THROUGH THE LOOKING GLASS: HST SPECTROSCOPY OF FAINT GALAXIES LENSED BY THE FRONTIER FIELDS CLUSTER MACSJ0717.5+3745

The emergence of the first galaxies from the mist of cosmic dawn is one of the major outstanding questions in current astrophysics. Measurements of the cosmic microwave background anisotropy indicate that the universe was reionized during the redshift range z ∼ 7–12 (e.g., Planck-Collaboration et al. 2013; Bennett et al. 2013; Hinshaw et al. 2013). The first galaxies were most likely the source of reionizing photons, although this has not been conclusively proven given the observational and theoretical uncertainties (Robertson et al. 2013; Schmidt et al. 2014).

Finding and spectroscopically confirming the first galaxies is essential not only to identify the sources of reionization, but also to understand the physical mechanisms at work in the early star-forming regions, the interstellar and circumgalactic media. Much progress has been achieved with photometric studies. Large samples of candidate galaxies at z ≳ 8 have been identified in blank fields or behind clusters using the Lyman break technique or spectral energy distribution (SED) fitting techniques (Bouwens et al. 2010; Ellis et al. 2013; Oesch et al. 2013; Coe et al. 2013; Schmidt et al. 2014).

In contrast, spectroscopic confirmation has been much harder to achieve. In spite of numerous attempts on modern and sensitive ground-based spectrographs, very few Lyα detections have been reported at z ∼ 7 (e.g., Pentericci et al. 2011; Bradač et al. 2012; Ono et al. 2012; Schenker et al. 2012; Capak et al. 2013; Caruana et al. 2013; Treu et al. 2013) and only one at z ≳ 7.5 (Finkelstein et al. 2013, at z = 7.51). The decrease in Lyα flux coming from Lyman break galaxies (LBGs) at z ≳ 6 has been interpreted as evidence for an increased hydrogen neutral fraction corresponding to a long ending tail of cosmic reionization (e.g., Fontana et al. 2010; Jensen et al. 2013; Taylor & Lidz 2014). Alternative explanations for the decrease in Lyα emission include changes in the local circumgalactic medium (e.g., Dijkstra et al. 2007; Jones et al. 2012) or perhaps a dramatic increase in the fraction of interlopers amongst LBGs. Sensitive and complete spectroscopic surveys sensitive to Lyα at z ≳ 6 are needed to make progress.

A powerful alternative to ground-based spectroscopy is slitless spectroscopy with the Hubble Space Telescope (HST). Basic advantages over the ground are the absence of night sky emission lines and of atmospheric absorption. Furthermore, slitless spectroscopy does not require preselection of targets for inclusion in masks and it is therefore straightforward to obtain complete samples. These two advantages make Hubble competitive for the spectroscopic study of galaxies at the epoch of cosmic reionization provided that long enough exposures are obtained, assisted by lensing magnification (Treu et al. 2012).

Studying the very high-redshift universe with the combination of the HST grism spectroscopy and gravitational lensing magnification is one of the key goals of the Grism Lens-Amplified Survey from Space (GLASS), which we introduce in this Letter. We briefly present the survey, its goals, and the observational strategy and Wide Field Camera 3 (WFC3) infrared observations of the first targeted cluster MACSJ0717.5+3745. In order to demonstrate the performance of the HST grism in a crowded field, we concentrate on spectroscopy of faint targets, including multiply-imaged emission-line galaxies and photometrically selected galaxy candidates at z ≳ 6. We also present additional ground-based spectroscopy obtained with the MOSFIRE spectrograph on the W. M. Keck-I 10 m Telescope.

We adopt a standard cosmology with Ω_m = 0.3, Ω_Λ = 0.7, h = 0.7.

GLASS¹⁶ (GO-13459; PI: Treu) is a 140 orbit spectroscopic survey with HST. Using WFC3's G102 and G141 infrared grisms, GLASS will obtain slitless spectroscopy of the cores of 10 galaxy clusters, with uninterrupted wavelength coverage in the wavelength range 0.8–1.7 μm. Pre-imaging through filters F105W and F140W is obtained before each spectral exposure to assist in the extraction of the spectra and modeling the contamination from neighboring sources. The total exposure time per cluster is 10 orbits in F105W+G102 and 4 in F140W+G141. Each cluster is observed at two orientations, differing by approximately 90°, in order to facilitate deblending and extraction of the spectra. Parallel observations with the Advanced Camera for Surveys through filter F814W and grism G800L are carried out for each cluster in order to map the cluster infall regions. The three key goals of GLASS are (1) the study of galaxies at the epoch of reionization through the detection of Lyα at z ≳ 6; (2) the study of the cycling of gas and metals in and out of galaxies at 2 < z < 4; and (3) the study of star formation and metallicity as a function of environment in the foreground clusters and infall regions. Magnification by the foreground clusters greatly assists the first two science goals as well as a number of other ancillary science goals (e.g., lensed supernovae, red quiescent galaxies, etc.).

