RELICS: Reionization Lensing Cluster Survey

TL;DR

RELICS is a wide, shallow survey of 41 galaxy clusters using Hubble and Spitzer, aimed at discovering and studying high-redshift galaxies from the universe's first billion years, including the brightest and most distant known objects.

Contribution

This survey extends previous gravitational lensing studies by providing high-resolution infrared imaging of more clusters, enabling the discovery of the brightest and most distant high-redshift galaxy candidates.

Findings

Discovered 322 high-redshift galaxy candidates at z ~ 6-10.

Identified the brightest known z ~ 6 galaxy and the most distant spatially-resolved z ~ 10 arc.

Detected 11 supernovae, including 3 lensed supernovae.

Abstract

Large surveys of galaxy clusters with the Hubble and Spitzer Space Telescopes, including CLASH and the Frontier Fields, have demonstrated the power of strong gravitational lensing to efficiently deliver large samples of high-redshift galaxies. We extend this strategy through a wider, shallower survey named RELICS, the Reionization Lensing Cluster Survey. This survey, described here, was designed primarily to deliver the best and brightest high-redshift candidates from the first billion years after the Big Bang. RELICS observed 41 massive galaxy clusters with Hubble and Spitzer at 0.4-1.7um and 3.0-5.0um, respectively. We selected 21 clusters based on Planck PSZ2 mass estimates and the other 20 based on observed or inferred lensing strength. Our 188-orbit Hubble Treasury Program obtained the first high-resolution near-infrared images of these clusters to efficiently search for lensed high-redshift galaxies. We observed 46 WFC3/IR pointings (~200 arcmin^2) with two orbits divided among four filters (F105W, F125W, F140W, and F160W) and ACS imaging as needed to achieve single-orbit depth in each of three filters (F435W, F606W, and F814W). As previously reported by Salmon et al., we discovered 322 z ~ 6 - 10 candidates, including the brightest known at z ~ 6, and the most distant spatially-resolved lensed arc known at z ~ 10. Spitzer IRAC imaging (945 hours awarded, plus 100 archival) has crucially enabled us to distinguish z ~ 10 candidates from z ~ 2 interlopers. For each cluster, two HST observing epochs were staggered by about a month, enabling us to discover 11 supernovae, including 3 lensed supernovae, which we followed up with 20 orbits from our program. We delivered reduced HST images and catalogs of all clusters to the public via MAST and reduced Spitzer images via IRSA. We have also begun delivering lens models of all clusters, to be completed before the JWST GO call for proposals.

Tables

Table 1. Table 1: Recent Large HST  Cluster Lensing Surveys

	CLASH	Frontier Fields	RELICS
Clusters	25	6	41
HST orbits per cluster	20a	140	5a
Total HST orbits	524	840	188
Supernova orbitsb	50	0	20
HST Filters	16	7	7
Depth in F160Wc	27.5	28.7	26.5
HST Cycle Numbers	18 – 20	21 – 23	23
HST Begin	Nov.’10	Oct.’13	Oct.’15
HST End	Jul.’13	Sept.’16	Apr.’17

Table 2. Table 2: RELICS Clusters

		R.A.b	Decl.b		Planck Mass $M_{500}$
Index	Clustera	(J2000)	(J2000)	Redshift	Rankc	(${10}^{14}$$M_{\odot }$)	E(B-V)
1	Abell 2163 NE	16:15:48.3	$-$06:07:36.7	0.203	1	${16.12}_{-0.29}^{+0.30}$	0.2972
	Abell 2163 SW	16:15:42.6	$-$06:09:22.1
2	PLCK G287.0+32.9	11:50:50.8	$-$28:04:52.2	0.390d	2	${14.69}_{-0.42}^{+0.39}$	0.0813
3	MACSJ0417.5-1154	04:17:33.7	$-$11:54:22.6	0.443	4	${12.25}_{-0.55}^{+0.53}$	0.0320
4	Abell 697	08:42:58.9	$+$36:21:51.1	0.282	10	${11.00}_{-0.37}^{+0.37}$	0.0333
5	RXS J060313.4+4212 N	06:03:12.2	$+$42:15:24.7	0.228	11	${10.76}_{-0.43}^{+0.45}$	0.1933
	RXS J060313.4+4212 S	06:03:25.6	+42:09:53.6
6	MACS J0308.9+2645	03:08:55.7	$+$26:45:36.8	0.356	12	${10.76}_{-0.65}^{+0.63}$	0.1776
7	ACT-CL J0102-49151 NW	01:02:53.1	$-$49:14:52.8	0.870	13	${10.75}_{-0.47}^{+0.48}$	0.0086
	ACT-CL J0102-49151 SE	01:03:00.0	$-$49:16:22.2
8	RXC J0600.1-2007	06:00:09.8	$-$20:08:08.9	0.460	14	${10.73}_{-0.54}^{+0.51}$	0.0433
9	PSZ2 G209.79+10.23	07:22:23.0	$+$07:24:30.0	0.677	15	${10.73}_{-0.66}^{+0.63}$	0.0375
10	PLCK G171.9-40.7	03:12:56.9	$+$08:22:19.2	0.270	16	${10.71}_{-0.50}^{+0.49}$	0.4477
11	RXC J2211.7-0350	22:11:45.9	$-$03:49:44.7	0.397	17	${10.50}_{-0.49}^{+0.50}$	0.0832
12	PLCK G004.5-19.5	19:17:04.5	$-$33:31:28.5	0.540	19	${10.36}_{-0.72}^{+0.68}$	0.0790
13	PLCK G308.3-20.2	15:18:49.9	$-$81:30:33.6	0.480	20	${10.32}_{-0.58}^{+0.57}$	0.2348
14	RXC J0018.5+1626	00:18:32.6	$+$16:26:08.4	0.546	24	${9.79}_{-0.53}^{+0.53}$	0.0501
15	SPT-CL J0254-5857	02:54:16.0	$-$58:57:11.0	0.438	26	${9.69}_{-0.38}^{+0.37}$	0.0183
16	PSZ2 G138.61-10.84	02:27:06.6	$+$49:00:29.9	0.702	27	${9.48}_{-0.53}^{+0.67}$	0.1830
17	RXC J0142.9+4438	01:42:55.2	$+$44:38:04.3	0.341	28	${9.02}_{-0.64}^{+0.60}$	0.0783
18	Abell 1300	11:31:54.1	$-$19:55:23.4	0.308	30	${8.97}_{-0.45}^{+0.46}$	0.0440
19	WHL J013719.8-082841	01:37:25.0	$-$08:27:25.0	0.566	31	${8.93}_{-0.70}^{+0.65}$	0.0286
20	RXC J1514.9-1523	15:15:00.7	$-$15:22:46.7	0.223	33	${8.86}_{-0.46}^{+0.41}$	0.0869
21	Abell 665	08:30:57.4	$+$65:50:31.0	0.182	34	${8.86}_{-0.32}^{+0.32}$	0.0400
22	MACS J0553.4-3342	05:53:23.1	$-$33:42:29.9	0.430	36	${8.77}_{-0.46}^{+0.44}$	0.0357
23	SMACS J0723.3-7327	07:23:19.5	$-$73:27:15.6	0.390	43	${8.39}_{-0.34}^{+0.33}$	0.1893
24	RXC J0949.8+1707	09:49:50.9	$+$17:07:15.3	0.383	48	${8.24}_{-0.46}^{+0.46}$	0.0255
25	Abell 1758a NW	13:32:39.0	$+$50:33:41.8	0.280	50	${8.22}_{-0.28}^{+0.27}$	0.0122
	Abell 1758a SE	13:32:53.4	+50:31:31.0
26	Abell 1763	13:35:18.9	$+$40:59:57.2	0.228	51	${8.13}_{-0.27}^{+0.26}$	0.0073
27	Abell 2813	00:43:25.1	$-$20:37:14.8	0.292	52	${8.13}_{-0.38}^{+0.37}$	0.0178
28	Abell 520 NE	04:54:19.0	$+$02:56:49.0	0.203	65	${7.80}_{-0.41}^{+0.40}$	0.0402
	Abell 520 SW	04:54:04.2	+02:53:41.9
29	RXC J0032.1+1808	00:32:11.0	$+$18:07:49.0	0.396	85	${7.61}_{-0.63}^{+0.57}$	0.1052
30	RXC J0232.2-4420	02:32:18.1	$-$44:20:44.9	0.284	91	${7.54}_{-0.32}^{+0.33}$	0.0165
31	Abell 3192e	03:58:53.1	$-$29:55:44.8	0.425	114	${7.20}_{-0.50}^{+0.52}$	0.0071
32	MACS J0159.8-0849	01:59:49.4	$-$08:50:00.0	0.405	115	${7.20}_{-0.68}^{+0.61}$	0.0207
33	MACS J0035.4-2015	00:35:27.0	$-$20:15:40.3	0.352	133	${7.01}_{-0.50}^{+0.45}$	0.0187
34	RXC J0911.1+1746	09:11:11.4	$+$17:46:33.5	0.505	136	${6.99}_{-0.79}^{+0.73}$	0.0359
35	Abell S295	02:45:31.4	$-$53:02:24.9	0.300	156	${6.78}_{-0.36}^{+0.37}$	0.0445
36	SPT-CL J0615-5746	06:15:54.2	$-$57:46:57.9	0.972	157	${6.77}_{-0.54}^{+0.49}$	0.0362
37	MACS J0257.1-2325	02:57:10.2	$-$23:26:11.8	0.505	227	${6.22}_{-0.74}^{+0.70}$	0.0251
38	Abell 2537	23:08:22.2	$-$02:11:32.4	0.297	376	${5.52}_{-0.51}^{+0.51}$	0.0798
39	MS 1008.1-1224	10:10:33.6	$-$12:39:43.0	0.306	504	${4.94}_{-0.60}^{+0.57}$	0.0601
40	MACS J0025.4-1222	00:25:30.3	$-$12:22:48.1	0.586	$\cdots$	$\cdots$	0.0239
41	CL J0152.7-1357	01:52:42.9	$-$13:57:31.0	0.833	$\cdots$	$\cdots$	0.0126

