Subaru High-z Exploration of Low-Luminosity Quasars (SHELLQs). II. Discovery of 32 quasars and luminous galaxies at 5.7 < z ≤ 6.8

Abstract

1 Introduction

High-z quasars are a unique and useful probe of the early Universe in many aspects. The progress of cosmic reionization has been measured by the strength of H i absorption in background quasar spectra, which is very sensitive to the neutral fraction of the intergalactic medium (IGM: Gunn & Peterson 1965; Fan et al. 2006b). Stringent constraints on the seed population and initial growth of supermassive black holes (SMBHs) can be obtained from their mass function, in particular their maximum mass, as a function of redshift (e.g., Volonteri 2012; Ferrara et al. 2014; Madau et al. 2014). We can also learn about the formation of their host galaxies, which is thought to have happened in the highest density peaks of the underlying dark matter distribution in the early phase of cosmic history.

There has been great progress in the quest for high-z quasars in the last two decades.¹ This progress was made possible by the advent of wide-field (≥1000 deg²) surveys in the optical or near-infrared (IR) bands, such as SDSS (York et al. 2000), the Canada–France–Hawaii Telescope Legacy Survey, the United Kingdom Infrared Telescope (UKIRT) Infrared Deep Sky Survey (UKIDSS; Lawrence et al. 2007), the Panoramic Survey Telescope & Rapid Response System 1 (Pan-STARRS1; Kaiser et al. 2010) 3π survey, the Dark Energy Survey (Dark Energy Survey Collaboration 2016), and the Visible and Infrared Survey Telescope for Astronomy (VISTA) Kilo-degree Infrared Galaxy (VIKING). High-z quasar discoveries from the above and other projects are reported in Fan et al. (2000, 2001a, 2003, 2004, 2006b), Jiang et al. (2008, 2009, 2015, 2016), Willott et al. (2005, 2007, 2009, 2010a, 2010b), Mortlock et al. (2011), Bañados et al. (2014, 2016), Reed et al. (2015, 2017), Venemans et al. (2013, 2015a, 2015b), Goto (2006), Carnall et al. (2015), Kashikawa et al. (2015), Kim et al. (2015), Wu et al. (2015), and Wang et al. (2017).

More than 100 high-z quasars are known today (Bañados et al. 2016), with the most distant object found at z = 7.085 (Mortlock et al. 2011). However, most of these quasars have redshifts z < 6.5 and absolute magnitudes M₁₄₅₀ < −24 mag, while higher redshifts and fainter magnitudes are still poorly explored. The known quasars must be just the tip of an iceberg predominantly comprising faint quasars and active galactic nuclei (AGNs), which may be a significant contributor to reionization, and may represent the more typical mode of SMBH growth in the early Universe.

This paper is the second in a series presenting the results of the Subaru High-z Exploration of Low-Luminosity Quasars (SHELLQs) project, which exploits multi-band photometry data produced by the Hyper Suprime-Cam (HSC) Subaru Strategic Program (SSP) survey. The first results were presented in Matsuoka et al. (2016, hereafter Paper I), where we described the motivation and strategy of the project in detail, as well as the discovery of 15 quasars and luminous galaxies at 5.7 < z < 6.9 from the initial 80 deg² of the survey. In the present paper, we report the discovery of an additional 24 quasars and eight luminous galaxies at z > 5.7, from about 430 deg² (including the previous 80 deg²) of the survey. The spectroscopic follow-up campaign on the present survey area is still ongoing, and its results will be presented in forthcoming papers. We are also working to construct a quasar luminosity function at z ∼ 6, which will be presented in a separate paper.

This paper is organized as follows. We describe the photometric candidate selection briefly in section 2, while a more complete description is found in Paper I. The spectroscopic follow-up observations are described in section 3. The quasars and galaxies we have discovered are presented and discussed in section 4. The summary appears in section 5. We adopt the cosmological parameters H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_M = 0.3, and Ω_Λ = 0.7. All magnitudes in the optical and NIR bands are presented in the AB system (Oke & Gunn 1983). Magnitudes refer to point spread function (PSF) magnitudes (see subsection 2.1) corrected for Galactic extinction (Schlegel et al. 1998), unless otherwise noted. In what follows, we refer to z-band magnitudes using the subscript “AB” (“z_AB”), whereas redshift z is written without a subscript.

