Quasar Sightline and Galaxy Evolution (QSAGE) survey – I. The galaxy environment of O vi absorbers up to z = 1.4 around PKS 0232−04

ABSTRACT

1 INTRODUCTION

The peak epoch in the volume-averaged star formation rate (SFR) for galaxies is at z = 1–2 with the SFR in typical galaxies being an order of magnitude higher than in the local Universe (e.g. Lilly et al. 1996; Madau et al. 1996; Sobral et al. 2013). A picture is emerging in which star formation at this epoch is very different to that at the present day. Rather than the subdued formation of stars that is the norm in today’s Universe, violent episodes of star formation are driven by the formation of superstar clusters within unstable gas-rich discs (e.g. Swinbank et al. 2010; Tacconi et al. 2010).

Theory suggests that the peak in the star formation rate density is the result of a higher rate of gas accretion in the early Universe (e.g. Dekel et al. 2009; Lagos et al. 2011; van de Voort et al. 2011). Star formation at high redshift should be critically dependent on the inflow rate of cold gas, whilst the present-day galaxy stellar mass function is most readily explained through an efficient feedback mechanism such as strong supernovae driven outflows. In order to fully understand the interplay and balance of these factors (e.g. Bouché et al. 2010; Davé, Finlator & Oppenheimer 2012), it is crucial that we test these theories by detecting diffuse baryons inside and around the haloes of galaxies, charting their abundance and physical properties as a function of redshift (e.g. Fumagalli et al. 2011; Davé et al. 2012; Hummels et al. 2013; Rahmati et al. 2013, 2015; Correa et al. 2018; Oppenheimer et al. 2018b). The approximate area of influence, within which a galaxy deposits and receives baryonic material, is commonly referred to as the circumgalactic medium (CGM).

Extensive studies have been performed using the powerful combination of galaxy surveys with high-resolution quasar sightline data, to analyse the nature of gas around galaxies at z ≲ 1 (e.g. Bergeron & Boissé 1991; Morris et al. 1991; Morris et al. 1993; Steidel, Dickinson & Persson 1994; Chen et al. 1998, 2001; Steidel et al. 2002; Chen & Mulchaey 2009; Crighton et al. 2010; Prochaska et al. 2011; Tumlinson et al. 2011; Bouché et al. 2012; Kacprzak, Churchill & Nielsen 2012; Nielsen, Churchill & Kacprzak 2013; Tejos et al. 2014; Finn et al. 2016; Prochaska et al. 2017) and at 2 ≲ z ≲ 4 (primarily using the Lyman Break method; e.g. Adelberger et al. 2003, 2005; Simcoe et al. 2006; Chen et al. 2010; Crighton et al. 2011; Rakic et al. 2012; Prochaska et al. 2013; Prochaska, Lau & Hennawi 2014; Tummuangpak et al. 2014; Turner et al. 2015; Fumagalli et al. 2016, 2017b; Turner et al. 2017; Bielby et al. 2017b). These have provided important insights for understanding the flow of baryons around galaxies and how this impacts galaxy evolution. However, the available data between these two epochs (i.e. z ≈ 1–2) is more limited, and primarily focused around either Ly α absorption or relatively low-ion absorbers such as Mg ii and C iv (e.g. Chen et al. 1998; Bowen et al. 2006; Morris & Jannuzi 2006; Rubin et al. 2010; Lovegrove & Simcoe 2011; Farina et al. 2014; Johnson, Chen & Mulchaey 2015b). There are an increasing number of studies focusing on the relevance of large-scale structure (i.e. voids, filaments, groups, and clusters) with respect to observations of the CGM (e.g. Penton, Stocke & Shull 2002; Aracil et al. 2006; Tejos et al. 2012, 2016; Stocke et al. 2014; Burchett et al. 2016, 2018; Fumagalli et al. 2016, 2017b; Bielby et al. 2017a; Péroux et al. 2017; Fynbo et al. 2018; Pessa et al. 2018). Such studies are adding important context to for interpreting CGM observations, but are again primarily limited to z ≲ 0.5 or z > 2.

The O vi doublet absorption feature offers important insights into galaxy evolution (Tripp, Savage & Jenkins 2000). Low-redshift studies have shown that O vi absorbers often trace 0.1–1 L* star-forming galaxy haloes (Chen & Mulchaey 2009; Wakker & Savage 2009). The nature of the material traced by O vi absorption in quasar sightlines potentially traces a broad range of physical manifestations: as outflowing material, indicative of ongoing or recent star formation (e.g. Tumlinson et al. 2011; Meiring et al. 2013; Werk et al. 2016); diffuse halo gas at the ambient halo temperature (e.g. Oppenheimer et al. 2016; Werk et al. 2016; Nielsen et al. 2017; Oppenheimer et al. 2018a; Kacprzak et al. 2019); intra-group material (e.g. Stocke et al. 2014, 2017; Pointon et al. 2017); or more widely spread warm-hot intergalactic medium (e.g. Shull, Smith & Danforth 2012).

In this paper, we present results from the first field analysed as part of the Quasar Sightline and Galaxy Evolution (QSAGE) survey. We have performed a grism spectroscopic survey around 12 bright quasars at z > 1.2, all with existing archival HST/STIS, and in some cases HST/COS, spectroscopy. By selecting quasar sightlines as target fields, we place the scientific focus on studying the CGM, whilst providing 12 effectively independent pencil beam surveys suitable for a wide range of galaxy evolution science. Whilst the HST/STIS and HST/COS data probe gas in absorption along the central sightline at z ≲ 1.44, our new HST WFC3 grism observations survey the galaxy population ‘blindly’ (i.e. without any preselection aside from flux limits) around these sightlines across a comparable redshift range (predominantly via H α emission). In an earlier paper (Bielby et al. 2017a), we presented an analysis of Mg ii absorption associated with a group environment detected in MUSE observations on one of the fields, prior to having acquired WFC3/grism data. Here we focus on a second field, this time incorporating both MUSE IFU and WFC3 grism data, whilst also presenting the survey strategy, data-processing methods, and an analysis of the relationship between O vi absorption and galaxies at 0.68 < z < 1.4 data. The quasar that forms the basis for this study is PKS 0232−04 (Shimmins et al. 1966, originally designated PHL 1377; Haro & Luyten 1962) at a redshift of z = 1.44.

In Section 2 we present the observations and data reduction methods (including optical and NIR imaging, grism spectroscopy, IFU spectroscopy with MUSE, and high-resolution sightline spectroscopy); in Section 3, we present an analysis of the distribution of galaxies around O vi absorbers as a function of galaxy properties and environment; we discuss our results in Section 4; and in Section 5 we give our conclusions and a summary. Throughout this paper, we use the AB magnitude system and the Planck 2015 cosmology (Planck Collaboration XIII 2016). Unless otherwise stated, all distances are given in proper coordinates.

2 OBSERVATIONS

The QSAGE survey targets 12 z ≳ 1.2 quasars with the aim of mapping and characterizing the galaxy populations in their foregrounds (HST Large Programme 14594; PIs: R. Bielby, J. P. Stott). The full survey consists of 96 orbits with the WFC3 in imaging and grism mode (i.e. 8 h per quasar sightline). In this paper we focus on the first field to be fully analysed, centred on the z = 1.44 quasar PKS 0232−04, chosen due to the extensive data available to complement the WFC3 grism and STIS sightline data: Canada–France–Hawaii Telescope Legacy Survey (CFHTLS) optical imaging, MUSE IFU data and HST/COS sightline spectroscopy. In the following section we describe in detail the methods used in taking and processing the various observations on this first target field, and which going forward will be used across the entirety of the 12 survey fields.

We note that this sightline was previously observed with HST + FOS (Jannuzi et al. 1998), which showed that the sightline is free of LLS and therefore maximizes the path-length for FUV ions such as Ne viii and Mg x. However, this pre-selection also means that the survey imposes a strong bias against some critical components of the CGM at z > 0.4 (see discussion in Tumlinson et al. 2013 for example), namely LLS and stronger H i systems. Many LLS exhibit coincident O vi absorption and so this is relevant even for a focus on high-ionization state gas (though a small fraction of O vi absorbers globally, LLS are common at ≲150 kpc from >0.1 L* galaxies of all types).

2.1 HST WFC3 data

For a single field, the QSAGE HST observations consist of eight grism exposures alongside supporting NIR imaging. The imaging is required for the identification of accurate source coordinates, which act as the basis for the extraction of object spectra from the grism data. The observations for a given field were split into four ‘visits’, with each visit providing observations at a different position angle. Each visit consisted of two orbits, with each orbit scheduled to acquire a single F140W image of ≈250 s, two exposures of ≈1000 s using the G141 grism, and a single ≈250 s F160W exposure. These exposures were taken using the ‘SPARS25’ sampling sequence in the case of the basic imaging exposures and the ‘SPARS100’ sampling sequence in the case of the grism exposures. The corresponding exposure times were tweaked to optimally fill the time afforded by a given orbit and all exposures were taken with the ‘grism1024’ aperture. The total exposure times across the four visits for this field were 2123 s with F140W; 2123 s with F160W; and 16 047 s with the G141 grism. These data represent some of the deepest observations using the WFC3 G141 grism thus far. As such, the different position-angle visits are necessary for the optimal extraction of spectra with the removal of contamination from overlapping sources.

2.1.1 WFC3 NIR imaging

Both the direct imaging and grism data were reduced using grizli (Brammer, in preparartion¹), a custom software package dedicated to the reduction and analysis of slitless spectroscopic datasets which builds on previous software packages such as axe (Kümmel et al. 2009) and threedhst (Brammer et al. 2012; Momcheva et al. 2016). For the direct images, individual exposures were corrected for small astronomic offsets using tweakreg before astrodrizzle was used to produce background-subtracted drizzled images in both the F140W and F160W filters. The F140W and F160W sets of images were each stacked using swarp (Bertin et al. 2002) using a median combination. The resulting F140W stacked image is shown in Fig. 1. Photometric zero-points for the F140W and F160W imaging were taken directly from the STScI guidelines [m_ZP(F140W) = 26.45 and m_ZP(F160W) = 25.95].

Figure 1.

