is shown in the right panel of Fig. 6. Similarly as for Si ii, the majority of the Si iv λ1393 equivalent widths are small (82 percent have Wr(1393) < 300 mÅ). From the observed frequency of Si iv absorption in the Si iii-selected systems we estimate (Si iv))≥ 1.5 for log N(Si iv)≥ 12.6. The occurrence of Si iv absorption without associated Si iii absorption in the CGM/IGM is possible, in principle, but is relevant only for absorbers at relatively low gas densities (log nH ≤ −4.2; see Sect. 7). Still, we formally can only give a lower limit for (Si iv) from our survey.

thumbnail Fig. 6

Distribution of restframe equivalent widths of H i Ly α (left panel), Si ii λ1260 (middle panel), and Si iv λ1393 (right panel) associated with intervening Si iii absorbers.

Open with DEXTER

thumbnail Fig. 7

Comparison between restframe equivalent widths of Si iii, Si ii, Si iv, and H i for 69 absorbers. First panel (left): Si iii λ1206 vs. H i Ly α; second panel: Si iii λ1206 vs. Si ii λ1260; third panel: Si iii λ1206 vs. Si iv λ1393; fourth panel: Si ii λ1260 vs. Si iv λ1393.

Open with DEXTER

4.4. Correlation plots

In Fig. 7 we show correlations between the measured (logarithmic) equivalent widths of Si iii λ1206, H i Ly α, Si ii λ1260, and Si iv λ1393 for absorbers in which the absorption in these ions is aligned in velocity space within one COS resolution element (20 km s-1). The interpretation of the observed correlations involves the possible presence of different gas phases in the absorbers that may or may not be co-spatial within the overall gas structures.

4.4.1. Si III vs. H I

For log [ Wr(Ly α) ] ≤ 3 the Si iii λ1206 equivalent width rises steeply with the mildly increasing equivalent width of the fully saturated H i Ly α absorption (Fig. 7; left panel). This trend indicates (as expected) that Si iii and H i trace the same physical regions that span a large range in neutral (and total) gas column densities. For log [ Wr(Ly α) ] > 3 the correlation turns over into a somewhat flatter regime because the Si iii λ1206 absorption itself becomes saturated at such high total gas columns.

4.4.2. Si III vs. Si II

The equivalent widths of Si iii λ1206 and Si ii λ1260 clearly are correlated with each other in those absorbers, where both ions are detected (Fig. 7; second panel from left, filled circles). This demonstrates that part of the Si iii arises in the same gas phase as Si ii, i.e., in predominantly neutral and/or mildly ionized gas. Some of the systems with upper limits in Wr(λ1260) (open circles/arrows) have relatively strong Si iii absorption without a Si ii counterpart, implying that these absorbers consist of predominantly ionized gas that is traced by higher ions (e.g., Si iv, see below).

4.4.3. Si III vs. Si iv

For systems that show both Si iii as well as Si iv absorption the equivalent width of Si iii λ1206 is also correlated with that of Si iv λ1393, as can be seen in the third panel of Fig. 7 (filled circles). This trend indicates that some part of the Si iii resides in a more ionized gas phase that is traced by Si iv (and other high ions such as C iv). The relation is mildly steeper than the one for Si iii/Si ii, but has as similar level of scatter.

4.4.4. Si II vs. Si iv

In the right panel of Fig. 7 we have plotted the equivalent width of Si ii λ1260 against that of Si iv λ1393 for the absorbers where both ions are detected. The total number of systems that show aligned Si ii and Si iv absorption is relatively small (38 percent). This implies that both ions predominantly trace complementary gas phases. For the systems detected in both ions (filled circles) the correlation between Wr(Si ii λ1260) and Wr(Si iv λ1393) is weak and shows a relatively large scatter.

4.4.5. Interpretation

We conclude that the correlation plots between the equivalent widths of Si iii λ1206, H i Ly α, Si ii λ1260, and Si iv λ1393 are in line with (and further support) the idea that the Si iii absorption in intervening metal-systems traces metal-enriched gas within a wide range of physical conditions including a) a denser (partly neutral) phase also traced by Si ii and other low ions; and b) a more diffuse (predominantly ionized) gas phase also traced by Si iv and other high ions. We further discuss these aspects in Sect. 7 where we model the ionization conditions in Si iii/Si iv absorbers in detail.

In a recent C iv-selected survey of absorbers at z ≤ 0.16 Burchett et al. (2015) find very similar relations between Si ii, Si iii, and Si iv in their absorber sample (their Figs. 15 and 16).

5. On the expected cross section of metal-enriched gas in galaxy halos

Before we investigate in detail the observed relation between Si iii absorbers and galaxies in a statistical sense, we first set up a geometrical model to estimate the expected cross section of metal-enriched gas in the extended gaseous halos of galaxies at z = 0.

Under the assumption that all intervening Si iii absorbers at low z are related to metal-enriched gas situated in the extended halos of galaxies, the observed number density of Si iii absorbers can be directly linked to the space density of galaxies, φ, and the effective (geometrical) cross section of the absorbing gas, which is a product of the projected area covered by the gaseous halo and the mean covering fraction fc of the gas phase that is seen in absorption:

(2)The Hubble parameter is defined as H(z) = H0m (1 + z)3 + ΩΛ)1/2, which is appropriate for a matter-dominated flat Universe with a cosmological constant. Eq. (2), and modified versions of it, have been commonly used to estimate the sizes of galaxy halos and the covering fractions of individual ions from QSO absorption-line observations (e.g., Kapzcrak et al. 2008; Richter et al. 2011; Prochaska et al. 2011).

We here take the opposite point of view and pose the following question: what would be the expected number density of intervening metal absorbers, if all galaxies at z = 0 contain detectable metal-enriched gas in their halos that extends exactly to their respective virial radii?

Using Eq. (2) it is indeed relatively straight-forward to set up a “toy model” for the absorption cross section of extended halo gas taking into account the observed space density and luminosity/mass distribution of galaxies at low redshift. The motivation for such a reverse approach is rather simple: if we could know the maximum contribution of metal-enriched gas that is gravitationally bound to galaxies to the number density of intervening metal absorbers, we would have an important reference value for the interpretation of the observed number densities of intervening Si iii systems and their origin in the CGM and/or IGM.

Both the galaxy density φ as well as the distribution of the galaxies’ virial radii at z = 0, i.e., the most important parameters to calculate via Eq. (2), can be obtained indirectly from the local galaxy luminosity function. We here adopt the g-band SDSS luminosity function from Montero-Dorta & Prada (2009), who give Schechter parameters of α = −1.10 and φ = 1.25 × 10-2h3 Mpc-3 for h = 1.0. We transform these parameters to the cosmological frame defined in Sect. 3.3. and calculate φ = φ Γ(α + 1,L/L) for different luminosity bins, where Γ is the incomplete gamma function and L is the characteristic luminosity that characterizes the cut-off for the power-law component in the Schechter luminosity function (Schechter 1976).

Table 1

Models for absorber cross sections based on Eq. (2).

thumbnail Fig. 8

Predictions from the halo model described in Sect. 5. Upper panel: the virial radius of galaxies as a function of galaxy luminosity (adapted from Stocke et al. 2014). Middle panel: Galaxy space density as a function of galaxy luminosity (from Montero-Dorta & Prada 2009). Lower panel: expected number density of CGM absorbers as a function of galaxy luminosity.

Open with DEXTER

The key assumption in our toy model is that all halos of galaxies with luminosities L ≥ 0.001 L contain metal-absorbing gas within their virial radii (Rvir). We initially assume a unity covering fraction of the absorbing gas, so that the effective absorption cross section is simply . Also the virial radius of a galaxy can be inferred indirectly from its luminosity. As recently discussed by Stocke et al. (2014), halo matching models provide a relatively well-defined relation between Rvir and L that can be used to define a scaling relation between these two parameters. We define C = log  (L/L) and approximate the relation between L and Rvir shown in Stocke et al.  (their Fig. 8) by the fit (3)Equipped with these relations we can now assign to each luminosity bin a galaxy space density φ(L) and an effective absorption cross section , which is all we need to calculate the expected number density of absorbers for the same bin. Integration over all galaxy luminosity bins then provides the expected total number density of absorbers arising from the halos of these galaxies.

