SDSS J2222+2745: A GRAVITATIONALLY LENSED SEXTUPLE QUASAR WITH A MAXIMUM IMAGE SEPARATION OF 151 DISCOVERED IN THE SLOAN GIANT ARCS SURVEY^*

The spectroscopic observations described in this paper were obtained using the Andalucia Faint Object Spectrograph and Camera (ALFOSC) at the NOT. We used grism 4, covering a wavelength range 3800–9000 Å with a resolution of about R = 350 and dispersion 3.0 Å pixel⁻¹. The ALFOSC pixels were binned by a factor of two in the dispersion direction but kept unbinned at the native pixel scale of 0188 pixel⁻¹ in the spatial direction. The slit width was 13. Redward of ∼7000 Å strong fringing is present in the spectra, effectively restricting the upper limit of the spectral range to this value unless special care is taken during the observations to suppress the effects of fringing. An order-sorter filter was used to suppress second-order contamination in the spectra.

On 2012 August 21–22, a series of ALFOSC spectra were obtained in the area of SDSS J2222+2745: three 1200 s spectra were obtained with the rotation angle set to provide simultaneous coverage of the A and B images. In addition, two single spectra of 1000 s exposure time were obtained to simultaneously cover image C and galaxy G1 and image B and image C, respectively. Finally, the blue giant arc and the stellar source superposed on it (labeled "WD" in Figure 1) were observed for a total exposure time 3 × 1000 s. The spectra were wavelength calibrated using He+Ne arc lamps and flux calibrated based on observations of the spectrophotometric standard star BD+33 2642. As shown in Figure 2, the three images A, B, and C all have identical spectra modulo noise, containing typical spectral features of a quasar (e.g., Vanden Berk et al. 2001) with a set of prominent emission lines, redshifted to z = 2.82. This demonstrates that the lensing interpretation is correct for these three quasar images.

Zoom In Zoom Out Reset image size

Despite having a significantly redder color, the stellar source superposed on the giant arc was also targeted for ALFOSC spectroscopy as a potential additional image of the quasar. Although a significant continuum was detected, emission lines were completely absent at the wavelengths of the prominent quasar emission lines in the A, B, and C component spectra. A more careful analysis of the spectrum reveals the typical signatures of a DZ white dwarf (WD; e.g., Koester et al. 2011).

Having confirmed the three images A–C as gravitationally lensed images of the same quasar, two alternative image configurations may be considered: SDSS J2222+2745 is either a naked-cusp lens with three quasar images, similar to SDSS J1029+2623 (Oguri et al. 2008) or a five-image system similar to SDSS J1004+4112 (Inada et al. 2008). Being able to distinguish between the two cases is important, not only for the understanding of the SDSS J2222+2745 lens system itself, but also in the context of using the statistics of cluster-lensed quasars as a cosmological tool. Naked cusps, i.e., lensing configurations where the source lies on a tangential caustic that extends outside a radial caustic, are extremely rare among galaxy-scale lenses, the only known candidate among ∼100 such lens systems being APM 08279+5255 (Lewis et al. 2002). However, due to their shallower central matter density slopes, naked-cusp lenses are expected to be much more common for clusters than for galaxies. Oguri & Keeton (2004) predicted that 30%–60% of all cluster lenses will be of this type, the fraction depending on the central slope and triaxiality of clusters.

To further investigate the lensing system configuration of SDSS J2222+2745, we searched our initial MOSCA imaging for potential stellar sources with blue colors consistent with the quasar. A blue source, labeled "D" in Figure 1, was identified just south of, and partially superposed on, galaxy G2. The source was apparently stellar, but the modest exposure time and overlap with G2 made this conclusion somewhat uncertain. A set of deeper ugri images in better seeing was obtained at the NOT in 2012 September (see Table 1). These images revealed image D to have a clearly stellar point-spread function (PSF), motivating further spectroscopic follow-up of this image. On 2012 September 16–17 UT, a series of 8 × 2400 s integrations were made using ALFOSC, using the same setup as for the earlier observations described in Section 3.1. The telescope pointing was dithered along the slit between exposures to improve fringe suppression at the red end of the spectrum. The spectra were flux calibrated based on observations of the spectrophotometric standard star BD+33 2642.