Stacked HST/WFC3 broad-band F140W image from the four separate orbits observing PKS 0232−04. Galaxies with a successfully identified redshift from either VLT/MUSE (blue ellipses) or HST/WFC3 grism observations (dark red ellipses) are marked. Objects detected in the imaging, but without a reliably identified redshift are marked by the pale orange ellipses. Dotted rings indicate scales of 100 kpc up to 600 kpc in steps of 100 kpc at z = 1. The dashed box denotes the field of view of the VLT/MUSE observations.

Open in new tabDownload slide

Extraction of the grism spectra requires a catalogue of sources with accurate positions from the associated imaging data. We produced this catalogue from the F140W stacked image using SExtractor (Bertin & Arnouts 1996). We used a detection threshold of 1.5σ, a minimum area for detection of five contiguous pixels, and we adjusted the deblending parameter to optimally extract any objects visually identified as being blended with the quasar point spread function (⁠|${\rm DEBLEND\_MINCONT}=0.005$|⁠).

The number counts of sources detected in the WFC3 F140W and F160W imaging are given in the final two panels of Fig. 2, whilst the corresponding 80 per cent completeness depths are given in Table 1. The completeness estimates were made by iteratively placing model sources (Gaussian profiles with the instrument full width at half-maximum – FWHM) directly into the imaging data, and re-running SExtractor with our given parameters on these modified images.

Figure 2.

Galaxy number counts estimated for all of the broad-band imaging used in this work, obtained from single image source extraction using SExtractor. Vertical lines show the estimated 80 per cent completeness limit for point sources. Where both CFHTLS and WHT/ACAM data are available, the dashed line denotes the CFHTLS limit and the dotted line shows the WHT/ACAM limit.

Open in new tabDownload slide

2.2 WFC3 G141 grism data

The catalogues and segmentation images generated with SExtractor were then used to extract the spectrum of each object using grizli. First, each G141 exposure was divided through by the F140W flat-field, neglecting any wavelength-dependent flat-field effects which are at most a few per cent (Momcheva et al. 2016). An initial background was subtracted using the ‘master sky’ images from Brammer, Ryan & Pirzkal (2015). These master images are necessary to account for the variation in background structure due to variations in zodiacal continuum, scattered light, and He emission across the sky (Momcheva et al. 2016). The remaining residuals (typically |$0.5\!-\!1 {{\ \rm per\ cent}}$| of the initial background levels) were then removed by subtracting average values of the sky pixels in each column. Finally 1D and 2D spectra were extracted from the background-subtracted grism images at the individual exposure level, which, for the qsage observations, resulted in 16 individual spectra per object.

An unavoidable feature of slitless spectroscopy is contamination from nearby sources. Despite the fact that this effect is substantially mitigated by the four independent grism orientations employed in the qsage survey, we nevertheless estimated the quantitative contamination for each individual spectrum. grizli initially generates models for each object in the catalogue assuming a linear continuum, based on their observed magnitudes. These models are then refined using a second-order polynomial fit directly to the observed spectra after subtracting off the initial contamination estimate. The contamination model of each spectrum can then be used to mask severely contaminated pixels when producing the final stacked spectrum across all orientations as discussed below.

Redshifts for sources in the grism spectra were primarily identified from the 1D spectra using a combination of H α, H β, [O ii], and [O iii] emission features. Given the low-resolution nature of the grism spectra, significant blending of lines can occur, e.g. [O iii] 4953 Å with [O iii] 5007 Å  and H α 6562 Å with [N ii] 6583 Å. In the higher signal-to-noise spectra these blends exhibit an asymmetry in the overall line profile, allowing (alongside the presence of H β emission) reliable differentiation between these lines. This asymmetry (and any H β emission) is often lost in lower signal-to-noise spectra and as such it can be difficult to differentiate between H α at z ∼ 0.9 and [O iii] at z ∼ 1.6. We therefore perform the line identification in two stages. First, lines are identified using an automatic fitting algorithm which then determines matches in wavelength ratios to common emission lines. We then visually inspect each spectrum, identify reliable detections, and attribute each detection a quality rating on a scale of Q_w = 1 to Q_w = 4 (where 4 represents a high S/N, multiple emission line detection and 1 represents a low-S/N less secure redshift). We mitigate the effect of contamination between nearby sources by simultaneously inspecting: a median stacked spectrum of all 16 exposures; the 4 separate median spectral stacks from the different roll-angle positions; and a mean-combination stack where contaminated pixels in individual exposures are rejected based on the contamination model described above (see Fig. 3). Whilst the observations are most sensitive to detecting emission line objects at z > 0.68, in the absence of such emission lines, galaxies are also identifiable via absorption features and, at z ≳ 2, the 4000 Å break (examples of which are shown in Fig. 3).

Figure 3.

Example objects from the HST WFC3 grism data. In each case the top-left panel shows the median spectrum (orange) and contamination-masked mean spectrum (dark red). The lower-left panel in each case shows the corresponding spectra taken at each roll angle, whilst the right hand panel in each case shows a thumbnail taken from the WFC3 F140W stacked image.

Open in new tabDownload slide

We produce a catalogue of secure sources which require at least one of the following criteria to be met: multiple lines clearly observed in the grism spectrum; clearly asymmetric lines indicating blended [O iii] emission or H α/[N ii] emission; or confirmed redshift from optical spectroscopy (i.e. VLT/MUSE data described below). The secure redshifts are used as a calibration/training set in our photometric redshift fitting analysis described below. With the photometric redshifts in hand, we then constrain the single blended emission line objects to the redshift most closely matched by the available photometric data. These objects are assigned a quality flag of Q_w = 2. For the subset of targets falling within the VLT/MUSE data field of view, we also use the optical spectra as a guide – both using the existence or the lack-of expected emission line features in the MUSE data to guide the identification of the grism data in this region.

We show the on-sky distribution of spectroscopically identified sources in Fig. 1, whilst example spectra from the grism data are shown in Fig. 3. The magnitude distribution of sources identified in the WFC3 G141 grism spectroscopy is given in the bottom panel of Fig. 4 and the redshift distribution is shown in Fig. 5.

Figure 4.

Magnitude distribution of sources in the MUSE (top panel) and WFC3 (lower panel) field of views. In both panels, the fainter circles denote number counts for all detected sources, whilst the darker circles show those with successfully identified redshifts from the MUSE and WFC3 grism spectroscopic data.

Open in new tabDownload slide

Figure 5.

Redshift distribution of the MUSE spectroscopic redshifts (top panel) and the HST/WFC3 grism redshifts (lower panel). The vertical dashed line marks the redshift of the central quasar, PKS 0232−04.

Open in new tabDownload slide

2.3 VLT MUSE IFU spectroscopy

2.3.1 Overview and data reduction

The VLT/MUSE data for this field were taken as part of the Guaranteed Time MUSE Program (096.A-0222, PI: Schaye). MUSE provides IFU data across a 1 arcmin × 1 arcmin field of view, with a pixel scale of 0.2 arcsec pixel⁻¹ (Bacon et al. 2010). All the observations were taken in the normal spectral mode with a wavelength coverage of 480–930 nm and resolution ranging from R = 1770 at the lower end of the wavelength range to R = 3590 at the higher end. A total of 12 exposures were taken, each of 900 s, totalling 3 h of integration on source. We have reduced the entirety of these data using the ESO MUSE pipeline (Weilbacher et al. 2014) with custom python scripts² (Fumagalli et al. 2014, 2017a) to optimize the illumination correction across different CCDs and using zap (Zürich Atmosphere Purge; Soto et al. 2016) to optimize the sky subtraction.

2.3.2 Source extraction and identification

Given the deep nature of the WFC3 NIR imaging, we use this as the basis for the source identification in the MUSE data. We first match the astrometry between the MUSE cube and the WFC3 F140W image (again using scamp, Bertin 2006) and then run SExtractor on the F140W image. From the resulting catalogue, we extract 1D spectra from the MUSE cube along with the variance and sky background measured for each spectrum. We use marz (Hinton et al. 2016) to measure redshifts from the spectra based on template fitting and visual inspection. For this we use A. Griffiths (September 2016) fork of the marz tool,³ which includes templates covering redshifts up to z ∼ 6 (for our wavelength coverage). Each identification is allocated a confidence flag, Q_m based on the following categorization:

• 1 – Low confidence/low S/N (can be single or multiple possible features);

• 2 – Single line emitter, low-S/N continuum or continuum fit with weak or no emission lines;

• 3 – Single line emitter, some hint of absorption lines in continuum or multiple emission lines with some at low signal-to-noise;

• 4 – Multiple high-S/N emission and/or absorption lines;

• 6 – Star.

The single line emitters are predominantly [O ii] emission at 0.88 ≲ z ≲ 1.5 (which is partially resolved in the MUSE data) or Ly α at z ≳ 2.9 (which can be identified as such via asymmetries and continuum shape).

The magnitude distribution of sources in the MUSE field of view are shown in Fig. 4 (top panel). The pale blue circles show the total number of sources extracted from the WFC3 data in the MUSE field of view, whilst the dark blue circles show the magnitude distribution of sources identified with confidence flags of Q_m = 2, 3, 4, and 6. The spectroscopic identification is |$\gt 50{{\ \rm per\ cent}}$| complete up to i < 25.8 (where the i-band photometry is taken from the broad-band imaging described later). The redshift distribution of sources identified in the MUSE cube is given in the top panel of Fig. 5. The quasar redshift is marked by the vertical dashed line and coincides with an overdensity of [O ii] emitting galaxies at z ≈ 1.4. The redshift distribution appears to show some structure with a particular peak at the redshift of the QSO. For a detailed discussion of the QSO environments please see Stott et al. (in preparation).

Using the MUSE spectroscopic data, we are able to measure the accuracy of the WFC3 grism derived redshifts. The grism data are at a low resolution – R = 130, which equates to a rest-frame velocity uncertainty of ≈1000 km s⁻¹ at z ≈ 1. With a resolution of R ≈ 1800−3500, MUSE has greater redshift accuracy (i.e. ≈50 km s⁻¹ at z ≈ 1), allowing a robust quantification of the grism line-fitting process’s accuracy. A comparison of the redshifts determined using WFC3 and MUSE as a function of redshift is shown in Fig. 6, with v_m − v_w giving the velocity offset between the MUSE and WFC3 redshifts in km s⁻¹. The dashed curves show the velocity uncertainty derived simply taking the G141 resolution limit of R = 130. From the 19 objects matched between the two samples, we find a velocity uncertainty on the G141 data of σ_v = 680 km s⁻¹ (at z = 1). A closer inspection of the two outliers (|Δv| > 2000 km s⁻¹) shows that one is a very extended bright object, where the emission line is heavily ‘smeared’ in the grism spectrum due to the internal kinematics of the galaxy; whilst the second appears to be a statistical outlier due to spectral noise.