The GLASS data presented provide the first orientation (i.e., half of the final data) of the cluster MACSJ0717.5+3745 and were carried out on 2013 December 24 and 30. The total exposure times are 10029 and 3812 s with the WFC3 G102 and G141 grisms, after removal of a few reads affected by significantly elevated backgrounds (G. B. Brammer et al. 2014, in preparation). In addition, we obtain 1979 and 712 s of direct imaging in the WFC3 F105W and F140W filters, aiding the alignment of the grism exposures.

MACSJ0717.5+3745 is part of the Frontier Fields initiative¹⁷ and has been observed extensively by CLASH (Postman et al. 2012). We take advantage of the CLASH photometry when selecting our LBG samples as described in Section 5. In the top panel of Figure 1, we show a false-color image using the CLASH imaging of MACSJ0717.5+3745 with the position of the GLASS MACSJ0717.5+3745 field of view (FOV) presented here marked by the magenta square.

Zoom In Zoom Out Reset image size

The GLASS observations were designed to follow the 3D-HST observing strategy (Brammer et al. 2012), and were processed with the 3D-HST reduction pipeline.¹⁸ In the following, we summarize the main reduction steps but refer to Brammer et al. (2012) for further details.

The data were taken in a four-point dither pattern identical to the one shown in Figure 3 of Brammer et al. (2012). At each individual dither position, a F105W-G102 or F141W-G141 pair of exposures was taken, in order to optimize rejection of bad pixels and cosmic rays and improve sampling of the WFC3 point-spread function.

The individual exposures were turned into a mosaic using MultiDrizzle (Koekemoer et al. 2003) adjusting the alignments using the PyRAF routine, tweakshifts. The background was subtracted from the direct images fitting a second-order polynomial to each of the source-subtracted exposures. For the G102 background subtraction we used the master background presented by Kümmel et al. (2011), whereas for the G141 grism we used the master backgrounds developed by Brammer et al. (2012) for 3D-HST.

The individual background-subtracted exposures were then combined to produce a mosaic with a final pixel scale of roughly 006 pixel⁻¹ as described by Brammer et al. (2013; ∼half a native pixel). This corresponds to ∼12(22) Å pixel⁻¹ for the G102(G141) dispersion. In the bottom panels of Figure 1, we show the resulting interlaced full FOV G102 (left) and G141 (right) grism images.

From these interlaced images, the position of each individual spectrum can be predicted by dispersing each pixel of each individual object in the SExtractor (Bertin & Arnouts 1996) segmentation maps created from the direct F105W and F140W images. In this way, individual spectra can be extracted and spectral contamination, i.e., flux from overlapping spectra from neighboring objects, can be accounted for.

3.1. Additional Ground-based Spectroscopy

We extracted spectra for all the sets of multiple images identified by Limousin et al. (2012; see also Zitrin et al. 2009; Medezinski et al. 2013, for additional models of the cluster) that lie within the GLASS FOV and do not have complete

sets of redshifts from the literature. For sets 3, 4, 12, and 14 as defined by Limousin et al. (2012), we detect the same emission lines from each image (see Table 1), thus confirming the lensing hypothesis. The postage stamp images and two-dimensional spectra of the four systems are shown in Figures 2 and 3. Contaminating spectra from nearby objects have been modeled and subtracted from these spectra. MOSFIRE spectra are shown together with GLASS HST spectra when available. We also detect with MOSFIRE the [O iii] 4959, 5007 doublet in emission from image 6.1, which falls outside of the GLASS FOV, placing it at z = 2.393. The redshifts of image sets 3 and 14 agree with the value for 3.2 and 14.1 (z = 1.85) previously measured by Limousin et al. (2012).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Note the spatial extent of the emission lines, e.g., for 4.1. In such cases, the combination of HST's angular resolution and lensing magnification will allow us to measure spatially resolved metallicity and star formation gradients with sub-kiloparsec resolution (T. A. Jones et al. 2014, in preparation).

The confirmation of the lensing hypothesis and the spectroscopic redshifts being close to the photometric redshift estimates (2.0 ± 0.1, 2.1 ± 0.1, and 1.8 ± 0.1, respectively, for sets 4, 6, and 12) further validates the lens models published and available as part of the Frontier Fields initiative. When the GLASS data set of MACSJ0717.5+3745 is complete, we will carry out a systematic search for new sets of multiple image systems based on spectroscopic data. The use of spectroscopic data greatly reduces the space of possible counter images and therefore should enable the unambiguous identification of faint multiple images. In combination with upcoming deep imaging from the Frontier Fields initiative, the additional data will allow us to further refine the lens models, the magnification maps, and study the distribution of luminous and dark matter in the cluster itself.

