3. Nebular properties
The basic source of ionization in an H ii region is the UV radiation field from young massive stars. The emission line spectra resulting from a pure photoionization field in a Strömgren sphere can be studied using diagnostic diagrams (Baldwin et al. 1981; Evans & Dopita 1985; Veilleux & Osterbrock 1987; Osterbrock 1989; Dopita & Sutherland 1995; Rola et al. 1997). These standard diagrams are based on nebular line ratios like [O iii]/H , [O i]/H , [N ii]/H , and [S ii]/H . For high redshift galaxies, the red part of the spectrum is shifted to the near-infrared. If optical studies are needed, other diagnostic lines like [O ii] 3727 or [Ne iii] 3869 can also be relied upon (Rola et al. 1997). These diagnostic diagrams are extremely useful in distinguishing between normal photoionized regions and regions with another ionization mechanism (e.g. high-velocity shocks, hard UV fields). The line ratios above are also independent of the reddening correction and depend only on the accuracy achieved for the spectrophotometry. However, despite the usefulness of these diagnostic diagrams for determining whether another ionization mechanism is present or not, it is very difficult to identify this mechanism. For this, sophisticated nebular models must be used (e.g. Stasinska 1990; Dopita & Sutherland 1995).
Fig. 1 illustrates the first diagnostic diagram with a correlation between [O iii]/H and [N ii]/H for the bar regions (dark symbols) and the disc regions of our control sample (open symbols). The dotted lines define the limits of the sequence of normal H ii regions that can be found in Osterbrock (1989) and Kennicutt et al. (1989). The full line is the separation between normal photoionized regions and regions with another ionization mechanism. We have also indicated the effect of high-velocity shocks and magnetic fields on these line ratios from models by Dopita & Sutherland (1995). The average location of Seyfert 2 and LINER galaxies in this diagram are also displayed. At this time, it is still not entirely clear that the spectral characteristics (mainly the high [N ii]/H ratio) of these last objects are due to high UV radiation or high-velocity shocks (see discussion by Dopita & Sutherland 1995). However, Fig. 1 shows that excepting for four regions, all the H ii regions in our sample are located inside the sequence of normal H ii regions . These four regions are located close to the nucleus in NGC 3504 (a starburst galaxy) and NGC 7479 (LINER). Thus, from this diagram alone, bar H ii regions do not exhibit any sign of high-velocity shocks or hard-UV radiation. Apart from these central peculiarities, there is no dependence on radius.
Although [N ii]/H is a good diagnostic ratio for high-velocity shocks or very hard ultraviolet radiation (Veilleux & Osterbrock 1987; Dopita & Sutherland 1995), other ratios like [S ii]/H or [O i]/H are more sensitive to these ionization mechanisms. Fig. 2 illustrates another diagnostic diagram: the correlation between [O iii]/H and [S ii]/H . The dashed and full lines are as in Fig. 1. Arrows also show the effect of the high-velocity shocks and magnetic fields. Although most of the H ii regions fall within the normal region sequence, there is a small number of bar regions outside the sequence. Fig. 3 shows the sequence between [O iii]/H and [O i]/H . The latter ratio was detected in about 67% of the sample of bar regions but only in about 40% of the disc region sample.
From these diagrams, it is clear that most bar H ii regions do not exhibit any systematic evidence of high-velocity shocks (150 km s-1) or very hard UV radiation . Only circumnuclear regions exhibit obvious signs of another ionization mechanism (fast shocks and/or hard-UV radiation), as expected from the work of Kennicutt et al. (1989). Nevertheless, if these conditions do not appear to be present in bar regions, Figs. 2 and 3 suggest that shocks with lower velocity or an abnormal UV photoionization field cannot be excluded. This possibility is discussed in Sect. 4.
3.2. Electronic density
The electronic density in H ii regions can be accessed with the line ratio of the [S ii] doublet at 6717-6731 Å, i.e. (Osterbrock 1989). The density can be derived from nebular models published by Blair & Kirshner (1985). Since the calibration depends on the nebular temperature (), we will assume K. This value is probably too high for the H ii regions in our sample with abundances higher than the solar value (see Sect. 3.5). However, for comparison purposes, this approximation is appropriate. The electronic density distributions for the bar and disc H ii regions are illustrated in Fig. 4. In both samples, several regions show line ratios that are very close or larger than the low-density limit (). The electronic densities derived in this regime are very uncertain because the doublet ratio becomes only weakly dependent on for values larger than about 1.3.