Figure 8 summarizes the main results from our modeling. The red solid line in the upper panel visualizes the LRvir-relation defined by Eq. (3), while the green-shaded histogram indicates the same relation in bins of log (L/L) = 0.5. In the middle panel of Fig. 8 the logarithmic galaxy density (log φ(L)) is plotted against log (L/L) (same binning), while in the lower panel we show the expected number density of absorbers (assuming a unity covering fraction, fc ⟩ = 1) as a function of log (L/L) for the same binning. By integrating over the desired luminosity range we obtain an estimate for the total number density of absorbers that arise in the gaseous halos of the galaxies in that luminosity range.

While our observational galaxy sample is restricted to relatively bright galaxies (see Sect. 3), we here consider the much larger range L/L ≥ 0.001 to evaluate the potential contribution of faint galaxies and their halos to the total cross section of intervening metal absorbers, even if they remain unseen in galaxy surveys. For L/L ≥ 0.001 and unity covering fraction we obtain a total absorber number density of . If we include the mean covering fraction as free parameter (which then represents an area-weighted mean), we can write more generally:(4)For realistic gas covering fractions in galaxy halos in the range fc ⟩ = 0.1−1.0 this absorber number density is remarkably close to the observed number density of intervening Si iii absorbers, as discussed in the previous section.

One serious concern about the relevance of the above given estimate is, how strongly the derived absorber number density depends on the input parameters from the adopted galaxy luminosity function. To explore this dependence we have calculated for a whole set of (realistic) Schechter parameters, as presented in Table 1. For this, we vary for a given slope α in each row the normalization density from over to (where Mpc-3 represents our preferred normalization; see above) and calculate the expected absorber number densities , , . For Schechter parameters in the range α = [ −1.0,−1.3 ] and φ = [ 0.0075,0.150 ] the expected number densities vary between and 8.35. It is therefore valid to claim that the expected contribution of the metal-enriched CGM to the cross section of intervening metal absorbers is (this conclusion holds for any metal ion that can be observed via QSO absorption spectroscopy).

One may argue that the values for estimated in this manner over-estimate the true absorption-cross section of gaseous halos because galaxy halos can overlap with each other (in particular satellite galaxies and their hosts), so that the effective halos cross section of a galaxy ensemble would be smaller than the sum of A for all individual galaxies in that ensemble. This effect is commonly investigated in studies that aim at characterizing the relationship between the stellar masses of galaxies and their superordinate dark matter (DM) halos, e.g., by using statistical methods such as the Halo Occupation Distribution (HOD) formalism and others (e.g., Peacock & Smith 2000; Moster et al. 2010). From these models follows that, on average, only the most massive halos (with DM halo masses >1011.8M, corresponding to L> 2 L) are occupied by more than one galaxy. These halos are rare, however, and their contribution to is small (see Fig. 8), so that a small change in the absorption cross section of such massive halos basically would have no influence on the estimate of the integrated number density of intervening absorbers.

A comparison between the estimated gas cross section of galaxy halos from our model with the observed number density of intervening Si iii absorbers allows us to draw some very interesting conclusions. The first important conclusion is that,because Si iiimodel) for log N(Si iii)> 12.2, the entire population of these Si iii systems can be explained by metal-enriched gas that is gravitationally bound to galaxies, i.e., there is no need for an intergalactic gas component that hosts Si iii absorbers at this strength. Turning this argument around, our model also predicts that intervening metal absorbers with number densities (e.g., O vi absorbers with ; Tripp et al. 2008) must partially arise from gas outside of the virial radius of galaxies. This aspect is further discussed in Sect. 8. A second conclusion is that for our preferred model with α = −1.1 and φ = 0.0125 h3 Mpc-3 (Montero-Dorta & Prada 2009) the mean covering fraction of Si iii within Rvir is expected to be fc ⟩ = 2.5 ± 0.4/3.6 = 0.69 ± 0.11, a value that is similar to the covering fraction of Si iii in the halo of the Milky Way, as recently derived from QSO absorption-line observations (Herenz et al. 2013; Lehner et al. 2012). In Sect. 6 we further investigate, whether the observed distribution of galaxies around Si iii absorbers is in line with these conclusions.

thumbnail Fig. 9

CGM absorption fraction as a function of impact parameter, as predicted from the halo model described in Sect. 5.

Open with DEXTER

Our model further allows us to predict the mean absorption fraction around galaxies as a function of impact parameter ρ. We here define the mean absorption fraction as the detection rate of intervening absorption from CGM gas for a given impact parameter for a population of galaxies with different virial radii. Only the most luminous galaxies with large Rvir are expected to contribute to the mean absorption at ρ> 100 kpc, while for small impact parameters ρ< 50 kpc both luminous as well as faint galaxies do contribute. Consequently, the mean absorption fraction around galaxies is expected to decline in a way that is determined by the distribution of Rvir (and thus by the shape of the galaxy luminosity function).

In Fig. 9 we show the mean absorption fraction plotted against ρ (in bins of 50 kpc), as derived from our preferred model with Schechter parameters α = −1.1 and φ = 0.0125 h3 Mpc-3. As expected, the absorption rate decreases rapidly with increasing impact parameter because the fraction of sightlines that pass galaxies at ρ >Rvir increases. Figure 9 can be directly compared to the observed absorption rate around low-redshift galaxies, as is presented in Sect. 6.

Another important piece of information that can be extracted from our model concerns the expected covering fraction of circumgalactic gas around a population of galaxies in a given luminosity range. It is useful to transform the absorption rate into an effective covering fraction, ⟨ ℱc(rmax) ⟩, that is normalized to a fixed radius rmax (instead of being normalized to Rvir, which depends on the galaxy’s mass and luminosity). To carry out this transformation we need to take into account that the covering fraction describes the detection rate per unit area. Since the area of each ring with thickness Δr and radius r is Δr 2πr, the absorption rates need to be weighted with r and integrated from r = 0 to r = rmax to obtain ⟨ ℱc(rmax) ⟩. In this way, it is possible to predict from our model the effective covering fractions for different values of rmax for a population of galaxies in a given luminosity range. In Table 2 we list the predicted effective CGM covering fractions for different values of rmax, fc, and different luminosity ranges based on the preferred parameters for the galaxy luminosity function (see Table 1). We would like to emphasize again that in this approach ⟨ ℱc describes the average CGM covering fraction of a population of galaxies within an annulus with a fixed radius, while fc describes the mean covering fraction of the galaxies’ CGM within Rvir.

Table 2

Model predictions for effective CGM covering fractions, ⟨ ℱc(rmax) ⟩, for different halo radii and galaxy luminosities (see Sect. 5).

thumbnail Fig. 10

Upper panels: distributions of impact parameters and absolute B-band magnitudes for galaxies that are located within | Δv | = 1000 km s-1 of the absorbers (galaxy/absorber samples). On the left-hand side we show the distribtions for all sightlines that exhibit Si iii absorption, while on the right-hand side only the bias-corrected set of sightlines is considered (see Sect. 4.2). Galaxies tend to cluster around the LOS in velocity regions where intervening absorption is observed. Lower panels: the same distributions for galaxies that have no absorption systems within | Δv | = 1000 km s-1 (composite control samples; see Sect. 6.2 for details). No clustering is observed in this case.

Open with DEXTER

6. Observed absorber-galaxy connection

6.1. Individual galaxies associated with Si iii absorbers

To characterize the true absorber-galaxy connection at z ≤ 0.1 we have calculated the geometrical impact parameters ρ between the sightlines that exhibit Si iii absorbers at z ≤ 0.1 in our COS data and nearby galaxies in our galaxy catalog (see Sect. 3.3).

Ideally, one would study the relation between absorbers and galaxies using the absorber-galaxy cross-correlation function together with a well-defined galaxy completeness function. Our galaxy sample, that is based on many different galaxy surveys and individual observations, is highly inhomogeneous with respect to the completeness limit at the faint end of the B-band magnitude distribution, however, and therefore a statistically meaningful galaxy completeness function cannot be obtained. We would like to point out, however, that the motivation for including these galaxy data in our study is not to provide a statistically complete sample of absorber/galaxy pairs, but to qualitatively check the plausibility of our hypothesis, in which most Si iii absorbers at z ≤ 0.1 arise in the extended gaseous halos of luminous low-redshift galaxies, for which the present galaxy data sample is sufficient in depth.

