A PILOT SURVEY FOR C III] EMISSION IN THE REIONIZATION ERA: GRAVITATIONALLY LENSED z ∼ 7–8 GALAXIES IN THE FRONTIER FIELDS CLUSTER ABELL 2744

The reionization of the intergalactic medium (IGM) represents a key phase in the evolution of the universe. Observations of high-redshift galaxies, which have charted a marked decline in the visibility of Lyα emission with redshift (Stark et al. 2010; Schenker et al. 2012), and those of $z\gt 5$ quasars which trace the redshift-dependent Gunn–Peterson absorption (Fan et al. 2006), indicate that cosmic reionization was largely complete by $z\sim 6$. The duration of the reionization process is constrained by the polarization of the microwave background due to Thomson scattering by electrons in the ionized era; recent data from the Planck satellite and results derived from the abundance and luminosity distribution of the $z\gt 6$ galaxy population now suggest that reionization was a rapid process which extended over $6\lt z\lt 10$ (Planck Collaboration et al. 2015; Robertson et al. 2015).

While Lyα emission (λ1216 Å) has proven to be the most valuable spectroscopic indicator for faint star-forming galaxies in the redshift range $4\lt z\lt 6$ (e.g., Stark et al. 2010, 2011), resonant scattering by neutral gas in the IGM likely renders this line ineffective as a reliable probe beyond $z\simeq 6.5$. Despite much observational effort, there are currently very few convincing cases of detected Lyα emission beyond $z\simeq 7$ (Ono et al. 2012; Finkelstein et al. 2013; Schenker et al. 2014; Vanzella et al. 2014a; Oesch et al. 2015) and several distant star-forming galaxies reveal no emission despite heroic exposure times (Vanzella et al. 2014b). As a result, Stark et al. (2014b) proposed it may be feasible to use metallic lines in the ultraviolet (UV) as alternative spectroscopic indicators. Examining the spectra of 17 gravitationally lensed low-luminosity galaxies at $z\simeq 1.5-3$, they discuss the feasibility of searching for C iii] ($\lambda \lambda$1907, 1909 Å) and C iv ($\lambda \lambda$1548, 1550 Å) emission. Although such metallic lines are normally much weaker than Lyα in luminous systems, in young metal-poor low luminosity systems characteristic of those at high redshifts, these lines may become relatively more prominent. In their sample of 17 $z\simeq 1.5-3$ galaxies, Stark et al. (2015) find C iii] emission has an equivalent width (EW) which correlates with that of Lyα and is typically 10 times weaker. As an encouraging proof of concept, Stark et al. (2014) recently claimed tentative detections of C iii] emission in two $J\sim 25.2$ galaxies with pre-determined Lyα emission at redshifts of z = 6.03 and z = 7.21.

However, as emphasized by Stark and collaborators, detecting C iii] emission in galaxies where its expected wavelength is a priori known from a Lyα redshift is less challenging than searching for emission across a wider range of wavelength governed only by a photometric redshift likelihood distribution, and given the density of skylines. Motivated by the interest in exploring the potential of this, possibly the only, immediate route to spectroscopic progress in the reionization era, we have embarked on a statistical search. Our plan is examine the spectra of a sample of gravitationally lensed sources in the redshift range $z\;\sim$ 6.7–8.5 derived from recent compilations in several massive clusters (e.g., Atek et al. 2014; Bradley et al. 2014; Coe et al. 2015; Zheng et al. 2014). Such a statistical approach is now possible due to the arrival of multi-slit near-infrared spectrographs such as Keck's Multi-Object Spectrometer For Infra-Red Exploration (MOSFIRE; McLean et al. 2012). In this paper, we examine the practicality of the method with realistic exposure times for the Frontier Field cluster Abell 2744, and also include preliminary data for two other CLASH/HFF clusters observed to shallower depths.

The paper is organized as follows. In Section 2 we overview the sample, observations, and data reduction. In Section 3 we discuss the results, summarized in Section 4. Throughout the work we use a standard ΛCDM cosmology with (${{{\Omega }}_{m0}}=0.3$, ${{{\Omega }}_{{\Lambda }0}}=0.7$, ${{H}_{0}}=100$ h km s⁻¹ Mpc⁻¹, with h = 0.7), and magnitudes are given using the AB convention. Cluster names are abbreviated to "A" or "M" for Abell (e.g., Abell et al. 1989) and MAssive Cluster Survey (MACS; e.g., Ebeling et al. 2010) clusters, respectively, followed by their ID.

