4 Discussion

The line fluxes presented in this paper suffer from a number of extra uncertainties, which break down into two groups. One has to do with the extinction correction of the fluxes, the other with the problem of different instrumental beams and the size of the sample objects. The problems connected with the extinction correction are the most important for the optical line fluxes. For the infrared data the extinction is negligible.

4.1 Extinction correction

The optical spectra given in this paper are all driftscan spectra. The surface integrated spectra make the comparison between the optical and infrared fluxes easier, but they also present a problem for the extinction correction of the line fluxes. This has to do with the non-uniform nature of the extinction across an object, which is especially a problem in very complex environments like star-forming regions, where the extinction can change drastically from one point to another. A good example in this respect is the SMC object N88A. In a recent paper, Heydari-Malayeri et al. (1999a) showed that even within an object with such a small angular size (3$\farcs$5), the H$\alpha$/H$\beta$ intensity ratio can change drastically within a few arcseconds. This general "patchiness'' of the extinction across the face of an object, however, is smeared in a driftscan spectrum. The resulting Balmer-decrements on which the extinction correction is based are therefore a complicated "weighted average'' that reflects the variation in extinction properties encountered across the area scanned.

Related to the problem of the non-uniformity of the extinction, is the issue of what extinction curve to apply. In this work, a general extinction curve was used for the reddening correction of the optical data. The assumption made in the correction applied was that all the extinction can be ascribed to foreground extinction. Furthermore, an implicit assumption made is that the properties of the absorbing matter are the same everywhere, which is reflected in the "unchangeable'' shape of the extinction curve adopted.

The assumptions about the origin of the extinction and the shape of the extinction curve are invalidated when extinction within the source itself also becomes important. The absorption and scattering properties of the matter inside the source can be different from those of the matter external to the source, this as a result of the drastically different environment within the source. In this case, the shape of the extinction curve will be modified. A situation where this can become important is when dust is mixed with the emitting gas of an object. Objects in the sample which are known to have internal dust are SMC-N88A (Kurt et al. 1999), and the LMC objects N159-5 and N160A1. Furthermore, any dust mixed with the ionized gas can decrease the Balmer-decrements found in the spectrum (Mathis 1983).

Both of the uncertainties described here will have repercussions for the line fluxes given. The smearing of the extinction in a driftscan spectrum will lead to an underestimate of the correction to apply, while the unknown variations in the shape of the extinction curve will give unknown wavelength dependent variations in the total extinction correction. One should bear in mind, however, that everything said in this Section will only affect the fluxes which have actually been corrected for extinction (i.e. the optical fluxes). The same problems also affect the extinction correction of the infrared lines but no correction has been applied here.

  
4.2 Telluric absorption

An additional problem affecting the optical line fluxes is the absorption from water in the atmosphere. The line fluxes red-wards from 8000 Å are affected by it, and it is especially important for the [S III] 9068 Å and [S III] 9532 Å lines. The presence of this absorption is clearly visible in the standard star spectra and in some of the object spectra. The amount of telluric absorption is difficult to asses, because it depends critically on the degree of overlap between the telluric and the spectral lines. Also, given the low spectral resolution of our spectra of about 4 Å, the narrow telluric lines are blended and only a broad absorption feature is seen. We can, therefore, only give a rough estimate of the extra uncertainty on our line fluxes introduced by this absorption, and did not attempt to actually correct the fluxes for this. We thereby focussed our attention exclusively on the crucial [S III] lines.

To get a rough idea of the magnitude of the telluric extinction, the depths of the absorption features seen in the standard star exposures were measured. The absolute values of the amount of absorption themselves were not very useful because of the changes in the atmospheric conditions from night to night, but it was possible to establish a ratio for the absorption near [S III] 9068 Å and near [S III] 9532 Å. It turned out that the absorption at [S III] 9532 Å was in all cases roughly three times larger than at [S III] 9068 Å. The absorption near [S III] 9068 Å was at most 0.1 mag.

