2. TRACKING THE NEUTRAL GAS RESERVOIR OVER COSMIC TIME

DLAs are historically defined as Lyα absorbers with neutral hydrogen column densities exceeding N(H I) = 2 × 10²⁰ cm⁻² (Wolfe et al 2005), corresponding to a surface mass density limit of Σ_atomic ≈ 2 M_⊙ pc⁻² for atomic gas (including helium). The large gas surface mass densities revealed in high-redshift DLAs are comparable to what is seen in 21 cm observations of nearby star-forming galaxies (e.g., Walter et al 2008, Leroy et al (2008)), making DLAs a promising signpost of young galaxies in the distant Universe (Wolfe et al 1986). In addition, the N(H I) threshold ensures that the gas is neutral under the metagalactic ionizing radiation field (e.g., Viegas 1995, Prochaska and Wolfe 1996, Prochaska et al 2002). Neutral gas provides the seeds necessary for sustaining star formation. Therefore, observations of DLAs not only help establish a census of the cosmic evolution of the neutral gas reservoir (e.g., Neeleman et al 2016a), but also offer a unique window into star formation physics in distant galaxies (e.g., Lanzetta et al 2002, Wolfe and Chen 2006).

While the utility of DLAs for probing the young Universe is clear, these objects are relatively rare (see the right panel of Fig. 2) and establishing a statistically representative sample of these rare systems requires a large sample of QSO spectra. Over the last decade, significant progress has been made in characterizing the DLA population at z ≳ 2, owing to the rapidly growing spectroscopic sample of high-redshift QSOs from the Sloan Digital Sky Survey (SDSS; York et al 2000). The blue points in the right panel of Fig. 2 are based on ∼ 1000 DLAs and ∼ 500 strong LLS identified at z ≈ 2−5 in an initial SDSS DLA sample (Noterdaeme et al 2009). The sample of known DLAs at z ≳ 2 has continued to grow, reaching ∼ 10,000 DLAs found in the SDSS spectroscopic QSO sample (e.g., Noterdaeme et al 2012).

The large number of known DLAs has led to an accurate characterization of the neutral gas reservoir at high redshifts. Figure 3a displays the observed N(H I) distribution function column density distribution function, f_DLA, based on ∼ 7000 DLAs identified at z ≈ 2−5 (Noterdaeme et al 2012). The plot shows that f_DLA is well represented by a Schechter function (Schechter 1976) at log N(H I) ≲ 22 following

(1)

with a shallow power-law index of α ≈ −1.3 below the characteristic H I column density log N_*(H I) ≈ 21.3 and a steep exponential decline at larger N(H I) (Noterdaeme et al 2009, 2012). At log N(H I) > 22, the observations clearly deviate from the best-fit Schechter function. However, DLAs are also exceedingly rare in this high-N(H I) regime. Only eight such strong DLAs have been found in this large DLA sample (Noterdaeme et al 2012), making measurements of f_DLA in the two highest-N(H I) bins very uncertain. In comparison to f_{N(H I)} established from 21 cm maps of nearby galaxies (Zwaan et al 2005), the amplitude of f_DLA at z ≳ 2 is ≈ 2 × higher than f_{N(H I)} at z ≈ 0 but the overall shapes are remarkably similar at both low- and high-N(H I) regimes (Fig. 3a; see also Sánchez-Ramírez et al 2016, Rafelski et al 2016).

