Witnessing Galaxy Assembly at the Edge of the Reionization Epoch^*

We inspected the XSHOOTER spectrum by eye, to detect prominent absorption systems. Absorption lines were identified and then fit with Voigt profiles using the context LYMAN of the ESO MIDAS software package (Fontana & Ballester 1995).

The most interesting system in the analyzed spectrum⁹ is a low ionization absorber at z = 5.938646 ± 0.000007 (Δv ≃−2746 km s⁻¹ from the quasar emission redshift) for which we detect strong transitions due to O i λ 1302 Å, C ii λ 1334 Å, Si ii λλ 1260, 1304 Å and Fe ii λλ 2344, 2382, 2586, 2600 Å (Figure 1). The absorption line due to Al ii λ 1670 Å is severely affected by a sky line. The presence of the likely saturated O i absorption suggests that this could be a DLA, because O i is a tight tracer of neutral hydrogen that has a very similar ionization potential. There is also a possible detection of the associated weak C iv doublet at a slightly different redshift, z = 5.9392 ± 0.0001, corresponding to Δv ≃ 24 km s⁻¹. A shift between low and high ionization transitions is often observed in damped systems.

Zoom In Zoom Out Reset image size

DLAs at these redshifts generally lack the information on the H i column density because the spectrum blueward of the quasar Lyα emission is completely absorbed. In the present case, the DLA falls in the region within 5000 km s⁻¹ of the quasar systemic redshift (called the "proximity region"). This allowed us to estimate the column density, N(H i), by fitting the damping wing at the position of the Lyα in emission (Figure 1). To fit the DLA profile, we estimated the continuum level in the region of the Lyα/Lyβ emissions with the following procedure. We considered all of the Principal Component Analysis continua computed for the quasars in the XQ-100 sample (López et al. 2016). The continua were redshifted to the systemic redshift of J2310 and compared to its XSHOOTER spectrum. Only two continua matched reasonably well the large width of the observed emission lines of J2310 (Figure 1, upper panel). We fit the DLA H i Lyα and Lyβ absorption lines using the two continua separately to estimate a more reliable uncertainty on the measure. The H i lines were centered at the redshift determined from the fit of the low ionization heavy element transitions. The adopted column density and its error, $\mathrm{log}\,N$(H i) = 21.05 ± 0.10 [cm⁻²], are the average of the two column density measurements and the half width of the interval spanned by the measurements and their 1σ errors, respectively. Figure 1 reports the results of the fit for all detected transitions. Absorption lines due to O i, C ii, and Si ii were fitted simultaneously adopting the same Doppler parameter, b = 8.0 ±0.4 km s⁻¹. This is a reasonable assumption, as metal lines are not resolved in the XSHOOTER spectrum. Detected lines of the Fe ii multiplet were fitted together (excluding the noisy Fe ii 2600 Å line) resulting in a Doppler parameter b = 10.5 ± 1.5 km s⁻¹. Column densities for the observed ions and relative chemical abundances are reported in Table 1 and discussed in Section 4.1.

Table 1.  Properties of the DLA

Ion	$\mathrm{log}\,N$	[X/H]a	[X/Fe]
H i	21.05 ± 0.10	⋯	⋯
C iib	14.73 ± 0.18	≥−2.95	≥0.1
O ib	15.00 ± 0.12	≥−2.9	≥0.2
Si ii	13.70 ± 0.09	−2.86 ± 0.14	0.22 ± 0.12
Fe ii	13.47 ± 0.06	−3.08 ± 0.12	⋯

Notes.

^aSolar abundances and corresponding uncertainties from Asplund et al. (2009). ^bDetected absorption lines are possibly saturated.

Download table as:  ASCIITypeset image

Table 2.  Properties of Serenity-18

ALMA J231038.44+185521.95
R.A. (J2000)	23:10:38.44
Decl. (J2000)	18:55:21.95
Redshift of CO(6–5) emission	5.93957
Impact parameter [arcsec]	6.7
FWHMCO(6–5) [km s−1]	155 ± 30
$\int {S}_{\mathrm{CO}(6\mbox{--}5)}{dv}\,[\mathrm{Jy}\,\mathrm{km}\,{{\rm{s}}}^{-1}]$	0.06 ± 0.012
L'CO(6–5) [K km s−1 pc−2]	(2 ± 0.4) × 109
M(H2) [M⊙]	(5.4 ± 0.5) × 109
Mdyn sin2(i) [M⊙]	≤5.6 × 109
LFIR [L⊙]	≈1012
SFR [M⊙ yr−1]	≈115

Download table as:  ASCIITypeset image

Following the identification of the proximate DLA absorber in the spectrum of J2310, we analyzed our ALMA data and found an emission line at a frequency 99.642 GHz, corresponding to a velocity of −2710 km s⁻¹ with respect to the quasar host galaxy (z_em = 6.0025). The line is located at the very edge of the bandpass, therefore the continuum-level can be estimated only on the right-hand side of the line (Figure 2, right panel) and there is the possibility that part of the emission is falling outside the observed bandpass. This implies that the integrated flux of the emission line and its derived quantities might be lower limits. A Gaussian fit to the emission line gives an FWHM = 155 ± 30 km s⁻¹, and an integrated intensity of 0.053 ± 0.01 Jy km s⁻¹.

