Astron. Astrophys. 346, 769-777 (1999)

4. Discussion

Degree of ionization. As discussed in Sect. 3.1, bar H ii region spectra do not exhibit any obvious signs of high-velocity shocks or hard UV radiation. However, there is marginal evidence from Figs. 2 and 3 that the ionization might be different for some bar regions. The degree of ionization at a specific position in a nebula can be accessed through the ionization parameter:

where is the number of ionizing photons per unit time by the central source, r is the position in the nebula, and c the speed of light (Osterbrock 1989). The most obvious evidence that bar regions are different from disc regions is seen in the [O i] 6300 line. As described in Evans & Dopita (1985), the [O i] line is emitted in the transition zone of an H ii region which contains a significant fraction of neutral hydrogen. The line is then stronger when U is lower (or the ionizing stellar temperature is lower). Since we have detected the [O i] line in a much larger fraction of bar regions than in disc regions, this suggests that the ionization parameter could be lower in the former population. Fig. 9 shows the correlation between [O i] 6300/[O iii] 5007 and [O ii] 3727/[O iii] 5007 which is particularly dependent on U (Evans & Dopita 1985). A correction for the interstellar extinction has been applied to these line ratios. The bulk of bar H ii regions is located at . For the disc regions, the scatter is quite large but on average , larger than the value for bar regions. Unfortunately, our sample is not large enough to firmly confirm that U is indeed different for both populations of H ii regions.

Fig. 9. Diagnostic diagram showing the relation between the [O i] 6300/[O iii] 5007 and [O ii] 3727/[O iii] 5007 nebular line ratios. Both ratios have been corrected for the extinction. The curves represent models from Evans & Dopita (1985). The horizontal dashed lines show three values for U, 0.03, 0.003, and 0.0003. The vertical lines indicate the stellar temperatures used in the models, 50 000 K, 40 000 K and 37 000 K.

If a difference in the ionization parameter is really present, this could be due to many factors: differences in the initial mass function, age, richness of the OB associations, or spatial distribution of the ionized material (Evans & Dopita 1985). As recently shown by Rozas et al. (1999), the luminosity function (LF) of the bar regions is much less regular than the LF of the disc regions in the strongly barred spiral NGC 7479. Their result, combined with our study on the nebular excitation, suggest strongly that the properties of the OB associations formed in bars differ from the normal associations of the disc. More work comparing LFs and nebular properties of a larger sample of bar and disc regions would allow us to investigate the origin of this difference. In the final paper in this series (Friedli & Martin, in preparation), we will also examine the properties of the clusters formed in diverse bar environments with high-spatial numerical simulations.

Mixing and element production. The estimated timescale given by numerical simulations for which the star formation activity phase lasts in bars is yr (Martin & Friedli 1997). How does this compare with mixing timescale?

It is possible to roughly quantitatively evaluate the timescale of mixing of the ISM due to radial flows. Roy & Kunth (1995) have discussed the diverse mixing mechanisms of the oxygen abundance in the ISM in galaxy discs. Assuming a pure radial mixing due to gas flows funnelled in the bar, the upper limit for the time for gas to diffuse a length scale, , in the radial direction is:

where v is the radial flow velocity, and l is the mean free path for molecular clouds. In a typical bar, kpc (see Table 1) and km s^-1. The value of the mean free path for the gas clouds is not a well-defined quantity in bars. In galaxy discs, pc (Roberts & Hausman 1984; Roy & Kunth 1995). If we assume pc for bars, we find yr. However, since radial flows in bars are not stationary, the real mixing timescale could be even of the order of , i.e. yr. Putting all this together yields the following reasonable interval for the mixing timescale: yr. Thus, is shorter than meaning that the abundance content in H ii regions formed during this phase must be homogenized.

It is also instructive to make rough (i.e. close-box) estimates of the global abundance increase during as well as of the abundance fluctuations in H ii regions which should result from their age spread. If is the initial mean gaseous abundance in the bar region, the global star formation efficiency, and the net yield for the species considered, then the final mean abundance is given by:

Interestingly enough, does not depend on the initial gas mass fraction. For instance, for the oxygen with an yield , an initial solar abundance , and a typical SF efficiency , then . The global increase of oxygen abundance is thus only about 0.1 dex. Eq. 3 can in fact also be applied to each individual H ii regions with exactly the same numbers; the fluctuations in the oxygen abundance are thus expected to be of the order 0.1 dex, which is indeed what is observed (Sect. 3.5). However, we do not observe any clear trend between the age and metallicity for bar H ii regions. In dwarf galaxies, the metal enrichment of H ii is not observed and metals are probably locked in the hot phase. (see e.g. Tenorio-Tagle 1996; Kobulnicky 1998). The situation could be similar for star forming regions in bars.

© European Southern Observatory (ESO) 1999

Online publication: June 17, 1999
helpdesk.link@springer.de