Open Access
Issue
A&A
Volume 654, October 2021
Article Number A79
Number of page(s) 18
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/202140790
Published online 15 October 2021

© J. Greiner et al. 2021

Licence Creative CommonsOpen Access article, published by EDP Sciences, under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Open Access funding provided by Max Planck Society.

1. Introduction

The clustering of quasars provides one of the most important observational constraints to probe their physical properties, their formation and evolution (Haiman & Hui 2001), and the mass of their dark matter (DM) halos (Sheth et al. 2001). As easily detectable objects throughout the Universe, quasars trace the underlying dark matter distribution (e.g., Haehnelt & Nusser August 1998), and thus provide a powerful test of hierarchical structure formation theory (e.g., Fang 1989). How this structure formation in the early Universe proceeds in detail is far from understood. How many massive galaxies form in a single massive dark matter halo? Do all of these massive galaxies have a massive black hole (BH)? Do all these BHs accrete from the rich gas reservoir, and what are the duty cycles of QSO activity? Do massive galaxies form sub-structures and/or multiple massive BHs?

Particularly pressing issues over the last years are the formation channels and efficiencies of massive BHs, and their growth rates with cosmic time (e.g., Johnson et al. 2013; Valiante et al. 2017, 2018; Inayoshi et al. 2020; Kroupa et al. 2020). Quantifying the number of QSO pairs at high redshift is one observational avenue to constrain a combination of parameters of BH formation and BH growth rate. High-z QSOs can be as massive as the most massive BHs found in the local Universe (e.g., Bañados et al. 2018; Yang et al. 2020; Wang et al. 2021), and so must have grown several orders of magnitude in less than 800 Myr. Black hole formation mechanisms predict the formation of BH seeds with masses of a few ∼105−6M at most (Volonteri 2010; Inayoshi et al. 2020). If the growth of quasars was mostly through gas accretion, then they must have grown at the Eddington rate for a large fraction of their lives. Finding QSO pairs, or the absence of them, can help us understand if BH mergers play a role in their growth, and if dense environments can sustain efficient growth for several QSOs in the same region of the sky or if baryonic processes (e.g., AGN feedback) prevent them from doing so (Wyithe et al. 2005; Kashikawa et al. 2007; Utsumi et al. 2010; Costa et al. 2014; Habouzit et al. 2019, for QSO feedback preventing galaxy formation and/or growth in their surroundings). Radiation from a QSO can impact a region from sub-Mpc to tens of Mpc, depending on the QSO mass, accretion rate, and properties of its surrounding environment, and could therefore also inhibit the accretion activity of nearby BHs. In addition, quantifying the number of QSO pairs also provides us with constraints on BH formation efficiency (i.e. whether BH formation takes place in a large number of halos), but also on whether the formation of massive seeds could be clustered. The existence of many pairs of QSOs could indeed imply that massive seeds can form close to each other. Constraints from QSO pairs on QSO origins and growth with time are likely degenerated, but are key to improving our understanding of these extreme objects.

Quasar clustering has been investigated at different scales and depths, among others using the Sloan Digital Sky Survey (SDSS) (Shen et al. 2007; Hennawi et al. 2010), the 2dF QSO Redshift and Dark Energy Survey (DES) (Porciani et al. 2007; Chehade et al. 2016), or the Canada-France-Hawaii Telescope Legacy Survey (CFHTLS) (McGreer et al. 2016). Studies at different depths for a given redshift range are important since a large luminosity dependence is theoretically predicted at higher redshifts (Hopkins et al. 2007). Clustering at large scales, sampling megaparsec or larger distances, is stronger at high redshifts (z > 3.5) than at low redshifts (2.9 < z < 3.5) (Shen et al. 2007). Clustering at small scales, sampling the 10 kpc–1 Mpc environment, also increases from lower (z ∼ 3) to higher (z ∼ 4) redshift, but shows a much stronger amplitude on kiloparsec scales than on megaparsec scales (e.g., a factor 4 stronger at z ∼ 2) (Eftekharzadeh et al. 2017).

Clustering measurements at high redshift (z > 3) come up against a problem: close quasar pairs are extremely rare. For comparison, the mean quasar separation at z > 3 is ∼150 h−1 Mpc (given the standard luminosity function of Φ ∼ 10−6 Mpc−1 mag−1 for M(AB) < −24; e.g., Willott et al. 2010; Ross et al. 2013; Matsuoka et al. 2018; Kulkarni et al. 2019), and on small scales (≲1 Mpc) the correlation function does not increase as fast as the volume decreases (Hennawi et al. 2010). However, direct observations of pairs might suggest differently; the three QSO pairs at the highest redshifts were all found serendipitously: (i) Schneider et al. (2000) reported a pair at z = 4.26 with a separation of 33″ (corresponding to 160 h−1 kpc proper on the sky) which was found in the slit while spectroscopically confirming another QSO candidate; (ii) Djorgovski et al. (2003) discovered a z = 4.96 quasar at a separation of 196″ (or proper transverse separation of 0.9 h−1 Mpc) from a z = 5.02 quasar discovered by Fan et al. (1999); and (iii) McGreer et al. (2016) reported a pair at z = 5.02 with a separation of 21″ (90 h−1 kpc), discovered in a faint z ∼ 5 quasar (but not pair) search down to i = 23 mag, and selected simultaneously with identical colour criteria. Given the relatively small sample sizes which these discoveries are based upon, they suggest that quasars are even stronger clustered at redshifts > 4 on small scales. This in turn has been used to argue that feedback is very inefficient at high redshift (Willott et al. 2010; McGreer et al. 2016), since inefficient feedback would correspond to the correlation length to flatten out at high redshift, while concordant QSO and host halo mass growth would imply a sharply rising correlation length. In addition, merger models of galaxy formation predict a large fraction of binary QSOs, at least over some fraction of their life time (e.g., Volonteri et al. 2003a), with the fraction of binary (off-centre and dual) QSOs increasing with redshift (Volonteri et al. 2016). This state of knowledge, and the importance of the conclusions that can be drawn, certainly warrants further searches for clustering properties at high redshift.

Here we search within a sample of 116 spectroscopically confirmed QSOs at redshift 6 (collected from the compilation of Bañados et al. 2016) for a companion. Throughout this paper we use the best-fit Planck cosmological model with Ωm = 0.286, ΩΛ = 0.714, and h = 0.69 (Planck Collaboration XIII 2016). At a mean sample redshift of 6, an angular separation of 1″ corresponds to a proper (comoving) transverse separation of 4 h−1 kpc (28 h−1 kpc). All magnitudes are in the AB system.

2. Selection criteria, observations, and data analysis

2.1. Selection

From the sample of 173 spectroscopically confirmed QSOs at z > 5.6 compiled in Bañados et al. (2016) we selected all 144 sources south of Dec. <  + 30°, thus being visible from La Silla (Chile). We added six objects from Mazzucchelli et al. (2017b), and two object from Jiang et al. (2016). For the observing campaigns we prioritised brighter sources, so that it was easier (less time consuming) to reach limiting magnitudes fainter than the spectroscopically confirmed QSOs. After weather and technical losses, we obtained useful new Gamma-ray Burst Optical/Near-infrared Detector (GROND) observations for 77 sources, in addition to the 42 objects from the Bañados et al. (2016) sample, which had already been observed with GROND in 2013–2015 during the selection process towards their spectroscopic sample. Excluding in the following two QSOs at z > 6.6, our sample of 116 QSOs spans redshifts in the range 5.6–6.6 and z′-band brightness values in the range 18.6–23.1 mag.

2.2. Imaging observations and data analysis

2.2.1. Strategy, data analysis and source detection

Simultaneous imaging in grizJHKs with GROND (Greiner et al. 2008) of the 77 new fields was performed at six different time slots: 2016 August 14 – September 3, 2016 October 11–25, 2017 January 24 – February 7, 2017 April 22 – May 2, 2017 September 18–21, and 2018 March 23. Since most fields have SDSS or Pan-STARRS coverage, no further observations for photometric calibration were obtained. The three remaining fields were calibrated against the SkyMapper Survey (DR1.1).

Since the field of view (FOV) of the four visual channels griz′ is smaller than that of the near-infrared (NIR) channels (Fig. 1), the former define the search distance from the spectroscopically identified QSO to 80″ distance. Thus, our maximum search radius for a QSO companion is 0.471 h−1 Mpc proper or 3.3 h−1 Mpc comoving (see Fig. 1). This accounts for the fact that most of the primary QSOs are not exactly centred in the GROND FOV.

thumbnail Fig. 1.

Schematic of the GROND optical (griz′) and NIR (JHK) channel spatial coverage at a redshift of 6. The dashed boxes are the respective griz′ FOV for each individual telescope dither position; a similar pattern applies to the JHK channels (not shown). With most of our targets being within 1 arcmin from the centre of the FOV, and since we require the riz′ bands, our conservative search radius is 80″, corresponding to about 471 h−1 kpc proper, or 3.3 h−1 Mpc comoving.

The question to answer before the start of our observing campaign was how much deeper to look for a companion, relative to the prime QSO. We chose a 2 mag difference as the threshold, based on three factors: (i) given the QSO luminosity function (Willott et al. 2010; Ross et al. 2013; Shen et al. 2020), this corresponds to a factor of more than 10 in the number of objects, and should imply a large enough sample depth; (ii) optical variability is rare above ∼0.5 mag amplitude, ensuring that the effective difference is not diminished substantially; and (iii) the pair at z = 5 found by McGreer et al. (2016) has a similar i′-band magnitude difference (19.4 versus 21.4 mag). Thus, exposure lengths of typically 20 minutes were decided upon, ensuring that the 3σ limiting magnitude in the z′ band was at least 2 mag deeper than the brightness of the spectroscopically identified QSOs (see below). While this 2 mag difference might suggest that the search could have been done just on Pan-STARRS data with its typical 5σ depth of z′ < 22.3 mag, only 11 PS1 QSOs are bright enough to also survive the colour cut of i′– z′ > 1.5 mag at that faint z′ brightness (see below).

