Annu. Rev. Astron. Astrophys. 2005. 43: 861-918 Copyright © 2005 by Annual Reviews. All rights reserved

8. STAR FORMATION IN DAMPED Ly SYSTEMS

Several lines of evidence imply that damped Ly systems experience ongoing star formation. At z < 1.6 the evidence is unambiguous, since more than half of the sample is associated with galaxies of stars. At z 1.6 the evidence is indirect because only one damped Ly system is identified with a resolved object of galactic dimensions that emits starlight. While starlight likely ionizes the gas that gives rise to Ly emission in two other high-z damped Ly systems, the starlight itself has not been directly detected. The expectation is that most of the metals seen in damped Ly systems were generated through star formation in their host systems, but the instantaneous (SFRs) characterizing this important population of objects remained unknown until recently.

8.1. Direct Emission Measurements of Star-Formation Rates in Damped Ly Systems with z > 1.6

Table 1 summarizes results including SFRs for the three positive detections and for objects with upper limits on SFR at z 1.6. Observed lower limits on L(H) were coverted to upper limits on SFR using the Kennicutt (1998) calibration, SFR (M year^-1) = 7.9 × 10^-42 L (H )(ergs s^-1). Detections of Ly emission ⁷ were combined with the expression for L (H) and case B radiative recombination to find lower limits of SFR (M year^-1) = 1.1 × 10^-42 L(Ly) (ergs s^-1). The total magnitude integrated over DLA2206 - 19A of V = 23 and the very sensitive B > 27 limit on DLA0953 + 47A (A. Bunker 2004, priv. comm.) were combined with the rest-frame UV Kennicutt (1998) calibration, SFR (M year^-1) = 1.4 × 10^-28 L (ergs s^-1 Hz^-1), to set a lower limit on the SFR for DLA2206 - 19A and a rough upper limit for DLA0953 + 47A assuming no dust extinction for the latter object. An estimated dust extinction correction for DLA2206 - 19A (Wolfe et al. 2004) suggests an upper limit on SFR of 50 M year^-1, which is comparable to the SFRs of the more luminous Lyman Break Galaxies (Shapley et al. 2003).

8.2. Star-Formation Rates from the C II* Technique

A method recently developed by Wolfe, Prochaska & Gawiser (2003) makes it possible to infer the SFR per unit area for individual damped Ly systems. The basic method is to use measurements of the C II* 1335.7 column density to measure the [C II] 158 µm cooling rate in the neutral gas producing the damped Ly absorption. This is possible because the C II* 1335.7 transition arises from the excited ²P_3/2 state in C⁺, and spontaneous photon decay of the ²P_3/2 state to the ²P_1/2 state results in [C II] 158 µm emission, which is the principal coolant of neutral gas in the Galactic ISM (Wright et al. 1991). Under the presumed condition of thermal balance, the cooling rate equals the heating rate and it is possible to calculate the SFR per unit area that generates the implied heating rate.

By analogy with the Galactic ISM, Wolfe, Prochaska & Gawiser (2003) adopt the grain photoelectric effect as the principal heating mechanism for damped Ly systems. In that case FUV radiation (h 6-13.6 eV) ejects photoelectrons from grain surfaces, which heat ambient electrons through Coulomb interactions (Bakes & Tielens 1994, Weingartner & Draine 2003). The heating rate from the grain photoelectric effect is proportional to the FUV radiation intensity J, which consists of a contribution from the FUV background radiation plus a local contribution from hot stars located inside the galaxy that is proportional to the instantaneous SFR per unit area, _*.

