5. Column density measurements
Fitting of absorption lines with Voigt profiles was done interactively using FIT/LYMAN in MIDAS to estimate column densities and line widths b. Different lines from the same ion were fitted simultaneously, but we fitted different elements and different ionization stages independently. We did not fit the 11 components simultaneously, instead we selected regions for individual fitting, i.e. the three components at the highest redshifts were always fitted simultaneously but independently of the other components.
The C IV 1548 absorption profile is blended with C IV 1550 absorption from the complex absorber system at . A comparison of the Si II line profiles reveals that Si II 1190, 1193 must be blended, too. We did not try to fit the saturated region of Si III absorption. For component 4 we omitted fitting of Si IV and C IV because the data yield no useful constraints. For some components we fitted O I 1302 absorption, but since it shows a different velocity profile it might arise in a different gas phase or might be contaminated by H I Ly absorption. The regions of corresponding Ly, Fe II 1608 and N V 1238 absorption are just shown for comparison in Fig. 3.
A comparison of the fit results in Fig. 3 with the observations might hint at further hidden components. There seems to be an additional component at +20 km s-1 when inspecting C IV , C II and Si II absorption lines. The logarithmic column densities, Doppler parameters and redshifts for the individual components are listed in Table 2. Derived values are uncertain due to intrinsic line blending, the unknown number of components involved and the line fitting process by itself, since the fit results are not unique. The velocities v of the components are given relative to . However, the total column densities derived for each ion in the high, intermediate, and low ionization groups respectively coincide remarkably well with the respective column densities measured by the apparent optical depth method (to within dex 0.15).
Table 2. Column densities of 11 components of the LLS in comparison to model results for different radiation fields. f indicates that the value was fixed. A colon indicates an error km s-1. M1: power law, ; M2: Haardt & Madau (1996); M3: power law, , factor 10 break.
We fitted different ions independently with the intention to examine possible systematic changes in line widths or redshifts. Different b- and z-values of different ions of the same element could hint at an origin in gas phases with different ionization mechanisms, i.e. collisional ionization, photoionization by local sources or the extragalactic radiation field. Furthermore, the differences in b-values for different elements of the same component can yield the contributions of turbulent and/or thermal broadening of the lines.
For the following discussion of redshifts and b-values we will neglect the observed O I 1302 absorption. The errors in the central wavelengths as given by FIT/LYMAN are normally 0.01-0.02 Å, with a maximum error of 0.03 Å. We find differences in z up to 5 10-5 for lines of the same component. We do not find a systematic trend in redshift with ionization level, except for components 6, 7 and 8, where C IV and Si IV lines differ in redshift from the singly ionized elements. But especially these components are strongly blended and the ions might differ in the number of components involved.
Much more difficult is the derivation of the line widths. More than half of the b-values are marked by a colon which indicates that the fit error was greater than 1 km s-1. For the most reliable fits no systematic trend is found for b-values for ions of different ionization level which must mean that the line widths are dominated by nonthermal components. The b-values are not particularly sensitive to the fitting procedure since, as can be seen from Fig. 3, it is always the combination of lines on different parts of the curve of growth that determines the b-values, even in case of SiIV and CIV where the line strengths of the doublet components are sufficiently different (not saturated). The main source of error is the insufficient S/N of the spectra, in particular for the weaker lines.
If there was a tail of shock-heated gas at high temperatures away from the curve of photoionization (Haehnelt et al. 1996), it should manifest itself in higher b values in CIV, which is not seen, or in NV (not seen) and OVI absorption. OVI absorption is probably seen (Fig. 4) although not unambiguously identified due to its location in the Ly forest, which cannot be explained by photoionization and which indeed could be what Haehnelt et al. (1996) have predicted.
Especially for the weaker lines a much better signal-to-noise ratio would improve this analysis. However, if the complex line profiles are due to spatial correlations in stochastic velocity fields (so called mesoturbulence) the b-values derived by standard Voigt profile fitting could be grossly in error (see Levshakov et al. 1997 and references therein).
5.1. The ionization of the clouds
The maximum temperature derived from the largest b-value of Si IV in component 5 (C IV in component 1) is 2 105 K (8 104 K) when assuming pure thermal broadening. At a temperature of 2 105 K, C II should be completely destroyed by collisional ionization (Sutherland & Dopita 1993). But for component 5 strong C II absorption is visible, so that there is bulk motion in addition to thermal motion and the gas is rather photoionized. A hot, highly ionized phase does not seem to exist, otherwise we would expect to see absorption by N V or Ne VIII . We do see absorption at the expected positions of O VI 1031, 1037, but the identification is doubtful due to a possible contamination by H I Ly. On the other hand, Kirkman & Tytler (1997) detected O VI and C IV but no N V absorption in a subcomponent of a LLS at and concluded that this component is rather collisionally ionized than photoionized.