Figure 6.

Rest-frame velocity offsets between redshifts for the same objects measured separately with VLT/MUSE and the HST/WFC3 G141 grism (red hexagons and histogram). The dashed curves denotes the rest-frame velocity scale given by ±c/R/(1 + z), where R = 130 is the instrumental spectral resolution of the G141 grism. The achieved velocity accuracy of σ_v = 682 km s⁻¹ (at z = 1) is represented by the solid red curves.

Open in new tabDownload slide

2.4 Optical imaging data

Supporting imaging is an important element in the analysis of the grism data, in particular in providing complementary photometric redshifts and galaxy properties. As described above, the photometric redshifts are needed for robust identification of objects detected with only single line features in the grism data. Below we describe the imaging datasets available on this field: (1) targeted William Herschel Telescope (WHT) auxiliary-port camera (ACAM) observations; and (2) coverage from the CFHTLS Wide-1 field.

2.4.1 WHT ACAM

The WHT ACAM imaging provides i ∼ 25 mag depth imaging, with the SDSS g, r, and i filters. ACAM is mounted at the folded-Cassegrain of the WHT and covers a field of view with diameter ⊘ |$\approx 8\,\mathrm{ arcmin}$| at a pixel scale of 0.25 arcmin pixel⁻¹. All observations were taken in observing runs W13BN5 (2013 November 24–25, PI: R. Crain), W14AN16 (2014 April 6–7, PI: R. Crain), and W17AP6 (2017 March 3–5, PI: R. Bielby). Seeing conditions were generally good with an image quality of ≈0.8–0.9 arcsec.

The ACAM data were reduced using standard image reduction methods. Master bias and flat-field frames were produced by stacking individual calibration frames, which were then applied to the science images. Individual weight maps for use with astromatic⁴ software were produced from the flat-field images combined with a bad pixel mask. Astrometric solutions were derived for the images in each field using scamp and these were then used in constructing stacked images using swarp.

Photometric calibration was performed by matching to the available (shallower) SDSS photometry in the region. Number counts for each of the filters are shown in Fig. 2 (dark blue squares), based on source catalogues extracted using SExtractor on the individual images. The corresponding depths, based on 80 per cent completeness of artificial point sources placed in the images, are given in Table 1.

2.4.2 CFHTLS wide

The central quasar studied here falls ≈3 arcmin from the Eastern edge of the CFHTLS (Cuillandre et al. 2012) W1 (wide) field (specifically in the ‘W1.+4 + 3’ region). The CFHTLS Wide survey provides relatively uniform imaging across a large area with the CFHLT ugriz filter set. Image quality is constrained to ≈0.9 arcsec across the survey, with the imaging reaching depths of u, g, r, i ≈ 25 and z ≈ 24 (AB), i.e. well suited to our needs. We note for reference that almost all of the W1 field has been surveyed spectroscopically to i ≈ 22.5 as part of the VIPERS survey (Scodeggio et al. 2018), however the central quasar studied here lies marginally outside the extent of that survey (12 arcmin from the nearest VIPERS spectroscopic data point).

We downloaded the publicly available median stacked images for each of the five bands in W1.+4 + 3 from CADC (Canadian Astronomy Data Centre⁵). Using scamp, these were then matched to the WFC3 imaging astrometry and then resampled and cropped using SWarp to match the WFC3 stack field of view. The image seeing and 80 per cent point-source completeness levels are reproduced in Table 1 (taken directly from the CFHTLS synoptic table⁶). We show the number counts for objects detected in each of the bands across the WFC3-stack field of view in Fig. 2, based on catalogues extracted from the images using SExtractor.

2.5 Galaxy properties

2.5.1 Collated photometric catalogue

In addition to the individual F140W catalogue for grism extraction, we also produce a collated catalogue of all 10 ACAM, CFHTLS, and WFC3 filters. To do so, all images were astrometrically matched and resampled to a common grid using a combination of the scamp and swarp software. SExtractor was then run in dual image mode using the F140W image as the detection image.

2.5.2 Photometric redshifts

Photometric redshifts are determined using the le phare photometric fitting code (Arnouts et al. 1999; Ilbert et al. 2006), using the ‘COSMOS’ set of SED templates (as used by Ilbert et al. 2010). The template SED fitting was performed on all the available imaging data and using all secure spectroscopic galaxy redshifts from the MUSE IFU and WFC3 grism data (i.e. Q_w,m = 3 and Q_w,m = 4) as a training set. As discussed above, the photometric redshifts for the single line emitters with non-secure redshifts were then incorporated into optimizing the redshift identification for these objects.

A comparison between the le phare photometric redshifts and the available spectroscopic data from WFC3 and MUSE is shown in Fig. 7 for magnitude limits of i_mag < 24, i_mag < 25 and i_mag < 26 (top, middle, and bottom panels, respectively). The red hexagons denote spectroscopic redshifts from WFC3, whilst the blue squares denote those from MUSE. In both cases, we only use those with confidence flags of 3 or 4 (given that objects with flags of 2 have partially been assigned redshifts with the photometric redshifts as a prior). Where an object is detected with both WFC3 and MUSE, we take the more accurate MUSE redshift. For each sample we quote the resultant photometric redshift accuracy, σ_{z/(1 + z)}, and outlier percentage, η.

Figure 7.

Photometric (z_p) versus spectroscopic (z_s) redshifts for all galaxies identified with confidence Q_op = 3 or Q_op = 4 in the HST/WFC3 grism (red hexagon points) or the VLT/MUSE (blue square points) data. The dashed line shows z_p = z_s and the dotted lines show the 1σ on the distribution of (z_p − z_s)/(1 + z_s).

Open in new tabDownload slide

The photometric fitting shows reliable results at i < 24, with an accuracy of σ_{z/(1 + z)} = 0.053 and an outlier rate of |$\eta =5.9{{\ \rm per\ cent}}$|⁠, whilst the results degrade at i > 24 as the photometric uncertainties increase. Evident from the comparison is a large scatter in the photometric redshift accuracy at z = 1.44. These are galaxies in the local environment of the bright QSO, many of which appear to have spectral line ratios indicative of significant active galactic nucleus (AGN) activity (which we investigate further in a forthcoming paper). Whilst these specific objects have little impact on the intervening absorber science in this paper, we note that any AGN in the foreground of the QSO will have accurate spectroscopic redshifts from their strong line features. We note that the varying image quality across the imaging bands could cause issues for the photometric fitting, this is countered to some degree by the adaptive fitting incorporating the spectroscopic training set, which determines and applies constant offsets (of Δm ≲ 0.1) to the input photometry.

2.5.3 Star formation rates

Over the main redshift range of interest (0.68 < z < 1.44), we primarily use the H α fluxes measured from the grism data to estimate SFRs for the individual galaxies in our sample. Using the relation given in Kennicutt (1998), divided by a factor of 1.8 to convert from Salpeter to Chabrier IMF. The galaxy intrinsic H α luminosities are estimated using an extinction correction of A_H α = 0.818A_V (Cardelli, Clayton & Mathis 1989), where we use a value of A_V = 1 (consistent with the mean absorption estimated from the template fitting with le phare and with Garn & Best 2010 for galaxies of mass M_⋆ ≈ 10^9.5–10 M_⊙). We remove the estimated [N ii] contribution to the blended H α-[N ii] emission assuming log([N ii]/H α) = −1.0 (Faisst et al. 2018).

We supplement the H α derived SFRs with [O ii] SFR estimates where [O ii] emission is observed either in the grism data or the MUSE data. SFRs are estimated from the [O ii] fluxes based on the relationship given by Kennicutt (1998) adapted for the non-local galaxy population by Kewley, Geller & Jansen (2004) and using A_V = 1 and A_[O ii] = 1.54A_V based on Cardelli et al. (1989). We note that for the objects where both [O ii] and H α are detected, we find the two indicators give broadly consistent measures of the SFR (with an average offset of ≈0.03 dex and a scatter of ≈0.4 dex). In these instances, where we have both [O ii] and H α measures of the SFR, we favour the H α for the analyses presented here given that it will be less affected by extinction and have a lesser dependence on metallicity and ionization parameter (e.g. Kewley et al. 2004; Gilbank et al. 2010). The resulting galaxy SFRs are given as a function of redshift in the upper-left panel of Fig. 8. We find a Hα-derived SFR systematic lower limit in the primary redshift range of interest in this study (i.e. 0.68 < z < 1.44) of ≈0.2–0.8 M_⊙ yr⁻¹. For galaxies with upper limits on the H α and/or [O ii] fluxes, we propagate these upper limits to determine upper limits on the SFR.

Figure 8.

Derived photometric properties of the spectroscopically identified galaxy sample. SFRs were estimated primarily from measured H α emission where available and [O ii] where H α was not covered by the spectral range of the data. Galaxy masses were estimated using photometric fitting to the available photometric data. The error bars along the right-hand axes of the left-hand panels show the mean estimated uncertainty on the points in bins of 1 dex. In the right-hand panels we limit the data to the prime redshift range of interest (0.68 < z < 1.4) and colour code points by redshift (as defined by the associated colour-bars). For reference, we also show the M_⋆−SFR relations derived at z ≈ 0.5, z ≈ 0.9, and z ≈ 1.4 given by Karim et al. (2011) (solid curves colour-coded by redshift, with the line widths indicating the uncertainties on the mean SFR measurements).

Open in new tabDownload slide

2.5.4 Galaxy stellar masses

Given the catalogue of spectroscopic and photometric redshifts for our sample, we next derive galaxy stellar masses also using le phare (via the gazpar online interface⁷). Following Ilbert et al. (2010), we use a set of SED templates calculated using the stellar population synthesis models of Bruzual & Charlot (2003). We assume a Chabrier (2003) IMF and an exponentially declining star formation history. Dust extinction is then applied to the templates using the Calzetti et al. (2000) law, with E(B − V) in the range 0–0.7.