2 Photometric candidate selection

Our quasar candidates were selected in essentially the same way as in Paper I, so we describe the relevant procedure only briefly here, highlighting a few minor changes we made. The reader is referred to Paper I for a more complete description of our selection.

2.1 The Subaru HSC-SSP Survey

The SHELLQs project is based on the imaging data collected by the Subaru SSP survey with the HSC (Miyazaki et al. 2012; Miyazaki et al. 2018), a wide-field camera installed on the Subaru 8.2 m telescope. HSC has a nearly circular field of view of 1|${^{\circ}_{.}}$|5 diameter, covered by 116 2 K × 4 K Hamamatsu fully depleted CCDs, with a pixel scale of 0|${^{\prime\prime}_{.}}$|17. The survey has three layers: the Wide layer observes 1400 deg² mostly along the celestial equator, with the 5σ target depths of (g_AB, r_AB, i_AB, z_AB, y_AB) = (26.5, 26.1, 25.9, 25.1, 24.4) mag measured in 2|${^{\prime\prime}_{.}}$|0 apertures, while the Deep and the UltraDeep layers observe smaller areas (27 and 3.5 deg²) down to deeper limiting magnitudes (r_AB = 27.1 and 27.7 mag, respectively). A full description of the survey may be found in Aihara et al. (2018a). The first public data release (DR1) took place in 2017 February, which includes the data taken in the first 1.7 yr (2014 March to 2015 November) of the survey, covering 108 deg² of the Wide layer and the Deep and UltraDeep layers to intermediate depths (Aihara et al. 2018b). The median seeing during the above observing period was 0|${^{\prime\prime}_{.}}$|5–0|${^{\prime\prime}_{.}}$|8, depending on filter. The DR1 Wide layer reaches limiting magnitudes consistent with the above target values, while we are still accumulating exposures in the Deep and UltraDeep layers to reach their final target depths.

The results presented in this paper were drawn from imaging data observed before 2016 April, i.e., a newer dataset than is contained in DR1. We used about 430 deg² of the Wide layer, in which we have more than one exposure in the i, z, and y bands.² The total integration time per pointing in each of the (i, z, y) bands in the Wide layer is 20 min, divided into six individual exposures with different dither positions. Data reduction was performed with the dedicated pipeline hscPipe (versions 4.0.1 and 4.0.2; Bosch et al. 2018) derived from the Large Synoptic Survey Telescope software pipeline (Jurić et al. 2015), for all the standard procedures including bias subtraction, flat-fielding with dome flats, stacking, and astrometric and photometric calibrations, as well as source detection and measurements. The astrometric and photometric calibrations are tied to the Pan-STARRS1 system (Schlafly et al. 2012; Tonry et al. 2012; Magnier et al. 2013; Chambers et al. 2016; Flewelling et al. 2016). We use the PSF magnitude (m_AB) and the CModel magnitude (m_{CModel, AB}), which are measured by fitting the PSF models and two-component, PSF-convolved galaxy models to the source profile, respectively (Abazajian et al. 2004). These magnitudes have been corrected for Galactic extinction (Schlegel et al. 1998).

2.2 Candidate selection

Fig. 1.

HSC i_AB − z_AB and z_AB − y_AB colors of high-z quasars (blue line) and galaxies (black line), as well as Galactic stars and brown dwarfs (crosses and dots). The models and references used to compute these colors are described in Paper I. The open circles along the blue and black lines represent redshifts in steps of 0.1, with z = 6.0 and 6.5 marked by the larger circles. The hatched area in lower left indicates the color space excluded from the SSP database query in the first step of the quasar selection (subsection 2.2). The grey scale represents the Bayesian quasar probability |$P_{\rm Q}^{\rm B}$| (the color bar is found at the top) over this plane, for a source magnitude of z_AB = 24.0 mag and 5σ limiting magnitudes of (i_AB, z_AB, y_AB) = (26.5, 25.5, 25.0) mag. Galaxy models are not included in our algorithm at present.

Open in new tabDownload slide

Next, an automatic image-checking procedure is run on stacked images and per-visit images (i.e., images of individual exposures before stacking) of the above sources. This procedure uses Source Extractor (Bertin & Arnouts 1996) in the double-image mode, with the stacked image as the detection reference. If any of the per-visit photometric measurements deviate from the stacked photometry by more than three times the measurement error, the source is eliminated. We also reject sources with too compact, diffuse, or elliptical profiles to be celestial point sources on the stacked images. The eliminated sources are mostly cosmic rays, moving or transient sources, and image artifacts.

