SPECTROSCOPIC CONFIRMATION OF A z = 6.740 GALAXY BEHIND THE BULLET CLUSTER^*

The epoch of reionization, which marks the end of the "dark ages" and the transformation of the universe from opaque to transparent, is poorly understood. It is thought that z > 6 faint proto-galaxies were responsible for this transformation, but recent observations of z ≳ 7 objects (e.g., Robertson et al. 2010 for a review) complicate that scenario. Finding robust samples of sources, representative of the population contributing a significant amount of energetic photons, is crucial.

Wide Field Camera 3 (WFC3) on Hubble Space Telescope (HST) enables a preliminary identification of such galaxies. Substantial progress has been made in detecting z ≳ 7 galaxies using the dropout technique (Steidel et al. 1996), both in blank fields (HUDF, CANDELS, e.g., Bouwens et al. 2012; Oesch et al. 2012; Finkelstein et al. 2012; McLure et al. 2011), and behind galaxy clusters (e.g., Kneib et al. 2004; Egami et al. 2005; Bradley et al. 2012; Richard et al. 2011; Zheng et al. 2012; Zitrin et al. 2012, and references therein). One of the most obvious limitations of the dropout technique, however, is that unambiguously confirming the object's redshift usually requires spectroscopic follow-up. This is hard to do for typically faint high-z sources, and it is thus an area where gravitational lensing magnification helps greatly, as demonstrated in this Letter.

In addition to the redshift confirmation, spectroscopy provides information on properties of the interstellar and intergalactic media (ISM and IGM). In particular, Lyα emission from sources close to the reionization era is a valuable diagnostic given that it is easily erased by neutral gas within and around galaxies. Its observed strength in distant galaxies is a gauge of the time when reionization was completed (Robertson et al. 2010). Furthermore, we expect Lyα emitters (LAEs) to be predominantly dust-free galaxies; hence, their numbers should increase with redshift until the state of the IGM becomes neutral, at which point their numbers should decline.

Significant progress has been made in detecting LAEs in narrow band and spectroscopic surveys at z ≳ 6 (e.g., Kashikawa et al. 2006; Rhoads et al. 2012; Schenker et al. 2012; Clément et al. 2012; Curtis-Lake et al. 2012; Ono et al. 2012; Stark et al. 2011; Pentericci et al. 2011). Most studies see a decline in the LAE population at z > 7, but not all do (Krug et al. 2012; Tilvi et al. 2010). The declining fraction of LAEs within the Lyman break galaxy (LBG) population (Stark et al. 2010; Kashikawa et al. 2011; Pentericci et al. 2011) is consistent with this decline being due to changes in the ISM/IGM, specifically to an increased amount of neutral gas. However, current studies only probe the bright end of the luminosity function of LBGs.

Furthermore, as noted by Dijkstra et al. (2011) and Dayal & Ferrara (2012), measuring the rest-frame equivalent width (EW) distribution of LAEs as a function of redshift and luminosity is a powerful tool to study reionization. The EW distribution changes with redshift and source luminosity. Simulations suggest that reionization is the key factor driving this trend (Dayal & Ferrara 2012), because unlike continuum photons, the Lyα photons that escape the galactic environment are attenuated by the H i in the IGM. With a measurement of the EW distribution in LAEs we can therefore help distinguish between effects of ISM dust and neutral IGM and study the epoch of reionization (see also Treu et al. 2012). The main missing observational ingredient is a measurement of the EW distribution for both luminous and sub-L* galaxies at the redshifts of reionization and this can only be achieved with spectroscopy.

Current spectroscopic observations unfortunately fall short of matching the extremely deep near-IR HST/WFC3 data for a significant sample of z ≳ 7 dropout selected galaxies. Even with state of the art facilities (e.g., the new spectrograph MOSFIRE on Keck; McLean et al. 2010) this will be a challenge. While samples of ≲ L* galaxies at z < 6.5 are steadily increasing (e.g., Richard et al. 2011; Schenker et al. 2012; Labbé et al. 2010 for spectroscopic and imaging detections), to date, very few ≲ L* galaxies at z ≳ 6.5 are spectroscopically confirmed. The only examples are a lensed z = 7.045 galaxy and a marginal detection at z = 6.905 by Schenker et al. (2012). At ∼L* a z = 6.944 galaxy was detected by Rhoads et al. (2012). At z ∼ 8 Lehnert et al. (2010) report a marginal detection of an emission line, but independent observations do not detect it (A. J. Bunker et al. 2012, in preparation). Other surveys (Shibuya et al. 2012; Ono et al. 2012; Vanzella et al. 2011; Pentericci et al. 2011; Fontana et al. 2010) target mostly brighter sources. It is important to increase the sample at z > 6.5 and compare it to z < 6.5 because the timescale for changing the number of LAEs and their observed EW distribution close to the reionization epoch is shorter than the interval of cosmic time between z ≃ 6 and z ≃ 7 (Dayal & Ferrara 2012).

