A Very Large (θ_E ≳ 40'') Strong Gravitational Lens Selected with the Sunyaev–Zel'dovich Effect: PLCK G287.0+32.9 (z = 0.38)

Galaxy clusters are the most massive gravitationally bound objects in the universe, sitting at the high end of the mass function. Their total masses and mass profiles can be inferred in various complementary ways, including from cluster galaxy kinematics (Diaferio & Geller 1997), weak lensing (WL; Bartelmann & Schneider 2001), X-ray measurements (Sarazin 1986), or the Sunyaev–Zel'dovich effect (Sunyaev & Zeldovich 1972).

The abundance of the most massive clusters is important for scaling the mass function and probing cosmological models. N-body numerical simulations predict a universal mass profile form for virialized dark matter (DM) halos (e.g., Navarro et al. 1996), with increasing concentrations toward lower redshifts (as structure evolves) and for smaller masses (Duffy et al. 2008). Generally, the Einstein radius of a lens increases with its overall mass (explicitly, with the inner projected mass density) and with concentration (Sadeh & Rephaeli 2008). Given the shape of the concentration–mass relation, more massive, highly concentrated clusters become rarer. Assuming a mass function and a cosmological model, predictions for the distribution of Einstein radii in the universe can be made (Oguri & Blandford 2009; Waizmann et al. 2012; Redlich et al. 2014). Indeed, more factors can play an important role in determining the actual critical area size, such as the effective ellipticity and distribution of projected substructure (Redlich et al. 2012), or the halo's triaxiality. These factors can be accounted for statistically so that the comparison of the abundance of large lenses to ΛCDM predictions remains interesting.

To date, less than a handful of clusters are known to have critical areas exceeding 1.4 arcmin² for ${z}_{s}=2$, i.e., an effective Einstein radius (defined as $\sqrt{A/\pi }$, with A being the critical area) exceeding 40'' (and only few others are known with such a large critical area even for high-redshift sources). The largest strong lens currently known, with ${\theta }_{{\rm{E}}}\sim 55^{\prime\prime}$ (Zitrin et al. 2009a), is MACS J0717.5+3745, a merging, massive (Medezinski et al. 2013), X-ray selected (Ebeling et al. 2010) galaxy cluster at z = 0.55. In fact, most massive clusters with known large Einstein radii have been chosen for follow-up based on optical classification (e.g., richness and luminosity), or X-ray measurements (Ebeling et al. 2010). However, with the recent abundance of SZ data and related cluster catalogs, such as those from Planck (Planck Collaboration et al. 2011a), the South Pole Telescope (SPT; Vanderlinde et al. 2010), and the Atacama Cosmology Telescope (ACT; Hasselfield et al. 2013), more potentially massive galaxy clusters have been detected.

Here, we present the first strong-lensing (SL) analysis of the Planck galaxy cluster PLCK G287.0+32.9 (PLCKG287 hereafter), which has the second-highest SZ-inferred mass in the Planck catalog¹¹ (${M}_{\mathrm{SZ},500}={13.89}_{-0.54}^{+0.53}\times {10}^{14}\,{M}_{\odot }$; Planck Collaboration et al. 2015).

This Letter is organized as follows: in Section 2, we summarize the observations and data reduction. In Section 3, we review the SL modeling of the cluster. In Section 4, we discuss the analysis results and their potential implications, and in Section 5, we conclude the work. Throughout, we use a standard ΛCDM cosmology with ${{\rm{\Omega }}}_{{\rm{m}}0}=0.3$, ${{\rm{\Omega }}}_{{\rm{\Lambda }}0}=0.7$, ${H}_{0}=100$ h km s⁻¹Mpc⁻¹, and h = 0.7. Magnitudes are given using the AB convention. 1'' equals 5.21 kpc at the redshift of the cluster. Unless noted otherwise, errors are $1\sigma$.

PLCGK287 was discovered in early SZ observations by the Planck telescope (Planck Collaboration et al. 2011a). Follow-up X-ray observations with XMM-Newton yield a temperature of ${T}_{{\rm{X}}}=12.86\pm 0.42\,\mathrm{keV}$ and mass of ${M}_{500}=15.72\,\pm 0.27\times {10}^{14}\,{h}_{70}^{-1}\,{M}_{\odot }$ (Planck Collaboration et al. 2011b).

