!

Spectroscopic Exploration at the Cosmic Dawn

Pascal Oesch Geneva Observatory

Exploring the Universe with JWST II P. Oesch, Observatoire UniGE October 2016 The history of astronomy is a history of receding horizons. E. P. Hubble

Epoch of Reionization

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 2 Unprecedented Galaxy Samples at z>=4 (from HST’s blank fields only)

3.5 10500 z~4-10 galaxies 3 (Bouwens, Illingworth, Oesch+15)

2.5

2

1.5 log Numbers 1 6 (z~10) 0.5 6000 (z~4) 3000 (z~5) 900 (z~6) 480 (z~7) 220 (z~8) 0 3 4 5 6 7 8 9 10 11 Redshift

Almost 1000 galaxies in the epoch of reionization at z>6 Current frontier: z~9-10

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 3 Mass and SFR Build-up

Time [Gyr] 10 5 2 1 0.7 0.5 9 Oesch+14 8 50% ]

3 10% − 7 Mpc

1% ⊙ 6 + [M *

ρ 0.1%

log 5

z<4: logM > 8 z>4: MUV < -18

stellar mass density of galaxies 4 from Marchesini+09 from Stark+12

0 2 4 6 8 10 Redshift

Are witnessing the assembly of the first 0.1% of local stellar mass density at z>8. The first two Gyr are a very active epoch of galaxy assembly.

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 4

8Our Oesch Current et al. Spectroscopic Census

Age of the Universe [Gyr] 0.9 0.8 0.7 0.6 0.5 0.4 TABLE 2 −23 Summary of Measurements for GN-z11 Oesch+16 4 R.A. 12 : 36 : 25.46 GN-z11 Dec. +62 : 14 : 31.4 UV −22 +0.08 a (z=7)]

2 * Redshift zgrism 11.09 0.12 UV Luminosity MUV 22.1 0.2 Half Light Radiusb 0.6 0±.3kpc c ± −21 1 log Mgal/M 9.0 0.4 log age/yr c 7.6 ± 0.4 ± 1 SFR 24 10 M yr 0.5 ± AUV < 0.2mag −20 Absolute Magnitude M

d zspec UV Luminosity [L/L UV slope (f ) 2.5 0.2 / ± zphot 0.25 a Age of the Universe at z =11.09 using our cosmol- ogy: 402 Myr −19 b From Holwerda et al. (2015) 6 7 8 9 10 11 c Uncertainties are likely underestimated, since our Redshift photometry only partially covers the rest-frame opti- Fig. 7.— The redshift and UV luminosities of known high- cal for GN-z11 Spectroscopic samples are severaly lagging behind! d See also Wilkins et al. (2016) redshift galaxies from blank field surveys. Dark filled squares corre- spondOnly to a spectroscopicallydozen spectroscopic confirmed confirmations sources, while at smallz>7 to gray date. dots are photometric redshifts (Bouwens et al. 2015b). GN-z11 clearly to estimate how many such galaxies we could have ex- stands out as one of the most luminous currently known galaxies at Montreal, Oct 2016 all redshifts z>6andisbyfarthemostdistantmeasuredgalaxyP. Oesch, Observatoire UniGE 5 pected based on (1) the currently best estimates of the with spectroscopy (black squares; see Oesch et al. 2015b,forafull UV LF at z>8 and (2) based on theoretical models and list of references). Wider area surveys with future near-infrared simulations. telescopes (such as WFIRST) will be required to determine how Our target was found in a search of the GOODS fields, common such luminous sources really are at z>10. which amount to 160 arcmin2. However, in a sub- sequent search of the⇠ three remaining CANDELS fields no similar sources were found with likely redshifts at TABLE 3 Assumed LFs for z 10 11 Number Density Estimates z & 10 (Bouwens et al. 2015a). We therefore use the ⇠ 2 full 750 arcmin of the CANDELS fields with match- 5 Reference /10 M ↵ Nexp ing WFC3/IR and ACS imaging for a volume estimate, ⇤ 3 ⇤ [Mpc ][mag] (< 22.1) which amounts to 1.2 106 Mpc3 (assuming z = 1). ⇥ Bouwens et al. (2015b)1.65-20.97-2.380.06 Using the simple trends in the Schechter parameters of Finkelstein et al. (2015)0.96-20.55-2.900.002 the UV LFs measured UV at lower redshift (z 4 8) Mashian et al. (2016)0.25-21.20-2.200.03 and extrapolating these to z = 11, we can get an⇠ empir- Mason et al. (2015)0.30-21.05-2.610.01 ical estimate of the number density of very bright galax- Trac et al. (2015)5.00-20.18-2.220.002 ies at z 11. This amounts to 0.06 (Bouwens et al. Note.—Theparameters , M ,and↵ represent the three 2015b) or⇠ 0.002 (Finkelstein et al. 2015) expected galax- parameters of the Schechter UV⇤ LF taken⇤ from the di↵erent papers. ies brighter than MUV = 22.1 in our survey correspond- ing to less than 0.3 per surveyed square degree. Simi- larly, recent empirical models (Mashian et al. 2016; Ma- z11. Our 2D data show clear flux longward of 1.47 µm son et al. 2015; Trac et al. 2015) predict only 0.002 0.03 exactly along the trace of the target galaxy⇠ and zero galaxies as bright as GN-z11 in our survey or 0.01 0.2 flux at shorter wavelengths, thanks to our comprehensive per deg2. All the assumed LF parameters together with and accurate treatment of contamination by neighboring the resulting estimates of the number of expected bright galaxies. The interpretation that we indeed detect the galaxies Nexp are listed in Table 3. continuum flux from GN-z11 is supported by the mor- The above estimates illustrate that our discovery of phology of the spectrum, the fact that the counts fall o↵ the unexpectedly luminous galaxy GN-z11 may challenge exactly where the sensitivity of the G141 grism drops, as our current understanding of galaxy build-up at z>8. well as the consistency of the observed counts with the A possible solution is that the UV LF does not follow H-band magnitude of GN-z11 (see e.g. Fig 3). a Schechter function form at the very bright end as has The grism spectrum, combined with the photometric been suggested by some authors at z 7(Bowler et al. constraints, allows us to exclude plausible low-redshift 2014), motivated by inecient feedback⇠ in the very early SEDs for GN-z11 at high confidence. In particular, we universe. However, current evidence for this is still weak can invalidate a low redshift SED of an extreme line emit- (see discussion in Bouwens et al. 2015b). Larger area ter galaxy at z 2 (see section 3 and Fig 4). Instead, studies will be required in the future (such as the planned the grism spectrum⇠ is completely consistent with a very +0.08 WFIRST High Latitude Survey; Spergel et al. 2015)sur- high-redshift solution at zgrism = 11.09 0.12 (see Figures veying several square degrees to determine the bright end 3 and 5). This indicates that this galaxy lies at only of the UV LF to resolve this puzzle. 400 Myr after the Big Bang, extending the previous redshift⇠ record by 150 Myr. 5. SUMMARY GN-z11 is not only⇠ the most distant spectroscopically In this paper we present HST slitless grism spectra measured source, but is likely even more distant than for a uniquely bright z>10 galaxy candidate, which all other high-redshift candidates with photometric red- +0.6 we previously identified in the GOODS-North field, GN- shifts, including MACS0647-JD at zphot = 10.7 0.4 (Coe Spectroscopic Features of High-z Galaxies

68 SHAPLEY ET AL. Vol. 588 to estimate stellar systemic redshifts for individual galaxies, we applied the formulae presented in Adelberger et al. Shapley+03 Lyα is the most promising feature of (2003), which predict a value of the systemic redshift for Lya high-redshift spectra three separate cases: when there is only a Ly emission red- Other emission lines very weak, but shift, when there is only an interstellar absorption redshift, possible to detect (e.g. Stark+15) and when there are both Ly emission and interstellar 10 Steven L. Finkelstein absorption redshifts. While no reference was made to stellar photospheric fea- spectroscopically confirm the redshifts of galaxies se- tures in the estimate of the systemic redshifts of individual lected to be at z>6. Figure 3 highlights the redshift of 8 Lyα galaxies, the rest-frame composite spectrum presented in the most distant spectroscopically-confirmed galaxy as non−Lyα a function of the year of discovery. the next section indicates a mean systemic velocity of The most widely used tool for the measurement of 1 Finkelstein 16 Dv 10 35 km s for the three strongest stellar fea- HeII 6 ¼À Æ À spectroscopic redshifts for distant star-forming galax- tures, which we have identified as C iii 1176, O iv 1343, CIII] ies is the Lyα emission line, with a rest-frame vacuum and S v 1501.4 It is worth noting possible contributions to wavelength of 1215.67 A.˚ While at z< 4, confirma- the O iv 1343 absorption from Si iii 1341 and to the S v 4 tion via interstellar medium absorption lines is possible 1501 absorption from Si iii 1501 if there is a significant B- (e.g., Steidel et al. 1999; Vanzella et al. 2009), the faint star component in the composite spectrum. However, at nature of more distant galaxies renders it nearly im- possible to obtain the signal-to-noise necessary on the Spectroscopic Redshift least the Si iii 1501 contribution will not change the 2 continuum emission to detect such features. Emission Advent of inferred negligible systemic velocity of the feature that we lines are thus necessary, and at z>3, Lyα shifts into have identified as S v 1501. The insignificant velocities of Advent of Sensitive Near-IR MOS Astronomical CCDs the optical, while strong rest-frame optical lines, such the stellar features demonstrate the success of the systemic 0 as [O iii] λλ4959,5007 and Hαλ6563 shift into the mid- 1950 1960 1970 1980 1990 2000 2010 2020 infrared at z>4, where we do not presently have sen- redshift estimates for the LBGs included in composite spec- Year of Discovery tra, at least on average. Also, since the stellar features sitive spectroscopic capabilities. Additionally, Lyα has appear at roughly zero velocity, the redshifts and blueshifts proven to be relatively common amongst star-forming Fig. 2.—Composite rest-frame UV spectrum constructed from 811 indi- Figure 3. The highest redshift spectroscopically confirmed galaxies at z>3. Examining a sample of ∼800 galax- of other sets of spectral features measured relative to the rest vidual LBG spectra. Dominated by the emission from massive O and B galaxy plotted versus the year of discovery. There are cur- ies at z ∼ 3, Shapley et al. (2003) found that 25% con- rently only three galaxies with robust spectroscopic redshifts frame of the composite spectrum should offer a true repre- Montreal,stars, the Oct overall 2016 shape of the UV continuum is modified shortwardP. Oesch, of Observatoire Ly UniGE 6tained strong Lyα emission (defined as a rest-frame EW i at z> 7.5: z =7.51fromFinkelsteinetal.(2013),z =7.73 sentation of the average kinematic properties of the large- by a decrement due to intergalactic H absorption. Several different sets of from Oesch et al. (2015b) and z =8.68fromZitrinetal. > 20 A),˚ while this fraction increases to !50% at z = scale galactic outflows in LBGs. The establishment of the UV features are marked: stellar photospheric and wind, interstellar low- (2015). Data prior to 1999 were taken from the review of 6 (Stark et al. 2011). velocity zero point from the stellar lines in composite spec- and high-ionization absorption, nebular emission from H ii regions, Si ii* Stern & Spinrad (1999), with the references listed therein. Ob- At higher redshifts, Lyα is frequently the only observ- fine-structure emission whose origin is ambiguous, and emission and jects at later times come from Hu et al. (2002); Kodaira et al. able feature in an optical (or near-infrared) spectrum tra represents a significant improvement over the kinematic absorption due to interstellar H i (Ly and Ly ). There are numerous weak (2003); Taniguchi et al. (2005); Iye et al. (2006); Vanzella et al. of a galaxy. While in principle a single line could be information contained in individual rest-frame UV spectra, features that are not marked, as well as several features blueward of Ly (2011); Ono et al. (2012); Shibuya et al. (2012); Finkelsteinetal. where only the strongest interstellar outflow-related that become visible only by averaging over many sight lines through the (2013); Oesch et al. (2015b); Zitrin et al. (2015). The shadedre- a number of possible features, in practice, the nearby gions denote roughly the time when CCDs became widely used, spectral break observed in the photometry (that was features are detected. IGM. [See the electronic edition of the Journal for a color version of this figure.] as well as when MOSFIRE (the first highly-sensitive near-infrared used to select a given galaxy as a candidate) implies multi-object spectrograph) was commissioned on Keck. Major that any spectral line must be in close proximity to such jumps in the most-distant redshift are seen to correspond with ii these technological advancements. a break. This leaves Lyα and [O ] λλ3726,3729 as the likely possibilities (Hα and [O iii], while strong, reside in 4. LBG REST-FRAME UV SPECTROSCOPIC FEATURES 4.1. Stellar Features relatively flat regions of star-forming galaxy continua). sources and true high-redshift galaxies; as discussed in Figure 2 shows a composite spectrum that is the average The C iii 1176, O iv 1343, and S v 1501 stellar photo- While most ground-based spectroscopy is performed at Finkelstein et al. (2015a), this requires imaging at 1 µm high enough resolution to separate the [O ii] doublet, of our entire spectroscopic sample of 811 LBGs, combined spheric lines discussed in 3.1 are marked in Figure 2. Also when working at z = 6–8 (see also Tilvi et al. 2013, for x the relative strength of the two lines can vary depend- in the manner described in 3.5 Rest-frame UV spectra of note (although not marked) is the large number of weak a discussion of the utility of medium bands). Using a x ˚ ing on the physical conditions in the ISM, thus it is of LBGs are dominated by the emission from O and B stars absorption features between 1400 and 1500 A.These combination of object colors and spatial extent, it is possible only a single line could be observed. True Lyα with masses higher than 10 M and T 25; 000 K. The include blends of Fe v, Si ii,Siiii, and C iii photospheric likely that space-based studies are relatively free of stel- lines are frequently observed to be asymmetric (e.g., overall shape of the UV spectrum is modified by dust extinc- absorption lines from O and B stars (Bruhweiler, Kondo, & lar contamination. This may be more of a problem with Rhoads et al. 2003), with a sharp cutoff on the blue side McClusky 1981; de Mello, Leitherer, & Heckman 2000). In ground-based studies, though with excellent seeing even and an extended red wing, due to a combination of scat- tion internal to the galaxy and, at rest wavelengths shorter bright z>6 galaxies can be resolved from the ground than 1216 A˚ , by intergalactic H i absorption along the line addition to photospheric absorption features, the spectra of tering and absorption within the galaxy (amplified due (e.g., Bowler et al. 2014). Future surveys must be cog- to outflows), and absorption via the IGM. An observa- of sight. Composite spectra contain the average of many dif- the most massive hot stars indicate the presence of stellar nizant of the possibility of stellar contamination, and winds of 2000–3000 km s 1 due to radiation pressure tion of a single, asymmetric line is therefore an unam- ferent lines of sight through the intergalactic medium À choose their filter set wisely to enable rejection of such biguous signature of Lyα. However, measurement of line (IGM). Therefore, spectral features that are intrinsic to the (Groenewegen, Lamers, & Pauldrach 1989). These wind fea- contaminants. asymmetry is only possible with signal-to-noise ratios galaxy at wavelengths shorter than Ly ,andwhichcanbe tures appear as broad blueshifted absorption for weaker of >10, which is not common amongst such distant ob- winds or as a P Cygni–type profile if the wind density is high jects (e.g., Finkelstein et al. 2013). Lacking an obvious completely wiped out by individual Ly forest systems 4Spectroscopyofz>6Galaxies along a specific line of sight, become visible in the composite enough (Leitherer, Robert, & Heckman 1995). The most asymmetry, other characteristics need to be considered. spectra. While we regain spectroscopic information by aver- prominent stellar wind features are N v 1238, 1242, Si iv While photometric selection is estimated to have a rel- For example, for very bright galaxies, the sheer strength 1393, 1402, C iv 1548, 1550, and He ii 1640. The atively low contamination rate, it is imperative to fol- of the Lyman break can rule out [O ii] as a possibility, aging over many different sight lines, we still, however, see ˚ ii shape of the N v wind profile, especially the absorption lowup a representative fraction of a high-redshift galaxy as the 4000 A break (which would accompany an [O ] the average decrement of the Ly forest, DA.Inthefollow- sample with spectroscopy, to both measure the true red- line) is more gradual (see discussion in Finkelstein et al. ing section, we describe the spectroscopic features contained component, is affected by its close proximity to the Ly shift distribution, as well as to empirically weed out 2013). For fainter galaxies, with a weaker Lyman break, in the composite spectrum of Figure 2, which trace the region of the spectrum and is therefore difficult to character- contaminants. In this section, I discuss recent efforts to and no detectable asymmetry, a robust identification of photospheres and winds of massive stars, neutral and ion- ize in detail, although we do see both emission and absorp- PASA (2015) ized gas associated with large-scale outflows, and ionized tion qualitatively consistent with a P Cygni–type profile. doi:10.1017/pas.2015.xxx gas in H ii regions where star formation is taking place. While clear of the large-scale continuum effects of Ly ,the Si iv and C iv transitions contain a combination of stellar wind and photospheric absorption plus a strong interstellar 4 The precise wavelengths are C iii 1175.71, O iv 1343.35, which is a blend of lines at  1342:99 and  1343:51 A˚ ,andSv 1501.76. absorption component, which are difficult to disentangle. 5 The composite¼ LBG spectrum¼ is available in electronic form from The stellar wind feature only becomes apparent in Si iv for http://www.astro.caltech.edu/~aes/lbgspec. blue giant and supergiant stars, while, in contrast, the C iv Vol 443|14 September 2006|doi:10.1038/nature05104 LETTERS