The slit orientation is illustrated in Figure 3, providing simultaneous spectroscopic coverage of images B and D, galaxies G2 and G3, and the giant arc. Inspection of the reduced two-dimensional spectrum (see Figure 4) reveals the strong spectral line in the quasar B image around 4640 Å, corresponding to Lyα at the redshift of the quasar. An Lyα line is also clearly evident at this wavelength in the spectrum of the D image, confirming its identification as a gravitationally lensed image of the quasar. More surprisingly, an Lyα line at the same redshift is also seen superposed on the spectrum of galaxy G3, providing clear indication of a fifth (E) quasar image spatially coincident with this galaxy. The extracted one-dimensional spectra of images D and E in a ∼1200 Å wide region surrounding the Lyα line of the quasar are plotted in Figure 5: a clear peak is seen at the location of the Lyα line in both the D and E image spectra. Redward of this line, there is some visible discrepancy in the slopes of the two spectra. This discrepancy is caused by the spatial superposition of image E and the red galaxy G3, the latter having a continuum which is rising toward longer wavelengths.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 6 shows a scaled subtraction of the second-epoch i-band image from the g-band image (see Table 1). The PSF of the i-band image was matched to that of the g-band image by convolving it with an elliptical Gaussian kernel. The i-band image was then scaled so that the subtraction optimally removes the light of galaxies G1–G3. Any remaining flux, shown in the negative in Figure 6, is bluer than galaxies G1–G3, and should be considered candidate images of the quasar.

Zoom In Zoom Out Reset image size

In Figure 6, image D is obvious. A comparison of this residual image to the residual quasar images A–C shows that it has a PSF consistent with a point source, as expected. Additionally, there is excess blue flux at the spectroscopically indicated position of image E; this flux is obviously extended, however. Figure 7, which shows the u-band image with contours from the g − i subtraction overlaid, shows the reason for this; the excess blue light is a combination of light from the core of galaxy G3, and offset from this, excess blue light attributable to image E.

Zoom In Zoom Out Reset image size

Finally, in Figure 6 there is also an additional blue excess source, near galaxy G1, which we label F in Figures 6 and 7. This source is consistent with being a point source. One of the slit placements from the initial spectroscopic observations includes this possible quasar image, and Figure 5 includes a spectrum extracted at that location. The extracted spectrum suggests the presence of Lyα emission; taken together, the imaging and spectroscopic data suggest that this source is a sixth image of the quasar.

To achieve the most accurate photometry of images of the quasar, we have constructed a model of the light distribution in the cluster core region using GALFIT (Peng et al. 2010). The input PSF for each of the models discussed below was in all cases constructed by fitting bright, isolated and unsaturated stars using a multi-component Moffat profile using GALFIT. Typically three Moffat components were used; we have explored the effects of PSF reference star selection and find that it introduces an uncertainty of less than 0.01 mag for the brighter quasar images, with no significant biases.

The GALFIT model of the light distribution was initially constructed using a stack of the second-epoch g, r, and i images. Sources were initially placed at locations of obvious galaxies, and at the six quasar image positions discussed above. Quasar images are PSF components, and all galaxies are Sérsic profile components. We continued adding components to the model iteratively at the locations of any significant residuals revealed by the modeling process, until a final complete model incorporating all significant sources was achieved. This model was then translated to individual epoch and filter images by transforming the image coordinates to match the image to be analyzed, fixing those coordinates, and then refitting the model component magnitudes, sizes, and shapes to the image to be analyzed.

The modeling process described above reveals that the images of the quasar include a contribution from the underlying host galaxy, as is seen in the two other quasars strongly lensed by galaxy clusters (Inada et al. 2005; Oguri et al. 2013). The data show that the host galaxy is blue—unlike the red hosts seen in the other two quasar-cluster lenses—and is most clearly seen in the brightest (and thus most likely the most magnified) image A. Figure 8 shows residuals from the full GALFIT model around image A; only the components for image A have been removed to specifically highlight residuals at that location. Two model variants are shown—one in which only a PSF component is used, and a second model for A consisting of a PSF component plus a single Sérsic component; this second model is a much better fit at that location.