The cluster was found to show also a non-thermal radio emission characterized by a double relic, indicating the system has undergone a recent major merger (Bagchi et al. 2011; Bonafede et al. 2014). The projected separation of the double radio relic is ∼4 Mpc, the largest known at comparable redshifts. From the XMM-Newton data a separation of ∼400 kpc is measured between the cluster BCG and the X-ray emission peak, providing additional evidence for merger activity.

PLCKG287 was previously imaged with the WFI on the 2.2 m MPG/ESO telescope in VRI filters. In a systematic WL study of SZ-selected clusters, Gruen et al. (2014; see also K. Finner et al. 2017, in preparation) confirmed the large mass estimates from SZ and X-ray and found the system to be the most massive among their sample, with ${M}_{\mathrm{WL},500}\,={19.5}_{-3.2}^{+3.3}\times {10}^{14}\,{h}_{70}^{-1}{M}_{\odot }$. In these optical data, several arc-like features were identified, spread out over a large field around the cluster core, with a radius exceeding an arcminute.

Following these findings, we obtained Hubble Space Telescope (HST) data in Cycle 23 to perform a dedicated SL analysis of this cluster (PI: Seitz, program ID 14165). PLCKG287 was observed on 2016 August 3 in the optical, ACS bands F475W, F606W, and F814W, with total exposure times of 1, 1, and 2 orbits, respectively. The cluster was also observed on 2016 May 18 with the WFC3/IR F110W filter, with four exposures per orbit over four orbits, centered on two locations, and adopting a half-pixel dither pattern. Data reduction was performed using advanced drizzle techniques (Koekemoer et al. 2011), including CR rejection, full astrometric alignment, weighting by the inverse variance, and drizzling, to produce the final set of mosaics.

We ran SExtractor (Bertin & Arnouts 1996) in dual-image mode to obtain the photometry of objects in the cluster field, using the F814W image as reference. Red-sequence cluster members, needed for our modeling, were chosen by a color–magnitude diagram, using the F814W and F606W filters, down to 23 AB. We then ran the Bayesian Photometric Redshift program (BPZ; Benítez 2000) with the photometric catalogs as input, to derive redshift estimates and help in the identification of multiple images.¹²

We had also observed PLCKG287 with VIMOS¹³ on the VLT/UT3 (Le Fèvre et al. 2003) in service mode on four nights in 2015 February–March. We used the medium-resolution red grism and the GG475 filter, covering the $\approx 5000\mbox{--}{\rm{10000}}\,\mathring{\rm A}$ spectral range, with a resolution of $\approx 600$. A total of twenty 1135 s exposures were obtained. These data have been reduced with a mix of custom pipeline and IRAF tasks (apall, onedspec). IRAF rvsao (Kurtz & Mink 1998) was used to estimate the redshifts. From $\geqslant 220$ member galaxies we estimate z = 0.380 for the cluster redshift (updating the z = 0.39 Planck estimate).

We cross-check our red-sequence selection with the VIMOS data to maximize the number of verified members included and remove red non-members. Out of the 248 red-sequence galaxies we consider members for the modeling, 35 within the HST FOV are spectroscopically confirmed. Additional objects suspected as interlopers in a visual inspection were removed.

We use the light-traces-mass (LTM) method by Zitrin et al. (2009b; see also Broadhurst et al. 2005; Zitrin et al. 2015) for the SL analysis of PLCKG287.

The photometry of red-sequence cluster galaxies (Section 2) is the starting point for the model. A power-law surface mass–density distribution is assigned to each galaxy, scaled in proportion to its luminosity (we use here the F606W magnitudes for reference). For the BCGs we typically assign ${\text{}}{elliptical}$ power-law surface mass–density distributions with a core. The power-law exponent is the same for all galaxies and is the first free parameter of the model. The resulting map from the superposition of these galaxies is then smoothed with a Gaussian kernel, whose size is the second free parameter. The smoothed map constitutes the DM component of our model. The two components are then added with a relative weight, which is also iterated for, and scaled with an overall normalization that is the fourth free parameter. To allow for further flexibility, we also add a two-parameter external shear. The total number of free parameters principally is thus six. In addition, we typically also allow for the weight (and possibly core radius and ellipticity) of the brightest galaxies to be freely optimized in the minimization procedure.

