Astron. Astrophys. 357, 37-50 (2000)

3. Absorption line parameters and identifications

Tables 1 and 2 list absorption lines found at the level in HS 1216+5032 A and B, respectively. Lines in the normalized spectra were fitted with Gaussian profiles without constraints in central wavelength, width or amplitude. In the case of obvious blends two or more Gaussian components were fitted simultaneously.

Table 1. Absorption Lines in HS 1216+5032 A.
Notes:
1 Absorption in the ISM.
2 Or SiIV at .
3 Blended with HI at if km s^-1.
4 Or SiII at .
5 Or SiIII at .
6 Blended with HI at .
7 Or SiII at .
8 Associated system. Ly is blended with line A21.
9 Lines mimic CIV doublet but no corresponding HI line is detected.
10 Small contribution from HI at to .
11 Strong blend. The identification is supported by the detection of HI (A4 and A5)

Table 2. Absorption Lines in HS 1216+5032 B.
Notes:
1 Absorption in the ISM.
2 Probably blended with Ly-.
3 Blended with Ly at .
4 Blended with HI BAL.
5 Also SIV BAL because the line is too strong compared with IS FeII .
6 Or HI at if km s^-1.
7 Lines mimic CIV doublet but no corresponding HI line is detected.
8 The fit procedure is not able to resolve the second doublet lines; therefore, the doublet ratios are probably not real.
9 Associated system. Ly falls in the OVI BAL troughs.
10 Blended with . See Sect. 5 for the classification as BAL system.
11 No line at can be identified with Ly.

The fit routine attempts to minimize between model and data, considering flux errors (flat-fielding, background subtraction and photon statistics noise); but the uncertainties introduced by the continuum placement are not included.

An inherent feature of minimization is the non-uniqueness of the solution due to the eventual presence of more than one minimum in the parameter space. In those cases, the fit can lead not only to wrong parameter estimations but also to underestimated parameter errors. To handle with this problem, instead of defining a rejection parameter we simply performed several fit trials taking different wavelength ranges to see how robust a particular solution was. In general, there were no significant changes in the resulting parameters due to a different choice of the fitting interval, thus giving us confidence on the results. However, some fits yield . Assuming that the flux errors are not overestimated, such behaviour can occur by chance, especially when too few pixels are considered in the fit.

The equivalent-width error estimates are generally larger than the propagated error if the flux values are integrated along a line. This is because the former errors consider the correlation between all three free parameters. This is not the case for multicomponent fits, where the errors in equivalent width may be overestimated since the fit algorithm does not consider that the strengths and widths of neighbour lines in blends are anti-correlated.

Concerning the line widths, note that several FWHM values in Tables 1 and 2 are slightly larger than the nominal width of the line spread function (LSF), 2 Å. This is caused by the difficulty in resolving absorption structures separated by less than km s^-1, since single-line fits of closely lying line complexes (blends) will inevitably lead to large widths. On the other hand, in few cases the measured lines are narrower than the LSF. While this might partly be caused by a too low placement of the continuum near emission lines and BAL troughs in the B spectrum, it is certainly not the case for all the narrow lines in the A spectrum. In the latter case fitting weak lines also leads to low FWHM values (with large associated errors), thus reflecting the limitations of the minimization using too few pixels.

BAL profiles in the B spectrum have been fitted with multicomponent Gaussians to better constrain their equivalent widths, but these fits do not necessarily represent the actual nature of the BAL profiles (see Sect. 5). For consistency, Tables 1 and 2 include the Gaussian parameters of BAL profiles.

Absorption lines were identified manually, using vacuum wavelengths and oscillator strengths taken from Verner et al. (1994). The identification of lines is not easy because of the low resolution and the lack of optical data; but it is especially difficult in the B spectrum because narrow intergalactic lines are hidden among the BALs. We looked first for interstellar absorption lines in both spectra. The MgII doublet and the FeII lines at 2586 Å and 2600 Å are detected in both spectra; the FeII lines at 2344 Å and 2382 Å are detected at only in the A spectrum (while in the B spectrum FeII would appear at the position of BAL CIII ). The next step was to search for CIV systems and their associated Ly and prominent metal lines. In addition, two MgII systems at (Sect. 4.2) and 1.14 could be identified in the spectrum of B through the positions of the doublet lines and their relative strengths. Due to their secure identification, CIV and MgII lines with have been retained in Tables 1 and 2, which otherwise contain only lines with . Ly lines with redshifts of were identified through the corresponding Ly and Ly lines, if present. Lines A45, A46, B37, and B38 mimic a CIV doublet at , but since no significant absorption by HI is detected at this redshift, they have been counted as Ly lines. The remaining lines have been considered to be Ly lines and will be discussed in Sect. 6.

© European Southern Observatory (ESO) 2000

Online publication: May 3, 2000
helpdesk.link@springer.de