5. Oxygen abundances
Fig. 2 shows the reddening corrected positions of the measured HII regions in the diagram. In order to avoid spuriously large values of, especially, we have omitted all HII regions where the peak value of the measured H line was less than 3 above the noise of the continuum. These regions are denoted in Table 2 by having their identification number (column 2) between brackets. The errorbars are determined by taking into account the uncertainties in the fluxes of the relevant lines using the continuum-offset procedure (see above). Note that the calibration uncertainties of the model (Sect. 4) are much larger than the formal errors in 23 and o32.
All points are consistent with the model computed assuming an upper mass cut-off in the Initial Mass Function of presented here. We therefore find no strong evidence for the existence of super-massive stars in the HII regions.
Fig. 3 shows the reddening corrected [N II ]/[O II ] ratio as a function of 23 for our HII regions, along with those given in McCall (1985). We see that most of our HII regions have , which implies that they are on the low-abundance lower branch of the model grid.
Oxygen abundances and ionization parameters are presented in Table 5, for those spectra where [N II ]/[O II ] could be determined. The errors were determined by projecting the continuum-offset uncertainties in and on the - -grid, and adding these in quadrature to the calibration uncertainties in and at these positions. In most cases the errors are dominated by the calibration uncertainties.
Table 5. Oxygen abundances
Table 6 presents the results for those spectra where [N II ]/[O II ] was not measured, and the abundance determination was ambigious. Both upper and lower-branch values are given.
Table 6. Ambiguous abundances
Fig. 4a shows histograms with the distribution of oxygen abundances. The top panel gives the distribution of abundances of HII regions from our sample, the middle panel shows the distribution of oxygen abundances of HII regions from the sample of McGaugh (1994) (his Table 3). The bottom panel shows the distribution of both samples combined. Both samples have a few galaxies in common, but the number of HII regions that are in both samples is a few at most and any overlap will not influence the eventual results.
The peak in the histograms at = -3.6 in Fig. 4a is at least partly artificial. This is due to the fold at = -3.6 in the model grid of Fig. 2. We have attempted to correct for this in Fig. 4b. Here each of the histogram bins has been replaced by a gaussian with a width (= 0.43 FWHM) equal to the uncertainty of the model grid at that abundance (e.g., the gaussian at = -3.6 has ). The peak value was determined keeping the area under the gaussian equal to that of the corresponding histogram bar. As a result the peak at = -3.6 has been smeared out, and the distribution is almost flat, with maybe a slight peak at . This peak might be the result of our selection method. Low abundance regions tend to be more ionized, and therefore brighter (e.g., Campbell 1988; Dopita & Evans 1986). Regardless of whether this effect is present or not, it is clear that within the abundance range shown by LSB galaxies, there is no preferred value.
With regard to the lowest abundances, Kunth & Sargent (1985) note that values of = -4.3 can already be reached after the first generation of massive stars. This might explain the cut-off in the abundance distribution around = -4.3 in Fig. 4. LSB galaxies are not primordial objects, but some of them appear to be very unevolved. Furthermore, to retain such low abundance values at the present epoch, these galaxies must have been quiescent over their entire life time, and some of them may not evolved significantly since their first epoch of star formation.
© European Southern Observatory (ESO) 1998
Online publication: June 18, 1998