ALMA Reveals Metals yet No Dust within Multiple Components in CR7

We observed CR7 (10:00:58.00 +01:48:15.3, J2000) in band 6 with ALMA during cycle 3 with configuration C43–4, aimed to achieve a $0\buildrel{\prime\prime}\over{.} 3$ angular resolution (program ID #2015.1.00122.S). The target has been observed with between 35–43 antennas for a total on source integration time of 6.0 hr on 22, 23, 24 May and 4, 8 November 2016, with precipitable water vapor ranging from 0.3–1.4 mm. We observed in four spectral windows centered at 249.94592, 247.94636, 234.94917, and 232.9496 GHz with a bandwidth of 1875 MHz, of which the first is centered at [C ii] line emission (${\nu }_{z=6.604}=249.9395$ GHz). The central frequencies of the spectral windows correspond to rest-frame frequencies between 1771.7 GHz and 1901.0 GHz, with a velocity resolution of 19.4 km s⁻¹. The phase calibrations have been performed on the source J0948 + 0022. The quasar J1058 + 0133 has been observed regularly as a bandpass and flux calibrator, resulting in typical flux calibration errors of ∼10%. Data have been reduced using Casa version 4.7.0 (McMullin et al. 2007) with natural weighting and channel averaging resulting in a 38.8 km s⁻¹ velocity resolution. We use uv tapering of the visibilities with a Gaussian width FWHM = 07 to optimize the S/N ratio. The observational set-up and data reduction results in a beam size of $0.35\times 0\buildrel{\prime\prime}\over{.} 33$ (Gaussian σ; $0.82\times 0\buildrel{\prime\prime}\over{.} 77$ FWHM), with a position angle of $-49\buildrel{\circ}\over{.} 5$. We have also performed the reduction with uv tapering with different smoothing kernels ranging from 04–08, but find that all changes in luminosity, line-width, and size measurements are consistent within the error-bars. A simple estimate of the noise level of the pixels within a radius of $20^{\prime\prime}$ of the center results in rms = 0.06 mJy beam⁻¹ in 38.8 km s⁻¹ channels.

As described in Section 4, we detect two separate clumps at high significance in the [C ii] image collapsed over all velocities at which significant line-emission is detected, see Figure 1. These [C ii] clumps are offset by $0\buildrel{\prime\prime}\over{.} 25$ to the east with respect to the UV positions of clumps A and B, while the offset in declination is $\lt 0\buildrel{\prime\prime}\over{.} 10$, respectively. The alignment of both clumps, and their separation is also similar in the UV and in [C ii]. As these two [C ii] clumps resemble the geometry of clumps A and B, we assume that the offsets between the UV positions and [C ii] positions are due to errors in the astrometry. This offset is similar in magnitude as reported in e.g., Dunlop et al. (2017), but in the R.A. direction instead of the decl. direction (see also Carniani et al. 2017 and references therein for offsets in $z\sim 6\mbox{--}7$ galaxies). For the rest of the analysis, we align the ALMA data with HST and Subaru narrowband data by applying an offset of $0\buildrel{\prime\prime}\over{.} 25$ to the west. We note that the positions of two serendipitous IR continuum-detections of foreground galaxies are also in agreement with this offset (see Section 5.1).

Zoom In Zoom Out Reset image size

Because the noise is correlated spatially and spectrally, we measure the sensitivity of the observations and significance of detections as follows. In each collapsed image (which has a size of $51\buildrel{\prime\prime}\over{.} 2\times 51\buildrel{\prime\prime}\over{.} 2$, centered on CR7), we measure the noise over the same spatial scales as those used for measurements (typically a beam size, $r=0\buildrel{\prime\prime}\over{.} 35$). We place circular apertures on 100,000 random positions on the image and integrate the flux over the scales confined by these apertures. As the image size is much larger than the source size, we do not mask any central region. We then compute the 1σ detection significance using the r.m.s., which results in 16.9 mJy beam⁻¹ km s⁻¹ for beam-sized apertures on the collapsed image shown in Figure 1.

