Structure, occurrence, and ionization of the diffuse warm gas and especially the presence of large ionized shells in dwarf galaxies were discussed in a number of recent publications (e.g. Martin 1997, 1998; Hunter & Gallagher 1997; Hunter et al. 1993). Here I will concentrate mostly on the dynamics of the ionized gas and its implications for the fate for the hot gas.
First it is important to note here, that the warm ionized gas cannot be used for the metallicity determination of the outflows. We have seen that there are signs of at least some mixing at the boundary layers between hot and cold gas and therefore a metal enrichment of warm gas. This makes the observation of the optical emissions lines of the shells and outflows tempting as alternative way to estimate the metal content of the outflows. Still the spectra cannot be treated with the usual methods as used for normal HII region. The line ratios clearly indicate a complex interplay of normal stellar photoionization, photoionization by a diffuse photon field, shock ionization, photoionization by X-ray, and even turbulent mixing layers (Hunter & Gallagher 1997; Martin 1998, Tüllmann & Dettmar 2000). With insufficient information to even determine the ionization mechanism(s), a direct derivation of ionic abundances from optical emission lines for the diffuse ionized gas is out of the question (at least for now). A helpful alternative method may be the use of interstellar UV absorption lines to background QSOs. Unfortunately, no usable chance alignment of a dwarf galaxy outflow with a sufficiently bright background source has been found.
When one finds a dwarf galaxy showing large H shells, which extend to a size larger than the stellar body of the dwarf galaxy, the first critical question is if this ionized gas is really expanding away from the dwarf galaxy. Observationally the dynamics of the ionized gas can be studied either with high-dispersion, long-slit spectroscopy, or using a Fabry-Perot interferometer. While the long-slit approach only uses one spatial axis, it is stable against changes in instrumental and weather conditions, relatively easy to reduce, and can provide high spatial and spectral resolution (~ 10 km s-1). The one-dimensional spatial axis allows also a very good sky subtraction and therefore sensitivity to faint features. Producing a 2-d map on the other hand, requires offsetting the slit and is therefore very telescope time intensive. The scanning Fabry-Perot approach delivers a real data cube with two spatial and a spectral axis, but is somewhat susceptive to changes in the observing conditions and the data relatively hard to handle. Unfortunately, no Fabry-Perot dataset on dwarf galaxies has delivered yet at the same time the high spectral resolution and sensitivity of the best long-slit data. The alternative use of the Fabry-Perot as spectrometer delivers very high sensitivity at the cost of low spatial resolution and is therefore of only limited use for the study of outflows.
As an example I present here an analysis of the galaxy NGC 1705. NGC 1705 contains several large H shells (Meurer et al. 1992), two of which show up in a deep ROSAT PSPC pointing as very soft, diffuse X-ray sources (Hensler et al. 1998). The data presented here were taken with the ESO VLT and the ESO NTT and are going to be discussed in detail in Bomans et al. (2001b, in prep.). In Fig. 4 the continuum subtracted H image of NGC 1705 is shown with contours of the Gunn-r continuum image overlayed. The body of the galaxy is relatively smooth, indicating a long phase of low star formation activity, and one bright knot, which harbors a young globular cluster (Melnick et al. 1985; Ho & Filippenko 1996). The about 1.8 kpc large stellar body (Holmberg diameter) of the galaxy is nearly filled with HII regions and large ionized shells and filaments extend radially out to more than 2.5 kpc from the center of the dwarf galaxy. The structure appears roughly bipolar, consistent with hydrodynamical simulations, but shows a very complex substructure with interlocking and overlapping individual shells and filaments. One should keep in mind that we see in the image a 2-d projection of the 3-d structure, which may explain at least some of the shells-inside-shells as line-of-sight projection. The currently available HI synthesis map is of to low spatial resolution to study the detailed dynamic of the gas, but shows that NGC 1705 is embedded in a large HI envelope.
Figure 4. Continuum corrected VLT H image of NGC 1705 with contours of the continuum emission overplotted.