A powerful way to detect emission lines from faint sources is to use galaxy clusters as cosmic telescopes (e.g., Treu 2010 for a recent review). Gravitational lensing magnifies solid angles while preserving colors and surface brightness. Thus, sources appear brighter than in the absence of lensing. The advantages of cosmic telescopes are that we can probe deeper (due to magnification), sources are practically always enlarged, and identification is further eased if sources are multiply imaged. Typically, one can gain several magnitudes of magnification, thus enabling the study of intrinsically lower-luminosity galaxies that we would otherwise not be able to detect with even the largest telescopes. Indeed the highest redshift sub-L* LBG currently spectroscopically confirmed is the z = 7.045 galaxy lensed by a cluster A1703 (Schenker et al. 2012). Observations using galaxy clusters as cosmic telescopes are consistently delivering record holders in the search for the highest redshift galaxies (Kneib et al. 2004; Bradley et al. 2008; Zheng et al. 2012). For this reason we have started a large campaign of spectroscopic follow-up of z > 6.5 candidates behind the best cosmic telescopes. In this Letter we present the first spectroscopic confirmation from this campaign: A z = 6.740 ± 0.003 galaxy behind the Bullet Cluster.

This Letter is structured as follows. In Section 2 we describe the data acquisition and reduction, in Section 3 we present the spectrum of the galaxy and summarize our conclusions in Section 4. Throughout the Letter we assume a ΛCDM cosmology with Ω_m = 0.3, Ω_Λ = 0.7, and Hubble constant H₀ = 70 km s⁻¹ Mpc⁻¹. Coordinates are given for the epoch J2000.0, magnitudes are in the AB system.

Our targets were selected from deep ACS/WFC3 HST observations and are presented by Hall et al. (2012). We targeted all 10 z_850LP-band dropouts with the FORS2 spectrograph on the ESO Very Large Telescope, Program ID 088.A-0542 (PI: Bradač). Here we present the first detection. The rest of the objects were not detected, inferences obtained from the non-detections and the spectra of filler slits will be presented in a subsequent paper.

The data were taken in service mode during 2011–2012 November–January. We used the 600Z holographic grating providing the highest sensitivity in the spectral range of 8000–10000 Å with a resolution R ∼ 1390 and a sampling of 1.6 Å pixel⁻¹ for a 1'' × 6'' slit. The spectra presented here come from the co-addition of 42 exposures of 1400 s of integration each, with median seeing around 07. Series of spectra were taken with two different masks, but all our main targets were placed on both masks for a total integration time of 16.3 hr. Standard flat-fielding, bias subtraction, sky subtraction, and wavelength calibration have been applied as in Vanzella et al. (2009, 2011). To perform sky subtraction we fit a polynomial to the partial spectra extracted from the slit just above and below the target and apply the fit to the full spectrum. All the sky-subtracted two-dimensional spectra (of a mask) were co-added in the pixel domain. Finally, spectra were flux calibrated using observations of spectrophotometric standards (Fontana et al. 2010). Slit losses are small, given the extremely compact size of the targets and good seeing conditions, and have been neglected in the subsequent discussion.

Most of the HST imaging data of the object are presented by Hall et al. (2012) and summarized in Table 1. In addition, we have reduced data from GO 11591 (PI: Kneib) and do not detect the object in I_814W. All 1σ limiting magnitudes are given in Table 1. We augment the HST data with imaging data from HAWK-I (Clément et al. 2012). The object is undetected at 3σ in the Y, NB1060, J, and K_s bands, which is consistent with our interpretation of the HST detections. For the purpose of determining photometric redshift we only use the HAWK-I non-detection in the K_s band (the other bands overlap with and are shallower than HST detections). The source is also undetected in all four Spitzer IRAC bands ([3.6 μm], [4.5 μm], [5.8 μm], and [8 μm]) with effective exposure times of 4 ks in each filter. This is not surprising given our detection limit for this object is $m_{H_{\rm 160W}}-m_{[3.6\,\mu \rm m]}=4$ at 3σ. Unfortunately, existing Spitzer images are too shallow to add to the quality of the photometric redshift estimate, and we only use the data listed in Table 1 in the spectral energy distribution (SED) fit (Figure 1).

