SUBARU HIGH-z EXPLORATION OF LOW-LUMINOSITY QUASARS (SHELLQs). I. DISCOVERY OF 15 QUASARS AND BRIGHT GALAXIES AT 5.7 < z < 6.9^{∗ †}

The era from the birth of the first stars to cosmic reionization is one of the key subjects in astronomy and astrophysics today. While the formation of the first stars is observationally out of reach at present, the epoch of reionization is being explored with several different approaches. The latest measurements of the cosmic microwave background by the Planck space mission suggest a reionization optical depth of τ = 0.058 ± 0.012, which implies that the midpoint of reionization occurred between z = 7.8 and 8.8 (Planck Collaboration 2016). This value is marginally consistent with the rapid decline of the H i neutral fraction of the intergalactic medium (IGM) from z ∼ 8 to 6, inferred from the evolving Lyα luminosity function of galaxies (e.g., Ouchi et al. 2008, 2010; Konno et al. 2014; Bouwens et al. 2015; Choudhury et al. 2015). The universe is in the final phase of reionization at z ∼ 6, as suggested by the small but non-zero H i transmission fraction of the IGM implied by the Gunn & Peterson (1965, GP hereafter) troughs of luminous high-z quasars (Fan et al. 2006a).

The main source of the ultraviolet (UV) photons that caused the reionization of the universe is still under debate. It has been argued that star-forming galaxies observed in deep surveys are not able to produce a sufficient number of photons to sustain reionization (e.g., Robertson et al. 2010, 2013), while the revised Planck results with later reionization than previously thought may alleviate this problem (e.g., Robertson et al. 2015). Active galactic nuclei (AGNs) have been studied as a possible additional source of ionizing photons (e.g., Lehnert & Bremer 2003; Fontanot et al. 2012; Grissom et al. 2014; Giallongo et al. 2015; Madau & Haardt 2015), but the results are still controversial, largely due to the lack of knowledge about the numbers of faint quasars and AGNs residing in the reionization era.

High-z quasars are also a key population to understanding the formation and evolution of supermassive black holes (SMBHs). If the assembly of an SMBH is predominantly via gas accretion onto a seed black hole with mass M_BH,0, then the time needed to grow to the mass M_BH is

$$\begin{eqnarray*}t & = & {t}_{{\rm{Edd}}}\left(\displaystyle \frac{\epsilon }{1-\epsilon }\right){\lambda }_{{\rm{Edd}}}^{-1}\ {\rm{ln}}\left(\displaystyle \frac{{M}_{{\rm{BH}}}}{{M}_{{\rm{BH,0}}}}\right)\\ & = & 0.043\ {\rm{ln}}\left(\displaystyle \frac{{M}_{{\rm{BH}}}}{{M}_{{\rm{BH,0}}}}\right)\ {\rm{Gyr}},\end{eqnarray*}$$

where ${t}_{{\rm{Edd}}}=0.44{\mu }_{e}^{-1}$ Gyr is the Eddington timescale, μ_e is the mean molecular weight per electron, is the radiative efficiency, and λ_Edd is the Eddington ratio (Shapiro 2005; Madau et al. 2014); here we assume that λ_Edd is constant in time. We adopt μ_e = 1.15,  = 0.1, and λ_Edd = 1.0 to derive the second line of the equation. For example, a seed with M_BH,0 = 100 M_⊙ will take 0.7 Gyr to form a quasar with M_BH = 10⁹ M_⊙. Because this timescale is comparable to the cosmic time that elapsed between z = 20 and z = 6, the SMBH mass function at z > 6 conveys critical information about the mass distribution of the seed black holes and the mode of subsequent growth, including super-Eddington accretion (Kawaguchi et al. 2004). Indeed, recent discoveries of the luminous quasars ULAS J1120+0641 with M_BH ∼ 2 × 10⁹ M_⊙ at z = 7.085 (Mortlock et al. 2011) and SDSS J0100+2802 with M_BH ∼ 1 × 10¹⁰ M_⊙ at z = 6.30 (Wu et al. 2015) have made a significant impact on such models (e.g., Volonteri 2012; Ferrara et al. 2014; Madau et al. 2014).

Furthermore, high-z quasars might be a signpost of galaxies and high density peaks in the dark matter distribution in the early universe. The stellar and gaseous properties in and around the host galaxies can be studied in the optical/infrared (e.g., Kashikawa et al. 2007; Goto et al. 2009, 2012; Willott et al. 2011) or at sub-mm/radio wavelengths (e.g., Maiolino et al. 2005; Wang et al. 2007; Venemans et al. 2012, 2016; Wang et al. 2013; Willott et al. 2013, 2015), giving a unique probe of galaxies in the reionization era. The chemical enrichment, and thus the preceding star formation history, can be measured with strong metal emission lines arising from ionized gas around the quasar nuclei (e.g., Jiang et al. 2007; De Rosa et al. 2011, 2014). On the whole, no or little chemical evolution of quasars has been observed from z ∼ 7 to the local universe. A good example of this is ULAS J1120+0641, whose emission-line and continuum spectrum is strikingly similar to those of the local quasars except for the deep GP trough.

