The Discovery of a Gravitationally Lensed Quasar at z = 6.51

Luminous quasars at z > 6 allow detailed studies of the evolution of supermassive black holes (SMBHs) and the intergalactic medium (IGM) at early cosmic times. To date, ∼150 quasars have been discovered at z > 6, with the highest redshift at z = 7.54 (Bañados et al. 2018). Detections of such objects indicate the existence of billion solar mass (M_⊙) SMBHs merely a few hundred million years after the first star formation in the universe and provide the most stringent constraints on the theory of early SMBH formation (Volonteri 2012).

Much of our understanding of the nature of high-redshift quasars assumes that their measured luminosities are intrinsic to the quasars themselves. However gravitational lensing can substantially brighten quasar images. This effect is particularly important in flux-limited surveys, which are sensitive to the brightest sources; the resulting magnification bias (Turner 1980) could cause a significant overestimation of the SMBH masses powering these objects. A large lensing fraction among the highest-redshift luminous quasars has long been predicted (Comerford et al. 2002; Wyithe & Loeb 2002a) and was suggested as a solution to the difficulty in forming billion M_⊙ SMBHs in the early universe. However, the two highest-redshift-known lensed quasars are at z ∼ 4.8 (McGreer et al. 2010; More et al. 2016), discovered in the Sloan Digital Sky Survey (SDSS); no multiple imaged systems were discovered at 01 resolution among the more than 200 quasars at z = 4–6.4 observed in two Hubble Space Telescope (HST) programs (Richards et al. 2006; McGreer et al. 2014). The lack of high-redshift lensed quasars has been a long-standing puzzle. The solution could be either a reduced magnification bias due to a flat quasar luminosity function (Wyithe 2004) or a strong selection effect against lensed objects arising from the morphology or color criteria used in quasar surveys (Wyithe & Loeb 2002b).

In our wide-area survey of luminous z ∼ 7 quasars (Wang et al. 2017), we discovered an ultraluminous quasar UHS J043947.08+163415.7 (hereafter J0439+1634) at z = 6.51. Subsequent HST imaging shows that it is a multiple imaged gravitationally lensed quasar, the most distant strongly lensed quasar yet known. We present the initial discovery and follow-up imaging observations that confirm its lensing nature in Section 2. In Section 3, we present the lensing model in detail. In Section 4, we discuss the possibility of a large number of high-redshift lensed quasars missed in previous surveys due to bias in color selection. We use a ΛCDM cosmology with Ω_M = 0.3, Ω_Λ = 0.7, and H₀ = 70 km s⁻¹.

2.1. Photometric Selection and Initial Spectroscopy

J0439+1634 was selected by combining photometric data from the UKIRT Hemisphere Survey (UHS; Dye et al. 2018) in the near-infrared J band, the Pan-STARRS1 survey (PS-1; Chambers et al. 2016) at optical wavelengths, and the Wide-field Infrared Survey Explorer (WISE; Wright et al. 2010) archive in the mid-infrared. It was chosen as a high-redshift quasar candidate based on it having a z-band dropout signature with z_AB = 19.49 ± 0.02, y_Vega = 17.63 ± 0.01, and a red z_AB − y_AB = 1.86 ± 0.02, along with a blue power-law continuum (J_Vega = 16.52 ± 0.01, y_AB − J_Vega = 1.11 ± 0.02), and a photometric redshift of z ∼ 6.5. The object has a weak i-band detection in PS-1 (i_AB = 21.71 ± 0.05), but is strongly detected in all bands in the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), at J_Vega = 16.48 ± 0.12, H_Vega = 15.96 ± 0.17, and J_Vega = 15.06 ± 0.13, respectively, as well in all four WISE bands, with Vega magnitudes of 13.98 ± 0.03, 13.24 ± 0.03, 10.28 ± 0.08, and 7.17 ± 0.13, respectively, from W1 to W4 (Schlegel et al. 1998). This object is in an area of the sky with high galactic extinction with E(B−V) = 0.60 (Schlegel et al. 1998); the J_Vega magnitude in UHS after correcting for extinction becomes 15.98.