Zoom In Zoom Out Reset image size

Table 1. Imaging and Spectroscopic Properties of z₈₅₀-band Dropout 10 from Hall et al. (2012)

R.A.	104.63015
Decl.	−55.970482
$m_{H_{\rm 160W}}$	26.37 ± 0.16
$m_{J_{\rm 110W}}$	26.5 ± 0.3
(J110W − H160W)	0.10 ± 0.15
(z850LP − J110W)	1.57 ± 0.68
$m_{V_{\rm 606}}$a	>28.75 (texp = 2336 s)
$m_{i_{\rm 775W}}$	>28.60 (texp = 10150 s)
$m_{I_{\rm 814W}}$	>29.00 (texp = 4480 s)
$m_{K_{\rm s}}$	>26.65 (texp = 3.75 hr)
μ	3.0 ± 0.2
$m^{\rm int}_{H_{\rm 160W}}$	27.57+0.17− 0.17
λ	9412 Å
z	6.740 ± 0.003
fb	(0.7 ± 0.1 ± 0.3) × 10−17 erg−1 s−1 cm−2
fλ, c	3.3+1.0− 0.8 × 10−20 erg−1 s−1 cm−2 Å−1
fint	(0.23 ± 0.03 ± 0.10 ± 0.02) × 10−17 erg−1 s−1 cm−2
fintλ, c	1.1+0.4− 0.3 × 10−20 erg−1 s−1 cm−2 Å−1
Wrest(Lyα)	30+12− 21 Å

Notes. ^aAll upper limits are 1σ limiting magnitudes calculated in 063 × 063 square apertures. t_exp is the exposure time in corresponding band. ^bf is the integrated line flux, f_{λ, c} is the continuum flux estimated from the J_110W magnitude, while int denotes corresponding intrinsic (unlensed) values.

Download table as:  ASCIITypeset image

For this we use the BPZ code (Benítez 2000) and assume uniform priors on the spectral types and redshift. The observed SED is best fit (reduced χ² = 0.3) by a young starburst (5 Myr) galaxy model at z_phot = 6.8^+1.6_{− 0.8} (95% confidence) which is in excellent agreement with the spectroscopic data presented below (Figure 2). The probability that the object is at low redshift given the set of templates used by BPZ is zero (Figure 1, right). Forcing the solution to z < 5 the best-fit SED is an elliptical galaxy template at z_phot = 1.4^+0.7_{− 0.6}. To further discard the low-z solution would require either extremely deep F475W (m_F475W = 30) or deeper Spitzer data, neither of which exists for this object. However, the reduced χ² of the fit for z < 5 increases by a factor of 10 compared to the high-redshift solution; in addition the object is unlikely an elliptical galaxy given the presence of an emission line (see below). We therefore conclude that the object is most likely at z > 5.

Zoom In Zoom Out Reset image size

We also performed an independent fit to the photometric data using the HyperZ code (Bolzonella et al. 2000), which has the advantage of treating internal extinction as a free parameter. The results are very similar and the probability that the object is at high redshift is >99%. Our conclusions are therefore robust even to cases of heavily reddened, dusty star-forming galaxies like the one in Gonzalez et al. (2010).

Finally, as noted in Hall et al. (2012) it is difficult to estimate the size of the object given the uncertainty in measuring low surface brightness objects. The object is resolved and we estimate the FWHM of ∼026 in J_110W and ∼021 in H_160W. In physical units and correcting for lensing this translates to ∼0.8 kpc which is consistent with the compact sizes reported at these high redshifts (e.g., Oesch et al. 2010).

The object we present here is the only one for which an emission line was detected, out of the 10 z-band dropout candidates listed in Hall et al. (2012). We detect an emission line at 9412 Å with >5σ significance (Figure 2). The line is detected in two different masks and is broader than cosmic rays or residuals due to sky subtraction, hence we are confident that the line is not an artifact. The integrated flux of the line is f = (0.7 ± 0.1 ± 0.3) × 10⁻¹⁷ erg⁻¹ s⁻¹ cm⁻², where the first error corresponds to statistical uncertainty in the detection. The second (larger) is due to systematic uncertainty in absolute flux calibration and due to flux losses in the proximity of skylines and was estimated based on previous multiple observations of various standard stars. No other emission lines are detected in the spectrum (7700 Å–10000 Å), which one would expect for some of the possible low-z solutions as discussed below. Based on this and on the SED fit we exclude other alternative explanations and conclude that the line is most likely Lyα at z = 6.740 ± 0.003. This agrees extremely well with the peak redshift probability distribution described in Section 2.

