Two more, bright, z > 6 quasars from VST ATLAS and WISE

ABSTRACT

1 INTRODUCTION

High-redshift quasars are key probes of the Universe within the first billion years to redshift z ∼ 6. First, they provide observational constraints on the evolution of the quasar luminosity and black hole mass functions to the highest redshifts (Willott et al. 2005, 2010a,b, Jiang et al. 2009, 2016). The existence of a high spatial density of luminous quasars at z > 6 would be a challenge to the standard cosmological model, which is limited in its capacity to produce large black hole masses, assuming Gaussian initial conditions (e.g. Rosas-Guevara et al. 2016).

Secondly, analysis of damped Ly α systems (DLAs) and Lyman limit systems along quasar lines of sight at intermediate redshifts (z ≈ 3) provide invaluable constraints on galaxy evolution by probing conditions within the interstellar medium (ISM) and circumgalactic medium (CGM; e.g. Fumagalli, O'Meara & Prochaska 2016; Lehner et al. 2016; Wotta et al. 2016 and references therein), via absorber metallicities, covering factors and kinematics. Some progress has been made in extending these analyses to higher redshifts (e.g. Rafelski et al. 2014; Crighton et al. 2015; Wang et al. 2015).

Finally, high-redshift quasars cast light on conditions in the intergalactic medium (IGM) during the epoch of reionization. Studies have been made of the Ly α emission-line profiles of these objects, their ionized near zones and the Gunn–Peterson troughs in their spectra (e.g. Becker et al. 2001; Fan et al. 2006; Carilli et al. 2010; Venemans et al. 2015b; Barnett et al. 2017) producing a general consensus view of a patchy reionization that ended around z = 5–6.

There has also been much debate about the relative contributions of active galactic nucleus (AGN) activity and star formation to the reionization of the IGM, and the luminosity function of high-redshift quasars is a crucial variable in this context. Whilst, it has been shown (McLeod et al. 2015; 2016) that star formation in high-redshift galaxies can provide the bulk of the ionizing photons necessary for reionization, the actual fractional contributions from AGN and star formation are still unknown, with Madau & Haardt (2015) presenting a plausible scenario in which AGN activity provides the majority of the ionizing photons. The major extra AGN contribution in their work is due to the increased slope of the faint end of the quasar luminosity function found in the CANDELS survey by Giallongo et al. (2015). However, using the same CANDELS data, Parsa, Dunlop & McLure (2018) report no evidence for a significant AGN contribution. Although here we shall be addressing the bright end only, the importance of determining the quasar luminosity function at high redshift is clear not only for the source of reionizing photons but also for models of black hole seeding, growth, feedback, and evolution. Clearly, bright, high-redshift, quasars are also easier to follow-up for the above ISM, CGM, and IGM absorption studies.

The techniques employed to detect high-redshift quasars are based on essentially the same Lyman-dropout technique as has been applied for many years, where objects with a very red colour, often implied by a non-detection in a bluer band, are selected from large area sky surveys. One of the first applications of this technique was by Shanks, Fong & Boyle (1983) who looked for U and B dropouts relative to their R-band magnitudes to find a quasar at z = 3.63. This method has evolved more recently to redder dropout bands in order to probe higher redshifts. However, due to the decreasing spatial density of quasars at increasingly high redshifts, contaminant objects such as cool L and T dwarf stars become more and more problematic. Due to this, a variety of new methods have recently been developed for cleaning photometrically selected samples before spectroscopic follow-up observations are made to confirm the nature of the sources.

The first study to be sensitive to quasars above z = 6 was reported by Fan, Narayanan & Lupton (2001), Fan et al. (2006), and Jiang et al. (2016) using data from the Sloan Digital Sky Survey (SDSS). They looked for i-band dropouts relative to their z-band magnitudes and then obtained follow-up J-band imaging, selecting quasars as objects with blue z–J colour to exclude cool dwarf stars. The same approach has also been applied to other survey data e.g. by Willott et al. (2010a), Venemans et al. (2015a), and Wang et al. (2017). More recently, large area Y-band surveys have been used instead of follow-up J-band photometry by Bañados et al. (2014, 2016) and Reed et al. (2015). A Bayesian statistical approach to candidate selection has also been developed by Mortlock et al. (2012) and Matsuoka et al. (2016). Most recently, the dropout method has been applied to the z band by Venemans et al. (2013, 2015b) to begin selecting quasars at z ≳ 6.5, with the current record holder being a quasar at z = 7.54 discovered by Bañados et al. (2018).

In this paper, we follow (Carnall et al. 2015, hereafter C15) who found two z > 6 quasars using the combination of the new Very Large Telescope (VLT) Survey Telescope ATLAS (VST ATLAS; Shanks et al. 2015) and Wide-field Infrared Survey Explorer (WISE; Wright, Eisenhardt & Mainzer 2010) surveys, by applying a similar method to a greater area of VST ATLAS imaging that yielded the discovery of two more bright z > 6 quasars.

In Section 2, we review the survey data, and in Section 3 we review our quasar selection technique. In Section 4, we report on follow-up spectroscopic observations of two further high-redshift quasar candidates, which we confirm to be two new quasars at z = 6.32 and 6.07. In Section 5, we make an approximate estimate of the bright end of the quasar luminosity function from our new quasars plus the others found within our search area. We present our conclusions in Section 7.