Because of the above mentioned limitations we cannot simply combine the impact-parameter distributions of the individual sightlines to statistically investigate the overall absorber-galaxy connection, but instead have to evaluate the connection between absorbers and galaxies along each sightline individually before combining the results. In the following, we further specify our strategy how to do this.

For each sightline that contains an intervening Si iii absorber we first selected all galaxies that have recession velocities within | Δv | = 1000 km s-1 of the absorber (hereafter referred to as “LOS absorber sample”) and identified the galaxy with the smallest impact parameter to the sightline. For the same line of sight we then created a set of nine galaxy control samples covering the adjacent velocity range | Δv | = 1000−10 000 km s-1, where each LOS control sample spans a velocity range of | Δv | = 1000 km s-1, similar as for the LOS absorber sample. We again marked the galaxies that are nearest to the LOS. The galaxy samples constructed in this way have the same magnitude completeness for each sightline.

If intervening absorbers would trace the CGM of galaxies that are close to the sightlines (as we suspect), one would expect for a large-enough absorber/galaxy sample (as provided here) that the impact parameters of the nearest galaxy to the LOS in the absorber samples are, on average, substantially smaller than in the control samples. Such a trend is indeed seen in our data: for 76 percent of the LOS that exhibit an intervening Si iii absorber, the galaxy with the smallest impact parameter is located within 1000 km s-1 of the absorption redshift.

In the upper panels of Fig. 10 (gray-shaded histograms) we show the distribution of ρ for the nearest galaxies in the LOS absorber samples (“galaxy/absorber sample”) for all sightlines with Si iii absorbers (left column) and for the bias-corrected set of sightlines with Si iii absorbers (right column). Obviously, galaxies cluster around sightlines if an Si iii absorber is present. This trend is evident in both galaxy/absorber samples (total sample and bias-corrected sample), but in the (biased) total sample there are more data points in the lowest ρ-bin, as is expected from the design of the Bordoloi et al. (2014) COS-Dwarfs survey (see Sect. 4.2).

Still, in the bias-corrected galaxy/absorber sample, 56 percent of the galaxies are located at impact parameters ρ< 200 kpc and for the same percentage the velocity difference between the galaxy and the absorber is | Δv | ≤ 300 km s-1. From our model we would expect that the maximum impact parameter in an idealized galaxy sample is ρ ≈ 300 kpc, as this length scale corresponds to the virial radius of the most luminous galaxies that are expected to significantly contribute to the number density of Si iii absorbers before the galaxy-density distribution breaks down for higher luminosities (Fig. 8, middle panel). However, since our galaxy sample is incomplete and highly inhomogeneous along the different LOS with respect to galaxies at the faint end of the luminosity function, the observed distribution of ρ extends much further out. With the green-shaded histogram we show in Fig. 10 the absolute B-band magnitudes for the galaxies in the galaxy/absorber samples, as calculated from their cosmological distance. The dotted line indicates the absolute B magnitude of an L galaxy (adopted from Montero-Dorta & Prada 2009).

thumbnail Fig. 11

Restframe equivalent widths of Si iii λ1206 are plotted against the impact parameters of the galaxies nearest to the absorbers. For increasing impact parameters the maximum value for Wr is decreasing from Wr ≈ 900 mÅ down to values <100 mÅ.

Open with DEXTER

To check the statistical significance of this clustering trend we analyzed the galaxies in the respective control samples in a similar manner. We find that the galaxies do not cluster around the sightlines in any of the control samples, but their impact parameters are more or less evenly distributed. To demonstrate this, we show in the lower left and right panels of Fig. 10 the impact-parameter distributions in the control samples, i.e., for galaxies that have no absorption systems within | Δv | = 1000 km s-1. For this, we combined for each of the two data sets (total sample/bias-corrected sample) the nine individual control samples to one “composite control sample”, respectively, which contains all galaxies within the velocity window | Δv | = 1000−10 000 km s-1 with respect to the absorber redshift (see above).

The distributions of absolute B-band magnitudes in the composite control samples (Fig. 10, lower left panel, green-shaded areas) are similar to those in the galaxy/absorber samples, proving that the different galaxy samples have the same completeness in B.

In summary, the observed absorber-galaxy relation strongly suggests that the galaxies “know” about the presence of nearby Si iii absorption systems. This is exactly what would be expected if the absorption were (predominantly) caused by metal-enriched gas in the extended halos and the superordinate cosmological environment of these galaxies.

For our bias-corrected galaxy/absorber sample we calculate effective covering fractions of ⟨ ℱc(rmax) ⟩ = 0.08 for rmax = 400 kpc and ⟨ ℱc(rmax) ⟩ = 0.01 for rmax = 1000 kpc. These effective covering fractions are comparable to those expected for a population of L> 0.1L galaxies that are surrounded by a CGM that reaches out to Rvir with fc ⟩ = 0.75 (Table 2).

6.2. Absorption strength vs. impact parameter

thumbnail Fig. 12

Absorption fraction (fractional abundance of intervening absorbers per galaxy) as a function of impact parameter, as measured in our galaxy/absorber sample.

Open with DEXTER

In Fig. 11 we plot the restframe equivalent widths of Si iiiλ1206 versus the impact parameters of the galaxies nearest to the absorbers. For ρ< 200 kpc the equivalent width scatters strongly in the range 10−1000 mÅ. For increasing impact parameters the maximum value for Wr (hereafter referred to as Wr,max) is decreasing to values <100 mÅ  for ρ> 500 kpc.

The interpretation of this plot is not as simple as it may look like: because of the incompleteness of the galaxy sample, the largest values for ρ (which are far beyond the virial radii of the most massive galaxies) do not represent true impact parameters to the nearest galaxies, but rather indicate the impact parameters to the nearest luminous galaxies. From observations of strong Mg ii absorbers it is known that the mean Mg ii λ2796 equivalent width in the CGM of luminous galaxies is larger than for low-luminosity galaxies (e.g., Nielsen et al. 2013). Assuming that a similar trend holds for Si iii, the large scatter for Wr(Si iiiλ1206) at ρ< 200 kpc can be interpreted by the large range in luminosities of the galaxies whose circumgalactic gas causes the absorption, where Wr,max is is determined by the most luminous galaxies. The decline of Wr,max for increasing ρ then simply reflects the decreasing likelihood to miss luminous galaxies around the absorbers that are responsible for for the strongest absorbers. For large values of ρ, only absorbers with small Wr values remain, as they belong to low-luminosity galaxies that are too faint to be covered in our galaxy sample.

6.3. QSO sightlines associated with individual galaxies

We now investigate the absorber/galaxy-connection from the perspective of the galaxies. For each galaxy in our sample we first identified QSO sightlines that pass the galaxy at impact parameters ρ< 1000 kpc. We then further selected only galaxies in redshift ranges that are covered by each relevant COS spectrum at a S/N that is high enough to detect Si iii absorption at column densities log Nlim(Si iii)≥ 12.2. Finally, we calculated for each 100 kpc wide impact-parameter bin the absorption fraction (i.e., the fraction of galaxies having a Si iii absorber above this column density limit in this impact-parameter range). The result from this analysis is shown in Fig. 12. As can be seen, the observed absorption fraction is substantially smaller than the one derived from the idealized halo model (Fig. 9), but extends to much larger impact parameters even beyond the expected virial radii of massive galaxies. We again interpret this behavior as a sign for the incompleteness in our galaxy data, i.e., we suspect that sightlines that have apparent galaxy impact parameters in the range 300 ≤ ρ< 1000 kpc are arising from halo gas within the virial radii of unseen galaxies.

7. Ionization conditions

7.1. Model setup

To gain insight into the physical properties of the intervening Si iii absorbers and to estimate the total gas mass that they trace we studied the ionization conditions in the absorbers using the ionization code Cloudy (v13.03; Ferland et al. 2013). The absorbers are modelled as plane-parallel slabs with fixed neutral gas column densities; they are exposed to the UV background radiation with log Γ = −13.6 (see Sect. 2 for more details on the used UV background field) and are assumed to be optically thin in H i.