Table 3. Table 3: RELICS Cluster Notes

Index	Cluster	Notes
1	Abell 2163	Most massive cluster according to both Planck PSZ2 and the X-ray analysis by Mantz et al. (2010) (also see Arnaud et al. 1992)
2	PLCK G287.0+32.9	Not observed by Hubble or Spitzer prior to RELICS. Lensed arcs 80″ from the BCG (Gruen et al., 2014, Figure 18). Discovered in early SZ catalog (Planck Collaboration et al., 2011a)
3	MACSJ0417.5-1154	One of the most X-ray luminous clusters discovered by MACS: 2.9$\times {10}^{45}$ erg s-1 between 0.1–2.4 keV (Ebeling et al., 2010; Pandge et al., 2019); JWST GTO target (CANUCS; PI Willott)
4	Abell 697	Highest SZ mass Abell cluster yet to be observed by ACS or WFC3/IR (WFPC2 only); high mass in X-ray analyses (Mantz et al., 2010; Piffaretti et al., 2011); more modest in weak lensing (Mahdavi et al., 2013)
5	RXS J060313.4+4212 = “Toothbrush”	Radio relic merger shock wave discovered by van Weeren et al. (2012) and studied further (Rajpurohit et al., 2018)
6	MACS J0308.9+2645	MACS half-orbit F606W, F814W
7	ACT-CL J0102-49151 = “El Gordo”	Discovered by Menanteau et al. (2012); excellent lens (Zitrin et al., 2013); weak lensing $M_{200}=3$$\times {10}^{15}$$M_{\odot }$ (Jee et al., 2014); ACS $2\times 2$ mosaic F606W; JWST GTO target (PI Windhorst)
8	RXC J0600.1-2007	MACS half-orbit F814W
9	PSZ2 G209.79+10.23	no HST, Chandra, XMM
10	PLCK G171.9-40.7	no HST
11	RXC J2211.7-0350	MACS half-orbit F606W
12	PLCK G004.5-19.5	$z=1.601$ arc at 30″ (Sifón et al., 2014)
13	PLCK G308.3-20.2 = SMACSJ1519.1-8130	MACS half-orbit F606W
14	RXC J0018.5+1626 = MS0015.9+1609	Known strong lens (Zitrin et al., 2011); ACS $2\times 2$ mosaic; CCCP $M_{2500}=6.2\times {10}^{14}{M}_{\odot }$ 4th highest measured of 50 clusters (Hoekstra et al., 2015); shallow off target NICMOS
15	SPT-CL J0254-5857	One of the most massive clusters discovered by SPT (Williamson et al., 2011)
16	PSZ2 G138.61-10.84	Optical counterpart is 4.5′ from PSZ1 coordinates (2:27:06, +49:05:10) from Planck Collaboration et al. (2016e)
17	RXC J0142.9+4438	CIZA (Kocevski et al., 2007)
18	Abell 1300	“rivals A2744” Mann & Ebeling (2012); Ebeling WFPC2 F606W
19	WHL J013719.8-082841 = WHL J24.3324-8.477	Discovered by Wen et al. (2012) based on photometric redshifts in SDSS-III. Renamed in the PSZ2 catalog with coordinates given in decimal degrees.
20	RXC J1514.9-1523	radio halo (Giacintucci et al., 2011)
21	Abell 665	similar in mass and redshift to A1689 (excellent lens); ACS F850LP + grism
22	MACS J0553.4-3342	Bullet-like merger (Mann & Ebeling, 2012); very hot and luminous merger (Ebeling et al., 2017; Pandge et al., 2017); big red triple arcs; bright star but probably okay
23	SMACS J0723.3-7327	MACS half-orbit F606W, F814W; 2 cycles
24	RXC J0949.8+1707	Wong et al. (2013) #7: 7th most powerful lensing line of sight expected based on LRGs identified in SDSS
25	Abell 1758a	northern component of merger; WtG $M_{2500}=6.1\times {10}^{14}{M}_{\odot }$ (Applegate et al. 2014; as compiled by Sereno 2015)
26	Abell 1763	Rines et al. (2013) highest mass; Mantz et al. (2010) #15; WFPC2 good cluster
27	Abell 2813	arcs close in; half-orbit F606W (PI Smith)
28	Abell 520 = “Train Wreck”	ACS $2\times 2$ mosaic (PI Clowe); merger like Bullet Cluster but with claimed (and disputed) dark core (Mahdavi et al., 2007; Okabe & Umetsu, 2008; Clowe et al., 2012; Jee et al., 2014)
29	RXC J0032.1+1808	merger like MACS0717 or MACS0416
30	RXC J0232.2-4420 = “RBS-0325”	arcs to large radius identified in ground-based imaging (Kausch et al., 2010)
31	Abell 3192 & MACSJ0358.8-2955	HST observations centered on MACS0358 ($z=0.428$); includes some contribution from A3192 ($z=0.168$); Mantz et al. (2010); Hamilton-Morris et al. (2012); Hsu et al. (2013)
32	MACS J0159.8-0849	arcs to large radius; not many cluster galaxies; Wong et al. (2013) #59
33	MACS J0035.4-2015	arcs to large radius
34	RXC J0911.1+1746	MACS
35	Abell S295 = ACT-CL J0245-5302	Known strong lens (Zitrin et al., 2011); ACT SZ detection (Menanteau et al., 2010); Ruel et al. (2014); Bayliss et al. (2016); ACS observations proprietary before inclusion in RELICS
36	SPT-CL J0615-5746	our highest redshift cluster at $z=0.972$; ACS $2\times 2$ mosaic in F606W and F814W
37	MACS J0257.1-2325	Known strong lens (Zitrin et al., 2011); Frontier Fields contender; F555W, F814W; WtG $M_{2500}=5.2\times {10}^{14}{M}_{\odot }$ (Applegate et al. 2014; as compiled by Sereno 2015)
38	Abell 2537	Frontier Fields contender; nice arcs in F606W; CCCP $M_{2500}=4.6\times {10}^{14}{M}_{\odot }$ (Hoekstra et al., 2015)
39	MS 1008.1-1224	weak lensing and X-ray analyses (Lombardi et al., 2000; Athreya et al., 2002; Ettori & Lombardi, 2003); ACS $2\times 2$ mosaic
40	MACS J0025.4-1222 = “Baby Bullet”	well studied merger similar to the Bullet Cluster (Bradač et al., 2008; Zitrin et al., 2011)
41	CL J0152.7-1357	z=0.8330 still merging at filaments; well studied (Jee et al., 2005; Demarco et al., 2005; Umetsu et al., 2005; Maughan et al., 2006; Murata et al., 2015)

Table 4. Table 4: Archival HST  ACS and WFC3/UVIS imaging of RELICS clusters

Cluster	Abbreviated	F390W	F435W	F475W	F555W	F606W	F625W	F775W	F814W	F850LP
Index	Namea	(s)	(s)	(s)	(s)	(s)	(s)	(s)	(s)	(s)
1	abell2163		4664[2]			4667[2]			9192[2]
2	plckg287+32			2160		2320			4680
3	macs0417-11					7152∗			1910
4	abell697
5	rxs0603+42	5294∗[2]				5068∗[2]			10194[2]
6	macs0308+26					1200			1440
7	act0102-49					7680[4]	4688[2]	5024[2]	1916	5032[2]
8	rxc0600-20								1440
9	plckg209+10
10	plckg171-40
11	rxc2211-03					1200
12	plckg004-19
13	plckg308-20					1200
14	rxc0018+16				4500	17920[4]		4623	4560	2540
15	spt0254-58
16	plckg138-10
17	rxc0142+44
18	abell1300
19	whl0137-08
20	rxc1514-15
21	abell665						2680
22	macs0553-33		4452			2092			4572
23	smacs0723-73					1200			1440
24	rxc0949+17					1200			1440
25	abell1758		5072[2]			5088[2]			10000[2]
26	abell1763
27	abell2813					1200
28	abell520		9296[4]			9328[4]			18320[4]
29	rxc0032+18					1200			1440
30	rxc0232-44
31	abell3192		4500			3320			4620
32	macs0159-08					1200
33	macs0035-20					1200			1440
34	rxc0911+17				4470				8825
35	abells295		1920			1936			3944
36	spt0615-57					8880[4]			12720[5]
37	macs0257-23				4500				8858
38	abell2537					2080
39	ms1008-12					17856[4]		2440		2560
40	macs0025-12				4140				4200
41	cl0152-13						19000[2]	23452[2]		19000[2]

Table 5. Table 5: RELICS HST  Imaging with ACS and WFC3

		Cluster				Parallel
Cluster	Abbreviated	F435W	F606W	F814W	WFC3/IR	WFC3
Index	Name	(orbits)	(orbits)	(orbits)	(orbits)	(orbits)
1	abell2163				4[2]
2	plckg287+32*	1	*	*	2
3	macs0417-11	1			2
4	abell697	1	1	1	2	3
5	rxs0603+42				4[2]
6	macs0308+26	1	0.5	0.5	2
7	act0102-49	1			4[2]
8	rxc0600-20	1	1	1	2	3
9	plckg209+10	1	1	1	2	3
10	plckg171-40	1	1	1	2	3
11	rxc2211-03	1	1	1	2	3
12	plckg004-19	1	1	1	2	3
13	plckg308-20	1	1	1	2	3
14	rxc0018+16	1			2
15	spt0254-58	1	1	1	2	3
16	plckg138-10	1	1	1	2	3
17	rxc0142+44	1	1	1	2	3
18	abell1300	1	1	1	2	3
19	whl0137-08	1	1	1	2	3
20	rxc1514-15	1	1	1	2	3
21	abell665	1	1	1	2	3
22	macs0553-33				2
23	smacs0723-73	1	0.5	0.5	2+6
24	rxc0949+17	1	0.5	0.5	2
25	abell1758				4[2]
26	abell1763	1	1	1	2+7	3
27	abell2813	1	1	1	2	3
28	abell520				4[2]
29	rxc0032+18	1	0.5	0.5	2
30	rxc0232-44	1	1	1	2	3
31	abell3192				2
32	macs0159-08	1	1	1	2	3
33	macs0035-20	1	0.5	0.5	2
34	rxc0911+17	1			2
35	abells295				2
36	spt0615-57	1			2
37	macs0257-23	1			2
38	abell2537	1		1	2
39	ms1008-12	1			2
40	macs0025-12	1			2
41	cl0152-13	1			2+5

Table 6. Table 6: HST WFC3/IR and UVIS Follow-Up Imaging of RELICS Supernovae

Cluster	Supernova	Total Orbits	F105W	F125W	F140W	F160W	F350LP
abell1763	Nebra	7	0.3	2.7		3.7	0.3
clj0152-13	Nimrud	5		2.5		2.5
smacs0723-73	Yupana	6	2	1	2	1

Table 7. Table 7: Nominal Exposure Times and Expected Depths

		$\lambda$a	Exp.b	Depthc	Sens.d
Camera	Filter	($\mu$m)	Time	(AB)	(nJy)
HST ACS/WFC	F435W	0.43	1952 s	27.2	9
HST ACS/WFC	F606W	0.59	1991 s	27.6	7
HST ACS/WFC	F814W	0.81	2123 s	27.1	11
HST WFC3/IR	F105W	1.06	1361 s	26.6	16
HST WFC3/IR	F125W	1.25	686 s	26.0	29
HST WFC3/IR	F140W	1.39	686 s	26.2	25
HST WFC3/IR	F160W	1.54	1861 s	26.5	19
SST IRAC	ch1	3.6	5 hours	25.0	76
SST IRAC	ch2	4.5	5 hours	24.6	108

Table 8. Table 8: Nominal HST Exposures for a Cluster and Parallel Field with no Archival Imaging

Epoch /	Dithera	Prime	Exposure	Parallel	Exposure
Orbit	Position	Filter	time (s)	Filter	time (s)
1 / 1	A	F435W	370	F350LP	454
1 / 1	B	F435W	667	F350LP	701
1 / 1	C	F435W	667	F350LP	701
1 / 1	D	F435W	371	F350LP	496
1 / 2	A	F814W	516	F140W	603
1 / 2	B	F814W	607	F105W	703
1 / 2	C	F814W	607	F160W	703
1 / 2	D	F814W	516	F125W	603
1 / 3	W	F140W	178
1 / 3	W	F105W	353
1 / 3	Y	F105W	353
1 / 3	Y	F140W	178
1 / 3	Z	F125W	178
1 / 3	Z	F160W	503
1 / 3	X	F160W	503
1 / 3	X	F125W	153
2 / 4	A	F606W	516	F125W	503
2 / 4	B	F606W	607	F160W	703
2 / 4	C	F606W	607	F105W	703
2 / 4	D	F606W	516	F140W	503
2 / 5	Z	F140W	178
2 / 5	Z	F105W	353
2 / 5	X	F105W	353
2 / 5	X	F140W	203
2 / 5	W	F125W	178
2 / 5	W	F160W	503
2 / 5	Y	F160W	453
2 / 5	Y	F125W	230