A galaxy at a redshift z 5 6.96 Masanori Iye1,2,3, Kazuaki Ota2, Nobunari Kashikawa1, Hisanori Furusawa4, Tetsuya Hashimoto2, Takashi Hattori4, Yuichi Matsuda5, Tomoki Morokuma6, Masami Ouchi7 & Kazuhiro Shimasaku2

When galaxy formation started in the history of the Universe NB973. We considered these to be z 7 LAE candidates (hereafter remains unclear. Studies of the cosmic microwave background referred to as IOK-1, the brighter of the¼ two in NB973, and IOK-2). indicate that the Universe, after initial cooling (following the Big Their images and photometric properties, as well as the candidate Bang), was reheated and reionized by hot stars in newborn selection criteria, are shown in Fig. 1 and Table 1. galaxies at a redshift in the range 6 < z < 14 (ref. 1). Though To confirm whether these candidates are real LAEs, we carried out several candidate galaxies at redshift z > 7 have been identified follow-up spectroscopy on 14 and 15 May, 1 June 2005, and 24 April photometrically2,3, galaxies with spectroscopically confirmed red- 2006, using the Faint Object Camera and Spectrograph (FOCAS)17. shifts have been confined to z < 6.6 (refs 4–8). Here we report a Total exposures were 8.5 h and 3.0 h for IOK-1 and IOK-2, respect- spectroscopic redshift of z 5 6.96 (corresponding to just 750 Myr ively. Figure 2 shows the spectrum of IOK-1 with OH skylines after the Big Bang) for a galaxy whose spectrum clearly shows subtracted. We identified this object as a LAE at z 6.96, which Lyman-a emission at 9,682 A˚ , indicating active star formation at makes it, to our knowledge, the most distant galaxy¼ ever spectro- 21 a rate of ,10M ( yr , where M ( is the mass of the Sun. This scopically confirmed. Our discovery of this galaxy provides direct demonstrates that galaxy formation was under way when the evidence that at least one galaxy already existed only ,750 million Universe was only ,6 per cent of its present age. The number years after the Big Bang. density of galaxies at z < 7 seems to be only 18–36 per cent of the Three-hour integration on IOK-2 did not reveal significant spec- density at z 5 6.6. tral features, and our photometric detection could have been due to 6 2 Narrowband filter surveys searching for redshifted Lyman-a noise. There are 10 independent 2 00 apertures in the 876 arcmin emitting galaxies (LAEs) through dark atmospheric windows in field, and the probability of noise exceeding 5j for a gaussian which the foreground telluric OH bandLong-Term emission is not prohibitively Spectroscopicdistribution is 3 1027, so that the Redshift expected number of Holder 5j noise bright have been successful in isolating galaxies at high redshift4,5. £ Some of the discovered LAEs are gravitationally lensed objects behind clusters of galaxies. To obtain unbiased information about the population of LAEs, systematic surveys at various redshifts have 6–8 been performed, for example, in the Subaru Deep Field (SDF) . NATURE|Vol 443|14 September 2006 LETTERS Such studies have been limited to a redshift range of z , 6.6. New attempts are in progress9–12 to search for galaxy populations at 7 , z , 10, using deep narrowband imaging observations with near-infrared cameras, aiming for high sensitivity despite the limited survey volume. We have taken another approach to survey z 7 LAEs, specifically using the last OH window accessible with a wide-¼ field charge-coupled device (CCD) camera. Figure 1 | Multi-waveband 20 00 3 20 00 images of the z 5 6.96 Lyman a We developed a narrowband filter NB973 for the Subaru Suprime- emitter IOK-1 and the unidentified candidate IOK-2. Deep broadband 13,14 ˚ ˚ images (Kron–Cousin B, V, R bands, and SDSS i 0 and z 0 bands, with limiting Cam , with a transmission bandwidth of 200 A centred at 9,755 A. Iye+06 To investigate the population of z 7.0 LAEs, we conducted an magnitudes 28.45, 27.74, 27.80, 27.43 and 26.62 in 3j, respectively) were ¼ taken during 2002–2004 (ref. 8). Dithered NB973 exposures (15–30 min NB973 imaging survey with 15 h exposure (NB973 , 24.9 mag in each; 15 h total) were combined to remove fringing produced by foreground 5j) on the photometric nights of 16 and 17 March 2005. An effective sky OH emission lines and yielded a 2 00 aperture limiting magnitude of 2 area of 876 arcmin and co-moving depth of 58 Mpc, corresponding NB973 24.9 (5j). All the images were convolved to a common seeing size 5 3 ¼ to z 6.94–7.11, led to a total probed volume of 3.2 10 Mpc . of 0.98 00 . Of the 41,533 objects detected down to NB973 24.9 (total ¼ £ ¼ (Throughout, we adopt a cosmology with Q M 0.3, QL 0.7, and magnitude in this case), we selected those satisfying the following criteria: 21 21 ¼ ¼ 0 0 H 0 70 km s Mpc and refer to magnitudes in the AB system (1) B, V, R, i ,z , 3j, for z 7.0 LAE candidates IOK-1 and IOK-2, and (2) ¼ 0 0 0 ¼ measured with a 2 00 aperture unless otherwise specified.) We B, V, R ,3j,i–z .1.3, z -NB973 .1.0, for z 6.4–7.0 LAE candidate ¼ 15,16 measured the NB973 fluxes of all detected objects and also measured IOK-3. These criteria were determined by model calculations . The i 0 –z 0 Figure 3 | Decline of the number density of LAEs between 6.6 < z < 7.0. colour eliminates interlopers such as lower-z [O II], [O III] line-emitting The number density n (a), luminosity density r and star-formation-rate their broadband fluxes using images previously obtained by the SDF LLya galaxies while the colour z 0 -NB973 picks up objects with NB973 flux excess project8. The colours of objects were derived from these measure- density SFRD (b) derived from Lyman-a line fluxes detected in several with respect to the z 0 continuum flux. In addition, null (,3j) detection in B, Suprime-Cam LAE surveys are plotted for the three epochs of z 5.7 ments and compared to the expected colours of a z 7.0 LAE, 6 18 7 ¼ Vand R (or B, V, R, i 0 and z 0 ) total magnitudes are required since continuum (square and triangle : other survey fields), 6.6 (open circle : also SDF), ¼15 computed using a stellar population synthesis model and an fluxes blueward of the Lyman-a line should be mostly (or completely) and 7.0 (filled circles: our result) down to our detection limit 16 42 21 intergalactic absorption model . Of the 41,533 objects detected in L Lya 9.9 10 erg s calculated from NB973 24.9 (5j). The absorbed by the neutral IGM. The criterion-(2) object, IOK-3, was identified ¼ £ ¼ 22 volumes probed by these surveys are of a similar order of magnitude NB973, we found only two objects with significant NB973 flux that as a z 6.6 LAE by an independent spectroscopic survey and the 5 3 were not detected in any of the broadband images blueward of possibility¼ of a z 7.0 LAE was excluded. (,2–3 10 Mpc ). The large filled circle at z 7.0 represents the values ¼ derived£ for the single LAE discovered in the present¼ study, and the small 1 2 filled circle those for two LAEs, including the unconfirmed candidate National Astronomical Observatory, 2-21-1 Osawa, Mitaka, Tokyo 181-8588, Japan. DepartmentFigure of Astronomy, 2 | Combined University spectrum of Tokyo, of z 5 Hongo,6.96 galaxy, Bunkyo-ku, IOK-1. TokyoThe 113-0033, bottom panel Japan. 3 4 5 IOK-2. All the vertical error bars in both panels include the errors from The Graduate University for Advanced Studies,Montreal, Mitaka, Oct Tokyo 2016 181-8588, Japan. Subaru Telescope,shows 650 the removedNorth A’ohoku OH sky Place,P. Oesch, emission Hilo, Observatoire Hawaii lines. 96720, An UniGE echelle USA. grismDepartment (175 lines of per 7 6 7 both poissonian statistics for small numbers and cosmic variance. The Astronomy, Graduate School of Science, Kyoto University, Kyoto 606-8502, Japan. The Institutemillimetre, of Astronomy, resolution University< 1,600) of Tokyo, with Mitaka, z 0 filter Tokyo and 0.8 181-0015,00 slit was Japan. used toSpace obtain horizontal error bars indicate the surveyed redshift ranges. Note that Telescope Science Institute, 3700 San Martin Drive, Baltimore, Maryland 21218, USA. 11 spectra of 30-min exposure each, dithered along the slit by ^1 00 . The large and small filled circles are shifted slightly horizontally from z 7.0 Lyman-a emission peak is located at 9,682 A˚ . This feature was confirmed to 186 to make them easier to see. ¼ follow the dithered shifts along the slit. The Lyman-a emission has a total flux of 2.0 10217erg s21 cm22, consistent with the estimate from the NB973 image£ (2.7 10217erg s21 cm22), and a full width at half- ˚ £ 43 21 maximum of 13 A. The Lyman-a luminosity is L Lya 1.1 10 erg s , Those transient objects, if any existed, would not show conspicuous 21 ¼ £ corresponding to SFRLya 10 M ( yr . The emission feature is significant emission line features in the NB973 waveband. at the 5.5j level. The asymmetric¼ emission line profile matches the We note, however, that even for the brighter IOK-1, it was necessary composite template line profile (dashed line) produced from 12 Lyman-a to combine the eleven 30-min exposures with the highest signal-to- emitters at z 6.6 (ref. 22) normalized and shifted to z 6.96. This noise ratio under the best sky conditions selected from the 17 spectra ¼ ¼ ˚ emission cannot be unresolved z 1.6 [O II] doublet lines with 7.0 A taken to detect the z 6.96 Lyman-a emission shown in Fig. 2 clearly. separation (3,726 A˚ , 3,729 A˚ at rest¼ frame), as finer sky lines (about Because our spectral¼ detection limit for 3 h did not reach the depth 5.1–6.2 A˚ ) are clearly resolved. Similarly, the line is not Hb,[OIII] 4,959 A˚ , a [O III] 5,007 A˚ ,Ha,[SII] 6,717 A˚ or [S II] 6,731 A˚ , because our spectrum required to deny the presence of weak Lyman emission, we retain shows none of the other lines that should accompany any of these lines in the IOK-2 as another possible but unconfirmed LAE. Therefore, we conclude that the detected number of LAEs at z 7.0 in our survey observed wavelength range. The possibility that OH lines mask other lines ¼ was carefully examined for each case and ruled out. There appears to be an volume is at least one, and possibly could be two at most. extremely weak emission at around z 7.02 in the 3 h integration spectrum The observed LAE number density is expected to decline beyond ¼ of IOK-2, but this needs further integration for confirmation. the redshift at which the reionization had completed, because the increasing fraction of intergalactic neutral hydrogen absorbs the Lyman-a photons from young galaxies. However, no significant events is 0.3. Another possible origin for IOK-2 is a transient object decline of LAE number densities has been yet observed for such as a supernova or active galactic nucleus that happened to 3 , z , 6.6 (ref. 26). Figure 3 shows the volume densities of the brighten in March 2005, but was inconspicuous in broadband images LAE numbers, Lyman-a line luminosity, and star formation rate at 6,7,18 taken 1–2 years earlier. Counts of variable objects in i 0 band images three redshifts down to our detection limit. These physical from the SDF taken over 1–2 years indicate that the expected number parameters are useful for tracing the properties of LAEs in an of variable objects having brightness amplifications large enough to environment with an increasing fraction of neutral hydrogen in the exceed our detection limit (NB973 24.9) is at most about three. intergalactic medium (IGM). Errors shown in the figure include (1) ¼