In Sect. 4 we have suggested that Si iii traces at least two characteristic gas phases, a more ionized, diffuse phase that is also traced by Si iv, and a more dense (partly neutral) phase that is also detected in Si ii. For our Cloudy modeling we therefore have focused on the relation between the column densities of H i, Si ii, Si iii, and Si iv as a function of the ionization parameter U, the ratio between ionizing photon density and total particle density (i.e., U = nγ/nH). For an assumed ionizing radiation field one can calculate nγ and thus can relate U with the gas density nH.

In Figure 13 we have plotted (as an example) the expected H ii, Si ii, Si iii, and Si iv column densities against log nH for a (typical) H i column density of log N(H i)= 15, based on a set of Cloudy models assuming solar abundances of all heavy elements (Asplund et al. 2009). We also show the expected trend for Mg ii, which follows Si ii very closely over the entire density range. This similarity justifies our previous approach to estimate the number density of strong Mg ii using Si ii as a proxy (Sect. 4.3.2). We have set up a grid of Cloudy models in which we have varied the H i column density in the range log N(H i)= 14−19 to provide column-density predictions for the above listed ions.

The most important conclusion from this modeling is that the shapes of the column-density curves for H ii, Si ii, Si iii, Si iv, and Mg ii (and their positions relative to each other) basically do not depend on log N(H i), implying that the observed column density ratios of the Si ions can be used to constrain nH in optically thin H i/Si iii absorbers even without knowing N(H i).

Figure 13 further indicates that Si iv becomes the dominant ion of Si only for relatively low gas densities (log nH< −3.5), while Si iii dominates in the density range −3.5 < log nH ≤ −2.0. Si ii is dominant at log nH> −2.0, thus at densities that are expected to be relevant only for DLAs and sub-DLAs in the neutral gas disks and in predominantly neutral gas structures in the inner halos of galaxies (e.g., in high-velocity clouds).

7.2. Single-phase model

thumbnail Fig. 13

Predictions of H ii, Si ii, Si iii, Si iv, and Mg ii column densities as a function of the gas density (log nH) for an H i column density of log N(H i)= 15, based on a Cloudy ionization model using the local UV background and assuming absolute and relative solar abundances of all elements. The gray-shaded area indicates the density that is typical for the Si iii/Si iv phase (see Sect. 7.3).

Open with DEXTER

One may assume that the simultaneous absorption of Si ii, Si iii, and Si iv at similar radial velocities, such as observed in some systems, stems from a single gas phase in the absorbers. For that case, our Cloudy models provide some firm predictions for the allowed column-density ranges for these three ions that can be summarized by the following parametrization of the expected column-density ratios Si iii/Si iv and Si iii/Si ii in a single gas phase:

(5)In Fig. 14 we have plotted the observed column-density ratios of these ions (and relevant limits) together with predictions from the Cloudy model as given in Eq. (5) (the model is indicated in Fig. 14 by the black solid line). For none of the systems that have measured column densities for Si ii, Si iii, and Si iv (Fig. 14 filled circles) do the data points lie on the expected relation for the single-phase model. For all systems for which we have measured values in these three ions the Si ii and/or Si iv column densities are too high for the observed Si iii column density to match the single-phase model. While we cannot exclude that at least some of the sytems for which only lower limits for Si iii/Si iv and Si iii/Si ii are available are in accordance with the single-phase model, the observations clearly do not favor such a scenario, but rather point towards a more complex multiphase nature of gas, as considered below.

thumbnail Fig. 14

Column-density ratios of Si iii/Si ii vs.  Si iii/Si iv are plotted together with predictions from the single-phase Cloudy model (Eq. (5), indicated by the black solid line). Measured values are plotted with filled circles, limits are indicated with open circles. For none of the systems do the data points lie on the expected relation for the single-phase model, implying that the gas is multiphase.

Open with DEXTER

7.3. Multi-phase model

The alternative (and probably more realistic) model for the absorbers is that of a multiphase gas, where Si ii and Si iv predominantly trace different gas phases (and different physical regions) that coexist within the same overall absorbing gas structures. In the Milky Way halo, the existence of multiphase halo gas as traced be various low, intermediate, and high ions is well established (e.g., Sembach et al. 2003; Fox et al. 2006; Collins et al. 2009; Shull et al. 2009; Richter et al. 2009; Herenz et al. 2013).

For the following we assume that Si iii absorption arises in both a diffuse ionized gas phase traced by Si iv/Si iii as well as a somewhat denser (partly neutral) gas phase traced by Si ii/Si iii. To further investigate the characteristic densities of these two phases we have plotted in Fig. 15 the column densities log N(Si iii) vs. log N(Si iv) (upper panel) and log N(Si iii) vs. log N(Si ii) (lower panel). Data points from measured column density limits in these ions are plotted with open circles. Because both gas phases defined above may be present in an absorber, the plotted values of log N(Si iii) have to be regarded as an upper limit for the Si iii column in each phase.

thumbnail Fig. 15

Distribution of logarithmic column densities in the Si iii absorbers. Upper panel: Si iv vs. Si iii (absolute values indicated with filled circles, limits with open circles). Lower panel: Si ii vs. Si iii. The dotted/solid lines indicate ratios of + 0.5,0,−0.5 dex, respectively.

Open with DEXTER

For the absorbers that are detected simultaneously in Si iii and Si iv (upper panel of Fig. 15, filled circles) the data points scatter within ~0.7 dex around the N(Si iii)= N(Si iv) line. For the absorbers that are detected simultaneously in Si iii and Si ii (lower panel of Fig. 15, filled circles) the measured Si ii column densities are typically lower than that of Si iii, but (again) note that a considerable fraction of the Si iii column may arise in the Si iv phase, so that log [N(Si ii)/N(Si iii)] could be much higher locally.

Because of the unknown intrinsic structure of each absorber it is challenging to provide firm predictions for the gas density for each individual system that has measured Si ii/Si iii/Si iv column densities. From the exploration of the parameter space the Cloudy models deliver, however, a characteristic gas density that separates the Si iv/Si iii from the Si ii/Si iii phase in the absorbers (see also Fig. 13). We find that the Si iv/Si iii phase traces gas with densities log nH ≤ −3.0 (in Fig. 13 indicated with the gray-shaded area), while the Si ii/Si iii phase has higher densities in the range log nH> −3.0.

7.4. Total gas mass and baryon budget

Our observations and Cloudy models imply that Si iii traces diffuse (predominantly ionized) gas in the extended gaseous halos (i.e., in the circumgalactic medium) of galaxies. An interesting question is, how much mass is contained in such gas and what is the overall baryon budget of intervening Si iii absorbers and the metal-enriched CGM. To derive the total gas mass of the Si iii systems in our sample we need to calculate the amount of ionized hydrogen in each absorber. Because of the much higher ionization fraction in the Si iv/Si iii systems compared to the Si ii/Si iii phase we here concentrate on the estimate of N(H ii) in low-density Si iii absorbers that are associated with Si iv.

For the range log N(H i)= 14−17 our Cloudy model grids imply a relatively simple relation between the minimum H ii column density and the Si iii column density in each absorber:(6)where the parameter Y, that solely depends on the gas density, needs to be determined from the observed Si iv/Si iii ion ratios (Fig. 13). Since we do not know what fraction of the Si iii column density can be assigned to the Si iv phase, we can only place a lower limit to Y (see above). Similarly, because of the unknown metallicity of the CGM, we have to assume an upper limit for the silicon abundance (Si/H) in the gas. Equation (6) allows us to derive a lower limit for N(H ii) in each Si iv/Si iii absorber, from which the integrated (=total) column density, N(H ii)tot, can then be determined.

The cosmological mass density of the Si iv/Si iii absorbers in terms of the current critical density, ρc, can be estimated by(7)with μ = 1.3, mH = 1.673 × 10-27 kg, H0 = 69.7 km s-1 Mpc-1 (Hinshaw et al. 2013), and ρc = 3H02/ 8πG. The comoving path length ΔX available for the detection of Si iv/Si iii absorbers along each sightline is given by:

(8)The integration over all sightlines then delivers ΔXtot. For our bias-corrected Si iv/Si iii absorber sample with log N(Si iii)≥ 12.2 we have N(H ii)tot ≥ 5.4 × 1020 cm-2 and ΔtotX = 20.65. The cosmological mass density can then be written as Ωb(Si iii)≥ 4.6 × 10-4ZSi, where ZSi is the Si abundance in the gas relative to the solar value, log (Si/H) = −4.49 (Asplund et al. 2009). Thus, if the CGM traced by Si iii would have a solar Si abundance, it would contain roughly as much mass as the neutral ISM within galaxies at z = 0 (Ωb(H i)= 4.7 ± 0.7 × 10-4; Zwaan et al. 2005).