Stellar masses for the spectroscopically identified galaxy sample are given in the lower-left panel of Fig. 8. The depths of our NIR imaging correspond to approximate limiting masses of M_⋆ ≈ 10^7.5–8 M_⊙ across the redshift range 0.68 < z < 1.44. In the right-hand panels of Fig. 8 we show the SFR and sSFR versus stellar mass for the galaxy sample in the redshift range 0.68 < z < 1.44. Points are colour-coded by each galaxy’s redshift. For comparison, the plotted line shows the z ≈ 1 star formation main-sequence trend reported by Karim et al. (2011). From this, we can see that the galaxies lie on or around the star-forming main sequence.

2.5.5 Inferred galaxy halo properties

2.6 High-resolution ultraviolet absorption spectroscopy

The HST spectroscopic observations with COS were obtained as part of the COS Absorption Survey of Baryon Harbors (CASBaH, HST Programs 11741 and 13846, PI: T. Tripp). Full details about the CASBaH programme design and data handling are reported by Tripp et al. (in preparation); here we briefly summarize some important aspects of the data set. CASBaH was designed to obtain high-resolution spectra with complete coverage from observed wavelength λ_ob = 1150 Å to the redshifted wavelength of the Ly α emission line of each target QSO. This was achieved using the far-UV G130M and G160M COS gratings (resolving power R = 18 000−22 000 or FWHM ≈15 km s⁻¹), the near-UV COS G185M and G225M gratings (R = 16 000−24 000; Osterman et al. 2011; Green et al. 2012), and the STIS E230M echelle spectrograph (R = 30000; Kimble et al. 1998; Riley et al. 2018; Woodgate et al. 1998). Table 2 provides a log of the COS and STIS observations.⁸

The FUV G130M and G160M spectra were obtained to survey very weak absorption lines such as the Ne viii doublet and accordingly required higher signal-to-noise (S/N) ratios. The NUV were obtained to record the stronger H i and longer-wavelength metal species (e.g. C iii, Si iii, and O vi) affiliated with Ne viii absorbers; for the detection of these stronger lines, lower S/N was acceptable. For PKS 0232−042, the COS G130M and G160M spectra have median S/N ratios of 13 and 17, respectively, whilst the COS G185M and G225M spectra have median S/N = 5. The STIS E230M spectrum has median S/N  = 7.

We reduced the FUV data as described in Meiring et al. (2011) and Tripp et al. (in preparation). In brief, we used the calcos pipeline (version 3.1.7) to carry out the initial reduction steps leading to one-dimensional extractions of the spectra from the individual exposures, and we aligned the individual exposures by comparing the positions of well-detected lines with distinctive component structure. COS employs a photon-counting detector with very low backgrounds, so we next coadded the individual exposures to accumulate the total gross and background counts in each pixel, which we used to determine the uncertainty in the flux based on counting statistics. For this absorption study, the absolute flux of the QSO is irrelevant, so no absolute flux calibrations were applied to the FUV spectra. The COS pipeline produces highly oversampled data, so we also binned the spectra to two pixels per resolution element to optimally sample the spectra. For the NUV COS and STIS data, we similarly used calcos and calstis (v. 2.22) to extract 1D spectra. However, for these data it was more convenient to flux calibrate the data (e.g. to account for the echelle blaze function of STIS), and then we aligned and coadded the individual spectra with the weighting method described by Tripp et al. (2001).

Absorption lines in the PKS 0232−042 spectrum were identified by inspecting every single line in the spectrum in many ways with a variety of tools. For every line, we checked whether the line would be an O vi candidate based on the velocity spacing and relative strength of the O vi doublet. This search for O vi did not require that H i is detected or not; we only searched for two lines with the signature spacings of O vi. Interestingly, detection of O vi without H i mainly occurs in absorbers with z_abs ≈ z_quasar, i.e. this seems to be a signature of ejected or at least ‘proximate’ absorption systems (Tripp et al. 2008). Examples of individual O vicomponents in intervening absorbers without H i at the same velocity can be found in Tripp et al. (2008) and Savage et al. (2010). However, in these examples, H i is clearly detected in other components of the same absorption systems. We corroborated the line identifications by searching for H i and other metal lines detected in the O vi absorption systems. Affiliated metals and H i lines often have similar (or identical) component structure as the O vi absorption profiles (Tripp et al. 2006, 2008, 2011), which gives additional credence to the identifications. Information on the other lines detected in these systems can be found in Tripp et al. (in preparation). Note that Tripp et al. (in preparation) provide information on additional, weaker O vi candidates; in this paper we concentrate on the well-detected O vi systems recorded at high significance (with upper limits used in place of the low-confidence weak lines).

Using the Voigt-profile fitting software developed by Burchett et al. (2015), we fitted multicomponent absorption-profile models to constrain the column density (N_O vi), velocity width (b), and velocity centroid of each component, and for purposes of comparison, these Voigt-profile parameters were used to calculate the equivalent width of each component using a full curve of growth. We use the convention of Tripp et al. (2008), i.e. we use the velocity centroid of the strongest O vi component to define the systemic redshift of the absorber. The resulting measurements for each O vi absorption doublet are given in Table 3. In general, we fit all available resonance transitions of a specific species simultaneously (e.g. the O vi 1031.93 and 1037.62 Å lines are fitted jointly and must have the same parameters), but different species are fitted independently (e.g. the H i profile parameters are entirely independent of the O vi parameters). We do not attempt to extract information about turbulent versus thermal broadening of the lines in this paper. In some instances, the O vi (or other) lines of interest are blended with interloping lines from other redshifts. In these cases, we include the interloping lines in the fit, but again, the interloper parameters are allowed to freely vary and are entirely independent of the parameters of the target O vi lines. Because the CASBaH data have broad wavelength coverage extending from 1152 Å to the Ly α emission line of the QSO, the interloping lines are often well constrained by other lines in the CASBaH data. For example, the O vi 1031.93 Å line at z = 0.173 55 is blended with an H i 915.82 Å line at z = 0.322 50; the parameters of this interloper are tightly constrained because many other H i Lyman series lines at this redshift are recorded in the CASBaH data, and those other Lyman lines are included in the overall fit.

We show the absorption systems, and the associated model fits, used in this paper in Fig. 9. In each panel, the black line shows the entire Voigt-profile model including all blends that were included in the fit, and the red line shows the Voigt-profile model for only the species of interest (indicated in each panel). We show three panels per absorption system: one of the H i Lyman series lines (often there are many to choose from; we selected the one that was most informative, but all available Lyman series lines were used in the fit); and both of the lines of the O vi doublet. Whilst we are only studying O vi in this paper, it is helpful to see the corresponding H i to corroborate the O vi identification. Tripp et al. (in preparation) gives the full suite of lines detected and measured in each of these systems. Note that for the O vi absorber at z_abs = 0.173 55, the O vi 1038 Å line is lost in the Milky Way Ly α  + geocoronal Ly α complex, so we do not show the 1038 Å data in that stack. Instead, we show the Ly α, Ly β, and O vi 1032 Å lines.

Figure 9.

Spectra (blue step histogram) of O vi and H i absorption lines (as labelled) overlayed with the absorption-profile fits. In each panel, the black line shows the full fit (including blends from interloping lines from other redshifts), and the red line shows the Voigt-profile model for the O vi or H i only.

Open in new tabDownload slide

3 THE GALAXY ENVIRONMENT OF O vi ABSORBERS

The blind nature of both the HST-WFC3 grism and VLT-MUSE large field IFU data provide a large quantity of information on the galaxy environment directly surrounding the gaseous environments probed along the quasar sightline. We aim to utilize these data to probe the galaxy environments of the individual O vi absorption systems along the line of sight.

We consider the systems in two separate redshift ranges: i.e. a low-redshift sample (i.e. z < 0.68), where the galaxy population is primarily probed up to ≈0.5 arcmin from the sightline by the MUSE data; and a high-redshift sample (0.68 < z < 1.44), where the population is revealed over a wider field of view up to ≈1.5 arcmin with the grism data. For the analyses that follow, we define a velocity window within which to consider an absorber to be associated with a local galaxy. We consider a number of physical factors: (1) the rotational velocity of galaxies within the mass range covered by our sample; (2) the typical range of velocities measured for outflowing material from galaxies; (3) the velocity dispersion of galaxies within galaxy groups and clusters; and (4) the velocity uncertainties on the galaxy redshifts themselves.

Taking typical galaxy rotation curves for guidance, galaxies at z ≈ 1 with masses of ≈10¹⁰ M_⊙ have median rotational velocities of ≈120–160 km s⁻¹, although rotational velocities of up to ≈2 × the median are common (e.g. Johnson et al. 2018). Velocity offsets of up to ≈320 km s⁻¹ therefore seem reasonable if any O vi absorption is tracing co-rotating material within the galaxy halo. Considering galactic winds driven by star formation activity, outflow velocities of v_w ≈ 200–400 km s⁻¹ are typical at the redshifts we probe (e.g. Mathes et al. 2014), although wind velocities of up to ≈600 km s⁻¹ have also been reported (e.g. Shopbell & Bland-Hawthorn 1998; Shapley et al. 2003; Tripp et al. 2011). In terms of large-scale galaxy structures, massive galaxy clusters can have velocity dispersions up to ≈1000 km s⁻¹ (e.g. Struble & Rood 1999). Such a large velocity window is clearly not appropriate for draw associations with the average galaxy population, however, so such a large offset will only be considered in this work where a galaxy group or cluster is detected. Allowing for the potential range in velocity offsets, we set a velocity window for associating absorbers with galaxies of Δv_max = max{400 km s⁻¹, σ_v}, where σ_v is the velocity uncertainty on any given galaxy (i.e. ≈50–80 km s⁻¹ for galaxies with MUSE redshifts and ≈682 km s⁻¹ for galaxies with only grism redshifts). As a reference point, low redshift studies (Werk et al. 2016; Nielsen et al. 2017; Kacprzak et al. 2019) find O vi absorber–galaxy pairs, at z = 0.2, associated in velocity space within |Δv| ≲ 200 km s⁻¹.