In the last two decades there has been great progress in the quest for high-z quasars.²² The SDSS (York et al. 2000) provided the first opportunity to search for high-z quasars over wide fields (>1000 deg²), resulting in several tens of objects published to date (Fan et al. 2000, 2001a, 2003, 2004, 2006b; Jiang et al. 2008, 2009, 2015). The Canada–France High-z Quasar Survey (CFHQS; Willott et al. 2005, 2007, 2009, 2010a, 2010b) explored fainter magnitudes than the SDSS and found a few tens of new quasars, including one very faint object (CFHQS J0216–0455; z_AB = 24.4 mag at z = 6.01) discovered in the Subaru XMM-Newton Deep Survey (Furusawa et al. 2008) area. However, these optical surveys are not sensitive to redshifts beyond z = 6.5, where quasars become almost invisible at observed wavelengths λ_obs < 0.9 μm due to strong IGM absorption. The first quasar discovered at z > 6.5 was ULAS J1120+0641, which was selected from near-infrared (NIR) data of the United Kingdom Infrared Telescope (UKIRT) Infrared Deep Sky Survey (UKIDSS; Lawrence et al. 2007).

With the advent of the Visible and Infrared Survey Telescope for Astronomy (VISTA) Kilo-degree Infrared Galaxy (VIKING) survey, more z > 6.5 quasars have been discovered in recent years (Venemans et al. 2013, 2015). New optical wide-field surveys such as the Panoramic Survey Telescope & Rapid Response System 1 (Pan-STARRS1; Kaiser et al. 2010) 3π survey and the Dark Energy Survey (Dark Energy Survey Collaboration et al. 2016) are equipped with a y-band filter centered at 9500–10000 Å, and are starting to deliver many more quasars at 6 ≤ z ≤ 7 (Bañados et al. 2014; Reed et al. 2015). In addition, various smaller projects have succeeded in identifying high-z quasars (e.g., Goto 2006; Carnall et al. 2015; Kashikawa et al. 2015; Kim et al. 2015). In total, the above surveys have identified about a hundred high-z quasars published to date. Most of the quasars are located at z < 6.5 and z_AB < 22.5 mag, while the higher redshifts and fainter magnitudes are still poorly explored. The known bright high-z quasars must be just the tip of the iceberg predominantly composed of faint quasars and 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 describes our ongoing project, SHELLQs (Subaru High-z Exploration of Low-Luminosity Quasars), which is the first 1000 deg² class survey for high-z quasars with a 8 m class telescope. The project exploits multiband photometry data produced by the Subaru Hyper Suprime-Cam (HSC) Subaru Strategic Program (SSP) survey. We present the results of the initial follow-up spectroscopy of photometric candidates, performed in the 2015 Fall and 2016 Spring semesters, which delivered 15 high-z objects including both quasars and bright galaxies. This paper is organized as follows. We introduce the Subaru HSC-SSP survey in Section 2. The details of the photometric candidate selection are presented in Section 3. The spectroscopic follow-up observations are described in Section 4. The quasars and galaxies we have discovered are presented and discussed in Section 5. The summary appears in Section 6. 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 Section 2) unless otherwise noted.

The Subaru HSC-SSP survey (M. Takada et al. 2016, in preparation) is a large collaborative project with contributions from researchers in Japan, Taiwan, and Princeton University. The project started in early 2014, and will include 300 nights, lasting until around 2019. It uses the HSC (Miyazaki et al. 2012; S. Miyazaki et al. 2016, in preparation), a wide-field camera newly installed on the Subaru 8.2 m telescope on the summit of Maunakea. HSC is equipped with 116 2K × 4K Hamamatsu fully depleted CCDs, of which 104 CCDs are used to obtain science data. The pixel scale is 017. The camera has a nearly circular field of view of 15 diameter, which enables it to image 1.77 deg² of the sky in a single shot. Five broad-band filters (g, r, i, z, and y) and several narrow-band filters are currently available.

The HSC-SSP survey has three layers with different combinations of area and depth. The Wide layer aims to observe 1400 deg² mostly along the celestial equator through the five broad-band filters. The present paper is based on this Wide-layer data. The total exposure times range from 10 minutes in the g- and r-bands to 20 minutes in the i-, z-, and y-bands, divided into individual exposures of ∼3 minutes each. The target 5σ limiting magnitudes are (g, r, i, z, y) = (26.5, 26.1, 25.9, 25.1, 24.4) mag measured in 20 apertures. The Deep and Ultra-Deep layers observe 27 and 3.5 deg², respectively, within and around popular deep survey fields. Five broad-band filters and four narrow-band filters are used, aiming to reach the 5σ limiting depth of r = 27.1 mag (Deep) or r = 27.7 mag (Ultra-Deep).

The SHELLQs project exploits the exquisite HSC survey data to search for low-luminosity quasars at high redshift. Assuming the quasar luminosity function at z ≥ 6 presented by Willott et al. (2010b), the expected numbers of newly identified quasars in the Wide layer are ∼500 with z_AB < 24.5 mag at z ∼ 6 and ∼100 with y_AB < 24.0 mag at z ∼ 7. The former magnitude limit corresponds to M₁₄₅₀ < −22 mag for a typical quasar spectral energy distribution (SED), thus allowing us to explore ∼2 mag lower luminosity than any previous wide-field survey at z ∼ 6 (e.g., Jiang et al. 2009). Our filter set is sensitive to quasars with redshifts up to z ∼ 7.4 (i.e., beyond the current quasar redshift record). However, the detection capability sharply drops at z > 7 where the GP trough comes into the y-band, hence the survey is limited to intrinsically very luminous objects at those redshifts. The deep optical data produced by the HSC survey will also provide opportunities to explore even higher redshifts when combined with wide and deep NIR surveys.

