Proceedings of the Workshop
"The Magellanic Clouds and Other Dwarf Galaxies"
of the Bonn/Bochum-Graduiertenkolleg

The Violent Interstellar Medium in Dwarf Galaxies: Atomic Gas

Elias Brinks1 and Fabian Walter2

1Departamento de Astronomía, Guanajuato, México
2Radioastronomisches Institut, Bonn, Germany

Received 24th May 1998
Abstract. We review the morphology of the warm, neutral ISM as observed in the 21 cm line of neutral hydrogen (H I) of several nearby, gas-rich (dIrr) galaxies. The H I emission is dominated by shell-like structures, most likely superbubbles produced by the combined effects of strong stellar winds followed by supernova explosions of stars more massive than M > 8 Msun within a region of massive star formation, an OB association. Somewhat counter-intuitively, H I superbubbles grow to larger dimensions in dwarf galaxies than in large spiral galaxies like our own or M 31. This can be explained as follows. In dwarf galaxies, the gravitational potential is lower than in spirals. Hence, for similar H I velocity dispersions, the scaleheight of the H I layer will be larger. Then, for comparable H I surface densities, the volume density will be lower. Both effects facilitate the growth to large dimensions of shells in dwarf galaxies and explains why such shells are much less likely to break through the H I layer into the halo. Moreover, dwarf galaxies lack density waves and tend to be dominated by solid-body rotation. As a result, shells will persist much longer than in spirals. A comparison with other galaxies shows that the energies needed to create H I supershells are the same for all types of galaxies, the energy output of a typical star-forming region therefore not being related to its galactic environment, at least to first order. The overall statistical properties of the H I holes and shells in galaxies show clear trends with Hubble type (or rather mass), such as in their diameter distribution, expansion velocities and ages.


1. Introduction

Since the discovery by Heiles (1979, 1984) of huge, shell-like structures in the H I distribution of our Galaxy similar such structures have been found in several nearby galaxies, such as in M 31 (Brinks & Bajaja 1986), M 33 (Deul & den Hartog 1990), Holmberg II (Puche et al. 1992) and the Small Magellanic Cloud (Staveley-Smith et al. 1997). Such maps would not have been possible without powerful aperture synthesis radio telescopes such as the Very Large Array (VLA), the Westerbork Synthesis Radio Telescope (WSRT) or the Australia Telescope's Compact Array (ATCA). One of the most stunning examples of what is possible these days is, without doubt, the image shown by Sungeon Kim of the LMC (Kim, this volume). Both the LMC and SMC data are a vast improvement over the old Parkes single dish maps.

All these observations indicate that the interstellar medium (ISM) of medium to late-type galaxies is dominated by features, which have variously been described as shells, rings, holes, loops, bubbles or cavities. The origin of these structures is generally thought to lie in the combined effects of stellar winds and supernova explosions produced by young stellar associations. For review articles see Tenorio-Tagle & Bodenheimer (1988), and van der Hulst (1996, and references therein). Based on a simple model of holes created by O and B stars, Oey & Clarke (1997) successfully predict the observed number distribution of SMC holes, lending support to this hypothesis.

Much progress in our understanding of the detailed processes involved has been made through numerical simulations (e.g., Palous et al. 1990; Silich et al. 1996; Palous et al. 1994; Mac Low et al. 1998). The reader is referred to Palous (this volume) and Mac Low (this volume) for the latest developments and refinements.

Just a word of caution, not all H I shells are necessarily the product of a major star formation event. In the case of the largest observed shells, their energy requirements seem to exceed the output of stellar winds and supernovae. To explain those structures an alternative mechanism was proposed, the infall of gas clouds. Tenorio-Tagle et al. (1987) present a numerical simulation and van der Hulst & Sancisi (1988) provide what is probably the best observational evidence for infall, the case of one of the largest holes in M 101.

There is another potential problem with the hypothesis that the H I holes are the result of an evolving OB association in which the most massive stars, through their winds and supernovae, create the observed supershells. Using reasonable assumptions one still expects a substantial population of A and F main sequence stars to be present. However, searches by Radice et al. (1995) and Rhode et al. (1997) in galaxies like Ho II have not led to the expected result. A possible alternative explanation has been proposed by Efremov et al. (1998) who suggest that Gamma Ray Bursts, which have been conclusively associated with objects at cosmological distances, might provide the required energy, and occur frequently enough, to explain the observed H I supershells.