Figure 3. Summary of known DLA properties: (a) evolving neutral hydrogen column density distribution functions column density distribution function, fN(H I) from DLAs at z = 2−3 (Noterdaeme et al 2012) to H I galaxies at z ≈ 0 (Zwaan et al 2005); (b) declining cosmic neutral gas mass density with increasing Universe age (or decreasing redshift) from observations of DLAs (solid points from Noterdaeme et al 2012, open circles from Prochaska and Wolfe 2009, open squares from Crighton et al 2015, and open triangle from Neeleman et al 2016a) following Eq. (2), local H I galaxies (green shaded box, a compilation from Neeleman et al 2016a), and molecular gas (blue shaded boxes, Decarli et al 2016), in comparison to increasing cosmic stellar mass density in galaxies with increasing Universe age (grey asterisks, a compilation from Madau and Dickinson 2014); (c) gas-phase metallicity (Z) relative to Solar (Z⊙) as a function of redshift in DLAs (grey squares for individual absorbers and blue points for N(H I)-weighted mean from Marc Rafelski, Rafelski et al 2012, Rafelski et al 2014), IGM at z ≳ 2 (orange circles, Aguirre et al 2008, Simcoe 2011), ISM of starburst galaxies at z ≈ 2−4 (light magenta boxes, Pettini et al 2001, Pettini 2004, Erb et al 2006a, Maiolino et al 2008, Mannucci et al 2009), intracluster medium in X-ray luminous galaxy clusters at z ≲ 1 (red triangles, Balestra et al 2007), H I-selected galaxies (green box, Zwaan et al 2005), and stars at z = 0 (dark purple box, Gallazzi et al 2008); and (d) molecular gas fraction, fH2 versus total surface density of neutral gas scaled by gas metallicity for high-redshift DLAs in triangles (Noterdaeme et al 2008, 2016), γ-ray burst host ISM in star symbols (e.g., Noterdaeme et al 2015), and local ISM in the Milky Way (Wolfire et al 2008) and Large and Small Magellanic Clouds (Tumlinson et al 2002) in dots, blue circles, and cyan squares, respectively.

At log N(H I) > 21, numerical simulations have shown that the predicted shape in f_DLA is sensitive to the detailed ISM physics, including the formation of molecules (H₂) and different feedback processes (e.g., Altay et al 2011, , Bird et al 2014). Comparison of the observed and predicted f_DLA therefore provides an independent and critical test for the prescriptions of these physical processes in cosmological simulations. However, the constant exponentially declining trend at N(H I) ≳ 2 × 10²¹ cm⁻² between low-redshift H I galaxies and high-redshift DLAs presents a puzzle.

At z = 0, the rapidly declining f_{N(H I)} at N(H I) ≳ N_*(H I) has been interpreted as due to the conversion of atomic gas to molecular gas (Zwaan and Prochaska 2006, Braun 2012). As illustrated at the end of this Section and in Fig. 3d, the column density threshold beyond which the gas transitions from H I to H₂ depends strongly on the gas metallicity, and the mean metallicity observed in the atomic gas decreases steadily from z ≈ 0 to z > 4 (Fig. 3c). Therefore, the conversion to molecules in high-redshift DLAs is expected to occur at higher N(H I), resulting in a higher N_*(H I) with increasing redshift. However, this is not observed (e.g., Prochaska and Wolfe 2009, Sánchez-Ramírez et al 2016, Rafelski et al 2016; Fig. 3a). Based on spatially resolved 21 cm maps of nearby galaxies with ISM metallicity spanning over a decade, it has been shown that f_{N(H I)} established individually for these galaxies does not vary significantly with their ISM metallicity (Erkal et al 2012). Together, these findings demonstrate that the exponential decline of f_DLA at N(H I) ≳ N_*(H I) is not due to conversion of H I to H₂, but the physical origin remains unknown.

Nevertheless, the observed f_DLA immediately leads to two important statistical quantities: (1) the number density of DLAs per unit survey pathlength, obtained by integrating f_DLA over all N(H I) greater than N₀ = 2 × 10²⁰ cm⁻² and (2) the cosmic neutral gas mass density, contained in DLAs, Ω_atomic, which is the N(H I)-weighted integral of f_DLA following

(2)

where µ = 1.3 is the mean atomic weight of the gas particles (accounting for the presence of helium), H₀ is the Hubble constant, c is the speed of light, and ρ_crit is the critical density of the Universe (e.g., Lanzetta et al 1991, Wolfe et al 1995). The shallow power-law index α in the best-fit f_DLA, together with a steep exponential decline at high N(H I) from the Schechter function in Eq. (1), indicates that while DLAs of N(H I) < N_*(H I) dominate the neutral gas cross-section (and therefore the number density), strong DLAs of N(H I) ∼ N_*(H I) contribute predominantly to the neutral mass density in the Universe (e.g., Zwaan et al 2005). A detailed examination of the differential Ω_atomic distribution as a function of N(H I) indeed confirms that the bulk of neutral gas is contained in DLAs of N(H I) ≈ 2 × 10²¹ cm⁻² (e.g., Noterdaeme et al 2012).