On the one hand, Ωb(Si iii) could be much higher, if the mean metallicity of the gas was substantally lower than solar. This appears plausible if most of the gas originates in the IGM, e.g., as material stemming from accretion flows. Lehner et al. (2013) studied the metallicity distribution of LLS at z ≤ 1 and found two distinct populations of absorbers with mean metallicities of 0.03 solar (population I) and 0.50 solar (population II). They suggest that population I absorbers represent metal-poor circumgalactic gas from accretion streams, while population II systems trace metal-enriched halo material from galaxy outflows and winds. In view of these results, we assume a value of ZSi = 0.5 as a conservative upper limit for the metallicity of the Si iii absorbers in our survey. This value leads to Ωb(Si iii)≥ 9.2 × 10-4, roughly twice the value of the neutral ISM within galaxies.

Earlier theoretical and observational studies that addressed the outflow of gaseous material from galaxies and the enrichment of the IGM suggested, on the other hand, that the metals produced within galaxies escape in the form of metal-rich (super-solar, eventually) gas pockets (Mac Low & Ferrara 1999; Rauch et al. 2001; Schaye et al. 2007). Such gas pockets could mimic the absorption properties of intervening metal absorbers, but they would carry only very little mass. While it cannot be excluded that some of the Si iii absorbers presented in this study belong to such metal patches with above-solar metallicities, the large overall cross section of circumgalactic Si iii together with the expected cosmological metal-mass density at z = 0 (e.g., Calura & Mateucci 2004) excludes that a dominating fraction of the circumgalactic Si iii absorbers have metallicites above solar.

From the above considersations we conclude that the Si iv/Si iii bearing gas phase in the CGM of low redshift galaxies contains (possibly substantially) more baryonic matter than their neutral ISM. This conclusion is supported by previous estimates of the baryon content of the cool, photoionized CGM in the local Universe from the COS-Halos survey (Werk et al. 2014).

8. Discussion

8.1. On the origin of intervening Si III absorbers

Our study indicates a tight spatial correlation between intervening Si iii absorbers at z ≤ 0.1 and the local galaxy population at impact parameters 400 kpc, suggesting that the majority of these absorption systems arise in the extended gaseous halos of these galaxies. This interpretation is supported by recent HST/STIS and HST/COS observations of Si iii absorption in the CGM of the Milky Way (Shull et al. 2009; Collins et al. 2009; Richter et al. 2009; Lehner et al. 2012; Herenz et al. 2013), who show that doubly-ionized Si has the largest absorption cross section of all low, intermediate and high ions with a sky covering fraction of fc ⟩ ≈ 0.7. While the distances and the space distribution of Si iii absorbers in the Milky Way halo still are unclear (owing to our internal vantage point), Lehner et al. (2015) demonstrated that M31 also exhibits an extended, metal-enriched gaseous halo that gives rise to Si iii absorption out to ~200 kpc. Finally, other groups that have studied the absorption properties of the CGM for individual galaxies (e.g., Keeney et al. 2013), or pre-selected galaxy samples such as the COS-Halos survey (e.g., Tumlinson et al. 2011, 2013; Werk et al. 2013; Peeples et al. 2013) demonstrate that Si iii is ubiquitous in the extended CGM of low-redshift galaxies. The existence of such discrete Si iii bearing gas complexes with temperatures T< 105 K (thus below the virial temperature of their DM host halos) in the extended circumgalactic environment of galaxies can be understood in terms of the steady circulation of hot and cold gas through the CGM (“circumgalactic fountain”) as part of the ongoing galaxy evolution. In this scenario, star formation in galaxies drives out large amounts of (hot) metal-enriched gas. Even if this material leaves the potential well of an individual galaxy, it is trapped by the superordinate cosmological structure. From there, the gas may (slowly) fall back onto the galaxy of its origin in the form of discrete gas structures, as the gas never reaches hydrostatic equilibrium during such a circulation cycle (Ford et al. 2014). Alternatively, the gas may be accreted by a different galaxy nearby, or heated up to the virial temperature of the superordinate DM halo, in which case it may remain extragalactic forever.

While the observed frequency of Si iii absorbers with log N(Si iii)≥ 12.2 and the derived impact parameter distribution are in line with a scenario, in which all Si iii absorbers are located within the virial radius of intervening galaxies and thus are gravitationally bound to them, it cannot be excluded that there exists a population of intervening Si iii absorbers that traces gas at larger distances. This is actually expected, since in many cases large, gas-rich spiral galaxies like the Milky Way do not represent isolated systems, but are part of a galaxy group, which binds metal-enriched diffuse gas within its own virial radius in the form of an intragroup medium.

Stocke et al. (2014) recently have studied a class of “warm” H i/O vi absorbers using COS spectra. They suggested that these systems trace an extended (Mpc scale) T = 105−106 K intragroup medium in spiral-rich galaxy groups. If such warm gas was typical for group environments, it appears plausible that one would find cooler (T< 105 K) gas patches embedded in such a medium that could give rise to intervening Si iii/Si iv absorption. Some of the galaxy/absorber pairs that have apparent impact parameters in the range ρ = 200−1000 kpc may belong to such group absorbers. The Cloudy modeling predicts that Si iii/Si iv absorbers trace gas down to thermal pressures of P/k = nHT ~ 1 cm-3 K. For gas that is gravitationally bound to individual galaxies, such low gas pressures would be expected only in the outermost regions of galaxies near their virial radius (see also discussion in Shull 2014). However, such a value for P/k also would be in agreement with the expected range of gas pressures in galaxy groups (Stocke et al. 2013).

Our survey suggests that the number density distribution of Si iii absorbers breaks down for column densities log N(Si iii)≤ 12.2 (Fig. 4, right panel), which is not a completeness effect in our data (see Sect. 4.1). These low-column density systems may represent the prime candidates for metal-enriched cloudlets that arise in regions with low gas densities (and pressures), such as in group environments and in the IGM. As pointed out by Stocke et al. (2013), the expected number density of galaxy groups is ten times less than the space density of L galaxies (Berlind et al. 2006). Interestingly, the number density of Si iii absorbers with log N(Si iii)≤ 12.2 is also ~10 times less than the number density of absorbers with log N(Si iii)≥ 12.2, supporting a scenario in which high-column density Si iii systems trace the CGM of galaxies, while low-column density Si iii systems arise in the intragroup gas of galaxy groups.

thumbnail Fig. 16

Left panel: distribution of galaxies and H i in the Dorado galaxy group (see Kilborn et al. 2005, and references therein). Intervening Si iii absorption is observed at z = 0.0037 (cz = 1109 km s-1) towards the QSO 1H4019577 (black filled circle). The absorption most likely is associated with the CGM of one group member or with the intergroup medium. Right panel: zoom-in for the region around the two group members F157081 and NGC 1574; the absorber could be associated with either or both of these galaxies given its velocity and projected distance to each of the systems.

Open with DEXTER

With the occurrence of several galaxies at impact parameters ρ< 1000 kpc to the QSO sightlines and with radial velocities close to the observed absorber velocities (such as expected for group environments), the interpretation of the observed absorption with respect to origin and nature of the gas in its galaxy environment generally is tricky. While a more systematic investigation between intervening Si iii absorbers and galaxy group environments clearly is beyond the scope of this paper, we show in Fig. 16 as a prominent example the galaxy group environment of the z = 0.0037 Si iii absorber towards the QSO 1H0419577 (IRAS F042505718). This sightline passes two galaxies in the Dorado galaxy group (Kilborn et al. 2005; Maia et al. 1989) at impact parameters ρ< 220 kpc. In principle, the observed Si iii absorption at cz = 1109 km s-1 could be associated with either of these two galaxies, which have radial velocities of vr = 1050 and 1215 km s-1 (Kilborn et al. 2005). The observed gas may be infalling or outflowing gaseous material; the metallicity of the absorber lies in the range between 0.1−1.0 solar (as derived from the fit of the H i Ly α absorption together with the Cloudy model of the Si iii/Si iv absorption), thus in line with both scenarios. However, the absorber could also be located outside the virial radii of these two galaxies and may belong to a faint (unseen) dwarf galaxy in the same group or may represent a metal-rich gas patch embedded in the intragroup medium of the Dorado group. The interpretation of individual absorber/galaxy pairs thus remains inconclusive with magnitude-limited galaxy data. The statistical connection between galaxies and intervening Si iii systems, as studied here, yet demonstrates that both class of objects trace the same Mpc-scale environment within the cosmic web.