All VST ATLAS magnitudes are given on the AB system, all other magnitudes are on the Vega system. All cosmological calculations assume the density parameters Ω_m = 0.3, |$\Omega _\Lambda =0.7$|⁠, and Hubble Constant h = 0.7, with h measured in units of 100 km s⁻¹ Mpc⁻¹.

2 SURVEY DATA

2.1 VST ATLAS

The VST is a 2.6 m wide-field survey telescope with a 1° × 1° field of view. The OmegaCAM camera (Kuijken et al. 2004) consists of 32 CCDs with 2 k × 4 k pixels, resulting in 16 k × 16 k image with a pixel scale of 0.21 arcsec. The VST ATLAS is a nearly completed photometric survey that will cover ≈4700 deg² of the southern extragalactic sky with coverage in ugriz bands. The survey takes two subexposures of 2 × 60 s (u), 2 × 50 s (g), 2 × 45 s (r), 2 × 45 s (i), 2 × 45 s (z), per 1 degree field with a small dither in X and Y to cover most interchip gaps. The subexposures are then reduced and stacked by the Cambridge Astronomy Survey Unit (CASU). Their pipeline outputs catalogues cut at ≈5σ and provide fixed aperture fluxes and morphological classifications of detected objects (see Shanks et al. 2015 for more details). Here, for stellar photometry, we use a 1 arcsec radius aperture (i.e. aper3). ATLAS photometry is calibrated using the APASS nightly zero-points (NIGHTZPT in the FITS headers, see Shanks et al. 2015). The star–galaxy classification is that supplied as default in the CASU catalogues and discussed in detail by González-Solares et al. (2008).

VST ATLAS provides ugriz photometry to similar depths as SDSS for galaxies and up to 0.7 mag fainter (e.g. in the z band) for stars, mainly due to its ≈40 per cent better seeing (e.g. 0.81 arcsec versus 1.18 arcsec in i; Shanks et al. 2015). In terms of VST-ATLAS progress at the time of this study, the survey area had increased to 3577 deg² from the 2060 deg² available to C15, partly by only using riz band-merged CASU catalogues. This maximizes the searchable area because the u and g sky coverage is less than for riz. Further, the CASU data base has higher sky coverage than the public accessible Edinburgh Wide Field Astronomy Unit (WFAU) archive. The average 5σ limits of the ATLAS riz catalogue for point sources as measured in a 1 arcsec radius aperture are r < 22.7, i < 22.0, and z < 20.9 mag. See Shanks (2016) for a comparison of ATLAS parameters with various other surveys.

The main issues with the CASU catalogues are that they are currently CCD based rather than tile based (as at the WFAU). This leads to increased incompleteness in the matching at the edges that increases the numbers of spurious r- and i-band dropouts.

2.2 WISE

The WISE (Wright et al. 2010) survey covers the mid-IR 3.4 (W1), 4.6 (W2), 12 (W3). and 22 (W4) micron bands. The advantage of WISE is its 100 per cent coverage of ATLAS at the present time and the excellent matching of both the W1 and W2 bands’ depths to VST ATLAS. The approximate 5σ limits for AllWISE¹ point sources are W1 = 16.90 and W2 = 15.95 mag in the Vega system. The W1 and W2 bands have point spread functions (PSFs) of 6.1 and 6.4 arcsec, respectively, compared with ≈0.85 arcsec in the VST ATLAS riz bands. Astrometric tests² between WISE and USNO CCD Astrograph Catalog (UCAC3) positions shows <0.5 arcsec rms offset between the two catalogues to the W1 limit. ATLAS optical photometry was matched to the publicly available ALLWISE Catalogue using a 3 arcsec matching radius. For the sky density of WISE sources at |b| > 30°, we calculate that ≈4 per cent of candidates identified in WISE will have a blended WISE source within 3 arcsec. In looking for rare, high-redshift, quasars, WISE blends (and other artefacts) will therefore have to be eliminated by visual inspection.

3 PHOTOMETRIC QUASAR SELECTION

C15 used ATLAS i–z colour to select dropout candidates, then the z − W2: W1 − W2 colour plane to discriminate between dwarf stars and high-redshift quasars, as shown in Fig. 1. W1–W2 colour is the main discriminator against L dwarfs and z–W2 is used to discriminate between T dwarfs and quasars, providing good separation out to z = 6.3 before the quasar colour track crosses the T dwarf locus. Application of this technique to the first 2060 deg² of VST ATLAS imaging resulted in the discovery of two z > 6 quasars, ATLAS J029.9915−36.5658 at z = 6.02 ± 0.03 and ATLAS J025.6821−33.4627 at z = 6.31 ± 0.03 and the re-discovery of VIKING/KiDS J0328−3253 at z = 5.86 ± 0.03.

Figure 1.

The z–w2: W1–W2 colour plane with the positions of four ATLAS quasars marked. Also marked are the positions of L and T dwarf stars and other quasars from the literature. The dashed-line box shows the colour selection of C15. The dotted line extends this selection bluewards in W1–W2 to W1–W2 > 0.3, including two previously discovered quasars from SDSS and with the aim of testing the completeness of the C15 selection. The black line is predicted quasar redshift track marked in 0.1 redshift intervals from z = 5.8 to 6.6.