As illustrated in Figure 2, the distribution of fluxes measured on random positions approximately follows a Gaussian. It can be seen that the negative image (where we have inverted all counts) has slightly lower number counts than the positive image at high flux levels. This is due to random positions around the center of CR7, or real sources around CR7. These sources (such as a $\sim 4\sigma$ detection $4^{\prime\prime}$ from the southeast of CR7, see Figure 1) will be discussed and investigated in a future paper. When integrating over a larger spatial scale, the noise estimate is slightly higher due to spatially correlated noise. The noise level is lower when data is collapsed over a smaller velocity range. In individual velocity channels (with resolution 38.8 km s⁻¹), we measure a 1σ sensitivity between 30–40 μJy when integrating flux over beam-sized scales. Throughout the paper, we convert measured fluxes to [C ii] luminosities following, e.g., Solomon et al. (1992) and Carilli & Walter (2013):

$$\begin{eqnarray}&&{L}_{[{\rm{C}}{\rm{II}}]}/{L}_{\odot }=1.04\times {10}^{-3}\times {S}_{\nu }{\rm{\Delta }}v\times {\nu }_{\mathrm{obs}}\times {D}_{L}^{2},\end{eqnarray} \tag{ 1 }$$

where ${S}_{\nu }{\rm{\Delta }}{\rm{v}}$ is the flux in Jy km s⁻¹, ${\nu }_{\mathrm{obs}}$ is the observed frequency in GHz ($\approx 250.0\,\mathrm{GHz}$ for CR7), and ${{\rm{D}}}_{L}$ is the luminosity distance in Mpc (64457.8 Mpc at z = 6.60).

In order to search for [C ii] emission around CR7, we inspect the data cube of the spectral tuning centered at the frequency where [C ii] is expected to be observed. As shown in Figure 1, resolved [C ii] emission is clearly detected.¹⁴ We show the collapsed data cube with velocities between −614 and +96 km s⁻¹ with respect to the Lyα redshift, which includes the full velocity range over which we detect [C ii] emission. Within the 3σ significance contours, we measure ${L}_{[{\rm{C}}{\rm{II}}]}=2.17\pm 0.36\times {10}^{8}\,{L}_{\odot }$ in the collapsed image.¹⁵ The spectrum extracted over this region is shown in Figure 3 and is best fitted with a line-width of ${v}_{\mathrm{FWHM}}=299\pm 27$ km s⁻¹ that is offset by −167 ± 27 km s⁻¹ with respect to the Lyα redshift, see Table 1. The 3σ significance contours span a total area of 2.2 arcsec² and have a maximum diameter of $1\buildrel{\prime\prime}\over{.} 8$, which corresponds to 9.7 kpc at z = 6.60. This detection of [C ii] emission in a large region in/around CR7 clearly indicates the presence of metals, ruling out the possibility that a large fraction of the gas is primordial.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

As illustrated in Figure 1, the [C ii] emission is clearly not homogeneous over the 3σ contour and reveals already two local peaks, resembling the rest-frame UV morphology as observed by HST. In order to show structure in both the spatial and spectral dimensions, we create collapsed [C ii] maps in three different velocity ranges of 100 km s⁻¹ each and show these in Figure 4. As the noise level varies slightly at different velocities and is smaller for images that are collapsed over a smaller velocity range, we compute the noise level in each image independently as described in Section 3.3.

Zoom In Zoom Out Reset image size

The three velocity channels reveal that the [C ii] emission in CR7 consists of multiple, well-separated components. Around the systemic velocity (−130 km s⁻¹ with respect to Lyα), two sources are clearly detected at the UV positions of clumps A and B. At a blueshift of ≈250 km s⁻¹ a relatively faint, resolved component is detected in between the three UV clumps, but closest to clump C. No significant UV emission is detected at that ALMA position. We name this component C-2. Another compact subcomponent is detected at an even larger blueshift of ≈420 km s⁻¹ close to the position of clump B (well within the beam radius), and hence named B-2. In these narrower velocity slices, all detections are at the $\gtrsim 4\sigma$ significance level, see Figure 4.

Zoom In Zoom Out Reset image size

Hence, we detect several components of [C ii] emission in the CR7 system that are significantly offset in both the spatial and spectral dimension. Before discussing the implications of these detections in Section 7.1, we first measure their luminosities and [C ii] line-profiles.

