Astron. Astrophys. 333, 841-863 (1998)

6. Discussion

One first goal of our study was to observe the Ly line from our candidate high systems. All of them turn out to have larger than  cm^-2, which confirms the validity of the Fe II /Mg II criterion used to select some of the systems (in EX 0302-223 and Q 1209+107).

6.1. Metal abundances

The metal abundance estimates discussed in Sect. 4.1 have been gathered in Table 6. For all systems, relative abundances are given using our own determination of . The Zn II and Cr II column densities given by Meyer & York (1992) and Steidel et al. (1995) have been adjusted according to the f values used in this paper (the N(Fe II) derived by these authors from a curve of growth analysis have been adopted although they rely in part on f values for Fe II 2249 and Fe II 2260 which were larger than the revised ones; these N(Fe II) may then be slightly underestimated, by 0.1 to 0.2 dex). Thus, our abundance estimates can be directly compared to those given by Lu et al. (1996) and Pettini et al. (1997a). The uncertainty is typically 0.1 - 0.2 dex except for 3C 196 and Q 1209+107 which have a less accurate measurement. As is customary, relative abundances have been computed assuming and (Fe), etc (except for Ca which may be partly in the form of Ca III).

Table 6. Relative metal abundances in damped Ly systems at ([X/H] except for Ca II for which is given)

We first compare the pattern of relative abundances observed at high redshift by Lu et al. (1996) (see also Prochaska & Wolfe 1996) and discussed by Pettini et al. (1997b) and Kulkarni et al. (1997) to that of low redshift absorbers. The estimates reported in Table 6 appear in rough agreement with the compilation of [X/Zn] and [X/Fe] values presented by Kulkarni et al. (1997) for high redshift systems. Although covering a broad range, values for [Mn/H] and [Ni/H] are roughly centered on the medians found at higher redshift, [Mn/Zn] [Ni/Zn] (Kulkarni et al. 1997), whereas the values (including one upper limit) derived for [Fe/Zn] are all lower than that at higher redshift [Mn/Zn] . Then, to first order, depletion onto dust grains seems to be effective in low redshift absorbers as it is at high redshift (Pettini et al. 1994; Pettini et al. 1997b; Kulkarni et al. 1997); since in our Galaxy Zn is only slightly depleted, we shall follow Pettini et al. (1994) and use [Zn/H] as a metallicity indicator. The [Fe/H] value in PKS 1229-021 is atypical in the sense that the upper limit inferred from our data appears very low when compared to [Mn/H] or, to a lesser extent, [Ni/H]. We note that in a recent study, Vladilo et al. (1997) find a [Mn/Fe] ratio of 0.4 in a candidate DLAS while our results on PKS 1229-021 imply [Mn/Fe] ; both values appear high when compared to the high z estimates ([Mn/Fe] ). However, given the difficulties quoted above for the measurement of [Fe/H] in the PKS 1229-021 DLAS, any conclusion (regarding e.g. the intrinsic nucleosynthetic pattern involved) would be premature and we stress that complementary observations (e.g. of Fe II 2367 or Fe II 2374) would be very valuable.