GROND data were reduced in the standard manner (Krühler et al. 2008) using pyraf/IRAF (Tody 2019; Küpcü et al. 2008). The optical imaging data (griz′) was calibrated against the Sloan Digital Sky Survey (SDSS)1 (Eisenstein et al. 2011), Pan-STARRS12 (Chambers et al. 2016) or the SkyMapper Survey3 (Wolf et al. 2018) catalogues, and the NIR data (JHKs) against the 2MASS catalogue (Skrutskie et al. 2006). This results in typical absolute accuracies of ±0.03 mag in griz′ and ±0.05 mag in JHKs. Since the GROND dichroics were built after the Sloan filter system (Greiner et al. 2008), the colour terms are very small, below 0.01 mag, except for the i′ band which is substantially narrower than the SDSS i’ band. All of our analysis was done in the SDSS system, and the following transformations were used: , , , . For the fields calibrated against Skymapper and PS1, we used their conversion to SDSS (Scolnic et al. 2015). The PS1 z′ band has a substantial colour term ( = − 0.214 × () due to the missing tail beyond 920 nm, but this becomes important only for redshifts above 6.6 when Lyman-α emission (Ly-α) moves beyond the z′-band limit.

The calibrated z′- and J-band images for each field were stacked after re-sampling due to their different pixel sizes (015 for the visual, and 059 for the NIR); for the four objects at z > 6.5 we just used the J-band images. A source detection procedure on each stacked image then provided a list of all objects per field, which was then used as input for ‘forced’ photometry on the individual seven-band images. This provides upper limits for the bands where the sources are not detected; otherwise it performs point spread function (PSF) photometry for griz′, and aperture photometry in JHK (with a radius of 1 × FWHM) at the position of the source in the master catalogue which is allowed to re-centre by ±0.3 of the PSF width. Table A.1 contains the details of the observation for each QSO, including the 3σ limiting magnitude in the z′-band. Table A.2 contains the foreground AV-corrected AB magnitudes of all 116 QSOs in the individual GROND filter bands.

2.2.2. Colour-colour selection of candidates

With these GROND data in hand, the goal was to search the field around each spectroscopically confirmed QSO for a nearby source with a similar colour, taking advantage of the simultaneous multi-filter optical–NIR spectral energy distributions (SEDs) of all objects inside the FOV.

We implicitly assume that the searches leading to the discovery of the QSOs pick the brighter of the pairs; this is corroborated by modelling which suggests that current flux-limited surveys of QSOs at high redshift preferentially detect objects at their peak luminosity, and thus miss a substantial population of similarly massive BHs accreting at lower accretion rates (Costa et al. 2014). Thus, here we are looking for objects fainter than the original QSO.

In all 116 fields we apply the following two colour selections (all in AB): i′– z′ > 1.5 mag, and z′– J > −0.7 mag. We checked that there are no detections in the g′ and r′ bands. For the z > 6.5 objects, we adjust the z′– J colour to −0.5, add J − H < −0.5, and require a i′ non-detection. This returns 74 new high-z quasar candidates. We excluded 48 candidates through visual inspection of the seven-channel image cut-outs of these candidates, based on either source confusion (i.e. a nearby source with overlapping PSF or an extended source that could be an interloper galaxy) or a non-resolved equally bright pair of objects. The remaining 26 candidate sources are shown in Fig. 2 as red dots (light and dark), overplotted on the distribution of the spectroscopically confirmed quasars (shown in blue), which by itself show that our colour cut criteria recover all but 13 objects, most of which fail due to non-detection (i.e. upper limits in the i′-band, but deeper i′-band exposure would likely recover those as well).

thumbnail Fig. 2.

Colour-colour diagram in the GROND-intrinsic izJHKs bands of the prime quasars (in blue; triangles denote limits) compared to normal foreground stars and galaxies (in grey) of the same fields, with selected brown dwarfs from earlier GROND observations (in green and purple, and labelled with their spectral types). Our 26 candidates are shown in red; the nine objects which have ALLWISE detections are in light red. The yellow-shaded region indicates the search region for the QSO pair. The arrow at the top right of each panel shows the amount by which a correction for AV = 1 would shift an object. Except for the grey objects, all are foreground AV-corrected.

For 9 of the remaining 26 candidate sources we found catalogue entries in the ALLWISE all-sky catalogue (Wright et al. 2010; Mainzer et al. 2011) within 1 arcsec distance, and after a visual check of the images we accepted them as matching objects. Tables A.4 and A.5 contain the photometry of these 26 candidates, separately for those with and without ALLWISE matches.

We then performed Le PHARE (Arnouts et al. 1999; Ilbert et al. 2006) fits to the photometry of the 26 candidates (including the ALLWISE magnitudes when available). We included an alternative IGM opacity model (Songaila 2004) in addition to the default one (Madau et al. 1999), but otherwise used the default spectral templates except for adding about 200 cloned high-z QSO templates based on stacked observed SDSS spectra of low-z QSOs, similar to Willott et al. (2007a). The goal of these Le PHARE fits was to distinguish late-type dwarfs from QSO candidates, where the distinction is based on two observational features: the steepness of the SED in the i′-z′-J range, and the SED shape in the HK range which is much redder for dwarfs. Since dwarfs in our sensitivity range would be at distances of order 100 pc, their photometry would not be affected by interstellar dust. We thus created two SEDs for each candidate, one without AV-correction (for Le PHARE fits with dwarf templates) and one with full galactic AV-correction (for Le PHARE fits with QSO templates). For 14 of the 26 candidates the best-fit template is a dwarf (see last column of Tables A.4 and A.5), for 10 candidates a QSO at high redshift (labelled in boldface in Table A.5), for 1 candidate a galaxy at intermediate redshift, and for 1 candidate (SDSSJ2054−0005_4) the reduced χ2 does not provide a clear favorite. The Le PHARE fits of the ten high-z quasar candidates (i.e. with AV-correction applied) are shown in Fig. 6, which also contains the best-fit photo-z (sixth field in the red line of labels), all within 13% of the spectroscopic redshift except for PSO J055.4244-00.8035 (the source whose spectrum we did not obtain).

2.3. Spectroscopic observations

We obtained spectroscopy of 11 objects (9 of the 10 QSO candidates, the galaxy, and the unidentified object), mostly with the GMOS instruments of the Gemini Observatory (under proposal IDs GN-2019B-FT-201 and GS-2019B-FT-202). One object was observed with the Folded-port InfraRed Echellette (FIRE) spectrograph in Prism mode at the Magellan Baade telescope. A log of the observations is given in Table A.3. At Gemini-North (South), a Hamamatsu detector was used with the R150 grating and filter G5308 (G5326). The spectra cover the range from 615 (750) to 1080 nm. Wavelength calibration was provided by CuAr comparison lamps. Flux calibration was done against the standard stars G191B2B (LTT7987). For the FIRE spectrum, we used archival A0V stars for telluric and flux calibration. The data were reduced in the standard manner, using GMOS specific routines provided within the IRAF package (Tody 2019), and a custom-made python code for the FIRE observation.

For object PSOJ071.0322−04.5591 no trace is visible in the 2D spectrum; for the other sources the optimally extracted 1D spectra are shown in Fig. 7. None shows a sign of the Ly-α ‘step’ at the wavelengths suggested by the photometric redshifts (vertical lines in Fig. 7).

3. Results and discussion

3.1. Pair candidates

We do not find a single QSO pair candidate brighter than M1450(AB) < −26 mag in our GROND data of 116 spectroscopically confirmed redshift 6 quasars. Except for two candidate objects (near PSOJ055.4244−00.8035 and PSOJ071.0322−04.5591) we obtained optical spectroscopy for all the other Le PHARE-derived QSO pair candidates, but could not confirm the quasar nature suggested by the colour selection and SED fitting. This is somewhat surprising, since the (ex post facto) photometric redshifts we derived with Le PHARE for the prime QSOs where in good agreement with the spectroscopic redshifts reported by Bañados et al. (2016), with only a 5% fraction mismatch. We consider it very unlikely that there is an unrecognised quasar among our identified candidates in Tables A.4 and A.5. This is in contrast to our empirical expectation (at the start of the project) of three pairs in our sample, based on the serendipitously discovered pair at z = 4.26 (Schneider et al. 2000) and the two pairs at z = 5.02 (Djorgovski et al. 2003; McGreer et al. 2016) (see details in the introduction), which when averaged together (1/100, 1/14, 1/47) suggest one pair every 40 QSOs, or 2.5%. This line of reasoning includes a few simplifications. First, it uses ex post facto arguments and guesses on the search volume of the above serendipitous discoveries. Next, it assumed that the quasar clustering fraction does not evolve with redshift, the question to be answered by the project. It also ignored the different luminosity ranges sampled at different redshifts; the integral over the luminosity function is certainly the most sensitive component contributing to the wp statistic (see below).