4.2.1. Measurements of Individual Clumps

Except for clumps A and C-2, all clumps are unresolved (see Section 4.3). For unresolved sources, we measure the [C ii] luminosity within a circular aperture-region with radius of $0\buildrel{\prime\prime}\over{.} 35$ (corresponding to the beam semimajor axis). For clumps A and C-2, we use $0\buildrel{\prime\prime}\over{.} 55$ and $0\buildrel{\prime\prime}\over{.} 50$ apertures, respectively, to account for the fact that these sources are slightly extended. For these clumps, we find that these apertures retrieve the same fraction of the flux as is retrieved for modeled point sources with a $0\buildrel{\prime\prime}\over{.} 35$ aperture. Measurements at HST/UV positions are done on the collapsed image over the full velocity range (e.g., Figure 1), while measurements at ALMA positions (B-2 and C-2) are done in the collapsed velocity range in which these clumps are detected (e.g., Figure 4).

In Figure 6, we show the ALMA spectra of the different clumps (extracted over their respective apertures) in the rest-frame of clump A (z = 6.601) and compare these to the Lyα profile measured with VLT/X-SHOOTER at the peak of Lyα emission (close to clump A). We measure the velocity offset between [C ii] and Lyα and the [C ii] line-width and amplitude by fitting Gaussian profiles. All measurements are summarized in Table 1.

Zoom In Zoom Out Reset image size

We find that clump A, which is the brightest in the UV and closest to the peak of Lyα emission, is also the brightest in [C ii] line-emission, with ${L}_{[{\rm{C}}{\rm{II}}]}=(0.96\pm 0.20)\times {10}^{8}\,{L}_{\odot }$. We caution that we measure a factor of two lower luminosity if we measure the luminosity of clump A with a beam-sized aperture. Clump A is blueshifted by 146 ± 27 km s⁻¹ with respect to Lyα. This is slightly smaller than typical offsets between Lyα and nebular lines for sources with similar UV luminosities (e.g., Stark et al. 2017), although Carniani et al. (2017) shows a wide spread of velocity offsets in a compilation of ALMA detections (i.e., $-150\lt {\rm{\Delta }}v\lt +400$ km s⁻¹ for galaxies with $10\lt \mathrm{SFR}\lt 100\,{M}_{\odot }$ yr⁻¹). The [C ii] line-width is 259 ± 24 km s⁻¹, similar to the Lyα line-width (see, e.g., Figure 6).

At the UV position of clump B, we measure [C ii], with ${L}_{[{\rm{C}}{\rm{II}}]}=(0.34\pm 0.09)\times {10}^{8}\,{L}_{\odot }$, which contains the contribution from both the ALMA detections B and B-2 (with ≈33% of the flux being due to the subcomponent B-2, see Table 1). The brightest [C ii] component of clump B is at $z=6.600\pm 0.001$, similar to the redshift of clump A, but has a narrower line width. Clump B-2 is blueshifted by 290 ± 30 km s⁻¹ with respect to the systemic redshift and has a relatively narrow line width of 92 ± 22 km s⁻¹.

