Dust in the Reionization Era: ALMA Observations of a z = 8.38 Gravitationally Lensed Galaxy

A deep ALMA Band 7 map (ID 2015.1.00594, PI: Laporte) of the FF cluster Abell 2744 centered at 0.84 mm (f_c = 356 GHz) was observed on 2016 July during 2.5 hr. The data were reduced using the CASA pipeline (McMullin et al. 2007) with a natural weighting and a pixel size of 004. Figure 1 reveals a source with greater than $4.0\sigma$ significance with a peak flux of 9.9 ± 2.3 × 10⁻⁵ Jy/beam (uncorrected for magnification). The uncertainty and significance level was computed from the rms measured across a representative ≈2 × 2 arcmin field. The signal is seen within two independent frequency ranges (center panels in Figure 1) and the significance level is comparable to that claimed for Watson et al's $z\sim 7.5$ lensed system, although its observed flux is six times fainter. Taking into account the different magnification factors (see later), the intrinsic (lensing-corrected) peak fluxes are similar at ≈5 × 10⁻⁵ Jy. Dividing the exposure into two independent halves, the significances of 3.2 and 3.4 are consistent with that of the total exposure.

Zoom In Zoom Out Reset image size

To identify the likely source, we examined the final version of the reduced HST data of Abell 2744 (ACS and WFC3) acquired between 2013 November and 2014 July as part of the Frontier Fields program (ID: 13495, PI: Lotz), combining this with archival data from previous campaigns (ID: 11689, PI: Dupke; ID: 13386, PI: Rodney). Although there is some structure in the ALMA detection, it lies close to the source A2744_YD4 (F160W = 26.3) at R.A. = 00:14:24.9, decl. = −30:22:56.1(2000) first identified by Zheng et al. (2014). Correcting for an astrometric offset between HST positions and astrometry measured by the Gaia telescope (Gaia Collaboration et al. 2016), we deduce a small physical offset of $\simeq 0.2$ arcsec between the ALMA detection and the HST image.

Deep K_s data are also available from a 29.3 hr HAWK-I image taken between 2013 October and December (092.A-0472, PI: Brammer), which reaches a 5σ depth of 26.0. Spitzer IRAC data obtained in channels 1 (${\lambda }_{c}\sim 3.6$ μm) and 2 (${\lambda }_{c}\sim 4.5$ μm) with 5σ depths of 25.5 and 25.0, respectively, carried out under DDT program (ID: 90257, PI: T. Soifer). We extracted the HST photometry on PSF-matched data using SExtractor (Bertin & Arnouts 1996) v2.19.5 in double image mode using the F160W map for the primary detection (Figure 1). To derive the total flux, we applied an aperture correction based on the F160W MAG_AUTO measure (see, e.g., Bouwens et al. 2006). The noise level was determined using several 0.2 arcsec radius apertures distributed around the source. The total K_s magnitude of 26.45 ± 0.33 was obtained using a 0.6 arcsec diameter aperture applying the correction estimated in Brammer et al. (2016). The uncertainty was estimated following a similar procedure to that adopted for the HST data. The Spitzer data were reduced as described in Laporte et al. (2014) using corrected Basic Calibrated Data (cBCD) and the standard reduction software MOPEX to process, drizzle, and combine all data into a final mosaic. As shown in Figure 2, four other galaxies are close to A2744_YD4, but only the other source within the X-shooter slit is comparably bright to A2744_YD4. We used GALFIT (Peng & Ho 2002) to deblend the two sources and to measure their IRAC fluxes. We fitted both IRAC ch1 and ch2 images assuming fixed positions, those measured from the F160W image. Our photometry of A2744_YD4 is consistent with that published previously by the AstroDeep team (Zheng et al. 2014; Coe et al. 2015 and Merlin et al. 2016).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We used several SED fitting codes to estimate the photometric redshift of A2744_YD4 and hence its implied association with the ALMA detection. In each case, we fit all the available photometric data (HST-ACS, HST-WFC3, VLT HAWK-I, Spitzer).

First, we used an updated version of Hyperz (Bolzonella et al. 2000) with a template library drawn from Bruzual & Charlot (2003), Chary & Elbaz (2001), Coleman et al. (1980), and Leitherer et al. (1999) including nebular emission lines as described by Schaerer & de Barros (2009). We permitted a range in redshift ($0\lt z\lt 10$) and extinction ($0\lt {A}_{v}\lt 3$) and found the best solution at z_phot = 8.42${}_{-0.32}^{+0.09}$ (${\chi }^{2}\sim 1$), with no acceptable solution at lower redshift. Restricting the redshift range to $0\ \lt z\lt 3$ and increasing the extinction interval to ($0\lt {A}_{v}\lt 10$), we found a low-redshift solution at ${z}_{\mathrm{phot}}^{\mathrm{low}-z}$ = 2.17${}_{-0.08}^{+0.03}$ but with a significantly worse ${\chi }^{2}\sim 9$ (Figure 3).

