SFDGs are characterized by low chemical abundances. For a review of metal poor SFDGs, see Kunth and Östlin (2000). In H II regions, O/H varies from ~ 3% solar to about half solar (using the solar oxygen abundance from Asplund 2005) for the most massive SFDGs (Garnett 2002). As shown by Skillman et al. (1989), there is a correlation between luminosity and metallicity among local SFDGs. As more data are added though, the scatter increases, in particular if the luminosities are based on emission in the blue part of the spectrum, dominated by young stars (e.g., Guseva et al. 2009). This is partly due to inhomogeneities in the data but cannot be the full explanation as becomes evident if we compare the location of LSBGs in the L-Z diagram with that of BCGs (Fig 2a). Again the data sample is selected in a subjective way and is limited in size since we included only galaxies with both H I masses and O abundances. We see a few striking details. First of all, there is a clear separation between the LSBG sample and the BCG/dI sample. They seem to form two parallel sequences, shifted in luminosity by almost 3 magnitudes. We see this division also in Fig. 2b. A least square fit through the whole sample makes no sense, it is only confusing since apparently we are dealing with two different categories of galaxies. In Fig 2c, showing O/H vs. MHI / LB, the two types merge. If the B band mainly reflects the star formation rate in the galaxies, then there is no difference between BCGs and LSBGs as concerns the star formation efficiency per gas mass. The major reason for the difference shown in Fig a and b is probably the scale size and surface density of the neutral hydrogen gas. At the same metallicity the LSBGs are larger, brighter and more extended, making the star formation proceed at a slow speed compared to normal galaxies. Perhaps one could say that the difference in distribution between LSBGs and BCGs+dIs reflects a division line between dwarfs and normal galaxies. Only disk galaxies can produce metal rich galaxies, the others remain below a metallicity ceiling of 12 + log(O/H) = 8.3. Tidal dwarfs (Duc and Mirabel 1999) form an exception to the rule and are found at O abundances ~ 8.5. Among the brighter galaxies we find a small number of BCGs with higher metallicities, mixed with the LSBGs. Possibly we find progenitors to more luminous (MB < ~ -16) BCGs among the LSBGs but hardly among the fainter BCGs since either the LSBG has to get rid of more than 50% of the gas or increase the metallicity with a factor of a few.
Figure 2. Oxygen abundances of Irr, BCG and LSB galaxies as function of absolute B magnitude (a), H I mass (b) and (H I mass)/(B luminosity) (c). The samples in the three diagrams are identical and are obtained from Masegosa et al. (1994), Roennback and Bergvall (1995), van Zee et al. (1997), Lisenfeld and Ferrara (1998), Izotov and Thuan (1999), Burkholder et al. (2001), Hirashita et al. (2002), Bergvall and Östlin (2002), Salzer et al. (2002), Pustilnik et al. (2002), Huchtmeier et al. (2005), Geha et al. (2006), Begum et al. (2008) and additional scattered information in the literature.
Before one draws too strong conclusions based on the L-Z or mass-metallicity diagrams one should consider the influence of mass exchange between galaxy and the environment. Kunth et al. (1994) realized the importance of determining the abundances in the neutral H I envelope to better understand the chemical evolution. Kunth and Lebouteiller (2010) find a correlation between metallicities in the neutral gas and the H II gas, indicating that mixing between the newly produced metals and the H I envelope is taking place. There is however a systematic difference in metallicity between the two states. In the L-Z diagram the most violent starburst dwarfs are found in the lower rightmost wing of the BCG distribution. One of the reasons may be that the stellar population in these galaxies manages to ionize a large part of the gas in the central region and consequently the measured metallicity will be lower than for a milder starburst. The most metal poor galaxies, the low luminosity dwarfs I Zw 18 and SBS 0335-052, have metallicities of a few % solar.This is close to the metallicity derived for the neutral gas in these and other BCGs using UV spectroscopy (Izotov and Thuan 1999, Izotov and Thuan 1999, Lecavelier des Etangs et al. 2004, Thuan et al. 2002, Heckman et al. 2001, Lee and Skillman 2004). It appears as if there is a minimum plateau defined by the mean metallicity of the intergalactic medium, i.e. log(O/H) ~ -5 (Thuan et al. 2002, Thuan et al. 2005).