At the position of clump C, we measure relatively faint [C ii] emission at the $\approx 3\sigma$ level, see Figure 1 and Figure 5. The aperture at the position of clump C is contaminated slightly by this clump B-2 (which may also be associated with C), explaining the bluest line, but the redder [C ii] component C-2 is offset spatially (see Figure 4). At the position of C-2, we measure ${L}_{[{\rm{C}}{\rm{II}}]}=(0.27\pm 0.07)\times {10}^{8}\,{L}_{\odot }$ with a line width of 181 ± 30 km s⁻¹ and an offset of −232 ± 27 km s⁻¹ with respect to Lyα. This offset is bluer than clumps A and B, but redder than B-2. Due to its large blueshift, component C-2 also dominates the velocity map shown in Figure 7. Due to this strong contribution of component C-2 to the velocity map, we find that there is no clear evidence for ordered rotation over the full [C ii] extent. This shows that the spatial resolution of our observations is crucial, as we may have not been able to distinguish clumps C-2 and B/B-2 with lower resolution, and could have misinterpreted the velocity map as rotation. It is challenging to investigate whether ordered rotation is present inside individual clumps because they are either unresolved or only marginally resolved. Similarly as for clump A, we note that the [C ii] luminosity of clump C-2 would be reduced by a factor of $\approx 2$ if we use a smaller aperture of $0\buildrel{\prime\prime}\over{.} 35$, which is used for unresolved sources. Therefore, the luminosities of clumps A and C-2 should be interpreted with caution. The total [C ii] luminosity that is associated to clumps is $(1.57\pm 0.23)\times {10}^{8}\,{L}_{\odot }$, which is the sum of the aperture measurements at the HST positions. This corresponds to a fraction of 0.72 ± 0.18 of the total observed [C ii] luminosity.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Starting from the measured velocity width and size of the emitting regions, it is possible to obtain a rough estimate of the dynamical mass of the various [C ii] components (e.g., Wang et al. 2013). First, we measure the size by performing two-dimensional Gaussian fits in the collapsed 2D image. Clumps A and B are simultaneously fitted in the image that is collapsed over the full velocity range, while clumps B-2 and C-2 are fitted in the collapsed image in the velocity range as shown in Figure 4. In our fitting procedure, we fixed the position angle and the ratio between the dispersion in both dimensions corresponding to those of the beam semimajor and semiminor axis. We also fix the positions to those indicated in Figure 4. Hence, the free parameters are the amplitudes and widths of the Gaussian profiles. In order to take the (correlated) noise into account, we add the noise from a randomly selected cut-out region in the collapsed image, avoiding the source itself, before performing the fit. This is repeated 5000 times in order to estimate the statistical reliability of the fit. We find that more than $\gt 50 \%$ of the measurements for clumps B and B-2 result in a size that is equal to or smaller than the beam, meaning that these clumps are unresolved and hence have a size ${r}_{1/2,[{\rm{C}}{\rm{II}}]}\lt 2.2\,\mathrm{kpc}$. We measure an observed size of ${r}_{1/2,[{\rm{C}}{\rm{II}}],\mathrm{obs}}={3.7}_{-0.6}^{+0.6}$ kpc for clump A (4% of the measurements result in an unresolved size, meaning that clump A is resolved at $\approx 2\sigma$ significance) and ${r}_{1/2,[{\rm{C}}{\rm{II}}],\mathrm{obs}}={4.4}_{-0.6}^{+0.9}$ kpc for clump C-2 (resolved at $\approx 3.5\sigma$ significance). We deconvolve the sizes of A and C-2 to obtain the intrinsic size with ${r}_{\sigma ,[{\rm{C}}{\rm{II}}]}=\sqrt{{r}_{\mathrm{obs}}^{2}-{2.2}^{2}}$. Resulting sizes are listed in Table 1. Dynamical masses are computed following Wang et al. (2013):

$$\begin{eqnarray}&&{M}_{\mathrm{dyn}}/{M}_{\odot }{(\mathrm{sini})}^{2}=1.94\times {10}^{5}\times {v}_{\mathrm{FWHM},\ [{\rm{C}}\,{\rm{II}}]}^{2}\times {r}_{1/2,[{\rm{C}}{\rm{II}}]},\end{eqnarray} \tag{ 2 }$$

where ${M}_{\mathrm{dyn}}$ is the dynamical mass in ${M}_{\odot }$, i is the inclination angle, ${v}_{\mathrm{FWHM},[{\rm{C}}{\rm{II}}]}$ is the line-width in km s⁻¹, and ${r}_{1/2,[{\rm{C}}{\rm{}}\mathrm{II}]}$ is the size in kiloparsecs (half-light radius). This results in dynamical masses (uncorrected for inclination) ranging from $(3.9\pm 1.7)\times {10}^{10}\,{M}_{\odot }$ for component A, $(2.4\pm 1.9)\times {10}^{10}\,{M}_{\odot }$ for clump C-2, and $\lt 0.7\times {10}^{10}\,{M}_{\odot }$ for $\lt 0.4\times {10}^{10}\,{M}_{\odot }$ for clumps B and B-2, respectively. These dynamical mass estimates are lower than typical quasar host galaxies at $z\approx 6$ (Wang et al. 2013), and comparable to star-forming galaxies at $z\approx 7$ with similar SFRs as CR7 (e.g., Pentericci et al. 2016; Smit et al. 2017).

We combine the flux in all four spectral windows from our ALMA coverage to search for dust continuum emission. In the entire image, we find two detections with ${\rm{S}}/{\rm{N}}\gt 3$, but they are not associated with CR7. One detection is $3\buildrel{\prime\prime}\over{.} 5$ northeast of CR7 (associated with ID number 339509 in the catalog from Laigle et al. (2016), $\mathrm{photo} \mbox{-} z=0.84$), while the other is $18\buildrel{\prime\prime}\over{.} 5$ to the southwest (ID number 335753 in Laigle et al. (2016), $\mathrm{photo} \mbox{-} z=3.10$). The positions of these foreground galaxies confirm the astrometric correction described in Section 3.2. We note that we detect a tentative (3σ) line at 250.484 GHz at the position of ID 339509 that is identified as CO(4–3) with ${\nu }_{0}=461.041\,\mathrm{GHz}$ at z = 0.841, perfectly consistent with its photometric redshift.