The red HSC sources selected above are matched to the UKIDSS and VIKING catalogs within 1|${^{\prime\prime}_{.}}$|0, which provides J-, H-, and K-band photometry. The present work utilizes the UKIDSS data release 10 and the VIKING data release 4. The UKIDSS data cover almost the entire HSC survey fields, while the available VIKING data cover 30%–40% of the HSC survey footprint used in this work.

We use a template-based Bayesian probabilistic method, which selects quasar candidates from the above sample of red HSC sources. Similar methods of spectral energy distribution (SED) fitting have been used in other quasar surveys (e.g., Reed et al. 2017), while more advanced Bayesian treatment, e.g., with surface density priors (Mortlock et al. 2012), is not frequently seen in the literature. For each source, we calculate the Bayesian probability (⁠|$P_{\rm Q}^{\rm B}$|⁠) of being a high-z quasar rather than a contaminating brown dwarf, based on models of SED and surface density as a function of magnitude. Galaxy models are not included in the algorithm at present. The quasar SED models were created by stacking the SDSS spectra of 340 bright quasars at z ≃ 3, where the quasar selection is fairly complete (Richards et al. 2002; Willott et al. 2005), and correcting for the effect of IGM absorption (Songaila 2004). The quasar surface density was modeled based on the luminosity function taken from Willott et al. (2010a). A more detailed description of our Bayesian algorithm may be found in Paper I.

We keep those sources with |$P_{\rm Q}^{\rm B} > 0.1$| in the sample of candidates, while removing sources with lower quasar probability. The rather low value of the threshold |$P_{\rm Q}^{\rm B} = 0.1$| was chosen to ensure that we would not throw away any possible candidates. We found that the actual |$P_{\rm Q}^{\rm B}$| distribution is bimodal, with only a small fraction falling in |$0.1 < P_{\rm Q}^{\rm B} < 0.9$| (see subsection 4.5), so our results are insensitive to the exact value of this cut. As we will see below, the bimodal |$P_{\rm Q}^{\rm B}$| distribution is populated mostly by high-z quasars at |$P_{\rm Q}^{\rm B} \simeq 1$| and brown dwarfs at |$P_{\rm Q}^{\rm B} \simeq 0$|⁠. This indicates that quasars and brown dwarfs occupy distinct regions of the color space, and HSC (and UKIDSS/VIKING) photometry is sufficiently accurate to separate these two populations, down to the limiting magnitude of our quasar survey.

Finally, we inspect the images of all the candidates by eye and reject additional problematic sources. Both stacked and per-visit images are used for this purpose. The sources rejected at this stage include those close to bright stars, cosmic rays, and moving objects overlooked in the above automatic procedure.

In the present survey area covering 430 deg², we had roughly 300000 red point sources meeting the database query conditions [expressions (1) and (2)] and undetected in the g and r bands. The vast majority of them (∼97%) were eliminated by the automatic image checking procedure, and the Bayesian algorithm selected 261 final candidates with |$P_{\rm Q}^{\rm B} > 0.1$|⁠. Among them, we put highest priority for follow-up spectroscopy on the 60 candidates that had the reddest colors (i_AB − z_AB > 2.0 or z_AB − y_AB > 0.8; see figure 1), relatively bright magnitudes (z_AB < 24 mag), and were detected in more than a single band or a single exposure.

2.3 Recovery of known objects

In the 430 deg² surveyed in this paper, there are 10 high-z quasars discovered by previous surveys. We have recovered seven of these quasars (CFHQS J0210−0456, CFHQS J0216−0455, CFHQS J0227−0605, SDSS J1602+4228, IMS J2204+0012, VIMOS2911001793, and SDSS J2228+0110, following the naming convention of Bañados et al. 2016), all of which have |$P_{\rm Q}^{\rm B} = 1.00$|⁠. On the other hand, two quasars (SDSS J0836+0054 and VIK J1148+0056) have relatively low redshifts (z = 5.81 and 5.84, respectively) and bluer HSC colors (i − z < 1.5) than our selection threshold. The remaining quasar (VIK J0839+0015) was dropped from the selection, because it has r-band detection (r_AB = 25.16 ± 0.11 mag) in the HSC catalog. This quasar is at z = 5.84 and bright also in the i band (i_AB = 22.86 ± 0.01 mag), so the r-band detection may be real. Alternatively, the r-band flux peak may be an artifact due to the halo around a saturated star, at ∼13″ away from the quasar.