Kennicutt et al. (1989) found that nuclei H ii regions tend to possess higher electronic densities on average than disc regions, with both classes showing a large range of densities. As seen in Fig. 4, the case of H ii regions located in bars is different. No significant difference is observed between the distributions of electronic densities of both populations . On average, (bar regions) and (disc regions). Bar H ii regions do not show any compactness with respect to disc regions.
As discussed in the Sect. 2.2, the visual interstellar extinction of individual H ii regions can be derived from the H/H line ratio. The values derived, however, are approximate since in reality, the real extinction is probably not distributed uniformly but is patchy. Fig. 5 presents the distribution of the visual interstellar extinction from both the bar and disc H ii regions. There is a considerable extinction in both populations of H ii regions. The mean values are and visual magnitudes for the bar and disc regions, respectively. This difference is probably not significant since the extinction derived in bar regions, based on the assumption that the nebular temperature is K, might be overestimated by about 0.1 to 0.2 magnitude.
The extinction values for the bar regions show a large dispersion. Most of the H ii regions contributing to the highest values are located in the bars of NGC 7479 and NGC 5921. As noticed in MF97, these regions are located close to the strong dust lanes seen in these bars. Nevertheless, the overall behavior shows that the interstellar extinction as derived from the Balmer decrement is similar for bar and disc regions. However, using IRAS observations, Phillips (1993) has shown that for circumnuclear regions the extinction derived from the H/H ratio can be underestimated by as much as 2 magnitudes. The "uniform screen" model assumed for the extinction is probably over-simplistic. Any line ratios (e.g. ) or other quantitative properties (e.g. integrated fluxes) severely affected by the interstellar extinction should be interpreted with caution for regions located in the inner parts of galaxies.
3.4. H equivalent widths
The equivalent widths (EW) of the Balmer emission-lines provide a measure between the number of ionizing and continuum photons emitted in the H ii region. As such, the EWs depend strongly on the stage of evolution of the ionizing stars, the initial mass function (IMF), and the metallicity (Dottori 1981; McCall et al. 1985; Copetti et al. 1986; Bresolin & Kennicutt 1997; Bresolin et al. 1999; Leitherer et al. 1999). As shown by Copetti et al. (1986) and more recently by Leitherer et al. (1999), EW(H) and EW(H) can both be used as age indicators for H ii regions. In disc galaxies, the distribution of EW(H) extends from about 100 Å to 1500 Å with a median value around 400 Å. No obvious correlation with the Hubble type is found (Bresolin & Kennicutt 1997). Assuming an instantaneous burst of star formation and a solar metallicity, these values correspond to an age range between 1 Myr to 7 Myr (Leitherer et al. 1999).
The accuracy of the Balmer line EWs is mostly determined by the uncertainty in the level of the nebular continuum which is severely contaminated by the galactic continuum. Because we could directly measure the contribution from the galactic continuum on our "off" band images used in MF97, we have only measured the EWs for the H line. The EW(H) is also less affected by the interstellar extinction and the underlying absorption. The fraction of the galaxy light contributing to the nebular continuum was estimated from two photometric apertures: one covering the integrated light of the H ii regions, and the other located on nearest area devoid of any H emission (determined from the H images). A correction factor was then applied to the EW values measured directly from the spectra. These correction factors vary from 1.1 to 20 depending on the location of the H ii region.
Fig. 6 illustrates the EW(H) distributions of the bar and disc H ii regions. The mean values for the distributions differ by about a factor of two: EW(H) 250 Å (bar regions) and EW(H) 560 Å (disc regions). Following the models of Leitherer et al. (1999) for an instantaneous burst of star formation with a Miller-Scalo mass function and solar metallicity, the mean age of bar regions is about 5.3 Myr while disc regions are about 4.0 Myr old. However, the difference in age is less ( yr) when a Salpeter function is used to describe the IMF. Also, no age gradient seems to exist along the sequence of bar H ii regions. This is an indication that H ii regions should be ignified all along the bar and not only at bar ends with a subsequent migration towards the center.