Table 9. Table 9: Spitzer IRAC Imaging Programs

Program	PI	TAC	hoursa	depthb	clustersc
12005	Bradač	GO	99.9	1.2	26
12123	Soifer	DDT	290	5	34
13165	Bradač	DDT	167	30	3
13210	Bradač	DDT	55.5	30	1
14017	Bradač	GO	333.2	30	6

Table 10. Table 10: Large Surveys Including RELICS Clusters Motivated in part by RELICS

Name	PI	Observatory	Instrument	Time	Wavelength	RELICS / Total Clusters
RELICSa	Coe	Hubble	ACS, WFC3	115 hours	0.4 – 1.7 $\mu$m	41
S-RELICSb	Bradač; Soifer	Spitzer	IRAC	945 hours	3 – 5 $\mu$m	34f
GAME	Mercurio	VST	OmegaCAM	300 hours	ugri	9 / 12
GCAVc	Nonino	VISTA	VIRCAM	540 hours	YJKs	13 / 20
Witnessing…d	Arnaud & Ettori	XMM-Newton	EPIC	833 hours	0.15 – 15 keV	18 / 118
ALCS	Kohno	ALMA	12-m array	95 hours	1.1 mm	16 / 33
Probing…e	van Weeren	VLA	25-m array	85 hours	2 – 4 GHz	34g

Table 11. Table 11: SExtractor parameters used in the RELICS Hubble  source catalogs

SExtractor parameter	acs-wfc3	wfc3ir	Description
DETECT_MINAREA	9	9	Contiguous pixels required above detection threshold
DETECT_THRESH	1	1	Detection threshold ($\sigma$ above background RMS)
BACK_SIZE	128	64	Background cell size
BACK_FILTERSIZE	5	3	Background grid size
DEBLEND_NTHRESH	32	64	Number of threshold levels
DEBLEND_MINCONT	0.0015	0.0001	Minimum contrast ratio
BACKPHOTO_TYPE	LOCAL	LOCAL	Method for measuring background
BACKPHOTO_THICK	24	24	Width of rectangular annulus around each object

Table 12. Table 12: Galactic Reddening Extinction Correction Coefficients A λ subscript 𝐴 𝜆 A_{\lambda} for Each Filter

Camera	Filter	Coefficient
WFC3/UVIS	F390W	4.514
ACS	F435W	4.117
ACS	F475W	3.747
ACS	F555W	3.242
ACS, WFC3/UVIS	F606W	2.929
ACS	F625W	2.671
ACS	F775W	2.018
ACS	F814W	1.847
ACS	F850LP	1.473
WFC3/IR	F105W	1.015
WFC3/IR	F110W	0.876
WFC3/IR	F125W	0.757
WFC3/IR	F140W	0.609
WFC3/IR	F160W	0.470

Table 13. Table 13: HST  Source Catalog Content

column	parameter	description
1	id	Object ID number
2	RA	Right Ascension in decimal degrees (J2000)
3	Dec	Declination in decimal degrees (J2000)
4	x	x pixel coordinate
5	y	y pixel coordinate
6	fwhm	Full width at half maximum (arcsec)
7	area	Isophotal aperture area (pixels)
8	stel	SExtractor “stellarity” (1 = star; 0 = galaxy)
9	ell	Ellipticity = $1-B/A$
10	theta	Position angle (CCW wrt x axis; degrees)
11	nf5sig	Number of filters with a 5-sigma detection
12	nfobs	Number of filters observed for this object (in the field of view and without bad pixels)
13	f435w_mag	F435W isophotal magnitude (99 = non-detection; –99 = unobserved)
14	f435w_magerr	F435W isophotal magnitude uncertainty (or 1-sigma upper limit for non-detection)
15	f435w_flux	F435W isophotal flux (e–/s)
16	f435w_fluxerr	F435W isophotal flux uncertainty (e–/s)
17	f435w_fluxnJy	F435W isophotal flux (nJy)
18	f435w_fluxnJyerr	F435W isophotal flux uncertainty (nJy)
19	f435w_sig	F435W detection significance
$\cdots$	$\cdots$	(photometry in other filters)
62a	bright_mag	Brightest magnitude in any filter
63	bright_magerr	Brightest magnitude uncertainty
64	zb	BPZ most likely Bayesian photometric redshift
65	zbmin	BPZ lower limit (95% confidence)
66	zbmax	BPZ upper limit (95% confidence)
67	tb	BPZ most likely spectral type (1–5 elliptical; 6–7 spiral; 8–11 starburst)
68	odds	P(z) contained within zb $\pm 0.04\left(1+z\right)$
69	chisq	${\chi }^2$ poorness of BPZ fit: observed vs. model fluxes
70	chisq2	Modified ${\chi }^2$: model fluxes allowed uncertainties (Coe et al., 2006)
71	M0	Magnitude used as BPZ prior: F775W or closest available filter
72	zml	Maximum Likelihood (flat prior) most likely redshift
73	tml	Maximum Likelihood (flat prior) most likely spectral type

Table 14. Table 14: HST  Source Catalog: Detection and Shape Measurement

	${\alpha }_{J2000}$	${\delta }_{J2000}$	$x$	$y$	FWHM	area		ellipticity	$\theta$
ID	(deg)	(deg)	(pixels)	(pixels)	(′′)	(pixels)	stellarity	($1-b/a$)	(deg)	$N_{f,5\sigma }$	$N_{f,obs}$
1	130.7320691	36.3412393	3212.886	1118.780	0.356	13	0.35	0.26	11.0	1	7
2	130.7567476	36.3856742	2020.442	3784.849	0.320	474	0.03	0.09	–42.8	7	7
3	130.7567488	36.3851756	2020.382	3754.932	0.754	125	0.00	0.33	58.4	7	7
4	130.7545731	36.3848770	2125.475	3737.004	0.173	42	0.98	0.06	–10.0	5	7
5	130.7535624	36.3842826	2174.292	3701.339	0.378	59	0.01	0.21	80.2	7	7

Table 15. Table 15: HST  Source Catalog: Photometry in Each Filter

Magnitude				Flux			Flux
(AB)				($e^{-}$/s)			(nJy)	S/N
28\@alignment@align.4629	0.4579  +/–	0\@alignment@align.0760	0.0399	14\@alignment@align.9573	7.8473  +/–	1\@alignment@align.9100
24\@alignment@align.2190	0.0567  +/–	3\@alignment@align.7868	0.2031	745\@alignment@align.4248	39.9764  +/–	18\@alignment@align.6500
25\@alignment@align.5322	0.1058  +/–	1\@alignment@align.1298	0.1156	222\@alignment@align.3921	22.7527  +/–	9\@alignment@align.7700
99\@alignment@align.0000	28.6848  +/–	-0\@alignment@align.1387	0.0619	-27\@alignment@align.3093	12.1918  +/–	-2\@alignment@align.2400
25\@alignment@align.6583	0.0796  +/–	1\@alignment@align.0059	0.0765	198\@alignment@align.0041	15.0577  +/–	13\@alignment@align.1500

Table 16. Table 16: HST  Source Catalog: Photometric Redshifts

$z_{BPZ}$	$z_{min}$	$z_{max}$	$t_{BPZ}$	ODDS	${\chi }^2$	${\chi }_{mod}^2$	$M_0$	$z_{ML}$	$t_{ML}$
2.741	0.225	3.470	9.5	0.148	0.940	1.334	28.102	2.730	9.5
0.222	0.145	0.255	4.1	0.850	0.933	0.158	22.015	0.190	4.2
1.033	0.631	1.115	9.6	0.481	0.163	0.212	25.013	1.040	9.6
4.501	4.315	4.698	8.2	0.974	1.292	1.367	25.835	4.520	8.2
0.778	0.461	0.961	10.3	0.473	0.112	0.770	25.346	0.780	10.2

Table 17. Table 17: RELICS Supernovae and HST Follow-Up Imaging

Cluster	Supernovaa	Abbreviationb	R.A. (J2000)	Decl. (J2000)	Notes
rxc0949+17	Eleanorc	RLC11Ele	09:49:47.97	+17:07:24.9	cluster member
rxc0949+17	Alexanderc	RLC11Ale	09:49:48.07	+17:07:24.0	cluster member
rxc0949+17	Antikythera	RLC15Ant	09:49:48.01	+17:07:23.0	cluster member
rxc0142+44	Makapansgat	RLC16Mak	01:43:16.326	+44:33:50.65	parallel field
abell1763	Nebra	RLC16Neb	13:35:15.13	+41:00:15.8	lensed
macs0025-12	Quipu	RLC16Qui	00:25:31.977	–12:23:31.80	cluster member
macs0257-23	Cheomseongdae	RLC16Che	02:57:07.795	–23:27:11.69	lensed or cluster member
plckg171-40	Kukulkan	RLC16Kuk	03:12:59.148	+08:22:43.60	cluster member
clj0152-13	Nimrud	RLC16Nim	01:52:40.352	–13:57:44.81	lensed
rxc0600-20	William	RLC17Wil	06:00:12.227	–20:07:23.91	cluster member
smacs0723-73	Yupana	RLC17Yup	07:23:28.40	–73:27:03.6	lensed or cluster member

Full text

early universe – galaxies: high redshift – cosmology: dark ages, reionization, first stars – galaxies: evolution – gravitational lensing: strong – galaxies: clusters

††journal: ApJ

1 Introduction

Gravitational lensing magnification by massive galaxy clusters has a long history of helping astronomers discover the most distant galaxies known with the Hubble and Spitzer Space Telescopes (see Kneib & Natarajan 2011 §5.8 for a review). Twenty years ago, a $z=4.92$ galaxy lensed by the cluster MS1358+62 was the most distant known (Franx et al., 1997). Ten years ago, that record belonged to A1689-zD1 at $z\sim 7.5$ (Bradley et al., 2008; Watson et al., 2015). More recently, the Cluster Lensing And Supernova survey with Hubble (CLASH; Postman et al. 2012) yielded the triply-imaged $z\sim 10.8$ candidate MACS0647-JD (Coe et al., 2013; Pirzkal et al., 2015; Chan et al., 2016). This redshift has been surpassed only by the grism measurement of $z=11.1$ (Oesch et al., 2016) for GN-z11 discovered by Oesch et al. (2014) in blank-field imaging by the CANDELS program (Grogin et al., 2011; Koekemoer et al., 2011). Both $z\sim 11$ galaxies are similarly bright at 1.6$\mu$m (F160W AB mag 25.9), making them excellent targets for follow-up study with the James Webb Space Telescope (JWST).

