Astron. Astrophys. 345, 73-80 (1999)

## 5. Physical conditions in the zabs = 2.198 system

In this section we study the physical conditions in the zabs = 2.198 system in greater detail. The column density per unit velocity interval at any velocity, v, with respect to the centroid of the line is given by,

and from Eq. (1) the optical depth is given by,

Assuming equal covering factor for the two absorption lines of a doublet, , we obtain for each doublet using Eq. (2). We then derive the column density per unit velocity interval from Eqs. (3) and (4). The total column density, N, is obtained by integrating over the velocity interval covered by the absorption line profile. For various species we give in Table 2 the absorption parameters (N and b) obtained from Voigt profile fitting (Columns 3 and 4 respectively) and the column density obtained by integration of the column density per unit velocity interval over the velocity profile (Column 7). The velocity range (Column 5), and the mean covering factor estimated over this range (Column 6) are also given. In the case of single lines we use a conservative lower limit of 0.7 for the covering factor. Note that the column density obtained with the Voigt profile fits are lower limits as corrections are not incorporated to take into account partial coverage.

Table 2. Parameters for the associated system at =2.198
Notes:
a Same as for Ne VIII

Ly from this system is weak and we could not derive any bound on the covering factor from the Ly line only. C III 977 is not detected and we derive a two sigma upper limit from the continuum rms at the expected position of the line. Voigt profile fits to the C IV lines are taken from Savaglio (1998). As the estimated covering factor is low, the column density derived by fitting the profile is less by a factor of 5 compared to the estimated column density using the column density per unit velocity interval. N III and Si III are not detected whereas a line is present at the expected position of N IV 765. This line clearly shows two components and the blue component is clearly absent in the N V profile. The further absence of N III suggests that this component is not real. Thus we have fitted the N IV blue component as an intervening Ly line and used only the column density obtained from the red component in our analysis. Although the resolution of the spectrum is not very high ( 50 km s-1), we have fitted the O VI lines using three components. There is a line at the expected position of O IV 787. However this line could possibly be Ly from the system at zabs = 2.5907. By carefully fitting the Ly, Ly and Ly lines in this system, we removed the contribution of the Ly and fit the residual as O IV 787. As discussed before, the estimate of column densities can be somewhat uncertain due to limited S/N ratio (especially for the Ne VIII lines) and possible contamination by intervening lines.

The derived column densities for H I , C IV , N V and O VI are characteristic of associated systems (Hamann 1997). We run photo-ionization models using Cloudy (Ferland 1996) to study the ionization structure of a plane parallel cloud with neutral hydrogen column density 1014 cm-2, solar metallicities, and illuminated by ionizing radiation fields with different spectra. Although a one-zone model is questionable in such medium, we believe that this is a reasonable representation of the absorbing cloud producing at least C IV , N V and H I (see below for O VI and Ne VIII ) because the kinematics of the lines are very similar (see Figs. 1 and 5) and the partial coverages discussed in the previous sections indicate small sizes for the cloud.

Results for a typical AGN spectrum given by Mathews & Ferland (1987) and two power-law spectra are given in Fig. 6. In the framework of these models, it is apparent that the column densities are reproduced easily. The  = -1 power-law ionizing spectrum is favored as it minimizes the N(C IV )/N(O VI ) ratio. Although metallicities cannot be much less than solar (because of C IV and N V ), it would be difficult to argue for metallicities much larger than solar (especially for oxygen). However, contrary to what is observed, the predicted C IV column density is always larger than that of N V . This suggests that nitrogen is over-abundant compared to carbon. It is interesting to note that the N V lines of the zabs = 2.207 system is also stronger than the lines of C IV suggesting a similar abundance pattern. Indeed, using N V /C IV emission line ratios, Hamann & Ferland (1992) have shown that nitrogen is over-abundant by a factor of in high-redshift QSOs (). They suggested rapid star-formation models to boost the nitrogen abundance through enhanced secondary production in massive stars. Korista et al. (1996) have also noticed overabundance of nitrogen with respect to carbon and oxygen in the well-studied BAL system in Q 0226-1024. They could obtain a much better fit of the lines after taking into account the abundance pattern due to rapid star-formation. The overabundance of nitrogen with respect to oxygen and/or carbon, does not seem to be seen in every associated system however. Indeed, we have good data for the associated systems in Q 0207-003 and Q 0138-381 showing partial coverage. Although solar metallicities are needed to explain the line ratios, there is no indication of enhanced nitrogen abundance.

 Fig. 6. The logarithm of column densities for species indicated next to each curve is given versus the logarithm of the ionization parameter for ionizing spectra as given by Mathews & Ferland (1987; top panel); or power-laws with index  = -1.5 (middle panel) and -1 (bottom panel).

From the observed N V to N IV column density ratio, it can be seen that log  -0.5 (the ionization parameter U is the ratio of the ionizing photons density to the total hydrogen density). However such a value for the ionization parameter implies that the gas producing N V and N IV can not account for the observed value of the Ne VIII column density without an unrealistically large neon abundance. Another and more likely explanation is that there are two distinct regions with different ionization parameters and that Ne VIII predominantly originates from the more highly ionized region. This is supported by the fact that the Ne VIII absorption is spread over a larger velocity range than the N V and O VI absorptions. We can check that, even in that case, C III 977 would not be detectable. Indeed, the expected C III column density is of the order of 1013 cm-2 (see Fig. 6), thus (C III 977)  0.2 Å which is about twice below the detection limit at   3100 Å.

Since the low-ionization region by itself can account for the observed H I and O VI column densities, the high-ionization region should have N(Ne VIII ) much larger than N(O VI ) which means log U larger than 0.5 and most probably 1.

If we suppose that this component is similar to warm absorbers detected by O VII and O VIII edges in the X-rays, then the condition that the optical depth of these edges are larger than 0.1 implies log N(O VII 17.55 and log N(O VIII 18.00. Photoionization models with solar abundances, fail to produce such high O VII and O VIII column densities for H I column densities in the range 1013 to 1014 cm-2. Indeed, the ratios NeVIII/Ne, OVII/O and OVIII/O are all maximized over the range of ionization parameters U  10-100 (see Hamann et al. 1995) where log HI/H  -6. For log N(H I ) = 14, this corresponds to log N(H)  20 and, assuming solar abundances, implies that log N(O VII ) and log N(O VIII ) are less than 17. Thus, it is most likely that in QSO J 2233-606, the region producing the Ne VIII absorption can not be a warm absorber. It is thus of first importance to study the intrinsic spectral energy distribution of J 2233-606 especially in the X-rays.

© European Southern Observatory (ESO) 1999

Online publication: April 12, 1999