In order to put this in context, we follow Hennawi et al. (2006) and Shen et al. (2010, in particular their Eq. (4)) in estimating the projected correlation function, or wp statistics, as defined by Davis & Peebles (1983), among others. Given the mean seeing in our images and correspondingly the ability to separate two sources, we take an inner radius of 100 h−1 kpc (corresponding to 3″) for the projected comoving area of the cylindrical annulus (the outer one being our search radius of 3.3 h−1 Mpc). We use the latest luminosity function description of Kulkarni et al. (2019), as well as their publicly provided tools4 to convert our observed z′(AB) magnitude limit and mean redshift to M1450, and applying the corresponding bolometric correction (using their ‘global fit A’; we verified our procedure against their Fig. 5). This results in a number density of 9 × 10−10 Mpc−3 for quasars with M1450 < −26 mag, for our redshift range 5.6 < z < 6.6. We obtain an upper limit of wp < 110 000 h−1 Mpc (see Fig. 3), where we use a formal upper limit of our search of one pair out of the 116 objects searched for, given the missing spectroscopy for PSOJ055.4244−00.8035. The figure also includes measurements at lower redshifts from Schneider et al. (2000), Shen et al. (2010), and McGreer et al. (2016) (modified by the updated luminosity function and corrected for our search scale according to the γ = 2 power-law dependence of wp(R)∝Rγ from Shen et al. 2010). Overall, our upper limit corresponds to a similar wp as previous findings at lower redshift, suggesting that there is no strong rise of clustering towards high redshift.

thumbnail Fig. 3.

Our upper limit (red triangle) on the projected correlation function of z ∼ 6 QSOs, as derived from the GROND observations. The two filled circles are the corresponding findings by Shen et al. (2010) for 7 and 8 pairs, respectively, modified for the different luminosity function (and cosmology) used over time, and adapted to our search scale of 0.1–3.3 h−1 Mpc according to the −2 power-law scale dependence. Open symbols indicate the serendipitous discoveries from Schneider et al. (2000), Djorgovski et al. (2003), and McGreer et al. (2016), also extrapolated to our search scale. The horizontal error bars show the corresponding redshift range of the studies. The labels at the data points report the limiting i-/z-band AB magnitude of the search (see Col. 5 of Table 1) and the corresponding absolute M1450(AB) limiting magnitude. The two black dashed lines depict wp(R) for the same rate of 7 and 8 pairs, as in Shen et al. (2010), extrapolated to higher redshift with their same M1450(AB), and unchanged luminosity function, suggesting higher wp(R) at higher redshift. Similarly, the dotted line shows the extrapolation for the case of constant limiting z-band magnitude. The red line depicts the backwards extrapolation of our upper limit to lower redshifts, at our M1450(AB) < −26 mag sensitivity, demonstrating the mismatch with respect to the findings at lower redshift. All the lines are computed via integration over the luminosity function of Shen et al. (2020) in redshift steps of 0.25, and proper redshift-dependent transformation between observed z′ and absolute magnitudes.

Table 1.

Comparison of search volumes and depth for previous, serendipituously found high-z QSO pairs.

Using Eq. (5) in Shen et al. (2007) and γ = 2, our upper limit on wp corresponds to a limit of the auto-correlation length ro < 1000 h−1 Mpc.

3.2. Completeness and biases

In the past clustering studies have typically relied on selecting pairs from large catalogues (e.g., SDSS), and the most important question is about the completeness of the underlying catalogue (e.g., Hennawi et al. 2006; Eftekharzadeh et al. 2017). Here, we obtained follow-up observations of known quasars at known redshifts. Thus, in our case the question of completeness reduces to the questions of (i) whether or not the selection criteria which led to the sample of these known redshift quasars had a bias against selecting binary quasars and (ii) whether our search strategy misses QSOs. As to item (i), one potential bias is that the seeing limit in the original survey data affects the discovery of QSO on 1–2″ scales due to nearby contaminating sources. As this is a statistical problem, it is not expected to affect the fraction of binary QSOs in the spectroscopically confirmed QSO sample.

As to item (ii), this splits into two sub-questions. First, what is the completeness of our colour-colour cuts? Since we ran our colour-colour selection blindly on all sources in each observation, we can answer this with the recovery rate of the spectroscopically confirmed QSO: out of the 116 QSOs, we missed 9: 3 not detected in z′, and the other 6 due to insufficiently deep i′-band limits. Since the three non-detections are due to variability, they are not counted here (see below on the variability bias), so our incompleteness of the colour-colour selection is about 5%. Second, what is the completeness of the Le PHARE fits in recognizing the QSO nature? Again, this can be answered with the sample of the spectroscopically confirmed QSOs. Ignoring sources with detections in fewer than three photometric bands, we ran Le PHARE fits on the remaining 97 spectroscopically confirmed QSOs in the same manner as with the candidates (i.e. two runs, one with and one without AV correction). Only 5 sources are classified as dwarfs, the other 92 as QSOs. We double-checked that the colours and photometric errors of the 5 sources are not outliers with respect to the sample distribution of the 92 correctly classified QSOs. So this corresponds to only < 5% incompleteness.

Further, we note that it is unlikely that even substantial optical variability of the QSOs would affect the fraction of pairs, since statistically the same number of ‘risers’ versus ‘faders’ is expected. The only exception would be effects due to Malmquist bias, but since only a handful of objects are detected with a z′-band uncertainty larger than 0.2 mag, this imbalance is unimportant.

Lastly, a potential bias could be due to dust obscuration of AGN, in our case obscuring the companion of the spectroscopically confirmed QSO. Again, this is not expected to be a major issue. First, previous suggestions were that obscuration could primarily happen in the highest mass accretion phases; one would first expect the brighter object of the pair to be obscured, rather than the lower-luminosity companions of the unobscured spectroscopicall confirmed QSOs. Second, while ALMA observations revealed substantial dust at high redshift, there is no evidence that these dusty objects are obscured AGN (see Sect. 3.4 below). Finally, theory of dust production fails to explain dust in large amounts at z > 5 (Leśniewska & Michalowski 2019), preventing quantitative estimates or simulations.

Another potential bias could come in if BHs of similar mass have different Eddington ratios at different redshifts, such that only the optically brightest QSOs are found at high redshift. However, Mazzucchelli et al. (2017b) show that (at least for the biased high-z sample they considered) BH masses and Eddington ratios of the z > 6.5 and SDSS 0.35 < z < 2.25 samples are consistent in a luminosity-matched sample. Thus, our assumption was that the clustering fraction at redshift 4-5 (Schneider et al. 2000; McGreer et al. 2016) is not much different than at redshift 6, leading to our expectation of more pairs compared to our findings.

Furthermore, our search comes with the biases that the luminosity difference between the two members of the pair can be larger than our selected 2 mag, for at least two different reasons. Firstly, the BH masses and/or accretion rates can differ substantially, and secondly the Ly-α emission, which contributes to the z-band magnitude, can be drastically different between the two members of the pair, replicating the differences in the spectroscopically confirmed QSO sample (see e.g., Bañados et al. 2016). Regarding the first reason, simulations suggest differently. Based on the large-scale cosmological hydrodynamical Horizon-AGN simulations, Volonteri et al. (2016) find that the luminosity ratio between the central BH (the more massive one) and an off-centre BH is within a factor of ten (rms = 5x) for the cases of Lbol > 1043 erg s−1 and BH masses up to 108 M. While this corresponds to the ratio we employed, it should be noted that this ratio depends on selected input parameters for the simulation, among others a mass ratio < 1:6 for the merging galaxies at which AGN activity is triggered in both BHs. Similarly, Bhowmick et al. (2019) find that in the MassiveBlack II simulations the satellite quasar luminosity is similar to that of central quasars. Finally, we note that for 41 of our targets the 3σ limiting z′ magnitude reached in our observations is actually 3 mag fainter than that of the spectroscopically confirmed QSO, which still would lead to an expectation of one pair in our sample.

3.3. Comparison to simulations

The upper limit to the projected correlation function for bright QSO pairs at z ∼ 6 presented in Sect. 3.1 serves as an important constraint for galaxy evolution models. Currently, cosmological-scale simulations typically either focus on quasar properties below z ∼ 5 due to improved statistics (e.g., DeGraf & Sijacki 2017; Bhowmick et al. 2019; Habouzit et al. 2021) or above z ∼ 7 due to reduced computational limitations (e.g., Di Matteo et al. 2017; Tenneti et al. 2019; Marshall et al. 2020). Zoom-in simulations of halos containing QSOs at z ∼ 6−7 have been performed (e.g., Costa et al. 2014; Lupi et al. 2019); however, they do not readily provide statistics on large scales. Consequently, model predictions for the total number of bright QSO pairs at the redshift of our sample are still lacking in the literature.

At lower redshifts models suggest that bright QSO pairs are rare. For example, in the MASSIVEBLACK-II simulation box of volume (100 h−1 Mpc)3, Bhowmick et al. (2019) find no pairs of QSOs with g < 20.85 within a search radius of 4 Mpc at z = 2, and only three such systems by z = 0.6.

At higher redshift, work on the BLUETIDES simulation has shown that QSOs bright enough to match observations are also very rare, even when accounting for large BH seeds (Di Matteo et al. 2017). There is only one BH of mass above 108 M present in the BLUETIDES (400 h−1 Mpc)3 box at z = 8, which has a luminosity of Lbol ∼ 2.2 × 1046 erg s−1 (Di Matteo et al. 2017; Tenneti et al. 2019).

However, despite the lack of pairs, galaxy evolution models still find that brighter QSOs cluster more strongly than their fainter counterparts, particularly on smaller scales. For example, on scales of ∼10 kpc at z = 4, DeGraf & Sijacki (2017) find the two-point auto-correlation function, ξ(r), for AGN with Lbol ≥ 1044 erg s−1 in the ILLUSTRIS simulation is higher by a factor of three compared to that for all AGN. On the larger scales probed by our observational measurement (∼0.1−3.3 h−1 Mpc) this enhancement is reduced, but still remains present for the brightest objects, as also seen in the L-GALAXIES semi-analytic model (Bonoli et al. 2009) and the HORIZON-AGN simulation (Volonteri et al. 2016). All three of these models report a value of ξ(r = 1.2 h−1 Mpc)∼10 at z = 4 for AGN with Lbol ≥ 1044 erg s−1, with the expectation that this value will be higher at z = 6, as also seen in the observations discussed in Sect. 1.