Zoom In Zoom Out Reset image size

By integrating over the line width, and collapsing the corresponding spectral channels, we find a 5σ emission on the map at a position R.A., decl. (23:10:38.437, 18:55:22.00), or about (−6.3,+2.2) arcsec offset from the quasar position (Figure 2, left panel). This is well within the primary beam of the ALMA antennas in band 3. The sky frequency corresponds to the expected frequency of the CO(6–5) emitted at the redshift of the DLA absorption system, therefore it can be identified with the CO(6–5) emission from a galaxy associated with the DLA. We estimate the probability of a chance coincidence between a foreground CO emitter (CO(3–2) at z ≃ 2.47 or CO(4–3) at z ≃ 3.63) and the DLA absorption, based on the expected number of emission lines in ALMA band 3 with flux similar to the one observed here (Walter et al. 2016). Assuming that a redshift difference Δv ≲ ±100 km s⁻¹, comparable with the rotation velocity of a disk at these redshifts, is required to associate the CO emission and the DLA absorption lines, we obtain a probability of chance alignment of ∼1%–2%.

We have also measured the CO line in the Fourier plane, after averaging visibilities over the line width and shifting the corresponding visibility table to the phase center. A fit with a point source gives a zero spacing flux density of 251 ± 51 μ Jy, which corresponds to an integrated intensity of Sdv = 0.06 ± 0.012 Jy km s⁻¹, consistent with that derived from the fit of the spectrum. The source may appear slightly more extended than the synthesized beam on the integrated map (Figure 2, central panel). To verify this possibility, we also fitted the visibilities with a circular or elliptical Gaussian function, but the fit does not converge. We conclude that the source is not resolved in our data.

The galaxy is not detected in the 3 mm continuum, down to a 3σ upper limit of 16.2 μJy/beam.

We found no detection in the rest-frame 150 μm continuum probed by band 6 ALMA data, down to a 3σ upper limit of 150 μJy/beam. For a typical spectral energy distribution (SED) of a star-forming galaxy with a dust temperature in the range T_dust = 30–50 K, and located at this redshift, this limit corresponds to a total infrared (IR) luminosity L(FIR) ≈ 1.2 × 10¹² L_⊙. We note that the bandpass does not include the expected frequencies of other far-infrared (FIR) emission lines, included the brightest one, namely [C ii]. The properties of Serenity-18 are summarized in Table 2.

On the basis of the large H i column density measured for the DLA system, we assumed that no ionization corrections were needed and carried out the computation of the relative chemical abundances based on the column densities and the solar abundances from Asplund et al. (2009). The iron and silicon abundances, [Fe/H] = −3.08 ± 0.12 and [Si/H] = −2.86 ± 0.14, place this DLA absorber in the very metal-poor regime as defined by Cooke et al. (2011). The absorption lines due to C ii λ 1334 Å and O i λ 1302 Å could be saturated, thus we could derive only lower limits to the abundances of C and O. Previous studies of DLAs (e.g., Vladilo et al. 2011; Rafelski et al. 2012) have shown that below metallicities [Fe/H] ≈ −2 dust corrections are negligible; as a consequence, we are computing abundances assuming a dust-free gas.

The studied absorber has abundances in very good agreement with those measured for the sample of z ≈ 6 DLAs by Becker et al. (2012).

As already discussed in Becker et al. (2012), the chemical abundances observed for our system are also consistent with the 95% confidence interval of the abundances for the sample of very metal-poor DLAs selected by Cooke et al. (2011) at z ∼ 2–4.¹⁰ The observed abundance pattern is well explained by the predictions for PopII progenitors with M ∼ 20 M_⊙.

The column density of H i requires self-shielding of the gas from the cosmic ultraviolet (UV) background, which in turn implies that the absorbing gas density should be larger than ≃0.1 cm⁻³ (Rahmati et al. 2013); this limit translates into a limit on the physical size of ≲4 kpc. Together with the low metallicity, this suggests that we are seeing a gas filament/clump that has been recently forming from the intergalactic medium.

The CO-emitting galaxy is unresolved in our data. We estimate its upper limits size as D ≤ FWHM_beam = 0.6 arcsec, which corresponds to ≈3.6 kpc at the redshift of the galaxy. The inclination on the line of sight cannot be estimated either, because the source is unresolved. We derive the dynamical mass, modulus the inclination, by applying the relation, M_dyn sin²(i) =1.16 10⁵ × (0.75 × FWHM_CO)² × D (Wang et al. 2013; Feruglio et al. 2018), where FWHM_CO = 155 km s⁻¹, and D is the source size in kpc (diameter). From this relation we find, M_dyn sin²(i) ≤ 5.6 10⁹ M_⊙.