2.1. Target Selection

We constructed a sample of high-z candidates magnified by the galaxy cluster A2744 using photometry from Hubble Frontier Fields program (Lotz et al. 2014). Candidates for inclusion in our multi-slit mask were derived from Zheng et al. (2014), Coe et al. (2015) and Atek et al. (2015, 2014, see also Ishigaki et al. 2015). We pre-selected targets of known magnification down to an apparent magnitude of ${{H}_{160}}\sim 28$ with photometric redshifts in the range $6.7\lt z\lt 8.5$, corresponding to the visibility of C iii] within MOSFIRE's H-band filter. We also included the $z\sim 9.8$ candidate from Zitrin et al. (2014) to explore the option of detecting the C iv $\lambda \lambda$ (1548, 1550) Å doublet in the same band. This follows a promising detection of C iv emission at $z\simeq 7.05$ by Stark et al. (2015). Photometric redshift likelihood distributions, P(z), were obtained with the Bayesian Photometric Redshift code (BPZ; Benítez et al. 2004; Coe et al. 2006; see also Zheng et al. 2014), covering the full redshift range available to HST (from z = 0.01 to z = 12 in ${\Delta }z=0.001$ increments). Due to slit-mask positioning constraints, only a subset of good candidates could be included on the MOSFIRE mask and priority was given to brighter galaxies. The final mask included eight high-z candidates as summarized in Table 1. The C iii] candidates lie mainly in the magnitude range $26.2\lt H\lt 27.5$ and have photo-z uncertainties of $\delta \;z\simeq 0.2-0.3$. The rest-frame UV luminosities corrected for lensing magnifications have a mean of ${{M}_{{\rm UV}}}\simeq -18.8$ and standard deviation of 1.4. As part of this campaign we also have begun observations of two further lensing clusters, MACS 0416 and A2261, drawing candidates from the catalogs of Bradley et al. (2014) and Coe et al. (2015, see also McLeod et al. 2014). As the photometric redshift distributions of these galaxies are somewhat less secure (especially for A2261), and since our observations of these clusters are significantly shallower, they currently provide less useful constraints on the presence of C iii], although we incorporate the results in this paper. We also list these additional sources in Table 1. Figure 1 summarizes the redshift and UV luminosity distribution of the total sample in the context of the H-band window available for detecting C iii] with MOSFIRE.

Zoom In Zoom Out Reset image size

Table 1.  The Sample

ID	α(degree)	δ(degree)	Phot-z	H160	μ	β	${{M}_{{\rm UV},1500}}$
A2744-YD7a,b	3.603397	−30.382256	$8.3_{-0.1}^{+0.1}$	26.17 ± 0.03	$1.4_{-0.1}^{+0.7}$	−1.38 ± 1.86	$-20.65_{-0.08}^{+0.54}$
A2744-ZD3a,b,c	3.606477	−30.380993	$7.7_{-0.3}^{+0.2}$	26.45 ± 0.04	$1.3_{-0.1}^{+1.0}$	−1.14 ± 0.26	$-20.19_{-0.19}^{+0.85}$
A2744-ZD9a	3.603208	−30.410368	$7.0_{-0.2}^{+0.2}$	26.48 ± 0.04	$3.4_{-0.8}^{+0.8}$	−1.17 ± 0.23	$-18.88_{-0.37}^{+0.37}$
A2744-ZD7A2a	3.592160	−30.409925	$7.3_{-0.5}^{+0.2}$	28.18 ± 0.04	$6.4_{-2.2}^{+7.8}$	−1.29 ± 1.22	$-16.63_{-0.42}^{+1.34}$
A2744-YD8a,b,c	3.596096	−30.385832	$8.1_{-0.1}^{+0.2}$	26.65 ± 0.04	$1.9_{-0.2}^{+1.0}$	−1.84 ± 1.64	$-19.86_{-0.13}^{+0.57}$
Atek-3772c	3.5978343	−30.395960	$7.0_{-0.6}^{+0.3}$	27.45 ± 0.05	∼6.8	−1.77 ± 1.00	$-17.51_{-0.24}^{+0.24}$
Atek-5918c,e	3.5951375	−30.381131	$7.7_{-0.6}^{+0.6}$	26.92 ± 0.02	∼3.5	−1.07 ± 0.19	$-18.65_{-0.27}^{+0.27}$
A2744-JDBd	3.5950200	−30.400750	$9.8_{-0.4}^{+0.2}$	27.30 ± 0.07	$11.3_{-2.5}^{+4.8}$	⋯	$\sim -17.6$
A2261-0450f	260.6124593	32.1438429	$6.8_{-0.3}^{+0.2}$	25.5 ± 0.06	∼5.6	−1.85 ± 0.15	$-19.50_{-0.23}^{+0.23}$
A2261-0731f	260.6232556	32.1393984	$6.9_{-5.9}^{+1.0}$	27.9 ± 0.22	∼7.7	−1.00 ± 0.67	$-16.65_{-0.34}^{+0.34}$
A2261-0772f	260.6059024	32.1388049	$6.5_{-5.4}^{+0.8}$	27.4 ± 0.19	∼6.3	−2.17 ± 0.64	$-17.35_{-0.30}^{+0.30}$
A2261-0187f	260.6073833	32.1495175	$7.5_{-1.2}^{+0.4}$	27.0 ± 0.13	∼2.9	−1.18 ± 1.47	$-18.89_{-0.25}^{+0.25}$
MACS 0416-0036f	64.0260447	−24.0509958	$7.0_{-6.0}^{+1.2}$	26.8 ± 0.16	∼1.3	−0.56 ± 0.45	$-19.46_{-0.45}^{+0.45}$
Zheng-4008g	64.0603330	−24.064960	$7.7_{-0.3}^{+0.3}$	27.85 ± 0.08	$2.2_{-0.3}^{+0.3}$	−3.49 ± 1.33	$-18.59_{-0.28}^{+0.28}$
FFC2-1151-4540b,g	64.0479780	−24.081678	$8.3_{-0.2}^{+0.2}$	26.59 ± 0.03	$1.8_{-0.5}^{+0.5}$	−1.44 ± 0.69	$-19.93_{-0.31}^{+0.31}$