To date, most authors decided to concentrate on large spiral systems. However, there are several advantages in using dwarf galaxies to study the structure of the ISM and how this is shaped by rapidly evolving, massive stars. Dwarfs are slow rotators, generally show solid body rotation, and lack density waves. This implies that once features like shells have formed, they won't be deformed by galactic shear and therefore tend to be long lived. Moreover, the overall gravitational potential of a dwarf is much smaller than in a normal spiral. The same amount of energy input of a star forming region therefore has a more pronounced impact on the overall appearance of the ISM, as shown by Puche et al. (1992).

2. Recent Results

Fig. 1 shows an example of what the neutral gas in a dwarf galaxy looks like. It is an H I surface brightness map of IC 2574, a dwarf irregular galaxy which belongs to the M 81 group (assumed distance 3.2 Mpc). The data were obtained with the NRAO* Very Large Array (VLA). The spatial resolution of our maps is about 6" (or ∼95 pc linear resolution); the velocity resolution is 2.6 km s-1. Note that the beamsize is so small that it can hardly be properly reproduced. A total of 48 supershells could be identified in this object. More information on IC 2574 is given by Walter & Brinks (this volume). A paper presenting a full analysis of the H I holes has been submitted (Walter & Brinks 1998).

[Click here to see Fig. 1!]

Another interesting system is the dwarf galaxy DDO 47 which is located at a distance of about 4 Mpc. Walter & Brinks (this volume) present a preview of this system. The observations were also made with the VLA at the same spatial and velocity resolution as IC 2574. In this object somewhat fewer H I shells were detected, nineteen in total. Their extent and location are displayed in Fig. 2.

[Click here to see Fig. 2!]

Wilcots & Miller (1998) have collected VLA data on IC 10. The H I surface brightness map is displayed in Fig. 3. They find a total of eight H I holes. Based on the energies involved most of the holes are wind-blown, not SN driven (Wilcots, priv. comm.). A very young bubble coincides with the region which was discovered by Yang & Skillman (1993) at radio continuum wavelengths. A position-velocity diagram through that location shows a kink which is indicative of an expansion velocity of 25 km s-1. A first analysis suggests that about 10 SNe are providing the energy needed to blow this structure (Wilcots, priv. comm.).

[Click here to see Fig. 3!]

In addition to the objects listed in the introduction and the maps shown here, several more dwarf galaxies have been observed. Walter & Brinks are working on data on NGC 3077 and Holmberg I. Van Dyk et al. (1998) have recently submitted a paper on Sextans A. Wilcots and collaborators have data on two more galaxies, IC 1613 and NGC 4449, the former object being completely dominated by H I holes and shells, much like IC 2574 and Ho II. Within the coming year a lot more material should thus become public, allowing for comparative studies to be performed.


* The National Radio Astronomy Observatory (NRAO) is operated by Associated Universities, Inc., under cooperative agreement with the National Science Foundation.

3. Some Surprises

As was first noticed by Puche et al. (1992) in Ho II, the scale height of the H I layer in dwarf galaxies is larger than in big spiral galaxies. This holds true not only in relative terms, but in absolute terms as well! The aspect ratio of dwarf galaxies tends to be of the order of 10:1 rather than 100:1. Let h be the 1 σ scale height of a presumed Gaussian H I distribution perpendicular to the disk. We find for Ho II: h=625 pc (Puche et al. 1992); NGC 5023: h=460 pc (Bottema et al. 1986); IC 2574: h=350 pc (Walter & Brinks 1998) and DDO 47: h≅700 pc, the largest value derived thusfar (Walter & Brinks, this volume). This, at first sight somewhat surprising result, can be understood as a direct consequence of the fact that the gravitational potential of a dwarf galaxy is lower whereas the measured velocity dispersion of the gas is within the same range (6-9 km s-1) as what is found in spiral galaxies.

For similar H I column densities perpendicular to the plane, the above result implies that the volume density is reduced accordingly. The smaller gravitational potential in dwarf galaxies therefore helps the growth of H I shells in various ways. The gravitational pull on an expanding shell is lower and the ambient density is lower, both effects favouring large shells. Moreover, because of the increased scaleheight, holes have to grow to yet larger dimensions until they will break out of the thick H I layer. This all helps explaining the somewhat counter-intuitive finding why the shells in dwarf galaxies are larger, in absolute terms, than in large spiral galaxies.