In the two cases where bright intervening spiral galaxies are present (systems in 3C 196 and Q 1209+107), the available Mn and Fe measurements suggest a relatively high metallicity. Assuming that the abundance of these elements relative to Zn is similar to that at high z, one gets estimates for [Zn/H] of -0.2 and -0.6 in 3C 196 and Q 1209+107 respectively. For the former, the value estimated for [Zn/H] could be higher by 0.6 dex if the N(H I) value given by Cohen et al. (1996) had been adopted instead of ours. Similarly, for the latter, [Zn/H] could also be higher by about 0.3 dex as it was derived using [Fe/H] (see above and Table 6). Using the conservative estimates of [Zn/H] given above together with the measurement in PKS 1229-021 ([Zn/H] 0.5), we find that the new low z absorbers studied in this paper have [Zn/H] . Pettini et al. (1997a) have plotted all available measurements of [Zn/H] as a function of redshift (their Fig. 3). In this diagram, the three values quoted above indicate that systems with higher metallicities are present at lower redshift, as expected from cosmic chemical evolution (Pei & Fall 1995). Pettini et al. (1997a) also present a binned version of this plot (their Fig. 4); in the low redshift bin which includes four systems at , the column density weighted average metallicity is [Zn/H] at . Including our three new estimates yields a higher value, [Zn/H] at . A few other DLAS have been measured recently. In 3C 336, Steidel et al. (1997) find [Fe/H] and  cm^-2 at . Two more DLAS have also been discovered in the spectra of the bright QSOs HE 1122-1649 (at ) and HE 0515-4414 (at ) with values for [Zn/H] of and -0.85 and for of  cm^-2 and  cm^-2 respectively (de la Varga & Reimers 1998). We checked that inclusion of the latter does not modify the above value for [Zn/H] (this is because they involve low H I column densities and because it is appropriate to consider the N(H I) weighted average).

To summarize, we find that extending the sample at lower redshifts brings [Zn/H] somewhat closer to the prediction of a metallicity approaching Solar values as goes to zero (Pei & Fall 1995), and further, that a large scatter is present, as at high redshift. The latter is likely to reflect the large variety of the observed absorber morphologies (Paper I) and the spread in impact parameters (Phillips & Edmunds 1996).

6.2. The bias induced by dust and physical conditions in the absorbers

Several authors have considered that the extinction induced by dust within intervening galaxies may affect the statistics of distant QSOs (Ostriker & Heisler 1984; Boissé & Bergeron 1988; Fall & Pei 1993). Fall & Pei (1993) stressed that the same bias also affects the statistics of DLAS themselves, especially of those with the largest and developed a method to correct for these effects. Boissé (1994, 1995) pointed out that, because dust is closely linked to metallicity and to the formation of molecules, extinction probably results in a preferential selection of QSOs with systems displaying a low metallicity and H₂ content.

It is noteworthy that the dust obscuration bias is expected to have the strongest effect at z 1, precisely in our range of interest (Fall & Pei 1995; their Fig. 5); this is because metallicity decreases while the extinction per dust grain in the observer's frame increases (due to the rising extinction law) when going at higher redshift. Indeed, in the present study, we find that the two faintest QSOs (which could not be observed with the HST at a resolution appropriate to detect metal lines) display systems with relatively high metallicities. We also find no very large values (e.g.  cm^-2) while this could have been expected given the low impact parameters involved (e.g. in 3C 196). As discussed by Boissé (1995), the absence of such systems which would be quite easily recognized in the numerous low resolution optical spectra obtained so far, further supports the reality of a cut-off caused by extinction.

These results alone are only indicative but after the extensive work by Pettini et al. (1994, 1997a) and Lu et al. (1996), there now exists a reasonably large sample of systems with and [Zn/H] measurements (again, since Zn does not deplete onto grains, it is a good indicator of the amount of metals available to form dust) and effects of the extinction bias might become apparent in the data themselves. We then plot [Zn/H] as a function of H I column density, since the latter should determine to first order the strength of the effect (Fig. 19 ; the system in PKS 0528-250 has not been included). We note a very clear deficiency of systems having at the same time a large and a high metallicity (the absence of data points in the bottom left part is just due to observational limitations). The available measurements appear to be distributed in the diagram as if the amount of metals along the line of sight, as estimated from N(Zn), were constrained to be less than about  cm^-2, which corresponds to the line drawn in Fig. 19. This result is all the more striking as, in an unbiased sample, one would have rather expected the opposite trend since the large values should correspond to the innermost parts of galaxies where the metallicity is presumably higher. To verify that the evolution of [Zn/H] in redshift does not affect our analysis (through a possible variation of with ), we consider separately measurements at (the median of values) and ; no tendency is seen for high z systems to cluster at large .