Table 1 | Properties of z < 7 Lyman-a emitter candidates 43 21 21 ID Position (J2000) i 0 (mag) z 0 (mag) NB973 (2 00 )NB973(total)LLya (10 erg s ) SFRLya (M( yr ) IOK-1 a 13 h 23 min 59.8 s, .27.84 .27.04 24.60 24.40 1.1 ^ 0.2 10 ^ 2 ¼ d 278 24 0 55.8 00 IOK-2 a¼þ13 h 25 min 32.9 s, .27.84 .27.04 25.51 24.74 ,1.1 ^ 0.2 ,10 ^ 2 ¼ d 278 31 0 44.7 00 ¼þ Magnitudes (SDSS i 0 ,z0 filters and NB973) are in the AB system, measured with a 2 00 aperture. The NB973 total magnitudes integrated over the entire image are also given. The Lyman-a luminosity LLya of IOK-1 was derived from spectroscopy, while the upper limit of LLya for IOK-2 was evaluated using NB973 total magnitude by assuming that 68% of the NB973 flux is in the 27 Lyman-a line, on the basis of simulations. Luminosities were then converted into corresponding Lyman-a star-formation rate SFRLya using the relation derived from Kennicutt’s equation with case B recombination theory. 1j errors are also given for L Lya and SFRLya IOK-1 was spectroscopically identified as a z 6.96 Lyman-a emitter. For IOK-2, the photometric estimation of Lyman-a luminosity should be regarded as an upper limit because we were unable to make a spectroscopic confirmation.¼ 187 Lyman Alpha in the Reionization Epoch

Spectroscopic confirmation of sources in the reionization epoch has proven very difficult, even with new efficient NIR spectrographs. 6 Schenker et al.

Schenker et al. 2011 Treu et al. 2013

EW0 > 25 Å in emission α

+ fraction of LBGs withfractionLBGs Ly of

Fig. 5.— The redshift-dependent fraction of color-selected Ly- Fraction of Lymanα line break emitters galaxies that drops reveal Ly αacrossin emission, z=6.5X(Lyα to), adjusted z=7: as discussed in the text to approximate one within a similar lumi- imprintnosity range of with cosmic a rest-frame reionization? EW in excess of 25 A.˚ Data points Fig. 4.— The expected number of detected Lyα emission lines for the galaxies with −21.75 < MUV < −20.25 are displaced by with greater than or equal to 5σ significance in the combined Keck +0.1inredshiftforclarity.Dataover46.3arederivedfrom lihood function for 10,000 Monte Carlo realizations assuming the the present paper, including sources discussed by Fontana etal. intrinsic line emission propertiesMontreal, Oct follow 2016 the luminosity dependence (2010). TheP. curves Oesch, shownObservatoire represent UniGE the aggregate redshift proba- 8 ′ seen in our 5.5 extinction averaged over the entire population. dict what global neutral hydrogen fraction, XHI would To compute the most likely value of f,weundertake be required to account for this decline, we find XHI ≃ Monte Carlo simulations using the previously described 0.44, and XHI ≃ 0.51, respectively. The models of EW distributions, but with f now added as a free param- Dijkstra et al. (2011), which provide a more comprehen- eter. We vary f from 0 to 1 in steps of 0.01, and compute sive treatment of Lyα radiative transfer through out- N=1000 simulations for each step. We can then calcu- flows, result in an increased value for XHI in both cases. late the probability distribution for f given our Nobs=2 confirmed sources using Bayes’ theorem: 4. DISCUSSION p(N =2|f)p(f) obs Although we consider the most likely explanation for p(f|Nobs =2)= 1 (1) !0 p(Nobs =2|f)df our observed decrease in the number of LBGs which show observable Lyα emission to be an increase with redshift Here, p(f) is the prior probability for f, which we take in the neutral fraction of the IGM, it is important to re- to be uniform for 0 ≤ f ≤ 1, and p(Nobs =2|f) is the member our assumptions. Foremost we have assumed probability, drawn from our Monte Carlo simulations, that all of our 26 targets have true redshifts beyond that we would find Nobs = 2 sources for a given value z ≃ 6.3. Should there be low redshift interlopers or of f. Assuming that the intrinsic EW distribution for Galactic stars in our new sample, we will overestimate our observed sources is that of Paper II at z =6,we the decline in the Lyα fraction. Secondly, we have as- find f =0.45 ± 0.20, while using the z = 7 extrapolated sumed the DEIMOS spectra from Paper II constitute a +0.24 distribution yields f =0.34−0.15. In the Figure 5, we representative sample for calculating the expected EW plot the value of X(Lyα) in the same luminosity bins of distribution for 6.3 6.3 data from our analysis. Our Monte Carlo simulations are (Bouwens et al. 2010a; Dunlop et al. 2011), we consider 4 Oesch et al.

wavelength as expected by our photometric redshift es- −23 timate z =7.7 0.3. We therefore interpret this line phot −22.5 Previous LBG+LAEs as Ly↵, but we discuss± other possibilities in section 4.3. EGS-zs8-1 We fit the line using a Markov Chain Monte Carlo −22 (MCMC) approach based on the emcee python library (Foreman-Mackey et al. 2013). Our model is based on a −21.5 Bright z~8 Galaxies with SpectroscopicUV Redshifts truncated Gaussian profile to account for the IGM ab- UV −21 M sorption and includes the appropriate instrumental res- M M* (UV LF) olution. The model also includes the uncertainty on the −20.5 background continuum level. The MCMC output pro- vides full posterior PDFs and uncertainties for the red- −20 shift,F606W line flux,F814W significance,F105W F125W and lineF160W width. [3.6] [4.5] −19.5 The line corresponds to a redshift of zLy↵ =7.7302 Small sample of four IRAC excess 43± −19 0.0006, with a total luminosity of LLy↵ =1.2 0.2 10 6 sources6.5 from CANDELS/WIDE7 7.5 8 1 ± ⇥ erg s , and a total detection significance of 6.1.This (see SpectroscopicRoberts-Borsani+15)Redshift Redshift is23 somewhat lower, but consistent with a simple estimate 5 4 pixel binned S/N of 7.2 detection significance from2 integrating the 1D ] Night 1 EGS-zs8-1 zbest=7.7 (χ = 1.7) 1 4 − Mask Layout extracted pixel flux over the full2 extent of the line (i.e., 3 24 zlow=1.8 (χ = 34.2) 0.5 z~8 Galaxies 2 not accounting for background continuum o↵sets). mag CANDELS+BoRG 2 Note that the redshift of the line is determined from − (BouwensEGS-zs8-1 et al. 2014) 1

25 0.4 binned S/N our model of a truncated Gaussian profile and is thus 0 corrected for instrumental resolution and the asymme- −1 AB 26 1 0.3 Night 2 −2 m try arising from the IGM absorption. The peak of the (2 hrs) 1.0555 1.06 1.065 1.07 zphot=7.7±0.3 Observed Wavelength [µNightm] 2 observed line ( = 10616 A)˚ thus lies 2.5 A˚ to the red 4 of27 the actual determined redshift.0.5 ⇠ 0.2 EGS−zs8−1 3 Given the brightness of the target galaxy, the detected 2 normalized p(z) 0.1 Night 1 N 1 line28 corresponds to a rest-frame0 equivalent width EW0 = 0 2.5 5 7.5 10 12.5 (2 hrs) binned S/N 0 ˚ Redshift Surface Density [arcmin LBG candidates 21 4 A. This is less than the strong Ly↵ emitter criterion −1 29± ˚ 0 4 arcsec of0.5 EW0 > 25 A set1.0 in recent analyses2.0 that use4.0 the Ly↵ 6.0 24.5 25 25.5 26 26.5 27E 27.5 −2 28 H AB mag 1.055 1.06 1.065 1.07 fraction among Lymanobserved Break wavelen galaxiesgth [µm] to constrain the 160 Observed Wavelength [µm] reionization process (e.g. Stark et al. 2011; Treu et al. Fig. 4.— Top – UV absolute magnitudes of spectroscopically 2013). confirmed Lyman break galaxies and Ly↵ emitters in the cosmic reionization epoch, at z>6. Our target, EGS-zs8-1 (red square), Spitzer/IRAC4.2. Line Properties colors allow us to exploitrepresents very wide the highest-redshift area imaging source and data is the brightest galax- ies currently confirmed. For reference, the gray dashed line shows Di↵erent quantitiesto search of the detectedfor rare, line ultra-luminous are tabulated z~8the evolution galaxy of thecandidates characteristic magnitudewith M of the UV LF ⇤ in Table 1. In particular, we compute therobust weighted photometric skew- (Bouwens redshifts et al. 2014). The galaxies shown as black squares are as- ness parameter, Sw, as outlined in Kajisawa et al. (2006). sembled from a compilation from Jiang et al. (2013b); Finkelstein et al. (2013); Shibuya et al. (2012); Ono et al. (2012), and Vanzella For this emission line, we measure Sw = 15 6 A,˚ which et al. (2011). Bottom – Surface density of the full sample of z 8 puts the line well above the 3 A˚ limit found± for emission galaxies in the combined CANDELS and BoRG/HIPPIES fields⇠ Montreal,lines Oct at2016 lower redshift (see also section 4.3). P. Oesch, Observatoire(Bouwens UniGE et al. 2014,grayhistogram).EGS-zs8-1isthebrightest9 and also one of the most massive sources at these redshifts. Note The full-width-at-half-maximum (FWHM) of the line that all three z 8candidateswithH 25.0mag( 0.5mag is quite broad with FWHM = 13 3 A,˚ corresponding brighter than the⇠ rest) are identified in the⇠ CANDELS/WIDE⇠ sur- ± +89 1 vey area where ancillary Y105 imaging is generally not available. to a velocity width of VFWHM = 376 70 km s .Our Our spectroscopic confirmation is thus especially valuable. galaxy thus lies at the high end of the observed line width distribution for z 5.7 6.6 Ly↵ emitters (e.g. Ouchi ⇠ the redshift of this galaxy would be zOII =1.85. This et al. 2010), but is consistent with previous z>7 Ly↵ is very close to the best low redshift SED fit shown in lines (Ono et al. 2012). Figure 1. However, that SED requires a strong spectral break caused by an old stellar population, for which no 4.3. Caveats emission line would be expected. Additionally, the low- While the identification of the detected asymmetric redshift solution can not explain the extremely red IRAC emission line as Ly↵ is in excellent agreement with the color (used to select this galaxy), and predicts significant expectation from the photometric redshift, we can not detections in the ACS/F814W band, as well as in the rule out other potential identifications. As pointed out ground-based WIRDS K-band image (Bielby et al. 2012). in the previous section, Kajisawa et al. (2006) find that No such detections are present, however, resulting in a 7 weighted asymmetries S > 3 A˚ are not seen in lower likelihood for such an SED of < 10 (see also Fig 1). w L redshift lines, but almost exclusively in Ly↵ of high- Thus all the evidence points to this line being Ly↵ at redshift galaxies. However, at the resolution of our spec- z =7.73. tra, the observed asymmetry is also consistent with an O II line doublet in a high electron density environment, 5. DISCUSSION i.e., with a ratio of [O II]3727 / [O II]3729 > 2, and In this Letter we used Keck/MOSFIRE to spectro- 1 with a velocity dispersion of v & 100 km s . scopically confirm the redshift of one of the brightest If the observed line is an [O II]3727, 3729 doublet, z 8 galaxies identified by Bouwens et al. (2014)over ⇠ 4 Stark et al.