In order to estimate the molecular gas mass from CO(6–5) we need to make some assumptions about the S_CO(6–5)/(1–0) ratio and on the luminosity-to-mass conversion factor. We adopt the following, S_CO(6–5)/(1–0) = r₆₁ = 20, which is the average value measured for star-forming galaxies from z = 0 to 4 (Carilli & Walter 2013). We note that an indication of a higher excitation at z ∼ 6, r₆₁ = 70–150, comes from recent hydrodynamical simulations by Vallini et al. (2018).

Concerning the α_CO, the Milky Way value (i.e., 4.3 K km s⁻¹ pc⁻² ${M}_{\odot }^{-1}$) is probably unphysical for high-z galaxies. We therefore adopt a lower value based on Vallini et al. (2018), α_CO = 1.5 K km s⁻¹ pc⁻² ${M}_{\odot }^{-1}$. Based on these assumptions we infer a molecular gas mass of M(H₂) =(5.4 ± 0.5) × 10⁹ × (α_CO/1.5) M_⊙. For different excitation properties of the gas (e.g., Vallini et al. 2018) the mass budget has to be decreased accordingly. For an inclination i = 90° (edge-on) and 50°, the molecular gas fraction in the galaxy would be μ = M(H₂)/M_dyn ∼ 0.9 and 0.6, respectively.

We note that the FWHM estimated by fitting a 1D Gaussian profile to the CO line, FWHM = 155 ± 30 km s⁻¹, is similar to those estimated for [C ii] in typical galaxies at z ≈ 6–7 (FWHM ≈ 150 km s⁻¹, Maiolino et al. 2015; Knudsen et al. 2016; Pentericci et al. 2016; Bradač et al. 2017; Matthee et al. 2017; Carniani et al. 2018a, 2018b). We cannot exclude, however, that part of the line may have been missed out of the bandpass (see Figure 2). In this case, our determinations of the gas mass and dynamical mass would be lower limits.

From the calibration of Greve et al. (2014) between L_FIR and L' CO(6–5), we derive a FIR luminosity of L_FIR ≈ 10¹² L_⊙. This is a very rough estimate with an uncertainty of ∼1 dex due to the scatter of the correlation. We convert L_FIR to a star formation rate (SFR) using the Kennicutt et al. (1998) conversion factor corrected for a Kroupa (2011) initial mass function. We find SFR ≈ 115 M_⊙ yr⁻¹. This value is consistent with the upper limit on the 150 μm continuum estimated from band 6 data. In fact, by assuming a typical SED of a star-forming galaxy with dust temperature in the range T_dust ∼30–50 K, we would expect a 3σ detection of the continuum for a SFR = 200 M_⊙ yr⁻¹ (see also Pallottini et al. 2017; Behrens et al. 2018).

The properties that we have derived for Serenity-18 allows us to state that this is the first detection of molecular gas emission from a typical main sequence galaxy at the end of the reionization epoch (see e.g., Santini et al. 2017).

The difference in redshift between the absorbing gas of the DLA and the emission line is of ∼50 km s⁻¹. Their angular separation is of 6.7 arcsec, corresponding to an impact parameter of ≈40 kpc at the DLA redshift, in a configuration similar to previous results for lower redshift DLAs (e.g., Neeleman et al. 2017, 2018). This confirms the association between the CO-emitting galaxy and the metal-poor absorber. Furthermore, the relatively small redshift difference between the galaxy/DLA system and the quasar (Δv ≃ −2746 km s⁻¹ or ≈4 proper Mpc) suggests that they could all be part of the same large-scale structure.

State-of-the-art cosmological hydrodynamical simulations can help visualize the overall picture, which could give rise to our observations. We note that the luminosity of the CO(6–5) line, the inferred M_dyn and SFR of Serenity-18 are consistent with the predictions for the simulated z ≃ 6 galaxy Althæa (Pallottini et al. 2017; Vallini et al. 2018). In Figure 3, we show the H i column density and metallicity maps for the filament embedding Althæa. In this context, the DLA can be identified with either one of the satellites of Althæa, or with a gas condensation/filament surrounding the galaxy and possibly feeding it with fresh fuel for star formation. This picture is also in agreement with observations of clumpy galaxy assembly at z ≳ 6 (Ouchi et al. 2010; Jiang et al. 2013; Bowler et al. 2017; Matthee et al. 2017; Carniani et al. 2018a, 2018b).

Zoom In Zoom Out Reset image size

The Serenity-18/DLA complex opens a new window in the study of typical galaxies in the early universe. It represents an ideal target for deeper, multiwavelength (UV/optical/NIR/mm) observations both in imaging and spectroscopy to understand the process of galaxy assembly. It suggests that also metal-poor DLAs could be associated with CO-emitting galaxies: this could be true only at the highest redshifts, but it should be tested also at the lower redshifts where it is easier to select DLAs by metallicity.