The actual estimation of the extra uncertainty on our line fluxes involved the use of the theoretical line flux ratio of the two [S III] lines. The theoretical ratio [S III] 9532/[S III] 9068 is 2.44, the ratio of the transition probabilities. The observed [S III] 9532/[S III] 9068 ratio was compared with the theoretical one, and the correction factor needed to arrive at the theoretical ratio was derived. The deviation of the observed ratio from the theoretical one is (most likely) due to absorption in both the [S III] lines. The correction factor described above, therefore, depends on the effect of the (unknown) absorption on both the lines. Taking this dependency into account, we derived a matrix of correction factors for the [S III] 9532 Å and the [S III] 9068 Å line flux from which the uncertainty was estimated.

Taking into account the relative difference in absorption near the two lines and the fact that the [S III] 9068 Å line appeared more reliable than the [S III] 9532 Å line, the investigation of the parameter space of correction factors led us to estimate that there is an extra uncertainty in the line fluxes of up to 20%. This uncertainty is not included in the error given for these lines in Tables 3 and 4.

  
4.3 Apertures and beams

An important issue, concerning the infrared line fluxes, is the question whether or not the sample objects are pointlike or extended compared to the ISO-SWS/LWS apertures. Any deviation of the source morphology from a perfect point source gives rise to a set of inter-related problems, resulting from the multitude of apertures and beams involved with the ISO spectrometers.

The total spectral coverage of the ISO spectrometers is subdivided into several subregions or bands, and all these different parts of the spectrum are observed through different apertures. The four rectangular ISO-SWS apertures through which the different spectral ranges are observed cover an area on the sky of 14 $\arcsec\times$ 20$\arcsec$ (2.38-12.0 $\mu$m), 14 $\arcsec\times$ 27$\arcsec$ (12.0-27.5 $\mu$m), 20 $\arcsec\times$ 27$\arcsec$ (27.5-29.0 $\mu$m) and 20 $\arcsec\times$ 33$\arcsec$ (29.0-45.2 $\mu$m), respectively. The circular ISO-LWS aperture is about 80$\arcsec$ in diameter. The result of the subdivision is that, depending on wavelength, different spatial areas of a sample object are observed.

All these apertures also have their own wavelength dependent throughput or beam profile. The beam profiles drop off sharply when one moves away from the centre of the aperture, which can lead to a significant loss of flux. In the ideal cases of an infinitely extended and homogeneous object or a point source, a correction for the loss of flux can be made. However, for a realistic object, where the resultant flux is a convolution of the beam profile and the source morphology, it is very difficult to make such corrections.

With these two facts in mind, it will also be clear that not only the source morphology is of importance but also the position of the source in the aperture. If an object is located at the edge of one or more of the apertures, the flux losses can be severe. These losses can hardly be traced or restored. It is therefore important that the intended target is in the centre of the aperture. Objects in our sample not satisfying this requirement are SMC-N81 and SMC-N88A, which are lying near the edge of the LWS aperture (see Fig. 3, center).

In this respect, a comment must be made about the pointings of the two ISO spectrometers for the sample sources. As can be seen in Table 1 and even more clearly in Figs. 2 and 3, there are sometimes considerable differences between the pointings of SWS and LWS, despite the fact that the pointings were meant to be exactly the same. The most serious cases are LMC-N11A (Fig. 2, lower left), where LWS was actually pointed at LMC-N11B, and the object LMC-N79A which falls outside the SWS beam completely (Fig. 3, upper left). In many cases, though, the pointing differences are only slight.

The problems described above raise the issue of flux (in)compatibility. This is a serious handicap for the analysis done with the line fluxes, which is primarily based on flux ratios of different fine-structure and recombination lines. The impact of the flux incompatibility on the analysis depends on the lines used. For lines observed through the same aperture, where the degree of incompatibility is likely to be small (e.g. Br$\beta$ and [S IV] 10.5 $\mu$m), the result will be more reliable than for lines observed through different apertures (e.g. [S III] 18.7 $\mu$m and [S III] 33.5 $\mu$m). It is obvious that any analysis based on the infrared line fluxes should be done with these caveats in mind.