Source zLy↵ zphot RA DEC Date of Observations H160 Filters UV lines targeted Ref

+0.36 EGS-zs8-1 7.730 7.92 0.36 14:20:34.89 +53:00:15.4 12-15 Apr 2015 25.0 H CIII] [1], [2] ...... ...... 11 June 2015 . . . H CIII] [1], [2] +0.26 ↵ EGS-zs8-2 7.477 7.61 0.25 14:20:12.09 +53:00:27.0 12-15 Apr 2015 25.1 Y, H Ly , CIII] [1] Bright. . . z~8 . . .Galaxies . . . . . .with . . .Spectroscopic 11 June 2015 . . . HRedshifts CIII] [1] +0.12 COS-zs7-1 7.154 7.14 0.12 10:00:23.76 +02:20:37.0 30 Nov 2015 25.1 Y Ly↵ [1] Table100% 1. Galaxies spectroscopic targeted with Keck/MOSFIRE success spectroscopic rate via observations. Lyα detection The final column in provides luminous the reference galaxies! to the article where each galaxy was first discussed in the literature. The photometric redshifts shown in column three are taken from the discovery papers. References: [1] Roberts-Borsani et al. (2016); [2] Oeschsee: et Roberts-Borsani+15 al. (2015) (z=7.48), Zitrin+15 (z=8.68), Stark+16 (z=7.15)

1.05 1.055 1.06 1.065 1.07 1.075 Stark+16

0.2 EGS-zs8-12.0 EGS−zs8−1 Oesch+15 0.8

) Ly [CIII] CIII] ) 1 1 − ] zLy=7.733 − Å Å

Å 0.6 /Å] / 1 1.5 1 2 2 −

0.1 − s s 2 2 − 0.4 1.0 −

erg/s/cm 0.2 erg/s/cm 0 0.5 17 erg cm − erg cm 18 − 18 0.0

0.0 −

−0.1 (10

(10 −0.2

z=7.7302 0.0006 λ

F ± Flux [10 −0.5 −17 2 F f(Ly) = 1.7±0.3 10 erg/s/cm −1.0 −0.4 −0.2 1.05 1.055 1.0551.06 1.0601.065 1.0651.07 1.0701.075 1.650 1.655 1.660 1.665 1.670 1.675 ObservedObserved Wavelength Wavelength [µm] (µm) Observed Wavelength (µm)

Figure 1. Keck/MOSFIRE spectraEGS-zs8-1 of EGS-zs8-1, now a z =7 has.733 galaxya three that was line originally redshift spectroscopically z=7.73. confirmed in Oesch et al. (2015). (Left:) Two- dimensional and one-dimensional Y-band spectra centered on the Ly↵ emission line. Data are from Oesch et al. (2015). (Right:) H-band observations showing detection of the [CIII], CIII] 1907,1909Very high doublet. EW The top CIII] panels emission show the two dimensional (W0=22+-2 SNR maps Å (black), is positive), and the bottom panel shows the flux calibrated one-dimensionalimplies extractions. strong radiation field and low metallicity

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 10 3.2 EGS-zs8-2 is situated. However the separation of the individual components of the doublet is such that at least one of the two lines must be EGS-zs8-2 is another bright (H160=25.1) galaxy identified in CAN- located in a clean region of the spectrum. We estimate 3 upper DELS imaging by RB16. The IRAC color of EGS-zs8-2 ([3.6]- 18 2 1 limits of 2.3 10 erg cm s for individual components. The [4.5]=0.96 0.17) is redder than EGS-zs8-1, likely reflecting ⇥ ± non-detection suggests that the total flux in the CIII] doublet must yet more extreme optical line emission. We estimate a rest-frame be less than 62% of the observed Ly↵ flux, fully consistent with [OIII]+H equivalent width of 1610 302 A˚ is required to re- ± the ratio observed in EGS-zs8-1 and in extreme CIII] emitters at produce the flux excess in the [4.5] filter. A 4.7 emission feature lower redshift. We place a 3 upper limit on the doublet rest-frame was identified by Roberts-Borsani et al. (2016) at a wavelength of equivalent width of <14 A.˚ Deeper data may yet detect CIII] in 1.031µm. RB16 tentatively interpret this feature as Ly↵. EGS-zs8-2. We obtained a Y-band spectrum of EGS-zs8-2 with the goal of verifying the putative Ly↵ detection. The spectrum we obtained shows a 7.4 emission line at 1.0305 µm (Figure 2a), confirming 3.3 COS-zs7-1 that EGS-zs8-2 is indeed a Ly↵ emitter at zLy↵ =7.477. The mea- 18 2 1 sured line flux (7.4 1.0 10 erg cm s ) is less than half Prior to this paper, COS-zs7-1 was the only source from Roberts- ± ⇥ that of EGS-zs8-1. We calculate the Ly↵ equivalent width using Borsani et al. (2016) lacking a near-infrared spectrum. Similar to the broadband SED to estimate the underlying continuum flux. The the other galaxies from RB16, COS-zs7-1 is bright in the near- resulting value (WLy↵=9.3 1.4 A)˚ is the smallest of the RB16 infrared (H160=25.1) and has IRAC color ([3.6]-[4.5]=1.03 0.15) ± ± galaxies. that indicates intense optical line emission. In addition to RB16, the The MOSFIRE H-band spectrum covers 14587 to 17914 A,˚ galaxy has been reported elsehwere (e.g., Tilvi et al. 2013; Bowler corresponding to rest-frame wavelengths between 1720 and 2113 A˚ et al. 2014). We estimate an [OIII]+H rest-frame equivalent width for EGS-zs8-2. In Figure 2b, we show the spectral window centered of 1854 325 A˚ based on the [4.5] flux excess, making COS- ± on the [CIII], CIII] doublet. No emission lines are visible. There are zs7-1 the most extreme optical line emitter in the RB16 sample. two weak sky lines in the wavelength range over which the doublet RB16 derive a reasonably well-constrained photometric redshift

c 2016 RAS, MNRAS 000, 1–13 –5–

Different Way Forward: Continuum Break Redshifts

If Lyα disappears, need different technique to measure redshifts: 18 continuum breaks (as done for QSOs)

Note: at z>6 these are the Lyα continuum breaks Fig. 1.— Direct imaging of PEARS-N-101687 from the GOODS survey (Giavalisco et al.

2004). The object is e↵ectively undetected in the B450, V606,andi775 bands (left three panels), while it is clearly seen in the z850 band (right panel) with magnitude z850 =26.16 (AB). Each panel is 1.500 1.500 in size. ⇥ Tanvir+09 Rhoads+13 zLyα+Ly-break = 6.573 zLy-break = 8.2

Fig. 2.— Upper: The 2D PEARS spectrum of PEARS-N-101687, displayed in inverse video

(dark = more flux), with all three position angles coadded and with a 2 pixel (0.100 80A)˚ ⇥ FWHM smoothing applied. A short segment of continuum is evident, with a blue edge near ˚ 9230AduetotheLyman-↵ forest break, and a red cuto↵ imposed by the fallo↵ of instrumen- 2 1 1 tal eciency. Lower: A 1D extraction of the PEARS spectrum, in ergscm s A˚ . The thin black curve is a best-fittingProblem continuum: +the IGM background absorption model with in redshift thez NIR=6.6 is very high from the ground and a continuum level of AB =25.7magontheredsideoftheLyman-and faintness of↵ galaxiesbreak.Figure The 2. blue compared to QSOs curve is a model based on the line flux from the Keck spectrum, and the continuum flux level andMontreal, line spreadOct 2016 function width expected based on HST imaging. P. Oesch, Observatoire UniGE 11 CANDELS/GOODS-North

GN-z10-1 H160=25.95 zphot = 10.2 ± 0.4

HST stamp

very bright z~10 sample from Oesch+14 is within reach of the WFC3/IR grism IRAC detected Montreal, Oct 2016 P. Oesch, Observatoire UniGE 12 Neighbor Contamination in Grism Spectra

Even in a blank field, it’s difficult to identify orientations with minimal contamination. PreviousAn HST AGHAST Grism Redshift spectra at heavilyz = 11 contaminated. 3 An HST Grism Redshift at z = 11 3 Contamination Model WFC3/IR Grism Observation Setup 1 1

TraceTrace0 0 ofTrace GN ofTrace− ofGNz11 GN−− ofz11 GN−z11 Epoch 2 −1 Epoch 1 −1 Epoch 1