In order to better probe the exact epoch and search volume of our observational QSO sample, we carried out a census of bright AGN pairs at z = 6 in the ILLUSTRIS-TNG300 (hereafter TNG300) and L-GALAXIES 2020 galaxy evolution models. TNG300 (Pillepich et al. 2018; Springel et al. 2018; Naiman et al. 2018; Marinacci et al. 2018; Nelson et al. 2018) is a magneto-hydrodynamical simulation run in a (205 h−1 Mpc)3 box, using a Bondi-Hoyle-Lyttleton BH accretion model with no boost factor but an assumed BH seed mass of 8 × 105 M (see Habouzit et al. 2021). L-GALAXIES 2020 (Henriques et al. 2020) is a semi-analytic model run on the larger (480 h−1 Mpc)3 MILLENNIUM-I box (Springel et al. 2005), but with a more simplified phenomenological model of quasar- and radio-mode BH accretion and no BH seed mass (see Croton et al. 2006). In both cases we assume that all AGN are radiatively efficient when calculating their luminosities, which provides an upper limit on the expected number of bright pairs. Therefore, the AGN bolometric luminosity for both models is calculated here as

(1)

where ϵr is the assumed radiative efficiency of the accretion disc, ṀBH is the BH accretion rate, and c is the speed of light. Given that the detection limit for the fainter of the two QSOs in our observational search is MB ∼ −26 mag (i.e., LB ∼ 8.25 × 1045 erg s−1), we assume a minimum bolometric luminosity of Lbol,min = 4 × 1046 erg s−1 when searching for BH pairs in the models, given the bolometric conversion factor of 5.13 recommended by Duras et al. (2020).

For the same search area as our observational sample (i.e., an inner radius of 125 h−1 kpc and outer radius of 3.3 h−1 Mpc comoving, see Sect. 3.1), we find no BH pairs with Lbol ≥ 4 × 1046 erg s−1 in either TNG300 or L-GALAXIES 2020 at z = 6. This holds true even when doubling the assumed radiative efficiency from ϵr = 0.1 to 0.2. There are no BHs at all at that luminosity within the TNG300 box above z = 5 when assuming ϵr = 0.2. This theoretical result is nicely consistent with the null result returned by our observational search at z = 6, especially when considering the relatively large box sizes of L-GALAXIES 2020 and TNG300.

We do not find any BH pairs at z = 5 in TNG300 or L-GALAXIES 2020 that match the serendipitous discovery by McGreer et al. (2016) of a bright pair with M1450 ≤  − 25.4 (i.e., Lbol ≥ 2.4 × 1046 erg s−1) separated by 21″ (i.e., 816 ckpc) at z = 5.02. These simulations are therefore somewhat in tension with this discovery, although the exact search area to use is not well constrained. However, we do find pairs in TNG300 that match the serendipitous discovery on a BH pair with MB ≤ −23.7 mag (i.e., Lbol ≥ 5.1 × 1045 erg s−1) by Schneider et al. (2000) at z = 4.26. When assuming ϵr = 0.1, and a standard search volume around the mean deprojected separation of r = 1.92 cMpc with inner radius r − (r/4) = 1.44 cMpc and outer radius r + (r/4) = 2.40 cMpc, we find 14 BH pairs with Lbol ≥ 5.1 × 1045 erg s−1 at z = 4 in TNG300. This equates to an upper limit on the pair fraction (i.e., the number of BH pairs divided by the total number of BHs at that luminosity) of ∼3.1%. We are also able to reproduce a fairly redshift-invariant pair fraction of ∼1% between z ∼ 4 and 1 (see Fig. 4) for BH pairs with a primary BH luminosity of Lbol,pri ≥ 1 × 1045.3 erg s−1, when selecting systems as Silverman et al. (2020) do for the HORIZON-AGN simulation (see their Sect. 5.1). No such pairs were found in TNG300 above z = 4.

thumbnail Fig. 4.

Fraction of BH pairs over redshift in TNG300, for two values of ϵr. Following Silverman et al. (2020), pairs are selected to have primary BH luminosities of Lbol,pri ≥ 1045.3 erg s−1, secondary BH luminosities of Lbol,sec ≥ Lbol,pri/10, a minimum BH mass of 108 M, and a minimum host-galaxy stellar mass of 1010 M. At redshifts > 5 there are AGN, but no pairs.

In conclusion, the observed number of bright QSO pairs seen at high redshift appears largely consistent with that predicted by current cosmological-sized galaxy evolution models, although further modelling efforts at the redshift of our observational sample are still required to confirm this.

3.4. Number of QSO pairs as a constraint on massive BH origins and assembly

The origin of MBH ∼ 108−9 M BHs in massive galaxies at z ∼ 6−7 is largely unconstrained (e.g., Johnson et al. 2013; Valiante et al. 2017, 2018; Inayoshi et al. 2020). The cosmic time since the Big Bang is likely too short for light seeds with masses of ∼100 M (e.g., the PopIII remnant model, Madau & Rees 2001; Bromm & Loeb 2003; Volonteri et al. 2003b) to grow to such masses even if the seeds accrete at the Eddington rate most of their lifetimes. Instead, more massive seeds with masses of order ∼104 M would have a somewhat less challenging growth history to reach the masses of the high-z QSOs.

Several theoretical models have been put forward to explain the formation of such massive BH seeds. By analogy with the nuclear clusters that we commonly observe in low-redshift galaxies, compact nuclear clusters in metal-poor environments are predicted to form very massive stars by runaway stellar collisions (Omukai et al. 2008; Devecchi & Volonteri 2009; Regan & Haehnelt 2009). These very massive stars would collapse onto BH seeds of MBH ∼ 102 − 103M. Seeds of about the same mass can be formed by runaway BH mergers in metal-poor star clusters in the centre of high-redshift galaxies. The direct collapse model predicts the most massive seeds, called direct-collapse BHs (DCBHs): atomic cooling halos form single supermassive stars, which can collapse onto massive seeds of MBH ∼ 104 − 106M (Loeb & Rasio 1994; Bromm & Loeb 2003; Spaans & Silk 2006; Begelman et al. 2006; Lodato & Narayan 2006; Dijkstra et al. 2008; Visbal et al. 2014; Latif & Volonteri 2015; Habouzit et al. 2016b). In order to avoid the fragmentation of the gas into multiple and less massive stars, the presence or formation of efficient gas coolants (i.e. molecular hydrogen and metals) must be prevented. Therefore, the formation of the first DCBHs depends on the number of surrounding halos because the Lyman-Werner radiation (photons with energy in the 11.2–13.6 eV range) needed to prevent H2 formation comes from the surrounding star-forming galaxies (Habouzit et al. 2016a).

These mechanisms yield different galaxy occupation fractions (i.e. the probability for a galaxy to host a BH), and therefore could lead a priori to different predictions on the number of QSO pairs at high redshift. While the compact nuclear stellar cluster model and the model of runaway BH mergers in clusters predict potentially large occupation fractions (i.e. the presence of BHs in many galaxies), the number density of direct collapse BHs is predicted to be small (e.g., Dijkstra et al. 2008; Habouzit et al. 2016b). However, DCBHs could appear clustered. Whether the large Lyman-Werner radiation is produced by several or a single star-forming halo, the high intensity required could irradiate several nearby halos, potentially forming DCBHs in some of these halos if metal-poor conditions are met. The need for clustered halo environments is not required for all the variations of the direct collapse model (e.g., the synchronised pair model, Visbal et al. 2014). It has also been proposed that DCBHs themselves could ignite a runaway process of further DCBH formation (Yue et al. 2014). DCBHs are expected to be Compton thick, and their reprocessed radiation outshines small high-redshift galaxies by more than a factor of 10. Thus, once the first DCBHs form, their Lyman-Werner radiation could help to provide surrounding halos with the needed conditions to form new DCBHs (Yue et al. 2014). This process comes to an end when the host atomic-cooling halos are rapidly photo-evaporated by ionizing photons. An analytical model of these processes (Yue et al. 2014) shows that a highly clustered spatial distribution of halos is the pre-requisite to triggering the formation of several DCBHs in the same region.

The different mechanisms of BH formation are not a priori mutually exclusive in the Universe, which therefore makes the interpretation of the QSO clustering difficult. Moreover, successive mergers of halos with time alter the initial halo occupation fraction and clustering of BHs at birth.

Observing several QSO pairs in our sample would likely suggest that massive BH seeds can form in the same regions (on scales of up to 1 Mpc), and that they can grow efficiently in the dense environments in which we often find high-redshift QSOs. The fact that we find no QSO pairs (i.e. no bright companion to any of our 116 spectroscopically confirmed QSOs) could mean that (i) most QSOs are isolated objects with no massive galaxy in their surroundings; (ii) QSOs are often located in dense environments, but nearby galaxies are devoid of BHs or devoid of massive BHs; or (iii) QSOs are often located in dense environments, and nearby galaxies also host massive BHs; however, these BHs have high variability and/or low duty cycle and no companion BHs were active when observed.

