3. RESULTS AND DISCUSSION

From the line intensities of the sulphur lines presented in Table 1 we have calculated the sulphur abundance parameter S₂₃ defined as

This parameter is analogous to the R₂₃ parameter defined for the optical oxygen lines and hereafter we will refer to the oxygen and sulphur abundance parameters as O₂₃ and S₂₃ respectively. Both parameters are listed in Table 2.

The relation between O₂₃ and the oxygen abundance for the objects in Table 2 can be seen in the upper panel of Fig. 1 (solid dots) together with similar data for HII galaxies compiled by Díaz (1999) (open circles). The figure illustrates the problems mentioned in the introduction, the most important one being the two-valued nature of the relation which makes an accurate metallicity calibration virtually impossible, more so for objects with logO₂₃ 0.8. Also notice the position of the nucleus of NGC 3310 which is probably due to its higher than usual density (n_H = 8000 cm^-3; Pastoriza et al. 1993). On the contrary, and as expected, the relation between S₂₃ and oxygen abundance for the objects of Table 2, shown in the lower panel of the figure, remains single valued up to a metallicity close to solar. In both figures the data corresponding to different observations of IZw18, NGC 5471 in M101, and S5 in M101 have been joined together to show the importance of internal errors. The solar metallicity regions have also been labelled.

Figure 1. Oxygen abundance parameter (upper panel) and sulphur abundance parameter (lower panel) vs oxygen abundance for the objects in Table 2 and HII galaxies in Díaz (1999).

The dependence of O₂₃ on the degree of ionization of the nebula is partially responsible for the large scatter found in the O₂₃ vs metallicity relation. Figure 2 (upper panel) shows the relation between logO₂₃ and log([OII]/[OIII]), which can be taken as a good ionization parameter indicator for ionizing temperatures larger than about 35000 K (see Díaz 1999), for the objects in Table 1. A positive correlation between ionization parameter and O₂₃ is evident for objects with O₂₃ between 0.2 and 1 (higher excitation objects are at the left in the plot). On the contrary, no relation between logS₂₃ and log([OII]/[OIII]) is readily apparent in the lower panel of the figure. The peculiar position of IZw18 in the plots is due to its very low metallicity.

Figure 2. Relation between logO23 (upper panel) and logS23 (lower panel with log([OII]/[OIII]), taken as ionization parameter indicator, for the objects in Table 1.

Based on the above considerations, we have attempted a calibration of oxygen abundance through the sulphur abundance parameter S₂₃, using the data of Table 2 with the exclusion of the three objects for which no direct determinations of the electron temperature exist: S5 in M101, CC93 in M33 and the nucleus of NGC 3310. A linear fit to the data, taking into account the observational errors, give:

with a correlation coefficient of 0.88 and a value of = 0.15.

The data points corresponding to IZw18 deviate slightly from linearity. This is due to the high excitation of this object which increases the fraction of S⁺⁺ converted to S³⁺ thus decreasing S₂₃. A quadratical fit:

provides a better fit to these extremely low metallicity data. The two fits are shown in the upper panel of Figure 3 by a solid and a dashed line respectively. The corresponding residuals are shown in the lower panel as a function of the oxygen abundance. Solid dots correspond to the linear fit while open circles correspond to the quadratic one.

Figure 3. Empirical calibration of oxygen abundance through the sulphur abundance parameter S23. The ressiduals of the fits are shown in the lower panel of the figure (see text for details).

Figure 4 shows the relation between S₂₃ and O₂₃. The largest value of S₂₃ corresponds to one of the observations of region S5 in M101 (S₂₃ = 0.48), and the solar metallicity objects have S₂₃ between 0.28 and 0.48. On the other hand, the object with the lowest metallicity known, IZw18, has a value of logO₂₃ = 0.47. These two facts taken together imply that objects with logO₂₃ 0.47 and -0.5 < logS₂₃ 0.28 will necessarily have oversolar abundances. These objects are 48 out of the 196 listed in Table 1 and therefore constitute about a quarter of the total sample. Most of them are circumnuclear star forming regions, HII regions in inner galactic discs and HII regions in Virgo cluster galaxies.

Figure 4. Relation between O23 and S23 for the objects in Table 1.

For logO₂₃ < 0.45 it can be seen that for the lowest metallicity objects there is a positive correlation between O₂₃ and S₂₃; NGC 5471, with logO₂₃ 1.00 and logS₂₃ 0.10, would lie at the end of this correlation. Then, there is a trend of increasing S₂₃ with decreasing O₂₃ which corresponds to the "upper metallicity branch" of the O₂₃ abundance calibration. For values of logO₂₃ between 0.45 and 0.00, the relation between the two abundance parameters is rather flat. Finally, for values of logO₂₃ < 0.0 a trend of decreasing S₂₃ with decreasing O₂₃ is apparent, indicating that, for the metallicities involved, the expected reversal of the S₂₃ vs metallicity relation has already taken place.

Figure 5. Variation of log(S23/O23) with galactic radius for M101, which actally mimics a logarithmic abundance gradient through the disc.

The behaviour of the O₂₃ and S₂₃ parameters is related with the different energies involved in the two sets of transitions and therefore a combination of both should, in principle, be a better indicator of metallicity than either of them alone. Figure 5 shows a plot of log(S₂₃/O₂₃) for the HII regions in M101 (Kennicutt & Garnett 1996). A single line logarithmic gradient is found. This might actually constitute a purely observational way to quantify galactic disc abundance gradients without the need to rely on theoretical photoionization models.