1 GN-z11 AGHAST 0 1 −1 Scale [arcsec] 0

−1 Scale [arcsec] 1 Epoch 2

0

−1 1 N 1.1 1.2 1.3 1.4 1.5 1.6 1.7 0 Observed Wavelength [µm] 5’’ Fig. 2.— Our model of the contaminating flux from neighbor- E ing− sources1 AGHAST in the slitless grism spectra around the trace of our source1.1 (i.e., the1.2 line along1.3 which we1.4 expect its1.5 flux; indicated1.6 by1.7 red lines). From1.1 topObserved to bottom,1.2 Wavelength the1.3 panels show 1.4[µm] the final1.5 model 1.6 1.7 Fig.Oesch+16 1.— CANDELS H-band image around the location of our contamination in our spectra in epochs 1 and 2 and in the pre- target source GN-z11. The arrows and dashed lines indicate the existing AGHAST spectra. NoteObserved that the contamination Wavelength model[µm] in- direction along which sources are dispersed in the slitless grism cludes emission lines for the neighboring sources as calibrated from spectra for our two individual epochs (magenta and blue) and for ourFig. two-epoch 2.— data.Our The model high contaminating of the contaminating flux in the AGHAST flux from neighbor- the pre-existing AGHASTPerform data (green). full The 2D latter contamination are significantly dataing modelling makessources these in spectra the and slitless inadequate neighbor grism for studying spectra subtraction GN-z11. around Our the trace of our contaminated by bright neighbors along the dispersion direction of orientations1.1 were chosen1.2 based on1.3 extensive simulations1.4 to1.5 mini- 1.6 1.7 GN-z11 (see Fig 2). mizesource such (i.e., contamination the line from along neighbors. which However, we expect some its zeroth flux; indicated by (based on 3D-HST grism pipeline;orderred lines). flux in Brammer+12, epoch From 2 could top notObserved to be bottom, avoided Momcheva+15) Wavelength while the at panels the same show [µ timem] the final model Fig. 1.— CANDELS H-band image around the location of our makingcontamination the observations in our schedulable spectra with inHST epochsin cycle 1 22. and 2 and in the pre- target source GN-z11. The arrows2.2. Slitless and dashed Grism lines Data indicate the existing AGHAST spectra. Note that the contamination model in- directionMontreal, along whichOctThe 2016 primary sources data are dispersedanalyzed in in this the paper slitless are 12 grism orbit P. Oesch,cludeswhere Observatoire emissionfcontam UniGEis lines the contamination for the neighboring model flux sources in a par- as calibrated from spectra for our twodeep individual G141 grism epochs spectra (magenta from our andHST blue)program and forGO- ticularour two-epoch pixel and data. is the The per-pixel high contaminating RMS taken from flux the in the AGHAST WFC3/IR noise model (c.f., 3.4.3 from Rajan & et al. the pre-existing AGHAST13871 (PI: data Oesch). (green). These The spectra latter were are taken significantly at two dif- data makes these spectra§ inadequate for studying GN-z11. Our contaminated byferent bright orients neighbors in two along epochs the of dispersion six orbits direction each on 2015 of 2011orientations). We have were tested chosen that basedis accurately on extensive characterized simulations to mini- GN-z11 (see Fig February2). 11 and April 3 (see Figure 1). The data ac- asmize demonstrated such contamination by the pixel from flux distribution neighbors. function However, some zeroth quisition and observation planning followed the success- inorder the 2D flux residual in epoch spectra 2 could (see not Appendix be avoided Fig 11 while). at the same time ful 3D-HST grism program (Brammer et al. 2012; Mom- makingA local the background observations is subtracted schedulable from with the 2DHST framein cycle 22. cheva et al. 2015). Together with each G141 grism expo- which is a simple 2nd order polynomial estimated from 2.2. Slitless Grism Data the contamination-cleaned pixels above and below the sure, a short 200 s pre-image with the JH140 filter was The primarytaken data to determineanalyzed the in zeroth this order paper of the are grism 12 spectra. orbit tracewhere of thefcontam target source.is the contamination1D spectra are then model com- flux in a par- The JH images were placed at the beginning and end putedticular using pixel optimal and extraction is the on per-pixel the final 2D RMS frames, taken from the deep G141 grism spectra140 from our HST program GO- weighting the flux by the morphology of the source as of each orbit in order to minimize the impact of variable WFC3/IR noise model (c.f., 3.4.3 from Rajan & et al. 13871 (PI: Oesch).sky background These spectra on the grism were exposure taken due at to two the bright dif- measured in the WFC3/IR imaging (Horne§ 1986). Earth limb and He 1.083 µm line emission from the upper 2011The). GOODS-N We have field, tested which that contains is ouraccurately candidate characterized ferent orients in two epochs of six orbits each on 2015 GN-z11, has previously been covered by 2 times 2- February 11 andatmosphere April ( 3Brammer (see Figure et al. 20141).). A The four-point data dither ac- as demonstrated by the pixel flux distribution function pattern was used to improve the sampling of the point- orbit deep grism data from “A Grism H-Alpha Spec- quisition and observation planning followed the success- Troscopicin the 2D survey” residual (AGHAST; spectra GO-11600; (see Appendix PI: Wiener). Fig 11). spread function and to overcome cosmetic defects of the A local background is subtracted from the 2D frame detector. However, the spectrum of our target GN-z11 is signif- ful 3D-HST grism program (Brammer et al. 2012; Mom- icantly contaminated in the AGHAST observations by cheva et al. 2015The). Together grism data with were reduced each G141 using the grism grism expo- reduc- which is a simple 2nd order polynomial estimated from tion pipeline developed by the 3D-HST team. The main severalthe contamination-cleaned nearby galaxies given their orientation pixels above (see Fig- and below the sure, a short 200 s pre-image with the JH filter was ure 2). Furthermore, the observations were severely af- reduction steps are explained in detail140 by Momcheva trace of the target source. 1D spectra are then com- taken to determineet al. ( the2015 zeroth). In particular, order of the the flat-fielded grism spectra. and global fected by variable sky background (Brammer et al. 2014). Theputed final using data used optimal in this paper extraction therefore do onnot theinclude final 2D frames, The JH140 imagesbackground-subtracted were placed at grism the imagesbeginning are interlaced and end to produce 2D spectra for sources at a spectral sampling theweighting AGHAST the spectra, flux which by were the not morphology optimized for GN- of the source as of each orbit in order to minimize the impact of variable z11 that was discovered 3–4 years after they were taken. of 23 A,˚ i.e., about one quarter of the native resolution measured in the WFC3/IR imaging (Horne 1986). sky background⇠ on the grism exposure due to the bright We confirmed that including the AGHAST data would of the G141 grism. The final 2D spectrum is a weighted notThe a↵ect GOODS-N our results, however, field, which because containsof the down- our candidate Earth limb andstack He 1.083of the dataµm fromline emission individual visits. from the In particular, upper weightingGN-z11, of heavily has previously contaminated been pixels in covered our stacking by 2 times 2- atmosphere (Brammerwe down-weight et al. pixels2014 which). A are four-point a↵ected by neighbor- dither procedure. pattern was usedcontamination to improve using the sampling of the point- orbit deep grism data from “A Grism H-Alpha Spec- 2 2 1 w =[(2 f ) + ] (1) Troscopic survey” (AGHAST; GO-11600; PI: Wiener). spread function and to overcome⇤ contam cosmetic defects of the detector. However, the spectrum of our target GN-z11 is signif- The grism data were reduced using the grism reduc- icantly contaminated in the AGHAST observations by tion pipeline developed by the 3D-HST team. The main several nearby galaxies given their orientation (see Fig- reduction steps are explained in detail by Momcheva ure 2). Furthermore, the observations were severely af- et al. (2015). In particular, the flat-fielded and global fected by variable sky background (Brammer et al. 2014). background-subtracted grism images are interlaced to The final data used in this paper therefore do not include produce 2D spectra for sources at a spectral sampling the AGHAST spectra, which were not optimized for GN- z11 that was discovered 3–4 years after they were taken. of 23 A,˚ i.e., about one quarter of the native resolution ⇠ We confirmed that including the AGHAST data would of the G141 grism. The final 2D spectrum is a weighted not a↵ect our results, however, because of the down- stack of the data from individual visits. In particular, weighting of heavily contaminated pixels in our stacking we down-weight pixels which are a↵ected by neighbor- procedure. contamination using 2 2 1 w =[(2 f ) + ] (1) ⇤ contam Lyman Break Detection at z=11 NEW CONTAM! 6 Oesch et al.

Lyα Redshift Oesch+16 9 10 11 12 1.0 0.5 0.0 Trace § 12 orbits of HST grism spectra with −0.5

scale [arcsec] −1.0 WFC3/IR

1.1 1.2 1.3 1.4 1.5 1.6 600 § Detect UV continuum (at 5.5σ) and a GN-z11 HST WFC3/IR G141 Grism Spectrum break at λ > 1.47 µm 400 § Rule out potential lower redshift 200 solutions (quiescent galaxy at z~2 or

0 strong emission line source) Flux Density [nJy] Flux Density [nJy]

−200 Model at § Best-fit redshift: z=11.09+-0.10 +0.08 zgrism=11.09−0.12 contamination GN-z10-1 ➔ GN-z11 −400 1.1 1.2 1.3 1.4 1.5 1.6 Wavelength [µm]

Fig. 5.— Grism Spectrum of GN-z11. The top panel shows the (negative) 2D spectrum from the stack of our cycle 22 data (12 orbits) with the trace outlined by the dark red lines. For clarity the 2D spectrum wasMost smoothed bydistant a Gaussian source indicated by ever the ellipse seen in the lower right corner. The bottom panel is the un-smoothed 1D flux density using an optimal extraction rebinned to one resolution element of the G141 grism (93 A).˚ The black dots showBuild-up the same further of binned massive to 560 A,˚ whilegalaxies the blue line well shows underway the contamination at level 400 that was Myr after Big Bang subtracted from the original object spectrum. We identify a continuum break in the spectrum at =1.47 0.01 µm. The continuum flux at > 1.47 µm is detected at 1 1.5 per resolution element and at 3.8 per 560 Abin.Afterexcludinglowerredshiftsolutions(see˚ ± Montreal,⇠ Oct 2016 +0.08 P. Oesch, Observatoire UniGE 14 text and Fig 4), the best-fit grism redshift is zgrism =11.09 0.12.TheredlinereflectstheLy↵ break at this redshift, normalized to the measured H-band flux of GN-z11. The agreement is excellent. The fact that we only detect significant flux along the trace of our target source, which is also consistent with the measured H-band magnitude, is strong evidence that we have indeed detected the continuum of GN-z11 rather than any residual contamination. tinuum break with J125 H160 > 2.4(2), without higher than the observed mean, while at > 1.47 µmthe a spectrum, we could not exclude contamination by a expected flux is too low compared to the observations. source with very extreme emission lines with line ratios Overall the likelihood of a z 2 extreme emission line ⇠ 6 reproducing a seemingly flat continuum longward of 1.4 SED based on our grism data is less than 10 and can µm(Oesch et al. 2014). be ruled out. The previous AGHAST spectra already provided Note that in very similar grism observations for a some evidence against strong emission line contam- source triply imaged by a CLASH foreground clus- ination (Oesch et al. 2014), and we also obtained ter, emission line contamination could also be excluded Keck/MOSFIRE spectroscopy to further strengthen this (Pirzkal et al. 2015). We thus have no indication cur- conclusion (see appendix). However, the additional 12 rently that any of the recent z 9 11 galaxy candidates orbits of G141 grism data now conclusively rule out that identified with HST is a lower⇠ redshift strong emission GN-z11 is such a lower redshift source. Assuming that all line contaminant (but see, e.g., Brammer et al. 2013, for the H-band flux came from one emission line, we would a possible z 12 candidate). have detected this line at > 10. Even when assuming ⇠ a more realistic case where the emission line flux is dis- 3.3. Excluding a Lower-Redshift Dusty or Quiescent tributed over a combination of lines (e.g., H +[OIII]), Galaxy we can confidently invalidate such a solution. The lower left panel in Figure 4 compares the measured grism spec- Another potential source of contamination for very high redshift galaxy samples are dusty z 2 3 sources trum with that expected for the best-fit lower redshift ⇠ solution we had previously identified (Oesch et al. 2014). with strong 4000 A˚ or Balmer breaks (Oesch et al. 2012; A strong line emitter SED is clearly inconsistent with Hayes et al. 2012). However, the fact that the IRAC data the data. Apart from the emission lines, which we do for GN-z11 show that it has a very blue continuum long- not detect, this model also predicts weak continuum flux ward of 1.6 µm, together with the very red color in the across the whole wavelength range. At < 1.47 µm, this is WFC3/IR photometry, already rules out such a solution (see SED plot in Figure 4). Nevertheless, we additionally Physical Properties of GN-z11 in Line with Models

Monsters in the Dark 3 GNz11 analogues in DRAGONS 3 Waters+16

Table 1. The properties of the GN-z11 analogues, DR-1 and DR-2, at both Mutch+16 z=11.1 and 5. For comparison, the observationally determined values for GN-z11 itself are also shown (Oesch et al. 2016). GNz11 z=11.1 z=5 GNz11 GN-z11 DR-1 DR-2 DR-1 DR-2

M 22.1 0.2 -22.4 -21.2 -24.3 -23.2 UV ± log10 (M / M ) 9.0 0.4 9.3 9.0 11.0 10.6 ⇤ ± log10 (Mvir/ M ) – 11.1 10.8 12.2 11.5 1 SFR [ M /yr ] 24.0 10 66.4 19.0 237.6 94.5 ± r [kpc] 0.6 0.2 0.3 0.2 0.9 0.4 half-light ±

Figure 3. The average predicted spectral energy distribution of bright (M 22) galaxies at z =11.1inBlueTides compared fuv ⇡ with the observed fluxes of GN-z11. The two spectral energy dis- tributions shown denote both the pure stellar case (fesc,LyC =1) and the case in which the Lyman continuum escape fraction is e↵ectively 0. In the latter case the Lyman-↵ line has also been damped.