As is visible in Figure 8, there is a $\approx 2.7\sigma$ continuum detection $\approx 2^{\prime\prime}$ to the southeast of CR7, nearby a faint HST detection that we name potential clump D. This faint HST detection has F110W = 27.2 ± 0.2 in a $0\buildrel{\prime\prime}\over{.} 4$ aperture, similar to clump B. It is also detected at $\approx 3\sigma$ in the ACS/F814W filter (${\rm{F}}814{\rm{W}}\,=27.5\pm 0.2$) and F160W filter (${\rm{F}}160{\rm{W}}=27.0\pm 0.3$). It is not detected in NB921 and $z^{\prime}$ imaging, which have 3σ depths of 25.8 AB magnitude, and hence are consistent with the HST photometry. Potential clump D is not detected in the Subaru S-cam imaging in the $B,V,R$ filters with 2σ upper limits of $\lesssim 28.5\mbox{--}29.0$. Therefore, the photometry is consistent with the Lyman-break falling in the F814W filter, such that $4.8\,\lt {z}_{\mathrm{phot}}\lt 6.9$. There is no significant [C ii] detection at its position in any of the spectral tunings observed with ALMA. Further IFU observations are required to measure the redshift of clump D and test whether it is associated with CR7.

There is no IR continuum detected at the position of CR7's clumps A, B, and C, see Figure 8. Around the position of CR7, we measure an rms of 7 μJy beam⁻¹ at ${\lambda }_{0}\approx 160\,\mu {\rm{m}}$. Note that we do not remove contamination from the [C ii] line as we expect that this is negligible, as a result, our upper limit is therefore somewhat on the conservative side. We follow Schaerer et al. (2015) to convert this limit into a 3σ upper limit on the infrared luminosity (between 8–1000 μm) of ${L}_{\mathrm{IR}}\lt 3.14\times {10}^{10}\,{L}_{\odot }$, under the assumption that the dust temperature is 35 K and a modified black body SED with a power-law slope of 2.9 in the Wien regime, ${\beta }_{\mathrm{IR}}=1.5$ and after removing the contribution of the CMB to the dust heating, see da Cunha et al. (2013). In the case that the dust temperature is 25 (45) K, the limit on the IR luminosity is ${L}_{\mathrm{IR}}\lt 1.68(6.15)\times {10}^{10}\,{L}_{\odot }$. Combined with the total [C ii] luminosity, we find a lower limit of the the ratio log₁₀(${L}_{[{\rm{C}}{\rm{II}}]}/{L}_{\mathrm{IR}}$) $\gt \,-2.1$, which is similar to the highest values measured in local star-forming galaxies (e.g., Malhotra et al. 2001) and significantly higher than submillimeter galaxies and quasar host galaxies at $z\gt 3$ (see Ota et al. 2014 for a compilation). This could potentially indicate a lower dust-to-metal ratio, which in turn could result in a higher dust temperature due to the lower dust opacity (e.g., Cen & Kimm 2014).

The ratio of the ${\mathrm{IR}}_{160\mu {\rm{m}}}$ to ${\mathrm{UV}}_{0.15\mu {\rm{m}}}$ flux density is constrained to ($\nu {{\rm{F}}}_{\nu }{)}_{\mathrm{IR}}/{(\nu {{\rm{F}}}_{\nu })}_{\mathrm{UV}}\lt 0.04$ (3σ)). This limit is almost an order of magnitude stricter than other $z\sim 7$ sources observed by Maiolino et al. (2015). In the local universe, this ratio is related to metallicity (see the compilation and fitted relation in Maiolino et al. 2015), likely as a result of dust content increasing with metallicity (e.g., Rémy-Ruyer et al. 2014). If such a relation would be unchanged at z = 7 (e.g., Popping et al. 2017), the measurements imply a metallicity of $12+\mathrm{log}({\rm{O}}/{\rm{H}})\lesssim 7.5$ (corresponding to $\lesssim 0.1\,{{\rm{Z}}}_{\odot }$), which is only found in dwarf galaxies in the local universe with masses ${M}_{\mathrm{star}}\lesssim {10}^{8}\,{M}_{\odot }$ (e.g., Izotov et al. 2015; Guseva et al. 2017).