We should briefly like to mention another aspect, which applies equally to spiral and to dwarf galaxies. In IC 2574, for example, the holes are found to be much more prominent in the single velocity (or channel) maps than in the integrated H I map. In contrast, in Ho II holes can be identified far more easily in the surface brightness map. This we interpret as being due to the difference in inclination. In IC 2574, H I shells at different locations within the disk and possibly different velocities overlap along the line-of-sight, as is the case in M 31 (Brinks & Bajaja 1986). The total H I surface brightness map was therefore of little use for the identification of a hole. In the case of Ho II, because of its much lower inclination, only the integrated map was used. This implies that the approach to search for holes needs to be adapted to the orientation of the galaxy.

4. Statistical Properties of the Holes

As we alluded to already in the introduction, the properties of H I holes depend on the type of galaxy they find themselves in. As we will show, the effects of stellar winds and SN explosions become progressively more pronounced as a function of Hubble type. In what follows we will compare the observed and derived H I hole properties of IC 2574 with those found in M 31 (Brinks & Bajaja 1986), an example of a massive spiral galaxy similar to our own, M 33 (Deul & den Hartog 1990), a less massive spiral and Ho II (Puche et al. 1992), another dwarf galaxy in the same group of galaxies, yet four times less massive than IC 2574. In other words, the sequence M 31 - M 33 - IC 2574 - Ho II spans a large range of different Hubble types from massive spirals to low-mass dwarfs.

Despite the fact that a similar study, and at considerably higher sensitivity, is now available for the SMC (Staveley-Smith et al. 1997), we decided not to include their results. The main reason being that the linear scales (or spatial frequencies) sampled by the ATNF hardly overlap with those observed in the galaxies listed above. The linear resolution, at 28 pc, is almost four times higher than, for example, the VLA maps on IC 2574. At the other end of the spectrum, because of the lack of short spacing information, structures larger than a few hundred parsec will have been missed. Moreover, the SMC is a very disturbed system, being torn apart by tidal forces due to interactions with the LMC and the Galaxy. An additional argument for leaving out the SMC is that Staveley-Smith et al. used a different approach in searching and identifying the holes which makes a direct comparison difficult.

Other objects with catalogued H I holes, such as DDO 47 (Walter & Brinks 1998) and IC 10 (Wilcots & Miller 1998) have so few holes that a statistical analysis is not warranted. Limiting the comparison to the four galaxies listed above has some further advantages. The linear resolutions are very similar, as are the velocity resolutions with which they have been observed (see Table 1 for a summary). In addition, all four galaxies were examined in more or less the same fashion, one of the authors (EB) having taken part in the analysis of three of the four objects. We are aware that the results for the four galaxies suffer partially from low statistics and incompleteness due to personal bias and observational constraints (such as the beamsize). However, we feel that these effects, to first order, affect a comparison in a similar way and that it is valid to try to find global trends as a function of Hubble type. In order to remove the human factor, it would be interesting to apply an automated object recognition package such as that developed by Thilker et al. (1998) to all galaxies with sufficiently detailed observations.

Table 1. Summary of H I hole statistics in four nearby galaxies
Property M 31a M 33b IC 2574c Ho IId
Linear resolution (pc) 100 55 95 65
Velocity resolution (km s-1) 8.2 8.2 2.6 2.6
Average surface density (1020 cm-2) 5 9 4 10
Average volume density (cm-3) 0.6 0.45 0.15 0.2
Average velocity dispersion (km s-1) 8 8 7 7
Derived scaleheight (pc) 120 100 350 625
Number of holes 141 148 48 51
Sensitivity per channel (1020 cm-2) 0.3 1.0 0.5 2.4
aBrinks & Bajaja 1986
cWalter & Brinks 1998
bDeul & den Hartog 1990
dPuche et al. 1992

In Figs. 4-8 we compare the distributions of swept-up matter, the energies, diameters, expansion velocities and ages of the holes in the 4 galaxies. In each graph we plot in the form of histograms the relative number of holes, in percent, in order to make a direct comparison possible. In each histogram, the data for the individual objects are binned in the same way. To improve the presentation, however, the respective bins were slightly shifted when plotting the results.

Fig. 4 shows an overlay of the indicative masses of the H I holes, binned logarithmically. The values range from typically 104 to 5·106 Msun. We see that, within the errors, the distributions of the indicative masses are almost the same, with the masses for Ho II perhaps being slightly higher as compared to the other galaxies. This is surprising if one realises that the volume density in the plane for the ambient medium measured in a spiral galaxy is typically higher by a factor of three than that found in a dwarf galaxy (the H I column densities have a much smaller variation from object to object, though; see Table 1). As a warning, it should be mentioned that, in order to determine the H I densities we simply smeared out any fine scale structure, potentially adding somewhat to the errors.

