7.1 The Metallicity-Luminosity Relation, and Other Empirical Relations
As we saw in sections 3.1 and 3.2, the metallicity of dEs and dIs correlate positively with their luminosity (Aaronson 1986, Lequeux et al. 1979, Kinman and Davidson 1981, Skillman et al. 1989). Whether they follow the same relation is a subject of debate due to the uncertain scaling between [O/H] and [Fe/H], (see Sect. 4.2.1; and Richer and McCall 1995). It would be instructive to examine the location of BCGs and LSBGs in the metallicity-luminosity diagram. The luminosity is usually represented by the B-band absolute magnitude which might be significantly affected by ongoing star formation, especially in BCGs, and thus is not necessarily a good estimator of the stellar mass. Indeed, NIR magnitudes would be a better choice, but data are rather scarce. Earlier work indicates that BCGs follow such a metallicity-luminosity relation, but with considerable scatter (Kunth 1986, Campos-Aguilar et al. 1993).
In Fig. 10, we show the oxygen abundance vs. luminosity diagram for dIs (crosses), BCGs (filled symbols) and LSBGs (open symbols), based on data collected from the literature, with available abundances and integrated B magnitudes. In addition we show the location of candidate tidal dwarfs and dEs with measured nebular oxygen abundances. For the BCGs we have divided the sample into morphological subtypes according to the classification schemes by Telles et al. (1997) and Loose and Thuan (1986a). Galaxies with regular outer isophotes, classified as Type II according to Telles et al. (1997) or as iE or nE according to Loose and Thuan (1986a), will be referred to as Type II, while ``irregular'' ones will be referred to as Type I following Telles et al. (1997). If there is no classification in the literature, but published images are available, we classified the galaxies according to this scheme. The solid line shows the MB - Z relation for dIs by Richer and McCall (1995), while the short and long dashed lines shows the relation for dEs (see caption).
At first sight many BCGs do not appear extreme when compared to the normal dIs. Indeed some BCGs are more metal rich than dIs at a given luminosity, while the opposite would be expected if BCGs were bursting dIs. These BCGs may be in a post burst stage and the fresh metals may have become ``visible'' already. Secondly, some ``extreme BCGs'' appear much more metal-poor at a given luminosity. These extreme BCGs (XBCGs) are 3 magnitudes brighter or more at a given metallicity, or equivalently 0.5 dex less metal rich at a given luminosity as compared to the dI relation. There is a tendency for the Marlowe et al. (1999) blue amorphous galaxies to lie above the dI relation. Galaxies of Type II follow the dI relation in a broad sequence, while Type I have a tendency to fall below the dI relation. The same phenomenon is seen for four galaxies from the sample by Östlin et al. (1999a), of which three have irregular morphology (Type I). The most metal-poor galaxies also have irregular morphologies and fall far below the dI relation. Thus the BCGs span a factor of 10 in metallicity at a given luminosity. The intriguing XBCGs include IZw18 and SBS 0335-052, ESO 338-IG04 (= Tololo 1924-516), ESO 400-G43, Haro 11 (= ESO 350-IG38), ESO 480-IG12 (Östlin et al. 1999a, b; Bergvall and Östlin 1999); UM 133, UM 448, C 0840+1201 (Telles and Terlevich 1997), UM 420, UM 469 and UM 382 (Masegosa et al. 1994, and Salzer et al. 1989a). LSBGs occupy locations in the range from dIs to XBCGs (see also Bergvall and Rönnback 1994, Bergvall et al. 1998). While enriched galactic winds could possibly explain the extreme metal deficiency of the least massive dwarfs, the existence of XBCGs in general cannot be understood in this way. While suggestive of important differences between different samples and types, these trends should be regarded as preliminary and should be put on more solid ground.