The cosmic evolution of ρ_gas observed in DLAs, from Eq. (2), is shown in black points in Fig. 3b. Only measurements based on blind DLA surveys are presented in the plot ¹. These include an early sample of ≈ 700 DLAs at z = 2.5−5 in the SDSS Data Release (DR) 5 (open circles; Prochaska and Wolfe 2009), an expanded sample of ≈ 7000 DLAs in the SDSS DR12 (solid points; Noterdaeme et al 2012), an expanded high-redshift sample of DLAs at z = 4−5 (open squares; Crighton et al 2015), and a sample of ≈ 14 DLAs at z ≲ 1.6 from an exhaustive search in the Hubble Space Telescope (HST) UV spectroscopic archive (open triangle; Neeleman et al 2016a).

A range of mean H I mass density at z ≈ 0 has been reported from different 21 cm surveys (see Neeleman et al 2016a for a recent compilation). These measurements are included in the green box in Fig. 3b. Despite a relatively large scatter between different 21 cm surveys and between DLA surveys, a steady decline in Ω_atomic is observed from z ≈ 4 to z ≈ 0. For comparison, the cosmic evolution of the molecular gas mass density obtained from a recent blind CO survey (Decarli et al 2016) is also included as blue-shaded boxes in Fig. 3b, along with the cosmic evolution of stellar mass density measured in different galaxy surveys, shown in grey asterisks (data from Madau and Dickinson 2014). Figure 3b shows that the decline in the neutral gas mass density with decreasing redshift is coupled with an increase in the mean stellar mass density in galaxies, which is qualitatively consistent with the expectation that neutral gas is being consumed to form stars. However, it is also clear that atomic gas alone is insufficient to explain the observed order-of-magnitude gain in the total stellar mass density from z ≈ 3 to z ≈ 0, which implies the need for replenishing the neutral gas reservoir with accretion from the intergalactic medium (IGM) (e.g., Kereš et al 2009, Prochaska and Wolfe 2009). At the same time, new blind CO surveys have shown that molecular gas contributes roughly an equal amount of neutral gas mass density as atomic gas observed in DLAs at z ≲ 3 (e.g., Walter et al 2014, Decarli et al 2016), although the uncertainties are still very large. Together with the knowledge of an extremely low molecular gas fraction in DLAs (see the discussion on the next page and Fig. 3d), these new CO surveys indicate that previous estimates of the total neutral gas mass density based on DLAs alone have been underestimated by as much as a factor of two. An expanded blind CO survey over a cosmological volume is needed to reduce the uncertainties in the observed molecular gas mass densities at different redshifts, which will cast new insights into the connections between star formation, the neutral gas reservoir, and the ionized IGM over cosmic time.

Observations of the chemical compositions of DLAs provide additional clues to the connection between the neutral gas probed by DLAs and star formation (e.g., Pettini 2004). In particular, because the gas is predominantly neutral, the dominant ionization for most heavy elements (such as Mg, Si, S, Fe, Zn, etc.) are in the singly ionized state and therefore the observed abundances of these low-ionization species place direct and accurate constraints on the elemental abundances of the gas (e.g., Viegas 1995, Prochaska and Wolfe 1996, Vladilo et al 2001, Prochaska et al 2002). Additional constraints on the dust content and on the sources that drive the chemical enrichment history in DLAs can be obtained by comparing the relative abundances of different elements. Specifically, comparing the relative abundances between refractory (such as Cr and Fe) and non-refractory elements (such as S and Zn) indicates the presence of dust in the neutral gas, the amount of which increases with metallicity (e.g., Meyer et al 1989, Pettini et al 1990, Savage and Sembach 1996, Wolfe et al 2005). The relative abundances of α- to Fe-peak elements determine whether core-collapse supernovae (SNe) or SNe Ia dominate the chemical enrichment history, and DLAs typically exhibit an α-element enhanced abundance pattern (e.g., Lu et al 1996, Pettini et al 1999, Prochaska and Wolfe 1999).