This paper describes the results from the early HSC survey data taken before 2015 August. It covers roughly 80 deg² in the five broad bands in the Wide layer, with a median seeing of 06–08. Data reduction was performed with the dedicated pipeline hscPipe (version 3.8.5; J. Bosch et al. 2016, in preparation) derived from the Large Synoptic Survey Telescope (LSST) software pipeline (Ivezic et al. 2008; Axelrod et al. 2010; Jurić et al. 2015), for all the standard procedures including bias subtraction, flat fielding with dome flats, stacking, astrometric and photometric calibrations, and 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). We utilize forced photometry, which allows for flux measurements in all five bands with a consistent aperture defined in a reference band. The reference band is i by default and is switched to z (y) for extremely red sources with no detection in the i (z) and bluer bands. We use the PSF magnitude (m_PSF, or simply m) and the cModel magnitude (m_cModel), which are measured by fitting the PSF models and two-component, PSF-convolved galaxy models to the source profile, respectively (Abazajian et al. 2004). We measure fluxes and colors of sources with m_PSF, while the source extendedness is evaluated with ${m}_{{\rm{PSF}}}-{m}_{{\rm{cModel}}}$. All the magnitudes are corrected for Galactic extinction (Schlegel et al. 1998).

We performed a rough assessment of the completeness limits achieved in the early Wide survey as follows. In each subregion of the sky with an approximate size of 12' × 12', we select every source whose processing flag indicates clean photometry, and measure the number counts N(m). It typically follows a straight line in the m − log$[N(m)]$ plane at bright magnitudes, then peaks at m = m_peak. We fit a straight line $\widetilde{N}(m)$ to the m − log$[N(m)]$ relation at ${m}_{{\rm{peak}}}-5\lt m\lt {m}_{{\rm{peak}}}-1$, and define the 50% completeness magnitude at the point where N(m) falls to half of $\widetilde{N}(m)$. The resultant completeness magnitudes averaged over all fields are (26.5, 26.3, 26.4, 25.5, 24.7) mag in the (g, r, i, z, y)-band, respectively, with a typical field-to-field variation of 0.3 mag. The spatial pattern of the above estimates agrees with that of the seeing, in such a way that poor seeing is accompanied by worse-than-average depth. These completeness magnitudes are roughly 0.3 mag fainter than the target limiting magnitudes of the survey mentioned above, which is at least partly because of the differences in the adopted magnitudes (PSF magnitudes versus 20-aperture magnitudes) and in the definitions of the depths (50% completeness versus 5σ detection).

Our project also benefits from archival NIR data from UKIDSS and VIKING. The UKIDSS is a multi-tiered imaging survey using WFCAM, a wide-field camera mounted on the UKIRT 3.8 m telescope (Lawrence et al. 2007). The widest Large Area Survey covers most of the HSC survey footprint, with target 5σ limiting magnitudes of (Y, J, H, K) = (20.9, 20.4, 20.0, 20.1) mag measured in 20 apertures. VIKING is one of the public surveys of the European Southern Observatory with the VISTA 4.1 m telescope. This project aims to observe 1500 deg² of the sky, with target 5σ limiting magnitudes of (Z, Y, J, H, K) = (23.1, 22.3, 22.1, 21.5, 21.2) mag measured in 20 apertures. Roughly half of the HSC survey footprint will be covered by the VIKING at its completion. The present work uses the data from the UKIDSS data release 10 and the VIKING data release 4.

High-z quasars are characterized by extremely red optical colors caused by strong IGM absorption blueward of Lyα. This is demonstrated in Figure 1, which presents i − z and z − y colors of model quasars and other populations in the HSC passband system. There are three major sources of astrophysical contamination to the photometric selection of quasars. The first is Galactic brown dwarfs, which have been the most serious contaminants in past surveys because of their very red colors and point-like appearance. The second is red galaxies at z ∼ 1, whose 4000 Å break leads to red i − z colors, but we expect that the excellent image quality of the HSC will help identify those low-z galaxies morphologically. The third is faint Lyman-break galaxies (LBGs) at z ≥ 6, which are also affected by IGM absorption. Figure 2 displays the luminosity functions of high-z quasars (Willott et al. 2010b; Kashikawa et al. 2015) and LBGs (Bouwens et al. 2015; Bowler et al. 2015). We assume the galaxy UV spectral slope of β = −2 (e.g., Stanway et al. 2005) to convert the UV magnitudes in the literature to M₁₄₅₀. Although the luminosity functions are still poorly constrained at M₁₄₅₀ > −24 mag for quasars and at <−22 mag for LBGs, it is likely that they intersect at an apparent magnitude of ∼24 mag, with LBGs outnumbering quasars at fainter magnitudes. This is why previous surveys at brighter magnitudes did not suffer from severe LBG contamination. The LBG contamination in our project could be significant, particularly if the LBG luminosity function has a double power-law form instead of the Schechter function with its exponential cut-off at the bright end (Bowler et al. 2014, 2015).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

There are several methods to extract quasar candidates from multiband imaging data. Our sample of candidates is selected based on a Bayesian probabilistic approach, as detailed in Section 3.1. It computes the posterior probability for each source being a high-z quasar rather than a red star or dwarf, based on photometry in all the available bands as well as SED and surface density models of the populations under consideration. We use photometry in the HSC optical bands plus the NIR bands (Y, J, H, K) where available from UKIDSS or VIKING. While UKIDSS is too shallow for all but the brightest HSC sources, the VIKING data are useful to identify and remove brown dwarfs, because the mean colors of L–T dwarfs (${z}_{{\rm{AB}}}-{J}_{{\rm{AB}}}\sim 2\mbox{--}4$ mag) match the relative depths of the HSC survey and VIKING. The entire flow of the candidate selection from the HSC-SSP database to the final spectroscopic targets is described in Section 3.2. Our selection procedure efficiently removed contaminants, recovered all the known quasars imaged by the HSC so far (Section 3.3), and discovered a number of new objects as described in the following sections.