Open in new tabDownload slide

Following C15, we first produced a list of high-redshift (z > 5.7) quasar candidates by applying the following selection criteria to the CASU band-merged catalogues.

(i) Objects must have 18 < z_AB < 20 mag, measured in a 1 arcsec aperture (labelled as aper3 in the CASU and WFAU catalogues). Chehade (2016) shows that this aperture provides the best balance between S/N and accuracy of the aperture corrections for point sources. The z_AB > 18 cut reduces the number of candidates for visual inspection in a range where the star/quasar number ratio is expected to increase.

(ii) Objects must be identified as a point source by their curve of growth (see González-Solares et al. 2008). Morphology is stored under the flag Classification where −1 is for point sources and +1 is for extended sources. We disregard morphological classification in the i band, since at low S/N, the Classification is unreliable. Chehade et al. (2016) tests the CASU morphological classification and finds it to be |${\approx }\ 90\hbox{ per cent}$| complete for the g and r bands. We examine the z band specifically in Section 5.2.

(iii) Objects must either be undetected in the i band, or if detected must have i–z > 2.2. An additional selection was also carried out for which this limit was relaxed to i–z > 1.8, to fill observing gaps between higher priority targets. Objects must also be undetected in the u, g, and r bands.

(iv) Objects must have a corresponding source in the ALLWISE catalogue within a 3 arcsec radius with S/N > 3 in both the W1 (≲ 17.9 mag) and W2 (≲ 17.0 mag) bands.

(v) Objects must have colours in the range 0.55 < W1–W2 < 1.2 and 3.2 < z_AB − W2 < 4.5 (see Fig. 1).

This left ≈130 candidates for visual inspection. The candidates were largely obviously misidentified as dropouts due to defects in the band-merging process, or in areas where the i-band seeing was poor (>1.2 arcsec), meaning the dropout status could not be confirmed due to a lower limiting colour. We also limited our selection regions to the doubly exposed regions of the VST ATLAS stack (see Shanks et al. 2015) for this reason. This reduced the searchable area per tile from ∼1 deg² to ∼0.8 deg².

As well as our primary selection, secondary targets were identified in the bluer, 0.3 < W1–W2 < 0.55 colour region and tertiary targets in the 0.3 < W1–W2 < 0.55 and 1.8 < i–z < 2.2 region of colour space. The bluer WISE cut was motivated by the presence of two previously discovered SDSS quasars of this colour that our original selection criteria of C15 miss. Whilst the region lies closer to the L dwarf locus, it provided a test of the incompleteness of our more conservative W1–W2 selection.

In total, we identified two candidates of highest priority within our primary selection region. These candidates were identified as highest priority due to their high signal-to-noise ratio (SNR) detections. A further seven primary candidates in areas of higher background noise and/or near CCD edges were also identified, along with 10 secondary and five tertiary candidates in the VST ATLAS footprint, as observable from La Silla in our 2015 May New Technology Telescope (NTT) EFOSC2 observing run (see Section 4.2 below).

4 FOLLOW-UP SPECTROSCOPY

4.1 Keck LRIS spectroscopy

4.1.1 VST-ATLAS J158.6938−14.4211

On 2015 April 19, we observed one of our two highest priority candidates with the Low Resolution Imaging Spectrometer (LRIS; Oke et al. 1995; Rockosi, Stover & Kibrick 2010) on the Keck I telescope. The observers were X. Prochaska (P.I.), J. Hennawi, and C. Leibler. The target was VST-ATLAS J158.6938−14.4211, which was observed for 600 s through a 1 arcsec wide slit. We used the 400/8500 grating in the red arm giving ≈6.9 Å resolution and a dispersion of 1.16 Å pixel⁻¹ and the 400/3400 grism in the blue arm giving ≈7 Å resolution and a dispersion of 1.09 Å pixel⁻¹. The LRIS spectrum of J158.6938−14.4211 was reduced with the LowRedux pipeline.³ First, the pipeline processes the calibrations (bias, flat-fields, and sky flats) and computes the wavelength solution using arc lamps. Next, these calibrations are applied to the raw science frames, and a 1D spectrum is extracted from the 2D frames. The spectra are then flux calibrated using observations of a spectrophotometric standard star. The LRIS spectrum showed strong Ly α emission, cut off sharply towards blue wavelengths due to a strong Gunn–Peterson hydrogen absorption trough, confirming this as a quasar with an approximate redshift of z = 6.05 ± 0.03 (but see Section 4.3 and Table 2). Also detected in emission are Ly β, NV(1240 Å), O i(1304 Å), and Si iv (1398 Å). Other details of this quasar, including i, z, W1, and W2 magnitudes are given in Table 1.

4.2 ESO NTT/EFOSC2 spectroscopy

4.2.1 VST-ATLAS J332.8017−32.1036

On 2015 May 29, we observed our second highest priority candidate with the European Southern Observatory's Faint Object Spectrograph and Camera 2 (EFOSC2; Buzzoni et al. 1984) on the 3.58 m ESO NTT. The observers were B. Chehade and T. Shanks and the target was VST-ATLAS J332.8017−32.1036. The target was observed for 600 s through grism no. 2 (100 lines mm⁻¹) and a 1 arcsec wide slit giving a resolution of 49.6 Å full width at half-maximum (FWHM). The pixels were binned 2 × 2 in order to reduce readout time and noise. The resulting spatial scale was 0.25arcsec pixel⁻¹ and the dispersion was 13.2 Å pixel⁻¹. All stages of the data reduction were performed using standard iraf routines. The wavelength calibration was performed against helium–argon (HeAr) arc lamp spectra. The spectra cover the wavelength range of ≈5100–11 000 Å. Observations of the spectrophotometric standard star LTT3864 (Hamuy et al. 1992, 1994) were used for absolute flux calibration.