Zoom In Zoom Out Reset image size

In the g band, including the host galaxy makes the fitted quasar 0.086 mag fainter. In our judgment, none of the other images of the quasar allow for clear modeling of the host galaxy component—image B is bright enough, but complicated by a second galaxy less than 2 arcsec to the northwest, and images C–F are insufficiently magnified. Thus, for images B–D, in order to measure the apparent magnitude of each quasar image as robustly as possible, we proceed by measuring the quasar brightness using a simple PSF-only fit in GALFIT, and then correct each measured magnitude fainter by 0.086 mag—i.e., by the fractional host galaxy contribution as measured from image A. This presumes that the quasar and host galaxy magnifications at each image location are the same; providing the host is compact and not too close to a caustic this assumption is reasonable and this supposition is well supported by the lack of giant arc images associated with the quasar host. This approach also in principle presumes no temporal variation in the quasar intrinsic brightness, since such variations would produce variation in the fractional correction. While not formally true (see below), that effect is second order and insignificant; the simple procedure we follow here for correcting for host galaxy light avoids systematic noise that would be introduced by attempting to model the host galaxy independently in multiple images. For reference, in addition to the g band above, the correction for the host galaxy in r and i is measured to be 0.084 and 0.071 mag, respectively. Within the uncertainties, this correction is the same across bands—i.e., the host galaxy has a color comparable to the quasar. For the u band, the image is of insufficient depth to model the host galaxy directly, we adopt the same correction as the g band. The resulting photometry for images A–D, across four bands and two epochs, is reported in Table 3. Uncertainties are calculated by using the full GALFIT model to generate realizations of each measured frame with appropriate noise, which are then refit. The measured scatter in the photometry of quasar images, combined in quadrature with the estimated calibration uncertainty, is reported. For the brighter images, systematics due to uncertain treatment of the host galaxy are likely dominant, and not addressable with these limited data.

Images E and F must be treated with further care. The photometry of image E is complicated by the blue core of galaxy G3, and image F is simply very faint, and near the core of the brightest cluster galaxy, G1. Particularly at redder wavelengths, this makes its measurement highly uncertain. Furthermore, tests using the full GALFIT model, in which we remeasured noisy realizations of the full model, also show potentially strong biases—up to factors of three—in the photometry of image F. To attempt best possible photometry for these, we proceeded as follows.

First, note that if the colors of the various quasar images are the same, then the relative photometry of residual images in the g − i subtraction in Figure 6 can be used to measure the fainter images. Matched colors are not guaranteed, due both to intrinsic variability (e.g., Schmidt et al. 2012) coupled with time delays, as well as variations induced by differential magnification between the quasar and underlying host galaxy, and possible extinction of individual images due to intervening galaxies in the complex cluster core. However, we can validate the likely variability in color between the images by considering the three brightest and most isolated images (A–C); these show that the typical scatter about a fixed color is 0.05 in g − i and 0.03 in g − i. This uncertainty is much less than the systemic biases noted above in attempts at direct measurement, and so the fixed color assumption is sufficient given these data.

Thus for image F, we use difference images to measure the differential flux of image F relative to images A–C. Assuming that F has the same color as these images, the flux of image F is then given simply by the mean flux of images A–C times the flux ratio in the difference image. All magnitudes are measured using PSF fitting photometry for consistency. The difference image g − i provides the best measurement of the residual fluxes, and is the basis of the measurements reported in Table 3. As before, we estimate uncertainties by using the fitted GALFIT model to generate and refit noisy realizations of the measured image. In the case of image F, the final reported uncertainty is the scatter in the magnitude offset in the residual image, combined in quadrature with the color scatter between the brighter images.

For image E, we must further account for the residual flux from the core of galaxy G3. We proceed as for image F, but additionally use GALFIT to model a best-fitting Sérsic profile fixed at the center of G3 concurrent with the PSF component for image E.

Finally, we note that each frame is directly calibrated to the SDSS using many hundreds of objects; calibration uncertainties are 0.01 mag at worst.

The maximum separation of the quasar images is 151 (between quasar images A and C) clearly indicative of a cluster-sized lensing mass. The SDSS photometric redshift estimates of galaxies G1–G3 (see Table 2) are consistent with all three galaxies being located in a cluster at z ∼ 0.5. As part of the observations described in Sections 3.1 and 3.2 above, all three galaxies were targeted spectroscopically. Redshift estimates were obtained through cross-correlation with the elliptical galaxy template of Kinney et al. (1996), yielding the redshift estimates listed in Table 4. The redshifts are consistent within the measured uncertainties, and all three galaxies G1–G3 clearly belong to the same cluster at a redshift consistent with the SDSS photometric redshift estimates. To further investigate the properties of the lensing cluster, we applied the SExtractor software (Bertin & Arnouts 1996) for galaxy photometry to the stacked second-epoch MOSCA gri images. Prior to obtaining the photometric measurements, the stacked r- and i-band images were smoothed to match the resolution of the g-band image. Object detection was initially performed on the stacked g-band image, and SExtractor was subsequently run in dual-image mode on the stacked images in the other passbands, using the g-band image as a reference for object detection. The observed magnitudes were subsequently converted to AB magnitudes in the SDSS system using objects in the MOSCA field with cataloged SDSS photometry. Total magnitudes were determined using the SExtractor MAG_AUTO parameter, and colors were determined from aperture magnitudes (MAG_APER) within a 2'' diameter aperture. A g − i versus i color–magnitude diagram of non-stellar objects, selected based on their location in a magnitude–μ_max diagram (where μ_max is the central surface brightness of objects), is shown in Figure 9. A red sequence is evident at g − i ∼ 2.5, similar to the g − i colors of the three galaxies G1–G3 (see Table 4). Figure 10 shows a smoothed map of the light distribution of red sequence galaxies over the observed MOSCA field. A strong peak in the luminosity distribution is located at the position of galaxies G1–G3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 4. Photometry and Redshift Measurements for Galaxies in the Cluster Core