We use the upper limits on the IR continuum to place an upper limit on the dust mass. We follow the method outlined in Ota et al. (2014), which assumes a dust mass absorption coefficient ${k}_{\nu }=1.875{(\nu /239.84)}^{{\beta }_{\mathrm{IR}}}$ m² kg⁻¹, where ν is in GHz and ${\beta }_{\mathrm{IR}}=1.5$ and removes the contribution from the CMB. This results in ${M}_{\mathrm{dust}}\lt 8.1(27.5)\times {10}^{6}\,{M}_{\odot }$ if the dust temperature is 35 (25) K. These limits are consistent with expectations for CR7's UV luminosity, based on post-processed hydrodynamical simulations by Mancini et al. (2016), who find dust masses between $\approx 1\mbox{--}10\times {10}^{5}\,{M}_{\odot }$.

Following the prescription from Kennicutt (1998), we measure a 3σ upper limit of the dust-obscured ${\mathrm{SFR}}_{\mathrm{IR}}\lt 5.4\,{M}_{\odot }$ yr⁻¹ for a Salpeter (1955) IMF. This allows us to put a stronger constraint on the total SFR of different clumps. Because of these tight limits, the large errors in β (which for example allow relatively red UV slopes and hence relatively large dust obscurations) are mitigated. Combining these measurements (by constraining that the difference between dust-corrected SFR and dust-free SFR is $\lt 5.4\,{M}_{\odot }$ yr⁻¹ in each of the 10,000 realizations used to self-consistently compute the uncertainties as described in Section 2) results in $\mathrm{SFRUV}+\mathrm{IR}={28}_{-1}^{+1},{5}_{-1}^{+2},{7}_{-1}^{+1}\,{M}_{\odot }$ yr⁻¹ for clumps A, B, and C, respectively. For the total (ground-based) photometry, we find $\mathrm{SFRUV}+\mathrm{IR}={45}_{-2}^{+2}\,{M}_{\odot }$ yr⁻¹, which is roughly consistent with the sum of the HST detected clumps and indicates no other significant source of star formation.

Because of the initial nondetections of [C ii] emission in galaxies with strong Lyα emission (e.g., Ouchi et al. 2013; Ota et al. 2014), it has been speculated that the [C ii] luminosity at fixed SFR is related to the strength of Lyα emission (e.g., Knudsen et al. 2016; Smit et al. 2017), as [C ii] emission may be reduced in ISM conditions that favor high Lyα escape. Indeed, conditions that may lower the strength of [C ii] emission are typically found in Lyα emitters. Compared to the general galaxy population, strong Lyα emitters at $z\approx 2\mbox{--}3$ are observed to have low metallicities and high-ionization states (e.g., Erb et al. 2016; Nakajima et al. 2016; Trainor et al. 2016; Kojima et al. 2017). Our results, however, do not agree with such scenario, as CR7 has an extremely high Lyα luminosity and EW, but has [C ii] luminosities similar to those expected for the local relation (both integrated, and for individual clumps). Therefore, the strength of Lyα emission may not be the most important property driving the scatter in the $\mathrm{SFR}-{{\rm{L}}}_{[{\rm{C}}{\rm{II}}]}$ relation. Hence, indications of a potential relation between Lyα strength and [C ii] luminosity may have reflected a more fundamental correlation.

A property that is related to the strength of Lyα emission in star-forming galaxies across $z\approx 0\mbox{--}5$ is the UV slope, with bluer galaxies being brighter in Lyα (e.g., Atek et al. 2008; Matthee et al. 2016; Oyarzún et al. 2017). The UV slope β is closely related to the age, metallicity, and dust-attenuation law (e.g., Bouwens et al. 2012; Duncan & Conselice 2015; Mancini et al. 2016; Narayanan et al. 2017), and therefore indirectly with properties that determine the [C ii] luminosity at fixed SFR. We illustrate this in Figure 10, where we show the [C ii] luminosity as a function of the UV slope for galaxies from the literature with SFR between 15 and 35 ${M}_{\odot }$ yr⁻¹. This is the range of SFRs in Figure 9 that includes most galaxies at $z\approx 5\mbox{--}7$ that have currently been observed. We note that we explicitly investigate a relatively narrow range in SFR to prevent being diluted by relations between SFR and galaxy properties themselves (for example, it has been established that fainter galaxies have bluer colors, Bouwens et al. 2012). Although the error-bars on UV slope measurements are large,¹⁷ it can be seen that at fixed SFR, galaxies with redder colors (which are typically more evolved) have higher [C ii] luminosities. Since β depends on several parameters, direct measurements of these parameters (such as metallicity) are required to fully understand the origin of this trend, and which parameter is most important.