As with VST-ATLAS J158.6938−14.4211, strong Ly α emission was seen redwards of a Gunn–Peterson absorption trough, again confirming that this object was correctly identified as a high-redshift quasar with approximate redshift of z = 6.37 ± 0.03 (but see Section 4.3 and Table 2). The NV (1240 Å) emission line might also be detected at 9140 Å. There is also a hint of Ly β emission at 6520 Å. Further discussion of these spectral features is postponed to Section 4.3, in which we report higher resolution X-shooter spectroscopy. Magnitudes for this quasar are also reported in Table 1.

4.2.2 PSO J340.2041−18.6621

A third high-redshift quasar was observed with NTT EFOSC-2 on the night of 2015 June 01. Using the same VST ATLAS+WISE combined selection technique, we re-discovered the z = 5.98 quasar, PSO J340.2041−18.6621, independently found by Bañados et al. (2014) using the PanSTARRS survey. The spectrum confirmed the object as a quasar although we measured a slightly lower redshift of z = 5.98 ± 0.03, based on the Ly α emission line, than that measured by Bañados et al. (2014) of z = 6.03. We include this quasar in Table 1 for completeness and its NTT spectrum is shown in Fig. 2.

Figure 2.

NTT/EFOSC2 spectrum for the re-discovered quasar PSO J340.2041−18.6621 of Bañados et al. (2014). The positions of Ly α, Si iv, and C iv emission lines at z = 5.98 are shown.

Open in new tabDownload slide

4.3 VLT X-shooter spectroscopy

To increase the wavelength coverage for the spectroscopically confirmed quasars from this work and C15, we observed the four VST ATLAS quasars with the medium-resolution spectrograph X-shooter (Vernet et al. 2011) on the Cassegrain focus of the 8.2 m VLT Kueyen (UT2) on the dates shown in Table 2.

The three arms of the X-shooter instrument are the UVB (3000–5595 Å), the VIS (5595–10 240 Å), and the NIR (10 240–24 800 Å). Given the redshift of our targets, we disregard the UVB arm data in our present analysis. The slit width for both the VIS and NIR arms was 0.9 arcsec and slit length was 11 arcsec. Pixel scales are ∼0.15 Å pixel⁻¹ and ∼0.2 Å pixel⁻¹ for the VIS and NIR arms, respectively. The resolution of the observations were R ≈ 7400 in the VIS arm and R ≈ 5400 in the NIR arm. We observed each of the four targets for 1 h each, ≈45 min of which was on-sky exposure.

To reduce the X-shooter data, we built a tailor made workflow that incorporates both standard ESO recipes but also iraf tasks to improve data quality. The ESO pipeline was used to perform the reduction steps up to the 2D rectified and merged spectrum. This includes bias subtraction, flat-fielding, modelling of the spectral orders, resampling, and rectifying the spectra. We ensured that the default pipeline parameters were producing optimal results and adjusted them in a few cases, namely cosmic ray removal and sky background subtraction. To extract the QSO 1D spectra from the 2D spectra, we used the iraf task apall. Finally, we applied a telluric correction to the NIR arm and merged the VIS and NIR arms to produce the spectra displayed in Fig. 3. Regions with transmission of less than ≈20 per cent were masked from the spectra.

Figure 3.

The VLT X-shooter spectra for our four ATLAS quasars (see Table 2 for observational details). The spectra were flux calibrated to the observed z-band magnitudes from VST ATLAS and corrected for telluric absorption as described in Section 4.3. The positions of the Ly α, Si iv, C iv, and Mg ii emission lines are marked at the redshifts given in Table 2. We note that higher ionization lines such as C iv are frequently found to be blueshifted with respect to lower ionization lines like Ly α and this is seen to be the case for all four quasars.

Open in new tabDownload slide

We show the positions of the Ly α, Si iv, C iv, and Mg ii emission lines at the redshifts in Table 2; these are based on the X-shooter redshifts estimated from the position of Ly α indicated in Fig. 4.⁴ Note the slight differences in these redshifts from those measured in the discovery spectra (see Table 1 and table 1 of C15). The higher ionization lines such as C iv are frequently seen to be blueshifted with respect to the lower ionization lines (e.g. Coatman et al. 2016) and this appears to be the case for these four quasars. However, it is not ruled out that there could be some uncertainty in our redshifts due to variation in the Ly α emission-line profile. These issues are further discussed in Section 6 where redshifts are estimated from the C iv and Mg ii emission lines. The C iii] emission lines lie inside the range affected by atmospheric absorption in the X-shooter NIR data and so are unavailable to further test our redshifts. We note that quasar J332.8017−32.1036 may also show broad absorption features below the Si iv and C iv emission lines i.e. it may be a broad absorption line quasar (BALQ). We used these X-shooter spectra to measure M_1450 Å absolute magnitudes, as reported in Table 3. We also make a preliminary calculation of black hole masses in Section 6 but defer the detailed analysis and the fitting of Ly α emission and associated absorption line profiles (see Fig. 4) to future work (Diener et al., in preparation).