Galaxy	i	g − i	r − i	Redshift
		(20 Aperture)	(20 Aperture)
G1	18.74	2.71	0.97	0.493 ± 0.013
G2	18.97	2.53	0.96	0.496 ± 0.012
G3	19.37	2.45	0.93	0.489 ± 0.013

Download table as:  ASCIITypeset image

The identification of SDSS J2222+2745 as a candidate gravitational lens system was partially due to recognition of the giant arc just south of the cluster center. The extent of the arc in our deepest stacked g-band image (see Figure 3) is 10''. Our SExtractor photometry (see Section 3.4) of the arc yields g = 21.57, and u − g = 0.87, -g − r = 0.21, g − i = 0.41, where the colors are measured in 2'' diameter apertures centered on the brightest central portion of the arc. We note that the quoted magnitude of the arc can be considered a lower limit on its brightness, given the very extended morphology of the arc. Although SExtractor attempts to treat the arc and the WD as separate objects, a more careful modeling of the shape of the arc (including PSF modeling and removal of the superposed WD point source) would be required to obtain an accurate total magnitude. The spectrum of the arc shows a single emission line at λ_obs  =   ∼  4015 Å, which we interpret as Lyα at z = 2.30. The alternative interpretation of [O ii] (3727 Å) at z = 0.077 would put the arc in front of the cluster, which is obviously inconsistent with the lensed morphology of the arc. The lens model presented in the following section does not predict a counter-image of the arc.

We have identified SDSS J2222+2745 as a gravitational lens system producing six observed images of the same quasar. This is the third known cluster-scale lens and the only such lens producing more than five observed images. The only previously known six-image active galactic nucleus lens is B1359+154 (Rusin et al. 2001), where a radio source at z_s = 3.235 is being lensed by a compact group of three galaxies at z_l ∼ 1. However, the quasar image separations we observe in SDSS J2222+2745 are an order of magnitude larger than in B1359+154, and the lensed images are ∼3 mag brighter at optical wavelengths, making six images detectable in ground-based imaging of SDSS J2222+2745. The B1359+154 lensed image configuration is similar to what we find in SDSS J2222+2745, with the three fainter images located in close proximity to the three galaxies situated within the critical curves of the lens system, and three brighter images located further away from these galaxies. However, the lens model of B1359+154 presented by Rusin et al. (2001) does not allow a significant dark matter halo associated with the galaxy group in addition to the three halos corresponding to the three individual galaxies. In contrast, in our model the cluster halo component is the dominant contributor to the observed image splitting of SDSS J2222+2745. Specifically, the cluster halo comprises 61% of the total enclosed mass within R_{E, eff}, 78% of the mass within 100 kpc, and 87% of the mass within 200 kpc. However, as with any parametric multi-component cluster mass model, these numbers will be sensitive to the chosen mass profile parameterization for the different components.

Since cluster-scale quasar lenses and image multiplicities >5 are both rare, we would expect a lensing system such as SDSS J2222+2745 to be extremely rare. There are no appropriate quantitative predictions for the occurrence of this type of lens systems available in the literature. Evans & Witt (2001) calculate the incidence of sextuplet and octuplet image systems (with demagnified seventh and ninth images, respectively) and estimate that these constitute ∼1% of the population of quasars multiply imaged by galaxy-scale lenses. This is consistent with the single known sextuplet case of B1359+154 among the ∼100 known galaxy-scale quasar lenses. For the larger image separations produced by cluster lenses, published studies of the probabilities of different lensing multiplicities (Oguri & Keeton 2004; Li et al. 2007; Minor & Kaplinghat 2008) only include double, quadruple (with demagnified third and fifth images, respectively) and naked-cusp configurations. Li et al. (2007) state that they do not expect a significant fraction of sources to have image numbers exceeding 5.