3.1. Bayesian Algorithm

Our quasar candidates are selected with a Bayesian probabilistic algorithm, developed following Mortlock et al. (2012). Here we assume that galaxies and relatively blue stars with O to early-M spectral types have been removed in advance with color and extendedness cuts (see Section 3.2). Although this may not be the case for LBGs at z ≳ 6, we do not model this population at present because (i) they are hard to distinguish from high-z quasars by colors alone and (ii) their surface density at <24 mag is still poorly understood (see Figures 1 and 2).

For a detected source with the observed quantities ${\boldsymbol{d}}$, the Bayesian probability ${P}_{{\rm{Q}}}^{{\rm{B}}}$ of being a quasar is given by:

$$\begin{eqnarray}&&{P}_{{\rm{Q}}}^{{\rm{B}}}({\boldsymbol{d}})=\displaystyle \frac{{W}_{{\rm{Q}}}({\boldsymbol{d}})}{{W}_{{\rm{Q}}}({\boldsymbol{d}})+{W}_{{\rm{D}}}({\boldsymbol{d}})}\end{eqnarray} \tag{ 1 }$$

and

$$\begin{eqnarray}&&{W}_{{\rm{Q/D}}}({\boldsymbol{d}})=\displaystyle \int S({\boldsymbol{p}})\ {\Pr }({\rm{\det }}| {\boldsymbol{p}})\ {\Pr }({\boldsymbol{d}}| {\boldsymbol{p}})\ d{\boldsymbol{p}}\end{eqnarray} \tag{ 2 }$$

where the subscripts Q and D denote a quasar and a brown dwarf, respectively. The vector ${\boldsymbol{d}}$ represents the magnitudes in all the available bands in the present case, while ${\boldsymbol{p}}$ represents the intrinsic source properties (i.e., luminosity and redshift for a quasar and luminosity and spectral type for a brown dwarf). The functions $S({\boldsymbol{p}})$, ${\Pr }({\rm{\det }}| {\boldsymbol{p}})$, and ${\Pr }({\boldsymbol{d}}| {\boldsymbol{p}})$ represent the surface number density, the probability that the source is detected ("det"), and the probability that the source has the observed quantities ${\boldsymbol{d}}$, respectively, each as a function of ${\boldsymbol{p}}$.

We compute $S({\boldsymbol{p}})$ with the quasar luminosity function of Willott et al. (2010b) and the Galactic brown dwarf model of Caballero et al. (2008). The former is well determined at M₁₄₅₀ < −24 mag, while we extrapolate it to M₁₄₅₀ = −20 mag as shown in Figure 2 to match the HSC observations. The brown dwarf model takes into account the spatial density distributions and luminosities of late-M, L, and T dwarfs, and allows one to compute number counts for each spectral type at a given Galactic coordinate. At our quasar selection limit of ${z}_{{\rm{AB}}}=24.5$ mag (see below), L–T dwarfs within ∼1 kpc of the Sun are bright enough to enter our sample. The validity of these quasar and brown dwarf models will be evaluated with the results of our and other surveys in future work.

Since the HSC-SSP survey depth has not been fully analyzed yet (Section 2), we arbitrarily set ${\Pr }({\rm{\det }}| {\boldsymbol{p}})=1$ for z_AB < 26.0 or y_AB < 25.0 mag and 0 otherwise. The SED models required for ${\Pr }({\boldsymbol{d}}| {\boldsymbol{p}})$ are created as follows. The quasar model spectrum at z = 0 is first 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), after correcting for the foreground IGM absorption. The IGM H i opacity data are taken from Songaila (2004). This spectrum is then placed at various redshifts with the appropriate IGM absorption taken into account, and convolved with the filter transmission functions to compute colors. Because the model spectra redshifted to z > 6 do not extend beyond the J-band, we take the J − H and H − K colors from Hewett et al. (2006). The dwarf colors are computed with a set of observed spectra compiled in the SpeX prism library²³ and the CGS4 library.²⁴ Because of the discrete sampling of the brown-dwarf templates grouped into individual spectral types, the integration in Equation (2) is treated as a summation for spectral types. Finally, the flux errors are taken from the outputs of the HSC image processing pipeline, and are assumed to follow a Gaussian probability density distribution in fluxes.

3.2. Selection Flow

The present work is based on the HSC-SSP Wide-layer data included in the S15A internal data release, which happened in 2015 September. Forced photometry (see Section 2) on the stacked images is used. We first query the HSC-SSP database for non-blended²⁵ sources meeting the following criteria:

$$\begin{eqnarray}&&({z}_{{\rm{AB}}}\lt 24.5\ {\rm{and}}\ {\sigma }_{z}\lt 0.155\ {\rm{and}}\ {i}_{{\rm{AB}}}-{z}_{{\rm{AB}}}\gt 1.5\\ &&\quad {\rm{and}}\ {z}_{{\rm{AB}}}-{z}_{{\rm{cModel,AB}}}\lt 0.3)\end{eqnarray} \tag{ 3 }$$

or

$$\begin{eqnarray}&&({y}_{{\rm{AB}}}\lt 24.0\ {\rm{and}}\ {\sigma }_{y}\lt 0.155\ {\rm{and}}\ {z}_{{\rm{AB}}}-{y}_{{\rm{AB}}}\gt 0.8\\ &&\quad {\rm{and}}\ {y}_{{\rm{AB}}}-{y}_{{\rm{cModel,AB}}}\lt 0.3)\end{eqnarray} \tag{ 4 }$$

and without any critical quality flags assigned.²⁶ Throughout this paper i_AB, z_AB, and y_AB refer to PSF magnitudes. The conditions of Equation (3) select i-band dropouts at z ∼ 6, while those of Equation (4) select z-band dropouts at z ∼ 7. The color cuts are used to exclude relatively blue stars with O to early-M spectral types (see Figure 1), while the difference between the PSF and cModel magnitudes is used to exclude extended sources. After the database query, we further remove low-z interlopers with more than 3σ detection in the g- or r-band.