The initial identification spectrum, obtained on 2018 February 6, with the Binospec optical spectrograph (Fabricant et al. 2003) on the 6.5 m MMT telescope, shows a prominent break consistent with a strong Lyα line at z ∼ 6.5. Follow-up optical and near-infrared spectra were acquired with MMT/Binospec, the Low Resolution Imaging Spectrograph (LRIS; Oke et al. 1995) on the 10 m Keck I Telescope, and the GNIRS instrument (Elias et al. 2006) on the 8.2 m Gemini-North Telescope. The combined optical–IR spectrum is shown in Figure 1. Strong Mg ii emission is detected by GNIRS, yielding a redshift of z = 6.511 ± 0.003.

Zoom In Zoom Out Reset image size

J0439+1634 is roughly 40% brighter than the luminous z = 6.30 quasar SDSS J0100+2802 (Wu et al. 2015), making it the brightest quasar known at z > 5. It is also the brightest submillimeter quasar at z > 5; it is detected by the SCUBA-2 instrument (Holland et al. 2013) on the James Clerk Maxwell Telescope (JCMT) with a total flux of 26.2 ± 1.7 mJy at 850 μm. However, its high luminosity is likely not intrinsic, but instead boosted via gravitational lensing. The optical spectrum of J0439+1634 shows a faint, continuous trace at λ < 9000 Å, visible in the middle of the deepest region of quasar Gunn-Peterson absorption at 8500 Å < 9000 Å (z_abs > 6). This trace extends beyond the quasar Lyman Limit at λ < 6840 Å, blueward of the intergalactic medium (IGM) transmission spikes in the quasar Lyβ region. No quasar continuum transmission is expected at these wavelengths due to the extremely high IGM optical depth (Fan et al. 2006), indicating the presence of a foreground object within the 1'' spectroscopic slit. The lensing hypothesis is further supported by the presence of a very small quasar proximity zone (Figure 1) and an apparent super-Eddington accretion rate based on the Mg ii measured SMBH mass (Figure 2), both of which can be explained with a significant lensing magnification.

Zoom In Zoom Out Reset image size

2.2. High-resolution Imaging

J0439+1634 appears as an unresolved point source on archival PS-1 and UHS images (seeing of ∼15) and on deeper near-infrared images taken with the Fourstar instrument (Persson et al. 2013) on the 6.5 m Magellan-1 Telescope (seeing ∼08). To test the lensing hypothesis, we obtained a high-resolution K-band image using the Advanced Rayleigh guided Ground layer adaptive Optics System (ARGOS; Rabien et al. 2018) on the 2 × 8.4 m Large Binocular Telescope, with a ground-layer adaptive optics (AO) corrected FWHM of 024. This image (Figure 3(A)), taken with the LUCI (Buschkamp et al. 2012) instrument, marginally resolves J0439+1634 beyond the point-spread function (PSF; FWHM = 030 ± 001).

Zoom In Zoom Out Reset image size

Even more revealing are the high-resolution observations of J0439+1634 with the Advanced Camera for Surveys (ACS) on the HST, taken on 2018 April 3, using two intermediate-band (Δλ ∼ 200 Å) ramp filters (Figure 1). The FR782N observation is centered at 7700 Å, fully covers the quasar Lyβ emission, and is the shortest wavelength at which quasar emission is still detectable, thus providing the highest possible spatial resolution of 0075. The FR853N observation is centered at 8750 Å, within the Gunn-Peterson trough, and images only the foreground galaxy. The "galaxy+quasar" FR782N image (Figure 3(B)) clearly resolves the system into multiple components: there are at least two point sources separated by 02 and a faint, extended source ∼05 to the east, which we interpret as the lensing galaxy. The "galaxy-only" FR853N image (Figure 3(C)) shows only the lensing galaxy, best fit with an exponential profile, an ellipticity of ∼0.65, and an effective radius of ∼04.

2.3. Properties of the Lensing Galaxy

A purely photometric fit of the HST/ACS FR782N data using only two quasar images has a significant residual, suggesting a more complex lensing configuration. We fit a singular isothermal ellipsoid lensing model, fixing the lens position and ellipticity (e = 0.65) to match the observed galaxy in the FR853N image, while varying the lens mass and position angle along with the source position to reproduce the observed configuration (Keeton 2001). We vary the Einstein ring radius and position angle of the galaxy along with the position of the source. For each set of parameters, we solve the lens equation to predict the positions of the images, place copies of the HST PSF at those positions, and compare with the FR782N image to compute a χ² goodness of fit. We then use Markov Chain Monte Carlo methods to sample the parameter space. The resulting model is depicted in Figure 3 and the parameters are summarized in Table 1. To interpret the Einstein radius, we assume the galaxy is a thin rotating disk such that the projected ellipticity reflects the inclination, and we compute the corresponding circular velocity (Keeton & Kochanek 1998). A three-image model is preferred (Figure 3(D)), with a best-fit Einstein radius of θ_E = 017 ± 001, corresponding to a circular velocity of ${v}_{c}={160}_{-6}^{+8}\ {\mathrm{kms}}^{-1}$ and a high total magnification of 51.3 ± 1.4. In this model, the separation of the two brighter images is only 004, unresolved even by HST.