Note. Column 1: dropout's ID and references. The first work cited for each object represents the original source of photometric data and analysis, although in some cases we made adjustments to enhance consistency across the sample. Columns 2 and 3: R.A. and decl. in J2000.0. Column 4: photometric redshift and 95% errors. Column 5: HST's apparent H₁₆₀-band magnitude. Column 6: lensing magnification. If no error is listed a nominal $\sim 20\%$ error is adopted (Zitrin et al. 2015). Column 7: UV-slope, β ($1\sigma$ errors), calculated by a weighted least-squares fit. Column 8: absolute magnitude, M_UV, at $\lambda =1500$ Å, calculated from the said ${{F}_{\lambda }}\propto {{\lambda }^{\beta }}$ fit, where the error includes in quadrature the discrepancy from the absolute magnitude obtained by translating the flux in the WFC3 band containing the redshifted $\lambda =1500$ Å, and the propagated photometric and magnification errors.

^aZheng et al. (2014). ^bCoe et al. (2015). ^cAtek et al. (2015, 2014). ^dZitrin et al. (2014). C iv target. ^eOur independent photo-z estimate permits a solution at $z\simeq 2$. ^fBradley et al. (2014). ^gW. Zheng (private communication; see also Infante et al., in preparation).

Download table as:  ASCIITypeset image

2.2. Observations and Data Reduction

All reduced spectra were visually inspected given the expected wavelength range where C iii] might be visible according to the photometric redshift likelihood function. No convincing line was seen for any of the seven C iii] candidates. We thus seek to determine the likely range of fluxes and EW for C iii] consistent with our non-detections.⁶ In other words, we estimate the probability of detecting at least one C iii] line in our survey, as a function of a given mean total flux and EW. Since C iii] is a doublet, we assume a line ratio C iii] 1907/1909 Å of 1.4 (Stark et al. 2015). We consider total line strengths in the range $0-4\times {{10}^{-18}}$ ergs cm⁻² s⁻¹ in 5 × 10⁻²⁰ ergs cm⁻² s⁻¹ increments as illustrated in Figure 2. For each doublet line, and for each redshift step, we checked if its input flux would exceed a certain detection significance ($x\sigma$) in the corresponding wavelength in the observer frame, where x is a chosen signal-to-noise ratio. For a fixed line flux, the likelihood of having a detection with $x{{\sigma }_{k}}$ significance in the examined slit k is given by the number of spectral pixels with positive detections over the total number of pixels, weighted by the redshift probability function P(z). The relevant expression for slit k can be formulated as:

$$\begin{eqnarray}&&{{P}_{{\rm det} ,k}}({{F}_{{\rm in}}},x)=\frac{{{\sum }_{i}}{{P}_{k}}({{z}_{i}}){\Theta }\left( {{F}_{{\rm in}}},x,{{\sigma }_{k}},{{z}_{i}} \right)}{{{\sum }_{i}}{{P}_{k}}({{z}_{i}})}\ ,\end{eqnarray} \tag{ 1 }$$