In their study of nuclear H ii regions, Kennicutt et al. (1989) found that the EW(H) of the H ii region nuclei, with a median value around 25 Å, are approximately 20 times lower on average than that of normal disc regions. Such a difference cannot easily be explained by assuming that the correction for galactic continuum was underestimated. The authors rather favor the idea that the stellar continuum is high due to continuous star formation in the same region or an unusual stellar mass spectrum in the ionizing clusters. In the present case, however, it is difficult to completely discard the effect of the contamination of the galactic continuum to the nebular continuum to explain the discrepancy observed between bar and disc regions. The location of the aperture used to measure the galaxy continuum has a strong effect on the correction performed. The light distribution in bars is not uniform and some bars have very patchy dust features. A systematic error of a factor 2 cannot be ruled out. In any case, no firm conclusion based on the difference observed in the EW(H) of both populations of H ii regions can be drawn from the actual sample.
3.5. O/H abundance (within bars)
The oxygen abundance in H ii regions can be derived either through semi-empirical calibrations (e.g. Edmunds & Pagel 1984; McGaugh 1991; Pagel 1997) or directly when the temperature can be measured from the nebular lines [O i] 4363 or [N ii] 5755. The latter, however, are generally detectable only for H ii regions with low oxygen abundance. In our case, almost all the regions have solar or above-solar oxygen abundances; semi-empirical techniques have to be used. Even if the uncertainties related to these methods are generally quite large (0.2 dex), it is worthwhile to derive the O/H values to address the important question of mixing of the ISM in bars. It is now well established that bars induce large-scale mixing of the chemical composition in the disc of spirals (Martin & Roy 1994; Zaritsky et al. 1994; Friedli et al. 1994). The radial flows of gas formed by bars flatten the strong (negative) abundance gradients generally observed in unbarred late-type disc galaxies. The importance of the homogenization effect is related to the bar strength as shown by Martin & Roy (1994). Very strong gas flows ( km s-1) are taking place along bars; efficient mixing should be also observed and the O/H scatter between the bar H ii regions should be smaller than what is observed in normal galaxy discs (See Sect. 4).
The oxygen abundances for our sample of bar H ii regions were determined using three line ratios: [N ii] /[O iii] , [O iii]/H , and . The conversion to relative oxygen abundances was done using the calibration of Edmunds & Pagel (1984). Much has been written on the accuracy of these line ratios as abundance indicators (e.g. McGaugh 1991; Martin & Roy 1995; Stasinska 1998). For our sample, Fig. 7 compares the different O/H values derived with all three indicators. The O/H values derived from [N ii] /[O iii] are slightly higher (0.1 dex) than the values derived from [O iii]/H and for . For , that is, the abundance regime for most of the bars in our sample, the abundances from [O iii]/H and [N ii] /[O iii] are slightly below the values given by . These results were previously discussed by Martin & Roy (1994) and are due to discrepancies in the semi-empirical calibrations. For our purposes, we use the O/H values derived from the [N ii] /[O iii] indicator; our conclusions are not affected by this choice.
The distribution of the oxygen abundance along the bars of our sample is illustrated in Fig. 8. It is clear that the O/H scatter observed in these bars is well smaller than 0.1 dex or even less . Martin & Belley (1996, 1997) have shown that the azimuthal O/H dispersion observed in the discs of normal and barred galaxies is generally equal or larger than 0.2 dex. These abundance variations are probably not intrinsic but are in fact a combination of real inhomogeneities and uncertainties associated with empirical techniques (Roy & Kunth 1995). However, a comparative analysis remains possible when the same method is used to derive the O/H values (Martin & Belley 1997). Clearly, the chemical composition of H ii regions in a bar is well-homogenized. This indicates that an efficient mixing of the chemical composition is taking place in the bar region (see next section).
© European Southern Observatory (ESO) 1999
Online publication: June 17, 1999