Our quasar selection algorithm was run on all HSC data observed before 2016 April, so the present quasar candidates supersede those in Paper I. Because the HSC data reduction pipeline is improving continuously, and because we made a minor change to the selection criteria, as mentioned above, the candidate lists in a given observed field may vary from one round of selection to another. In Paper I, we reported spectroscopic identification of nine high-z quasars, six high-z galaxies, and one brown dwarf. We found that the present selection recovers eight quasars and two galaxies from the above sample. J2232+0012, which we classified as a quasar based on the very high luminosity (∼10^44.1 erg s⁻¹) of its narrow Lyα line, was dropped from the candidates, because its extendedness (z_AB − z_{CModel, AB} = 0.156) is slightly above our new threshold (0.150; see above). The two galaxies J0857+0142 and J0848+0045 were dropped from the quasar candidates for the same reason. The three remaining sources, i.e., the galaxies J0210−0523 and J0215−0555 and the brown dwarf J0850+0012, have lower |$P_{\rm Q}^{\rm B}$| values than in Paper I, and did not meet the selection criterion (⁠|$P_{\rm Q}^{\rm B} > 0.1$|⁠).

3 Spectroscopy

Since the discovery reported in Paper I, we took optical spectra of 48 additional unidentified quasar candidates, using the Optical System for Imaging and low-Intermediate-Resolution Integrated Spectroscopy (OSIRIS: Cepa et al. 2000) mounted on the Gran Telescopio Canarias (GTC), and the Faint Object Camera and Spectrograph (FOCAS: Kashikawa et al. 2002) mounted on Subaru. The observations were carried out in the 2015 fall to 2016 fall semesters. Roughly the brighter half of the candidates were observed with OSIRIS, while the fainter candidates were observed with FOCAS. We prioritized observations in such a way that the targets with brighter magnitudes and higher |$P_{\rm Q}^{\rm B}$| were observed at the earlier opportunities. The journal of these discovery observations is presented in table 1. The details of the observations are described in the following sections.

3.1 GTC/OSIRIS

GTC is a 10.4 m telescope located at the Observatorio del Roque de los Muchachos in La Palma, Spain. Our program was awarded 14.4 and 21.5 hr in the 2015B and 2016A semesters, respectively (GTC19-15B and GTC4-16A). We used OSIRIS with the R2500I grism and 1|${^{\prime\prime}_{.}}$|0-wide long slit, which provides spectral coverage from λ_obs = 0.74 to 1.0 μm with a resolution R ∼ 1500. The observations were carried out in queue mode on dark and gray nights, with mostly photometric (sometimes spectroscopic) sky conditions and seeing of 0|${^{\prime\prime}_{.}}$|6–1|${^{\prime\prime}_{.}}$|2. The data were reduced using the Image Reduction and Analysis Facility (IRAF). Bias correction, flat-fielding with dome flats, sky subtraction, and 1D extraction were performed in the standard way. The wavelength was calibrated with reference to sky emission lines. The flux calibration was tied to white dwarf standard stars (Feige 110, Feige 66, G191-B2B, GD 140, or Ross 640) observed on the same nights. We corrected for slit losses by scaling the spectra to match the HSC magnitudes in the z and y bands for the i- and z-band dropouts, respectively.

3.2 Subaru/FOCAS

Our program was awarded five nights each in the S16A and S16B semesters (S16A-076 and S16B-071I) with the Subaru 8.2 m telescope. The latter program (S16B-071I) has been approved as a Subaru intensive program, for which a total of 20 nights will be allocated during the S16B–S18A semesters. We used FOCAS in the multi-object spectrograph mode with the VPH900 grism and SO58 order-sorting filter. The widths of the slitlets were set to 1|${^{\prime\prime}_{.}}$|0. This configuration provides spectral coverage from λ_obs = 0.75 to 1.05 μm with a resolution R ∼ 1200. All the observations were carried out on grey nights. A few of these nights were occasionally affected by cirrus and poor seeing (≲ 2|${^{\prime\prime}_{.}}$|0), while the weather was fairly good with seeing 0|${^{\prime\prime}_{.}}$|5–1|${^{\prime\prime}_{.}}$|0 for the rest of the observations.