where the sum is over all the redshift steps z_i (z = 0 to z = 12 in 0.001 increments), and Θ is defined as

$$\begin{eqnarray*}{\Theta }\left( {{F}_{{\rm in}}},x,{{\sigma }_{k}},{{z}_{i}} \right)=\left\{ \begin{array}{ccccccccccccccc} 1 & {\rm if}\;\;\;{{F}_{{\rm in}}}\gt x{{\sigma }_{k}}(1907\;\overset{\circ}{\rm A} \cdot (1+{{z}_{i}})) \\ {} & {\rm or}\;\;\;{{F}_{{\rm in}}}/1.4\gt x{{\sigma }_{k}}(1909\;\overset{\circ}{\rm A} \cdot (1+{{z}_{i}})) \\ 0 & {\rm otherwise}, \\ \end{array} \right.\end{eqnarray*}$$

and we define ${{\sigma }_{k}}=\infty$ for z_i steps placing the line outside the MOSFIRE H band. ${{P}_{{\rm det} ,k}}$ therefore provides a conditional probability, i.e., the chance of detecting at least one of the two C iii] lines for a given target k with redshift probability distribution P_k(z), given the limiting noise in our spectra for the mask, ${{\sigma }_{k}}(\lambda )$ (see Section 2), and as a function of the input line flux F_in and the detection significance x. Finally, the probability of detecting at least one line of a given flux F_in over the entire sample is:

$$\begin{eqnarray}&&{{P}_{{\rm sample}}}({{F}_{{\rm in}}},x)=1-\mathop{\prod }\limits_{k}\left( 1-{{P}_{{\rm det} ,k}}({{F}_{{\rm in}}},x) \right),\end{eqnarray} \tag{ 2 }$$

where the product is over all slits.

Zoom In Zoom Out Reset image size

We repeat the above process also in terms of restframe EW, where in each iteration, instead of running over a range of input fluxes, we run over a range of restframe EWs, translated in each iteration, for each object individually, to the corresponding input flux.

Figure 3 shows the probability, for both 3 and 5σ detections, of finding at least one line in our A2744 survey as a function of the mean total C iii] flux and rest-frame EW. For example we have a 95% chance of detecting at least one C iii] line in our MOSFIRE survey at 5σ significance, if the typical C iii] λ1907 Å line flux is $\simeq 1.5\times {{10}^{-18}}$ ergs cm⁻² s⁻¹ (total doublet flux of $\simeq 2.6\times {{10}^{-18}}$ ergs cm⁻² s⁻¹), or equivalently, if the rest-frame EW for the combined C iii] doublet is 26 ± 5 Å or higher. In this estimate we have included limits from the shallower exposures on A2261 and M0416, but the results remain similar (to within typically 5%) if the sample is restricted to A2744—for which photometric redshift errors are typically smaller and the observations are significantly deeper. Errors were propagated assuming our adopted 20% uncertainty in the flux calibration.

Zoom In Zoom Out Reset image size

For comparison, Stark et al. (2015) detected, with $3.3\sigma$, a λ1909 Å C iii] line of $\simeq 4.2\pm 1.2\times {{10}^{-18}}$ ergs cm⁻² s⁻¹ (total estimated C iii] flux $\simeq 1.1\pm 0.3\times {{10}^{-17}}$) in a z = 6.03 galaxy (J = 25.2), and a $2.8\sigma$ C iii] detection of likely λ1909 Å of $\simeq 0.9\pm 0.3\times {{10}^{-18}}$ ergs cm⁻² s⁻¹ (total estimated C iii] flux $\simeq 2.3\pm 0.5\times {{10}^{-18}}$) in a z = 7.21 galaxy (J = 25.2). The total C iii] restframe EWs of these detections are 22.5 ± 7.1 Å and 7.6 ± 2.8 Å, respectively.

While our observational limits are deep enough to recover similar line fluxes to those found by Stark et al. (2015, see also Stark et al. 2014), our galaxies are fainter (∼27 AB), so that our limits on the rest-frame EWs are less constraining. Assuming a C iii] EW of 22.5(7.6) Å as was found by Stark, we have $\sim 99.9\%(10\%)$ chance of detecting at least one such line in our sample with 3σ, or $\sim 90\%(0\%)$ for 5σ. Thus, it is quite likely that the primary reason for the non-detection in our survey is that, on average, the present sample is significantly fainter than those targeted by Stark et al., which also had the benefit of secure Lyα-based redshifts. The main conclusion of our limits seen in Figures 3 and 4 is that even with a more ambitious spectroscopic campaign that would likely increase the exposure time by a factor ×4 (corresponding to a three night integration on one mask), only more luminous $z\simeq 7$–8 galaxies in the reionization era would appear to be amenable for study with any reliability. Alternatively, brighter and/or more highly magnified examples, such as those close to the critical line of a foreground cluster, might provide promising targets although generally such sources are rare. It is interesting to note the non-detection (and upper limit) on C iii] emission recently claimed by Watson et al. (2015) for a brighter source with H = 24.7, magnified by $\mu \sim 10$ at z = 7.5, showing that even for significantly brighter objects C iii] detection can be challenging. Searching for bright magnified dropouts in a very large sample of clusters, for example, is desirable for progress with current facilities and might help deliver James Webb Space Telescope with first light targets.