To infer _* and other properties from observations, Wolfe, Prochaska & Gawiser (2003) deduce the [C II] 158 µm spontaneous emission rate per H atom from the observational quantity:

(8)

where A_ul is the Einstein coefficient and h_ul is the energy of the 158 µm transition. These are known quantities, and N(C II^*) and N(HI) are measured from the absorption line spectra. The total heating rate includes inputs due to the grain photoelectric effect, X-ray photoionization, cosmic ray ionization, C I photoionization, and collisional heating. The total cooling rate includes cooling due to Ly, grain radiative recombination, and emission by fine-structure states of O⁰, Si⁺, Fe⁺, etc., in addition to the [C II] 158 µm fine-structure emission. Radiative excitation of the C⁺ fine-structure states by CMB radiation is included because it can be significant at high redshift. Wolfe et al. (2004) showed that the heating rates predicted for the Haardt & Madau (2003) backgrounds were significantly lower than the 158 µm cooling rates implied for damped Ly systems with detected C II* 1335.7 absorption, requiring active star formation to explain the observed cooling rates. On the other hand, Wolfe et al. (2004) also showed that heating by background radiation alone cannot be ruled out for systems in which C II* 1335.7 absorption was not detected. A summary of the _c data for damped Ly systems (and the Galaxy) is given in Figure 12. Wolfe, Prochaska & Gawiser (2003) argued that the heating rate in the Galaxy is much larger than that in damped Ly systems because the dust content in the Galaxy is at least a factor of 30 higher than that in damped Ly systems, whereas _* in the Galaxy is only a factor of 2 to 3 lower than that in damped Ly systems (see below).

Figure 12. c versus N(H I) for sample of 52 damped Ly systems. Red data points are positive detections, blue are 2 upper limits, and green are 2 lower limits. Small stars depict positive detections from sightlines in the Galaxy ISM. Large star depicts [C II] 158 µm emission rate per H atom averaged over the disk of the Galaxy. The latter is about 30 times higher than the average of the DLA detections.

This method was used by Wolfe et al. (2004) to analyze a sample of 45 damped Ly systems, 23 with measured C II* column densities and 22 for which C II* was not detected. In the redshift interval z = [1.6, 4.5] the average SFR per unit area for positive detections is <_*> = 11.3 × 10^-3 M year^-1 kpc^-2 for the CNM solution and 0.21 M year^-1 kpc^-2 for the WNM solution; by comparison with the Galaxy _* 4 × 10^-3 M year^-1 kpc^-2. However, the WNM solution is unlikely to be correct because the bolometric background intensity produced exceeds the observational limits (see Wolfe, Gawiser & Prochaska 2003). This conclusion is consistent with recent evidence from individual systems; Howk, Wolfe & Prochaska (2005) showed that the low upper limit on the optical-depth ratio of Si II* 1264.7 to C II* 1335.7 in a high-redshift damped Ly system results in an upper limit of 800 K for the temperature in the gas producing C II*, which implies that C II* absorption arises in CNM gas in this system. Furthermore, a pure WNM solution would significantly overpredict the observed rest-frame-UV luminosity from DLA2206 - 19A based on its measured _c, implying that C II* absorption in this damped Ly system arises in a CNM. It is significant that the predicted value for CNM, J^CII* = 1.7^+2.7_-1.0 × 10^-18 ergs cm^-2 s^-1 Hz^-1 sr^-1, is the largest mean intensity inferred from the entire C II* sample. This may help explain why DLA2206 - 19A is one of the rare damped Ly systems detected in emission.

On the other hand, the gas detected in absorption is more likely to be a WNM for damped Ly systems with upper limits on C II* absorption. In this case the gas could be a pure single-phase WNM heated by background radiation alone. These damped Ly systems would then be objects without significant star formation at the absorption epochs. Or the gas could be the WNM branch of a two-phase medium in which the SFR per unit area is similar to that found for the CNM solutions (Wolfe et al. 2004). While the absence of 21 cm absorption at z > 3 (Kanekar & Chengalur 2003) and the large C II/C I ratios (Liszt 2002) are consistent with the WNM hypothesis, both phenomena are also naturally explained within the context of the two-phase hypothesis. Clearly, it is important to determine which of these explanations is the correct one.

8.3. Implications

The C II* technique can also be used to obtain quantities with cosmological significance. Specifically, the SFR per unit comoving volume is given by _* = <_*(z)>(A_* / A_p)(d / dX)(H₀ / c), where A_* is the intrinsic cross-sectional area occupied by stars emitting FUV radiation and A_p is the projection of the intrinsic H I area, A_HI, on the sky. ⁸ In the uniform disk model, neutral gas and stars are uniformly distributed across A_HI, yielding the values for _* (z) given in Table 2, which also summarizes the systematic uncertainties discussed at length in Wolfe, Gawiser & Prochaska (2003).