We also made use of the Easy and Accurate Zphot from Yale (EAZY; Brammer et al. 2008) software. The SED fits adopted the standard SED templates from EAZY, as well as those from the galaxy evolutionary synthesis models (GALEV; Kotulla et al. 2009) including nebular emission lines as described by Anders & Fritze-v. Alvensleben (2003). Adopting a large redshift range ($0\lt z\lt 10$) with no prior assumptions on the extinction, the best fit has z_phot = 8.38${}_{-0.11}^{+0.13}$ in excellent agreement with that from Hyperz.

In summary, the photometry strongly supports a $z\gt 8$ solution. A low z solution is unlikely given the F814W-F125W > 3mag break as well as the low statistical likelihood.

Given the importance of confirming the presence of dust emission beyond $z\simeq 8$, we undertook a spectroscopic campaign using X-Shooter/VLT (ID: 298.A-5012, PI: Ellis). Between 2016 November 24 and 27, we secured 7.5 hr on-source integration with excellent seeing ($\approx 0.6$ arcsec). We used a 5 arcsec dither to improve the sky subtraction and aligned the slit so that a brighter nearby source could verify the blind offset (see Figure 2). The data were reduced using v2.8 of the ESO Reflex software combined with X-Shooter pipeline recipes v2.8.4.

We visually inspected all three arms of the X-Shooter (UVB, VIS, NIR) spectrum and identified one emission line at λ = 11408.4 Å with an integrated flux of f = 1.82 ± 0.46 ×10⁻¹⁸ erg s⁻¹ cm⁻². By measuring the rms at adjacent wavelengths, we measure the significance as ≈4.0σ. We checked the reliability of the line by confirming its presence on two independent spectral subsets spanning half the total exposure time (Figure 4). These half-exposures show the line with significances of 2.7 and 3.0, consistent with that of the total exposure. No further emission lines of comparable significance were found. We explore two interpretations of this line at 11408 Å. It is either (1) one component of the [O ii] doublet at a redshift $z\simeq 2.06$, or (2) Lyα at z = 8.38.

Zoom In Zoom Out Reset image size

For (1), depending on which component of the [O ii]λ3727, 3730 doublet is detected, we expect a second line at either 11416.9 Å or 11399.8 Å. No such emission is detected above the 1σ flux limit of 4.6 × 10⁻¹⁹ cgs. This would imply a flux ratio for the two components of $\approx 3.95$ (2.02) at 1(2)σ, greater than the range of 0.35–1.5 from theoretical studies (e.g., Pradhan et al. 2006).

For (2), although the line is somewhat narrow for Lyα (rest-frame width ≈20 km ${{\rm{s}}}^{-1}$), its equivalent width deduced from the line flux and the F125W photometry is 10.7 ± 2.7 Å, consistent with the range seen in other $z\gt 7$ spectroscopically confirmed sources (Stark et al. 2017). We detect no flux above the noise level at the expected position of either the CIV and [O iii] doublets at this redshift. However, at the expected position of the C iii] doublet, we notice a very marginal (≈2σ) feature at λ = 17914.7 Å (7.5 ± 0.35 × 10⁻¹⁹ erg s⁻¹ cm⁻²) seen on two individual sub-exposures. If this is C iii]λ1907 Å (normally the brighter component) at z_{C iii} = 8.396, the Lyα offset of 338 ± 3 km s⁻¹ would be similar to that for a z = 7.73 galaxy (Stark et al. 2017). The other component would be fainter than 5.0 × 10⁻¹⁹ erg s⁻¹ cm⁻², consistent with theoretical studies of this doublet (e.g., Rubin et al. 2004).

Previous spectroscopy of A2774_YD4 was undertaken by the GLASS survey (Schmidt et al. 2016), who place a 1σ upper limit on any Lyα detection at 4.4 × 10⁻¹⁸ erg s⁻¹ cm⁻², $\approx 2.4$ times above our X-Shooter detection.