Because we are looking for rare faint objects detected in only one or two bands, we are sensitive to false dropout sources of both astrophysical and non-astrophysical origins. However, we can take advantage of the fact that in the HSC-SSP Wide survey, each patch of the sky is visited and imaged several times with different dithering positions. Glitches in the data may become apparent upon comparing the individual per-visit exposures. For every candidate, we retrieve and perform photometry on the per-visit images with Source Extractor, version 2.8.6 (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 by more than three times the measurement error from the stacked photometry, the candidate is eliminated. This procedure is performed in the band in which the source photometry has the highest signal-to-noise ratio (S/N; typically the z-band for i dropouts and the y-band for z dropouts). We also reject candidates with profiles that are too compact, diffuse, or elliptical to be celestial point sources with the Source Extractor measurements on the stacked images. The eliminated sources are mostly cosmic rays, moving or transient sources, and image artifacts.

The candidates selected above are matched to the UKIDSS and VIKING catalogs within 10 in the overlapping survey area. They are then processed through the Bayesian probabilistic algorithm, and those with the quasar probability ${P}_{{\rm{Q}}}^{{\rm{B}}}\gt 0.1$ are added to the sample of candidates. 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 $0.1\lt {P}_{{\rm{Q}}}^{{\rm{B}}}\lt 0.9$, so our results are insensitive to the exact value of this cut.

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

In the present survey area covering 80 deg², we had roughly 50,000 red and point-like sources meeting the database query conditions (Equations (3) and (4)) and undetected in the g- and r-bands. The vast majority of them (∼97%) were eliminated by checking the per-visit photometry and source morphology as described above. From ∼2000 remaining candidates, the Bayesian algorithm selected 117 candidates with ${P}_{{\rm{Q}}}^{{\rm{B}}}\gt 0.1$. In this initial work, we further excluded candidates with bluer colors ($1.5\lt {i}_{{\rm{AB}}}-{z}_{{\rm{AB}}}\lt 2.0$ and ${z}_{{\rm{AB}}}-{y}_{{\rm{AB}}}\lt 0.8;$ see Figure 1), fainter magnitudes (z_AB > 24.3 mag), or with detection only in the z-band; we obtained 38 final candidates after visual image inspection.

We are also developing more classical methods of quasar color selection, such as simple color cuts and SED fitting with no Bayesian prior, to understand completeness and any possible bias in the Bayesian algorithm. A comparison of these selection techniques will be presented in a future paper. We found that the present sample of 38 candidates passed our current color cut and SED fitting selection criteria as well.

3.3. Recovery of Known Quasars

Figure 3 presents the magnitude histogram of the sample selected above. We found 36 i-band dropouts interpreted as z ∼ 6 quasar candidates, and two z-band dropouts as z ∼ 7 quasar candidates. The magnitudes of the former objects range from z_AB = 21.8 mag to the limiting magnitude of our selection, z_AB = 24.5 mag, while the latter objects are fairly bright in the y-band, 21.7 and 22.6 mag.

Zoom In Zoom Out Reset image size

Five high-z quasars in the present survey area were identified prior to our project; they are summarized in Table 1. We found that they all successfully pass our selection criteria and end up in the final quasar candidates, as marked by the dark blue cells in Figure 3. The Bayesian quasar probability is ${P}_{{\rm{Q}}}^{{\rm{B}}}=1.000$ in all cases. The figure implies that the success rate of our quasar selection is quite high at the brightest magnitudes; for example, three of the four candidates at z < 22.5 mag were indeed previously discovered high-z quasars. The five known quasars include CFHQS J0216–0455 with z_AB = 24.22 mag, which, as we discussed in Section 1, was the faintest high-z quasar known before this work.

Table 1.  Known High-z Quasars in the HSC-SSP S15A Footprint

R.A.	Decl.	iAB (mag)	zAB (mag)	yAB (mag)	Comment
02:10:13.19	−04:56:20.7	26.96 ± 0.42	22.34 ± 0.01	22.39 ± 0.04	z = 6.44; ref (1)
02:16:27.79	−04:55:34.1	27.61 ± 0.69	24.20 ± 0.07	25.81 ± 0.68	z = 6.01; ref (2)
02:27:43.30	−06:05:30.3	25.49 ± 0.12	22.05 ± 0.01	22.03 ± 0.03	z = 6.20; ref (2)
22:19:17.22	+01:02:49.0	27.78 ± 0.80	23.43 ± 0.04	23.29 ± 0.07	z = 6.16; ref (3)
22:28:43.52	+01:10:32.0	24.01 ± 0.02	22.40 ± 0.01	22.47 ± 0.02	z = 5.95; ref (4)

Note. Coordinates (J2000.0) and magnitudes were measured with the HSC data. All the objects have the Bayesian quasar probability ${P}_{{\rm{Q}}}^{{\rm{B}}}=1.000$.