Because of the relatively low signal-to-noise ratio (S/N) of the spectrum we cannot determine whether the line is asymmetric (which is expected for high-redshift LAEs where absorption happens mostly in the blue wing of the line) and thus further test our identification as Lyα. However, we use indirect arguments to test the alternative hypothesis that this galaxy is at a low redshift and the line is not Lyα but (1) [O ii] (3727 Å) at z = 1.525; (2) [O iii] (4959 Å, 5007 Å) at z = 0.898, 0.880; or (3) Hα (6563 Å) at z = 0.434, which are the prominent lines in emission line galaxies (e.g., Straughn et al. 2009).

• 1.  
  Any of the three low-redshift scenarios are strongly disfavored by photometry. As shown in Figure 3 (right), the fit to the photometric data while forcing z < 5 is very poor (z_phot = 1.4^+0.7_{− 0.6}). Forcing it to z = 1.525 (if line is [O ii]), or accounting for the emission due to the line, does not change the quality of the fit. The [O ii] doublet should nominally be resolved at our resolution; however, the sky emission at these wavelengths degrades S/N and our ability to distinguish individual components (Figure 3, left). Unfortunately, in either scenario (Lyα/[O ii]) no other lines are expected in the wavelength range covered by the spectrum (7700 Å–10000 Å). The line has the rest-frame EW of ∼100 Å if at z = 1.5 ([O ii] scenario) and ∼150 Å if at z = 0.43 (Hα). This is higher than typical [O ii] and Hα emitter EWs as measured by Straughn et al. (2009). Their median EWs are 36 Å and 73 Å for [O ii] and Hα, respectively. Out of 30 [O ii] emitters only 3 have EW > 100 Å. Hence, given the photometry and strength of the line we conclude that the [O ii] scenario is very unlikely.
• 2.  
  [O iii] at z = 0.880 is even more strongly disfavored as we would expect to detect the second line of [O iii] as well as Hβ line for typical line ratios. That part of the spectra is clear, hence this identification is ruled out by the data.
• 3.  
  Hα at z = 0.434 is also strongly disfavored by the photometry.

Zoom In Zoom Out Reset image size

Recently, Hayes et al. (2012) cautioned against using photometry only to select high-redshift galaxies. They targeted a J_110W-band dropout from Laporte et al. (2011) and discovered it was a low-redshift interloper. However, their situation is different from the one reported here. While their target is a J_110W-band dropout by their selection criteria, the object is detected blueward of Lyα in the i_775W band and the resulting best-fit model is poor. Even without the i_775W-band observations the photometry can be fit with a low-z template. In our case the high-redshift SED fit is good and the low-redshift fit is extremely poor.

We conclude that the line is most likely Lyα at z = 6.740 ± 0.003. The object has an AB magnitude of 26.5 ± 0.3 in the J_110W band (lensed, before correcting for magnification), which corresponds to a continuum flux of f_{λ, c} = 3.3^+1.0_{− 0.8} × 10⁻²⁰ erg s⁻¹ cm⁻² Å⁻¹. With an integrated line flux of f = (0.7 ± 0.1 ± 0.3) × 10⁻¹⁷ erg⁻¹ s⁻¹ cm⁻², the resulting line rest-frame EW is W^rest(Lyα) = f/f_{λ, c}(1 + z) = 30⁺¹²_{− 21} Å (Table 1). The distribution of EW (including non-detections) of the total sample of 10 dropouts to be presented in a future paper will help distinguish between scenarios of Lyα opacity around the epoch of reionization (Treu et al. 2012).

Using the data we also obtain a rough estimate for the star formation rates (SFRs) using both Lyα and UV continuum luminosities, and Kennicutt's relations (Kennicutt 1998). We first estimate the SFR from the Lyα luminosity L_Lyα for the case B recombination theory as SFR_Lyα = 9.1 × L_Lyα[M_☉ yr⁻¹], with L_Lyα in units of 10⁴³ erg s⁻¹. We obtain SFR ∼ (3.3 ± 1.5)/μ M_☉ yr⁻¹, where errors only include measurement errors on L_Lyα and no uncertainties in Kennicutt's relation. Note that this is a lower limit since L_Lyα is not corrected for absorption effects which depend on various parameters, including the neutral fraction of the IGM and the kinematic status of neutral hydrogen.