While the question of whether QSOs are embedded in the densest environments is still subject to debate in observations (Kim et al. 2009; Husband et al. 2013; Mazzucchelli et al. 2017a; McGreer et al. 2016; Habouzit et al. 2019, and references therein) hypothesis (i) where a large fraction of QSOs are isolated objects is unlikely. Theoretical models suggest that the first massive galaxies formed through mergers of gas-rich galaxies at very high redshift. Recently, Decarli et al. (2017) and Neeleman et al. (2019) identified six z > 6 QSOs with close gas-rich companions through ALMA observations of [CII] and dust emission; five of these sources are in our sample: SDSSJ0842+1218, PSOJ167.6415−13.4960, SDSSJ1306+0356, PSOJ308.0416−21.2339, and CFHQSJ2100−1715. These QSO–galaxy pairs prove that z > 6 QSOs are not living isolated from the surrounding matter in their halos. This QSO–galaxy clustering is consistent with the quasar/LBG clustering at z ∼ 4 (see Fig. 3 in Decarli et al. 2017). Recent deep X-ray observations of some of the above sources did not reveal potentially obscured quasars. To date, there is no strong evidence that these obscured companions of z ∼ 6 quasars are obscured AGN (Connor et al. 2020; Vito et al. 2021).

Therefore, our non-detection of bright QSO–QSO pairs could hint towards an evolutionary scenario that prevents the formation of a massive accreting BH in the potentially gas-rich QSO companion. This could be due to tidal interaction of the primary QSO being more rapid than the accretion on the central BH of the companion. Feedback from the QSO could also reduce the gas reservoir of the companion galaxy through its lifetime, therefore diminishing the ability of a BH to efficiently accrete and become as massive as the BH powering the QSO. Additionally, efficient stripping of gas from companion galaxies of massive systems is seen out to several virial radii at low redshift in cosmological simulations (Ayromlou et al. 2021a,b). If this effect is present in the neighbourhood of the brightest QSOs at high redshift, it could also contribute to a reduction in the gas reservoir of companion BHs. The absence of QSO pairs could also be explained in case of low efficiency of BH formation, so that early massive BH formation happens only in a small fraction of galaxies, or alternatively if the formation of light seeds is dominant in the Universe. Due to their low masses, light seeds would have a hard time growing and even sinking efficiently to the potential well of their host galaxies. Searching for companions of the ten most massive BHs in TNG300 reveals none with masses MBH ≥ 107.7 M at z = 5 and above. There is just one companion BH with a mass of MBH ∼ 107.3 M at z = 6, but with a luminosity of only Lbol = 1043.3 erg s−1, which is not even the brightest luminosity in the companion sample (which is 1045.1 erg s−1), and below our search sensitivity. With its limited volume of (205 h−1 cMpc)3 there is no environment in TNG300 that was able to build up two extremely massive and luminous BHs in the same region by z ∼ 6.

Finally, this leads us to hypothesis (iii) where very short coincident activity periods of both QSOs (see e.g., Haiman & Hui 2001 for an early account of the lifetime-dependence of QSO clustering). In this case QSO companion galaxies would also host massive BHs, but these BHs would not be efficiently accreting at the time of observation. In this case, the companion massive BHs would have grown by several orders of magnitude, potentially releasing a significant amount of energy in their host galaxies through cosmic time, which could explain an accretion history with a succession of ‘on and off’ phases. In the observations presented here we can only detect very bright objects. Further investigations at larger telescopes (e.g., with the GROND-descendant SCORPIO instrument at Gemini Observatory) will shed new light onto the degenerate scenarios to explain the absence of QSO pairs with M1450(AB) <  − 26 mag.

3.5. Incidence of brown dwarfs

The exposure for the spectroscopic observations was determined to provide > 5σ detections of flux redwards of Ly-α (if present) binned at 200 Å. Thus, the signal-to-noise ratio of these spectra is good enough to distinguish brown dwarfs and low-redshift galaxies from QSOs, but much too low to perform a spectral classification of the dwarfs or galaxies. We therefore revert to the results of the Le PHARE fits, and provide the spectral types of the best-fitting template in the last columns of Tables. A.4 and A.5. For the conversion from the Le PHARE best-fit temperature to spectral type we used the mean value from the compilation of Kellogg (2017). The associated error is ±2 sub-types over most of the range.

Since the unWISE (Schlafly et al. 2019) and CatWISE (Marocco et al. 2021) catalogues became available after the selection of our targets for spectroscopic follow-up, we ran a new cross-correlation of our targets against these catalogues. From our list of 17 candidates without ALLWISE counterparts (Table A.5) we found 6 objects with W1 and W2 entries. Re-running Le PHARE fits with this extended photometry yields consistent results for three dwarf identifications (within the above-mentioned ±2 sub-type error), but deviates for the other three objects: two SEDs, previously best fit with a QSO or M8 dwarf template are now best fit with a low-redshift galaxy (z ∼ 1.2 in both cases), similar to ULASJ0148+0600 (Table A.4), and another source, SDSSJ2054−0005_3, for which the best-fit template is a galaxy at redshift 6, which is inconsistent with our absolute photometry. The second-best template (nearly identical χ2) is a QSO at redshift 6, which is unlikely given our GMOS spectrum, and the third-best template is an L2 dwarf, but with substantially worse χ2. We therefore do not assign an identification in Table A.6. For the candidate PSOJ055.4244−00.8035, which also had a QSO template as best fit in Le PHARE, but for which we did not obtain a spectrum, we refrain from changing the ID, though Occam’s razor would argue against a QSO.

The combination of our spectroscopic results and the Le PHARE fits results in the following identifications: 20 dwarfs, 3 low-redshift galaxies, 1 QSO, and 2 unsettled cases (Tables A.4A.6). Their spectral range is rather wide, from M5 to T8, though accurate spectral typing remains to be done. The large incidence rate of such dwarfs is a surprise because our search was not tuned towards late dwarfs, and also because our small search size was just 0.52 square degrees (summing over all 116 candidates). For instance, Kakazu et al. (2010) found six faint (19 ≲ J ≲ 20) ultracool T dwarfs over a 9.3 deg2 area with a limiting magnitude of z′(AB) ≲ 23.3 mag. This is slightly deeper than our mean depth of z′(AB) ≲ 23.1 mag, yet we find two T dwarfs per 0.52 deg2, compared to their 0.6 deg−2, a factor of 6 higher. Given that we pointed at known high-z QSOs, our search area similarly avoids the galactic plane. In a more recent compilation, Ryan & Reid (2016) used thick and thin exponential disc models and the luminosity function of ultracool dwarfs to predict their surface density per spectral type. While this depends strongly on the galactic coordinates, the prediction for T dwarfs has only a small dependence on galactic latitude, suggesting an even higher excess in our data, about a factor of 20, with even the Kakazu et al. (2010) surface density being higher by a factor of ∼3. For early L-type dwarfs, the excess is smaller, and for the M8-9 dwarfs we find consistency. Figure 5 summarises the findings.

thumbnail Fig. 5.

Number of brown dwarfs among our sample of 26 candidates per total area of our search region over 116 QSOs, compared to the cumulative surface density as modelled by Ryan & Reid (2016), plotted here for their COSMOS sample field (central line) as well as the fields with the smallest and largest numbers (corresponding to different galactic latitudes) providing the covered range (shaded area). Our data points are plotted at the faintest magnitude per spectral type range, and the horizontal bar covers the magnitude range up to the brightest. While the observed number of M dwarfs is compatible with the known surface density, that of T dwarfs is substantially larger.

thumbnail Fig. 6.

Le PHARE fits for the ten objects from Tables A.4 and A.5, for which a QSO template provides the better fit, plus the one with a best-fit galaxy template (from top left to bottom right): ULASJ0148+0600, CFHQSJ1509−1749, PSOJ009.7355−10.4316, PSOJ011.3899+09.0325_C2, PSOJ045.1840−22.5408, PSOJ071.0322−04.5591, PSOJ261.0364+19.0286_C1, PSOJ267.0021+22.7812_C2, SDSSJ2054−0005_C3, SDSSJ2054−0005_C4, IMSJ2204+0012. These are the SEDs after correction of the galactic foreground AV, and most of the dwarf fits return an even worse reduced χ2 for the no-AV SEDs than shown here.

thumbnail Fig. 7.

Long-slit spectra binned at 160 and 400 Å (thin and thick lines, respectively) obtained with the GMOS spectrographs at the Gemini North (top panel) and South (lower panel) Observatories. The vertical solid lines denote the position of Ly-α of its ‘mother’ QSO, the vertical dashed line the position of Ly-α based on the Le PHARE fits. The dwarf templates are at arbitrary normalizations.

Since we observe a substantial excess of T dwarfs, search completeness cannot be an issue. Foreground extinction correction should also not be an issue since most of these dwarfs are within approximately 100 pc distance. Even if we adopt (systematic?) error in the temperature estimates from the Le PHARE fits that is twice as large, this would not change our numbers. One potential reason could be the appropriateness of the templates in Le PHARE, meaning that dwarfs are more numerous than low-redshift red galaxies, for example. This is contrary to the case of high-z QSOs, for which late dwarfs form a numerous class of false positives. This finding of a higher than expected T-dwarf number density is a surprising puzzle that warrants further study. Existing GROND data could provide a first step in this direction, though spectroscopic typing of the candidate dwarfs is the ultimate step.

4. Conclusions

No candidate QSO pair has been found in a search of 116 spectroscopically confirmed redshift 6 quasars within our 0.1–3.3 h−1 cMpc search radius with a companion brighter than M1450(AB) < −26 mag. Although this result is consistent with cosmological-scale galaxy evolution simulations (given the redshift, search volume, and luminosity limits), the serendipitously discovered pairs at z = 4.26 (Schneider et al. 2000) and z = 5.02 (McGreer et al. 2016) suggested one pair every 40 QSOs, or 2.5%, implying an empirical expectation of 3 pairs in our sample of 116 search targets. At first glance this result suggests little clustering at redshift 6. However, with updated knowledge of the faint end of the luminosity function, the wp statistics returns more than an order of magnitude higher value than inferred from the low-redshift clustered QSOs; thus, given the still low statistics in our search and the uncertainties (in particular in the unknown search volumes) of the serendipitous detections, we are not able to claim that the clustering at redshift 6 is indeed smaller than that up to redshift 5.