References. (1) Willott et al. (2010a), (2) Willott et al. (2009), (3) Kashikawa et al. (2015), (4) Zeimann et al. (2011).

Download table as:  ASCIITypeset image

Thus our candidate selection works quite efficiently at the brighter magnitudes. We now explore the remainder of the sample with new spectroscopic observations, as described in the next section.

We carried out spectroscopic follow-up observations of the quasar candidates in the 2015 Fall and 2016 Spring semesters. One of the candidates was observed with the Optical System for Imaging and low-intermediate-Resolution Integrated Spectroscopy (OSIRIS; Cepa et al. 2000) mounted on the Gran Telescopio Canarias (GTC) in 2015 September, which led to the first discovery of a high-z quasar from our project. We also observed 19 candidates with the Faint Object Camera and Spectrograph (FOCAS; Kashikawa et al. 2002) mounted on Subaru in 2015 November and December, and identified 14 more quasars and galaxies. We further obtained additional exposures for a few of the above objects in 2016 February. The journal of these discovery observations is presented in Table 2. The details of the observations are described in the following sections.

Table 2.  Journal of Discovery Observations

Target	Date
GTC/OSIRIS; 2015 Sep
	Sep 13	Sep 14
HSC J2216–0016	90 minutes	55 minutes
Subaru/FOCAS; 2015 Nov and Dec
	Nov 3	Nov 4	Dec 6	Dec 7
HSC J0210–0523	23 minutes	...	60 minutes	...
HSC J0210–0559	80 minutes	...	45 minutes	...
HSC J0215–0555	...	40 minutesa,b	120 minutes	60 minutes
HSC J0219–0416	80 minutes	...	...	60 minutes
HSC J0848+0045	80 minutesb	30 minutesa	30 minutes	...
HSC J0850+0012	...	...	30 minutes	...
HSC J0857+0142	...	...	...	100 minutes
HSC J0859+0022	...	...	30 minutes	...
HSC J1152+0055	...	...	15 minutes	...
HSC J1202–0057	...	...	...	60 minutes
HSC J1205–0000	...	...	100 minutesa	...
HSC J1207–0005	...	...	...	40 minutes
HSC J2216–0016	...	...	...	30 minutes
HSC J2228+0128	80 minutes	...	...	...
HSC J2232+0012	60 minutes	...	...	50 minutesc
HSC J2236+0032	25 minutesa	80 minutes	...	...
Subaru/FOCAS; 2016 Feb
	Feb 13	Feb 14	Feb 15	Feb 16
HSC J0210–0523	...	...	30 minutesc	40 minutes
HSC J1202–0057	50 minutes	...	60 minutes	...
HSC J1205–0000	...	...	80 minutes	100 minutesa
HSC J1207–0005	...	60 minutes	...	...

Notes. Observing conditions:

^aPoor seeing (10 ∼ 20). ^bLow transparency. ^cLow elevation (∼30°).

Download table as:  ASCIITypeset image

4.1. GTC/OSIRIS

4.2. Subaru/FOCAS

Figure 4 displays the spectra of the identified quasars and possible quasars, while Figure 5 displays those of the non-quasars (i.e., galaxies and a brown dwarf). Their photometric and spectroscopic properties are summarized in Table 3. We present short notes on the individual quasars in Section 5.1 and on the contaminating objects in Section 5.2. Discussion and future prospects are described in Section 5.3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 3.  Spectroscopically Identified Objects

Name	R.A.	Decl.	iAB (mag)	zAB (mag)	yAB (mag)	JAB (mag)	${P}_{{\rm{Q}}}^{{\rm{B}}}$	Redshift	M1450 (mag)
Quasars and Possible Quasars
HSC J1205–0000	12:05:05.09	−00:00:27.9	>26.69	>26.45	22.61 ± 0.03	21.95 ± 0.21	1.000	6.7–6.9	−24.35 ± 0.18
HSC J2236+0032	22:36:44.58	+00:32:56.9	>27.05	23.78 ± 0.08	23.23 ± 0.05	...	1.000	6.4	−23.66 ± 0.10
HSC J0859+0022	08:59:07.19	+00:22:55.9	27.89 ± 1.06	22.77 ± 0.01	23.65 ± 0.07	>21.77	1.000	6.39	−23.59 ± 0.15
HSC J1152+0055	11:52:21.27	+00:55:36.6	25.52 ± 0.09	21.83 ± 0.01	21.61 ± 0.02	21.66 ± 0.22	1.000	6.37	−24.97 ± 0.11
HSC J2232+0012	22:32:12.03	+00:12:38.4	27.76 ± 0.51	23.84 ± 0.04	24.26 ± 0.13	...	1.000	6.18	−22.56 ± 0.34
HSC J2216–0016	22:16:44.47	−00:16:50.1	26.05 ± 0.14	22.78 ± 0.02	22.94 ± 0.03	...	1.000	6.10	−23.58 ± 0.12
HSC J2228+0128	22:28:27.83	+01:28:09.5	27.56 ± 0.41	24.06 ± 0.05	24.50 ± 0.13	...	1.000	6.01	−22.40 ± 0.12
HSC J1207–0005	12:07:54.14	−00:05:53.3	26.39 ± 0.15	24.00 ± 0.03	23.87 ± 0.08	>21.77	1.000	6.01	−22.59 ± 0.08
HSC J1202–0057	12:02:46.37	−00:57:01.7	26.13 ± 0.13	23.82 ± 0.03	23.89 ± 0.10	>21.77	1.000	5.93	−22.54 ± 0.18
Galaxies
HSC J0219–0416	02:19:29.41	−04:16:45.9	>26.74	24.32 ± 0.06	24.14 ± 0.11	>21.78	1.000	5.96	−22.50 ± 0.07
HSC J0210–0523	02:10:33.82	−05:23:04.1	25.85 ± 0.17	23.78 ± 0.06	23.54 ± 0.11	...	0.940	5.89	−22.92 ± 0.20
HSC J0857+0142	08:57:23.95	+01:42:54.6	26.24 ± 0.26	24.14 ± 0.05	23.93 ± 0.09	>21.76	0.964	5.82	−22.52 ± 0.03
HSC J0210–0559	02:10:41.28	−05:59:17.9	26.61 ± 0.27	24.25 ± 0.07	24.10 ± 0.16	...	0.996	5.82	−22.37 ± 0.06
HSC J0848+0045	08:48:18.33	+00:45:09.6	26.27 ± 0.21	23.90 ± 0.06	24.00 ± 0.10	>21.76	1.000	5.78	−22.75 ± 0.08
HSC J0215–0555	02:15:45.20	−05:55:29.1	26.03 ± 0.15	23.98 ± 0.05	23.67 ± 0.10	...	0.929	5.74	−22.66 ± 0.02
Brown Dwarf
HSC J0850+0012	08:50:02.63	+00:12:10.0	28.40 ± 1.44	24.06 ± 0.06	23.22 ± 0.05	>21.76	0.171	0.00	...