Figure 3c presents a summary of gas metallity (Z) relative to Solar (Z_⊙) measured for > 250 DLAs at z ≲ 5 (grey squares from Rafelski et al 2012, ). The cosmic mean gas metallicity in DLAs as a function of redshift can be determined based on a N(H I)-weighted average over an ensemble of DLAs in each redshift bin (blue points), which is found to increase steadily with decreasing redshift following a best-fit mean relation of ⟨ Z / Z_⊙ ⟩ = [−0.20 ± 0.03] z −[0.68 ± 0.09] (dashed blue line, Rafelski et al 2014). For comparison, the figure also includes measurements for stars (dark purple box, Gallazzi et al 2008) and H I-selected galaxies (green box, Zwaan et al (2005)) at z = 0, iron abundances in the intracluster medium in X-ray luminous galaxy clusters at z ≲ 1 (red triangles, Balestra et al 2007), ISM of starburst galaxies (light magenta boxes) at z ≈ 2−3 (Pettini et al 2001, Pettini 2004, Erb et al 2006a) and at z = 3−4 (Maiolino et al 2008, Mannucci et al 2009), and IGM at z ≳ 2 (orange circles, Aguirre et al 2008, Simcoe 2011).

It is immediately clear from Fig. 3c that there exists a large scatter in the observed metallicity in DLAs at all redshifts. In addition, while the cosmic mean metallicity in DLAs is significantly higher than what is observed in the low-density IGM, it remains lower than what is observed in the star-forming ISM at z = 2−4 and a factor of ≈ 5 below the mean values observed in stars at z = 0. The chemical enrichment level in DLAs is also lower than the iron abundances seen in the intracluster medium at intermediate redshifts. The observed low metallicity relative to the measurements in and around known luminous galaxies raised the question of whether or not the DLAs probe preferentially low-metallicity, gas-rich galaxies and are not representative of more luminous, metal-rich galaxies found in large-scale surveys (e.g., Pettini 2004).

The large scatter in the observed metallicity in DLAs is found to be explained by a combination of two factors (Chen et al 2005): (i) the mass-metallicity (or luminosity-metallicity) relation in which more massive galaxies on average exhibit higher global ISM metallicities (e.g., Tremonti et al 2004, Erb et al 2006a, Neeleman et al 2013, Christensen et al 2014) and (ii) metallicity gradients commonly seen in star-forming disks with lower metallicities at larger distances (e.g., Zaritsky et al 1994, van Zee et al 1998, Sánchez et al 2014, Wuyts et al 2016). If DLAs sample a representative galaxy population including both low-mass and massive galaxies and probe both inner and outer disks of these galaxies, then a large metallicity spread is expected.

The observed low metallicity in DLAs, relative to star-forming ISM, is also understood as due to a combination of DLAs being a gas cross-section selected sample and the presence of metallicity gradients in disk galaxies (Chen et al 2005). A cross-section selected sample contains a higher fraction of absorbers originating in galaxy outskirts than in the inner regions, and the presence of metallicity gradients indicates that galaxy outskirts have lower metallicities than what is observed in inner disks (see Sect. 3 and Fig. 4 below for more details). Indeed, including both factors, a gas cross-section weighting scheme and a metallicity gradient, for local H I galaxies resulted in a mean metallicity comparable to what is observed in DLAs (green box in Fig. 3c; Zwaan et al 2005).