As a consequence of our search, we find a higher than expected rate of ultracool dwarfs. While the excess of L dwarfs is marginally consistent with statistics and could be due to the accidental lack of a single source about 1 mag fainter, this argument does not hold for the T dwarfs. The observed excess of a factor of ∼20 suggests (i) strong surface density variations over the sky, (ii) a lack of appropriate templates for other source types (e.g., low-redshift red galaxies) such that the T templates just happen to be the closest in shape to the observed SEDs, or (iii) an underestimate of the true rate of T dwarfs. A larger survey at the same depth would be needed to settle this question.


Acknowledgments

We thank the referee for the detailed comments. Part of the funding for GROND (both hardware as well as personnel) was generously granted from the Leibniz-Prize to Prof. G. Hasinger (DFG grant HA 1850/28-1). P. S. acknowledges support through the Sofja Kovalevskaja Award from the Alexander von Humboldt Foundation of Germany while at MPE. J. G. greatly acknowledges the GROND observers Thomas Krühler, Tassilo Schweyer, Simon Steinmassl, Helmut Steinle and Phil Wiseman. We explicitly acknowledge the open attitude of providing the detailed model predictions for dwarfs by Ryan & Reid (2016) for download, which only enabled the creation of Fig. 5. The reported spectroscopy is based on observations obtained at the international Gemini Observatory, a program of NSF’s NOIRLab, which is managed by the Association of Universities for Research in Astronomy (AURA) under a cooperative agreement with the National Science Foundation on behalf of the Gemini Observatory partnership: the National Science Foundation (United States), National Research Council (Canada), Agencia Nacional de Investigación y Desarrollo (Chile), Ministerio de Ciencia, Tecnología e Innovación (Argentina), Ministério da Ciência, Tecnologia, Inovações e Comunicações (Brazil), and Korea Astronomy and Space Science Institute (Republic of Korea). This paper includes data gathered with the 6.5 m Magellan Telescopes located at Las Campanas Observatory, Chile. This research has made use of the NASA/IPAC Infrared Science Archive, which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. For photometric calibration, the Sloan Digital Sky Survey, PanSTARRS, SkyMapper and 2MASS are acknowledged: Funding for SDSS-III has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, and the US Department of Energy Office of Science. The SDSS-III web site is http://www.sdss3.org/. SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration including the University of Arizona, the Brazilian Participation Group, Brookhaven National Laboratory, Carnegie Mellon University, University of Florida, the French Participation Group, the German Participation Group, Harvard University, the Instituto de Astrofisica de Canarias, the Michigan State/Notre Dame/JINA Participation Group, Johns Hopkins University, Lawrence Berkeley National Laboratory, Max Planck Institute for Astrophysics, Max Planck Institute for Extraterrestrial Physics, New Mexico State University, New York University, Ohio State University, Pennsylvania State University, University of Portsmouth, Princeton University, the Spanish Participation Group, University of Tokyo, University of Utah, Vanderbilt University, University of Virginia, University of Washington, and Yale University. 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). The national facility capability for SkyMapper has been funded through ARC LIEF grant LE130100104 from the Australian Research Council, awarded to the University of Sydney, the Australian National University, Swinburne University of Technology, the University of Queensland, the University of Western Australia, the University of Melbourne, Curtin University of Technology, Monash University and the Australian Astronomical Observatory. SkyMapper is owned and operated by The Australian National University’s Research School of Astronomy and Astrophysics. The survey data were processed and provided by the SkyMapper Team at ANU. The SkyMapper node of the All-Sky Virtual Observatory (ASVO) is hosted at the National Computational Infrastructure (NCI). Development and support the SkyMapper node of the ASVO has been funded in part by Astronomy Australia Limited (AAL) and the Australian Government through the Commonwealth’s Education Investment Fund (EIF) and National Collaborative Research Infrastructure Strategy (NCRIS), particularly the National eResearch Collaboration Tools and Resources (NeCTAR) and the Australian National Data Service Projects (ANDS). This publication makes use of data products from the Two Micron All Sky Survey, which is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.