Zoom In Zoom Out Reset image size

It is possible to compare our results in more detail to the results obtained for Himiko in Ouchi et al. (2013). Himiko consists of three UV components, with similar total UV and Lyα luminosity as CR7. Himiko has similar limits on its dust mass, but has not been detected in [C ii], with a limiting luminosity ${L}_{[{\rm{C}}{\rm{II}}]}\lesssim 5\times {10}^{7}\,{L}_{\odot }$ (Ouchi et al. 2013), which is well below the total [C ii] luminosity from CR7. However, unlike CR7, the individual components in Himiko have SFRs of $\approx 7,5$ and 8 ${M}_{\odot }$ yr⁻¹ (Ouchi et al. 2013), and are hence likely of lower mass than CR7s clump A. The SFRs of Himiko's components are comparable to CR7 clumps B and C, that have lower [C ii] luminosities than the existing [C ii] limits for Himiko (and the clumps in Himiko are all very blue, with $\beta \lesssim -2.0$). Deeper [C ii] observations at the resolution we obtained for CR7 could perhaps be able to detect those clumps. Another explanation could be that Himiko has a lower dynamical mass, such that supernova explosions disperse [C ii] emitting gas over larger areas, decreasing its observability.

As shown in Figure 11, the [C ii] emission in CR7 coincides spatially with the Lyα emission. This may not be unexpected if both Lyα and [C ii] originate from star formation. Although [C ii] is also emitted from gas that is (partly) neutral (due to its slightly lower ionization energy than hydrogen; for example, photo-dissociating regions), Lyα emission scatters in these neutral gas regions, such that it may also be observed from these regions.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Interestingly, the [C ii] line width and velocity offset of clump A are surprisingly similar to the best fitted shell model to CR7's Lyα profile (Dijkstra et al. 2016), which has an intrinsic line width of 259 km s⁻¹ and outflow velocity of 230 km s⁻¹ (compare to Table 1). Their best fitted shell model assumes negligible dust content (consistent with our constraints on the IR luminosity), and the resulting shell-model parameters are similar to Lyα emitters and analog galaxies at lower redshifts. Therefore, the resemblance of the shell-model parameters based on the Lyα profile with the observed velocity offsets indicates feedback processes are already present in CR7. Further detailed studies on how the Lyα and [C ii] line profiles are related spatially require high resolution Lyα IFU observations with, e.g., MUSE, as our current Lyα spectral information is limited to clump A.

As shown in Figures 4 and 6, CR7 consists of several different components of [C ii] emission, at different velocities. Similar clumpy structures are also observed in [O iii]${}_{88\mu {\rm{m}}}$ and [C ii] emission in the z = 7.1 galaxy BDF3299 (Carniani et al. 2017), and also in [C ii] emission in a relatively massive galaxy at z = 6.1 (Jones et al. 2017). Such relatively complex [C ii] structures and the observed velocity offsets are also observed in hydrodynamical simulations of galaxies at $z\approx 6\mbox{--}7$ with halo masses of $\sim {10}^{11}\,{M}_{\odot }$ (e.g., Vallini et al. 2013; Pallottini et al. 2017b). The line profiles do not show evidence for broad-line components due to outflows (e.g., Gallerani et al. 2016). Therefore, we are likely observing satellites (B, B-2, C, C-2) falling into a more massive central galaxy. These satellites have velocity widths similar to those of satellites in simulations by Pallottini et al. (2017a, 2017b), although simulated satellites typically have lower masses than those estimated for CR7's clumps. Clump A is the component that has the highest dynamical mass estimate (see Table 1, although we stress that these are rough estimates) and likely the central galaxy. The dynamical mass estimate of clump C-2 is on the same order of magnitude, potentially indicating a major merger. Observations at higher resolution and with improved sensitivity are required to robustly resolve different components and perform more detailed kinematical/dynamical analyses.

Here, we investigate what our new measurements mean for the interpretation of CR7. For a more in depth discussion on the updated detection significance of Heii and limits on other high-ionization UV lines, we refer to Sobral et al. (2017). The main conclusion from Sobral et al. (2017) is that the most luminous component of CR7 is likely undergoing a recent burst of star formation, with no clear evidence for AGN activity and no convincing detection of UV metal lines in clump A.