For a given observing strategy, lensed fields generally yield significantly more high-redshift galaxies than blank fields, especially at the highest redshifts and in relatively shallow imaging (e.g., Coe et al., 2015). Lensing does sacrifice high-redshift search volume due to the magnification as well as the foreground cluster light. But lensing more than compensates for the lost volume by magnifying many more faint galaxies into view whenever luminosity function number counts are steeper than $\phi\propto L^{-2}$ (Broadhurst et al., 1995). This is especially true brightward of $L^{*}$, for example, AB mag $\sim$28 at $z\sim 8$ (Bradley et al., 2012; Finkelstein, 2016).

CLASH demonstrated this lensing advantage, yielding significantly more galaxies at $z\sim 6-8$ than blank field surveys (Bradley et al., 2014). CLASH obtained 20-orbit HST imaging for each of 25 clusters in 16 filters, including F160W observed to AB mag 27.5 (5$\sigma$ depth; Postman et al. 2012; Molino et al. 2017). The first 18 clusters yielded 262 candidates at $z\sim 6-8$ (Bradley et al., 2014), plus a few at $z\sim 9-11$ (Zheng et al., 2012; Coe et al., 2013; Bouwens et al., 2014).

The Frontier Fields (Lotz et al., 2016) obtained deeper 140-orbit HST imaging for each of 6 clusters and 6 blank parallel fields in 7 filters, including F160W observed to AB mag 28.7. At these depths, $\phi\propto L^{-2}$ for high-redshift galaxies (the luminosity function faint end slope $\alpha\sim-2$ roughly at $z\sim 6-8$), so the lensed and blank field high-z counts were predicted to be similar (Coe et al., 2015). Indeed, the Frontier Fields yielded similar total numbers of lensed and blank field high-redshift galaxies, altogether 453 at $z\sim 6-9$ from one analysis (Kawamata et al., 2018), plus a few $z\sim 10$ candidates (Zitrin et al., 2014; Infante et al., 2015; McLeod et al., 2016; Oesch et al., 2018). Due to their magnifications, the lensed galaxies are intrinsically fainter than the blank field galaxies. By combining deep Hubble imaging with the power of gravitational lensing, the Frontier Fields revealed the faintest galaxies yet known, including some of those likely responsible for reionization at $z\sim 6$ (Livermore et al., 2016; Atek et al., 2018).

While CLASH and the Frontier Fields yielded many high-redshift candidates, neither survey was optimized to deliver high-redshift candidates observed brightly enough for detailed follow-up study with current and future observatories. Detailed studies are required to determine galaxies’ ages, stellar masses, compositions, and ionizing strengths (see Stark 2016 for a review). It is imperative to use existing facilities to discover the best targets for study in the early universe, ideally before JWST Cycle 1. Gravitational lensing offers the most efficient route to do this. The Planck all-sky survey delivered the PSZ2 catalog of $>$1,000 massive galaxy clusters (Planck Collaboration et al., 2016a) detected via their Sunyaev-Zel’dovich (SZ) effect (Sunyaev & Zeldovich, 1970) on the Cosmic Microwave Background (CMB). By searching the PSZ2 catalog, we found that many massive clusters (including presumably excellent lenses) lacked the combination of HST Advanced Camera for Surveys (ACS) and Wide Field Camera 3 infrared (WFC3/IR) imaging required to discover high-redshift candidates at $z>6$.

RELICS embarked on an efficient survey with Hubble and Spitzer to discover the best and brightest high-redshift galaxies (Coe, 2018). RELICS obtained 5-orbit HST imaging of 41 clusters in 7 filters (the same filters used by the Frontier Fields), including F160W to AB mag 26.5. Our relatively shallow imaging covered more area than CLASH or the Frontier Fields, yielding high-redshift candidates that are brighter, either intrinsically and/or due to lensing magnification. See Table 1 for a summary comparison of CLASH, the Frontier Fields, and RELICS. More recently, a new HST program BUFFALO (PIs Steinhardt & Jauzac; GO 15117) has begun extending the Frontier Fields to the wider area covered by deep Spitzer imaging.

Also notable and inspiring for this project are the HST Snapshot programs observing galaxy clusters discovered by the MAssive Cluster Survey (MACS; Ebeling et al. 2001). Shallower HST imaging ($\sim 1$ orbit per cluster) has been obtained for 86 clusters to date (Repp & Ebeling, 2018). For 29 of these clusters, 4-band imaging, including F140W to AB mag 26.6, was completed, yielding 20 candidates at $z\sim 7-9$ (Repp et al., 2016). Nine MACS clusters are included in RELICS. Prior to RELICS, they had archival HST ACS and/or WFPC2 imaging, but not WFC3/IR or NICMOS.

Meanwhile, large Spitzer programs such as SURFS UP (Bradač et al., 2014; Ryan et al., 2014; Huang et al., 2016) and the Frontier Fields have also surveyed many galaxy clusters, helping to identify high-redshift galaxies and study their properties (e.g., Hoag et al., 2019). The Grism Lens-Amplified Survey from Space (GLASS; Treu et al. 2015) delivered spectroscopic properties and high-redshift confirmations (e.g., Schmidt et al., 2016, 2017). The high-redshift searches in lensing programs complement searches in blank field surveys such as the UDF, CANDELS, and BoRG, constraining luminosity functions from $z\sim 4-10$ (Bouwens et al., 2015; Finkelstein, 2016; Oesch et al., 2018; Morishita et al., 2018).

Below we discuss the RELICS science drivers (§2), galaxy cluster targets (§3), observations with Hubble, Spitzer, and other observatories (§4), the image reductions and catalogs (§5), and results to date (§6), followed by a summary (§7).

We use the AB magnitude system, $m_{\rm AB}=31.4-2.5\log(f_{\nu}/{\rm nJy})$ (Oke, 1974; Oke & Gunn, 1983). Where needed, we assume a flat concordance $\Lambda$CDM cosmology with $h=0.7$, $\Omega_{m}=0.3$, and $\Omega_{\Lambda}=0.7$, where $H_{0}=100\,h$ km s*-1* Mpc*-1*. Galaxy cluster masses are given as $M_{500}$, the mass enclosed within $r_{500}$, within which the average density is 500 times the critical density of the universe at that redshift (e.g., Coe, 2010). These masses $M_{500}$ are less than the total virial cluster masses measured at larger radii.

2 Science

RELICS was primarily designed and optimized to search for brightly lensed high-redshift galaxies in the epoch of reionization. Ancillary science enabled by RELICS includes supernova searches, cluster mass measurements, and limits on the dark matter particle cross section. We discuss each of these science goals in turn.

2.1 High-Redshift Galaxies

Hubble’s WFC3 has revealed $\sim$2,000 $z\sim 6-11$ candidates from the Universe’s first billion years, but only a small fraction have been bright enough ($H<25.5$) for detailed follow-up study and spectroscopic confirmation (e.g., Bouwens et al., 2015; Salmon et al., 2017). RELICS was designed primarily to deliver

1. brightly observed high-redshift candidates, amenable to more detailed studies; and
2. a large sample of high-redshift candidates to improve luminosity function constraints. This combination is required to improve our understanding of galaxies and reionization in the first billion years.

2.1.1 Spectroscopic Studies of Brightly Observed High-Redshift Candidates

Spectroscopic redshift confirmations based on Lyman-$\alpha$ detections have proven increasingly difficult beyond $z>6$ (Stark et al., 2010; Bradač et al., 2012; Schenker et al., 2014; Schmidt et al., 2016; Hoag et al., 2019; Mason et al., 2019), likely due primarily to absorption by patchy neutral hydrogen before reionization was complete (Treu et al., 2013; Tilvi et al., 2014; Mason et al., 2018). A higher success rate was achieved with four luminous $z\sim 7-9$ galaxies inferred to have significant [O iii]+H$\beta$ equivalent width (EW) based on Spitzer photometry (Roberts-Borsani et al., 2016). All four yielded Lyman-$\alpha$ detections as distant as $z=8.68$ (Oesch et al., 2015; Zitrin et al., 2015a; Stark et al., 2017). This suggests that luminous galaxies with high ionization parameters carve out ionized gas bubbles allowing Lyman-$\alpha$ to stream free (Stark et al., 2017).

Alternatives to Lyman-$\alpha$ redshift confirmations include fainter, slightly redder UV metal lines N v$\lambda$1243Å, C iv$\lambda$1549Å, He ii$\lambda$1640Å, O iii]$\lambda$1663Å, and C iii]$\lambda$1908Å (Stark et al., 2014) and sub-mm lines including [C ii] 158$\mu$m and [O iii] 88$\mu$m (Inoue et al., 2014). Studying these lines also yields more information about the physical properties of the galaxies, including their ionizing strength.

The UV metal line C iii] has been detected in several $z>6$ galaxies as distant as $z=7.73$ (Stark et al., 2015, 2017; Laporte et al., 2017b; Mainali et al., 2018). At $z\sim 2$, Rigby et al. (2015) found that lower metallicity galaxies exhibit stronger C iii], likely explaining why C iii] appears far more often at $z\sim 6$ than at lower redshifts (see also Senchyna et al., 2017; Du et al., 2017; Le Fèvre et al., 2017). C iii] detections may be powered by low metallicity massive stars (Stark et al., 2015, 2017) or may require AGNs (Nakajima et al., 2018). Still lower metallicity stars ($<0.05Z_{\odot}$) produce higher ionization potentials yielding C iv. All three known $z>6$ C iv detections are lensed (Stark et al., 2015; Mainali et al., 2017; Hoag et al., 2017) and thus less massive and presumably lower metallicity than average $z>6$ galaxies from current surveys.

Larger samples of UV metal line observations including RELICS galaxies will enable us to quantify the prevalence of these intense ionizing sources in the reionization epoch, both directly and by extrapolating to larger samples with similar observed properties (luminosity, rest-frame UV slope $\beta$, and Lyman-$\alpha$). The observed equivalent widths will also be very informative to future JWST surveys planning to observe these spectral features.

A new window for studying high-redshift galaxies has been opened by the Atacama Large Millimeter/submillimeter Array (ALMA) in the south and the Plateau de Bure interferometer (PdBI) in the north, later upgraded and renamed the Northern Extended Millimeter Array (NOEMA). These telescopes (primarily ALMA) have detected high-redshift galaxies by their infrared continuum dust emission (Capak et al., 2015) as well as spectral lines [C ii] 158$\mu$m and [O iii] 88$\mu$m, which trace star formation (De Looze et al., 2014). Continuum observations show that $z\sim 5-6$ galaxies have at least an order of magnitude less dust than local starbursts with similar rest-frame UV colors (Capak et al., 2015). Dust is often not detected at these redshifts, but at least a few galaxies have been found to be dusty and thus evolved as early as $z=7.15$ (Hashimoto et al., 2018a), $z=7.5$ (Watson et al., 2015), $z=8.312$ (Tamura et al., 2018), and $z=8.38$ (Laporte et al., 2017b).

[C ii] 158$\mu$m is one of the brightest lines in local galaxies (Malhotra et al., 1997; Brauher et al., 2008) and the strongest observed ISM cooling line in $z\sim 1-2$ galaxies (Stacey et al., 2010). To date, [C ii] has been detected in 22 galaxies between $5.1533\leq z\leq 7.1453$ (Carniani et al. 2018 and references therein; Hashimoto et al. 2018a), including two spatially resolved galaxies at $z=6.81$ and 6.85 displaying disk-like rotation (Smit et al., 2018).