The data were reduced with IRAF using the dedicated FOCASRED package. Bias correction, flat-fielding with dome flats, sky subtraction, and 1D extraction were performed in the standard way. The wavelength was calibrated with reference to the sky emission lines. The flux calibration was tied to white dwarf standard stars (Feige 110 and G191-B2B) observed on the same nights as the targets. We corrected for slit losses in the same way as in the OSIRIS data reductions.

4 Results and discussion

Figures 2–8 present the reduced spectra of the 48 quasar candidates. As we describe in detail below, we identified 24 high-z quasars, eight high-z galaxies, two strong [O iii] emitters at z ∼ 0.8, and 14 brown dwarfs. Their photometric properties are summarized in table 2. Note that the astrometric accuracy of the HSC-SSP data is estimated to be ≲ 0|${^{\prime\prime}_{.}}$|1 (root mean square; Aihara et al. 2018b). Table 2 also lists the updated magnitudes and |$P_{\rm Q}^{\rm B}$| of the 16 objects presented in Paper I, measured with the present version of the HSC data reduction pipeline. Note that some of the objects in this table do not meet our latest quasar selection criteria, due either to the improvement of the HSC photometry or to our minor changes in the selection criteria (see subsection 2.2). We clarify this point below whenever necessary. We have now spectroscopically identified 64 HSC sources in Paper I and this work, which include 33 high-z quasars, 14 high-z galaxies, two [O iii] emitters, and 15 brown dwarfs. Six of these sources are detected in the J, H, and/or K bands, as summarized in table 3.

Fig. 2.

Reduced spectra of the first set of eight quasars and possible quasars discovered in this work, displayed in decreasing order of redshift. The object name and the estimated redshift are indicated at the top left-hand corner of each panel. The blue dotted lines mark the expected positions of the Lyα and N v λ1240 emission lines, given the redshifts. The spectra were smoothed using inverse-variance weighted means over 3–9 pixels (depending on the S/N), for display purposes. The bottom panel displays a sky spectrum, as a guide to the expected noise. (Color online)

Open in new tabDownload slide

Fig. 3.

Same as figure 2, but for the second set of eight quasars and possible quasars. (Color online)

Open in new tabDownload slide

Fig. 4.

Same as figure 2, but for the last set of eight quasars and possible quasars. (Color online)

Open in new tabDownload slide

Fig. 5.

Same as figure 2, but for the eight high-z galaxies. The expected positions of the interstellar absorption lines of Si ii λ1260, Si ii λ1304, and C ii λ1335 are marked by the red dotted lines. (Color online)

Open in new tabDownload slide

Fig. 6.

Same as figure 2, but for the two [O iii] emitters at z ∼ 0.8. The expected positions of Hγ, Hβ, and two [O iii] lines (λ4959 and λ5007) are marked by the dotted lines. (Color online)

Open in new tabDownload slide

Fig. 7.

Same as figure 2, but for the first seven brown dwarfs. The red lines represent the best-fitting templates, the spectral types of which are indicated at the top left-hand corner of each panel. (Color online)

Open in new tabDownload slide

Fig. 8.

Same as figure 7, but for the remaining seven brown dwarfs. (Color online)

Open in new tabDownload slide

4.1 Quasars and possible quasars

We identified 24 new quasars and possible quasars at 5.9 < z ≲ 6.8, as displayed in figures 2–4 and listed in the first section of table 4. The highest-z quasar, J1429−0104, was the only z-band dropout we took a spectrum of. This quasar has two emission peaks, which have wavelengths close to the expected positions of Lyα and N v λ1240 at z = 6.8. The dip between the two emission lines is likely caused by a broad absorption line (BAL) system of N v. The majority of the objects in figures 2–4 exhibit characteristic spectral features of high-z quasars, namely, strong and broad Lyα and in some cases N v λ1240, blue rest-UV continua, and sharp continuum breaks just shortward of Lyα. On the other hand, several objects have considerably narrower Lyα than do typical quasars. As discussed later in this section, we classify these objects with narrow Lyα as (possible) quasars, given their high Lyα luminosity, and possible mini BAL feature of N v found in their composite spectrum.