Figure 4.

The VLT X-shooter spectra for our four ATLAS quasars now ‘zoomed in’ to emphasize the H i absorption structure around the Ly α emission line. The positions of the Ly α emission lines as estimated from the maxima of the line profiles are indicated by the red arrows.

Open in new tabDownload slide

4.4 Other candidates and effect of relaxing the W1–W2 limit

Our original selection criteria as listed in Section 3 have produced the discoveries of four z > 6 quasars from the top of the ATLAS+WISE priority list. However, none of the other candidates that were also observed as part of our four night NTT run between 2015 May 29 and June 01, including our sample selected to be bluer in W1–W2 and i–z, displayed spectral features consistent with high-redshift quasars. These candidates had been marked as lower priority due to their lower signal-to-noise detections meaning a larger probability of these objects having scattered into our selection region from the stellar locus (see Fig. 1). We subsequently re-measured the i-band photometry for the i-band dropouts based in the z-band detections using the imcore routines supplied by CASU. The non-quasar targets were generally positively detected at below the 5σ limit of the ATLAS catalogue i.e. they were not i-dropouts.

5 CONSISTENCY WITH PREVIOUS QUASAR LUMINOSITY FUNCTION ESTIMATES

In this section, we will use our discovery of four (plus two re-discovered) z > 6 quasars from C15 and this work to check consistency with previous studies of the quasar luminosity function at bright absolute magnitudes. We estimate the completeness of our catalogues with respect to their depth and our ability to identify point sources. Next, we test the completeness of our colour selections from Section 3. Finally, we check our completeness estimates against the PanSTARRS survey of Bañados et al. (2016) before comparing our quasar space density to that expected from previous estimates of the quasar luminosity function.

5.1 Catalogue depths

To determine the completeness of VST ATLAS catalogues as a function of magnitude, we need to compare the number of sources to the expected number of sources. We take advantage of the overlap between VST ATLAS and the PanSTARRS Medium Deep Survey (MDS; Chambers et al. 2016). The MD02 field of the MDS survey is covered by the VST ATLAS survey. The MD02 centre is RA 03:32:24, Dec. −28:08:00 with a radius of ≈1|${^{\circ}_{.}}$|5. The approximate 5σ limits for this field in riz are 25.4, 25.8, and 25.3 mag (cf. 22.7, 22.0, and 20.9 mag for VST ATLAS).

We wish to compare VST ATLAS exposures that are representative of the survey. To do this, we select a single VSTATLAS tile in each band with seeing similar to median seeing for the survey (0.91, 0.69, 0.85 arcsec for the riz bands, respectively).

First, we assume that the MD02 field is complete and uniform to the depth that we test to [24th, 23rd, and 22nd magnitude (AB) for the riz bands, respectively]. Given the dither pattern of the VST ATLAS survey, different areas in the stack will either have two, one, or no exposures. For each detection brighter than the faint limit in the MDS catalogue, we compare to the VST ATLAS images, masking the MDS data according to the depth of VST coverage.

We compared aper3 magnitudes from VST ATLAS to the PanSTARRS PSF magnitudes and found offsets of +0.06 ± 0.03, −0.19 ± 0.02, and +0.08 ± 0.03 mag in the riz bands. After correcting for these differences, we can check VST ATLAS catalogue completeness.

In Fig. 5, we show the completeness of VST ATLAS catalogues compared to the PanSTARRS MDS as a function of magnitude. We find that the catalogue is 99.6 per cent complete to z_AB = 20, assuming that PanSTARRS is 100 per cent complete.

Figure 5.

Completeness of the r, i, and z bands (red, pink, and black, respectively) as a function of magnitude. Fractions are derived by comparing catalogue detections in VST ATLAS with detections in the PanSTARRS MD02 field. The vertical dashed line corresponds to the 5σ detection limit for the VST ATLAS z band. The catalogue is 99.6 per cent complete down to z_AB = 20.0 mag.

Open in new tabDownload slide

5.2 Point-source completeness

To remove contamination of our colour selection from local galaxies, we select only point sources as determined by their single band (z) morphological classification in the CASU catalogues. To test the completeness of the CASU morphological classification, we modelled the difference between Petrosian and aper1 (0.5 arcsec radius) aperture magnitudes in the z band for galaxies. We identified galaxies as all objects being 0.4 mag brighter in Petrosian than in the aper1 magnitude. We fit the distribution with a Gaussian model, allowing the centre, width, and height to vary. We did this for all targets within 4 mag ranges; 18 < z < 18.5, 18.5 < z < 19, 19 < z < 19.5, and 19.5 < z < 20 mag. We excluded targets identified as noise (by their CASU Classification), targets in the singly exposed region and targets in regions with poorly fit sky background and de-blended photometry.

