Astron. Astrophys. 342, 395-407 (1999)

7. Summary and discussion

We have studied a complex Lyman limit system at z = 1.9 towards the UV bright QSO HS 1103+6416 using a combination of ultraviolet HST spectra of low resolution with optical high-resolution spectra. This combination for the first time allows a quantitative study of abundances of several heavy elements (C, O, N, S, Si, S, Al) and of the ionization conditions by using the observed subcomponent splittings from the optical (CII, SiII, AlII,..) to synthesize lines from ions like OII, OIII, OIV, NII, NIII, SIII, etc.) at intrinsic EUV wavelengths.

We find that the complex absorption system that spans a velocity range of roughly 200 km s^-1 with at least 11 components can be subdivided into three groups: A low ionization subgroup L: components 2, 3, 6 with radial velocities v = -129 to -95 km s^-1, an intermediate ionization group I: components 4, 5, 7, 8 (v = -75... +2), and a high-ionization subgroup H with components 9, 10, 11 (v = +3.4... +57). Component 1 at -129 km s^-1 appears to belong to group H.

Ionization calculations show, that ion ratios in components L (ionization parameter log U -3.1) and I (log U = -2.9 to 3) are best fitted with a rather hard radiation field (M3; power law with = -0.5 and a break by factor 10 at the HeII edge) while for the high ionization component H the best fit is obtained with the metagalactic UV radiation background as predicted by Haardt & Madau (1996). The probable implication is that while the `Halo' component H is purely photoionized by the metagalactic UV background, the low and intermediate ionization clouds, which possibly differ only in electron density, need an additional source of ionization, which cannot be local stars, since a harder radiation field compared to the Haardt & Madau (1996) UV background is required. This result and the detection of OVI, which cannot be produced by photoionization models, both provide evidence for collisional ionization as a further ionization mechanism or deviations from strict photoionization equilibrium (Rauch et al., 1997). Our results on ionization are similar to what Boksenberg (1997) has found in his study of strong metal line systems of the QSO HS 1626+6433 (Reimers et al. 1995a). Besides the systematic variation of ionization with velocity, we do see distinctive abundance differences, in particular between the high ionization components H on one hand and the low and intermediate ionization components L and I on the other hand.

For the `Halo' components H we find [C/H] = -1.2 and for the -elements O, Si and S roughly [/C] = 0.4 0.1, while [N/C] = 0.2. For oxygen, a similar overabundance at low metal abundances is known from both QSO absorption systems at z = 2 (Vogel & Reimers 1995) and from metal deficient stars (e.g. Israelian et al. 1998). This is in accordance with SNII nucleosynthesis in massive stars before lower mass stars contribute significantly to heavy element production (see McWilliam 1997).

Components L and I show [C/H] -1 and only slight ([Si/C] = 0.2) or no ([O/C] [S/C] [Al/C] 0) overabundance, within the uncertainties. This is still consistent with what is seen in the galactic disk at 1/10 solar abundance (Fig. 3a in McWilliam 1997).

If all components (L, I and H) were ionized by the same radiation field, the overabundance of Si relative to C would be even larger up to [/C] = +1 in the highly ionized components, which is not consistent with what is known from metal deficient `Halo' stars.

While the systematic trends of ionization and abundances with velocity, i.e. different locations in a galaxy or group of galaxies, appear to be real and consistent with evidence from other sources of information (Halo stars, QSO absorption line systems) we have to remember that neither the fit results nor the model results are quantitatively unambiguous.

Possibly both low and high ions are a mixture from different gas phases with different ionization mechanisms as found for the Milky Way (Savage 1987). In this case, the observed column density ratios cannot be used to constrain photoionization models. Also, the assumption of individual, isolated clouds producing absorption lines might be inadequate (see Miralda-Escudé et al. 1997).

Optical data of higher signal-to-noise are needed to obtain more reliable information on column densities and b-values especially for weak absorption lines, and higher resolution in the UV range provided in the future by the COS spectrograph will allow to improve constraints on the models via the large number of intrinsic EUV quasar absorption lines.

© European Southern Observatory (ESO) 1999

Online publication: February 22, 1999
helpdesk.link@springer.de