Note. Coordinates are at J2000.0. The J-band magnitudes are taken from VIKING. The magnitude upper limits are placed at 5σ significance. The errors of M₁₄₅₀ do not include the uncertainty inherent in the assumed quasar and galaxy spectral slopes.

Download table as:  ASCIITypeset image

5.1. Notes on Individual Quasars

5.2. Notes on the Contaminating Objects

5.3. Discussion and Future Prospects

In summary, we obtained spectra of 19 out of the 38 candidates, and identified 15 new high-z quasars and galaxies. The above candidates also include 5 previously known quasars, which were recovered by our selection. The current status of the spectroscopic identifications is presented in Figure 3. Our selection is quite successful at magnitudes brighter than z_AB = 23.5 mag, where seven out of the nine candidates have turned out to be quasars and the remaining two are awaiting spectroscopy. This implies the following two points: First, the source detection and measurements with the HSC hardware and reduction software are highly reliable without serious systematic effects. Second, our quasar selection algorithms work quite efficiently as long as they are fed with correct photometry information.

Interestingly, we have started to find high-z galaxies as significant contaminants at z_AB > 23.5 mag. This is not surprising because the luminosity functions of quasars and LBGs are likely to intersect at ∼24 mag, with LBGs outnumbering quasars at fainter magnitudes (see Figure 2 and the discussion in Section 3). Figure 6 presents the difference between the PSF and cModel magnitudes, as a measure of source extendedness, for all identified objects in this work. Because the host galaxy contribution is not always negligible for the low-luminosity quasars we are looking for, and the extendedness is a noisy quantity at the faintest magnitudes, no clear cut can be defined to separate quasars and galaxies on this plane. Indeed, the three possible quasars with relatively narrow Lyα lines (J2232+0012, J2228+0128, J1207–0005; see Figure 4) appear to have larger extendedness than the remaining quasars, suggesting significant light from the host galaxies.

Zoom In Zoom Out Reset image size

Meanwhile, the discovered galaxies are an important probe of the reionization era. For example, stellar populations in such high-z bright galaxies can be studied in detail with high-quality spectra. Measurements of the interstellar absorption lines such as those observed in J0857+0142 and J0215–0555 have the potential to constrain the escape fraction of ionizing photons (Jones et al. 2013). Follow-up observations with facilities at other wavelengths (e.g., the Atacama Large Millimeter/submillimeter Array; ALMA), would also be useful to understand the nature of these galaxies in the high-z universe.

At the faintest magnitudes of our survey (24.0 < z_AB < 24.5 mag), the photometric selection and spectroscopic identification become more challenging. Along with the increasing fraction of galaxies, we found a contaminating brown dwarf, which is expected from its low Bayesian quasar probability (${P}_{{\rm{Q}}}^{{\rm{B}}}$ = 0.17; see Section 5.2). We were not able to confirm the nature of two candidates due to their low spectral S/N, although we typically spent a few hours per object with Subaru. They seem to have no strong emission lines and so may be galaxies or brown dwarfs, but weak-line quasars such as J2236+0032 are known to exist. Further analysis of these objects, possibly with additional observing time to increase the S/N, will be presented in a future paper.

With the five previously known objects and the nine newly identified objects (including the possible quasars J2232+0012, J2228+0128, and J1207–0005), fourteen high-z quasars are now known in the present survey area. This is roughly half of the expected number with our survey limit in 80 deg² (see Section 2), although the spectroscopic identification is still not complete. In the last column of Table 3, we report the absolute magnitudes of the discovered quasars as well as galaxies. For quasars, we measure the flux densities at the rest-frame wavelength λ_rest = 1270–1330 Å (except for J1205–0000 with the highest redshift, in which case we adopt λ_rest = 1250–1260 Å and assume z = 6.75) and estimate M₁₄₅₀ assuming the power-law continuum slope α = −1.5 (f_λ ∝ λ^α; e.g., Vanden Berk et al. 2001). For galaxies, we measure M₁₃₅₀ at λ_rest = 1320–1380 Å, and convert them to M₁₄₅₀ by assuming the UV spectral slope of β = −2 (Stanway et al. 2005). As expected, our survey has succeeded in identifying quasars (and galaxies) with luminosities approaching M₁₄₅₀ ∼ −22 mag (i.e., ∼2 mag lower luminosity than found in most of the previous large surveys; see Section 1).