Initially, Sobral et al. (2015) argued that different UV components in CR7 were in different evolutionary stages, with the majority of the stellar mass possibly being found in clumps B and C (see also Agarwal et al. 2016). This interpretation was mostly based on the different UV slopes, with clump A being the bluest and hence the youngest. However, there are two uncertainties related to this interpretation. The first is that UV slopes are relatively uncertain (see Table 1), and thus consistent with being the same for all sources. The second is that it is challenging to estimate where the majority of stellar mass is from Spitzer/IRAC observations, which has a relatively large PSF. For example, contrary to Sobral et al. (2015), Bowler et al. (2017) found that the majority of stellar mass is likely associated to clump A using a deconvolution of the IRAC data based on UV detections as a prior. The new observations point toward a scenario where most of the mass is in clump A as the dynamical mass estimate of clump A is higher than those of clumps B and C. Simultaneously, the constraints from the UV slope, IR continuum, and the Lyα luminosity indicates that clump A is also the youngest.

As the SFR-[C ii] ratios of different clumps are similar, there is evidence that all clumps have similar metallicities. However, [C ii] luminosity may be reduced due to photoevaporation by far-UV (6–13.6 eV) and ionizing ($\gt 13.6$ eV) photons emitted from young stars in the vicinity, which typically happens a few tens of megayears after the burst of star formation (e.g., Decataldo et al. 2017; Vallini et al. 2017), which can affect different clumps differently. The metallicities that we would infer from [C ii] ($\approx 0.1\mbox{--}0.2\,{{\rm{Z}}}_{\odot }$) are inconsistent with the metallicity inferred by Bowler et al. (2017), which is $\approx 0.005\,{{\rm{Z}}}_{\odot }$. However, their result depends on a Heii strength that is now ruled out (Sobral et al. 2017). Based on upper limits on UV metal lines, the SFR and photoionization modeling, Sobral et al. (2017) inferred a metallicity of $\approx 0.05\mbox{--}0.2\,{{\rm{Z}}}_{\odot }$, which is consistent with the metallicity indicated by the [C ii] luminosity. Therefore, it is ruled out that the luminosity in all clumps in CR7 is dominated by PopIII-like stars, but note that small amounts of PopIII-like stars in unpolluted pockets of gas could still be present (e.g., Pallottini et al. 2015). The fact that the SFR-[C ii] ratio is similar to other $z\approx 6\mbox{--}7$ galaxies and galaxies in the local universe also indicates that photoevaporation does not play a major role in lowering the [C ii] luminosity.

Due to its extreme Lyα luminosity $(\approx 10\,{L}^{\star }$; Matthee et al. 2015), CR7 could be powered by an AGN, particularly as the AGN fraction of Lyα emitters with similar luminosities at $z=2\mbox{--}5$ approaches 100% (J. Calhau et al. 2017, in preparation). However, these galaxies typically have (much) broader Lyα lines and C iv luminosities exceeding the observational limits for CR7 (e.g., D. Sobral et al. 2017, in preparation), such that they are not fully comparable. The data presented in this paper do not favor an AGN explanation, particularly for clump A, because AGNs typically have much lower log₁₀(${L}_{[{\rm{C}}{\rm{II}}]}/{L}_{\mathrm{IR}}$) ratios (e.g., Ota et al. 2014). Furthermore, Bowler et al. (2017) show that the UV emission in clump A is slightly extended, which could also be at odds with an AGN-dominated scenario. Therefore, it is likely that clump A in CR7 is powered by a burst of star formation with moderately low metallicity. It is remarkable that there is no UV nor IR continuum emission observed at the position of the ALMA detection C-2, which has a similar dynamical mass estimate as clump A. If potential UV emission is obscured, the dust is likely at a higher temperature (such that our IR continuum luminosity limit would be weak).

In order to improve our understanding of the CR7 system and other similar systems at $z\sim 7$, it is most important to obtain more accurate constraints on the stellar masses, ages, and metallicities of different clumps. While deeper HST imaging results in more strongly constrained UV slopes, the majority of progress will be made with integral field spectroscopic observations in the IR, which can trace the rest-frame optical. Future adaptive optics assisted observations with MUSE can further map out the relation between the Lyα and [C ii] kinematics, and simultaneously probe the large-scale environment on ∼0.3–7 Mpc scales.