Using preliminary LTM models, a method that has been shown to excel in predicting the appearance of multiply imaged galaxies (e.g., Broadhurst et al. 2005; Zitrin et al. 2009a, 2009b), we iteratively identify 20 sets of multiple images and candidates in the HST data (Section 2; see Figure 3). We set their redshift to the corresponding, best-value photometric redshifts, and use them as constraints for constructing the final model presented here. We only fix the redshifts for systems with relatively secure estimates (dropouts, or those where the photo-z agrees well among the different multiple images). The redshifts of the other systems were left free to be optimized in the minimization, as indicated in Table 1 listing the multiple images.

Table 1.  Multiple Images and Candidates

Arc ID	R.A.	Decl.	zphot [95% C.I.]	zmodel [95% C.I.]	Comments
1.1	11:50:49.448	−28:05:02.060	3.76 [3.29–4.23]	=3.75	Radial image
1.2	11:50:49.054	−28:05:04.792	3.72 [3.26–4.18]	''	''
2.1	11:5054.201	−28:05:52.064	1.67 [1.41–1.93]	1.38 [1.30–1.60]	Other iterations yield ${z}_{\mathrm{phot}}\sim 2.3$
2.2	11:50:54.060	−28:05:54.105	1.82 [1.54–2.10]	''	''
2.3	11:50:52.844	−28:06:03.226	1.67 [1.41–1.93]	''	''
3.1	11:50:51.549	−28:05:39.538	2.81 [0.04–3.18]	1.34 [1.21–1.46]	⋯
3.2	11:50:51.197	−28:05:42.271	0.35 [0.09–3.12]	''	⋯
3.3	11:50:53.830	−28:05:10.648	0.93 [0.13–2.56]	''	⋯
4.1	11:50:50.800	−28:05:43.124	1.04 [0.84–1.24]	2.55 [2.14–2.92]	⋯
4.2	11:50:50.667	−28:05:44.562	1.17 [0.96–1.41]	''	⋯
c4.3	11:50:54.469	−28:05:04.469	0.97 [0.13–1.25]	''	⋯
5.1	11:50:50.890	−28:04:31.287	2.55 [0.07–2.93]	1.53 [1.31–1.89]	⋯
5.2	11:50:51.303	−28:04:31.069	2.23 [0.18–2.81]	''	⋯
c5.3	11:50:47.248	−28:05:16.134	1.86 [1.58–2.41]	''	⋯
c5.3	11:50:46.243	−28:05:30.772	3.14 [1.14–3.57]	''	Most likely according to the model
c5.3	11:50:45.865	−28:05:39.327	0.46 [0.13–3.31]	''	⋯
c6.1	11:50:47.021	−28:04:51.560	4.43 [3.90–4.96]	⋯	⋯
c6.2	11:50:47.232	−28:04:42.851	4.51 [0.68–5.05]	⋯	⋯
7.1	11:50:49.252	−28:04:23.606	3.32 [2.90–3.74]	=3.4	⋯
7.2	11:50:52.875	−28:04:23.606	3.47 [3.03–3.91]	''	⋯
7.3	11:50:51.702	−28:05:17.047	3.33 [2.91–3.79]	''	⋯
7.4	11:50:46.982	−28:05:41.706	3.52 [3.08–3.96]	''	⋯
8.1	11:50:52.743	−28:04:48.339	2.44 [1.88 2.78]	0.99 [0.92–1.06]	⋯
8.2	11:50:51.290	−28:05:12.952	3.04 [3.59–3.44]	''	Other iterations yield ${z}_{\mathrm{phot}}\sim 2.4$
8.3	11:50:49.593	−28:05:22.032	1.06 [0.45–1.66]	''	''
c9.1	11:50:50.430	−28:05:26.016	0.87 [0.59–2.78]	⋯	⋯