Only a few far-infrared emission lines lines are detectable for sources in the reionization era (see, e.g., Combes 2013). Only the [O iii] 88 μm line at the z = 8.38 redshift of Lyα would be seen in the frequency range covered by our ALMA observations. Given the recent detection of this line in a $z\simeq 7.2$ Lyα emitter (Inoue et al. 2016), we examined our ALMA data for such a possibility. Searching our data in frequency space, we find a 4.0σ narrow emission line offset by 0.35 arcsec from the astrometric position of A2744_YD4 at a frequency of 361.641 GHz. Dividing the exposure time in half, the line is detected with independent significances of 2.8 and 3.2, consistent with that of the total exposure. Assuming this line is [O iii] 88 μm, the redshift would be z = 8.382 (see Figure 5), leading to a Lyα velocity shift of ∼70 km s⁻¹ in good agreement with that observed in a $z\sim 7.2$ galaxy (Inoue et al. 2016). Fitting the emission line with a Gaussian profile, we derive a modest FWHM = 49.8 ± 4.2 km s⁻¹ implying an intrinsic width of $\simeq 43$ km s⁻¹. The emission line luminosity is estimated at 1.40 ± 0.35 × 10⁸ L${}_{\odot }$ without correction for magnification, which is ≈7 times fainter than that detected in Inoue et al.'s z = 7.2 source. The peak line flux of A2744_YD4 is consistent with that computed from simulations in Inoue et al. (2014, see their Figure 3). Compared to the aforementioned lower-mass source at $z=7.2$, the narrower line width is perhaps surprising but may indicate its formation outside the body of the galaxy as inferred from the offset and recent simulations (Katz et al. 2016).

Zoom In Zoom Out Reset image size

Estimating the magnification is critical to determine the intrinsic properties of any lensed source. Several teams have provided mass models for each of the six clusters. Moreover, a web tool¹¹ enables us to estimate the magnification for Abell2744_YD4 from parametric high-resolution models, i.e., version 3.1 of the CATS model (Richard et al. 2014), version 3 of Johnson et al. (2014), version 3 of Merten et al. (2011) and version 3 of GLAFIC (Ishigaki et al. 2015). We took the average value with error bars corresponding to the standard deviation: μ = 1.8 ± 0.3.

The detection of dust emission in a z = 8.38 galaxy provides a unique opportunity to evaluate the production of dust, presumably from early supernovae in the first few 100 Myr since reionization began. The key measures are the dust and stellar masses and the likely average past SFR.

We first estimate some physical properties based on the ALMA continuum detection using a simple modified blackbody SED with the dust temperatures ranging from 35 to 55 K and the dust emissivity fixed at β ∼ 2. We found a total FIR luminosity ranging from 7.1 to 18.2 × 10¹⁰ ${L}_{\odot }$ and a dust mass ranging from 1.8 to 10.4 × 10⁶ ${M}_{\odot }$. These values are corrected for magnification and CMB heating.

We also ran an updated version of MAGPHYS (da Cunha et al. 2008) adapted for high-z galaxies (da Cunha et al. 2015). The code estimates the properties of A2744_YD4 in two steps. First, it generates a library of model SEDs at the redshift of our source ($z=8.38$) in our 11 bands (7 from HST, the deep HAWK-I K_s band, the two first IRAC channels, and a synthetic filter of the ALMA band 7) for a wide range of variables including the SFR and dust content. We generated a total of about 9 million models, including ≈25,000 IR dust emission models. MAGPHYS then derives the likelihood distribution of each physical parameter by comparing the observed SED with all the models in the library. In this way, we derived the following properties: SFR = 20.4${}_{-9.5}^{+17.6}$ ${M}_{\odot }$ yr⁻¹; a stellar mass ${M}_{\star }$ = (1.97${}_{-0.66}^{+1.45}$) × 10⁹ ${M}_{\odot };$ a dust mass M_dust = (5.5${}_{-1.7}^{+19.6}$) × 10⁶ ${M}_{\odot }$; and an extinction A_v = 0.74${}_{-0.48}^{+0.17}$ with a dust temperature ranging from 37 to 63 K. The error bars refer to 1σ uncertainties. These values estimated with MAGPHYS using the full SED are consistent with those deduced solely from the ALMA continuum flux.

Although the uncertainties in these physical properties are large, our target is similar to the lensed source A1689-zD1 at $z\simeq 7.5$ studied by Watson et al. (2015). Watson et al. (2015) reports ${M}_{\star }$ = 1.7${}_{-0.5}^{+0.7}$ × 10⁹ ${M}_{\odot }$ and SFRs ${9}_{-2}^{+4}$ ${M}_{\odot }$ yr⁻¹. Their specific star formation rates are thus similar (1.04${}_{-0.21}^{+10.2}$ × 10⁻⁸ yr⁻¹ for A2744_YD4 and ${0.6}_{-0.3}^{+1.1}$ × 10⁻⁸ yr⁻¹ for A1689-zD1) implying a mean lifetime for a constant SFR of ≈100 Myr. However, A1689-zD1 has a significantly larger dust mass of M_dust =${4}_{-2}^{+4}\,\times$ 10⁷ ${M}_{\odot }$. Possibly, this is a consequence of continuous star formation over a longer period together with more advanced chemical enrichment.