In the following section, we provide details of individual absorber–galaxy associations, discussing system properties and environment on a case-by-case basis. We collate the data and present the statistical properties of the sample as a whole in Section 3.2.

3.1 Census of O vi systems

An overview of the field is presented in Fig. 10, where all detected O vi absorbers (diamonds) and galaxies (squares and hexagons) at z < 1.5 are shown as a function of redshift and impact parameter to the quasar sightline (lower panel). The top panel shows a histogram of the galaxy population within 600 kpc of the sightline at 0.68 < z < 1.44. The vertical shaded regions highlight redshifts coincident with the O vi absorbers. The dotted and dashed horizontal lines in the top panel denote the median and twice the median galaxy density in the field, respectively. The data reveal a diversity of environments in terms of galaxy density and potential associations between O vi absorption systems and the galaxy population.

Figure 10.

Redshift versus impact parameter for all identified galaxies within the MUSE (inner dashed curve) and WFC3 grism (outer dashed curve) fields of view. O vi absorbers along the quasar sightline are marked by the shaded regions with ±1000 km s⁻¹ intervals to highlight any potentially associated galaxies or galaxy structures. The galaxy number density within Δv ± 100 km s⁻¹ bins (histogram) alongside the O vi absorber positions. The horizontal dashed line shows the median observed galaxy density in the WFC3 grism field of view (within the redshift range 0.68 < z < 1.38). The O vi absorbers are found in a range of galaxy environments, with at least one out of the five coinciding with a potential group environment (i.e. twice the mean density as given in the text) and two having no associated galaxy within 200 kpc (although all have associated galaxies within ≈350 kpc).

Open in new tabDownload slide

We now look at how the ionized oxygen is distributed around the galaxies, by analysing the column density of O vi absorbers as a function of impact parameter to individual detected galaxies within our given Δv_max constraints. This is shown in Fig. 11, where the filled red hexagons and dark blue squares denote the QSAGE O vi–galaxy pairs at z > 0.68 and z ≤ 0.68, respectively. We differentiate upper limits as pale hexagons and squares for the two samples respectively. These upper limits are calculated within Δv ≤ 50 km s⁻¹ of detected galaxies in the survey sample. The open hexagons show the impact parameters of galaxies that are not the nearest to the sightline, but do fall within the Δv_max velocity window. We find five O vi absorption systems associated with galaxies in our sample at z > 0.68 and a further three at z ≤ 0.68, given the Δv_max constraint on associations. These fall primarily at impact parameters of ≈100 kpc, but extend to ≈350 kpc in the most extreme case. Within this range in impact parameters, we found numerous non-detections with limits of N_O vi ≲ 10^13.5–14 cm⁻². These non-detections point to a large scatter in O vi column densities within ≈300 kpc of galaxies at z ∼ 1.

Figure 11.

O vi column density as a function of impact parameter. The low-redshift (z < 0.68) points from our dataset are shown by the blue squares, whilst the red hexagons denote the 0.68 < z < 1.44 sample (in both cases, the paler symbols denote upper limits). The open symbols show galaxies coincident with a detected absorber, where a galaxy at smaller impact parameters is also present. Data points from Kacprzak et al. (2015) and Pointon et al. (2017) are shown by the blue plus and orange times symbols, respectively. The data of Werk et al. (2016) are shown by the green triangles, diamonds, and circles (denoting broad, narrow, and no low-ionization subsets). The dashed curve and shaded region show a fit to the Werk et al. (2016) broad absorbers (extrapolated to larger impact parameters).

Open in new tabDownload slide

For comparison, we also show the COS-Haloes measurements taken from Werk et al. (2016, green points). The triangle, diamond, and circle points show broad, narrow, and ‘no low-ionization’ absorbers, respectively. The Werk et al. (2016) data consist primarily of isolated galaxies at z < 0.2, with impact parameters of ≲120 kpc (≲0.9 R_vir) and M_⋆ ≳ 10^9.6 M_⊙. The blue ‘+’ symbols show the results of Kacprzak et al. (2015). These show isolated galaxies probing down to a faintest galaxy near-infrared (AB) magnitude of ≈22.5 mag (corresponding to a minimum stellar mass of ≈10⁸ M_⊙). We estimate these data probe up to a maximum impact parameter in units of the virial radius of b/R_vir ≈ 1.8. Complementary to these, the orange ‘×’ symbols show column densities from Pointon et al. (2017), which correspond to galaxies identified to be in groups at z < 0.5. In collating these datasets, we only plot points for the galaxies that are deemed to be ‘associated’ with each individual absorber by the corresponding authors (i.e. we do not plot multiple galaxies for a single absorber). Combining these datasets here produces a broad mixture of galaxy properties and environments, such that caution is required in comparing any one sample to another. However, noting this, our own sample is effectively a blind selection incorporating whatever environments and galaxies happen to be found along the given sightline.

The dashed curve and shaded region show a fit calculated by Werk et al. (2016) for their broad absorber (triangle) data. The distinction between the broad and narrow categories here is identified using the Doppler parameter at a value of b ∼ 30 km s⁻¹. Werk et al. (2016) exclude narrow O vi absorption and absorbers with no associated low-ionization lines from this fit to the data, finding that these dominate the large scatter in the observed column densities. Quantitatively, 15 out of 16 of the Werk et al. (2016) broad absorbers lie on the relation, whilst 13 of 17 narrow and ‘no-low’ absorbers deviate significantly below the fit. Two of our own data points at z < 0.68 lie at impact parameters, both in proper distance and as a fraction of virial radius (r_vir), approximately equivalent to the Werk et al. (2016) data. These both lie well below the z ∼ 0.2 broad-absorber fit and are more consistent with the narrow absorbers of Werk et al. (2016). Indeed both have velocity widths of b ≲ 30 km s⁻¹, chosen by Werk et al. (2016) to classify between broad and narrow absorption systems. The third galaxy-absorber pair similarly corresponds to a narrow absorption feature and lies significantly below the Werk et al. (2016) fit to the column densities of broad absorbers extrapolated to impact parameters beyond the range of their data.

Considering the 0.68 < z < 1.4 absorber–galaxy pairs from our data, these generally lie at impact parameters extending beyond the range of the Werk et al. (2016) data, whilst also representing galaxies of different masses, SFRs and environments. Despite these caveats, the detections at z > 0.68 in our own data (which are predominantly b ≳ 30 km s⁻¹ absorbers) are consistent with the extrapolated low-redshift fit, suggesting some consistency between the picture at z < 0.2 and at z ∼ 1. The extent to which we identify absorbers at distances of up to b ≈ 300–400 kpc (or b ≈ 4–5R_vir) from the most proximate detected galaxy appears comparable to wide field low-redshift results such as those presented by Mathes et al. (2014), Kacprzak et al. (2015), Johnson, Chen & Mulchaey (2015a), and Pointon et al. (2017).

We now consider individual systems of interest: (1) those where O vi is detected; (2) those where a galaxy or set of galaxies falls within b ≲ 2R_vir of the quasar sightline but no O vi is detected; or (3) a candidate group/cluster environment is found within the WFC3 data field of view. In this instance, we focus on galaxies within impact parameter b/R_vir = 2 cut prompted in part by results at low redshift (Tumlinson et al. 2011; Shull 2014) but also for brevity in focusing on only the systems of most interest where we might have expected to find absorption systems based on previous studies. We provide a list of the proximate galaxies along the sightline and their properties in Table 4.

3.1.1 Absorbers at z < 0.68

Whilst the data at z < 0.68 do not probe the large physical scales afforded by the higher redshift sample, we do detect a number of galaxies coincident with absorbers in the sightline. These are shown in Fig. 12 alongside all z < 0.68 galaxies that lie within b = 2R_vir where no O vi absorption is detected.

Figure 12.

The velocity (top panels) and spatial (lower panels) distribution of galaxies (hexagons) around each of the z < 0.68 systems where one or more galaxies are found within b = 2R_vir of the sightline. The presence of any O vi absorption lines at each redshift is denoted by the vertical dashed lines in the top panels and a central blue star in the lower panels. Dashed circles centred on the galaxy positions illustrate the scales of 2R_vir and 2σ_v for each galaxy. The dotted square in each spatial panel indicates the extent of the MUSE FOV at the system redshift.

Open in new tabDownload slide

The absorbers at z = 0.1736, z = 0.3224, and z = 0.4343 are all found to lie within 2R_vir of galaxies detected in the MUSE data. The velocity offsets between the galaxies and absorbers in all three cases are ≲50 km s⁻¹, i.e. relatively small in comparison to the nominal halo σ_v estimates for the galaxies (≳200 km s⁻¹). We note that the absorbers detected at z = 0.218, z = 0.356, z = 0.364, and z = 0.512 show no detected galaxies in the deep MUSE data within a few 1000 km s⁻¹ and therefore are not shown in Fig. 12. The MUSE field of view at these redshifts corresponds to maximum probed impact parameters of b = 109 kpc (compared to a virial radius of R_vir,12 ≈ 195 kpc for a M_⋆ ≈ 10¹² M_⊙ mass galaxy), 154 kpc (R_vir,12 ≈ 176 kpc), 156 kpc (R_vir,12 ≈ 174 kpc), and 191 kpc (R_vir,12 ≈ 157 kpc), respectively. Given previous studies at low redshift have found ‘host’ galaxies of O vi absorption at impact parameters of up to b ≈ 2R_vir, and some deep surveys at low redshift have found O vi absorbers at very large distances from the closest galaxy (Tripp et al. 2006; Johnson, Chen & Mulchaey 2013), it is not unreasonable that associated galaxies for these absorbers may fall outside of the field of view. Indeed, from a search of the NASA Extragalactic Database (NED) and the soon to be published CASBAH data, we find galaxies associated with these absorbers outside the MUSE field of view. In relation to the z = 0.512 absorber, we find a galaxy at z = 0.511 listed in NED, which is also detected in the WFC3 grism data via absorption features.

Taking the reverse, i.e. galaxies with no detectable associated O vi absorption in the sightline data, we find four such galaxies within b ≈ 2R_vir of the sightline at z < 0.68. Three of these form a small grouping, lying within ≈600 km s⁻¹ of each other at z ≈ 0.295. The upper limit on any associated O vi coincident with this grouping is N_O vi < 10^13.9 cm⁻² (which we note is higher than the detected systems at z < 0.68).