c9.2	11:50:50.200	−28:05:27.985	0.85 [0.22–4.06]	⋯	Nearby parts yield ${z}_{\mathrm{phot}}\sim 2.5$
10.1	11:50:44.942	−28:05:02.578	0.98 [0.79–1.17]	0.89 [0.71–0.91]	⋯
10.2	11:50:44.929	−28:05:04.072	0.97 [0.78–1.16]	''	⋯
10.3	11:50:44.993	−28:05:06.556	1.02 [0.82–1.22]	''	⋯
c11.1	11:50:44.089	−28:05:09.461	0.10 [0.00–2.54]	⋯	⋯
c11.2	11:50:44.063	−28:05:10.642	0.09 [0.00–3.09]	⋯	⋯
c11.3	11:50:44.202	−28:05:17.428	0.09 [0.00–2.86]	⋯	⋯
12.1	11:50:43.705	−28:05:18.794	3.43 [3.00–3.86]	=3.4	⋯
12.2	11:50:43.598	−28:05:13.836	3.47 [3.03–3.91]	''	⋯
c12.3	11:50:43.603	−28:05:13.483	3.36 [2.93–3.79]	''	⋯
c12.4	11:50:43.821	−28:05:00.398	⋯	''	Blended with another object
c13.1	11:50:55.879	−28:04:25.578	0.81 [0.20–4.98]	⋯	⋯
c13.2	11:50:55.428	−28:04:19.980	4.16 [3.65–4.67]	⋯	⋯
c13.3	11:50:54.995	−28:04:14.134	4.77 [4.21–5.33]	⋯	⋯
14.1	11:50:55.751	−28:04:04.016	0.77 [0.60–1.05]	⋯	Not used as constraint
14.2	11:50:55.855	−28:04:05.816	0.92 [0.60–1.11]	⋯	''
14.3	11:50:56.109	−28:04:07.315	0.75 [0.50–0.92]	⋯	''
c15.1	11:50:50.470	−28:03:54.703	0.54 [0.39–0.69]	⋯	Other iterations yield ${z}_{\mathrm{phot}}\sim 3$
c15.2	11:50:50.239	−28:03:54.937	0.47 [0.33–0.61]	⋯	''
16.1	11:50:53.949	−28:06:17.632	2.97 [0.19–3.36]	⋯	Not used as constraint
16.2	11:50:53.919	−28:06:17.939	⋯	⋯	''
16.3	11:50:53.588	−28:06:19.559	3.37 [2.74–3.80]	⋯	''
c17.1	11:50:45.396	−28:05:00.931	0.83 [0.12–3.87]	⋯	⋯
c17.2	11:50:45.377	−28:05:08.390	3.19 [0.41–3.72]	⋯	Nearby parts yield ${z}_{\mathrm{phot}}\sim 3.7\mbox{--}4$
c18.1	11:50:45.271	−28:05:15.966	0.59 [0.12–3.46]	⋯	⋯
c18.2	11:50:45.255	−28:05:21.258	3.40 [0.69–3.83]	⋯	⋯
c17/18	11:50:46.578	−28:04:23.945	0.71 [0.47–0.97]	⋯	⋯
19.1	11:50:49.203	−28:04:14.689	6.90 [6.13–8.43]	=6.9	Dropout, not detected in ACS bands
19.2	11:50:52.784	−28:04:17.603	6.90 [6.13–8.43]	''	''
19.3	11:50:51.936	−28:05:16.296	6.90 [6.13–8.54]	''	''
19.4	11:50:46.185	−28:05:44.875	6.94 [6.16–8.32]	''	''
c19.5	11:50:50.443	−28:04:53.959	1.35 [0.95–8.83]	''	'', in BCG's light
c20.1	11:50:50.632	−28:05:03.137	0.98 [0.07–1.71]	⋯	In BCG light, radial image
c20.2	11:50:50.521	−28:05:00.919	5.88 [1.60–6.55]	⋯	''

Note. Column 1: ID . "c" stands for candidate where identification was more ambiguous, and image was not used as a constraint. Columns 2 and 3: R.A. and decl. in J2000.0. Column 4: photometric redshift and 95% C.L. from BPZ. Column 5: predicted and 95% C.L. redshift by our lens model, for systems whose redshift was left to be optimized in the minimization (otherwise, a fixed value is listed). Column 6: comments.