8.2. Comparison with previous Si III absorption-line studies

We are not aware of any other systematic studies in the literature that focus explicitly on Si iii-selected intervening absorption systems and their relation to galaxies. There are, however, several absorption-line surveys at low redshift (using various UV spectrographs) that provide information on the number density of Si iii and other intermediate ions (e.g., C iii) and the distribution of equivalent widths/column densities of these ions.

Tilton et al. (2012) compiled UV absorption-line data from HST/STIS and FUSE and prepared a catalog of UV absorbers for redshifts z< 0.4. From their data sample they derive a number density of intervening Si iii absorbers of for log N(Si iii)> 12.2 for this redshift range, based on a total redshift path of Δz ≈ 5.2. This value is more than twice the value derived in this study. The same authors recently published another absorption-line catalog of intervening absorption systems at z< 0.4, this time based on HST/COS data of 82 AGN (Danforth et al. 2016). In this new survey, the authors derive a number density of for log N(Si iii)> 12.2. Moreover, their study suggests an increase for (Si iii) for decreasing redshift in the local Universe (their Fig. 15). Their new value again is substantially higher than the value derived by us. To investigate the origin for this discrepancy we have carefully compared their absorber list with ours, as our COS data sample includes all of the 82 QSOs presented in Danforth et al. (2016). We suspect that the higher value for (Si iii) derived by Danforth et al. (2016) stems from the less stringent selection criteria for identifying metal absorbers and defining the relevant absorption path lengths along their sightlines. Also, a possibly existing selection bias in their QSO sample may be the reason for the higher value of (Si iii).

We identify three Si iii systems in their absorber list, whose identifications are based on absorption features that (in our opinion) most likely have different origins. We also have identified one candidate Si iii absorber that is not listed in the Danforth et al. (2016) paper. In Table A.4 in the Appendix we list these discrepant absorption systems together with a short description of the absorption characteristics.

As we discuss in one the following sections, state-of-the art hydrodynamical simulations of metal-enriched gas in the local Universe do not support values of for Si iii absorbers for log N(Si iii)> 12.2, but favour Si iii number densities of <4 for z ≈ 0.

8.3. Comparison with other CGM studies

As mentioned above, Si iii is frequently detected in the CGM of individual galaxies (Keeney et al. 2013; Lehner et al. 2015) and in the CGM of pre-selected galaxy samples (e.g., COS-Halos; Tumlinson et al. 2013; Werk et al. 2013; Liang & Chen 2014). Our study, which represents a statistical rather than a targeted study of the CGM in the local Universe, complements most of the previous observational CGM studies. Many of the COS sightlines that have been used by previous, targeted CGM studies are included in our sample. Because of the statistical nature of our survey, we refrain from comparing our results for individual galaxy/absorber pairs with results from other studies.

From an archival study of CGM absorption around 195 galaxies at z< 0.176 Liang & Chen (2014) find a mean Si iii covering fraction of within 0.54 Rhalo and for the range 0.54−1.02 Rhalo, but no significant Si iii absorption beyond. These covering fractions are lower than our estimate of fc ⟩ = 0.69 ± 0.11. This is not surprising, however, since their study considers only strong Si iii systems with Wr> 100 mÅ  (log N(Si iii)> 12.6), while our estimate is based on a more sensitive search including weak and strong absorbers with log N(Si iii)> 12.2. If we adopt our measured value of (Si iii)= 1.9 for N(Si iii)≥ 12.6 for our estimate of fc using Eq. (4), we obtain a value of within Rvir, thus in good agreement with the Liang & Chen (2014) estimate. Also Werk et al. (2013) consider only strong Si iii absorbers around LL absorbers with Wr> 100 mÅ, and find for their total galaxy with little dependence on the galaxy mass or star-formation rate (their Table 6). Within the given error bars, these estimates for fc (based on different methods) agree very well with each other.

In contrast, Bordoloi et al. (2014) studied the CGM around dwarf galaxies at z ≤ 0.1 and found no metal absorption beyond 0.5 Rvir. Their study suggests that the filling factor of warm circumgalactic gas is substantially smaller for dwarfs than for L galaxies. As described in Sect. 4.2, our full data set includes most of the sightlines from the Bordoloi et al. sample, so that we can analyze the Si iii absorption fraction in the CGM of these dwarf galaxies directly. We find that only 17 out of the according 40 sightlines show detectable Si iii with log N(Si iii)> 12.2 at the redshifts of the dwarf galaxies (), demonstrating that the absorption fraction of the warm CGM traced by Si iii around dwarf galaxies is only ~60 percent of that seen in galaxies with higher luminosities. The small CGM absorption fraction around dwarf galaxies is the reason for the almost identical number densities of intervening Si iii in our original data sample and in the bias-corrected sample (Sect. 4.2).

Turning to high-ionization species, Wakker & Savage (2009) have studied the relation between O vi and nearby galaxies at z< 0.017 based on UV and FUV data along 76 extragalactic sightlines, demonstrating that a substantial fraction of the O vi systems arise from gas within Rvir of these galaxies (see also Prochaska et al. 2011; Tumlinson et al. 2011; Stocke et al. 2013). However, in contrast to Si iii, O vi is also present beyond the virial radius of these galaxies, tracing hot metal-enriched gas (at higher gas temperatures and lower densities than Si iii) that presumably has been ejected by galactic winds and outflows in past epochs of active star formation. In fact, in view of our CGM modeling results presented in Sect. 5, the measured O vi number density at low redshifts of (O vi)≈ 16 (Tripp et al. 2008) indicates a very large absorption cross section of O vi that is clearly beyond the expected cross section of the CGM in the local Universe, even if non-standard Schechter parameters in the luminosity function are considered (Table 1, fourth row). The Milky Way possibly also contains a massive, extended O vi envelope that reaches deep into the Local Group potential well (Sembach et al. 2003; Wakker et al. 2003).

Interestingly, the detailed analysis of the COS-Halos sample suggests that the absorption properties of low and intermediate ions (such as Si iii) in star-forming galaxies do not significantly differ from those in passive galaxies, while O vi predominantly arises in the halo of active galaxies (Thom et al. 2012; Werk et al. 2013; Tumlinson et al. 2011). From their study, Liang & Chen (2014) conclude that the CGM becomes progressively more ionized from small to large radii. A similar conclusion was drawn by us from modeling the radial decline of the neutral gas fraction in the CGM in the Local Group (Richter 2012).

8.4. Comparison with hydrodynamical simulations

Cosmological, hydrodynamical simulations have recently become a powerful tool for studying intergalactic and circumgalactic gas at low and high redshifts (e.g., Oppenheimer & Davé 2006; Crain et al. 2013; Nuza et al. 2014). The line statistics from synthetic absorption spectra generated from such simulations can be directly compared to QSO absorption-line studies to investigate the different gas phases in the IGM and CGM and their redshift evolution (Fang & Bryan 2001; Richter et al. 2006; Fangano et al. 2007; Tepper-García et al. 2011, 2012, 2013; Oppenheimer et al. 2012; Smith et al. 2011; Churchill et al. 2015).

While most of these studies have focused on the properties of high-ion absorbers, such as O vi and Ne viii, to study the shock-heated warm-hot intergalactic medium or shock-heated gas in the CGM, intermediate ions such as Si iii and C iii so far have not been considered for a detailed comparison study of absorption-line statistics in simulations and observational data. From the simulations of Smith et al. (2011), Danforth et al. (2016) extracted the cumulative line density of intervening metal absorbers for a comparison with their COS absorption-line survey. While their own value for (Si iii) of ~6 for log N(Si iii)> 12.2 (see Sect. 8.2) lies far above the value predicted by Smith et al. for this column-density limit ((Si iii)≈ 3), the Si iii number density derived by us is in very good agreement with the results by Smith et al.