While DLAs exhibit a moderate level of chemical enrichment, searches for molecular gas in DLAs have yielded only a few detections (e.g., Noterdaeme et al 2008, Jorgenson et al 2014, Noterdaeme et al 2016). Figure 3d displays the observed molecular gas fraction, which is defined as f_H2 ≡ 2 N(H₂) / [N(H I) + 2 N(H₂)], versus metallicity-scaled total hydrogen column density for ≈ 100 DLAs at z ≈ 2−4 (triangles). The DLAs span roughly two decades in N(H I) from N(H I) ≈ 2 × 10²⁰ cm⁻² to N(H I) ≈ 2.5 × 10²² cm⁻². Strong limits have been placed for f_H2 for the majority of DLAs at f_H2 ≲ 10⁻⁵ with only ≈ 10% displaying the presence of H₂ and two having f_H2 > 0.1. In contrast, the ISM of the Milky Way (MW), at comparable N(H I), displays a much higher f_H2 than the DLAs at high redshifts.

The formation of molecules is understood to depend on two competing factors: (i) the ISM radiation field which photo-dissociates molecules and (ii) dust which facilitates molecule formation (e.g., Elmegreen 1993, Cazaux and Spaans 2004). Dust is considered a more dominant factor because of its dual roles in both forming molecules and shielding them from the ISM radiation field. In star-forming galaxies, the dust-to-gas mass ratio is observed to correlate strongly with ISM gas-phase metallicity (e.g., Leroy et al 2011, Rémy-Ruyer et al 2014). It is therefore expected that the observed molecular gas fraction should correlate with gas metallicity (e.g., Elmegreen 1989, Krumholz et al 2009, Gnedin et al 2009).

In the MW ISM with metallicity roughly Solar, Z ≈ Z_⊙, the molecular gas fraction is observed to increase sharply from f_H2 < 10⁻⁴ to f_H2 ≳ 0.1 at N(H I) ≈ 2 × 10²⁰ cm⁻² (see Wolfire et al 2008). The sharp transition from atomic to molecular is also observed in the ISM of the Large and Small Magellanic Clouds (LMC and SMC), but occurs at higher gas column densities of N(H I) ≈ 10²¹ cm⁻² for the LMC and N(H I) ≈ 3 × 10²¹ cm⁻² for the SMC (see Tumlinson et al 2002). The ISM metallicities of LMC and SMC are Z ≈ 0.5 Z_⊙ and Z ≈ 0.15 Z_⊙, respectively. These observations therefore support a simple metallicity-dependent transitional gas column density illustrated in Fig. 3d. Following the metallicity-scaling relation, it is clear that despite a high N(H I), most DLAs do not have sufficiently high metallicity (and therefore dust content) to facilitate the formation of molecules (Gnedin and Kravtsov 2010, Gnedin and Draine 2014, Noterdaeme et al 2015). This finding also applies to γ-ray burst (GRB) host galaxies (star symbols in Fig. 3d). With few exceptions (Prochaska et al 2009, Krühler et al 2013, Friis et al 2015), the ISM in most GRB hosts displays a combination of very high N(H I) and low f_H2 (e.g., Tumlinson et al 2007, Ledoux et al 2009). The observed absence of H₂ in DLAs, together with a large molecular mass density revealed in blind CO surveys (e.g., Walter et al 2014, Decarli et al 2016), shows that a complete census for the cosmic evolution of the neutral gas reservoir requires complementary surveys of molecular gas over a broad redshift range. In addition, as described in Sect. 4 below, the observed low molecular gas content also has important implications for star formation properties in metal-deficient, high neutral gas surface density environments.

¹ At z ≲ 1.6, DLA surveys require QSO spectroscopy carried out in space and have been limited to the number of UV-bright QSOs available for absorption line searches. Consequently, the number of known DLAs from blind surveys is small, ≈ 15 (see Neeleman et al 2016a for a compilation). To increase substantially the sample of known DLAs at low redshifts, Rao & Turnshek (Rao et al 2006) devised a clever space programme to search for new DLAs in known Mg II absorbers. Their strategy yielded a substantial gain, tripling the total sample size of z ≲ 1.6 DLAs. However, the Mg II-selected DLA sample also includes a survey bias that is not well understood. It has been shown that excluding Mg II-selected DLAs reduces the inferred Ω_atomic by more than a factor of four (e.g., Neeleman et al 2016a). For consistency, only measurements of Ω_atomic based on blind DLA surveys are included in the plot. Back.