Table 1.  Lens Model Parameters

	Fiducial Three-image Model	Alternate Four-image Model	Alternate Two-image Model
Image 1	(ΔR.A., ΔDecl.) ≡ (0, 0)	(0, 0)	(0, 0)
	μ = 5.4 ± 0.1	μ = 1.4	$\mu ={3.9}_{-0.1}^{+0.3}$
Image 2	(−0.032, −0.233)	(−0.027, −0.233)	(−0.033, −0.215)
	μ = 21.8 ± 0.7	μ = 5.1 ± 0.1	$\mu =-{19.3}_{-1.2}^{+0.8}$
Image 3	(−0.035, −0.192)	(−0.060, −0.203)	⋯
	μ = −24.2 ± 0.7	μ = −2.7 ± 0.1
Image 4	⋯	(0.045, −0.200)	⋯
		μ = −1.2 ± 0.1
Source	(0.215, 0.076)	(−0.005, −0.118)	(−0.025, −0.107)
	μtot = 51.3 ± 1.4	μtot = 10.4 ± 0.2	${\mu }_{\mathrm{tot}}={23.1}_{-0.8}^{+1.4}$
Lens	(0.438, 0.055)	(−0.004, −0.171)	(−0.028, −0.125)
	θE = 0168 ± 0001	θE = 0051 ± 0001	θE = 0095 ± 0001
	${v}_{c}={160}_{-6}^{+8}$ km s−1	${v}_{c}={88}_{-3}^{+4}$ km s−1	${v}_{c}={121}_{-4}^{+6}$ km s−1
	e = 0.65, PA = 103.1 ± 0.1	e = 0.8, PA = 101.8 ± 0.6	e = 0.2, ${\rm{PA}}={112.8}_{-7.5}^{+6.0}$

Download table as:  ASCIITypeset image

We estimate the observed optical luminosity at rest-frame 3000 Å to be (4.35 ± 0.09) × 10⁴⁷ erg s⁻¹ by fitting the calibrated spectrum. Applying an empirical factor (Shen et al. 2011) to convert the luminosity at 3000 Å to the bolometric luminosity gives L_bol = 2.24 × 10⁴⁸ erg s⁻¹ = 5.85 × 10¹⁴ L_⊙. After correction for magnification factor of 51.3, the bolometric luminosity of J0439+1634 is reduced to 1.14 × 10¹³L_⊙, and the SMBH mass to 4.29 ± 0.60 × 10⁸ M_⊙. This corresponds to an Eddington ratio of 0.83 ± 0.12.

However, this model seems to underpredict the flux of the faintest quasar image. It is not clear whether the discrepancy is due to limitations in the current data (e.g., in the HST PSF model) or to fundamental problems with this class of lens models. As an alternative, we consider the possibility that the lens galaxy could actually lie between the quasar images and be blended with them. In this scenario, the galaxy light detected in the HST image could be offset from the mass centroid, due perhaps to strong dust obscuration. For example, if the lensing galaxy is seen mostly edge-on, then we might have detected only the part of the galaxy with the highest surface brightness along the disk. The smallest residuals are obtained for a highly inclined galaxy with projected ellipticity e = 0.8, which produces four images and a total magnification of 10.4 ± 0.2 (see Figure 4 and Table 1). The implied circular velocity ${v}_{c}={88}_{-3}^{+4}$ km s⁻¹ is quite low, comparable to that of the Large Magellanic Cloud. The orientation is consistent with the hypothesis that the observed galaxy light is from part of the disk. It also possible that the nearby galaxy is not related to the lensing. In this case, the true lens galaxy is too faint for detection here, could lie between the quasar images, and be relatively round. We therefore test a third model with ellipticity e = 0.2, which produces just two images that have a total magnification of ${23.1}_{-0.8}^{+1.4}$. This model has a modest circular velocity of ${v}_{c}={121}_{-4}^{+6}$ km s⁻¹.