We here re-use synthetic spectra generated along random LOS from a simulation run (Model AGN) of the OverWhelmingly Large Simulations (OWLS; Schaye et al. 2010) to further investigate the frequency and physical conditions in Si iii absorbers at low redshift (z = 0.25). Similar simulations have been used by us previously to study absorption signatures of warm-hot gas in the local Universe (Tepper-García et al. 2011, 2012, 2013). The OWLS simulations were set up using the smoothed particle hydrodynamics (SPH) code GADGET III (a modified version of GADGET II; Springel 2005) and were carried in a cubic box of 100 h-1 comoving Mpc on a side, containing 5123 dark matter (DM) and 5123 baryonic particles. The initial mass resolution is 4.1 × 108h-1M (DM) and 8.7 × 108h-1M (baryonic), while the gravitational softening is set to 8 h-1 comoving kpc and is fixed at 2 h-1 proper kpc below z = 3. The cosmological simulations from the OWLS project are characterized by a particularly careful implementation of important physical processes that have been largely ignored in earlier studies (e.g., the influence of the photoionization on the cooling function of the gas). Details on the sub-grid physics used in OWLS can be found in Wiersma et al. (2009a; radiative cooling), Schaye & Dalla Vecchia (2008; star formation), Dalla Vecchia & Schaye (2008; stellar feedback), and Booth & Schaye (2009; AGN feedback). The strategy, how the synthetic absorption-line spectra from OWLS were set up and analyzed is presented in Tepper-García et al. (2011, 2012, 2013). In Fig. 17, upper panel, we show the cumulative number density of intervening Si iii absorbers as a function of the limiting equivalent width predicted by OWLS (Model AGN). The redshift in the simulation run is higher than the maximum redshift of our data. This is, however, not an issue since we do not expect the gas producing the observed absorption to evolve strongly from z = 0.25 to z = 0.10. The measured value from this study and from Tilton et al. (2012; T12) and Danforth et al. (2016; D16) are indicated with filled circles. The predicted number density is ~1.7 for log N(Si iii)> 12.2 (log [Wr(λ1206)/mÅ] ≥ 1.52), thus only slightly below the observed value of 2.5 ± 0.4 from our COS data set.

thumbnail Fig. 17

Upper panel: cumulative number density of intervening Si iii absorbers for different equivalent-width limits predicted by a simulation run from the OverWhelmingly Large Simulations (OWLS; Schaye et al. 2010), as presented in Tepper-García et al. (2012). The measured value from this study and from Tilton et al. (2012; T12) and Danforth et al. (2016; D16) are indicated with filled circles. Lower panel: phase diagram of intervening Si iii absorbers in the temperature-density plane, as predicted by OWLS.

Open with DEXTER

It is worth noting that differences between the predicted and observed line-number densities may be alleviated considering that the stellar yields used in the simulations to model the chemical evolution of the gas are uncertain by factors of a few (see Wiersma et al. 2009b). Thus, some modification of the abundances in post-processing would be justified, which would bring the predicted into better agreement with our measured value. However, we refrain from doing so because this could break the self-consistency of the simulation if it were to change the cooling rates significantly.

In the lower panel of Fig. 17 we show the phase-space distribution in the temperature-density plane of the Si iii-absorbing gas in the AGN run at z = 0.25 from OWLS. The typical Si iii absorber in the simulation has a temperature of T ≈ 15 000−20 000 K and a density of nH ≈ 10-4−10-2 cm-3, in very good agreement with the values estimated with Cloudy from the Si ii/Si iii/Si iv column-density ratios in our COS absorber sample. The resulting range for the expected thermal gas pressure is P/k ≈ 2−160 cm-3 K, thus in a range that is typical for the CGM (Cen 2013).

In conclusion, the AGN model from OWLS and the simulations from Smith et al. (2011) predict absorber properties that are compatible with our observations and in line with the idea, that the majority of the Si iii absorbers at z ≤ 0.1 arise in the extended halos of low-redshift galaxies. For the future, we are planning to further explore the relation between intermediate-ion absorbers and galaxies at low z using higher-resolution simulations from the EAGLE (Evolution and Assembly of GaLaxies and their Environments) project (Schaye et al. 2015).

9. Summary and conclusions

In this paper, we present a systematic study of intervening Si iii absorbers and their relation to galaxies in the redshift range z ≤ 0.1 along 303 QSO sightlines using archival UV spectra obtained with HST/COS. The main results are summarized as follows:

  • (1)

    We detect 69 intervening Si iii absorbers in the range z = 0.00014−0.09789 along a total redshift path of Δz ≈ 24. The restframe equivalent widths in the Si iii λ1206 line lie in the range Wr = 13−885 mÅ, corresponding to Si iii column densities log N(Si iii)> 11.9. We derive a bias-corrected number density of intervening Si iii systems of for absorbers with log N(Si iii)≥ 12.2, suggesting that Si iii-selected absorbers outnumber strong Mg ii at z ≈ 0 by a factor of ~3. Weak absorbers in the range log N(Si iii)= 11.8−12.2 turn out to be rare (); they possibly represent an absorber population that is distinct from the one with larger Si iii column densities.

  • (2)

    The observed Si iii absorption is typically accompanied by absorption from other low and/or high ions. The positive correlations between the equivalent widths of Si iii and Si iv and between Si iii and Si ii indicates that doubly ionized silicon traces gas in a range of physical conditions. This conclusion is supported by Cloudy ionization models of the absorbers, which indicate that Si iii arises either in diffuse, ionized low-density (log nH ≤ −3, typically) gas that is also observed in Si iv absorption, or in somewhat denser (log nH ≥ −3, typically), more neutral environments that are also traced by Si ii.

  • (3)

    We geometrically model the absorption cross section of diffuse gas within the virial radii (Rvir) of galaxies at z = 0 for galaxies in the luminosity range L/L> 0.001 using the SDSS galaxy luminosity function for z = 0 and scaling relations between Rvir and L. Our modeling implies that circumgalactic gas within the virial radius of galaxies contributes with to the total number density of intervening absorbers, even if a unity covering fraction of the absorbing gas is assumed. If we adopt the SDSS g-band luminosity function as reference (Montero-Dortas et al. 2009), we obtain as expected number density for CGM absorbers from L ≥ 0.001 L galaxies. In light of this result, the measured value of (Si iii)= 2.5 ± 0.4 is in line with the idea, that the majority of the intervening Si iii absorbers with log N(Si iii)≥ 12.2 at z ≈ 0 trace the circumgalactic medium at R<Rvir of nearby galaxies with an average covering fraction of fc ⟩ = 2.5/3.6 = 0.69, a value that is in excellent agreement with results from other studies (Liang & Chen 2014; Werk et al. 2013). Our model further predicts that intervening metal absorbers with number densities (e.g., O vi absorbers) must partially arise from gas outside of the virial radius of galaxies.

  • (4)

    We compare the Si iii absorber redshifts and positions with that of ~64 000 galaxies at z ≤ 0.1 using archival galaxy-survey data. For ~60 percent of the absorbers we find possible host galaxies within 300 km s-1 of the absorbers and at impact parameters ρ< 200 kpc, indicating that the spatial distributions of Si iii absorbers and galaxies are highly correlated. We verify the significance of this result by studying the distribution of galaxies around the same sightlines in velocity ranges without Si iii absorption and find no clustering of the galaxies, as expected. The observed Si iii absorption fraction around the galaxies scales differently with impact parameter than predicted by our model. We assign this behavior to the incompleteness in our galaxy sample to identify faint galaxies that are associated with Si iii absorbers.

  • (5)

    We estimate the baryon content of the Si iii-absorbing gas phase in the CGM at low z using Cloudy ionization models. Assuming that the metallicity of the gas is 0.5 solar, the cosmological mass density of the gas in Si iii absorbers is Ωb(Si iii)≥ 9.2 × 10-4. This lower limit is roughly twice the value derived for the H i mass density at z = 0, indicating that the diffuse CGM around galaxies, as traced by doubly-ionized Si, contains substantially more baryonic matter than their neutral ISM.