Fig. 19. [Zn/H] versus for 37 damped Ly systems; small symbols correspond to and large symbols to . The line in the upper right corresponds to  cm-2 or to Galactic material inducing (see text)

One alternative explanation would involve the opacity of Zn lines; in the upper right part of Fig. 19, N(Zn II) is large and the optically thin approximation could no longer be valid which would result in an underestimate of [Zn/H].

While this may be true in a few cases (two measurements in Fig. 19 are in fact given as lower limits) the high resolution data presented by Lu et al. (1996) indicates that the whole set of [Zn/H] values is unlikely to be affected by saturation effects. We then conclude that dust extinction is effective at inducing a preferential selection of QSOs with systems having a low metal content.

The upper bound on N(Zn) present in Fig. 19 can be interpreted as an upper bound on of about 0.3 for Galactic-type material; the latter is assumed to be characterized by [Zn/H] (Sembach et al. 1995) and we have adopted the ratio given by Bohlin et al. (1978). For larger values of , significant effects are indeed expected to occur since in the rest frame of the absorber, it is the UV part of the extinction curve that is involved. It is often argued that, given the low dust-to-gas ratio inferred from the observed systems (Pei et al. 1991), the effects of extinction are necessarily small. Indeed, if extinction effects were estimated assuming the average dust-to-gas ratio inferred from the reddening of QSOs with DLAS, one would predict negligible effects in Fig. 19. However, this reasoning does not take into account the properties of the systems that are missed and which are very likely to be those with the largest dust-to-gas ratio. Fall & Pei (1993) precisely pointed out that the (unknown) dispersion in the dust-to-gas ratio is an important limitation in our ability to correct for the effects of this bias. Metal abundances in DLAS (as in nearby galaxies) show a large scatter at any redshift, and similarly large variations can be expected for the dust-to-gas ratio itself, including the possibility that galaxies with dust-to-gas ratio and metallicities even larger than Galactic exist, since we have no reason to believe that the Milky Way is particularly dust and metal rich. Fig. 19 confirms that in the available samples, a selection with respect to absorbing gas properties is present.

A direct consequence of the bias induced by dust extinction is that estimates of , the total amount of gas (see e.g. Wolfe et al. 1995; Storrie-Lombardi et al. 1996), are necessarily highly uncertain. First, the relative contribution to of large systems is important and these are precisely the most heavily biased. Second, H₂, which is not taken into account, may represent a non-negligible fraction of the total gas mass. In the local Universe, this fraction is estimated to be in the range 30 - 50 % (Casoli et al. 1997). The evolution in redshift of this quantity is unknown but we already observe that by 1 - 2, galaxies form stars at a high rate which strongly suggests the presence of a large H₂ mass. This is also indicated by CO observations of distant galaxies (Solomon et al. 1992; Omont et al. 1996; Alloin et al. 1997) which probably give a lower limit to the true amount of H₂ because the CO to H₂ conversion factor is likely to be higher due to a lower metallicity. Note however that, to first order, the uncertainty on the H₂ amount does not affect the method developed by Fall & Pei (1993) since their estimate of extinctions is based primarily on the column density of metals, which is an observed quantity.

As noted above, the extinction bias is expected to have large effects at . This is also true at because the UV spectroscopic observations required to study these systems are very demanding in terms of QSO brightness. This tends to reduce the evolution of [Zn/H] (which further helps to understand the discrepancy discussed by Pettini et al. 1997a between the observed low z bin and model predictions) and might also mimic an evolution of the shape of the distribution, such as that observed by Wolfe et al. (1995).

Since molecules and dust are closely related, the extinction bias is likely to affect the apparent H₂ content as well. One point that has been overlooked when discussing absorption line studies is that such methods yield very little information on the densest phases of the interstellar medium where stars form. This is for two reasons. First, the surface coverage factor of the latter is quite small and therefore the probability to intersect intervening material of this type is very low. Second, any background QSO will be strongly dimmed by the associated extinction and will remain undetected or appear too faint for spectroscopic studies; this effect is reinforced by the marked tendency of molecules to assemble in dense opaque clouds, due to self-shielding. For galaxies at 1, we have ample evidence for the presence of dense and dusty molecular clouds like those in our own Galaxy (Wiklind & Combes 1996; Casoli et al. 1996). Nevertheless, when searching for dust and molecules from low z systems in QSO spectra, we get essentially the same null result as at higher redshift!