At higher redshifts, Inoue et al. (2014) predicted that [O iii] 88$\mu$m would yield more detections. While [C ii] is associated with neutral H i gas in photodissociation regions (PDRs), [O iii] is associated with ionized H ii gas, more prevalent in higher redshift, lower metallicity galaxies with higher ionization states (Harikane et al., 2018). As predicted, ALMA’s highest-redshift spectroscopic confirmations have come from [O iii], and the six $z>6$ galaxies targeted to date have all yielded [O iii] detections at $z=6.900$ (Marrone et al., 2018); $z=7.107$ (Carniani et al., 2017); $z=7.212$ (Inoue et al., 2016); $z=8.312$ (Tamura et al., 2018); $z=8.382$ (Laporte et al., 2017a); and $z=9.11$ (Hashimoto et al., 2018b). The two highest redshift detections are lensed galaxies.

Based on this previous work, we expect RELICS to deliver ALMA confirmations and science at $z\sim 6-10$ (see §6.1).

2.1.2 Luminosity Functions of Galaxies at the Epoch of Reionization

Improving constraints on the $z\sim 9$ luminosity function (and the evolution from $z\sim 10$ to 6) is another primary science goal of RELICS. This is required to determine the numbers of faint galaxies available to reionize the universe.

Planck constrained the reionization history by measuring the column density of free electrons to the CMB, or the Thomson scattering optical depth $\tau=0.058\pm 0.012$ (Planck Collaboration et al., 2018a). This value, lower than previous estimates, implies a later reionization halfway complete by $z\sim 8$, before fully completing by $z\sim 6$. It also means reionization can be achieved by galaxies producing less Lyman-continuum (LyC) flux and with lower escape fractions $f_{esc}$. Direct measurements of these $f_{esc}$ values have been obtained recently for $\sim$40 galaxies with detected LyC leakage, including half extending from the local universe (e.g., Leitherer et al., 2016) to $z\sim 0.3$ (e.g., Izotov et al., 2018), with the other half at $z\sim 2-4$ (e.g., Shapley et al., 2016; Rivera-Thorsen et al., 2017; Vanzella et al., 2018). Assuming these measurements hold at higher redshifts, and given high-redshift luminosity functions, low-mass galaxies could have produced most of the flux required to reionize the universe (e.g., Robertson et al., 2015; Madau, 2017; Ishigaki et al., 2018; Mason et al., 2018; Finkelstein et al., 2019). There may also have been significant contributions from low-luminosity AGN jets at $z\gtrsim 6$ (Bosch-Ramon, 2018).

At the highest redshifts ($z\sim 9-12$), Hubble and Spitzer imaging programs have yielded fewer candidates than expected. This has left luminosity functions highly uncertain at these redshifts, while also hinting at accelerated evolution in the first 600 Myr (Bouwens et al., 2012; Oesch et al., 2018). Luminosity functions are fairly well constrained at $z\sim 4-8$ (Bouwens et al., 2015; Finkelstein, 2016; Kawamata et al., 2018). Only recently were constraints placed on all three parameters ($\phi^{*}$, $M^{*}$, $\alpha$) of the $z\sim 9$ luminosity function (Morishita et al., 2018); these results are consistent with either accelerated or smoother (non-accelerated) evolution at $z>8$. Higher precision measurements, especially with greater leverage at higher redshifts, are required to better constrain this early evolution rate.

2.2 Strong Lens Modeling

Robust strong lens modeling is required to estimate magnifications of our lensed galaxies and the survey volume. The mass in galaxy cluster cores responsible for the lensing is predominantly dark matter. We must infer the distribution of this dark matter based on observed strong lensing that produces multiple images of more distant galaxies. We often add the assumption that luminous galaxy cluster members are good tracers of their dark matter halos.

To this end, RELICS obtained 7-band HST imaging of each cluster, and we are using ground-based telescopes to measure spectroscopic redshifts of strongly lensed galaxies. High-quality lens modeling generally requires multiband high-resolution HST imaging to reliably identify multiple images of strongly lensed galaxies based on their colors and morphologies. Spectroscopic redshifts are also crucial. Magnification accuracies improve with greater numbers of strongly lensed images with spectroscopic redshifts (Johnson & Sharon, 2016). Recent tests with simulated cluster lensing find that the best lens models accurately yield magnification estimates of 3 (10) with precisions of 15% (30%), with uncertainty increasing with magnification (Meneghetti et al. 2016; see also Zitrin et al. 2015b). Such precision is encouraging, as the vast majority (over 80%) of lensed high-redshift galaxies observed are magnified by factors of 10 or less (Coe et al., 2015).

Magnification uncertainties directly impact measurements of some physical properties such as luminosity, star formation rate, stellar mass, and size. However, since lensing is achromatic, other properties derived from galaxy colors are not affected by magnification; these include redshift, age, metallicity, extinction, and rest-frame UV slope. Magnifications do affect luminosity functions, but the uncertainties are mitigated by averaging over many galaxies. The resulting uncertainties on the total survey volume are subdominant compared to the current small number statistics at high redshifts (Coe et al., 2015).

In the case of a multiply-imaged high-redshift candidate, lens modeling can yield geometric support, distinguishing between low and high-redshift solutions based on the separation between the observed images (e.g., Coe et al., 2013; Zitrin et al., 2014; Chan et al., 2016). Lens modeling is required to study delensed (source plane) properties, including $\sim$100 parsec structures resolved in highly elongated arcs (e.g., Sharon et al., 2012; Johnson et al., 2017). Lens modeling is also required to constrain luminosity functions in lensed fields (e.g., Livermore et al., 2016).

RELICS lens modeling is yielding the overall lensing strength for each cluster. While we expect all RELICS clusters to be excellent lenses, our analyses will reveal which are truly among the strongest and best to use for efficient discoveries of the most distant galaxies known in future surveys. For RELICS results to date, see §6.2.

2.3 Galaxy Cluster Masses

Scaling relations linking mass estimates from lensing, X-ray, and SZ studies require understanding of observational systematics and scale-dependent cluster astrophysics. Improving the accuracy of these mass scaling relations will be key to realizing the full potential of future missions such as eROSITA, which anticipates detecting 100,000 clusters / groups out to $z\sim 1.3$.

Constraints on cosmological parameters (primarily $\Omega_{m}$ and $\sigma_{8}$) derived from the Planck CMB (Planck Collaboration et al., 2016b) are at odds with those derived from Planck SZ galaxy cluster counts (Planck Collaboration et al., 2016c). The latter paper calibrated the scaling relation between Planck SZ signal strength and cluster masses based on analyses of XMM-Newton X-ray observations assuming hydrostatic equilibrium (HSE). One may expect cluster masses to be $\sim$20% greater due to deviations from HSE, including non-thermal pressure support (e.g., Nagai et al., 2007; Rasia et al., 2012). But Planck Collaboration et al. (2016c) found that the cosmology results can be reconciled only if one assumes that SZ-derived cluster masses underestimate the true $M_{500}$ masses by $b\approx 42\%$ (where they reported the mass bias as $1-b=0.58\pm 0.04$). That is about 2-$\sigma$ greater than the $\sim$20% initially adopted in the PSZ1 analysis (Planck Collaboration et al., 2014).

Weak lensing analysis of clusters included in both the Planck analysis and the Weighing the Giants survey (WtG; von der Linden et al. 2014a) shows Planck cluster masses may indeed be underestimated by $\sim$42% for the most massive clusters ($>10^{15}M_{\odot}$), while Planck masses appear to be more accurate for less massive clusters ($\sim 5\times 10^{14}M_{\odot}$). Subsequent weak lensing analyses from various surveys (WtG, CCCP, LoCUSS, CLASH, CFHTLenS, RCSLenS, HSC-SSP) found a range of results, some consistent with WtG including bias increasing with mass (Sereno & Ettori, 2017), and others more consistent with the original Planck estimate of $\sim$20% bias (Hoekstra et al., 2015; Battaglia et al., 2016; Smith et al., 2016; Penna-Lima et al., 2017; Medezinski et al., 2018). Overall, the tension appears to be somewhat relieved, although not conclusively (Planck Collaboration et al., 2018b), especially after accounting for new Planck measurements of the reionization optical depth (Planck Collaboration et al., 2016d, 2018a).

SZ mass estimates are calibrated in part based on X-ray temperatures measured with spectroscopy out to $r_{500}$, which can be difficult, especially for higher redshift clusters. X-ray mass calibration at the smaller $r_{2500}$ radii probed by HST strong lensing may be key to improving mass scaling relations. These smaller radii are also important for ground-based SZ telescopes offering significantly higher resolution than Planck. Czakon et al. (2015) analyzed arcsecond-resolution Bolocam SZ imaging of 45 clusters and found the $Y_{SZ}$ signal does not scale self-similarly with $M_{2500}$ derived from X-ray observations. The cause of this mismatch may be due to AGN feedback regulating star formation and altering the gas properties in cluster cores. Through accurate lensing mass profiles measuring total mass, free from the assumption of gas HSE, and accurate gas mass profiles from Chandra, we will determine the radial dependence of $f_{gas}=M_{gas}/M_{tot}$ and characterize the efficiency with which AGN input energy into the intracluster medium.

Long term efforts to improve mass scaling relations focus on weak lensing analyses, for example from Euclid and LSST (Grandis et al., 2018). However, strong lensing analyses will also contribute. Simulations have shown that joint strong + weak lensing analyses yield significantly smaller biases ($\sim$2%) and uncertainties ($\sim$20%) in virial mass estimates than either weak or strong lensing alone (Meneghetti et al., 2010). A joint strong plus weak lensing analysis of 20 CLASH clusters based on HST and Subaru observations (Umetsu et al., 2016) shows that the $M_{2500}$ masses can be determined with a fractional total uncertainty of 25% per cluster at <$$M_{2500}$$> $=3.6\times 10^{14}M_{\odot}$ (or <$$M_{500}$$> $=10^{15}M_{\odot}$), accounting for the dominant contributions from intrinsic profile variations (Gruen et al., 2015). Thus a similar analysis of all 41 RELICS clusters would yield an overall mass calibration accuracy of about 4% from $r_{2500}$ out to the virial radii, providing a legacy mass-profile database for critical tests for models of background cosmology and structure formation.