that GN-z11 has little or no dust attenuation. This can be seen in Figure 3 where we show the average intrinsic spec- tral energy distribution of bright (Mfuv 22) galaxies at Figure 3. The UV magnitude vs. stellar mass distribution⇡ of M z =11.1. We note, however, that the predicted intrinsic UV galaxies at z 11. The thick black line and shaded region indicate the median ⇠ continuum slope also sensitive to the choice of stellar pop- and 95 pc confidenceulation intervals synthesis of the distribution model, initial as a mass function function, of stellar and mass. assumed TheThe top (/right) derived panelsLyman indicate physical continuum the marginalised escape properties fraction. log-probability Alternative distribution choices(SFR, can mass,Figure 4. Stellar and mass, age) star formation of rate GN-z11 and stellar ages versusare in very ˙ of stellar mass (/UV magnitude)result in bluer values. intrinsic The red UV point continuum with error slopes bars leaving indicate open UV luminosity for the galaxies at z =11inBlueTidesThe large the possibility of some dust attenuation (see Wilkins et al. black data point denotes the inferred constraints on GN-z11 from the position ofgood GN-z11 inagreement this plane, whilst the greywith squares expectations show the model O16. from large-volume simulations submitted). analogue galaxies, DR-1 and DR-2. Both analogue galaxies (as well as GN- Figure 2. The observed frame SED of DR-1 (green) and DR-2 (orange) z11) are rare outliers from the main distribution, but are approximately 5.1 Morphologies in units of flux density. Lyman-↵ absorption has been included, however, consistentHowever: with an extrapolation4.2 Stellar did of mass, thenot median SFR expect relation and stellar from to ages lower find masses. such a source in the current HST fields no dust extinction model has been applied. Red points show the observed In Figure 5 we show the stellar surface density (for a random BlueTides makes predictions for a number of properties of orientation) for a sample of five galaxies with MUV < 22 in GN-z11 HST photometric measurements and upper limits. Both model ana- GN-z11 which have have been inferred by O16. In Figure 4 the z = 11 snapshot of the BlueTidessimulation. Three of logue galaxies have a spectral shape in good agreement with theMontreal, GN-z11 Oct 2016 we show the stellar masses, star formation ratesP. andOesch, stellar Observatoire UniGE 15 model has been applied. For more details on the methodology used the galaxies closely match the brightess of GN z11 (on the ages as a function of UV luminosity for the z =11galaxies measurements. DR-1 also shares the same normalisation due to it having a left) and two are examples of brighter galaxies. Even though to construct the modelin BlueTides SEDs, see. TheLiu black et dataal. (2015 points). show Both the analogue correspond- very similar UV luminosity (see Table 1). massive and bright, the galaxies show irregular, disturbed ing values inferred for GN-z11 by O16. We can see that in galaxies possess spectra and UV slopes () in close agreement with morphologies and have typical sizes 1 kpc. Note that in the relevant magnitude range, BlueTides predicts stellar ⇠ the GN-z11 observations, supporting the claim that this observed Feng et al. 2015a we found from a visual and kinematic masses 109M , SFR of a few tens M /yr and stellar ages ⇠ analysis that the most massive galaxies at z = 8 are nearly higher redshifts may suggest (see Oesch et al. 2016, and references system is relativelyof dust a about free 30 (Oesch60 Myr et foral. galaxies2016). Thewith normalisationmUV 22. These ⇠ all classified as disks. We can see here that this does not values are fully conistent with all the O16 constraints. therein). of DR-1 spectrum further shows excellent agreement with GN-z11 appear to be the case as early as z =11. Having established that galaxies as UV-luminous as GN-z11 measurements due to the similar UV luminosities of these two exist in the output of DRAGONS, we selected the two brightest objects (c.f Table 1). 5 PROPERTIES: PREDICTIONS 5.2 Metallicity UV magnitude galaxies at redshift z=11.1 for detailed study. The DR-1 and 2 are the two most massive galaxies in the simulation properties of these objects1, hence-forth referred to as DR-1 and DR- at the redshift at whichAs galaxies they were with theselected presently (z=11.1) observed and characteristics are hosted of In Figure 6 we show predictions for both the star forming GN-z11 exist in BlueTides it is useful to investigate their gas and stellar metallicity of galaxies at z =11inBlue- 2, show good agreement with the best observational measurements by the two most massiveother properties. subhaloes.BlueTides Theyhas are high also enough rare spatial outliers reso- Tides. Galaxies in the simulation follow a strong luminos- of GN-z11 (Oesch et al. 2016), as can be seen in Table 1. This from the majority oflution the (180 model pc at galaxyz = 11) population to allow determination in terms of of their galaxy ity - metallicity relationship. For bright galaxies such as morphologies (see Feng et al. 2015a). The simulation also GB z11 we predict mean stellar metallicities 0.001 0.002 suggests that we can use these model galaxies as analogues with stellar masses, star formation rates and UV luminosities. However, tracks gas and stellar metallicities and includes modelling of (approximately 5 10% Z ) with star forming gas metallic- which to investigate the potential formation, evolution and fate of they remain broadlyblack consistent holes. Here with we make the mean predictions trends for displayed these aspects. by ities around a factor of 2 higher. GN-z11. galaxies at lower luminosities / masses, suggesting that their history As well as those properties listed in Table 1, DR-1 and 2 also is not particularly specialMNRAS or000 unique., 1–?? (2015) As an example, in Figure 3 we possess similar SEDs to GN-z11. In Figure 2 we plot the observed present the distribution of all M z=11.1 galaxies in the UV frame SEDs of DR-1 (green) and DR-2 (orange) in terms of their flux magnitude vs. stellar mass plane. The positions of DR-1 and DR-2 density between 0.2 and 6 µm. The red data points show the GN- are shown as grey squares, whilst GN-z11 is indicated by the red z11 HST photometric measurements and upper limits. Lyman-↵ point with error bars. Although these three objects lie out-with the absorption has been included in the model spectra and is manifested bulk of the main distribution, they are roughly consistent with the as the sharp drop in flux at .1.1 µm, however, no dust extinction median M1600–M trend extrapolated from lower masses. ⇤

1 The model galaxy half-light radii are calculated from the disk scale radius, rs, using rhalf-light=1.68rs for an axisymmetric exponential disk 4 THE ORIGIN AND FATE OF GN-Z11 profile, where rs is derived from the spin of the host dark matter subhalo The detection of such a massive and luminous galaxy at z 11 raises under the assumption of specific angular momentum conservation (Fall & ⇠ Efstathiou 1980; Mo et al. 1998; Mutch et al. 2015). a number of interesting questions. How do such massive systems form so rapidly? Is their extreme brightness merely a transient fea- ture of their evolution brought on by a merger or other significant

MNRAS 000, 1–7 (2016) The Current Spectroscopic Frontier

8 Oesch et al. NEW CONTAM! 6 Oesch et al. Age of the Universe [Gyr] Lyα Redshift 9 10 11 12 TABLE 2 0.9 0.8 0.7 0.6 0.5 0.4 −23 1.0 Summary of Measurements for GN-z11 0.5 4 0.0 Trace R.A. 12 : 36 : 25.46 GN-z11 −0.5 GN-z11 is by far the most scale [arcsec] Dec. +62 : 14 : 31.4 UV −22 −1.0 distant known source +0.08 a (z=7)] 6001.1 1.2 1.3 1.4 1.5 1.6

2 * Redshift zgrism 11.09 0.12 GN-z11 HST WFC3/IR G141 Grism Spectrum UV Luminosity MUV 22.1 0.2 Half Light Radiusb 0.6 0±.3kpc 400 c ± −21 1 log Mgal/M 9.0 0.4 log age/yr c 7.6 ± 0.4 200 ± 1 SFR 24 10 M yr 0.5 ± AUV < 0.2mag −20 0 Flux Density [nJy] Absolute Magnitude M Flux Density [nJy] d zspec UV Luminosity [L/L UV slope (f ) 2.5 0.2 / ± zphot −200 Model at a 0.25 +0.08 Age of the Universe at z =11.09 using our cosmol- zgrism=11.09−0.12 contamination ogy: 402 Myr −19 −400 b 6 7 8 9 10 11 1.1 1.2 1.3 1.4 1.5 1.6 From Holwerda et al. (2015) Wavelength [µm] c Uncertainties are likely underestimated, since our Redshift photometry only partially covers the rest-frame opti- Fig. 5.— Grism Spectrum of GN-z11. The top panel shows the (negative) 2D spectrum from the stack of our cycle 22 data (12 orbits) with the trace outlined by the dark red lines. For clarity the 2D spectrum was smoothed by a Gaussian indicated by the ellipse in the lower cal for GN-z11 Fig. 7.— The redshift and UV luminosities of knownright corner. high- The bottom panel is the un-smoothed 1D flux density using an optimal extraction rebinned to one resolution element of the d G141 grism (93 A).˚ The black dots show the same further binned to 560 A,˚ while the blue line shows the contamination level that was See also Wilkins et al. (2016) redshift galaxies from blank field surveys. Dark filled squaressubtracted corre- from the original object spectrum. We identify a continuum break in the spectrum at =1.47 0.01 µm. The continuum flux at > 1.47 µm is detected at 1 1.5 per resolution element and at 3.8 per 560 Abin.Afterexcludinglowerredshiftsolutions(see˚ ± spond to spectroscopically confirmedWith sources, surprising while small discovery gray dots of⇠ GN-z11, +0.08 text and Fig 4), the best-fit grism redshift is zgrism =11.09 0.12.TheredlinereflectstheLy↵ break at this redshift, normalized to the are photometric redshifts (Bouwens et al. 2015b). GN-z11measured clearly H-band flux of GN-z11. The agreement is excellent. The fact that we only detect significant flux along the trace of our target HST+Spitzer have alreadysource, reached which is also consistent into with the JWST measured H-band territory magnitude, is strong evidence that we have indeed detected the continuum of to estimate how many such galaxies we could have ex- stands out as one of the most luminous currently knownGN-z11 galaxies rather than at any residual contamination. all redshifts z>6andisbyfarthemostdistantmeasuredgalaxy pected based on (1) the currently best estimates of the tinuum break with J125 H160 > 2.4(2), without higher than the observed mean, while at > 1.47 µmthe with spectroscopy (black squares; see Oesch et al. 2015ba,forafull spectrum, we could not exclude contamination by a expected flux is too low compared to the observations. UV LF at z>8 and (2) based on theoretical models and list of references). Wider area surveys with future near-infraredsource with very extreme emission lines with line ratios Overall the likelihood of a z 2 extreme emission line ⇠ 6 reproducing a seemingly flat continuum longward of 1.4 SED based on our grism data is less than 10 and can simulations. telescopesMontreal, Oct 2016 (such as WFIRST) will be required to determineP. Oesch,µm( ObservatoireOesch how et al. UniGE2014). be ruled out. 16 common such luminous sources really are at z>10. The previous AGHAST spectra already provided Note that in very similar grism observations for a Our target was found in a search of the GOODS fields, some evidence against strong emission line contam- source triply imaged by a CLASH foreground clus- 2 ination (Oesch et al. 2014), and we also obtained ter, emission line contamination could also be excluded which amount to 160 arcmin . However, in a sub- Keck/MOSFIRE spectroscopy to further strengthen this (Pirzkal et al. 2015). We thus have no indication cur- ⇠ conclusion (see appendix). However, the additional 12 rently that any of the recent z 9 11 galaxy candidates sequent search of the three remaining CANDELS fields orbits of G141 grism data now conclusively rule out that identified with HST is a lower⇠ redshift strong emission no similar sources were found with likely redshifts at TABLE 3 GN-z11 is such a lower redshift source. Assuming that all line contaminant (but see, e.g., Brammer et al. 2013, for Assumed LFs for z 10 11 Number Density Estimatesthe H-band flux came from one emission line, we would a possible z 12 candidate). z 10 (Bouwens et al. 2015a). We therefore use the ⇠ have detected this line at > 10. Even when assuming ⇠ & a more realistic case where the emission line flux is dis- 3.3. Excluding a Lower-Redshift Dusty or Quiescent 2 tributed over a combination of lines (e.g., H +[OIII]), full 750 arcmin of the CANDELS fields with match- 5 Galaxy Reference /10 M ↵ we canNexp confidently invalidate such a solution. The lower left panel in Figure 4 compares the measured grism spec- Another potential source of contamination for very ing WFC3/IR and ACS imaging for a volume estimate, ⇤ 3 ⇤ high redshift galaxy samples are dusty z 2 3 sources [Mpc ][mag] (trum< with22. that1) expected for the best-fit lower redshift ⇠ which amounts to 1.2 106 Mpc3 (assuming z = 1). solution we had previously identified (Oesch et al. 2014). with strong 4000 A˚ or Balmer breaks (Oesch et al. 2012; Bouwens et al. (2015b)1.65-20.97-2.380.06A strong line emitter SED is clearly inconsistent with Hayes et al. 2012). However, the fact that the IRAC data Using the simple trends⇥ in the Schechter parameters of the data. Apart from the emission lines, which we do for GN-z11 show that it has a very blue continuum long- Finkelstein et al. (2015)0.96-20.55-2.900.002not detect, this model also predicts weak continuum flux ward of 1.6 µm, together with the very red color in the across the whole wavelength range. At < 1.47 µm, this is WFC3/IR photometry, already rules out such a solution the UV LFs measured UV at lower redshift (z 4 8) Mashian et al. (2016)0.25-21.20-2.200.03 (see SED plot in Figure 4). Nevertheless, we additionally and extrapolating these to z = 11, we can get an⇠ empir- Mason et al. (2015)0.30-21.05-2.610.01 ical estimate of the number density of very bright galax- Trac et al. (2015)5.00-20.18-2.220.002 ies at z 11. This amounts to 0.06 (Bouwens et al. Note.—Theparameters , M ,and↵ represent the three 2015b) or⇠ 0.002 (Finkelstein et al. 2015) expected galax- parameters of the Schechter UV⇤ LF taken⇤ from the di↵erent papers. ies brighter than MUV = 22.1 in our survey correspond- ing to less than 0.3 per surveyed square degree. Simi- larly, recent empirical models (Mashian et al. 2016; Ma- z11. Our 2D data show clear flux longward of 1.47 µm son et al. 2015; Trac et al. 2015) predict only 0.002 0.03 exactly along the trace of the target galaxy⇠ and zero galaxies as bright as GN-z11 in our survey or 0.01 0.2 flux at shorter wavelengths, thanks to our comprehensive per deg2. All the assumed LF parameters together with and accurate treatment of contamination by neighboring the resulting estimates of the number of expected bright galaxies. The interpretation that we indeed detect the galaxies Nexp are listed in Table 3. continuum flux from GN-z11 is supported by the mor- The above estimates illustrate that our discovery of phology of the spectrum, the fact that the counts fall o↵ the unexpectedly luminous galaxy GN-z11 may challenge exactly where the sensitivity of the G141 grism drops, as our current understanding of galaxy build-up at z>8. well as the consistency of the observed counts with the A possible solution is that the UV LF does not follow H-band magnitude of GN-z11 (see e.g. Fig 3). a Schechter function form at the very bright end as has The grism spectrum, combined with the photometric been suggested by some authors at z 7(Bowler et al. constraints, allows us to exclude plausible low-redshift 2014), motivated by inecient feedback⇠ in the very early SEDs for GN-z11 at high confidence. In particular, we universe. However, current evidence for this is still weak can invalidate a low redshift SED of an extreme line emit- (see discussion in Bouwens et al. 2015b). Larger area ter galaxy at z 2 (see section 3 and Fig 4). Instead, studies will be required in the future (such as the planned the grism spectrum⇠ is completely consistent with a very +0.08 WFIRST High Latitude Survey; Spergel et al. 2015)sur- high-redshift solution at zgrism = 11.09 0.12 (see Figures veying several square degrees to determine the bright end 3 and 5). This indicates that this galaxy lies at only of the UV LF to resolve this puzzle. 400 Myr after the Big Bang, extending the previous redshift⇠ record by 150 Myr. 5. SUMMARY GN-z11 is not only⇠ the most distant spectroscopically In this paper we present HST slitless grism spectra measured source, but is likely even more distant than for a uniquely bright z>10 galaxy candidate, which all other high-redshift candidates with photometric red- +0.6 we previously identified in the GOODS-North field, GN- shifts, including MACS0647-JD at zphot = 10.7 0.4 (Coe Moving Forward with JWST NIRSPEC