From C IV ) km s-1 follows a maximum temperature of 22 000 K. Since at such low temperatures no C IV will be formed by collisional ionization (Sutherland & Dopita 1993) we will examine photoionization models for all components of the LLS.
Already from inspection by eye we notice different ionization levels in the components. Line profiles show strong absorption by singly and doubly ionized species but very weak absorption of highly ionized species for component 3. For components 9, 10 and 11 one can see that C II and Si II absorption is much weaker compared to other components, but C IV and Si IV are still strong. Component 6 shows in comparison to components 5, 7 and 8 stronger absorption by singly ionized species. In Table 3 we list the observed column density ratios like Si IV /C IV , Si III /Si IV etc. for the 11 individual components of the LLS at .
In the immediate neighbourhood of the LLS at we find further complex MLSs at and 1.942, i.e. with velocity differences of 500 and 5000 km s-1, respectively. For comparison with the LLS at we performed Voigt profile fits to all heavy element absorption lines detected longward of Ly in emission. We identified in total 5 MLSs with redshifts between and 2.05, 3 of which split into subcomponents. Results for the column densities and column density ratios are given in Table 4. We find that the ratio Si IV /C IV shows strong variations not only for the components of the LLS (see Table 3) but also for all these MLSs with absorption lines located longward of Ly emission. Such variations are also detected for weaker MLSs at redshifts up to 3.5 (see Boksenberg 1997). However, Boksenberg (1997) could not confirm the change in the column density ratio Si IV /C IV for Ly clouds near as found by Songaila & Cowie (1996). A generally low Si IV /C IV ratio of the order of 0.03 as claimed by Songaila & Cowie (1996) for Lyman forest clouds with cannot be confirmed. This may be considered as evidence for additional ionizing mechanisms in MLS compared to Ly clouds.
Table 4. Column densities and column density ratios, both logarithmic, for 5 MLSs with heavy element absorption lines detected longward of Ly emission. 3 of them reveal a splitting in subcomponents. For comparison column density ratios for the LLS at are given in Table 3. f indicates that the value was fixed. A colon indicates an error km s-1.
To estimate the ionization level of the individual components of the LLS at we first simply compared our observed column density ratios like C II /C IV , Si III /Si IV , Si IV / C IV , O I /C II given in Table 3 with the diagrams shown by Bergeron & Stasinska (1986). From their photoionization calculations they noted that the ratio Si III /Si IV is almost independent of the H I column density and is therefore a good indicator of the ionization parameter U. Unfortunately, Si III is heavily saturated and blended for most of the components of the LLS. A reliable ratio Si III /Si IV is obtained only for components 1, 3 and 5. For the highly ionized components 9, 10 and 11 information is available only for the Si IV /C IV and Si II /Si IV ratios. Since the ratio Si IV /C IV also depends on the relative metal abundances we have chosen the ratio Si II /Si IV to constrain U or - when calculating photoionization models (see Sect. 6.1.1) - the hydrogen density.
For component 3 both C II /C IV and Si III /Si IV ratios hint at a cloud of low ionization with U between 2 10-3 and 2 10-4. For component 5 both ratios indicate a cloud of slightly higher ionization than component 3 with U slightly below 2 10-3. The O I /C II ratios do not fit into these models. Possible explanations are either differences in the relative abundances or different gas phases. For components 9, 10 and 11 Si IV /C IV ratios are a little bit higher than in the neighboring components of the same LLS. Si IV /C IV ratios are also much stronger compared to the neighbouring complex absorption line systems at and 1.94 (see Table 4). Either these systems are not exposed to the same ionizing radiation field or they differ in element abundances. Also, the Si II /Si IV ratio of component 9 is much lower than in other LLS components. Values for Si II and Si IV column densities in components 10 and 11 are much more uncertain, but indicate also high ionization.
Components 5, 7 and 8 are of intermediate ionization according to their column density ratios. Less information is available for components 1, 2 and 4. According to the Si II /Si III ratio component 4 is comparable to component 5 and component 2 is rather comparable to component 3. The Si III /Si IV ratio of component 1 yields a low ionization with U slightly below 2 10-3.
To perform model calculations we need to know the H I column densities of the individual components. Since O I can arise in a different gas phase, i.e. a mainly neutral phase, and the observed O I profile does not follow the profiles of the singly ionized elements we chose Si II as a tracer of the H I column density. From the Lyman edge in HST data we find a total H I column density of log N(H I ) 17.46 cm-2. From the apparent optical depth method (see Savage & Sembach 1991) we find a total Si II column density log N(Si II ) = 13.69 cm-2. H I column densities for the individual components are derived according to the ratios of the observed Si II column densities in the individual components to the total Si II column density. The total column density of the 11 components is then log N(H I ) 17.40 cm-2. The remaining hydrogen column density of 3.6 1016 cm-2 might be related to the O I absorption. Results for log N(H I ) are given in Table 2.
© European Southern Observatory (ESO) 1999
Online publication: February 22, 1999