2.4 Dark Matter Constraints

Self-interacting dark matter (SIDM) particles with cross sections of $\sigma_{\rm DM}/m\sim 0.1-0.6~{}{\rm cm}^{2}/{\rm g}$ have been invoked to explain several observational inconsistencies with collisionless cold dark matter (CDM) theory, including galaxy cluster density profiles, “missing” Milky Way satellites, and the “cusp-core” problem (see Robertson et al. 2018 and references therein). Baryonic processes can explain these “problems” with varying degrees of success. Merging galaxy clusters yield the most robust constraints on SIDM. As two clusters collide, galaxies pass straight through the collision, leaving the stripped cluster gas lagging behind. We expect dark matter to also pass straight through unless the particles have a significant self-interacting cross section. Joint HST lensing plus Chandra X-ray analysis of the Bullet Cluster (Clowe et al., 2006; Bradač et al., 2006) constrained $\sigma_{\rm DM}/m<1.25~{}{\rm cm}^{2}/{\rm g}$ (68% confidence; Randall et al. 2008). More recently, Harvey et al. (2015) improved this constraint to $\sigma_{\rm DM}/m<0.47~{}{\rm cm}^{2}/{\rm g}$ (95% confidence) by jointly analyzing HST + Chandra imaging of 30 cluster mergers (e.g., Bradač et al., 2008; Dawson et al., 2012). However, Wittman et al. (2018) subsequently analyzed more data (including more HST filters) on these same 30 clusters and claimed the constraint should be more relaxed: $\sigma_{\rm DM}/m\lesssim 2~{}{\rm cm}^{2}/{\rm g}$. RELICS includes 7 cluster mergers analyzed by these papers plus another 17 confirmed plus 9 possible mergers not in their sample. Of the confirmed mergers, all but one has existing Chandra imaging. Joint analyses of the lens models plus X-ray data of these massive cluster mergers will help constrain the SIDM parameter space toward the astrophysically and theoretically interesting limit of $0.1~{}{\rm cm}^{2}/{\rm g}$.

2.5 Supernovae

RELICS observed each cluster in two epochs separated by 30–60 days to identify supernovae (SNe) and other transient phenomena. This cadence was designed to catch $z\sim 1-2$ SNe near peak brightness in epoch 2. The RELICS proposal included 20 orbits for follow-up imaging to obtain light curves for the more interesting supernovae. Ultimately, RELICS discovered 11 supernovae (§6.3).

Our primary goal was to significantly contribute to the numbers of known distant and lensed supernovae discovered in previous surveys. CLASH and CANDELS discovered both Type Ia supernovae (SNIa) and Core Collapse supernovae (CCSN) out to $z=2.5$. The observed SNIa rates suggest that Type Ia progenitors are primarily double white dwarf systems (Graur et al., 2014) that do not explode quickly after formation (Rodney et al., 2014). High-z CCSN rates from these programs reinforced measurements of the cosmic star formation rate history and put constraints on the initial mass function for CCSN progenitors (Strolger et al., 2015).

Lensed Type Ia supernovae from CLASH and the Frontier Fields provided the first empirical tests verifying the accuracy of lens model magnification estimates (Patel et al., 2014; Nordin et al., 2014; Rodney et al., 2015b). The appearance of “SN Refsdal” (Kelly et al., 2015) was the first strongly-lensed SN observed as multiple images with measurable time delays (Rodney et al., 2016b). These time delays can be used to test lens models (Treu et al., 2016; Kelly et al., 2016) or as a cosmological distance measurement (Vega-Ferrero et al., 2018; Grillo et al., 2018). Strong lensing clusters have also revealed other faint transient phenomena, including extreme magnifications of individual stars or stellar eruptions in galaxies at $z=1.49$ (Kelly et al., 2018), $z=1.01$ (Rodney et al., 2018), and $z=0.94$ (Chen et al., 2019; Kaurov et al., 2019). RELICS observations will also provide a baseline for the longer term (decade) monitoring required to detect higher-redshift supernovae and other lensed transients, perhaps including individual Population III stars (Windhorst et al., 2018).

3 Galaxy Clusters

RELICS observed the 41 massive galaxy clusters listed in Tables 2 and 3. None had existing Hubble infrared imaging (WFC3/IR or NICMOS) prior to RELICS. We selected 21 of the clusters based on exceptionally high mass estimates from Planck and the other 20 based on other factors revealing or suggesting exceptional lensing strength.

Planck Collaboration et al. (2016a) identified 1653 Sunyaev-Zel’dovich (SZ; Sunyaev & Zeldovich 1970) sources in their PSZ2 all-sky ($|b|>15^{\circ}$) catalog111Planck PSZ2 catalogs are available and described at: http://irsa.ipac.caltech.edu/data/Planck/release_2/catalogs/ and https://wiki.cosmos.esa.int/planckpla2015/index.php/Catalogues, including 1203 confirmed clusters, 1094 of which had measured redshifts and SZ mass estimates. A few hundred of these were newly discovered clusters, including some of the most massive known.

We queried the HST observations of all the Planck clusters using the Mikulski Archive for Space Telescopes (MAST)222http://archive.stsci.edu/hst/search.php. Many had existing HST imaging prior to RELICS, but many more did not.

The 34 most massive clusters ($M_{500}>8.8\times 10^{14}M_{\odot}$) include some already well studied by Hubble and Spitzer, including 4/6 Frontier Fields clusters333The Frontier Fields cluster Abell 370 has a lower Planck mass $M_{500}=7.7\times 10^{14}M_{\odot}$ (though weak lensing analysis yields a far greater $M_{500}=1.9\times 10^{15}M_{\odot}$; Umetsu et al. 2011). MACSJ0416.1-2403 is PSZ2 G221.06-44.05, but the association was not initially made; without the redshift, no mass estimate was possible in the PSZ2 catalog. However, the relatively low Planck signal-to-noise ratio of 5.2 suggests a lower mass. and 5/25 from CLASH (with 3 common to both surveys). But others were less well studied. Nine of the top 34 had yet to be observed by Hubble or Spitzer at all, and 12 more had yet to be observed by Hubble in the near-infrared with NICMOS or WFC3. These 21 clusters comprise half the RELICS sample; RELICS obtained the first HST infrared imaging of these clusters, and Spitzer imaging as needed.

The remaining 20 RELICS clusters are primarily known strong lenses based on existing Hubble optical imaging (or ground-based imaging in the case of RXC0232-44; Kausch et al. 2010). Existing HST ACS imaging, where available, reduced our total orbit request. We also considered other factors in this selection, including X-ray mass estimates (Mantz et al. 2010; MCXC Piffaretti et al. 2011); weak lensing mass estimates (Sereno 2015 compilation including Weighing the Giants Applegate et al. 2014, von der Linden et al. 2014b; Umetsu et al. 2014; Hoekstra et al. 2015); SDSS clusters (Wong et al., 2013; Wen et al., 2012); and other SZ mass estimates from SPT (Bleem et al., 2015) and ACT (Hasselfield et al., 2013) as well as clusters nearly selected for the Frontier Fields444http://www.stsci.edu/hst/campaigns/frontier-fields/frontier-fields-high-magnification-cluster-candidate-list/.

The 41 RELICS clusters generally bear the names of the surveys which discovered them: Abell (1958); Abell et al. (1989); the Einstein Observatory Extended Medium-Sensitivity Survey (MS; Gioia et al., 1990); the ROSAT X-ray All-Sky Galaxy Cluster Survey (RXC; Ebeling et al., 1998, 2000a); the ROSAT MAssive Cluster Survey (MACS and SMACS; Ebeling et al., 2001, 2007, 2010, 2013; Mann & Ebeling, 2012; Repp et al., 2016; Repp & Ebeling, 2018); an extended radio source discovered in ROSAT (RXS; van Weeren et al., 2012); the Wen, Han, & Liu SDSS-III cluster catalog (WHL; Wen et al., 2012); the South Pole Telescope SZ survey (SPT; Bleem et al., 2015); the Atacama Cosmology Telescope SZ survey (ACT; Hasselfield et al., 2013); and the Planck all-sky SZ survey (Planck Collaboration et al., 2011b, 2016a). Cl J0152.7–1357 at $z=0.833$ was discovered by Ebeling et al. (2000b) in the Wide Angle ROSAT Pointed Survey (WARPS). Alternate cluster names include one also identified in the Rosat Bright Survey (RBS; Kausch et al., 2010) and another in Clusters in the Zone of Avoidance (CIZA; Kocevski et al., 2007).

The Abell clusters are numbered sequentially based on the original catalog. Numbers in other cluster names codify coordinates, often in J2000 RA HH:MM and Dec DD:MM. Planck cluster names, instead, give Galactic coordinates in longitude and latitude. Table 3 lists names and any alternate names for each RELICS target.

4 Observations

All data from our large Hubble and Spitzer observing programs had no proprietary period. We describe each of these programs in turn, followed by other subsequent large observing programs of RELICS clusters.

4.1 Hubble Imaging

Of the 41 RELICS clusters, 28 had been observed previously by HST with ACS and/or WFC3/UVIS (Table 4). Our 188-orbit HST Treasury Program (Cycle 23; GO 14096; PI Coe; Deputy PI Bradley) obtained additional observations with ACS and WFC3/IR (Table 5). Five clusters required two WFC3/IR pointings for a total of 46 new WFC3/IR images of strongly lensed fields. For each field, we observed 2 orbits WFC3/IR split among F105W, F125W, F140W, and F160W. And for each cluster, we observed 3 orbits ACS split among F435W, F606W, and F814W, minus any archival imaging. For the 18 RELICS clusters without any existing ACS imaging, we observed the full 3 orbits of ACS and, in parallel, 3 orbits of WFC3 on a blank field: 1 orbit WFC3/UVIS F350LP and 2 orbits WFC3/IR F105W, F125W, F140W, and F160W. The parallels use the same filters as in BoRG[z9-10] (Calvi et al., 2016; Morishita et al., 2018), adding area to our high-z search.

For each cluster, we split the observations into two epochs separated by about a month to enable searches for supernovae and other transients. Twenty orbits of RELICS were allocated to follow up such Targets of Opportunity (ToO). We executed these orbits on three of the nine supernovae discovered by RELICS (Table 6).

Total integration times for each HST orbit vary with the time available between Earth occultations and the time to acquire or reacquire guide stars. Table 7 gives the total integration times in each filter for Abell 1763, as a representative example. For most other clusters, the integration times were slightly longer (by up to 15%). Each integration was split into four exposures at different dither positions (pointings). We dithered across the ACS chip gap and WFC3/IR “death star” to fill these gaps in the data.

Table 8 gives the detailed break down of epochs, orbits, and exposure times at each of the four dither positions in each filter for a nominal target with no archival imaging. In the parallel fields, note that the WFC3/UVIS F350LP imaging consists of four exposures at four dither positions, but the parallel WFC3/IR imaging consists of only two exposures in each filter (one per epoch).

4.2 Spitzer Imaging

Altogether, RELICS has been awarded 945 hours of Spitzer observing time (Table 9). For each cluster, we obtained IRAC imaging as needed in the two warm mission filters, ch1 (3.6$\mu$m) and ch2 (4.5$\mu$m). Combined, these filters span approximately $3-5$ $\mu$m. About 100 hours of archival IRAC imaging were also available for 18 of the clusters in these two filters, most notably from GO 60034 (PI Egami). Below we summarize the complete Spitzer RELICS datasets, which will be detailed further by Strait et al. (2019, in preparation).

Our initial 100-hour SRELICS (Spitzer RELICS) program (Cycle 12; GO 12005; PI Bradač) was supplemented by a 290-hour Director’s Discretionary Time program (DDT 12123; PI Soifer). Combined, these programs observed all RELICS clusters (as needed) to achieve a total of 5 hours integration time (combining new and archival imaging) in each of the two IRAC filters. The five clusters requiring two WFC3/IR pointings also required two IRAC pointings.