3.1.2 Galaxy environments probed by the sightline at z > 0.68

Fig. 13 shows the galaxy distributions around each of the 5 O vi absorption systems detected at z > 0.68. In each case, galaxies within Δv = ±680 km s⁻¹ are marked with green points, galaxies at −1600 km s⁻¹ < Δv < −680 km s⁻¹ with blue points, and galaxies at 680 km s⁻¹ < Δv < 1600 km s⁻¹ with red points. Galaxies within this range that have been identified in the MUSE IFU data are given a black outline, whilst the estimated 2R_vir scale for each galaxy is illustrated by the dashed ellipses. In each case Δv = 0 km s⁻¹ is centred on the strongest component observed in the O vi absorption system at each redshift. The large dotted circle in each of the spatial panels shows the approximate field of view of the WFC3 observations at each redshift. As with the low redshift sample, we find the absorbers predominantly lie within ≈100 km s⁻¹ of a nearby galaxy, except in the case of the absorber at z ≈ 0.73. Additionally, the impact parameters in three of these cases are at scales of b < 2R_vir.

Figure 13.

The velocity (top panels) and spatial (lower panels) distributions of galaxies at redshifts centred on detected O vi absorbers at z > 0.68. The sightline position is marked by a star in each case, whilst the O vi absorption lines at each redshift are denoted by the vertical dashed lines in the top panels. Galaxies marked in green are within Δv < 680 km s⁻¹ of the absorber, whilst blue and red points denote galaxies blueshifted and redshifted by Δv > 680 km s⁻¹ respectively. Points with a black hexagonal outline have been detected in [O ii] emission in the MUSE datacube. We mark twice the inferred virial radii of each galaxy within Δv = 1800 km s⁻¹ of the detected absorber. The large dotted circle marks the extent of the WFC3 grism data field of view at the redshift of interest.

Open in new tabDownload slide

The z ≈ 1.089 strong (⁠|$N_{\rm OVI}=10^{14.43\pm 0.19}\, {\rm cm}^{-2}$|⁠) O vi absorber in the quasar spectrum coincides with a detected overdensity in the galaxy population. Four of the galaxies lie within b = 200 kpc of the quasar sightline, with three of these having velocities within a range of Δv ≈ 250 km s⁻¹ of each other (based on the MUSE [O ii] measured redshifts). Of these three most closely associated galaxies, we find that the stellar mass is dominated by a M_⋆ = 10^{10.0 ± 0.1} M_⊙ galaxy, which lies at Δv = +230 km s⁻¹ and b = 170 kpc from the absorption system. The remaining two closely associated galaxies are estimated to have masses of M_⋆ ≈ 10^8.5 M_⊙.

Given the spatial and velocity distribution, we assume these to be the primary members forming a triplet/low-mass group (with the remaining galaxies at larger separations spatially and in velocity potentially tracing the large scale structure environment). Given only three detected members, it is not possible to reasonably estimate a halo mass from the velocity dispersion, however from the stellar mass, we infer a halo mass for the central group galaxy of M_halo = 10^{11.8 ± 0.1} M_⊙. From this halo mass, we estimate a virial radius of R_vir ≈ 90 kpc, meaning the dominant galaxy lies at b = 1.9R_vir from the sightline.

For the z ≈ 0.73 system, no galaxy lies within Δ_v < 680 km s⁻¹ of the absorber in the data sample. Indeed, this system may be a prospective candidate as the sightline tracing the cosmic web outside of galaxy haloes as discussed by Pessa et al. (2018). Those authors set a limit of |$\frac{\Delta v}{\sigma _v}\frac{b}{R_{\rm vir}}\gt 2$| in their analysis to identify candidate cosmic web absorbers, which is clearly satisfied here (see Fig. 13 where the dashed ellipses in the top panels show the extent of 2σ_v in velocity space and 2R_vir in impact parameter). We note that the redshift uncertainties on these galaxies (which lack MUSE coverage) are ≈680 km s⁻¹ and as such the lack of alignment in velocity space could in part be due to the redshift accuracy of the galaxies.

Looking to the remaining overdense regions along the sightline as traced by the galaxy distribution. From the galaxy density, we find peaks above twice the median galaxy density at z ≈ 0.698, z ≈ 0.977, z ≈ 1.026, z ≈ 1.089, z ≈ 1.131, and z ≈ 1.396 (marked in Fig. 10). For each of these we plot the galaxy distribution, both spatially (lower panels) and in velocity space (top panels), around the sightline in Fig. 14 (except for the z ≈ 1.089 system which has already been discussed and shown in Fig. 13). We centre the velocity axis in the top panels on the nearest galaxy (in units of R_vir) to the sightline.

Figure 14.

As in Fig. 13, but for the galaxy overdensity regions marked in Fig. 10. The sightline position is marked by a ‘×’ symbol in each case.

Open in new tabDownload slide

Overall, we find little sign of O vi absorption features tracing overdensities along the sightline probed, finding only one case out of 6 in which there is an association between detected O vi and a galaxy overdensity (i.e. the z ≈ 1.089 system already discussed). Comparing to similar studies, Pointon et al. (2017) find |$33\pm 16{{\ \rm per\ cent}}$| of their 18 galaxy groups align with detected O vi absorption. The Pointon et al. (2017) survey is at lower redshift and incorporates a greater number of galaxies at smaller impact parameters than our own sample, but despite this the two survey samples show approximately the same general trend for only a relatively small fraction of galaxy groups to align with strong O vi absorption.

3.2 The properties of O vi associated galaxies

Matter in the Universe is distributed hierarchically, such that galaxies exist within shared haloes. Assuming this present paradigm of hierarchically ordered dark matter haloes and sub-haloes, the conditions at any point within large scale structure can be driven and influenced by a number of components of the surrounding structure, e.g. the nearest galaxy, individual galaxies in the local neighbourhood, and as the overall mass and size scale of the matter halo hosting both the gas and the galaxies. As such (and combined with the large scales over which the O vi ion is observed around galaxies) it is worth identifying any correlations between not just the properties of the nearest galaxy to a given absorption system, but also to the wider galaxy population and environment. As discussed, this is one of the primary advantages of the strategy invoked in this survey, that we take a blind snapshot of the galaxy population around the quasar sightline.

We take two primary approaches in this section: (1) what are the observed properties of the most proximate galaxies to the sightline? and (2) what are those same properties for the most massive galaxies detected at each given redshift probed? The motivation for the latter is that, by isolating the most massive galaxies, these more closely relate the total structure mass that is being probed, thus opening the possibility of discerning any correlation between absorber properties and parent-halo properties (e.g. Oppenheimer et al. 2016; Nelson et al. 2018).

We begin by analysing the properties of galaxies lying closest in impact parameter to each O vi absorber, within our defined velocity window around an absorber redshift. These are shown in Fig. 15, with the absorber column density versus SFR of the nearest galaxy in the top panel and the absorber column density versus nearest galaxy stellar mass in the lower panel. In both cases, the red squares show the galaxy–absorber pairs at z < 0.68, whilst the 0.68 < z < 1.44 pairs are split into two groups: those with b < 2R_vir (red hexagons); and those at 2R_vir < b < 5R_vir (grey triangles). In each case the paler points denote upper limits on O vi absorption, whilst the darker points denote significant detections. We show the z < 0.2 data-points of Werk et al. (2013) as the green diamonds in both panels.

Figure 15.

Sightline O vi absorber column density versus measured stellar mass (top panel) and SFR (lower panel) of the nearest galaxy within a velocity offset of Δv_max (as defined in the text). Red hexagons show O vi absorber–galaxy alignments at 0.68 < z < 1.44 within b = 2R_vir of the sightline; grey triangles denote absorber–galaxy alignments at 0.68 < z < 1.44 in the range 2R_vir < b < 5R_vir of the sightline; and blue squares denote z < 0.68 absorber–galaxy alignments. The paler points in each case highlight points where only upper limit measurements are available on the absorption. We show the z < 0.2 data points of Werk et al. (2013) as green diamonds in both panels for comparison.

Open in new tabDownload slide

At z < 0.68, we find three significant detections of O vi are coincident with galaxies. The galaxies cover a wide range of SFR (0.01 M_⊙ yr⁻¹ ≲ SFR ≲ 2 M_⊙ yr⁻¹) with stellar masses of log(M_⋆/M_⊙) ≲ 10^9.5, lower than those observed by Werk et al. (2013). We note that three of the associations are at b/R_vir ≲ 1 (i.e. equivalent to the Werk et al. 2013 sample), with the remaining two at 1 ≲ b/R_vir ≲ 2. Our low redshift data are consistent with the spread of N(O vi) versus SFR data points in the Werk et al. (2013) results, although they show no correlation taken on their own.

At z > 0.68 and impact parameters of b/R_vir < 2, our data show a large scatter in N_O vi for the star-forming galaxies, with several SFR ≈1 M_⊙ yr⁻¹ galaxies coincident with O vi absorption of N(O vi) ≲ 10¹⁴ cm⁻². This particular field does not contain any z ∼ 1 low-SFR galaxies, although we would have been able to identify such galaxies in the MUSE field of view, i.e. up to ≈300 kpc, to a magnitude limit of i ≈ 24 if any were present. We note that the space density of passive galaxies to i ≲ 24 at z ∼ 1 is n ≈ 0.0003 h⁻¹ Mpc (e.g. Bielby et al. 2014), equating to ≈1 in a 1 arcmin², 0.68 < z < 1.44 volume on average (i.e. not taking into account the highly clustered nature of such galaxies). Our data in this field can do little then to inform both the low-SFR range (≲10^−0.5 M_⊙ yr⁻¹) and by association the high-mass range beyond M_⋆ ≳ 10¹¹ M_⊙. However, the fact that we find such low limits on O vi absorption around star-forming galaxies has implications for any causal connection between ongoing star-formation in a galaxy and the detection of O vi ions in the galaxy’s vicinity. Indeed, the large impact parameters involved, alongside recent kinematical studies (e.g. Nielsen et al. 2017), suggest that a causal connection between current SFR and O vi absorption systems is not universally applicable. Comparing to the low redshift Werk et al. (2013) data, we find a higher fraction of SFR ≈1 M_⊙ yr⁻¹ galaxies with no detected coincident O vi absorption (at the N(O vi) ≲ 10¹⁴ cm⁻² level) in our z ∼ 1 data, although the larger impact parameters probed in our own sample compared to the Werk et al. (2013) data will inescapably be a significant contributing factor to this.