Fig. 6 shows the performance of the morphological separation of stars and galaxies in our faintest 19.5 < z_AB < 20.0 mag range. Subtracting the modelled distribution of galaxies in the Petrosian-aper1 plane to the distribution of all objects, we may estimate the number of stars not identified as point sources by CASU. We find the predicted extra number of stars in the above four bins (brightest to faintest) of 416, 837, 1277, and 704, implying 94, 90, 87, and 93 per cent point-source completeness, respectively. Increasing the aperture size to aper2 (0.7 arcsec radius) yields similar results for z < 19 but the separation in the Petrosian-aper2 plane is less well defined so poorly fits the distribution of galaxies. Including these ‘missed’ stars to those identified by CASU, we find an increase in the number of stars of 10 per cent. Based on this analysis, we estimate that our point-source completeness is 91 per cent. This number agrees well with the estimate of Chehade et al. (2016) where stars were colour selected using multiple VST ATLAS and WISE bands.

Figure 6.

We show the distribution of Petrosian-aper1 (0.5 arcsec radius aperture) magnitudes in the z band. The data consists of a single concatenation of 17 tiles. The images were taken on night 2014-06-17 with median 0.86 arcsec seeing (cf. 0.84 arcsec median for the survey). We plot the total histogram of objects (blue) between 19.5 < z_AB < 20.0 mag i.e. the faintest 0.5 mag of our target candidates. The fitted galaxy distribution is shown as a solid green line. The black and grey histograms show the point-like and extended sources (as identified by CASU), respectively.

Open in new tabDownload slide

5.3 Colour completeness

There are two methods used by other z ∼ 6 quasar searches to calculate the incompleteness of their samples. The first is to clone low-redshift quasar spectra and shift them to higher redshifts (see Willott et al. 2005; Findlay et al. 2012). The second method, which we use in this work, is to create model quasar spectra based on empirical models of quasar SEDs (Fan et al. 1999; McGreer et al. 2013). Both methods assume that the low-redshift UV properties of quasars, those redward of Ly α, do not evolve with redshift (Fan et al. 2004; Jiang et al. 2006).

To generate our model quasar spectra, we use the publicly available simqso⁵ code developed by McGreer et al. (2013). In this code, each quasar is assigned a power-law continuum with a break at 1100 Å. The slopes are drawn from normal distributions based on the results of Telfer et al. (2002). Emission lines with Gaussian shape are added to this continuum and the emission-line properties (wavelength, equivalent width, and FWHM) are also drawn from normal distributions. The emission-lines distributions are generated from composite BOSS spectra that have been stacked in different luminosity bins. The emission blueward of Ly α is based on the work by Worseck & Prochaska (2011). The model relies on the observed number densities of high column density systems (Songaila & Cowie 2010). A large number of sightlines are generated and the mean free paths are estimated by matching to the observations of Songaila & Cowie (2010). The column density distribution function is used to estimate the effective optical depth (τ_eff) that is checked for consistency with observations (Songaila 2004; Fan et al. 2006). Further details of the model may be found in McGreer et al. (2013). Note that we use the luminosity function parameters of Willott et al. (2010a). To accommodate WISE colour selection at higher redshifts, we follow Yang et al. (2016), adding three new spectral breaks at 5700, 10 850, and 22 300Å and including their assumed spectral slopes and dispersions. Emission-line parameters are derived from the composite quasar spectrum of Glikman, Helfand & White (2006).

We use simqso to generate 140 000 model quasars evenly distributed between 5.7 < z < 6.4 and −26.5 ≥ M_1450 Å ≥ −27.5. In Fig. 7, we show the resulting colour completeness using the photometric selections from Section 3. The completeness of the conservative selection rises sharply between redshifts 5.7 < z < 5.9 due to the i–z colour selection. When the colour selection is relaxed (shown by the dashed line), we see that the colour completeness is increased at lower redshifts. At higher redshifts (z ≳ 6.05), the completeness falls due to the faint limit of z_AB = 20 mag.

Figure 7.

Colour completeness as a function of redshift. The solid line shows the completeness when we apply the most conservative selection from Section 3 and the dashed line shows the completeness using the relaxed selection criteria.

Open in new tabDownload slide

It should be noted that the six quasars in Table 3 show a redshift distribution apparently skewed to higher redshifts than implied by Fig. 7. Indeed, the simqso quasar count model predicts that ∼2/3 of detected quasars should lie in the 5.7 < z < 6.0 range whereas only ∼1/3 of those in Table 3 do. Indeed, the quasars selected by Bañados et al. (2016) are more consistent with this prediction. However, lacking any other explanation, we note that this larger than expected z > 6 fraction for ATLAS at least remains within the bounds of statistical error.

5.4 Completeness estimates checked via PanSTARRS

Bañados et al. (2016) have searched for 5.7 ≤ z ≤ 6.7 quasars using the PanSTARRS1 (PS1) |$3\pi$| survey. This survey covers the ATLAS sky footprint except for the areas below Dec. < −30° in the SGC and NGC areas. Chambers et al. (2016) indicates that PS1 reaches ≈1.5 mag fainter than ATLAS in z, with a 5σ stellar limit of z_AB = 22.3 mag as opposed to z_AB = 20.87 mag for ATLAS. Bañados et al. (2016) quote a median limit of z = 22.0 mag for their PS1 survey. Five extra quasars were detected in the ATLAS overlap area with z_AB ≤ 20.0 mag; four of these had z < 6. One quasar was missed by ATLAS because it was classed as a galaxy (PSO J056.7168−16.4769), making an incompleteness of 1/7 or ≈14 per cent of the seven quasars that were detected by either survey in the ATLAS-PS1 overlap area. This is similar to our 10 per cent star–galaxy incompleteness estimate. One quasar was missed on account of its i − z colour being too blue (PSO J021.4213−25.8822) and two because either their W1 − W2 colour was too blue (PSO J183.2991−12.7676) or there was no WISE counterpart within 3 arcsec (CFHQS J1509−1749). This makes the colour incompleteness 3/7 or ≈43 per cent compared to our incompleteness estimates of between 10 and 60 per cent depending on redshift (see Fig. 7). We conclude that the extra quasars detected by Bañados et al. (2016) are not inconsistent with our star–galaxy separation and colour incompleteness estimates, although clearly there may still be further residual incompleteness associated with the PS1 survey.