Download table as:  ASCIITypeset images: 1 2

For the two central BCGs, we assign ellipticity values measured by SExtractor. We leave the weight of the two central BCGs, and two other bright galaxies, to be optimized in the minimization procedure.

The minimization of the model is performed with several thousand Monte Carlo Markov Chain steps, through a ${\chi }^{2}$ criterion quantifying the reproduction of the multiple image positions (adopting a positional uncertainty of 05, or 14 when calculating the errors). The final model has an image reproduction rms of 19. The resulting critical curves (${z}_{s}=3.4$) are seen in Figure 1 along with the multiply imaged galaxies. The best-model mass map and profile are seen in Figure 2.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

It should be noted that our results are based only on four-band photometric redshifts for the multiple images, and thus may vary strongly depending on the true redshift of the sources. In addition, there is some ambiguity with respect to few potential multiple images (Figure 1, Table 1; for example, which candidates are the true, third counter images of systems 5, 12, or 15). It is also at present unclear what the nature of the $22^{\prime\prime}$ long giant arc is (systems c17 and c18), whether or not it is multiply imaged, and whether or not there is an additional counter image north of the arc. Similarly, if candidate system 13 is a multiply imaged galaxy, this means that the eastern "arm" seen in the surface density map in Figure 2 creates a local critical curve larger that the simple LTM assumption yields. On the other hand, if systems 10 and 11 have a notably higher redshift than our current estimates imply, it might mean the western arm is in fact smaller and consists only of local critical curves around the respective galaxies. Additional uncertainty arises from the bright galaxy at [R.A., decl.] = [11:50:46.747, −28:03:56.560] that was chosen by our red-sequence criteria, but the VIMOS data suggest it is not officially a cluster member (it has a >10,000 km s⁻¹ lower velocity than the cluster average). The galaxy seems to contribute to some extent to the lensing signal, though, and so we leave it in the modeling but with a reduced weight. The lack of obvious multiple-image systems around that northern tip suggests this choice is reasonable.

Zoom In Zoom Out Reset image size

We did not use candidate images as constraints, and, in most ambiguities listed above, we generally favored the conservative choice, so that the actual critical area may, if anything, be larger than our estimate in Section 4.

Our SL analysis reveals a very big lens, with a critical area of 1.55 arcmin², for ${z}_{s}\simeq 2$, corresponding to an effective Einstein radius of ${\theta }_{{\rm{E}}}\simeq 42\pm 4$''. The mass enclosed within these curves is $3.1\pm 0.5\times {10}^{14}\,{M}_{\odot }$. For ${z}_{s}\simeq 3.4$, for example, the critical curves expand further as expected, reaching an effective Einstein radius of ${\theta }_{{\rm{E}}}\sim 49^{\prime\prime}$, and for ${z}_{s}\sim 10$, our model suggests they would reach ${\theta }_{{\rm{E}}}\sim 54^{\prime\prime}$.

To examine the effect of lack of accurate redshifts we ran four other models with significantly different combinations of fixed redshifts. The Einstein radii estimates between the different models is within 10% of each other, and the enclosed masses agree to 15%. The mass profile and correspondingly, predicted redshifts, are more sensitive to the exact redshifts initially adopted, and typically agree to $2\sigma \mbox{--}3\sigma$.

Few clusters are known to have Einstein radii above 40'' (for typical source redshifts of ${z}_{s}\sim 2$). These include the largest known gravitational lens, MACS J0717.5+3745, included also in the Cluster Lensing And Supernova with Hubble (CLASH; Postman et al. 2012) and Hubble Frontier Fields (HFF; Lotz et al. 2016) programs, with ${\theta }_{{\rm{E}}}\sim 55^{\prime\prime}$ (Zitrin et al. 2009a); Abell 1689 (${\theta }_{{\rm{E}}}\sim 45^{\prime\prime}$; Broadhurst et al. 2005); and the HFF cluster Abell 370 ($\sim 40^{\prime\prime}$; Richard et al. 2010). Also worth mentioning is RCS2 J232727.6−020437, for which Sharon et al. (2015) find ${\theta }_{{\rm{E}}}\simeq 40^{\prime\prime}$ for ${z}_{s}\simeq 3$ (but smaller—i.e., 26''—as expected, for another source at $z\simeq 1.4$).