To convert UV luminosity L_UV into SFR we use SFR_UV = 1.4 × L_UV[M_☉ yr⁻¹], where L_UV is in units of 10²⁸ erg s⁻¹ Hz⁻¹. We estimate L_UV = (6.2^+1.8_{− 1.5})10²⁸ erg s⁻¹ Hz⁻¹ using J_110W-band magnitude, as the central wavelength is very close to rest-frame 1500 Å, giving SFR_UV = 8.7^+2.6_{− 2.1}/μ M_☉ yr⁻¹. Both estimates are fully compatible within systematic and statistical uncertainties implying that the dust attenuation is very low (Verhamme et al. 2008), which is also indicated by the blue UV spectral slope from J_110W, H_160W, and K_s band photometry. Dividing SFR by magnification, the intrinsic SFR is SFR ∼ 2–3 M_☉ yr⁻¹, consistent with SFR from, e.g., Labbé et al. (2010).

We have presented deep VLT spectroscopy of a strongly lensed z_850LP-band dropout galaxy behind the Bullet Cluster. We detected an emission line at 9412 Å with >5σ significance, which we identify as Lyα at z = 6.740 ± 0.003 at >99% CL. Correcting for magnification (by a factor of μ = 3.0 ± 0.2 as discussed by Hall et al. 2012; Bradač et al. 2009), the intrinsic (unlensed) line flux is f = (0.23 ± 0.03 ± 0.10 ± 0.02) × 10⁻¹⁷ erg⁻¹ s⁻¹ cm⁻² (Table 1), which is ∼2–3 times fainter than the faintest spectroscopic detection of an LAE at z ∼ 7 (Schenker et al. 2012). Its intrinsic H_160W-band magnitude is $m^{\rm {int}}_{H_{\rm {160W}}}=27.57\pm 0.17$, corresponding to an intrinsic luminosity of 0.5 L* (where L* was calculated from the best-fit LBG luminosity function from Bouwens et al. 2011).

The source is undetected in the four IRAC bands, which is not surprising given that we would only be able to detect extremely red galaxies with $m_{H_{\rm {160W}}}-m_{{[3.6\, \mu \rm {m}]}}=4$ at 3σ in [3.6 μm] for sources this faint. For comparison, the z = 6.027 source behind A383 (Richard et al. 2011) with an unusually mature stellar population of ∼800 Myr and the multiply imaged z = 6.2 object (Zitrin et al. 2012) with a younger age ∼180 Myr have much bluer colors $m_{H_{\rm {160W}}}-m_{[3.6\, \mu \rm {m}]} \sim 1.5$. Deeper Spitzer data will be needed to probe the presence of mature stellar populations in the galaxy we present here and other systems at high redshift.

While this work presents only a single spectroscopic detection at z > 6.5, it nonetheless probes a very important region of parameter space. As noted above, measuring the EW distribution of LAEs as a function of redshift and luminosity is a very powerful tool to study reionization, because the latter is likely the key factor driving the trend of EW in luminosity (Dayal & Ferrara 2012). The main missing observational ingredient is a measurement of the EW distribution for both luminous and sub-L* galaxies at the redshifts of reionization.

Our source is the faintest one (in line flux) detected thus far and is only the second firm spectroscopic detection of a sub-L* source at z > 6.5. With future observations of dropouts magnified by cosmic telescopes we plan to further increase this sample. Once completed, this survey will help constrain the duration and physical processes occurring at the epoch of reionization.

We thank the anonymous referee for suggestions that greatly improved the Letter and Sam Schmidt, Michele Trenti, and Andy Bunker for stimulating discussions. Support for this work was provided by NASA through HST-GO-10200, HST-GO-10863, and HST-GO-11099 from STScI. E.V. acknowledges financial contribution from ASI-INAF I/009/10/0 and PRIN MIUR 2009 "Tracing the growth of structures in the universe: from the high-redshift cosmic web to galaxy clusters." T.T. acknowledges support from the NSF through CAREER/NSF-0642621, through a Sloan Research Fellowship, and by the Packard Fellowship. Part of the work was carried out by M.B. and T.T. while attending the program "First Galaxies and Faint Dwarfs" at KITP which is supported in part by the NSF under grant No. NSF PHY11-25915.