8 Oesch et al.

Age of the Universe [Gyr] 0.9 0.8 0.7 0.6 0.5 0.4 TABLE 2 −23 Summary of Measurements for GN-z11 4 R.A. 12 : 36 : 25.46 GN-z11 Dec. +62 : 14 : 31.4 UV −22 +0.08 a (z=7)]

2 * Redshift zgrism 11.09 0.12 400 JWST/NIRSpec UV Luminosity MUV 22.1 0.2 b ± z = 11.1 Half Light Radius 0.6 0.3kpc 300 m = 25.9 c ± −21 1 obs log Mgal/M 9.0 0.4 T = 1800 s c ± exp log age/yr 7.6 0.4 200 ± 1 SFR 24 10 M yr 0.5 ± AUV < 0.2mag −20 Flux nJy

Absolute Magnitude M 100 d zspec UV Luminosity [L/L UV slope (f ) 2.5 0.2 / ± z a phot 0.25 Age of the Universe at z =11.09 using our cosmol- 0 ogy: 402 Myr −19 b From Holwerda et al. (2015) 6 7 8 9 10 11 1 2 3 4 5 c Uncertainties are likely underestimated, since our Redshift Wavelength [micron] photometry only partially covers the rest-frame opti- cal for GN-z11 Fig. 7.— The redshift and UV luminosities of known high- d See also Wilkins et al. (2016) redshift galaxiesHowever, from blank JWST field surveys. spectroscopy Dark filled squares will completely corre- revolutionize this field! spond to spectroscopically confirmed sources, while small gray dots are photometric redshiftsJWST (Bouwens can et in al. principle2015b). GN-z11 get spectroscopic clearly redshifts for to estimate how many such galaxies we could have ex- stands out as one of the most luminous currently known galaxies at all redshifts z>6andisbyfarthemostdistantmeasuredgalaxyevery single source currently known with HST pected based on (1) the currently best estimates of the with spectroscopy (black squares; see Oesch et al. 2015b,forafull UV LF at z>8 and (2) based on theoretical models and list of references). Wider area surveys with future near-infrared simulations. telescopesMontreal, Oct 2016 (such as WFIRST) will be required to determineP. Oesch, Observatoire how UniGE 17 Our target was found in a search of the GOODS fields, common such luminous sources really are at z>10. which amount to 160 arcmin2. However, in a sub- sequent search of the⇠ three remaining CANDELS fields no similar sources were found with likely redshifts at TABLE 3 Assumed LFs for z 10 11 Number Density Estimates z & 10 (Bouwens et al. 2015a). We therefore use the ⇠ 2 full 750 arcmin of the CANDELS fields with match- 5 Reference /10 M ↵ Nexp ing WFC3/IR and ACS imaging for a volume estimate, ⇤ 3 ⇤ [Mpc ][mag] (< 22.1) which amounts to 1.2 106 Mpc3 (assuming z = 1). ⇥ Bouwens et al. (2015b)1.65-20.97-2.380.06 Using the simple trends in the Schechter parameters of Finkelstein et al. (2015)0.96-20.55-2.900.002 the UV LFs measured UV at lower redshift (z 4 8) Mashian et al. (2016)0.25-21.20-2.200.03 and extrapolating these to z = 11, we can get an⇠ empir- Mason et al. (2015)0.30-21.05-2.610.01 ical estimate of the number density of very bright galax- Trac et al. (2015)5.00-20.18-2.220.002 ies at z 11. This amounts to 0.06 (Bouwens et al. Note.—Theparameters , M ,and↵ represent the three 2015b) or⇠ 0.002 (Finkelstein et al. 2015) expected galax- parameters of the Schechter UV⇤ LF taken⇤ from the di↵erent papers. ies brighter than MUV = 22.1 in our survey correspond- ing to less than 0.3 per surveyed square degree. Simi- larly, recent empirical models (Mashian et al. 2016; Ma- z11. Our 2D data show clear flux longward of 1.47 µm son et al. 2015; Trac et al. 2015) predict only 0.002 0.03 exactly along the trace of the target galaxy⇠ and zero galaxies as bright as GN-z11 in our survey or 0.01 0.2 flux at shorter wavelengths, thanks to our comprehensive per deg2. All the assumed LF parameters together with and accurate treatment of contamination by neighboring the resulting estimates of the number of expected bright galaxies. The interpretation that we indeed detect the galaxies Nexp are listed in Table 3. continuum flux from GN-z11 is supported by the mor- The above estimates illustrate that our discovery of phology of the spectrum, the fact that the counts fall o↵ the unexpectedly luminous galaxy GN-z11 may challenge exactly where the sensitivity of the G141 grism drops, as our current understanding of galaxy build-up at z>8. well as the consistency of the observed counts with the A possible solution is that the UV LF does not follow H-band magnitude of GN-z11 (see e.g. Fig 3). a Schechter function form at the very bright end as has The grism spectrum, combined with the photometric been suggested by some authors at z 7(Bowler et al. constraints, allows us to exclude plausible low-redshift 2014), motivated by inecient feedback⇠ in the very early SEDs for GN-z11 at high confidence. In particular, we universe. However, current evidence for this is still weak can invalidate a low redshift SED of an extreme line emit- (see discussion in Bouwens et al. 2015b). Larger area ter galaxy at z 2 (see section 3 and Fig 4). Instead, studies will be required in the future (such as the planned the grism spectrum⇠ is completely consistent with a very +0.08 WFIRST High Latitude Survey; Spergel et al. 2015)sur- high-redshift solution at zgrism = 11.09 0.12 (see Figures veying several square degrees to determine the bright end 3 and 5). This indicates that this galaxy lies at only of the UV LF to resolve this puzzle. 400 Myr after the Big Bang, extending the previous redshift⇠ record by 150 Myr. 5. SUMMARY GN-z11 is not only⇠ the most distant spectroscopically In this paper we present HST slitless grism spectra measured source, but is likely even more distant than for a uniquely bright z>10 galaxy candidate, which all other high-redshift candidates with photometric red- +0.6 we previously identified in the GOODS-North field, GN- shifts, including MACS0647-JD at zphot = 10.7 0.4 (Coe F606W F814W F105W F125W F160W [3.6] [4.5] TheHigh Astrophysical Redshift Journal Letters,777:L19(6pp),2013November10 Galaxies Show Extreme Lines Labbeetal.´ bright z 7galaxies(Bouwensetal.2013). The best-fit sSFR ∼ +13 1 is then sSFR 7 4 Gyr− ,ingoodagreementwiththesSFR derived23 from W= −. Hα 2 We deriveEGS-zs8-1 integrated stellarzbest=7.7 mass (χ = densities1.7) of H-band se- Labbe+13 24 2 lected galaxies to faint UV-limitszlow=1.8 (χ by= 34.2) converting the stepwise UV–luminosity25 function (LF; Bouwens et al. 2011;Oeschetal. 2012)tostepwisemassdensities,usingthebest-fitM/Ltothe

emissionAB 26 line corrected SED. 1 m zphot=7.7±0.3 Note that the emission line correction is derived from Extremely Strong 1.5L 27(UV,z 8) galaxies, so0.5 we assume that the effect Rest-frame Optical Lines ∼ ∗ = of emission lines is similar for L7.2 or emission show line strengths extremely inferred fromred excess IRAC emission colors in the IRAC This paper presents new ultradeep IRAC data from26 the [3.6] filter of spectroscopically confirmed galaxies at z 4–5. The broadband 100 derived [O iii]+Hβ EW at z 8 from this Letter are shown∼ by the blue solid IUDF program (120 hr in [3.6] and [4.5] over the HUDF) in 2

χ 15 ∼ 27 symbols. They are placed on the same scale assuming the relative emission combination with a large samplezspec of 63=7.51Y105-dropout galaxies line strengthsRest-frame of Anders & Fritze-v. equivalent Alvensleben (2003widths) for sub-solar 0.2 Z at z 8tostudyaverageHST/WFC3 and Spitzer/IRAC (lower10 value) and solar (higher value). The dotted line shows the EW evolution⊙

Observed Flux Densities AB Magnitude on average are 700 Å! ∼ 28 (1 + z)1.8 derived by Fumagalli et al. (2012)forstarforminggalaxieswith SEDs,Flux Density (nJy) the effect of emission lines, and their impact on stellar Observed 2σ Limit ∝ population model fits. Best−fit Model (z=7.51) stellar5 mass 10.0 < log(M/M ) < 10.5at0Space Telescope.Butinthemeantimewiderand ∼ ∼ we compare to recent estimates for star forming galaxies from still deeper future IRAC observations will continue3,5,6,8 to provide but it is predicted to be about 5× fainter than [O III]andthusdoesnot 2.5σ significance. This confirms previous results at z ∼ 6.5, but other surveys (shown in Figure 5). While direct comparisons valuable insight into the properties of the z 8population, significantly affect the 3.6µmband. here we probe z>7. The lack of detectable Lyα emission∼ is unlikely are difficult due to different techniques and stellar mass ranges, with emerging detections of z 9–10 (e.g., Zheng et al. 2012) This galaxy is forming stars at a very high rate, with a “mass- to be due to sample contamination, as contamination= by lower redshift ourdoubling” results time confirm of at most the picture 4 Myr. The of mostan increasing recent estimates importance17 at z ≈ of7 interlopersgalaxies is likely probably not dominant not far behind. at z = 7giventhelowcontamina- 9 8 emissionfind that galaxies lines toward with stellar higher masses redshift of 5 ×10 (e.g.,M⊙ Shimtypically et al. have2011 spe-; tion rate at z = 6. To explain the low detection rate of Lyα,anIGM 3,8 Fumagallicific SFRs et (sSFR al. =2012SFR;Starketal. divided by stellar2013 mass)). The∼10− evolution8 yr−1.This of neutralWe fraction are grateful at z = 6.5 to asJarle high Brinchmann, as 60–90% has Steve been Finkelstein, proposed, Casey 1.2 0.25 20 thegalaxy rest-frame is a factor EW of of five H lessα follows massive,W yetα its sSFR(1 + isz) a factor ofover 30 implyingPapovich,DanStark,SimoneWeinmannforhelpfuldiscussions. a rapid increase from z = 6. However, most other observa- H ± 21,22 1 greater remains30× greater. constant If this (e.g., SFR Onefor alternativethis work explanation was provided for at by least NASA part of through the Lyα deficit an award may issued function is accurate, the expected space density per co-moving Mpc3 be gas within galaxies. A high gas-to-stellar mass ratio may be con- van Dokkum et al. 2010;Papovichetal.−5 2011). by JPL/Caltech. This work has further been supported by NASA for this galaxy would be ≪ 10 .Theimpliedrarityofthisgalaxy sistent with the very high SFR of z8 GND 5296, as galaxies should couldClearly, imply at that such it is high the progenitor EW emission of some lines of the cannot most massi be ignoredve sys- not havegrant SFRs HST-GO-11563.01. (for long periods) exceeding We also acknowledgetheir average gas fundingaccre- from whentems inderiving the high-redshift stellar masses universe. and However, ages the atz high=7.213 redshiftgalaxy (e.g., GN- tion rateERC from grant the HIGHZ IGM (which No. is 227749. set by the total baryonic mass). For 3 Schaerer108036 ,alsoinGOODS-North,alsohasanimpliedSFR & de Barros 2010;deBarrosetal.2012> ).100 After M⊙ the inferredFacilities: stellar mass Spitzer and(IRAC), redshift, z8HSTGND(WFC35296/ mustIR) have a gas −1 correctingyr .Whilethecurrentstatisticsarepoor,thepresenceofthes the SEDs for emission lines we find that the L∗(zetwo8) reservoir of about 50 times the stellar mass to give an accretion rate galaxiesgalaxies have in a relatively stellar masses small survey and areaM/Lthatare suggests that the3 abundancelower.= of comparable to the SFR.23 If true,REFERENCES this galaxy would have a gas surface galaxies with such large star formation+11 rates1 may have∼ previ× ously been density similar to the most gas-rich galaxies in the local universe, and The WHα based sSFR 11 5 Gyr− are 5 higher than thoseunderestimated. at z 2(Reddyetal. If the high= SFR2012− of z8), consistentGND 5296∼ withcontinues× an increase down to its SFRAnders, would P., &be Fritze-v. consistent Alvensleben, with local U. relations 2003, A&A between, 401, 1063 the gas and 10 24 z = 6.3, it= would1 have.5+0.7 a stellar mass of ∼5 × 10 M⊙,comparableto SFRBouwens, surface densities. R. J., Illingworth,The large G. D., gas-to-stellar Oesch, P. A., masset al. 2011, ratioApJ could, 737, be 90 sSFR (1 + z) 0.5 between 2 , 533,7galaxies,it 682 Dave,´ R., Oppenheimer, B. D., & Finlater, C. 2011, MNRAS, 451, 11 (2013)at2

