TRACING THE REIONIZATION EPOCH WITH ALMA: [C ii] EMISSION IN z ∼ 7 GALAXIES

In Table 2, we present our results: a [C ii] emission line is detected for the three sources observed only in Cycle-3, with an S/N = 4.5. For NTTDF6345 an emission line is observed separately in the Cycle-2 and Cycle-3 data sets with an S/N of 4.1 and 5.6, respectively, and an S/N > 6 is obtained in the combination. Although the S/N is modest, the spatial coincidence (or very close spatial association) between the [C ii] and the near-IR counterparts and the consistent small shift with respect to the redshift determined from the Lyα emission, all argue for the reality of the detections. In addition, in three out of four cases, the detections are spatially resolved. In Figure 1 for each source, we present the spectrum of the [C ii] region with a rebinning of 40 km s⁻¹ for NTTDF6345 and 20 km s⁻¹ for the other sources: the vertical dotted line at v = 0 km s⁻¹ indicates the redshift determined from the Lyα emission. The maps of the line emission are shown in Figure 2, extracted with a spectral width of 440 km s⁻¹ for NTTDF6345, 280 km s⁻¹ for COSMOS24108, 100 km s⁻¹ for UDS16291, and 120 km s⁻¹ for COSMOS13679. The black crosses indicate the centroid of the Y-band image for NTTDF6345 and of the HST H-band images for the other galaxies.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 2.  Galaxies ALMA Properties

ID	${\lambda }_{0}$[C ii])	[C ii] Flux	FWHM [C ii]	S/N	rmscont	tint	${\rm{\Delta }}\mathrm{pos}$	${\rm{\Delta }}v$	${M}_{\mathrm{dyn}}\ast \sin {(i)}^{2}$	SFRdust
	(mm)	(Jy km s−1)	(km s−1)		$(\mu \mathrm{Jy}\,{\mathrm{beam}}^{-1})$	(s)	(arcsec)	(km ${{\rm{s}}}^{-1})$	$({10}^{9}{M}_{\odot })$	$({M}_{\odot }{\mathrm{yr}}^{-1})$
COSMOS13679	1.28426 ± 0.00006	5.9 × 10−2	90 ± 35	4.5	14	2782	0.4	135	<2.5	<6.2
NTTDF6345	1.2143 ± 0.0002	1.6 × 10−1	250 ± 70	6.1	16	2087	0.0	110	<18	<5.7
UDS16291	1.20438 ± 0.00003	6.3 × 10−2	50 ± 15	4.5	20	2117	0.1	110	<0.7	<6.6
COSMOS24108	1.20249 ± 0.00007	9.2 × 10−2	150 ± 40	4.5	18	2177	0.8	240	<6.5	<6.2

Download table as:  ASCIITypeset image

Thermal far-infrared continuum is not detected in any of the galaxies: the limits on the total IR-luminosity convert into limits on the dust obscured ${\mathrm{SFR}}_{\mathrm{dust}}$ that are reported in Table 2 (assuming a Kennicutt 1998 relation with a Salpeter initial mass function).

In Table 2, we report the offset between the near-IR coordinates and the ALMA detections. In two cases, the shifts are consistent with the ALMA astrometric uncertainty (01–015), while for COSMOS24108 and COSMOS13679 they are larger. Dunlop et al. (2016) recently noted that the HST and ALMA astrometry of the HUDF field presented both a systematic shift of $0\buildrel{\prime\prime}\over{.} 25$ to the south and a random shift of up to $0\buildrel{\prime\prime}\over{.} 5$. Spatial offsets of up to $0\buildrel{\prime\prime}\over{.} 5$ are clearly evident in most of the LBGs observed by Capak et al. (2015) and for the $z\sim 6$ galaxy WMH5 observed by Willott et al. (2015; $\sim 0\buildrel{\prime\prime}\over{.} 4$), while in the other galaxy, CLM1, the [C ii] emission is co-spatial with the UV-continuum.