References

  1. Arnouts, S., Cristiani, S., Moscardini, L., et al. 1999, MNRAS, 310, 540 [NASA ADS] [CrossRef] [Google Scholar]
  2. Ayromlou, M., Nelson, D., Yates, R. M., et al. 2021a, MNRAS, 502, 1051 [NASA ADS] [CrossRef] [Google Scholar]
  3. Ayromlou, M., Kauffmann, G., Yates, R. M., et al. 2021b, MNRAS, 505, 492 [NASA ADS] [CrossRef] [Google Scholar]
  4. Bañados, E., Venemans, B. P., Decarli, R., et al. 2016, ApJS, 227, 11 [NASA ADS] [CrossRef] [Google Scholar]
  5. Bañados, E., Venemans, B. P., Mazzucchelli, C., et al. 2018, Nature, 553, 473 [NASA ADS] [CrossRef] [Google Scholar]
  6. Begelman, M. C., Volonteri, M., & Rees, M. J. 2006, MNRAS, 370, 289 [NASA ADS] [CrossRef] [Google Scholar]
  7. Bhowmick, A. K., DiMatteo, T., Eftekharzadeh, S., & Myers, A. D. 2019, MNRAS, 485, 2026 [NASA ADS] [CrossRef] [Google Scholar]
  8. Bonoli, S., Marulli, F., Springel, V., et al. 2009, MNRAS, 396, 423 [NASA ADS] [CrossRef] [Google Scholar]
  9. Bromm, V., & Loeb, A. 2003, ApJ, 596, 34 [NASA ADS] [CrossRef] [Google Scholar]
  10. Carnall, A. C., Shanks, T., Chehade, B., et al. 2015, MNRAS, 451, L16 [Google Scholar]
  11. Chambers, K. C., Magnier, E. A., Metcalfe, N., et al. 2016, ArXiv e-prints [arXiv:1612.05560] [Google Scholar]
  12. Chehade, B., Shanks, T., Findlay, J., et al. 2016, MNRAS, 459, 1179 [NASA ADS] [CrossRef] [Google Scholar]
  13. Connor, T., Bañados, E., Mazzucchelli, C., et al. 2020, ApJ, 900, 189 [Google Scholar]
  14. Costa, T., Sijacki, D., Trenti, M., & Haehnelt, M. G. 2014, MNRAS, 439, 2146 [NASA ADS] [CrossRef] [Google Scholar]
  15. Croton, D. J., Springel, V., White, S. D. M., et al. 2006, MNRAS, 365, 11 [NASA ADS] [CrossRef] [Google Scholar]
  16. Davis, M., & Peebles, P. J. E. 1983, ApJ, 267, 465 [NASA ADS] [CrossRef] [Google Scholar]
  17. Decarli, R., Walter, F., Venemans, B. P., et al. 2017, Nat, 545, 457 [NASA ADS] [CrossRef] [Google Scholar]
  18. DeGraf, C., & Sijacki, D. 2017, MNRAS, 466, 3331 [NASA ADS] [CrossRef] [Google Scholar]
  19. Devecchi, B., & Volonteri, M. 2009, ApJ, 694, 302 [NASA ADS] [CrossRef] [Google Scholar]
  20. Dijkstra, M., Haiman, Z., Mesinger, A., & Wyithe, J. S. B. 2008, MNRAS, 391, 1961 [NASA ADS] [CrossRef] [Google Scholar]
  21. Di Matteo, T., Croft, R. A. C., Feng, Y., et al. 2017, MNRAS, 467, 4243 [NASA ADS] [CrossRef] [Google Scholar]
  22. Djorgovski, S. G., Stern, D., Mahabal, A. A., & Brunner, R. 2003, ApJ, 596, 67 [NASA ADS] [CrossRef] [Google Scholar]
  23. Duras, F., Bongiorno, A., Ricci, F., et al. 2020, A&A, 636, 73 [Google Scholar]
  24. Eftekharzadeh, S., Myers, A. D., Hennawi, J. F., et al. 2017, MNRAS, 468, 77 [NASA ADS] [CrossRef] [Google Scholar]
  25. Eisenstein, D. J., Weinberg, D. H., Agol, E., et al. 2011, AJ, 142, 72 [Google Scholar]
  26. Fan, X., Strauss, M. A., Schneider, D. P., et al. 1999, AJ, 118, 1 [NASA ADS] [CrossRef] [Google Scholar]
  27. Fang, L. Z. 1989, Int. J. Mod. Phys. A, 4, 3477 [NASA ADS] [CrossRef] [Google Scholar]
  28. Greiner, J., Bornemann, W., Clemens, C., et al. 2008, PASP, 120, 405 [NASA ADS] [CrossRef] [Google Scholar]
  29. Habouzit, M., Volonteri, M., Latif, M., Dubois, Y., & Peirani, S. 2016a, MNRAS, 463, 529 [NASA ADS] [CrossRef] [Google Scholar]
  30. Habouzit, M., Volonteri, M., Latif, M., et al. 2016b, MNRAS, 456, 1901 [CrossRef] [Google Scholar]
  31. Habouzit, M., Volonteri, M., Somerville, R. S., et al. 2019, MNRAS, 489, 1206 [CrossRef] [Google Scholar]
  32. Habouzit, M., Li, Y., Somerville, R. S., et al. 2021, MNRAS, 503, 1940 [NASA ADS] [CrossRef] [Google Scholar]
  33. Haiman, Z., & Hui, L. 2001, ApJ, 547, 27 [NASA ADS] [CrossRef] [Google Scholar]
  34. Haehnelt, M., & Nusser, A., 1999, in Evolution of large scale structure: from recombination to Garching, Proc. MPA-ESO cosmology conf., August 1998, eds. A. J. Banday et al., 375 [Google Scholar]
  35. Hennawi, J. F., Strauss, M. A., Oguri, M., et al. 2006, AJ, 131, 1 [NASA ADS] [CrossRef] [Google Scholar]
  36. Hennawi, J. F., Myers, A. D., Shen, Y., et al. 2010, ApJ, 719, 1672 [NASA ADS] [CrossRef] [Google Scholar]
  37. Henriques, B. M. B., Yates, R. M., Fu, J., et al. 2020, MNRAS, 491, 5795 [NASA ADS] [CrossRef] [Google Scholar]
  38. Hopkins, P. F., Lidz, A., Hernquist, L., et al. 2007, ApJ, 662, 110 [NASA ADS] [CrossRef] [Google Scholar]
  39. Husband, K., Bremer, M. N., Stanway, E. R., et al. 2013, MNRAS, 432, 2869 [NASA ADS] [CrossRef] [Google Scholar]
  40. Ilbert, O., Arnouts, S., McCracken, H. J., et al. 2006, A&A, 457, 841 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  41. Inayoshi, K., Visbal, E., & Haiman, Z. 2020, ARAA, 58, 27 [NASA ADS] [CrossRef] [Google Scholar]
  42. Johnson, J. L., Whalen, D. J., Li, H., & Holz, D. E. 2013, ApJ, 771, 116 [NASA ADS] [CrossRef] [Google Scholar]
  43. Jiang, L., Fan, X., Bian, F., et al. 2009, AJ, 138, 305 [NASA ADS] [CrossRef] [Google Scholar]
  44. Jiang, L., McGreer, I. D., Fan, X., et al. 2016, ApJ, 833, 222 [NASA ADS] [CrossRef] [Google Scholar]
  45. Kakazu, Y., Hu, E. M., Liu, M. C., et al. 2010, ApJ, 723, 184 [NASA ADS] [CrossRef] [Google Scholar]
  46. Kashikawa, N., Kitayama, T., Doi, M., et al. 2007, ApJ, 663, 765 [NASA ADS] [CrossRef] [Google Scholar]
  47. Kellogg, K. 2017, Electronic Thesis and Dissertation Repository, 4857, https://ir.lib.uwo.ca/etd/4857 [Google Scholar]
  48. Kim, S., Stiavelli, M., Trenti, M., et al. 2009, ApJ, 695, 809 [NASA ADS] [CrossRef] [Google Scholar]
  49. Kim, Y., Im, M., Jeon, Y., et al. 2015, ApJ, 813, L35 [NASA ADS] [CrossRef] [Google Scholar]
  50. Kroupa, P., Subr, L., Jerabkova, T., & Wang, L. 2020, MNRAS, 498, 5652 [Google Scholar]
  51. Krühler, T., Küpcü, Yoldaş A., Greiner, J., et al. 2008, ApJ, 685, 376 [CrossRef] [Google Scholar]
  52. Kulkarni, G., Worseck, G., & Hennawi, J. F. 2019, MNRAS, 488, 1035 [NASA ADS] [CrossRef] [Google Scholar]
  53. Küpcü, Yoldaş A., Krühler, T., Greiner, J., et al. 2008, AIP Conf. Proc., 1000, 227 [CrossRef] [Google Scholar]
  54. Latif, M. A., & Volonteri, M. 2015, MNRAS, 452, 1026 [CrossRef] [Google Scholar]
  55. Leśniewska, A., & Michalowski, M. J. 2019, A&A, 624, L13 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  56. Lodato, G., & Narayan, P. 2006, MNRAS, 371, 1813 [NASA ADS] [CrossRef] [Google Scholar]
  57. Loeb, A., & Rasio, F. A. 1994, ApJ, 432, 52 [NASA ADS] [CrossRef] [Google Scholar]
  58. Lupi, A., Volonteri, M., Decarli, R., et al. 2019, MNRAS, 488, 4004 [NASA ADS] [CrossRef] [Google Scholar]
  59. Madau, P., & Rees, M. J. 2001, ApJ, 551, L27 [NASA ADS] [CrossRef] [Google Scholar]
  60. Madau, P., Haardt, F., & Rees, M. J. 1999, ApJ, 514, 648 [CrossRef] [Google Scholar]
  61. Mainzer, A., Bauer, J., Grav, T., et al. 2011, ApJ, 731, 53 [NASA ADS] [CrossRef] [Google Scholar]
  62. Marshall, M. A., Ni, Y., Di Matteo, T., et al. 2020, MNRAS, 499, 3891 [NASA ADS] [Google Scholar]
  63. Marinacci, F., Vogelsberger, M., Pakmor, R., et al. 2018, MNRAS, 480, 5113 [NASA ADS] [Google Scholar]
  64. Marocco, F., Eisenhardt, P. R. M., Fowler, J. W., et al. 2021, ApJS, 253, 8 [Google Scholar]
  65. Matsuoka, Y., Onoue, M., Kashikawa, N., et al. 2016, ApJ, 828, 26 [NASA ADS] [CrossRef] [Google Scholar]
  66. Matsuoka, Y., Strauss, M. A., Kashikawa, N., et al. 2018, ApJ, 869, 150 [NASA ADS] [CrossRef] [Google Scholar]
  67. Mazzucchelli, C., Bañados, E., Venemans, B. P., et al. 2017a, ApJ, 849, 91 [NASA ADS] [CrossRef] [Google Scholar]
  68. Mazzucchelli, C., Bañados, E., Decarli, R., et al. 2017b, ApJ, 834, 83 [NASA ADS] [CrossRef] [Google Scholar]
  69. McGreer, I. D., Eftekharzadeh, S., Myers, A. D., & Fan, X. 2016, AJ, 151, 61 [NASA ADS] [CrossRef] [Google Scholar]
  70. Naiman, J. P., Pillepich, A., Springel, V., et al. 2018, MNRAS, 477, 1206 [CrossRef] [Google Scholar]
  71. Neeleman, M., Bañados, E., Walter, F., et al. 2019, ApJ, 882, 10 [NASA ADS] [CrossRef] [Google Scholar]
  72. Nelson, D., Pillepich, A., Springel, V., et al. 2018, MNRAS, 475, 624 [NASA ADS] [CrossRef] [Google Scholar]
  73. Omukai, K., Schneider, R., & Haiman, Z. 2008, ApJ, 686, 801 [NASA ADS] [CrossRef] [Google Scholar]
  74. Pillepich, A., Springel, V., Nelson, D., et al. 2018, MNRAS, 473, 4077 [Google Scholar]
  75. Planck Collaboration XIII. 2016, A&A, 594, A13 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  76. Porciani, C., Magliocchetti, M., & Norberg, P. 2007, MNRAS, 355, 1010 [Google Scholar]
  77. Reed, S. L., McMahon, R. G., Banerji, M., et al. 2015, MNRAS, 454, 3952 [NASA ADS] [CrossRef] [Google Scholar]
  78. Regan, J. A., & Haehnelt, M. G. 2009, MNRAS, 393, 858 [NASA ADS] [CrossRef] [Google Scholar]
  79. Ross, N. P., McGreer, I. D., White, M., et al. 2013, ApJ, 773, 14 [NASA ADS] [CrossRef] [Google Scholar]
  80. Ryan, R. E., & Reid, I. N. 2016, AJ, 151, 92 [NASA ADS] [CrossRef] [Google Scholar]
  81. Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103 [NASA ADS] [CrossRef] [Google Scholar]
  82. Schlafly, E. F., Meisner, A. M., & Green, G. M. 2019, ApJS, 240, 30 [Google Scholar]
  83. Schneider, D. P., Fan, X., Strauss, M. A., et al. 2000, AJ, 120, 2183 [NASA ADS] [CrossRef] [Google Scholar]
  84. Scolnic, D., Casertano, S., Riess, A., et al. 2015, ApJ, 815, 117 [NASA ADS] [CrossRef] [Google Scholar]
  85. Shen, Y., Strauss, M. A., Oguri, M., et al. 2007, AJ, 133, 2222 [NASA ADS] [CrossRef] [Google Scholar]
  86. Shen, Y., Hennawi, J. F., Shankar, F., et al. 2010, ApJ, 719, 1693 [NASA ADS] [CrossRef] [Google Scholar]
  87. Shen, X., Hopkins, P. F., Faucher-Giguee, C. A., et al. 2020, MNRAS, 495, 3252 [NASA ADS] [CrossRef] [Google Scholar]
  88. Sheth, R. K., Mo, H. J., & Tormen, G. 2001, MNRAS, 323, 1 [NASA ADS] [CrossRef] [Google Scholar]
  89. Silverman, J. D., Tang, S., Lee, K.-G., et al. 2020, ApJ, 899, 154 [Google Scholar]
  90. Skrutskie, M. F., Cutri, R. M., Stiening, R., et al. 2006, AJ, 131, 1163 [Google Scholar]
  91. Songaila, A. 2004, AJ, 127, 2598 [Google Scholar]
  92. Spaans, M., & Silk, J. 2006, ApJ, 652, 902 [NASA ADS] [CrossRef] [Google Scholar]
  93. Springel, V., White, S. D. M., Jenkins, A., et al. 2005, Nature, 435, 629 [NASA ADS] [CrossRef] [PubMed] [Google Scholar]
  94. Springel, V., Pakmor, R., Pillepich, A., et al. 2018, MNRAS, 475, 676 [NASA ADS] [CrossRef] [Google Scholar]
  95. Tenneti, A., Wilkins, S. M., Di Matteo, T., et al. 2019, MNRAS, 483, 1388 [NASA ADS] [CrossRef] [Google Scholar]
  96. Tody, D. 2019, ASP Conf. Ser., 52, 173 [NASA ADS] [Google Scholar]
  97. Utsumi, Y., Goto, T., Kashikawa, N., et al. 2010, ApJ, 721, 1680 [NASA ADS] [CrossRef] [Google Scholar]
  98. Valiante, R., Agrarval, B., Habouzit, M., & Pezzulli, E. 2017, PASA, 34 [CrossRef] [Google Scholar]
  99. Valiante, R., Schneider, R., Graziani, L., & Zappacosta, L. 2018, MNRAS, 474, 3825 [NASA ADS] [CrossRef] [Google Scholar]
  100. Venemans, B. P., Findlay, J. R., Sutherland, W. J., et al. 2013, ApJ, 779, 24 [NASA ADS] [CrossRef] [Google Scholar]
  101. Venemans, B. P., Verdoes Kleijn, G. A., Mwebaze, J., et al. 2015, MNRAS, 453, 2259 [NASA ADS] [CrossRef] [Google Scholar]
  102. Visbal, E., Haiman, Z., & Bryan, G. L. 2014, MNRAS, 445, 1056 [NASA ADS] [CrossRef] [Google Scholar]
  103. Vito, F., Brandt, W. N., Ricci, F., et al. 2021, A&A, 649, A133 [NASA ADS] [CrossRef] [EDP Sciences] [Google Scholar]
  104. Volonteri, M. 2010, A&ARv, 18, 279 [Google Scholar]
  105. Volonteri, M., Haardt, F., & Madau, P. 2003a, ApJ, 582, 559 [Google Scholar]
  106. Volonteri, M., Madau, P., & Haardt, F. 2003b, ApJ, 594, 661 [NASA ADS] [CrossRef] [Google Scholar]
  107. Volonteri, M., Dubois, Y., Pichon, C., & Devriendt, J. 2016, MNRAS, 460, 2979 [Google Scholar]
  108. Wang, F., Yang, J., Fan, X., et al. 2021, ApJ, 907, L1 [Google Scholar]
  109. Willott, C. J., Delorme, P., Omont, A., et al. 2007a, AJ, 134, 2435 [NASA ADS] [CrossRef] [Google Scholar]
  110. Willott, C. J., Delorme, P., Reyle, C., et al. 2007b, AJ, 137, 3541 [Google Scholar]
  111. Willott, C. J., Albert, L., Arzoumanian, D., et al. 2010, AJ, 140, 546 [NASA ADS] [CrossRef] [Google Scholar]
  112. Wolf, C., Onken, C. A., Luvaul, L. C., et al. 2018, PASA, 35, 10 [Google Scholar]
  113. Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868 [Google Scholar]
  114. Wyithe, J. S. B., Loeb, A., & Carilli, C. 2005, ApJ, 628, 575 [NASA ADS] [CrossRef] [Google Scholar]
  115. Yang, J., Wang, F., Fan, X., et al. 2020, ApJ, 897, L14 [Google Scholar]
  116. Yue, B., Ferrara, A., Salvaterra, R., et al. 2014, MNRAS, 440, 1263 [NASA ADS] [CrossRef] [Google Scholar]
  117. Zeimann, G. R., White, R. L., Becker, R. H., et al. 2011, ApJ, 736, 57 [Google Scholar]