The distribution of Einstein radii, and in particular the high end of this distribution, is important to characterize, as the largest and most massive lenses help to probe cosmological models and structure formation scenarios (Oguri & Blandford 2009; Waizmann et al. 2012; Redlich et al. 2014). The high concentrations and large Einstein radii found for several massive clusters have been previously claimed to challenge ΛCDM (Broadhurst & Barkana 2008; Broadhurst et al. 2008), although updated analyses (and accounts of projection biases) have alleviated this tension (Merten et al. 2015; Umetsu et al. 2016). In a similar fashion, the amount of substructure within massive clusters can also be compared to numerical simulations and expectations from ΛCDM (Jauzac et al. 2016; Schwinn et al. 2017). Given its extreme properties, PLCKG287 is another useful laboratory for similar studies.

Most of the impressive lenses known to date, including the ${\theta }_{{\rm{E}}}\gt 40^{\prime\prime}$ lenses mentioned above as well as those observed by the CLASH and HFF programs, and that were found to be magnifying many high-redshift lensed galaxies (e.g., Bradley et al. 2014; Kawamata et al. 2016 and references therein), were selected for HST follow-up mostly due to their optical (e.g., richness) or X-ray signals. X-ray selection in particular has proved to be an excellent probe for locating merging clusters with large critical curves (Ebeling et al. 2010; Zitrin et al. 2015), as mergers tend to boost the critical area. Here, we concentrate on an SZ-selected cluster, which has the second-highest M_SZ estimate in the Planck cluster catalog. In addition to our observations, PLCKG287 will be observed in the ongoing Reionization Lensing Cluster Survey (RELICS; PI: Coe), an HST program to observe a large set of mainly SZ-mass selected clusters, designed to find high-redshift, bright dropout galaxies in the reionization epoch.

Indeed, recent SZ surveys have expanded our massive-cluster sample. One such notable example is ACT-CL J0102-4915, the "El Gordo" cluster, a high-redshift (z = 0.87) and likely the most massive, hottest, most X-ray luminous and brightest SZ effect cluster known at $z\gt 0.6$ (Menanteau et al. 2012; Jee et al. 2014). Zitrin et al. (2013) performed the first SL analysis of this cluster, revealing a large lens with a critical area exceeding $\simeq 1.4$ arcmin² for high-redshift sources (${z}_{s}\gtrsim 4$). This shows the power of SZ massive-cluster selection, especially at higher redshifts, much due to the fact that the SZ signal is not redshift dependent (nonetheless, X-ray and optical/near-infrared observations have also revealed high-redshift clusters, including out to $z\gt 1$ and $z\gt 2$, respectively; Rosati et al. 2009; Strazzullo et al. 2016).

One interesting question is whether there is a preferred selection that leads to larger or "better" strong lenses, particularly for high-z applications. Scaling relations between X-ray, SZ, luminosity, richness, and lensing masses are well established, and characterized with increasing precision (e.g., Rozo et al. 2014). However, high total mass is not sufficient to guarantee a large critical area, which is dependent on the exact central projected mass density distribution (Redlich et al. 2012). In this relation, efforts have been made in recent years to detect clusters with large critical areas more directly, based on the luminosity distribution of red cluster member galaxies, and calibrated with well-studies lensing clusters (Wong et al. 2012; Zitrin et al. 2012). It will be interesting to compare these selections to other probes (richness, SZ, X-ray) once relevant HST data become available.

