Classification criteria

The emission-line galaxies are classified using the line ratio diagnostic diagrams proposed by Baldwin et al. ([1981], hereafter BPT81) and Veilleux & Osterbrock ([1987], hereafter VO87). These diagrams allow to distinguish between narrow-line AGN ionization by a non-thermal power law continuum and HII region-like galaxies with ionization of gas by a large number of hot OB-stars.

According to VO87 a confident separation of the two ionization mechanisms and a classification of the galaxy can be obtained from its location on the [O III] $\lambda\,5007$ Å/H ${\beta}$ versus [N II] $\lambda\,6583$ Å/H $\alpha$ diagram. This diagram has the advantage of being reddening-insensitive and is used to classify the ELGs from the Calar Alto observations which cover a wide spectral region from 3700 Å to 8100 Å.

The observations of the 12 ELGs with the 6m telescope do not cover the red region. For these objects we use the [O III] $\lambda\,5007$ Å/H ${\beta}$ versus [O II] $\lambda\,3727$ Å/[O III] $\lambda\,5007$ Å diagram, which is also a good constraint on the ionization mechanisms, but it is reddening-sensitive, so that extinction corrections would be desirable. However this correction is unreliable for the 6m telescope observations due to poor S/N ratios near the H $\gamma$ emission line. Nevertheless for high-excitation BCGs the average reddening is not large (Izotov et al. [1993a], [1994], [1997b]), and therefore, even without extinction correction they occupy nearly the correct positions in this diagnostic diagram. For lower-excitation galaxies, however, the reddening is getting larger in average, and thus the possible uncertainty of their classification can increase, especially if an object falls into the transition area.

**Figure:** a,b Classification diagrams for ELGs. The dashed line separates regions of HII-type and AGN spectra, according to VO87 and BPT81. The dotted lines separate the region in which high T_eff star excitation takes place from the region of low ionization populated by DANS and SBN (after S89a)
$\begin{figure} \centerline{ \psfig{figure=FIGURE4.ps,width=9.7cm}} \vspace{-2.0cm}\end{figure}$

In Fig. 4 emission-line galaxies from the HSS sample are plotted in comparison with some BCGs from the SBS (Izotov et al. [1994], [1997b]; Thuan et al. [1995]) and ELGs from the Case survey (Weistrop & Downes [1988], [1991]; Augarde et al. [1987]; Salzer et al. [1995]; Ugryumov et al. [1998]). The branch of HII-galaxies is separated from AGN by a short-dashed line running in both figures from the left-top corners populated by objects showing strong emission lines to the right-bottom corners populated by objects with a low level of gas ionization.

It is clear from the data in Fig. 4 that the major part of the observed HSS emission-line galaxies belongs to HII-galaxies and only a few objects can be classified as probable AGN.

>From the two objects in the AGN region of Fig. 4a only one -- HS 1609+4902 -- can be definitely classified in both diagrams as a Sy2 galaxy. The second probable AGN-type object is HS 1304+4710. In Fig. 4a it is located not too far from the border line separating HII-type objects from AGNs. We tentatively classify it as a possible LINER (low-ionization nuclear emission-line region galaxies; Heckman [1980]). In Fig. 4b this object falls however within the HII-type region. A possible reason for this is the reddenning due to the dust extinction.

Among the second priority candidates we find a number of low-excitation massive galaxies with nondetected (in low S/N spectra) H ${\beta}$ and [OIII] $\lambda\,5007$ Å emission lines. They are tentatively classified as possible LINERs (marked as LINER? in Table 1) according to the flux ratios of [NII] $\lambda\,6583$ Å and H $\alpha$ (Heckman [1980]). Although these galaxies are not shown in Fig. 4, they, likely, populate the extreme right-bottom corner in Fig. 4a.

We use some additional criteria for the classification of ELGs in the cases where the S/N ratio is insufficient to use the diagnostic diagrams. As it was emphasized by Salzer ([1989a], hereafter S89a) there is a tight correlation between the ELG type and their mean global parameters such as luminosity and size. In particular, the ELGs with the highest excitation are low-luminosity, compact objects, while AGN-type objects are almost always high-luminosity, large galaxies.

Two more objects in Fig. 4b, HS 1038+4616 and HS 1443+5018, fall into AGN region. We suspect that it is due to poor S/N ratio for H ${\beta}$ . Therefore, their large [O III] $\lambda\,5007$ Å/H ${\beta}$ ratios are unreliable, and, probably these galaxies are low excitation BCGs. Indeed, their low luminosities ( $M_B = -17\fm2$ and $-17\fm1$ , respectively) are more compatible with a HII-type nature.

HII-galaxies can be splitted further into several classes depending on physical parameters such as sizes, absolute magnitudes, colors, metal abundance, equivalent widths (EW) of the [O III] $\lambda$ 5007 Å line and morphology. S89a have proposed the following sequence of HII-galaxy types (starting with the least luminous and most compact objects and ending with the most luminous, quite large spirals): ``Sargent-Searle (SS) objects'' (or ``Blue Compact Dwarfs (BCD)''), ``dwarf HII hotspot (DHIIH) galaxies'', ``HII hotspot (HIIH) galaxies'', ``Dwarf Amorphous Nucleus Starburst (DANS) galaxies'' and ``Starburst Nucleus (SBN) galaxies''. SBN-galaxies have seemingly also some input from shock excitation. There is also a significant correlation of the luminosity with other parameters, such as metallicity, EW of the [O III] $\lambda$ 5007 Å line and color.

The insufficient quality of our spectroscopy does not allow to perform such detailed classification. Therefore, in our case SS, DHIIH and HIIH galaxies are taken together as one class of blue compact/ HII galaxies (BCGs), which are objects with strong and moderate emission lines, caused by the large number of young OB-stars in current burst of star formation.

Following S89a, we separate the BCGs and low-excitation ELGs regions in the diagnostic diagrams with dotted lines (Fig. 4). This border is derived from the analysis and generalization of good-quality ELG observational data from the UM survey. To classify objects near the border we use additional parameters, such as EW([O III] $\lambda$ 5007 Å) and M_B, according to their characteristic values for corresponding ELGs from S89a.