Therefore, one must not draw any definite conclusion from the observed low amount of molecules and lack of clear signatures from dust grains in DLAS. Only the observation of fainter QSOs would lead to a more representative view of the interstellar medium in these distant galaxies (Boissé 1994, 1995). Even if H₂ molecules have not been detected in our QSO sample, the likely detection of C I in two systems suggests that physical conditions (radiation field, density, ...) are not very different from those in the Solar neighborhood, at least in some absorbers. A similar conclusion is reached by Ge & Bechtold (1997) for gas at .

6.3. Relation between damped absorber morphology and intervening gas properties

The galaxies proposed in Paper I as the absorbers have not yet been confirmed spectroscopically. This will be a difficult task because in ground-based observations a large fraction of the QSO light is superimposed onto the galaxy emission and a redshift can be measured only if the latter shows prominent spectral signatures. Additional problems arise from the detection of several metal systems, each of them implying the presence of a galaxy close to the QSO line of sight. Since we now have a detailed census of metal systems in four out of six QSOs from our sample, let us summarize the status of the proposed identifications and discuss their reliability:
- EX 0302-223: four objects are present at low impact parameter (all with a relatively compact morphology). Given its lower impact parameter, galaxy #2 is the best candidate. The other three galaxies, of which two are very close and probably form an interacting system, could give rise to the metal-rich systems at and 1.3284. The absorber is identified with a very bright galaxy at large impact parameter (Guillemin & Bergeron 1997).
- PKS 0454+039: there is little ambiguity in this case since only one object is seen near the QSO. The galaxy is very compact and the QSO line of sight probes regions which are outside the stellar component.
- 3C 196: the large spiral is very likely to be at 0.437 because its luminosity would have to be extremely large if it were at 0.87. This very extended galaxy does not show any [O II ]3727 emission but the Ca II absorption doublet has been tentatively identified; in galaxy # 3, strong emission lines at a redshift close to that of the QSO have been detected (Drouet d'Aubigny & Bergeron, in preparation). However, the latter may not be the absorber, since the intervening gas does not cover the whole broad line region and should thus be very close to the active nucleus (Cohen et al. 1996),
- Q 1209+107: the proposed identification with galaxy #2 is reliable because no other candidate is present close to the line of sight. Furthermore, its color is consistent with a spiral galaxy at ,
- PKS 1229-021: although we could reject objects # 2, #4 and #6 as potential absorbers since they are associated with knots in the QSO radio jet, two objects (a compact galaxy, #3, and a low surface brightness galaxy, #5) could be at 0.3950. Galaxy #3 is favored because of its lower impact parameter,
- 3C 286: three fairly bright objects (#2a, #2b and #2c) have been detected at less than 1 arcsec from the QSO line of sight. All three could contribute to the DLAS. Alternatively, since #2a and #2b are located roughly symmetrically to the QSO, they could be part of the QSO's host galaxy, #2c being then the absorber.

Some of these identifications could be erroneous if, as in the case of 3C336 (Steidel et al. 1997), none of the galaxies close to the QSO line of sight turned out to be at the DLAS redshift. However, from a statistical point of view, the presence of galaxies at low impact parameter in all the studied cases cannot be due to chance and very likely, most of the proposed identifications are correct.

Among the gas properties that may be connected to absorber morphology let us consider metallicity and kinematics. We already noted that the two cases with intervening spiral galaxies correspond to relatively high metallicities. The low [Zn/H] value measured in 3C 286 in spite of the small impact parameter is consistent with the proposed absorber of fairly low surface brightness. This case may then be similar to the DLAS studied by Steidel et al. (1997): no galaxy could be detected despite an intensive search which suggests a faint object centered on the QSO. The two compact absorbers display intermediate metallicities; these galaxies are reminiscent of the blue nucleated galaxies that appear at 0.5 in galaxy surveys (Schade et al. 1996). The most intriguing case is the DLAS in PKS 1229-021: although the amorphous aspect suggests an unevolved object, the Zn and Mn abundances are clearly higher than in the two previous cases, with a possibly unusual abundance pattern.