3 Rest-Frame Optical Emission Line Coverage

Ground-based multi-object spectrographs now trace galaxy build-up to z~2-4

Halpha 0.50.7 0.81.1 1.2 1.7 2.0 2.7

OIII 0.91.2 1.3 1.7 1.9 2.6 2.9 3.8

Y J H K Hbeta 1.0 1.3 1.4 1.8 2.0 2.7 3.0 3.9

OIII4363 1.2 1.6 1.6 2.1 2.4 3.1 3.5 4.5

OII 1.6 2.0 2.1 2.6 2.9 3.8 4.2 5.4

0 1 2 3 4 5 6 7 8 9 10 Redshift

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 19 Rest-Frame Optical Emission Line Coverage

JWST will seamlessly extend these surveys to z~7-9

Halpha 0.50.50.7 0.81.1 1.2 1.7 1.71.72.0 2.7 3.43.4 6.6

OIII 0.91.21.01.3 1.7 1.9 2.5 2.62.52.9 3.8 4.84.8 9.0

Y JBand1 H KBand2 Band3 Hbeta 1.01.1 1.3 1.4 1.8 2.0 2.6 2.72.63.0 3.9 5.05.0 9.3

OIII4363 1.21.3 1.6 1.6 2.1 2.4 3.0 3.13.03.5 4.5 5.65.6 10.5

OII 1.61.7 2.0 2.1 2.6 2.9 3.7 3.83.74.2 5.4 6.86.8

0 1 2 3 4 5 6 7 8 9 10 Redshift

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 20 JWST/NIRSpec: Unprecedented Spectra

Simulation based on z=7.73 source from Oesch+15 3000 T=0.5hr, R=1000 JWST will be extremely efficient in 2500 [OIII] • z=7.73 spectroscopic characterization of z>7 2000 galaxies; rest-frame optical lines will be

1500 [OII] Hβ extremely easy to detect Flux nJy 1000 Hγ • For brightest targets, like the recently 500 confirmed target EGS-zs8-1 at z=7.73, we will even be able to measure 0 3 3.2 3.4 3.6 3.8 4 4.2 4.4 4.6 absorption lines individually Wavelength [micron] 600 • Are working on collecting largest T=20hr, R=1000 possible samples of very bright 500 mobs = 25.0 SiII CIV CII SiIV galaxies at z>=8 SiII SiII 400 What is the ionization state of gas in

Flux nJy 300 early galaxies?

200 What is their dynamical state? How fast did they build up their metals? 100 1.05 1.1 1.15 1.2 1.25 1.3 1.35 Wavelength [micron]

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 21 When did the First Quiescent Galaxies Form?

The emergence of the quiescent galaxy population is a crucial test for galaxy formation models. quiescent = no star-formation

500

T=15hr, R=1000 400 z=5.5

300

Flux nJy 200 Hδ Hγ CaHK 100 Oesch+17, in prep. 0 2 2.2 2.4 2.6 2.8 3 Wavelength [micron]

Few candidate, passive galaxies exist at z>3. JWST will get redshifts for these and probe their star-formation histories with continuum spectroscopy. ➔ Indirect constraints on epoch of first galaxy formation.

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 22 Summary

• Combination of HST and Spitzer/IRAC allows us to probe the build up of both the star-formation rate and stellar mass density over 97% of cosmic history, revealing large samples of galaxies at z>6

• Current spectroscopic samples at z>6 are still extremely small due to a lack of strong Lyα emission lines in the EoR. But HST has recently extended spectroscopic frontier to z=11.1, only ~400 Myr after the Big Bang

• JWST will completely revolutionize this field. Strong lines will be detected out to the most distant and faintest sources found with HST. Will finally get a handle on the physics of early galaxy build up.

• Unique information will come from absorption line spectroscopy from luminous sources. Important to build up such samples before JWST launch.

Montreal, Oct 2016 P. Oesch, Observatoire UniGE 23 An HST Grism Redshift at z = 11 7

explore what constraints the grism spectrum alone can set on such a solution. 1.2 μm 1.4 μm 1.6 μm 4.5 μm 23 3 The expected flux for such a red galaxy increases grad- 10 ually across the wavelength range covered by the G141 Best-fitgrism, redshift unlike what weof observe combined in the data (lower right 25 in Fig 4). Compared to our best-fit solution (see next 2 2 spectrasection) + wephotometry: measure a = 15 z when=11.1 comparing ±0.1 the 10 data with the expected grism flux. Apart from the ex- 27 Magnitude [AB] (GN-z10-1tremely large ➜ discrepancy GN-z11!) with the IRAC photometry, we Flux Density [nJy] can thus exclude this solution at 98.9% confidence based Photometry of GN-z11 at zgrism=11.09 NEW CONTAM! 1 on the spectrum alone. 10 6Similar conclusions Oesch canet al. be drawn from the break 29 strength alone (see e.g. Spinrad et al. 1998). Assum- 0.5 1.0 2.0 4.0 6.0 Lyα Redshift Observed Wavelength [µm] ing that9 the observed10 break at 1.4711 µm corresponds12 to 4000 A˚ at z =2.7, a galaxy with a maximally old spec- 1.0 tral energy distribution (single burst at z = 15) would 300 Zoom-in: 0.5 short long Comparison with Grism Data show a flux ratio of (1 f⌫ /f⌫ ) < 0.63 when av- 0.0 Traceeraged over 560 A˚ bins. This is based on simple Bruzual 200 −0.5 & Charlot (2003) models without any dust. As men- scale [arcsec] −1.0 tioned earlier, the observed spectrum has a break of 100 (1 f short/f long) > 0.68 at 2, thus indicating again 6001.1 1.2⌫ ⌫ 1.3 1.4 1.5 1.6 that we can marginally rule out a 4000 A˚ break based 0 GN-z11on the HST spectrum WFC3/IR alone G141 withoutGrism Spectrum even including the pho- Flux Density [nJy] 400 tometric constraints. −100

4. DISCUSSION 1 1.2 1.4 1.6 1.8 200 4.1. Physical Properties of GN-z11 Observed Wavelength [µm] Despite being the most distant known galaxy, GN-z11 5 is relatively bright and reliably detected in both IRAC 0 just 3.6 and 4.5 µm bands from the S-CANDELS survey 4 grism Flux Density [nJy] Flux Density [nJy] (Ashby et al. 2015). This provides a sampling of its rest- 3 −200 frame UV spectral energy distributionModel and at even partially with +0.08 photo-z grism covers the rest-frame optical wavelengthszgrism=11.09 in the−0.12 IRAC 2 old 4.5 µm band (see Figure 6). contamination Probability F140W −400 The photometry of GN-z11 is consistent with a spectral 1 1.1 1.2 1.3 1.4 1.5 1.6 energy distributionWavelength (SED) of [µm] log M/M 9 using stan- 0 dard templates (Bruzual & Charlot 2003 ,⇠ see appendix). 10 10.5 11 11.5 The UV continuum is relatively blue with a UV spectral RedshiftR d hift Fig. 5.— Grism Spectrum of GN-z11. The top panel shows the (negative) 2D spectrum from the stack of our cycle 22 data (12 orbits) with the trace outlined by the dark red lines.slope For clarity= 2 the.5 2D0. spectrum2 as derived was smoothed from a by powerlaw a Gaussian indicated fit to the by the ellipse in the lower Montreal, Oct 2016 ± P. Oesch, Observatoire UniGE Fig. 6.— Top – The photometry (red) and best-fit24 spec- right corner. The bottom panel is the un-smoothedH160, K, 1Dand flux [3.6] density fluxes using only, an optimal indicating extraction very rebinned little to one dust resolution element of the G141 grism (93 A).˚ The black dots show the same further binned to 560 A,˚ while the blue line shows the contaminationtral level energy that was distribution (SED; gray) of GN-z11 at the measured subtracted from the original object spectrum.extinction We identify (see a also continuumWilkins break etin the al. spectrum2016). at Together =1.47 0 with.01 µm. Thegrism continuum redshift flux of z =11.09. Upper limits correspond to 1 non- at > 1.47 µm is detected at 1 1.5the per absence resolution of element a strong and at Balmer 3.8 per 560 break,Abin.Afterexcludinglowerredshiftsolutions(see˚ this is consistent± detections. The black squares correspond to the flux measurements ⇠ text and Fig 4), the best-fit grism redshift is z =11.09+0.08.TheredlinereflectstheLy↵ break at this redshift, normalizedof the best-fit to the SED. Inset negative images of 6 6 show the HST with agrism young stellar0.12 age of this galaxy. The best fit age 00⇥ 00 measured H-band flux of GN-z11. Theis agreement only 40 is excellent.Myr (< The110 fact Myr that at we 1 only). detect GN-z11 significant thus flux formed along the traceJ125 of, ourJH target140,andH160 bands as well as the neighbor-cleaned IRAC source, which is also consistent with the measured H-band magnitude, is strong evidence that we have indeed detected the4.5 continuumµmimage.GN-z11isrobustlydetectedinallbandslongward of GN-z11 rather than any residual contamination.its stars relatively rapidly. The inferred star-formation of 1.4 µm, resulting in accurate constraints on its physical param- rate is 24 10 M /yr. All the inferred physical parame- eters. The photometry is consistent with a galaxy stellar mass ± tinuum break with J125 H160 ters> 2 for.4(2 GN-z11), without are summarizedhigher than in the Table observed2. Overall, mean, while our at >of1. log47 M/Mµmthe 9.0withnoorverylittledustextinctionanda ⇠ a spectrum, we could not excluderesults contamination show that by galaxy a expected build-up flux was is toowell low underway compared at to theyoung observations. average stellar age. Middle – Azoom-inaroundthewave- source with very extreme emission lines with line ratios Overall the likelihood of a z 2 extremelength emission range line probed by the G141 grism. The rebinned grism data 400 Myr after the Big Bang. ⇠ are6 shown by the blue line with errorbars. The grism flux is con- reproducing a seemingly flat continuum⇠ longward of 1.4 SED based on our grism data is less than 10 and can µm(Oesch et al. 2014). be ruled out. sistent with the photometry (red points) and the best-fit SED at 4.2. The Number Density of Very Bright z>10 z =11.09 (gray line). The red horizontal errorbars represent the The previous AGHAST spectra already provided NoteGalaxies that in very similar grism observationswavelength for coverage a of the di↵erent HST filters, indicating that some evidence against strong emission line contam- source triply imaged by a CLASH foreground clus- the break of GN-z11 lies within the H160 band. Bottom – The ination (Oesch et al. 2014), andThe we spectrumalso obtained of GN-z11ter, emission indicates line contaminationthat its contin- could alsoredshift be excluded probability distribution functions, p(z), when fitting only Keck/MOSFIRE spectroscopy touum further break strengthen lies this within(Pirzkal the H160 et al.filter2015). (which We thus covers have no indicationto the broad-band cur- photometry (red) or when including both the conclusion (see appendix). However, the additional 12 rently that any of the recent z 9 11 galaxyphotometry candidates and the grism in the fits (blue). The photometric p(z) 1.4 1.7 µm; see Fig 6). The rest-frame UV continuum⇠ orbits of G141 grism data now conclusivelyflux⇠ of this rule galaxy out that is thereforeidentified with0.4 magHST brighteris a lower than redshift strongpeaks emission at significantly lower redshift, but contains an extended tail GN-z11 is such a lower redshift source. Assuming that all line contaminant⇠ (but see, e.g., Brammer etto al.z>201311., The for addition of the grism data significantly tightens the inferred from the H160 magnitude. The estimated abso- p(z)resultinginuncertaintiesofz 0.1. The fits that include the H-band flux came from one emission line, we would a possible z 12 candidate). ' have detected this line at > 10.lute Even magnitude when assuming is MUV = 22.1 ⇠0.2, which is roughly the old, shallower JH140 photometry are shown in gray, while those a more realistic case where the emissiona magnitude line flux brighter is dis- (i.e., a factor± 3 ) than the char- that use the grism data alone, without any photometric constraints, 3.3. Excluding⇥ a Lower-Redshift Dusty orare Quiescent shown in green. tributed over a combination of linesacteristic (e.g., H luminosity+[OIII]), of the UV luminosity functionGalaxy at we can confidently invalidate suchz a solution.7 8(Bouwens The lower et al. 2015b; Finkelstein et al. 2015). left panel in Figure 4 compares the measured⇠ grism spec- Another potential source of contamination for very With zgrism = 11.09, thehigh galaxy redshift GN-z11 galaxy is samples thus surpris- are dusty z detection2 3 sources of a galaxy this bright is not very constraining trum with that expected for theingly best-fit bright lower and redshift distant (see Figure 5). While one single ⇠given the large Poissonian uncertainties, it is interesting solution we had previously identified (Oesch et al. 2014). with strong 4000 A˚ or Balmer breaks (Oesch et al. 2012; A strong line emitter SED is clearly inconsistent with Hayes et al. 2012). However, the fact that the IRAC data the data. Apart from the emission lines, which we do for GN-z11 show that it has a very blue continuum long- not detect, this model also predicts weak continuum flux ward of 1.6 µm, together with the very red color in the across the whole wavelength range. At < 1.47 µm, this is WFC3/IR photometry, already rules out such a solution (see SED plot in Figure 4). Nevertheless, we additionally