We will continue to develop the SHELLQs project as the HSC-SSP survey continues. The present work only partially covers the first 80 deg² of the Wide layer, which will eventually observe 1400 deg². Our immediate goal is to complete the spectroscopic identification in this 80 deg² area and derive our first quasar luminosity function. In the long term, we will expand the survey area and significantly increase the sample size and luminosity range of known high-z quasars. As described in Section 1, the expected numbers of quasars over the whole Wide area are ∼500 with z_AB < 24.5 mag at z ∼ 6 and ∼100 with y_AB < 24.0 mag at z ∼ 7. We will also explore even lower luminosities with the Deep and Ultra-Deep layers of the HSC-SSP survey, although follow-up spectroscopy will become more challenging.

At the same time, it is important to follow up individual quasars in greater detail. The redshifts of some of the discovered quasars are poorly constrained at the moment, which should be improved. We are also planning deep optical and NIR spectroscopy to measure the near-zone size, SMBH mass, and metallicity for those quasars at lower luminosity than previously known at z > 6. These low-luminosity objects are expected to be much more numerous than the brighter ones, and hence possess critical information about the general properties of quasars in the early universe. They will also provide a useful constraint on the low-mass end of the SMBH mass function, and in turn, models of the formation and early evolution of SMBHs. In addition, X-ray observations will play a critical role to estimate the bolometric luminosity, the Eddington ratio, and the presence and properties of absorbing material. We also plan to conduct ALMA follow-up observations to study the gas and dust content, as well as the star formation activity in the host galaxies.

We present initial results from the SHELLQs project, a survey of low-luminosity quasars and AGNs at high redshift close to the reionization era. The project exploits the exquisite imaging data with five optical bands (g, r, i, z, and y) produced by the Subaru HSC-SSP survey, supplemented with NIR photometry where available from UKIDSS and VIKING. The limiting magnitudes of the quasar search are currently set to z_AB < 24.5 mag and y_AB < 24.0 mag, but these may change in the future. The candidates are selected by combining several photometric approaches, including a Bayesian probabilistic algorithm, which have turned out to be quite efficient in eliminating astrophysical contaminants such as stars and dwarfs, as well as cosmic rays, moving objects, and transient events. From the early HSC-SSP survey area covering 80 deg², we identified 38 candidate high-z quasars, which are the focus of this paper.

We carried out spectroscopic follow-up observations of 19 of these candidates, with GTC/OSIRIS and Subaru/FOCAS, in the 2015 Fall and 2016 Spring semesters. Nine objects were identified as quasars or possible quasars at 5.9 < z < 6.9, based on the sharp continuum breaks characteristic of GP troughs, broad Lyα and N v λ1240 lines, and/or blue continuum. Six objects are likely high-z galaxies with interstellar absorption lines of Si ii λ1260, Si ii λ1304, and C ii λ1335, and in some cases narrow Lyα emission lines. The remaining objects include a L0 dwarf, a moving or transient object, and two sources whose nature is still uncertain due to the low spectral S/N. In addition to these newly identified objects, five quasars were known prior to our survey among the 38 candidates. The success rate of our selection is quite high, and most of the objects we took spectra of were identified as high-z quasars or galaxies.

The SHELLQs project will continue as the HSC-SSP survey continues toward its goals of observing 1400 deg² in the Wide layer, as well as 27 and 3.5 deg² in the Deep and Ultra-Deep layer, respectively. We will soon deliver our first quasar luminosity function reaching down to M_AB ∼ −22 mag at z ∼ 6. Further follow-up observations of the discovered quasars and galaxies are being considered at various wavelengths from submillimeter/radio to X-ray.

We thank the referee for his/her useful comments and suggestions to improve this paper. We are grateful to everyone involved in the hardware development, observations, and data reduction for the HSC-SSP survey. We had a lot of great help from Chien-Hsiu Lee, Takashi Hattori, and other Subaru staff members for the FOCAS observations. NK acknowledges support from the Japan Society for the Promotion of Science (JSPS) through Grant-in-Aid for Scientific Research 15H03645. KI acknowledges support by the Spanish MINECO under grant AYA2013-47447-C3-2-P and MDM-2014-0369 of ICCUB (Unidad de Excelencia "María de Maeztu"). TN acknowledges financial support from the JSPS (KAKENHI grant no. 25707010) and also from the JGC-S Scholarship Foundation.

The HSC collaboration includes the astronomical communities of Japan and Taiwan, and Princeton University. The HSC instrumentation and software were developed by the National Astronomical Observatory of Japan (NAOJ), the Kavli Institute for the Physics and Mathematics of the universe (Kavli IPMU), the University of Tokyo, the High Energy Accelerator Research Organization (KEK), the Academia Sinica Institute for Astronomy and Astrophysics in Taiwan (ASIAA), and Princeton University. Funding was contributed by the FIRST program from Japanese Cabinet Office, the Ministry of Education, Culture, Sports, Science and Technology (MEXT), the Japan Society for the Promotion of Science (JSPS), Japan Science and Technology Agency (JST), the Toray Science Foundation, NAOJ, Kavli IPMU, KEK, ASIAA, and Princeton University.

This paper makes use of software developed for the LSST. We thank the LSST Project for making their code available as free software at http://dm.lsstcorp.org.

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, Eotvos Lorand University (ELTE), and the Los Alamos National Laboratory.

Facilities: Subaru - Subaru Telescope, GTC - .