5.5 Consistency with previous luminosity function studies

Following C15, we calculate M_1450 Å luminosities by scaling their SDSS composite template spectrum from Vanden Berk et al. (2001) and Songaila (2004) to the X-shooter spectra of each of our four ATLAS quasars in Fig. 3 using their observed Ly α redshift and z-band magnitude. The results are listed in Table 3, where we have also included the absolute magnitudes for our two re-discovered quasars quoted by Venemans et al. (2015a) and Bañados et al. (2014). We see that the six quasars occur in approximately one absolute magnitude bin centred on M_1450 Å ≈ −27 ± 0.5.

We have estimated the quasar space density at M_1450 Å ≈ −27 based on our observations. We assume a search area of 3119 deg², representing 80 per cent of the 3577 deg² area searched i.e. assuming 20 per cent area lost at CCD edges. From Fig. 7 and assuming the relaxed selection criteria of Section 3, we take the redshift range to be 5.7 < z < 6.4. We take the redshift-dependent colour completeness to be that given in Fig. 7, the magnitude completeness at z_AB < 20 to be 100 per cent and the star–galaxy separation completeness to be 90 per cent, implying we need to multiply our six detected quasars by a factor of 1.9 to account for these incompletenesses. Thus, our corrected quasar count is now 11.3 ± 4.6. We note that the total quasar count in the overlap area from Bañados et al. (2016) and ATLAS is 7 where ATLAS discovered 2, giving an ATLAS incompleteness factor of 3.5, compared to our modelled factor of 1.9. Applying this correction to the six ATLAS quasars gives a count of 21 ± 8.6 over 3119 deg². Although one out of two quasars available in Table 2 was missed by Bañados et al. (2016), we neglect this possible evidence for residual incompleteness in their sample on the grounds that the statistical error on the corrected count would be dominant. We conclude that the corrected number of quasars in our sample (11.3 ± 4.6) is in reasonable statistical agreement with those of Bañados et al. (2016) in a similar area (21 ± 8.6). As noted already, most of this difference is due to ATLAS missing 5.7 < z < 6 quasars.

With a 3119 deg² survey area, our assumed, Ω_m = 0.3, |$\Omega _\Lambda =0.7$|⁠, h = 0.7, cosmology indicates a volume of 17.4 Gpc³ between 5.7 < z < 6.4 and assuming a raw count of 6 ± 2.4 and an incompleteness corrected count of 11.3 ± 4.6 gives the quasar space density estimates shown in Fig. 8. Comparing these with the SDSS luminosity function of Jiang et al. (2016) and that of Willott et al. (2010a) at fainter magnitudes, we see there is good agreement between our observed number of quasars and the SDSS data that are bracketted by our raw and corrected space density estimates. Although our PS1 corrected space density would be ≈2× higher than the corrected value shown, its ≈40 per cent error means there is still no significant disagreement with previous luminosity functions.

Figure 8.

Approximate ATLAS quasar space densities at M_1450 Å = −27 compared to the SDSS luminosity function of Jiang et al. (2016) and the fainter result of Willott et al. (2010a) corrected to our assumed cosmology. ‘ATLAS raw’ refers to six 5.7 < z < 6.4 quasars detected in our 3119 deg² search and ‘ATLAS corrected’ refers to a corrected total of 11.3 quasars in the same volume.

Open in new tabDownload slide

As an alternative view of how our quasar counts compare to previous work, we use the simqso code of McGreer et al. (2013) with the luminosity function of Willott et al. (2010a) to predict the quasar count in the 15 < z_AB < 20 mag and 5.7 < z < 6.4 ranges. The predicted result is 8.1 ± 2.8 quasars in our 3119 deg² area. This compares to our raw, observed count of 6 ± 2.4 and our corrected count of 11.3 ± 4.6 quasars. We conclude that both these estimates are in good statistical agreement with the predicted value.

6 BLACK HOLE MASSES

Finally, we have made a preliminary estimate of the black hole masses powering these quasars. We therefore fitted Gaussian line profiles to the C iv and Mg ii lines in the NIR spectra from X-shooter using the irafsplot routine. Table 4 shows the measured FWHM and central wavelengths of the emission lines. As can be seen from Fig. 3, the S/N for the C iv line is generally higher than for Mg ii. The C iv lines can be seen to be blueshifted with respect to the redshifts based on the Ly α line as measured from the X-shooter spectra (see Fig. 4).

