These were made in the traditional way, paying attention to both wavelengths and line intensities within the multiplets.
The identifications were made with the help of the table of Moore (1959); for Fe II we also used Johansson's (1978) compìlation and for [NiII] Nussbaumer & Storey (1982). In addition we have used the Meinel et al. (1969) catalogue, for lines which we could not identify (see notes to Table 2).
On our spectra are present 217 features, of which 40% are pure absorption lines whereas the majority is constituted by emission lines. We were unable to identify 7% of the lines, which is about normal in identification work. Furthermore 33% of all lines are produced by Fe II and [Fe II] The complete list of identifications is given in Table 2. Parts of the spectrum are reproduced in Fig. 1 (click here). The Paschen region, at different epochs is illustrated in Fig. 3 (click here) and the O I region in Fig. 4 (click here).
From Table 2 and Figs. 1 (click here) and 3 (click here) it can be seen
that up to 4200 A the spectrum is mostly
formed by absorption lines, whereas beyond 4200 the emission lines start to predominate. After
6300 the absorption lines do constitute an exception. We shall thus fix, somewhat arbitrarily,
as the point where the shell starts to predominate over the underlying spectrum.
In what follows, we will describe first the identifications, element by element and then the structures present in the lines of different elements Since, as can be seen from Table 1 (click here), the material was taken at different dates, the discussion of the structures is directly linked with that of spectrum variability.
Figure 1: Reproduction of CCD spectra of HD 50138. Abscissae: wavelengths in A. Ordinates:
intensities. The continuum is set to unity. Important lines are identified.
The spectra correspond
to 
3925-4300 (from 25.10.91),
4300-4575 (from 23.10.91), 4725-5100 (from 25.10.91),
6500-6600 (from 23.2.92) and
9850-10225 (from 14.11.89)
Figure 2: Variations in the profil of O I 7774. From top to bottom, the spectra were taken on
07.01.90-29.12.90-29.10.91
-16.3.92 and 07.04.93. Two iron lines are also visible: l 7712 Fe
II and 
7732 Fe II+[Fe II]
Figure 3: Variations in the region P12 to P20. From top to bottom, the
spectra were taken on 15.11.89-27.12.90-21.2.92 and 31.03.93
Figure 4: The equivalent width of the Paschen lines at different dates. Abscissae: principal quantum
number n. Ordinates: equivalent widths in A