Moving to the O vi column density as a function of coincident galaxy stellar mass, the z < 0.68 sample shows all the absorber-galaxy pairs correspond to stellar masses of M_⋆ ≲ 10^9.5 M_⊙ and column densities of N(O vi) ≲ 10¹⁴ cm⁻². Interestingly, combining this log(M_⋆/M_⊙) ≈ 10^9.5 sample with the Werk et al. (2013) data (with the caveat that two of our data points are at larger impact parameters than present in the Werk et al. 2013 data), shows some suggestion of a transition from low to high column densities around a proximate galaxy stellar mass of M_⋆ ∼ 10^9.5 M_⊙. The higher redshift sample primarily probes stellar masses of ≈10^9–10 M_⊙, i.e. coincident with the possible transition region in the low redshift data. Indeed, the z ∼ 1 data tentatively support this picture of an increase in O vi column density at stellar masses in the range ≈10^9.5–10 M_⊙. In addition, we find a single galaxy-absorber pair corresponding to a galaxy with stellar mass M_⋆ ≈ 10¹¹ M_⊙. This higher mass point corresponds to an O vi upper limit of N(O vi) ≲ 10¹¹ cm⁻² and overlaps with the regime in the low-redshift data of Werk et al. (2013) where a large fraction of galaxies show no detectable sightline O vi absorption. The small number of data points in this first analysis is restrictive in making any strong conclusions however.

We now reproduce the plot of column density versus galaxy properties from Fig. 15, but with the SFR (top panel) and stellar mass (lower panel) of the most massive galaxies lying at the redshift of each absorption system in Fig. 16. The symbols used are consistent with those in Fig. 15. For clarity and given the smaller field of view probed by MUSE, we omit the z < 0.68 O vi absorber–galaxy pairs.

Figure 16.

As in Fig. 15 but taking the stellar mass and SFR properties for the most massive proximate galaxies (i.e. up to 5R_vir). Only the 0.68 < z < 1.44 sample is now shown, given the more limited field of view of the z < 0.68 data.

Open in new tabDownload slide

This change of ‘host galaxy’ selection affects two absorber–galaxy pairs: one at b/R_vir < 2 and one at b/R_vir ≈ 3 (≈350 kpc). The N(O vi) versus SFR distribution shows no indication of any correlations given the available data. The distribution of points shows a large scatter in column densities across the full SFR range probed. For the stellar mass, in the case of both shifted data points, the most massive galaxy detected at the absorber redshifts have estimated stellar masses of ≈10¹⁰ M_⊙. For the z ∼ 1 galaxies with stellar masses of ≲10^9.5 M_⊙, we effectively detect no O vi absorption unless a more massive galaxy (i.e. ≳10^9.5 M_⊙) is also present. At |$M_\star \gt 10^{10}\, {\rm M_\odot }$|⁠, our dataset from this first field poorly samples the galaxy population, with only a single massive galaxy in this range (which corresponds to an O vi limit of N_O vi ≈ 10^13.9 cm⁻²).

In summary, we find a marginal preference for the N(O vi) ≈ 10^14.5 cm⁻² absorbers to be associated with galaxies of stellar masses M_⋆ ≈ 10¹⁰ M_⊙ (for both galaxy–absorber pairs lying at b < 2R_vir and b < 5R_vir). The detections of O vi absorption in the sightline all lie at the redshift of a galaxy within a mass range of |$10^{9.5}\, {\rm M_\odot }\lt M_\star \lt 10^{10}\, {\rm M_\odot }$|⁠, whilst the non-detections that also lie in this mass range offer only relatively poorly constrained upper limits of N_O vi ≈ 10^14.3 cm⁻².

3.3 Covering fraction of O vi around z ≈ 1 galaxies

We now calculate the covering fraction of O vi around the galaxy population as a function of impact parameter, galaxy stellar mass and galaxy star-formation rate. For this, we include all galaxies around the sightline and not just the nearest or most massive. We measure the covering fraction for O vi column densities of N_O vi > 10¹⁴ cm⁻², such that the covering fraction is given by |$c=\frac{n_{\rm gal}(\rm \mathit{ N}\gt 10^{14})}{n_{\rm gal}}$|⁠. Galaxies found within Δ_v = max{σ_v, 400 km s⁻¹} of a column density upper limit of N_O vi > 10¹⁴ cm⁻² (with no corresponding significant detection) are not included in the calculation (i.e. as these are likely the result of contamination from other absorption lines masked in the analysis).

The resulting covering fraction of O vi absorbers with column densities of N > 10¹⁴ cm⁻² is shown as a function of impact parameter (in units of the virial radius) in the left-hand panel of Fig. 17. We find a trend for increasing covering fraction with decreasing impact parameter in the 0.68 < z < 1.5 sample, with covering fractions of c ≈ 0.2 up to b/R_vir = 4. Covering fractions based on Werk et al. (2013) and Kacprzak et al. (2015) are shown by the green diamonds and blue plus symbols respectively. As described previously, the Werk et al. (2013) data are primarily higher mass and lower redshift than our own data, whilst the Kacprzak et al. (2015) data lie at z ≲ 0.5 and are comparable in stellar mass range to our own data. Clearly our own sample predominantly probes large scales in impact parameter than these previous works. Agreement is seen where the QSAGE data overlap with the Kacprzak et al. (2015) data at 1 < b/R_vir < 2.

Figure 17.

The covering fraction of N_O vi > 10¹⁴ cm⁻² O vi absorbers as a function of impact parameter (left-hand panel); separated into mass bins at M_⋆ = 10^9.6 M_⊙ (centre panel); and separated into SFR bins at SFR = 10^0.25 M_⊙ yr⁻¹ (right-hand panel). The green diamonds shown in all three panels show covering fractions based on Werk et al. (2013), whilst the plus symbols in the left-hand panel show covering fractions from Kacprzak et al. (2015).

Open in new tabDownload slide

In the central panel of Fig. 17, we show the covering fraction for galaxies split into two mass bins: M_⋆ < 10^9.6 M_⊙ and 10^9.6 M_⊙ < M_⋆ < 10^10.6 M_⊙. We find a moderately significant relation between covering fraction and stellar mass at b/R_vir ≈ 2, with the higher mass galaxies showing a higher covering fraction. Again, we show covering fractions based on the data of Werk et al. (2013), now split in bins of 10^9.6 M_⊙ < M_⋆ < 10^10.6 M_⊙ (i.e. matching our own higher mass sample) and M_⋆ > 10^10.6 M_⊙ (i.e. masses higher than probed by our own data). This shows an inversion with respect to our own results, where the higher mass sample now has a lower covering fraction (albeit at only the 1σ level).

The right-hand panel shows the covering fraction split into samples based on an SFR limit of 10^0.25 M_⊙ yr⁻¹. In this case we find no sign of a difference of the covering fraction based on the two different samples, in contrast to the low-redshift/low impact parameter covering fractions from the COS-Haloes sample (Werk et al. 2013, green diamonds), where a significant dependence on SFR is seen.

4 DISCUSSION

Our results have shown tentative evidence for O vi absorption preferentially being found in the vicinity of moderate mass (M_⋆ ≈ 10^9.5–10 M_⊙) star-forming galaxies. The star-forming properties of these z ≈ 1 galaxies are comparable (SFR ≳ 1 M_⊙ yr⁻¹) to that observed for isolated L^⋆ galaxies at low redshift coincident with O vi absorption (Tumlinson et al. 2011; Werk et al. 2014; Kacprzak et al. 2015).

We have reported on O vi absorption systems found up to scales of ≈300–400 kpc (≈ 4R_vir) from the nearest galaxy, comparable to the largest field of view low redshift surveys of such absorbers (e.g. Mathes et al. 2014; Johnson et al. 2015a; Pointon et al. 2017)

Prochaska et al. (2011) at low redshift similarly find a picture whereby the association between galaxies and O vi shows a strong dependence on galaxy luminosity. They find O vi absorption most commonly to be found within b ≈ 200–300 kpc of a 0.1L^⋆ < L < 1L^⋆ (intermediate sub-L^⋆) galaxies, and that where a dwarf galaxy is found closest to O vi sightline absorption, there often also exists an intermediate sub-L^⋆ galaxy within b ≈ 300 kpc of that same absorber. However, it is clear from low redshift studies that O vi systems arise from a diverse array of galaxy environments ranging from dwarfs to sub-L*–L* galaxies and even more massive systems. Prochaska et al. (2011) show that some dwarf galaxies without more massive neighbours are associated with O vi absorption systems. Indeed, Johnson et al. (2013) showed that a fraction of O vi absorbers arise in low-mass galaxy groups with star-forming members that are all <0.1 L*. Moreover, Johnson et al. (2017) showed that star-forming field dwarfs of <0.1 L* without massive neighbours account for |${\approx }20{{\ \rm per\ cent}}$| of O vi absorption systems with rest equivalent widths stronger than ≈0.1Å, and that this phase of their CGM likely dominates the baryon budget of these low-mass systems. More recently, Chen et al. (2018) and Zahedy et al. (2019) showed that O vi absorption systems with logN(O vi) > 13.5 are common in the CGM of luminous, quiescent galaxies.

Two of the z > 0.68 absorption systems are coincident with multiple galaxies in our survey (a galaxy pair at z ≈ 0.868 and the galaxy triplet group at z ≈ 1.089), whilst around 3 we detect only a single galaxy given our survey limits. We find a more massive (M_halo ≈ 10^12.6 M_⊙) group system for which we detect no O vi absorption at z = 1.026. In this first QSAGE field, given the relatively small volume covered, we have only a single galaxy with which to probe the higher mass galaxy environment beyond M_⋆ ≳ 10^10.5 M_⋆. At M_⋆ ≈ 10^10.9 M_⋆, this galaxy and its associated group shows no detectable O vi in the sightline data.