We estimated the monochromatic continuum luminosity, λL_1350 Å, for the C iv estimate from the quasar's z_AB band magnitudes and λL_3000 Å for the Mg ii estimate from their K_AB magnitudes. We chose this route rather than using the X-shooter spectral fluxes from Fig. 3 partly on the grounds that the overall reliability of the spectrophotometry is uncertain and partly because the continuum at rest 3000 Å lies at longer wavelengths than available from Fig. 3. If we assume that the spectrophotometry at 1350 Å is accurate, then our λL_1350 Å estimates in Table 4 may be too large by a factor of ≈2 due to Ly α emission-line contamination, implying that our resulting C ivM_BH estimates may be too large by ≈40–50 per cent. There may be a similar systematic issue for the Mg iiM_BH estimates.

These compare to the (1.24 ± 0.19) × 10¹⁰ M_⊙ result for the z = 6.3 quasar of Wu et al. (2015), the brightest quasar so far found at z > 6. This quasar has a monochromatic continuum luminosity of λL_3000 Å = (3.15 ± 0.47) × 10⁴⁷ erg s⁻¹ some ≈5× brighter than J158−14 that is our brightest quasar by this measure. J158−14 is ≈3× less massive. Our four z > 6 quasars are generally similar in luminosity and mass to the z = 7.05 quasar ULAS J1120+0641 (Mortlock et al. 2011) and to the z = 6.89 quasar J0100+2802 (De Rosa et al. 2014), both with M_BH ≈ 2 × 10⁹ M_⊙. The recently discovered z = 7.54 quasar (Bañados et al. 2018) has M_BH ≈ 7.8 × 10⁸ M_⊙, significantly less massive than our four ATLAS quasars, likely due to lower continuum luminosity, it being ≈1 mag fainter in apparent K mag and ≈0.5 mag fainter given its luminosity distance and assuming zero K-correction.

These ATLAS quasars are therefore close to having some of the most massive black holes so far discovered and they lie at a redshift where they may cause significant difficulties due to the fast gravitational growth rates needed to explain their existence at such early times. A more detailed M_BH analysis, with improved error estimates and, e.g. where account is taken of the Fe ii emission surrounding Mg ii in measuring its line width, is postponed to a forthcoming paper.

7 CONCLUSIONS

Using similar combinations of VST ATLAS and WISE colours as C15, we have found a further two z > 6 quasars and rediscovered a third at a redshift z ∼ 6. Adding these to the previous three found by C15 makes six quasars discovered in the searched ATLAS area of 3119 deg², corresponding to a volume of ∼17.4 Gpc³ between 5.7 < z < 6.4. We confirmed the two new quasars using low dispersion spectroscopy at the ESO NTT and Keck telescopes and have presented X-shooter spectra for these and the two other quasars discovered in ATLAS. The initial criteria listed in Section 3 gave a 100 per cent success rate for these six quasars that had the highest priority at the spectroscopic stage, indicating high selection efficiency. This possibly led to lower completeness rates and we have estimated the magnitude, star–galaxy separation, and colour completenesses. When the joint effect of these incompletenesses are taken into account, the corrected number of quasars rises to 11.3 ± 4.6. This corrected number is reasonably consistent with the discovery of a further five quasars detected in the overlap between the ATLAS and PS1 surveys by Bañados et al. (2016). We have compared our observed quasar space density at M_1450 Å ≈ −27.0 ± 0.5 mag to the SDSS results of Jiang et al. (2016). Our raw and completeness corrected space densities bracket those of these authors and so can be considered in good agreement with the SDSS results. Finally, from preliminary virial analyses of the Mg ii and C iv broad emission lines, we find black hole masses in the range, M_BH ≈ 1–6 × 10⁹ M_⊙ for the four ATLAS quasars.

ACKNOWLEDGEMENTS

Based on VST and VLT data products from observations made with ESO Telescopes at the La Silla Paranal Observatory under programme ID's 177.A-3011(A,B,C,D,E.F,G,H,I,J), 294.A-5031(B), 095.A-0506(A), and 096.A-0418(A,B). Observations were also made at the Keck Telescope (Project U071LA, PI X Prochaska). This publication also makes use of data products from the WISE, which is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology, and NEOWISE, which is a project of the Jet Propulsion Laboratory/California Institute of Technology. WISE and NEOWISE are funded by the National Aeronautics and Space Administration.

We also thank the PanSTARRS1 collaboration for access to the MDS data in advance of publication. The Pan-STARRS1 Surveys (PS1) have been made possible through contributions of the Institute for Astronomy, the University of Hawaii, the Pan-STARRS Project Office, the Max-Planck Society and its participating institutes, the Max Planck Institute for Astronomy, Heidelberg and the Max Planck Institute for Extraterrestrial Physics, Garching, The Johns Hopkins University, Durham University, the University of Edinburgh, Queen's University Belfast, the Harvard–Smithsonian Center for Astrophysics, the Las Cumbres Observatory Global Telescope Network Incorporated, the National Central University of Taiwan, the Space Telescope Science Institute, the National Aeronautics, and Space Administration under Grant no. NNX08AR22G issued through the Planetary Science Division of the NASA Science Mission Directorate, the National Science Foundation under Grant no. AST-1238877, the University of Maryland, and Eotvos Lorand University (ELTE) and the Los Alamos National Laboratory.

We also acknowledge support from Science and Technology Facilities Council Consolidated Grant ST/P 000541/1.

Finally, we thank an anonymous referee for comments that have significantly improved the quality of this paper.

Footnotes

REFERENCES