  • (6)

    We discuss the origin of intervening Si iii absorption in the CGM of galaxies and in galaxy groups and compare our results with previous absorption-line catalogs and CGM studies at low redshift. We suggest that the majority of the strong Si iii absorbers with log N(Si iii)≥ 12.2 typically arise within the extended halos of galaxies within Rvir, while weak systems with log N(Si iii)< 12.2 possibly also trace gas at larger distances (e.g., group environments or the IGM). The comparison of our measurements with predictions from cosmological hydrodynamical simulations from the OWLS project and other simulations shows a fair agreement for the absorber number density and the estimated temperature-density range of intervening Si iii absorbers.


2

These values are consistent at the 2σ level with the latest measured values by the Planck Collaboration (Planck Collaboration XIII 2015).

Acknowledgments

This research used the facilities of the Canadian Astronomy Data Centre operated by the National Research Council of Canada with the support of the Canadian Space Agency. This research also has made use of the SIMBAD database, operated at CDS, Strasbourg, France. We thank Joop Schaye and an anonymous referee for helpful comments and suggestions.

References

Appendix A: Supplementary tables and figures

Table A.1

COS QSO sample.

Table A.2

Absorption-line systems.

Table A.3

Si iii candidate systems.

Table A.4

Comparison with (revised) Danforth et al. (2016) absorber list.

thumbnail Fig. A.1

Velocity profiles.

Open with DEXTER

All Tables

Table 1

Models for absorber cross sections based on Eq. (2).

Table 2

Model predictions for effective CGM covering fractions, ⟨ ℱc(rmax) ⟩, for different halo radii and galaxy luminosities (see Sect. 5).

Table A.1

COS QSO sample.

Table A.2

Absorption-line systems.

Table A.3

Si iii candidate systems.

Table A.4

Comparison with (revised) Danforth et al. (2016) absorber list.

All Figures

thumbnail Fig. 1

Ionization fractions of different Si ionization states for different gas densities and temperatures, as calculated from a Cloudy ionization model (including photoionization and collisional ionization) of circumgalactic gas at z = 0 using the local UV background. The upper panel shows a model for log T = 4.2, the lower panel a model for log T = 4.4. The Si iii ion represents the dominant ionization state in the density range that is characteristic for circumgalactic gas structures that are embedded in hot coronal gas (log nH = −2.0 to −3.5).

Open with DEXTER
In the text
thumbnail Fig. 2

Five examples for velocity profiles of Si iii-selected absorption line systems in our HST/COS survey with decreasing Si iii λ1206 absorption strengths (gray-shaded area; from left to right). Shown are the velocity profiles of H i, Si iii, Si ii, Si iv, C ii, and C iv (from top to bottom) plotted against the absorber’s restframe velocity. The names of the background QSOs and the absorption redshifts are indicated above each panel. A complete set of velocity plots for all detected Si iii absorbers discussed in this paper is available in the Appendix.

Open with DEXTER
In the text
thumbnail Fig. 3

Distribution of B-band magnitudes for ~41 000 galaxies in our sample.

Open with DEXTER
In the text
thumbnail Fig. 4

Left panel: redshift distribution of the 69 intervening Si iii absorbers in our sample. Middle panel: distribution of Si iii λ1206 restframe equivalent widths of the 69 absorbers, obtained from the pixel integration over the Si iii velocity profiles. Right panel: distribution of logarithmic Si iii column densities for the 69 absorbers, derived using the AOD method. The gray-shaded area indicates lower limits for log N(Si iii) for absorbers that have saturated Si iii λ1206 lines.

Open with DEXTER
In the text
thumbnail Fig. 5

Number densities, , of intervening Si iii absorbers for different limiting column densities.

Open with DEXTER
In the text
thumbnail Fig. 6

Distribution of restframe equivalent widths of H i Ly α (left panel), Si ii λ1260 (middle panel), and Si iv λ1393 (right panel) associated with intervening Si iii absorbers.

Open with DEXTER
In the text
thumbnail Fig. 7

Comparison between restframe equivalent widths of Si iii, Si ii, Si iv, and H i for 69 absorbers. First panel (left): Si iii λ1206 vs. H i Ly α; second panel: Si iii λ1206 vs. Si ii λ1260; third panel: Si iii λ1206 vs. Si iv λ1393; fourth panel: Si ii λ1260 vs. Si iv λ1393.

Open with DEXTER
In the text
thumbnail Fig. 8

Predictions from the halo model described in Sect. 5. Upper panel: the virial radius of galaxies as a function of galaxy luminosity (adapted from Stocke et al. 2014). Middle panel: Galaxy space density as a function of galaxy luminosity (from Montero-Dorta & Prada 2009). Lower panel: expected number density of CGM absorbers as a function of galaxy luminosity.

Open with DEXTER
In the text
thumbnail Fig. 9

CGM absorption fraction as a function of impact parameter, as predicted from the halo model described in Sect. 5.

Open with DEXTER
In the text
thumbnail Fig. 10

Upper panels: distributions of impact parameters and absolute B-band magnitudes for galaxies that are located within | Δv | = 1000 km s-1 of the absorbers (galaxy/absorber samples). On the left-hand side we show the distribtions for all sightlines that exhibit Si iii absorption, while on the right-hand side only the bias-corrected set of sightlines is considered (see Sect. 4.2). Galaxies tend to cluster around the LOS in velocity regions where intervening absorption is observed. Lower panels: the same distributions for galaxies that have no absorption systems within | Δv | = 1000 km s-1 (composite control samples; see Sect. 6.2 for details). No clustering is observed in this case.

Open with DEXTER
In the text
thumbnail Fig. 11

Restframe equivalent widths of Si iii λ1206 are plotted against the impact parameters of the galaxies nearest to the absorbers. For increasing impact parameters the maximum value for Wr is decreasing from Wr ≈ 900 mÅ down to values <100 mÅ.

Open with DEXTER
In the text
thumbnail Fig. 12

Absorption fraction (fractional abundance of intervening absorbers per galaxy) as a function of impact parameter, as measured in our galaxy/absorber sample.

Open with DEXTER
In the text
thumbnail Fig. 13

Predictions of H ii, Si ii, Si iii, Si iv, and Mg ii column densities as a function of the gas density (log nH) for an H i column density of log N(H i)= 15, based on a Cloudy ionization model using the local UV background and assuming absolute and relative solar abundances of all elements. The gray-shaded area indicates the density that is typical for the Si iii/Si iv phase (see Sect. 7.3).

Open with DEXTER
In the text
thumbnail Fig. 14

Column-density ratios of Si iii/Si ii vs.  Si iii/Si iv are plotted together with predictions from the single-phase Cloudy model (Eq. (5), indicated by the black solid line). Measured values are plotted with filled circles, limits are indicated with open circles. For none of the systems do the data points lie on the expected relation for the single-phase model, implying that the gas is multiphase.

Open with DEXTER
In the text
thumbnail Fig. 15

Distribution of logarithmic column densities in the Si iii absorbers. Upper panel: Si iv vs. Si iii (absolute values indicated with filled circles, limits with open circles). Lower panel: Si ii vs. Si iii. The dotted/solid lines indicate ratios of + 0.5,0,−0.5 dex, respectively.

Open with DEXTER
In the text
thumbnail Fig. 16

Left panel: distribution of galaxies and H i in the Dorado galaxy group (see Kilborn et al. 2005, and references therein). Intervening Si iii absorption is observed at z = 0.0037 (cz = 1109 km s-1) towards the QSO 1H4019577 (black filled circle). The absorption most likely is associated with the CGM of one group member or with the intergroup medium. Right panel: zoom-in for the region around the two group members F157081 and NGC 1574; the absorber could be associated with either or both of these galaxies given its velocity and projected distance to each of the systems.

Open with DEXTER
In the text
thumbnail Fig. 17

Upper panel: cumulative number density of intervening Si iii absorbers for different equivalent-width limits predicted by a simulation run from the OverWhelmingly Large Simulations (OWLS; Schaye et al. 2010), as presented in Tepper-García et al. (2012). The measured value from this study and from Tilton et al. (2012; T12) and Danforth et al. (2016; D16) are indicated with filled circles. Lower panel: phase diagram of intervening Si iii absorbers in the temperature-density plane, as predicted by OWLS.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.