To further investigate this issue, we looked at bright serendipitous sources detected in the continuum band in our fields. We found at least one source per field, and we measured a shift between $0\buildrel{\prime\prime}\over{.} 1$ and $0\buildrel{\prime\prime}\over{.} 6$ in random directions between the HST (or HAWK-I) counterpart and the ALMA detection. Given the depth of the CANDELS images ($H\sim 27$ at 5σ), it is unlikely that the ALMA detections are associated with other objects undetected in the H-band. Note that for NTTDF6345 where we only have HAWK-I images, the shift is negligible both for the serendipitous source and the LBG. We conclude that there are still substantial uncertainties in the relative ALMA-HST astrometry. In summary, for UDS16291 and NTTDF6345 the [C ii] detections are centered at the position of the near-IR sources, and for COSMOS13679 the offset (04) is within the range of those reported in the literature and measured for the serendipitous sources: for these three galaxies we conclude that the [C ii] emission comes from the same region as the bulk of the SF observed in the near-IR images. For COSMOS24108 the offset is slightly larger than the range reported in the literature, and while the [C ii] emission is almost certainly associated with the source, it could actually come from an external region of the galaxy, not coincident with the bulk of the star formation. A similar case was already observed in BDF3299 at z = 7.109 (Maiolino et al. 2015): for this galaxy we concluded that the [C ii] emission arises from an external accreting/satellite clump of neutral gas, in agreement with recent models of galaxy formation (Vallini et al. 2015). While the identification with another transition from a foreground galaxy, which by chance happens to be at $0\buildrel{\prime\prime}\over{.} 8$ from COSMOS24108, is unlikely, we cannot completely discard the possibility of a spurious detection given the low S/N.

The [C ii] line traces the systemic redshift of the sources, unlike the Lyα line, which is typically redshifted by up to several hundreds km ${{\rm{s}}}^{-1}$ (Erb et al. 2004; Trainor et al. 2015) consistent with the presence of outflowing gas, although the final observed Lyα profiles depend on many factors such as geometry, gas covering fraction, dust, and IGM ionization state. For our galaxies the velocity shifts are not very large, of the order of 100–200 km s⁻¹, smaller than those reported at ${\text{}}z\sim 3$ for galaxies with similar UV luminosities. Specifically, Erb et al. (2004) measure shifts of up to 1000 km s⁻¹ and average values of 400 km s⁻¹ for LBGs with ${M}_{\mathrm{UV}}\lt -21$. The mean shift is also lower than those reported by Willott et al. (2015), 430 and 275 km s⁻¹, respectively, for their two $z\sim 6$ galaxies.¹¹ The small shifts in our galaxies are particularly significant given that in general objects with low Lyα emission have larger velocity offsets. Similar tentative evidence for smaller Lyα velocities at $z\sim 7$ compared to $z\sim 3$ was recently reported by Stark et al. (2015) using the UV nebular C iii]$\lambda 1909$ doublet to determine the systemic redshift in two distant LBGs.

The velocity of Lyα compared to the systemic redshift is very relevant when interpreting the line visibility during the reionization epoch, in the presence of a partly neutral IGM (Dijkstra et al. 2011). Smaller velocity offsets imply that Lyα is closer to resonance and more easily quenched by a neutral IGM. If at redshift 6 and 7 the offsets are as large as those found for lower-redshift LBGs, the IGM must be very neutral to produce the drop in Lyα fraction that is observed between these two epochs (Pentericci et al. 2014). On the other hand, if the velocity offsets were much smaller, as our observations indicate, the drop in the Lyα visibility could be produced by an IGM that is still substantially more ionized (Mesinger et al. 2015).