Zoom In Zoom Out Reset image size

For completeness, we also calculate the limit on the C iv $\lambda \lambda$ (1548, 1550) Å doublet for the $z\simeq 9.8$ multiply imaged object discovered by Zitrin et al. (2014) behind A2744 (Figure 3). Stark et al. (2014) found prominent C iv emission in some of the $z\sim 2$ galaxies they targeted, and highlighted C iv as an additional promising diagnostic for high-redshift galaxies. Typically they found C iv line fluxes only a factor of about 2 weaker than those of C iii]. At the proposed redshift of the Zitrin et al. (2014) object, the doublet would be readily resolved and we assume both lines have equal strength. In this case, we determine that, with $\sim 90\%$ confidence, the line flux for either one of the two C iv lines for a detection significance of 5σ is less than $\simeq 3.6\times {{10}^{-18}}$ ergs cm⁻² s⁻¹, and the rest-frame EW less than $\simeq 32$ Å. This translates to a magnification-corrected, total C iv luminosity of $\lesssim 3.9\times {{10}^{41}}$ erg s⁻¹ at z = 9.8. It would be interesting to investigate further the properties of C iv emission in a larger sample. Note also Stark et al. (2015) have now detected a promising a C iv $\lambda 1548$ Å line in a $z\simeq 7.05$ object, corresponding to a restframe EW of $\simeq 18.1$ Å (a total C iv restframe EW of 38 Å). As in the C iii] case, despite reaching deep enough to detect a similar line flux, $\simeq 4.1\times {{10}^{-18}}$ ergs cm⁻² s⁻¹, in terms of EW the non-detection is consistent with the limits obtained from the single $z\sim 10$ object.

Given the attenuation of Lyα by neutral gas in the reionization era, the C iii] doublet has been proposed as a promising route toward spectroscopic verification and study of high-redshift candidates (Stark et al. 2015, 2014). We report results from a short campaign with Keck/MOSFIRE to assess the prospects of detecting C iii] lines in a sample of faint gravitationally lensed $z\sim 7-8$ galaxies where Lyα is not seen and thus the search window in wavelength is much larger than in earlier work. We observed 14 high-z candidates magnified by three galaxy clusters. For our deepest field (A2744, with seven C iii] candidates), we reached a 5σ(3σ) flux limit of $1.5(0.9)\times {{10}^{-18}}$ ergs cm⁻² s⁻¹ but did not detect any convincing line. Using a statistical method employing data from our collective campaign, we provide upper limits on the typical C iii] line flux and its rest-frame EW. Although our limits reach the line fluxes observed in some actual C iii] detections claimed in the recent literature, because our sample is significantly fainter in apparent magnitude, we only marginally reach the expected EWs based on these recent detections. This demonstrates the challenge of continuing the present investigation with current observing facilities unless either (i) brighter or more strongly-lensed sources are targeted and/or (ii) the C iii] is found to be more prominent in intrinsically fainter systems (e.g., Stark et al. 2014b). More data is needed to test this latter suggestion.

We thank the reviewer of this work for important comments. A.Z. is grateful for data products generated by Dan Coe and the CLASH team (PI: M. Postman), and for data supplied to us by Wei Zheng, Alberto Molino and group (HST grant AR 13279; PI: W. Zheng). A.Z. thanks Carrie Bridge and Adam Miller for helpful discussions, and Chuck Steidel for sharing his nominal calibration files. Support for this work was provided by NASA through Hubble Fellowship grant #HST-HF2-51334.001 A awarded by STScI, which is operated by the Association of Universities for Research in Astronomy, Inc. under NASA contract NAS 5-26555. This work is in part based on previous observations made with the NASA/ESA Hubble Space Telescope. The data presented herein were obtained at the W. M. Keck Observatory. The authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of Mauna Kea has always had within the indigenous Hawaiian community. We are most fortunate to have the opportunity to conduct observations from this mountain.