Fig. 5 shows a long-slit spectrum taken with a position angle of 310 deg, roughly aligned to the central star cluster of NGC 1705 and the bright foreground star in the north-west. The spectrogram runs top to bottom from north-west to south-east and shows a spectral range of 60 Å centered on the H line and the two [NII] lines. Note that the lines are redshifted due to the radial velocity of NGC 1705 and the [NII] lines quite weak due to the sub-solar metallicity of NGC 1705. Just below the bright star (the residuum from continuum subtraction running as white line left to right near the top the the spectrogram) the H line splits into two components and merges again at the beginning of the stellar body. This Doppler ellipse is the sign of an expanding bubble and coincides exactly with the bright loop in the north-west. The expansion velocity is 75 km s-1. The H line maintains a complex structure over the whole stellar body of NGC 1705, not surprisingly given the complicated structure of the H emission in this region visible in Fig. 4. At the bottom of the spectrogram the H line does not show a line split, but the width of the line is large, implying motions with velocities below the spectral resolution of ~ 30 km s-1. Interestingly, the slit cuts through another large shell, which is, contrary to the shells in the north-west, not detected in the deep ROSAT image. It is tempting to make a link between the lower expansion velocity of the shell in the south-east and the missing X-ray emission. Similar velocities are seen also in an spectrum with a position angle of 230 deg (Marlowe et al. 1995).
Figure 5. High-dispersion spectrogram of one slit position through NGC 1705.
Still, the situation is a bit more complicated, since neither the 75 km s-1 nor the 30 km s-1 shock speed would give rise to X-ray emission. It requires about 200 km s-1 to get post-shock temperatures above 106 K (McKee 1987). Simple projection effects are not likely to make a large effect for expanding bubbles. The probable explanation comes from the inherent selection effect of using the warm ionized gas as tracer of the flow: the gas detected with the spectrometer is the gas with highest surface brightness and therefore density, which is the least probable to show the highest velocities. We apparently measure the large scale expansion of the superbubble driven by the overpressure of the hot gas inside, which is heated by the supernovae. As we have seen in the case of N51D, a significant part of the X-ray emission results from recent supernovae shocks hitting the shell walls. Therefore there is a link between the temperature of the hot gas and the expansion velocity, but it is not the simple post-shock temperature from the large-scale shock, hitting the surrounding interstellar medium, but local heating near the inner shell wall. Indeed, we detect hot gas inside the shell walls not outside (Bomans et al. 2001b, in prep.).
This is also consistent with the missing correlation of the X-ray temperature with global galaxy parameters (Tab. 1). The visible clustering of the X-ray temperature of the diffuse gas near 0.2 and 0.8 keV could be an effect of the location of the local minima visible in the cooling curve (Boehringer & Hensler 1989) at low metallicity.
The shells of NGC 1705 are clearly expanding out of the galaxy and they are (at least in part) filled with hot gas. The question is now, what will be the fate of this bubbles? The faintest shells and filaments visible on our VLT H image reach the rim of the HI envelope of NGC 1705. If they are still expanding at this radius, they are likely going to speed up when hitting the gas density gradient. Unfortunately, the filaments are very faint and it is hard to measure their radial velocity. Even with a measured velocity, the lack of HI also means that the rotation curve and therefore the potential of the galaxy is not well constrained there. These problems are generic to the analysis method and should be kept in mind when evaluating similar analyzes of the mass loss from dwarf galaxies (e.g. Martin 1998).
There is another effect, which may be important for the processes discussed here. Ordered magnetic fields, if present, could have a large influence on the conditions of the outflows and the kinematics of the warm and hot gas. Unfortunately, the knowledge about magnetic fields in dwarf galaxies is very sparce. From theoretical considerations no large-scale ordered magnetic fields are to be expected in dwarf galaxies, but at least the LMC (Klein et al. 1993) and NGC 4449 (Chyzy et al. 2000) do show such magnetic fields. Still, both galaxies are at the upper mass and luminosity boundary for dwarf galaxies and the conditions in lower mass dwarf galaxies are unexplored yet.