Subsequently, based on our analyses of the HST and Spitzer images, we were awarded three more proposals (PI Bradač) to obtain deeper IRAC imaging (30 hours / band) requiring an additional 556 hours on the 10 clusters yielding the most high-z candidates at $z\sim 6-10$ (Salmon et al., 2017, 2018). DDT 13165 observed PLCK G287.0+32.9, PLCK G004.5–19.5, and A1763. DDT 13210 observed SPTCLJ0615–5746 and the $z\sim 10$ arc discovered by Salmon et al. (2018). And GO 14017 observed CLJ0152.7–1357, ACTCLJ0102–49151, PLCKG308.3–20.2, RXS J060313.4+421, MS1008.1–1124, and SMACSJ0723.3–7327.

4.3 Other Large Surveys of RELICS Clusters

In addition to the large HST and SST observing programs, RELICS has also motivated large surveys with other telescopes including VST, VISTA, XMM-Newton, ALMA, and VLA (see Table 10). The VLA survey exclusively observes RELICS clusters, while the other listed programs have most or just some clusters in common with RELICS. We have also carried out smaller follow-up programs with many other observatories including Keck MOSFIRE & DEIMOS, VLT MUSE & X-SHOOTER, Gemini GMOS, Subaru SuprimeCam & HSC, Magellan MegaCam & LDSS3, MMT Hectospec, and GMRT. Spectra from Magellan LDSS3 have already been used in several papers (Cerny et al. 2018; Paterno-Mahler et al. 2018; Mainali et al. 2019, in preparation). Also note that while we are advertising these newer surveys here, we also emphasize that many RELICS clusters have been observed previously by large programs carried out with Hubble, Spitzer, Chandra, XMM-Newton, and ground-based observatories.

5 Image Reductions and Catalogs

We have made our HST reduced images and catalogs publicly available via MAST555https://archive.stsci.edu/prepds/relics/ at https://doi.org/10.17909/T9SP45 (catalog doi:10.17909/T9SP45) and our Spitzer reduced images available via IRSA666https://irsa.ipac.caltech.edu/data/SPITZER/SRELICS/. Below we describe the procedures used to generate these data products.

5.1 Hubble Image Reductions

We reduced all HST images obtained by RELICS and all archival HST ACS and WFC3/IR images that overlap with the RELICS WFC3/IR images. Our HST image reduction procedure is similar to that performed on the Frontier Fields HST images (Lotz et al., 2016). Key differences are that we do not produce “self-calibrated” ACS images (due to insufficient numbers of exposures), nor do we correct for time-variable sky emission due to Helium line emission at 1.083$\mu$m which occasionally affects WFC3/IR F105W images (Brammer et al., 2014).

We correct all HST images for bias, dark current, and flat fields with up-to-date reference files. We manually identified and masked any satellite trails. ACS images were corrected for CTE (charge transfer efficiency) and bias striping. The multiple ACS exposures were used to automatically identify and reject cosmic rays. Each WFC3/IR MULTIACCUM exposure consists of multiple samples, enabling “up-the-ramp” cosmic ray rejection. We also masked bad pixels using updated identifications in WFC3/IR images from GO 14114 (G. Brammer, private communication).

We produced inverse variance maps (IVMs) quantifying the uncertainty in each pixel before accounting for correlated pixel noise (an additional 10-15%) and Poisson source noise. These IVMs were used as weights to drizzle-combine the images obtained in each filter.

For each cluster or parallel field, we aligned all processed HST images to two common grids with 0.06″ and 0.03″ resolution. We used procedures from DrizzlePac, specifically AstroDrizzle (Gonzaga et al., 2012) and as outlined in Koekemoer et al. (2002, 2011) and Lotz et al. (2016). We corrected the absolute astrometry of our images using the Wide-field Infrared Survey Explorer (WISE) point source catalog (Wright et al., 2010).

Finally, we produced automatically scaled color images using Trilogy (Coe et al., 2012). Figures 6 – 6 show HST ACS + WFC3 color images of all 46 WFC3/IR cluster fields observed by RELICS. The ACS color images extend to wider areas, which are not shown here, but are available on MAST. Also on MAST are color images of the 18 RELICS parallel fields observed with WFC3 UVIS and IR.

5.2 Hubble Detection and Photometry Catalogs

Based on the 0.06″ resolution HST images, we produced source catalogs using techniques similar to those employed for the public CLASH catalogs (Postman et al., 2012; Coe et al., 2013) and the Frontier Fields analysis presented in Coe et al. (2015).

We ran SExtractor (Bertin & Arnouts, 1996) version 2.8.6 in dual-image mode to detect objects in each field and define their isophotal apertures for photometry to be measured in each filter image. For each field, we produce two source catalogs:

• •

acs-wfc3 (or acs-wfc3ir): based on detections in a weighted stack of all HST images (ACS, WFC3/UVIS, and WFC3/IR), optimized to detect most objects

• •

wfc3ir: based on detections in a weighted stack WFC3/IR images only using a finer background grid and more aggressive deblending, optimized to detect smaller high-redshift galaxies

The stacked images are weighted sums; the weights are the IVMs produced by the drizzling software (§5.1). For SExtractor input, we also produce RMS maps equal to 1 / sqrt(weight).

Table 11 lists the SExtractor parameters used for the acs-wfc3 and wfc3ir catalogs. The latter is geared toward detections of small high-redshift galaxies, so we use a smaller background grid and more aggressive deblending to detect small objects near brighter ones. The acs-wfc3 catalogs cover the full (larger) ACS field of view and aim to detect whole objects, breaking them apart less often.

For detections in either catalog, 9 contiguous pixels are required at the level of the observed background RMS or higher. The acs-wfc3 (wfc3ir) background is calculated in 5$\times$5 (3$\times$3) grids of cells with 128$\times$128 (64$\times$64) pixels in each cell. We set the deblending of adjacent objects to 128 (64) levels of 0.0015 (0.0001) minimum contrast.

Each object’s detection defines an isophotal aperture, which SExtractor uses in dual-image mode to measure isophotal photometry in every filter in the aligned images. Just outside this aperture, we have SExtractor use a 24-pixel-wide rectangular annulus around each object to estimate and subtract the local background in each filter. We do not perform aperture corrections, as the PSF FWHMs only vary between $\sim 0.07^{\prime\prime}$–$0.15^{\prime\prime}$ and we use relatively large isophotal apertures (see discussion in Postman et al. 2012, §5.1).

Finally, we correct all photometry for Galactic extinction using the IR dust emission maps of Schlafly & Finkbeiner (2011)777Dust extinctions extracted using [http://irsa.ipac.caltech.edu/

applications/DUST/docs/dustProgramInterface.html](http://irsa.ipac.caltech.edu/applications/DUST/docs/dustProgramInterface.html). Table 2 gives the extinction E(B–V) for each cluster from those maps. These are multiplied by coefficients $A_{\lambda}$ (Table 12) for each filter to determine the extinction in magnitudes.

Table 13 summarizes the output in our HST catalogs available on MAST. Samples are provided from our Abell 697 IR-detection catalog for object detection and shape measurement (Table 14), photometry (Table 15), and Bayesian photometric redshifts (Table 16), which are discussed below.

As a caution to users, we note the source catalogs include image artifacts such as diffraction spikes as well as objects poorly segmented by SExtractor. Additionally, photometry is complicated in crowded fields, especially as brighter cluster members contaminate the light from fainter, more distant objects.

5.3 Photometric Redshifts

Based on the Hubble photometry, we measured photometric redshifts using Bayesian Photometric Redshifts (BPZ; Benítez 2000; Coe et al. 2006) and Easy and Accurate $z_{phot}$ from Yale (EAZY; Brammer et al. 2008), which were two of the top performing methods in controlled tests (Hildebrandt et al., 2010).

For BPZ, we used 11 spectral models shown in Figure 4 and described in Coe et al. (2013); Benítez et al. (2014); Rafelski et al. (2015). Briefly, the model spectral energy distributions (SEDs) are originally from PEGASE (Fioc & Rocca-Volmerange, 1997), but recalibrated, based on observed photometry and spectroscopic redshifts from FIREWORKS (Wuyts et al., 2008). The templates were selected to encompass ranges of metallicities, extinctions, and star formation histories observed for the vast majority of real galaxies. We allowed BPZ to interpolate 9 templates between each pair of adjacent templates, yielding 101 templates altogether. BPZ fit the photometry to a grid of these 101 templates and 1300 redshifts linearly spaced from $z=0.001$ to $z=13$. BPZ tempers those $\chi^{2}$ results with a Bayesian prior $P(z,T|m)$, which gives the likelihood of a redshift $z$ and template type $T$ given an observed magnitude $m$ in F814W. We used the original BPZ prior derived from the HDFN (Benítez, 2000) and plotted in Figure 5 for $m=24$, 26, and $\geq 32$, the latter including F814W non-detections. We do not attempt to correct magnitudes for lensing magnifications in our initial catalogs because those estimates are not available from the start. We note our prior’s dependence on magnitude is often gradual, but magnification should be accounted for a more accurate $P(z)$ estimate.

We used EAZY (Brammer et al., 2008) to obtain a second independent set of photometric redshift estimates. EAZY uses a different template set and allows interpolation between any pair of templates. The 9 templates used here include 7 from PEGASE, plus one very dusty and quiescent galaxy from Maraston (2005), and one extreme emission line galaxy (EELG) from Erb et al. (2010). We used a flat prior with EAZY as we found the default prior was systematically biased against high-redshift galaxies, strongly preferring lower redshift EELGs with worse fits to the photometry (see discussion in Salmon et al. 2017).

Salmon et al. (2017) presents a comparison of the templates and results from BPZ and EAZY for RELICS high-z candidates.

5.4 Spitzer Image Reductions

We reduced the Spitzer images using MOPEX (Makovoz & Khan, 2005) and generated catalogs using T-PHOT (Merlin et al., 2015). The reduced images are available via IRSA. Currently these include data from programs 12005 and 12123 (totaling 5 hours depth per filter). Deeper data from the additional three Spitzer programs (Table 9) will be included in future releases of reduced images. More details will be presented by Strait et al. (2019, in preparation). Due to the broader Spitzer PSF, extra care is required in obtaining aperture-matched Hubble + Spitzer photometry. Less careful Spitzer photometry can result in less accurate photometric redshifts (Hildebrandt et al., 2010). We did not use the Spitzer photometry in our initial photometric redshift catalog release. We did use the Spitzer photometry to vet our $z\sim 10$ candidates, and we will use it to study the properties of all our high-redshift candidates.

6 Results

To date, RELICS has delivered the following science results on high-redshift galaxies (§2.1), strong lens modeling (§2.2), and supernovae (§2.5). Reduced images, catalogs, and lens models are available via MAST and IRSA. Constraints on cluster masses (§2.3) and the dark matter particle cross section (§2.4) require weak lensing data and analyses on a longer timescale.

6.1 High-Redshift Candidates

RELICS yielded 321 high-redshift candidates at $z\sim 6-8$, including the brightest known at $z\sim 6$ (Salmon et al., 2017). These galaxies are lensed as brightly as F160W $H\sim 23$, enabling detailed studies of galaxy properties in the first billion years. Follow-up study is beginning to match some of these candidates as multiple images (Acebron et al., 2018a) and deliver spectroscopic confirmations (Cibirka et al., 2018). We are following up RELICS high-redshift candidates with ground-based telescopes and instruments including Keck/MOSFIRE, VLT/MUSE, Gemini/GMOS, Gemini-S/Flamingos-2, ALMA, and NOEMA/PdBI.