Incorporating low redshift results, Tumlinson et al. (2011) find 50 per cent of galaxies at M_⋆ ≳ 10^10.5 M_⊙ show associated O vi absorption, compared to 100 per cent at 10^9.5 M_⊙ ≲ M_⋆ ≲ 10^10.5 M_⊙. At the upper end of the latter range, Muzahid et al. (2015) report on an extremely strong O vi absorption system associated with a M_halo ≈ 10^12.5 M_⊙ star-forming galaxy. Stocke et al. (2014) report the alignment of warm O vi absorbers with several galaxy groups at z ≲ 0.2, with group velocity dispersions of σ_v ≈ 100–600 km s⁻¹. This equates approximately to a halo mass range of M_halo ≈ 10^12.5–14.5 M_⊙, suggesting groups of all masses, up to low-mass clusters have the potential to exhibit warm gas capable of being detected in O vi absorption. Pointon et al. (2017) find approximately a third of low-mass galaxy groups (≈2–5 members) in their sample exhibit O vi absorption, whilst Burchett et al. (2018) found no O vi absorption within cluster haloes at M_halo ≈ 10^14–14.5 M_⊙. The caveat remains, however, that it is not often clear whether the halo gas probed by a given sightline is reflecting the nature of the properties of a sub-halo or the overall group halo (e.g. Stocke et al. 2017), whilst star-forming winds remain a potentially significant influence on the presence of warm gas at small scales (e.g. Tripp et al. 2011; Tumlinson et al. 2011)

If some fraction of O vi absorbing gas is simply diffuse halo gas, collisionally ionized at approximately the virial temperature of the host dark matter halo, then we may expect to see some sort of peak in the column density distribution at some halo mass. Indeed, this is predicted by simulations such as shown by Mulchaey et al. (1996), Oppenheimer et al. (2016) and Nelson et al. (2018). Following this reasoning, we show in Fig. 18 the inferred halo masses for the z > 0.68 sample assuming the absorbers to be associated with the most massive proximate galaxy (i.e. within b = 2R_vir; dark red hexagons). We find that the N_O vi ≳ 10¹⁴ cm⁻² O vi absorption is generally found within the proximity of a galaxy corresponding to a 10^11.5 M_⊙ ≲ M_halo ≲ 10¹² M_⊙ halo mass. Upper limits on the O vi column density adjacent to other galaxies along the sightline (within 0.68 < z < 1.4) preferentially show upper-limit constraints indicating lower levels of the O vi ion.

Figure 18.

Column densities of O vi absorbers versus their estimated dark matter halo mass for the z > 0.68 galaxy sample based on identified associations from Fig. 16. Hexagons denote associated galaxies within b = 2R_vir, whilst triangles denote associated galaxy–absorber pairs at impact parameters of 2R_vir < b < 5R_vir. The dashed curve shows the predicted O vi ionization fraction for a given halo mass, assuming the halo gas being probed to be close to the virial temperature of the halo. Green diamonds and blue plus symbols show the results of Werk et al. (2013) and Mathes et al. (2014), respectively.

Open in new tabDownload slide

The dashed curve shows the predicted O vi ionization fraction assuming that the halo gas being probed is at temperatures of |${\approx }70{{\ \rm per\ cent}}$| of the virial temperature (i.e. accounting for predicted halo temperature profiles e.g. Komatsu & Seljak 2001). This provides a simple physical picture in which the detected gas represents a diffuse warm halo component (e.g. Mathews & Prochaska 2017). For comparison to our own data, we also show the low-redshift results of Werk et al. (2013, green diamonds) and Mathes et al. (2014, blue plus symbols). The Werk et al. (2013) sample is dominated by strong O vi detections at 10^11.5 M_⊙ ≲ M_halo ≲ 10¹² M_⊙, whilst the scatter increases at M_halo ≳ 10¹² M_⊙. Similarly, the strong absorbers (N(O vi) ≈ 10^14.5 cm⁻²) detected in the Mathes et al. (2014) sample are coincident with galaxies with estimated halo masses of M_halo ≲ 10¹² M_⊙. Similar to our own data and that of Werk et al. (2013), there is also a lot of scatter with a number of low-column density systems (or upper limits) coincident with comparable galaxy halo masses. The evidence is not clear-cut then, although the scatter may be the effect of a patchy/clumpy halo medium (Lopez et al. 2018), or AGN heating of the gas (Oppenheimer et al. 2018b).

The EAGLE hydrodynamical simulations (Crain et al. 2015; Schaye et al. 2015) predict that the O vi column density peaks (N_O vi ≈ 10^14.2 cm⁻²) for galaxies in haloes of masses ≈10^11.7–12.4 M_⊙, with the gas within lower mass haloes too cool, and the gas within higher mass haloes too hot, to lead to significant O vi absorption in quasar sightlines (albeit at z = 0.2; Oppenheimer et al. 2016). Similarly, the IllustrisTNG predicts, at z = 0, that O vi should be predominantly found in diffuse T ≈ 10^5.6 ± 0.2 K haloes of half-mass radii ≈10^2.1 ± 0.2 kpc around M_⋆ ≈ 10^{10.2 ± 0.3} M_⊙ galaxies (Nelson et al. 2018). Alternatively, the O vi gas may trace low-pressure gas photoionized by the UV background (e.g. Stern et al. 2016, 2018). In this case clouds are hierarchically embedded in the CGM structure, allowing for a patchy medium and potentially explaining the large scatter in column densities that we measure.

Finally, we note that the halo masses that we find associated with the O vi absorbers are comparable to the mass of the Local Group. Indeed, observations of the Milky Way’s halo have shown detection of O vi absorption (N_O vi ≳ 10^13.6 cm⁻²) in |${\approx }70{{\ \rm per\ cent}}$| of sightlines through the halo (e.g. Sembach et al. 2003; Wakker et al. 2003). The halo mass of the Milky Way is estimated to be M_halo ≈ 10¹² M_⊙ (e.g. Callingham et al. 2019), whilst the mass of the Local Group as a whole is estimated to be M_halo ≈ 2 × 10¹² M_⊙ (e.g. Courteau & van den Bergh 1999).

5 CONCLUSIONS

We have presented methods and results from the first field in the QSAGE survey – a blind HST/WFC3 grism survey of galaxies in the region of bright z > 1.2 quasars with archival HST/STIS and COS spectra. Our key results are as follows:

• We find O vi up to impact parameters of b ≈ 350 kpc from the nearest detected galaxy at 0.68 < z < 1.42. Column densities of absorbers over the impact parameters probed (100 < b < 400 kpc) show a large scatter, corresponding to covering fractions of c(O vi) ≲ 0.5 (for N(O vi) > 10¹⁴ cm⁻²).

• Whilst all five of the detected z ≈ 1 O vi absorbers are found to lie within b ≈ 400 kpc of a star-forming galaxy, we also find comparably star-forming galaxies within the same range in impact parameter with no detected O vi absorption. Taking a limit in impact parameter of b = 2R_vir, we find 50 per cent of the sample relate to upper limits on the sightline O vi column density of N_O vi ≲ 10^13.9 cm⁻².

• We identify a low-mass galaxy group at z = 1.08 coincident in redshift with significant O vi absorption in the quasar sightline, potentially probing the intra-group medium. The group consists of three confirmed members and estimate a group halo mass of M_halo ≈ 10^11.8 M_⊙. We find several further galaxy overdensities close to the sightline, with estimated halo masses of up to M_halo ≈ 10^12.8 M_⊙. None are coincident with detected O vi absorption to a detection limit of |$N_{\rm O\,{\small VI}}\lesssim 13.9\, {\rm cm}^{-2}$|⁠.

• Estimating the host halo masses of N_O vi ≳ 10¹⁴ cm⁻² absorbers suggests the majority of such absorbers are found in the proximity of haloes of mass M_halo ≈ 10^11.8 M_⊙, consistent with diffuse gas at the virial temperature of such haloes. Significantly, we find a higher covering fraction of N_O vi ≥ 10¹⁴ cm⁻² absorbers around higher mass star-forming galaxies (at ≈2σ) at impact parameters of ≲4R_vir.

This first of 12 fields from our HST WFC3 Large Programme, QSAGE, acts as a proof of concept of what the full survey can deliver. Whilst we have focused on tracing the properties of the galaxy population around O vi absorbers here, these data will provide a comprehensive basis for studies of the CGM across a range of absorption species, allowing insights into a broad range of phases of material within the CGM.

SUPPORTING INFORMATION

Figure S1 Example objects from the HST WFC3 grism data. In each case the top-left panel shows the median spectrum (orange) and contamination-masked mean spectrum (dark red). The lower-left panel in each case shows the corresponding spectra taken at each roll angle, whilst the right-hand panel in each case shows a thumbnail taken from the WFC3 F140W stacked image.

Please note: Oxford University Press is not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

ACKNOWLEDGEMENTS

We first thank S. Johnson, G. Kacprzak, and the anonymous referee for their input on this work. RMB, MF, RGB, and SLM acknowledge the Science and Technology Facilities Council (STFC) through grants ST/P000541/1 and ST/L00075X/1 for support. This project has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation programme (grant agreement no. 757535). TMT, JNB, and JXP received financial support for this research through NASA Grant HST-GO-11741 from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA Contract NAS5-26555. This work is based in part on observations obtained with MegaPrime/MegaCam, a joint project of CFHT and CEA/IRFU (Commisariat à l'énergie et atomique et aux énergies alternative/Institut de recherche de lois fondamentales de l'Univers), at the Canada–France–Hawaii Telescope (CFHT) which is operated by the National Research Council (NRC) of Canada, the Centre National de la Recherche Scientifique (CNRS) of France, and the University of Hawaii. The data used in this paper are available via the Mikulski Archive for Space Telescopes (HST COS, STIS, and WFC3 data), the Canadian Astronomy Data Centre (CFHTLS data), and the ESO archive (VLT/MUSE data). The WHT/ACAM imaging data are available at http://astro.dur.ac.uk/qsage/. We have produced data products using le phare and gazpar (located at the Laboratoire d’Astrophysique de Marseille) and we thank O. Ilbert for support in using these. This research made use of astropy, a community-developed core python package for Astronomy (Astropy Collaboration 2018). In the course of this work, we made use of A. Rohatgi’s webplotdigitizer (http://arohatgi.info/WebPlotDigitizer/).

Footnotes

REFERENCES