Zoom In Zoom Out Reset image size

We consider the fiducial three-image model to be the most likely lensing configuration because it naturally places the center of the lensing galaxy at the position of the detected galaxy image in the two HST bands. However, further observations are needed to clearly distinguish between the different models. Images that are deeper than the current HST observation could fully characterize the lensing galaxy, while observations with higher spatial resolution (possible only with JWST or ALMA) would reveal whether there are two, three, or four image components.

The probability that a luminous quasar is gravitationally lensed with magnification factor μ > 2 at z ∼ 6 ranges from ∼4%, if the bright end of the quasar luminosity function is Φ(L) ∝ L^−2.8 (Jiang et al. 2016), to ∼20%, if the quasar luminosity function is as steep as Φ(L) ∝ L^−3.6 (Yang et al. 2016). Yet J0439+1634 is the first strongly lensed quasar discovered at z > 5 among the several hundred quasars known at this redshift. A reexamination of the color selection used in previous high-redshift quasar surveys suggests a strong selection bias against lensed quasars.

Selecting z ≳ 6 quasars requires either a nondetection (Jiang et al. 2016; Wang et al. 2017) or a strong drop in the dropout band below the quasar Lyman break (Mazzucchelli et al. 2017). The presence of a lensing galaxy, however, introduces flux into the dropout bands when the image is not fully resolved. Most lensing galaxies are expected to be massive galaxies at z ∼ 0.5–1.5 and to have detectable r- or i-band flux in the SDSS or PS-1 survey. For example, among the 62 lensed z < 4 quasars in the SDSS sample (Inada et al. 2012) with measurements of the lensing galaxy, the faintest lens has i_AB = 21.64. On the other hand, the J0439+1634 lens is among the faintest lensing galaxies known, with i_AB = 22.47. The faintness of this lens, combined with the high apparent luminosity of the lensed quasar, minimizes the impact of lensing galaxy flux to the overall unresolved quasar+lens color used in candidate selection. If the lens were brighter by even 0.5 mag, J0439+1634 would not have been selected as a high-redshift quasar candidate by our color selection criteria (Wang et al. 2017; Mazzucchelli et al. 2017), suggesting that previous surveys have potentially missed the majority of lensed quasars at the highest redshifts due to their stringent dropout criteria. Thus, a full modeling of quasar+lens colors and selection procedure modifications are needed to cover the majority of the high-redshift lensed quasar population.

A statistical study of strong lensing properties using the Millennium Simulation (Hilbert et al. 2008) shows that for a source at z = 5.7, only 5% of the lensing optical depth is provided by galaxies with a halo mass lower than 7 × 10¹¹ M_⊙, comparable to J0439+1634's lensing galaxy (v_c = 160 km s⁻¹) in the fiducial three-image lensing model. This implies up to ∼20 lensed high-redshift quasars could have been missed in our survey due to contamination from lensing galaxy light. Benefiting from the boosted flux, an object such as J0439+1634 is a powerful probe of the physical properties of quasars and their host galaxies as well as serving as an ideal background source for studying high-redshift metal absorption lines and early IGM chemical enrichment.

We acknowledge the support of the staff at the MMT, Magellan, LBT, Keck, and Gemini Telescopes, and thank the Directors of LBTO, Gemini Observatory, JCMT, and STScI for granting us Director Discretionary time for follow-up observations of this object. X.F., J.Y., M.Y., and I.D.M. acknowledge support from US NSF grant AST-1515115, NASA ADAP grant NNX17AF28G, and HST-GO-13644 grant from the Space Telescope Science Institute. C.K. acknowledges support from US NSF grant AST-1716585. A.I.Z. acknowledges support from NSF grant AST-1211874. F.P. acknowledges support from the NASA Chandra award No. AR8-19021A and from the Yale Keck program No. Y144. B.P.V. and F.W. acknowledge funding through the ERC grants "Cosmic Dawn" and "Cosmic Gas." R.W. and X.-B.W acknowledge support from NSFC grant No. 11533001.

Facilities: UKIRT - United Kingdom Infrared Telescope, WISE - Wide-field Infrared Survey Explorer, PS-1 - Panoramic Survey Telescope and Rapid Response System Telescope #1 (Pan-STARRS), MMT - MMT at Fred Lawrence Whipple Observatory, Magellan - , LBT - Large Binocular Telescope Observatory, Gemini - , Keck - , JCMT - James Clerk Maxwell Telescope, HST - Hubble Space Telescope satellite.