Discussion

The major goal of our observational program described in §3 is to determine what is the formation mechanism of boxy/peanut-bulge spiral galaxies. Therefore, we will briefly discuss here the implications of the observations we have just described. While we only report on seven galaxies in this paper, our total sample comprises about 32 galaxies. Thus, our conclusions will be on much firmer ground when all our data is obtained, reduced, and analysed.

The NGC 5746 data we obtained (Fig. 1a) is so far unique. The PVD displays an enormous amount of detail; most interesting here, line-splitting is very strong in the inner regions and present in all the lines. Kuijken & Merrifield (1995) (see also Merrifield 1996) showed that this supports the bar buckling hypothesis by linking the line-splitting (``figure-of-eight'') with the presence of a bar. Our NGC 5746 data confirm their observational results. The other boxy/peanut-shaped bulge galaxies NGC 6722 (Fig. 1b), IC 5096 (Fig. 1c), and IC 2531 (Fig. 2) also show double-peaked PVDs. The peanut-bulge galaxy IC 4767 (Fig. 1d) does not show line-splitting, but this might simply be due to the fact that the gaseous emission does not extend far enough in the disk. In fact, a hint of line-splitting is seen in the strong H absorption line. In addition, for NGC 5746, NGC 6722, and IC 5096, the split in the emission lines (at least in [N II] at 6584 Å) extends out to about twice the bulge (peanut) length. Considering that simulations suggest (e.g.\ Sellwood 1981, Combes & Sanders 1981) that corotation occurs just outside the end of the bar, associating the peanut shape with a bar is consistent with Kuijken & Merrifield's (1995) model which predicts that the line-splitting extends to about two corotation radius.

Before claiming detections of triaxiality, one needs to consider whether other structural or dynamical components in an axisymmetric spiral galaxy could give rise to a feature similar to the ``figure-of-eight'' in the gas PVD. The most obvious candidate is a ring embedded in an axisymmetric disk. The superposition of the apparent solid-body rotation curve of an edge-on ring and an intrinsically more-or-less flat rotation curve would produce a double-peaked gas PVD, qualitatively similar to the gross features of the distribution shown in Fig. 1a for NGC 5746. We note, however, that the PVD for NGC 5746 is in detail much more complex than this simple picture would allow. Merrifield (1996: Figure 3) has shown that the PVD signature expected from an edge-on barred spiral galaxy can be similarly complex, depending on the viewing angle. It would be difficult to reproduce this level of complexity by simply adding or subtracting a dynamical component to an axisymmetric disk, a redistribution of the gas is really needed. In addition, a ring would produce a perfectly solid-body rotation curve, while the components of the PVDs observed are curved. Furthermore, rings can not account for PVDs of the type observed in IC 5096, where the ``solid-body'' component of the PVD is rotating faster than the disk. Therefore, based on Kuijken & Merrifield's (1995) and Merrifield (1996) work, we believe that our data represent the strongest case made so far in favour of the bar-buckling instability scenario to form boxy/peanut-shaped bulge spiral galaxies.

Although our data agree at least qualitatively with the modelling of Kuijken & Merrifield, some details are puzzling. First, the line-splitting region in the HI PVD of IC 2531 (Fig. 2) extends to about 5 peanut-lengths. Why is that? Many possibilities exist: maybe the HI does not trace the same structures as H and [N II], maybe there is a very extended weak bar in the disk of IC 2531 giving rise to the extended line-splitting, maybe we are seeing the signature of strong spiral arms, ... IC 2531 is particularly interesting as it looks very much like the Galaxy. Therefore, pinpointing its structure and dynamics would help us greatly to understand our own Milky Way. Another interesting result is the structure seen in the PVD of the late-type galaxy ESO 240-G 11 (Fig. 1e). It does seem to show line-splitting in the inner regions, although the S/N is poor. This is unexpected as the bulge of ESO 240-G 11 is not boxy or peanut-shaped. Of course, ESO 240-G 11 could have a very weak bar in its disk, not giving rise to a boxy/peanut-shape structure, but still producing a split PVD. Such a possibility is hard to verify but would mean that thin bars do exist.

On a different note, our observations raise a worrying prospect concerning rotation curves of galaxies. In our long-slit observations, three galaxies out of the four that have structure in their PVD show a different behaviour in the H and [N II] lines (the galaxies are NGC 6722, IC 5096, and ESO 240-G 11). For those three galaxies, in the region where a double-peaked PVD is seen in the [N II] 6584 Å line, the H line only shows the smaller rotation velocity peak (IC 5096 in Fig. 1c is the best example). This might simply be due to the fact that the H line is superposed on the corresponding stellar absorption line. That idea is supported by the fact that we see an identical behaviour in the H and [N II] lines only in the galaxy with the strongest emission lines (NGC 5746). Nevertheless, if we were to determine the rotation velocity of the gas from such data, without correcting for the stellar absorption, the [N II] line (unaffected by absorption) would yield a rapidly rising rotation curve which flattens abruptly at a relatively small radius, while the H line (affected by absorption) would yield a slowly rising almost solid-body rotation curve which flattens out at a relatively large radius. The two rotation curves thus obtained would be significantly different, even on a qualitative level. This fact has at least two important consequences. First, many rotation curves derived in the conventional manner using the H line might be wrong (at least for galaxies significantly inclined). Secondly, if some rotation curves are wrong, the galaxies mass models based on them will be as well. Therefore, our understanding of galactic dynamics and structure based on those H rotation curves could be seriously erroneous.

The observations also show the potential use of our data to determine the ionisation conditions and abundance of the gas in different regions of the galaxies, as we can measure the spatial behaviour of line ratios. Particularly, in addition to identifying nuclear activity, we can study large scale changes in the line ratios, such as those seen in some of our data where the whole bulge/bar region shows unusual physical conditions (NGC 5746 in Fig. 1a is the best example). Why does the [N II]/H ratio change so much in those regions? This may prove a really interesting research direction to follow as we get data on more sample galaxies. Hopefully, we will learn something about the effects of bars on the ISM, and their relevance to the fuelling mechanism of active galaxies.

© Copyright Astronomical Society of Australia 1997