TABLE 2. Global properties

Property	0.0-1.6	1.6-4.5
d / dX	....	0.077 ± 0.016 b
103 g(z)	0.96 ± 0.23 a	0.92 ± 0.21 b
log10 *(z)	<   -1.40	-0.70 ± 0.28c
< Z >	-0.81 ± 0.032	-1.33 ± 0.09
a Due to large uncertainties in individual measurements, error in mean given by propagation of experimental errors. b Because of systematic decrease in d / dX and g(z) with time, error in mean determined by standard deviation. cComputed by averaging over the "WD low" and other dust models discussed in Wolfe et al. (2003a) (see their Table 1), and by assuming that both DLAs with detected and undetected C II* absorption have same mean SFR per unit area. If the SFR per unit area of the non-detections were significantly lower, * would decrease by about 0.3 dex.

Wolfe, Gawiser & Prochaska (2003) also calculated the mass density of metals produced in damped Ly systems by using the conversion of SFR to metal formation rate suggested by Madau et al. (1996) and Pettini (1999), _metals = (1/42) _*. Extrapolating to the present day, including the contribution from star formation in normal galaxies at z < 1.6, yields a mass of metals a few times larger than that found in spiral bulges today, which seems feasible. Integrating under the DLA cosmic star-formation history at z 2.5 predicts a factor of 30 higher cosmic mean metallicity due to metal enrichment of neutral gas than is observed in damped Ly systems at z = 2.5 (Prochaska et al. 2003a). This sort of "missing metals" problem was first identified by Pettini (1999) for Lyman Break Galaxy SFRs compared to damped Ly system metallicities, but that problem could be solved by assuming that damped Ly systems are not the descendants of the star-forming Lyman Break Galaxies. The damped Ly system "missing metals" problem appears to be fundamental and may illuminate a fundamental flaw in our understanding of the relationship between metal enrichment and star formation at high redshift. The problem is a significant challenge not only to the C II* method of measuring damped Ly SFRs, but to most hierarchical models, which also produce too many metals (see Nagamine, Springel & Hernquist 2004b; Somerville, Primack & Faber 2001). ⁹ Given the other successes of the method it seems likely that the problem has a physical solution.

The simplest solution would be to hypothesize that the damped Ly systems forming stars at z 2.5 are no longer damped Ly systems at z = 2.5 but have used up most of their neutral gas and now have the expected metallicities; the Lyman Break Galaxies are an obvious candidate for their descendants. This is inconsistent, however, with timescale of 2 Gyr to use up most of the neutral gas implied by the observed SFRs, which is too long. Ejecting the metals into the IGM does not solve the problem, as the IGM metallicity predicted at z = 2.5 would be [M/H] = -1.5, i.e., about a factor of 30 higher than that observed in the Lyman forest, unless the ejected gas is so hot that most metals are in an unobservable ionization state, with the possible exception of oxygen, which might be detected in O VI 1031.9, 1037.6. An alternative possibility is that the metals are sequestered, with the most attractive solution being a model in which damped Ly systems represent neutral gas on the outskirts of actively star-forming "bulge" regions with the majority of the produced metals remaining in these regions rather than polluting the damped Ly gas or general IGM (Wolfe, Gawiser & Prochaska 2003). This hypothesis necessitates the conclusion that metal-rich bulge regions are underrepresented (or completely unrepresented) in the H I-weighted cosmic mean metallicity measured from DLAs. That would not be surprising given that the bulges may well have used up most of their neutral gas and/or have sufficient dust content to dim background QSOs out of optically selected samples.

⁷ Due to its extreme sensitivity to dust extinction, Ly emission provides a lower limit to the star formation rate as long as the AGN contribution to the emission line is negligible. Back.

⁸ Note that A_p equals A(N, X) (see eq. 1) averaged over the column-density interval N(H I) = [N_min, N_max] Back.

⁹ The model of Pei et al.(1999) avoids this problem by invoking significant obscuration corrections and low yields. Back.