Appendix A: Additional tables

Table A.1.

Details of the GROND observations of selected known QSOs, sorted by right ascension2.

Table A.2.

Full GROND photometry of all 116 QSOs observed3.

Table A.3.

Log of spectroscopic observations of QSO pair candidates in 2019. All are 900 sec exposures with grating R150 and a 09 slit, except for the FIRE spectrum (1″ slit).

Table A.4.

Pair candidates with existing ALLWISE channel W1/W2 photometry.

Table A.5.

Pair candidates without ALLWISE photometry.

Table A.6.

Pair candidates from Table A.5 with unWISE or CatWISE photometry (AB system).

All Tables

Table 1.

Comparison of search volumes and depth for previous, serendipituously found high-z QSO pairs.

Table A.1.

Details of the GROND observations of selected known QSOs, sorted by right ascension2.

Table A.2.

Full GROND photometry of all 116 QSOs observed3.

Table A.3.

Log of spectroscopic observations of QSO pair candidates in 2019. All are 900 sec exposures with grating R150 and a 09 slit, except for the FIRE spectrum (1″ slit).

Table A.4.

Pair candidates with existing ALLWISE channel W1/W2 photometry.

Table A.5.

Pair candidates without ALLWISE photometry.

Table A.6.

Pair candidates from Table A.5 with unWISE or CatWISE photometry (AB system).

All Figures

thumbnail Fig. 1.

Schematic of the GROND optical (griz′) and NIR (JHK) channel spatial coverage at a redshift of 6. The dashed boxes are the respective griz′ FOV for each individual telescope dither position; a similar pattern applies to the JHK channels (not shown). With most of our targets being within 1 arcmin from the centre of the FOV, and since we require the riz′ bands, our conservative search radius is 80″, corresponding to about 471 h−1 kpc proper, or 3.3 h−1 Mpc comoving.

In the text
thumbnail Fig. 2.

Colour-colour diagram in the GROND-intrinsic izJHKs bands of the prime quasars (in blue; triangles denote limits) compared to normal foreground stars and galaxies (in grey) of the same fields, with selected brown dwarfs from earlier GROND observations (in green and purple, and labelled with their spectral types). Our 26 candidates are shown in red; the nine objects which have ALLWISE detections are in light red. The yellow-shaded region indicates the search region for the QSO pair. The arrow at the top right of each panel shows the amount by which a correction for AV = 1 would shift an object. Except for the grey objects, all are foreground AV-corrected.

In the text
thumbnail Fig. 3.

Our upper limit (red triangle) on the projected correlation function of z ∼ 6 QSOs, as derived from the GROND observations. The two filled circles are the corresponding findings by Shen et al. (2010) for 7 and 8 pairs, respectively, modified for the different luminosity function (and cosmology) used over time, and adapted to our search scale of 0.1–3.3 h−1 Mpc according to the −2 power-law scale dependence. Open symbols indicate the serendipitous discoveries from Schneider et al. (2000), Djorgovski et al. (2003), and McGreer et al. (2016), also extrapolated to our search scale. The horizontal error bars show the corresponding redshift range of the studies. The labels at the data points report the limiting i-/z-band AB magnitude of the search (see Col. 5 of Table 1) and the corresponding absolute M1450(AB) limiting magnitude. The two black dashed lines depict wp(R) for the same rate of 7 and 8 pairs, as in Shen et al. (2010), extrapolated to higher redshift with their same M1450(AB), and unchanged luminosity function, suggesting higher wp(R) at higher redshift. Similarly, the dotted line shows the extrapolation for the case of constant limiting z-band magnitude. The red line depicts the backwards extrapolation of our upper limit to lower redshifts, at our M1450(AB) < −26 mag sensitivity, demonstrating the mismatch with respect to the findings at lower redshift. All the lines are computed via integration over the luminosity function of Shen et al. (2020) in redshift steps of 0.25, and proper redshift-dependent transformation between observed z′ and absolute magnitudes.

In the text
thumbnail Fig. 4.

Fraction of BH pairs over redshift in TNG300, for two values of ϵr. Following Silverman et al. (2020), pairs are selected to have primary BH luminosities of Lbol,pri ≥ 1045.3 erg s−1, secondary BH luminosities of Lbol,sec ≥ Lbol,pri/10, a minimum BH mass of 108 M, and a minimum host-galaxy stellar mass of 1010 M. At redshifts > 5 there are AGN, but no pairs.

In the text
thumbnail Fig. 5.

Number of brown dwarfs among our sample of 26 candidates per total area of our search region over 116 QSOs, compared to the cumulative surface density as modelled by Ryan & Reid (2016), plotted here for their COSMOS sample field (central line) as well as the fields with the smallest and largest numbers (corresponding to different galactic latitudes) providing the covered range (shaded area). Our data points are plotted at the faintest magnitude per spectral type range, and the horizontal bar covers the magnitude range up to the brightest. While the observed number of M dwarfs is compatible with the known surface density, that of T dwarfs is substantially larger.

In the text
thumbnail Fig. 6.

Le PHARE fits for the ten objects from Tables A.4 and A.5, for which a QSO template provides the better fit, plus the one with a best-fit galaxy template (from top left to bottom right): ULASJ0148+0600, CFHQSJ1509−1749, PSOJ009.7355−10.4316, PSOJ011.3899+09.0325_C2, PSOJ045.1840−22.5408, PSOJ071.0322−04.5591, PSOJ261.0364+19.0286_C1, PSOJ267.0021+22.7812_C2, SDSSJ2054−0005_C3, SDSSJ2054−0005_C4, IMSJ2204+0012. These are the SEDs after correction of the galactic foreground AV, and most of the dwarf fits return an even worse reduced χ2 for the no-AV SEDs than shown here.

In the text
thumbnail Fig. 7.

Long-slit spectra binned at 160 and 400 Å (thin and thick lines, respectively) obtained with the GMOS spectrographs at the Gemini North (top panel) and South (lower panel) Observatories. The vertical solid lines denote the position of Ly-α of its ‘mother’ QSO, the vertical dashed line the position of Ly-α based on the Le PHARE fits. The dwarf templates are at arbitrary normalizations.

In the text

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

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

Initial download of the metrics may take a while.