In Figure 3, we show the SFR–L[C ii] relation for our galaxies and previously observed sources. We plot COSMOS24108 with a different symbol because it is not certain whether its [C ii] emission is from the main galaxy or just from a clump in its outskirt, in which case the point would shift ∼1 dex to the left. We remark that the SFR for our sources as well as previous ${\text{}}z\sim 7$ ones are UV-based, with no correction for dust extinction. As stated above the upper limits on the ${\mathrm{SFR}}_{\mathrm{dust}}$ are very low, at least for our galaxies. The SFR for the Capak et al. (2015) sample includes both the UV and dust obscured contribution. Finally, for the Willot et al. sample we plot SED-derived SFR. Our galaxies are a factor of 2–3 less luminous in L[C ii] than $z\sim 5.5$ galaxies with similar SFRs. They also fall below the SFR–L[C ii] relation of low-redshift star-forming galaxies and low-metallicity galaxies (black solid and dashed lines; De Looze et al. 2014). Previous observations of $z\sim 7$ galaxies failed to detect the [C ii] emission. Some of the galaxies were fainter than ours, but few others were in the same range, and in these cases, the limits reached were deep enough to detect [C ii] if the emission was at the same level as in our sources (Maiolino et al. 2015; Schaerer et al. 2015).

Zoom In Zoom Out Reset image size

However, we note that previously observed sources were either selected as Lyα emitters or were LBGs, but their spectra showed a Lyα emission with very high EW (typically >40 Å). The only exception z8-GND-5296 at z = 7.5, which has modest Lyα emission, but in this case the L[C ii] limit is very shallow (Schaerer et al. 2015). Our four new sources have Lyα emission with low EW (Table 1). The Lyα emission strength is known to depend on the presence of dust and possibly metallicity (Raiter et al. 2010): although the derivation of metallicity is not easy, several studies indicate that LAEs are more metal-poor galaxies compared to the rest of the LBG population (e.g., Song et al. 2014). Metallicity plays an important role in shaping the SFR–L[C ii] relation: in Figure 3, we show different metallicity-dependent relations produced by a recent study of Vallini et al. (2015) based on high-resolution, radiative transfer cosmological simulations. The different lines have Z = 0.05, 0.1, and 0.2 ${Z}_{\odot }$, respectively, and for metallicity $\leqslant 0.1$ are consistently below the relation found for local galaxies. Therefore, the contradictory results for $z\sim 7$ galaxies might be due to different intrinsic metallicity.

Alternatively, galaxies could be caught in different evolutionary stages: it has been suggested that molecular clouds in the central parts of primordial galaxies could be rapidly disrupted by stellar feedback hence suppressing the emission of [C ii]. Clumps of neutral gas (or small satellites) in the outer regions of such galaxies, could survive photo-ionization and still show [C ii] emission, as a consequence of the diffuse far-UV radiation emitted by the primary galaxy. This could be the case of COSMOS24108, where the [C ii] emission is offset from the near-IR source and could come from one such clump. A similar occurrence was also observed in BDF3299 at z = 7.109 where the [C ii] emission was displaced by the near-IR source by 4 kpc. We refer to Maiolino et al. (2015) for an extensive discussion of this scenario.

Assuming the sources have ordered motions, we can estimate the dynamical masses based on the [C ii] velocity dispersion. We follow the method described in Wang et al. (2013), who approximate ${M}_{\mathrm{dyn}}=1.16\times {10}^{5}{V}_{\mathrm{cir}}^{2}D$, where V_cir is the circular velocity in km s⁻¹, D is the size in kpc, ${V}_{\mathrm{cir}}=1.763\sigma [{\rm{C}}\,{\rm{II}}]/\sin (i)$, and i is the disk inclination angle. Our sources are only marginally resolved so we assume an upper limit on D from the resolution of $0\buildrel{\prime\prime}\over{.} 9$. We obtain masses between $0.7\,-\,18\times {10}^{9}{M}_{\odot }$. We can compare these values to the total stellar masses determined by a classical SED fitting of the multi-wavelength photometry for the three CANDELS sources. We use the photometry from the CANDELS catalogs, including deep IRAC data essential for sampling the rest-frame optical emission at these redshifts. We include the contribution from nebular emission lines, which can strongly contaminate the IRAC bands. The stellar masses are reported in Table 1 and are very similar to the dynamical masses, especially considering the uncertainty due to the unknown inclination angle i.