The absorber in PKS 1229-021 is also remarkable for the kinematics of the absorbing gas. The asymmetrical velocity distribution revealed by the data obtained by Lanzetta & Bowen (1992) is the prototype of what Prochaska & Wolfe (1997) consider to be the signature of a massive disk. However, at first sight, such an interpretation of absorption profiles is not supported by the appearance of the damped absorber in HST images as discussed by Pettini et al. (1997a). Recent numerical simulations also confirm (at least for the high z DLAS) that the presence of an edge-leading asymmetry is not an unambiguous signature of a rapidly rotating disk (Haehnelt et al. 1997). We also note that in the other case of a proposed low surface brightness galaxy (in 3C 286), the kinematics seems to be very different since all or most of the absorption arises from the single narrow component which induces the 21cm absorption line (Meyer & York, 1992).

Detailed information on the kinematics is also available for the damped Ly system in PKS 0454+039 (Lu et al. 1996). Absorption is spread more or less continuously over 140  km s^-1 with four maxima and no characteristic asymmetry. No simple picture emerges from the yet very scarce systems for which the required data are available. A larger sample is necessary to check if the absorption profiles are related to the large-scale kinematics of the intervening galaxy itself (as proposed by Prochaska & Wolfe 1997) or rather are governed by e.g. the recent star formation activity (through supernovae explosions). In the present context, the two systems induced by bright spiral galaxies (in 3C 196 and Q 1209+107) are of much interest and would clearly deserve high resolution spectroscopic observations.

6.4. Perspectives

The results obtained in our study and the above discussion clearly show the need for complementary observations. Obviously, it is very desirable to confirm spectroscopically the candidate absorbers proposed in Paper I, especially when several galaxies are present at low impact parameter. Then, it is important to improve the accuracy of some measurements presented in this paper and extend the set of metal elements considered in order to compare in more details the abundance patterns seen at low and high z. In particular, it is important to confirm the high metallicity suggested for the systems caused by intervening spiral galaxies in the spectra of 3C 196 and Q 1209+107. For the two latter cases, the absorber properties are very well characterized by HST images; high resolution optical spectra would provide the velocity distribution of the gas and thus help to establish the phenomena governing the kinematics of the absorbing material.

Moreover, the observed broad range of absorber morphologies and absorbing gas properties indicate that a larger sample is needed to get a correct overall view of the origin of low z damped Ly systems. Possibly, some types of absorbers have not yet been detected; the recent study by Lanzetta et al. (1997) shows that early-type galaxies do also contribute to the damped Ly absorber population. Presently, only two or three absorbers of each type have been found and it is not possible to assess the relative contribution of the various classes of galaxies. Observations of several absorbers from the same class, but probed at different impact parameters and inclinations, would be crucial to understand the relation between the absorbing gas and the parent galaxy properties.

Finally, as discussed in Sect. 6.2, we strongly suspect that surveys of damped Ly samples drawn from the observation of fainter QSOs would reveal more systems with large N(H I) and high metallicity, whatever the redshift. However, the most reddened QSOs would not be observable in their UV rest-frame and the intervening gas properties should then be derived from observations at longer wavelengths. Nevertheless, searches for damped Ly absorption in QSOs fainter by two or three magnitudes could reveal absorption systems in which the presence of dust grains and molecules is much more conspicuous than in the systems investigated so far, and thus would allow a more comprehensive study of the evolution of the interstellar medium in galaxies. Spectroscopic surveys of fainter QSOs should then not only enlarge the present samples but also lead to a more representative view of the whole diversity of the damped Ly absorber population.

© European Southern Observatory (ESO) 1998

Online publication: April 28, 1998

helpdesk.link@springer.de