Among the multiply imaged galaxies we detect in PLCKG287, noteworthy is a likely high-redshift dropout, detected only in F110W. We identify four multiple images of this galaxy (Figure 4). An additional, demagnified fifth one is predicted next to the BCG core, and we list here one potential candidate (system 19; Table 1). The photometric redshifts for all four magnified images of the galaxy agree well, ${z}_{\mathrm{phot}}=6.9$ [6.13–8.43] (95% C.I.) and are corroborated by our lens model: the system symmetry follows that of system 7 at $z\simeq 3.4$, but at a larger radius (and requires a higher lensing distance ratio) so that the higher-redshift nature of this object is also supported geometrically from lensing.

Zoom In Zoom Out Reset image size

Due to its relative brightness (observed apparent magnitudes of ${J}_{{\rm{F}}110{\rm{W}}}=25.3\mbox{--}25.5\,\mathrm{AB}$ for the four images), this galaxy presents a promising case for the challenging spectroscopic follow-up of high-redshift sources. In addition to its brightness, the SNR of such observations can be significantly increased by observing the four magnified images simultaneously. Using the magnification factors implied by our model (μ between 2.4 and 4.8), the source intrinsic (i.e., demagnified) apparent magnitude is ${J}_{{\rm{F}}110{\rm{W}}}\sim 26.7\pm 0.4$ AB, and the intrinsic half-light radius is 0.4 ± 0.14 kpc, where the errors represent the range from the four images and their magnification values. In future iterations, the source's relative magnifications can also be used as additional constraints for the model.

Galaxy clusters constitute great cosmic telescopes. Extensive lensing surveys have taught us that essentially all massive clusters act as useful gravitational lenses, distorting, magnifying, and multiply imaging objects behind them. This has allowed us to construct dozens of mass maps and enabled the continuous discovery of hundreds of magnified, high-redshift galaxies. However, even with significant HST time devoted to recent lensing surveys, the data (and corresponding lensing models) currently available sample only a relatively small part of all massive clusters. There are expected to be a dozen to several dozen massive clusters in the sky with ${\theta }_{{\rm{E}}}\,\gt \,40^{\prime\prime}$ (Oguri & Blandford 2009; Zitrin et al. 2012), and finding the most massive lenses and the best cosmic telescopes among these is important for studying the cosmological model and structure formation and evolution, and for maximizing the chances to detect increasingly fainter objects at higher redshifts.

Here, we modeled the massive cluster PLCKG287 and identified 20 multiple-image families and candidates, including an apparently bright (${J}_{{\rm{F}}110{\rm{W}}}=25.3\mathrm{AB}$), quadruply imaged ${z}_{\mathrm{phot}}\sim 7$ galaxy. Our analysis reveals a large Einstein radius of $42\pm 4^{\prime\prime}$ for ${z}_{s}\simeq 2$, and $54\pm 5^{\prime\prime}$ for ${z}_{s}\sim 10$, adding to the short list of only a few similarly large lenses. Our results are based solely on photometric redshifts, and so a refinement of the model is warranted when more data—in particular, spectroscopic redshifts—become available (however, given the redshift range already spanned by the multiply imaged galaxies, we consider the Einstein radius robust). Interestingly, in contrast to most impressive lenses we are familiar with to date, that are X-ray or optically selected, and similar to the notable example of El Gordo, this cluster is SZ-selected. It will be intriguing to see in the near future if a certain selection (such as SZ, X-ray, luminosity, richness, or other, more direct lensing-based optical methods) is most efficient in locating the largest strong lenses and those best suited for studying the early universe.

We thank the anonymous reviewer of this work for very useful comments. This work is based on observations made with the NASA/ESA Hubble Space Telescope. Support for program ID 14165 (PI: Seitz) was provided by NASA from the Space Telescope Science Institute (STScI), which is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. S.S. thanks the DFG Transregio 33 "The Dark Universe" and the DFG cluster of excellence "Origin and Structure of the Universe" for support. Support for D.G. was provided by NASA through the Einstein Fellowship Program, grant PF5-160138. A.M. and M.N. acknowledge support by PRIN-INAF 2014 1.05.01.94.02. IRAF is distributed by the National Optical Astronomy Observatories, which are operated by the Association of Universities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. IRAF rvsao was developed at the Smithsonian Astrophysical Observatory Telescope Data Center.