5.1. Photometric Selection

5.2. Flux, Luminosity, and Equivalent Width Limits

Out of the 21 LBG candidates 14 fell within the FOV of the G102 and G141 GLASS pointings analyzed here. We do not detect any line emission. Nine of the GLASS spectra were free of contamination by spectra from other objects and were used to estimate the flux sensitivity.²⁰ For an aperture of 5 (spatial) by 3 (spectral) native pixels (∼06 × 72(132) Å for G102(G141)), the 1σ limiting flux is of the order of 10⁻¹⁷ erg s⁻¹ cm⁻² as shown in the top panel of Figure 4. The sensitivity is comparable with what is reached in blank fields (e.g., Atek et al. 2010) and agrees well with the exposure time calculator.

Zoom In Zoom Out Reset image size

One of the goals of GLASS is to take advantage of the lensing magnification of high-redshift sources to enhance the probability of detecting otherwise unreachable flux limits. To illustrate the benefits of magnification, the bottom panel of Figure 4 shows the limiting flux for the nine spectra corrected for the estimated magnification, assuming that they are at a redshift of 6, 6.4, or 7 depending on their selection method. Note that magnification changes very little at these redshifts, owing to the flatness of the angular diameter distance–redshift relation. It is clear that a significant improvement in the effective limiting fluxes is achieved from the cluster lensing magnification. However, it is also clear that a relatively high magnification is needed to obtain several-σ detections of line fluxes in the ∼10⁻¹⁸ erg s⁻¹ cm⁻² regime, which is the deepest practical limit for ground-based optical and near-infrared studies of field galaxies (e.g., Stark et al. 2011; Treu et al. 2013). Naturally, owing to the low and smooth background from space, the sensitivity of GLASS is more or less independent of wavelength except at the edges of the spectral coverage where the grisms throughput drop.

Assuming the SED of the sources to be approximately flat in f_ν over the G102 and G141 wavelength range, we can estimate the continuum flux for each source as a function of wavelength from a weighted average of the CLASH F125W, F140W, and F160W magnitudes. In combination with the limiting flux, the continuum flux gives the limiting equivalent width for each of the nine objects with suitable spectra. The rest-frame equivalent width limits at the fiducial redshift for each candidate are given in Table 1.

Given the intrinsic faintness of the candidates, the equivalent width limits are rather high and it is thus not surprising that we do not detect any Lyα emission for this sample. Even at z ∼ 6 only ≲ 4% of LBGs would be expected to have emission larger than 100 Å and thus be significantly detected in our exposures. However, the equivalent width limits are sufficient to rule out pure line emitters at lower redshift as contaminants (Atek et al. 2011).

We have introduced GLASS, an HST Large Program devoted to the study of the high-redshift universe by means of deep grism spectroscopy. Using the first WFC3 pointing in the center of the rich cluster MACSJ0717.5+3745, we have demonstrated the power of grism spectroscopy even in very crowded fields. We have shown that by carefully modeling the background contamination, one can obtain deep spectra of large numbers of sources at once, without the limitations due to preselection for inclusion in masks or wavelength gaps typical of ground-based studies. In practice, we have considered two sets of faint targets and obtained the following results.

• 1.  
  We have confirmed four sets (3, 4, 12, and 14) of candidate strong gravitational lens systems by identifying the same emission lines in each of the multiple images. Redshift for one additional system (6.1) was obtained from ground-based MOSFIRE spectroscopy. These five redshifts are close to the photometric redshift estimates used to construct the lens models provided by the Frontier Fields initiative. The confirmation of the strong lensing hypothesis and redshift thus provides additional validation of the models.
• 2.  
  We have compiled a list of 14 photometrically selected candidate galaxies at z ∼ 6–8 that fall in the GLASS FOV. No Lyα emission is detected.
• 3.  
  We have used the spectra of the nine candidates with no significant contamination from neighboring objects to measure the line flux sensitivity of GLASS and found it to be 0.5–1 × 10⁻¹⁷ erg s⁻¹ cm⁻². This is consistent with the limits obtained in blank fields. The completeness and sensitivity are going to improve with the second set of visits, which will deliver the same exposure time with a dispersion direction that is approximately orthogonal to that of the data presented here.

Even with just a fraction of the total observing time allocated for GLASS, the observations illustrate the power of HST grism spectroscopy combined with strong gravitational lensing. The complete data set will provide a treasure trove of spectroscopic information, which will be useful to many in addressing several outstanding scientific questions.

This Letter is based on observations made with the NASA/ESA Hubble Space Telescope, obtained at STScI. We acknowledge support through grants HST-13459, HST-GO13177, and HST-AR13235. This work utilizes gravitational lensing models produced by PIs Bradač, Ebeling, Merten, Zitrin, Sharon, and Williams funded as part of the HST Frontier Fields program conducted by STScI. STScI is operated by AURA, Inc. under NASA contract NAS 5-26555. The lens models were obtained from the Mikulski Archive for Space Telescopes (MAST). The Dark Cosmology Centre (DARK) is funded by the Danish National Research Foundation. Some of 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.