RELICS also delivered SPT0615-JD, the most distant lensed arc known (Salmon et al., 2018). At $z\sim 10$ and spanning a full $2.5^{\prime\prime}$ on the sky, SPT0615-JD provides by far the most detailed view we have of any galaxy in the first 500 million years. (The two known $z\sim 11$ galaxies are not spatially resolved, despite lensing magnification in the case of MACS0647-JD from Coe et al. 2013, and the relatively high intrinsic luminosity of GN-z11 from Oesch et al. 2016.) ALMA observations have been awarded (PI Tamura) to search for the [O iii] 88$\mu$m line in the $z\sim 10$ lensed arc, continuing the success of that research group at $z\sim 7-9$ as noted in §2.1 (Inoue et al., 2016; Tamura et al., 2018; Hashimoto et al., 2018b). An [O iii] detection would yield the highest spectroscopic redshift confirmation to date along with the earliest detection of heavy elements (oxygen).

RELICS Spitzer (SRELICS) imaging was crucial in distinguishing between bonafide $z\sim 10$ candidates and $z\sim 2$ interlopers. Salmon et al. (2018) actually identified three $z\sim 10$ candidates based on our HST imaging, but two turned out to be $z\sim 2$ interlopers based on the Spitzer photometry. Red $z\sim 2$ galaxies are significantly brighter at 3–5$\mu$m than bluer $z\sim 10$ galaxies. RELICS Spitzer imaging will also enable us to measure stellar masses for our $>300$ candidates at $z\sim 6-8$.

Improved constraints on the $z\sim 9$ luminosity function from RELICS will require adding simulating lensed galaxies to our images to quantify our detection efficiency as a function of magnitude, position, and redshift (e.g., Livermore et al., 2016; Carrasco et al., 2018). But based on our current lack of any strong $z\sim 9$ candidates and a single $z\sim 10$ candidate (Salmon et al., 2017, 2018), our yields appear lower than expected at these redshifts, suggesting support for the accelerated evolution scenario. Alternatively, more detailed study may shed light on why we have missed galaxies at these redshifts in our searches and photometric redshift analyses.

At lower redshifts, RELICS is studying compact, low metallicity dwarf galaxies that are excellent analogs to high-redshift galaxies, but can be studied in greater detail. Analysis of one RELICS $z=1.645$ galaxy shows that low metallicity stars are driving C iii] emission with the strongest rest-frame equivalent width ($\sim$20Å) yet observed at these redshifts, suggesting a more intense radiation field than assumed by most population synthesis models (Mainali et al. 2019, in preparation).

6.2 Lens Modeling

Strong lens models of RELICS clusters are primarily used to estimate magnifications of our lensed galaxies and to correct the surveyed volume for lensing magnification (§2.2). We use three lens modeling methods, all of which assume the observed cluster light traces some component of the cluster mass distribution: Lenstool (Kneib et al., 1996; Jullo & Kneib, 2009; Johnson et al., 2014); Zitrin LTM, or light-traces-mass (Zitrin et al., 2009, 2015a); and GLAFIC (Oguri, 2010). The use of multiple methods on an individual cluster yields a more accurate estimate of systematic uncertainties than using one method alone, as shown in analyses of Frontier Fields clusters (Coe et al., 2015; Livermore et al., 2016; Acebron et al., 2017; Remolina González et al., 2018).

RELICS has published strong lens modeling analyses of 14 clusters to date (Cerny et al., 2018; Acebron et al., 2018a, b; Cibirka et al., 2018; Paterno-Mahler et al., 2018; Mahler et al., 2018). Most of these analyses have revealed lensing strengths on par with Frontier Fields clusters, quantified in terms of cumulative area with magnification greater than some threshold. Strong lens models are currently available on MAST for 28 / 41 RELICS clusters, with the rest to be delivered in time for the JWST GO call for proposals. Our data products include maps of cluster mass as well as lensing deflection, shear, and magnification.

Paterno-Mahler et al. (2018) used all three methods (primarily Lenstool) to model SPT0615-57 and study the $z\sim 10$ candidate discovered by Salmon et al. (2018). They delivered magnification estimates and explained the current lack of observed counterimages. Acebron et al. (2018b) compared magnification estimates from two methods (Zitrin LTM and Lenstool) for the many (24) $z\sim 6-7$ candidates lensed by CL0152-13. Acebron et al. (2018a) modeled two clusters with the Zitrin LTM method, including MACS0308+26, which lenses one of the brightest $z\sim 6$ candidates known. This paper identified two multiple images of that galaxy lensed to $J\sim 23.2$ and 24.6 AB. Mahler et al. (2018) used spectroscopic redshifts of many arcs from VLT/MUSE (Jauzac et al., 2019) to produce a detailed Lenstool model of MACS0417-11, confirming it is a strong lens, despite curiously yielding no $z\sim 6-8$ candidates (Salmon et al., 2017). Cibirka et al. (2018) modeled four clusters using the LTM method and presented one lensed galaxy showing strong Lyman-$\alpha$ emission with a spectroscopic redshift $z=5.800$. Cerny et al. (2018) modeled five clusters using Lenstool and presented detailed estimates of the statistical and systematic uncertainties. They present new spectroscopic redshifts and a new method to mitigate modeling uncertainties due to photometric redshifts.

Wide field imaging for weak lensing analyses from Subaru and Magellan is in hand for many RELICS clusters, as are X-ray data from Chandra and XMM-Newton, along with Planck SZ mass measurements. Combining these with strong lensing mass measurements from HST, as in Umetsu et al. (2016) for CLASH clusters, will improve both precision and accuracy on the overall mass calibration of massive clusters (§2.3) and contribute to constraints on the dark matter particle cross section (§2.4).

6.3 Supernovae

The RELICS observing strategy (§2.5) yielded 11 supernovae, summarized here for the first time in Table 17. Most of these are cluster members, as expected. The first three supernovae, all found in the RXC0949+17 cluster, were announced in Rodney et al. (2015a). Three of the other RELICS supernovae are lensed, and we obtained follow-up HST imaging of them, using 20 orbits allocated to RELICS for this purpose (Table 6). The most distant, dubbed “Nebra”, is a $z\sim 2$ candidate type Ia lensed by Abell 1763 (Rodney et al., 2016a). If the redshift and type were confirmed, Nebra would be among the most distant SN Ia known. The current record holder at $z=2.22$ is also lensed (Rubin et al., 2018).

7 Summary

With RELICS observations complete, the 34 most massive Planck clusters ($M_{500}>8.8\times 10^{14}M_{\odot}$) now all have Hubble optical and near-infrared imaging as well as Spitzer infrared imaging. Based on this imaging we have discovered 322 $z\sim 6-10$ candidates, including the brightest galaxies known at $z\sim 6$ (Salmon et al., 2017) and the most distant spatially-resolved lensed arc known, SPT0615-JD at $z\sim 10$ (Salmon et al., 2018). These are among the best and brightest targets for detailed follow-up study from the first billion years after the Big Bang. Follow-up observations of RELICS fields are currently being carried out with facilities including Keck, VLT, Subaru, Magellan, MMT, GMRT, and ALMA.

At lower redshifts, RELICS is studying compact, low metallicity dwarf galaxies that are excellent analogs to high-redshift galaxies, but can be studied in greater detail. We have also discovered 11 supernovae (Table 17).

To date, we have published strong lens modeling analyses of 14 RELICS clusters (Cerny et al., 2018; Acebron et al., 2018a, b; Cibirka et al., 2018; Paterno-Mahler et al., 2018; Mahler et al., 2018). Many of the clusters modeled so far have proven to be comparable to the strongest lenses known. By combining our strong lensing analyses of the cluster cores with weak lensing analyses from ground-based imaging covering the full clusters out to their virial radii, we can derive robust mass profiles for these clusters. Our cluster mass measurements will help inform SZ mass scaling relations.

RELICS has proven once again that cluster lensing delivers distant galaxies more efficiently than blank field observations. The efficiency gains are greatest for discoveries of relatively bright galaxies. We expect these gains to be even greater at higher redshifts. Our lens modeling analyses of RELICS clusters will identify which cluster lenses are truly among the best to use going forward to efficiently search for the first galaxies.

RELICS Hubble reduced images, catalogs, and lens models are available via MAST at https://doi.org/10.17909/T9SP45 (catalog doi:10.17909/T9SP45). RELICS Spitzer reduced images are available via IRSA.

We thank Lindsey Bleem for providing Magellan Megacam LDSS3 images of SPT0254-58 and Abell S295 prior to RELICS to inform our HST observations of these clusters. We thank Florian Pacaud and Matthias Klein for discussions regarding our Abell S295 HST pointings. We thank Dale Kocevski for an image of RXC0142+44 obtained with the University of Hawaii 2.2-meter telescope. And we thank Stella Seitz et al. for sharing their HST observations of PLCKG287+32 obtained during the same cycle. We thank the STScI and SSC directors and time allocation committees for enabling these large observing programs. We are grateful to our HST program coordinator William Januszewski for implementing the RELICS HST observations. We thank Jennifer Mack for expert mentoring of our HST image reduction gurus R.A. and S.O. And we thank Gabriel Brammer for providing an updated WFC3/IR hot pixel mask derived from science observations from GO 14114. We are grateful to the University of Arizona for hosting our team meeting. The RELICS Hubble Treasury Program (GO 14096) consists of observations obtained by the NASA/ESA Hubble Space Telescope (HST). Hubble is operated by the Association of Universities for Research in Astronomy, Inc. (AURA), under NASA contract NAS5-26555. Data from the NASA/ESA Hubble Space Telescope presented in this paper were obtained from the Mikulski Archive for Space Telescopes (MAST), operated by the Space Telescope Science Institute (STScI). STScI is operated by the Association of Universities for Research in Astronomy, Inc. (AURA) under NASA contract NAS 5-26555. The Hubble Advanced Camera for Surveys (ACS) was developed under NASA contract NAS 5-32864. Spitzer Space Telescope data presented in this paper were obtained from the NASA/IPAC Infrared Science Archive (IRSA), operated by the Jet Propulsion Laboratory, California Institute of Technology. Spitzer and IRSA are operated by the Jet Propulsion Laboratory, California Institute of Technology under contract with NASA. We gratefully acknowledge support from JPL for the Spitzer analysis. MB and VS also acknowledge support by NASA through grant number 17-ADAP17-0247. Part of this work by WD performed under the auspices of the U.S. DOE by LLNL under Contract DE-AC52-07NA27344. KU acknowledges support from the Ministry of Science and Technology of Taiwan (grant MOST 106-2628-M-001-003-MY3) and from Academia Sinica (grant AS-IA-107-M01). OG is supported by an NSF Astronomy and Astrophysics Fellowship under award AST-1602595. SAR was supported by NASA grant HST-GO-14208 from STScI. which is operated by Associated Universities for Research in Astronomy, Inc. (AURA), under NASA contract NAS 5-26555. AM acknowledges the financial support of the Brazilian funding agency FAPESP (Post-doc fellowship - process number 2014/11806-9). JH was supported by a VILLUM FONDEN Investigator grant (project number 16599).

HST (WFC3, ACS); Spitzer (IRAC); Planck; MAST; IRSA