Both the quasar source redshift and cluster lens redshift of SDSS J2222+2745 are comparable to the two other cluster quasar lenses and in good agreement with theoretical expectations: the observed lens redshift in SDSS J2222+2745 is z = 0.49, which is similar to, but slightly lower than, the cluster lenses in SDSS J1004+4112 (z = 0.68) and SDSS J1029+2623 (z = 0.58). This redshift is very close to the peak (at z ≃ 0.5) of the predicted lens redshift distribution for photometrically selected wide-separation quasars in the SDSS (Li et al. 2007). The observed quasar source redshift in SDSS J2222+2745 is z = 2.82, slightly higher than the sources in SDSS J1004+4112 (z = 1.73) and SDSS J1029+2623 (z = 2.20). These three redshifts are all near the peak (around z = 2) of the predicted source redshift distribution for photometrically selected wide-separation quasars in the SDSS (Li et al. 2007).

The photometry reported in Section 3.3 shows significant variability in the quasar, demonstrating the feasibility of making time delay measurements in SDSS J2222+2745 from monitoring campaigns using ground-based 2–4 m class telescopes. We have recently initiated a monitoring program for SDSS J2222+2745 at the NOT for this purpose. The predicted time delays from our lens model (see Section 4) indicate that three independent time delays Δt_AB (≡ Δt_B − Δt_A), Δt_ED and Δt_FE in the system are all of the order of 100 days. This suggests that several time delay measurements could, in the best case, be obtained from a single season of monitoring. The predicted time delays between image C and the close pair (A–B) and between (A–B) and the central images (D–E–F) are of the order of 1400 days and 700–1000 days, respectively. These values are similar to the two longest predicted time delays for SDSS J1004+4112, of 1218 and 1674 days, respectively (Oguri 2010) and the measured longest time delay in SDSS J1029+2623 of Δt_AB = (744 ± 10) days (Fohlmeister et al. 2013).

Morphological, photometric, and spectroscopic evidence presented in this paper strongly suggest the presence of a sixth (F) image of the quasar located closest to the lensing mass center. The odd number theorem of strong gravitational lensing theory (McKenzie 1985; Petters & Werner 2010) will then require the presence of a seventh (G) image, for which our model predicts a location close to the center of cluster galaxy G2. In the case of galaxy-scale lenses, the faintest image of a multiply lensed quasar is typically highly demagnified and hence unobservable in the vast majority of cases. However, the shallower central slopes of the mass density profiles of clusters increase the chance of observing the final odd image, as exemplified by the identification of the demagnified fifth central image in SDSS J1004+4112 using HST imaging (Inada et al. 2005, 2008).

While the data presented in this paper do not provide conclusive evidence for the presence of the seventh image, there is a hint of Lyα emission from this image in the spectrum of G2. If this feature is indeed emission from the quasar, a firm confirmation would be possible through spectroscopic observations with 8–10 m class telescopes.

While time delay measurements would provide important additional modeling constraints on the SDSS J2222+2745 lens system, follow-up HST imaging will also be crucial to improve the modeling constraints from the observed locations of the quasar images and hopefully provide morphological and photometric confirmation of the predicted seventh (G) image close to the center of galaxy G2. Further constraints on the lens system could also come from spatially resolved, multiply imaged structures within the giant arc and from the identification of additional fainter, multiply imaged, galaxies as revealed by the HST observations of SDSS J1004+4112 (Sharon et al. 2005) and SDSS J1029+2623 (Oguri et al. 2013).

The significance of the discovery of SDSS J2222+2745 is twofold. First, the presence of six quasar images (and potential future confirmation of a seventh image), spread over a range of clustercentric radii, with good prospects for future time delay measurements, coupled with the presence of an additional lensed source producing a bright giant arc, can provide unique detailed measurements of the matter distribution in the core of a cluster. Such measurements provide a stringent test of the scenario of structure formation, including the properties of dark matter in cluster cores.

Second, adding our new discovery of SDSS J2222+2745 to the previous detections of SDSS J1004+4112 and SDSS J1029+2623 makes a significant contribution toward establishing quasars strongly lensed by galaxy clusters as a cosmological tool. The probability of such rare lenses depends very sensitively on cosmological parameters, in particular the normalization of the matter power spectrum σ₈ and the matter density Ω_M (e.g., Li et al. 2007). Even though no directly applicable predictions can be found in the literature, the occurrence of lens systems with image separations and image multiplicities as high or higher than SDSS J2222+2745 must necessarily be exceptionally rare, and this could conceivably be the only such system, provided that the current ΛCDM paradigm is correct.