Figure 10. The luminosity versus metallicity diagram for dwarf galaxies. The crosses represent dIs taken from Richer and McCall (1995) and Skillman et al. (1989). Filled symbols represent galaxies classified as BCGs or H II-galaxies: The small filled diamonds are ``regular'' galaxies which can be classified as Type II or iE/nE according to Telles et al. (1997) and Loose and Thuan (1986a), respectively, while filled circles are galaxies that can be classified as Type I (see text for description of types). Filled triangles are BCGs for which no classification or images were available. The three most metal-poor galaxies are labelled and shown as filled hexagons; their morphology are indicative of Type I. The filled stars are luminous BCGs from Östlin et al. (1999a, b), three of which are of Type I. The asterisks show the location of ``blue amorphous galaxies'' (Marlowe et al. 1999, except for II Zw40 which is the filled circle falling on the short dashed line). LSBGs are shown as open squares (blue LSBGs, Bergvall et al. 1999, Rönnback and Bergvall 1995) and open pentagons (McGaugh 1994, McGaugh and Bothun 1994). The open star is H I 1225+01, the H I-cloud in Virgo (Salzer et al. 1991). The boxes with plusses inside are quiescent (dI/LSBG) dwarfs from van Zee et al. (1997b, c). Candidate tidal dwarfs (Mirabel et al. 1992, Duc and Mirabel 1994, 1998) are shown as ``T'', and dEs are shown as ``dE'' (data from Mateo 1998). The solid line shows the MB - O/H relation for dIs from Richer and McCall (1995), while the dotted line shows the same relation offset by 3.5 magnitudes, indicating the location of XBCGs. The short dashed line shows the MV - Z relation for dE/dSph from Caldwell et al. (1998) assuming (B - V) = 0.75 and [O/H] = [Fe/H], while the long dashed line shows the same relation assuming [O/H] = [Fe/H] + 0.5. When necessary, we have rescaled absolute magnitudes to H0 = 75 km/s/Mpc. Metallicities of BCGs from: Izotov and Thuan (1999), Lipovetsky et al. (1999), Bergvall and Östlin (1999), Telles and Terlevich (1997), van Zee et al. (1998), Kunth and Joubert (1985), Alloin et al. (1978), Thuan et al. (1996), Masegosa et al. (1994). Absolute B-magnitudes for BCGs from: Telles and Terlevich (1997, assuming B - V = 0.5), Papaderos et al. (1996a, 1998), Östlin et al. (1999a), Thuan et al. (1996), Mazzarella and Boroson (1993), Salzer et al. (1989a), Schulte-Ladbeck et al. (1998).
Another important relation, between H I mass and blue luminosity, has been addressed by several authors. A positive correlation was found by Chamaraux (1977) for a sample of Zwicky BCGs. It has been found (e.g. Staveley-Smith et al. 1992) that for gas rich dwarfs, the hydrogen mass to blue luminosity decreases with increasing luminosity, i.e. low luminosity galaxies are more H I rich, and have apparently converted a smaller fraction of their neutral gas content into stars. Moreover, LSBGs have proportionally more H I than BCG in the sense that they have higher MHI / LB values (see e.g. fig 1a in Bergvall et al. 1998). BCGs and dIs seem to have similar MHI / LB ratios. McGaugh and de Blok (1997) showed that, for a sample of disc galaxies (extending from normal spirals to LSB dwarfs), MHI / LB increases systematically with decreasing surface brightness, and is typically MHI / LB = 1 (solar units) for LSBGs. LSBGs have low mass densities and have thus been inefficient in converting H I to stars and metals. Thus, to a first approximation, metallicity anticorrelates with the gas mass fraction (cf. Lequeux et al. 1979, Kinman and Davidson 1981, Pagel 1997) as expected from closed box models, but this cannot be the whole explanation (Matteucci and Chiosi 1983). Dust-to-gas ratios positively correlate with metallicity for dIs, while BCGs appear comparably more dust rich (Lisenfeld and Ferrara 1998), although in many BCGs one can only put upper limits to the dust content. The general relation between surface brightness and luminosity for dwarf and late type galaxies (Ferguson and Binggeli 1994), implies that dwarfs have low mass densities and/or are inefficient star formers. It has been argued that this could be a pure selection effect, since faint low surface brightness would be more difficult to detect (Phillipps et al. 1988).
Terlevich and Melnick (1981) report a positive correlation between the H luminosity, the H line width, and linear size for giant H II regions and H II galaxies. Subsequent work confirmed a dependence also on metallicity (Melnick et al. 1987, se also Campos-Aguilar et al. 1993). This also opened up the interesting possibility to use H II-galaxies as distance indicators (Melnick et al. 1987, 1988; see Sect. 8.6). If, as Melnick et al. (1987) argue, the H line-width is due to virial motions, this relation reflects an underlying dependence on galaxy mass. Whereas this relation cannot hold for dwarfs irregulars with low star formation activity, it is striking that low mass galaxies with strong star formation activity seems to form a well defined sequence. This relation may reflect an intrinsic upper limit, perhaps regulated by feedback, on the possible star formation rate (directly proportional to the H luminosity) as a function of mass or mass density. These galaxies would then represent the most efficient star formers with respect to their mass. Similarly, Meurer et al. (1997) find an upper limit on the bolometric surface brightness in starbursts at high and low redshifts, implying the existence of a physical mechanism limiting the global areal star formation rate in galaxies.