Fig. 5 is similar to Fig. 4, showing instead the distribution of the logarithm of the energies needed to create the observed H I holes. These energies refer to the initial total energy deposited (Chevalier 1974), not the currently observed kinetic energy in the H I shell. We find the very important result that the distributions are identical. Although the derived energies should be treated as order of magnitude estimates, it is clear that the global distribution over energy per galaxy is similar. The derived energies range from 1050 to 1053 ergs. So, from this we can conclude that regardless of the size, mass and type, the star clusters formed in a typical galaxy are, to first order, identical and deposit more or less the same amount of energy into the ISM.

Fig. 6 shows an overlay of the relative size distribution of the holes found in the four galaxies. In this and the following plots the bins are on a linear scale. Note that there is a clear sequence with Hubble type! The size distribution for holes in M 31 and M 33 cuts off sharply near 600 pc. In contrast, holes in IC 2574 and Ho II reach sizes of 1200 to 1500 pc, respectively. The lack of holes with sizes smaller than ∼100 pc is due to our resolution limit. As we explained in the previous section, holes are larger for ``later'' Hubble types because these smaller galaxies have lower masses and hence a lower mass surface density. So, for the same amount of energy deposited, an H I shell can grow much larger, both because of a lower gravitational potential and a lower ambient density. And because the H I layer is much thicker as well, shells are prevented from breaking out of the disk.

The age distribution of the holes (Fig. 7) also shows a clear dichotomy between the dwarfs and big spiral galaxies with no holes being found in spirals which are older than 30 Myr whereas the distribution in dwarfs is flatter and spans an age range of up to 120 Myr. This is not unexpected since in more massive galaxies shear and mixing due to the passage of a density wave passage tend to destroy holes. In dwarfs, unless a shell is hit by a neighbouring explosion, and subsequently expanding structure, a feature can potentially persist for a long time, until the rim disperses and mixes with the surrounding ISM.

Finally, the distribution of the expansion velocities is presented in Fig. 8. The predominantly low values for the dwarfs (especially Ho II) reflect, again, the fact that we see many large holes in the dwarfs which have almost stalled. A word of caution should be added as the velocity resolution of the IC 2574 and Ho II observations was a factor of about 3 better than those of M 31. Very young (i.e., smaller than ∼100 pc) and fast expanding holes are missed, of course, but this holds true for all galaxies in the sample.

Figure 9 displays an alternative way of looking at the properties of the H I holes on a galaxy-wide scale. It shows the expansion velocities of the shells in the four nearby galaxies plotted as a function of their diameter (crosses). Note that in those cases where no expansion velocity could be determined the velocity was set to 5 km s-1, slightly below the typical one-dimensional velocity dispersion in the quiescent ISM. Based on the hydrodynamical models for expanding shells (Chevalier 1974), we plotted lines of constant energy for the expanding structures in Fig. 9 (assuming a constant density of the ambient medium of 0.15 cm-3). The plotted curves represent energies at a level of 1049, 1050, 1051 and 1052 ergs, respectively (from left to right). The evolution of an individual hole of a given energy is along an equi-energy line from the top of the plot downwards. In addition, lines of equal ages are shown in the figure, ranging from 107 years (steepest slope) via 5·107 years to 108 years. The dashed line corresponds to the average age for each of the four galaxies. The mean values are: 15·106 years (M 31); 16·106 years (M 33); 47·106 years (IC 2574) and 69.0·106 years (Ho II).

[Click here to see Fig. 4-9!]

Ignoring the symbols plotted at 5 km s-1, there is a marked difference between the two spiral galaxies and the dwarfs. The H I shells in M 31 and M 33 are smaller and the range of expansion velocities is larger. In fact, when looking at the objects in order of decreasing mass (M 31 - M 33 - IC 2574 - Ho II) there seems to be a gradually shifting pattern. Whereas in the larger spirals there is an absence of older holes (older than a few ×107 years), holes in IC 2574 range between 5·107 and 108 years and holes in Ho II cluster near 108 years.

This suggests an alternative explanation for why we don't see younger holes in IC 2574 and Ho II. The SF histories of these dwarf galaxies could be quite different from the spirals. In the latter, the global SF rate is, to first order, constant with time. Therefore, there is always a population of young holes. In contrast, in gas-rich dwarf irregulars, SF occurs in bursts which transforms them into an H II or blue compact dwarf (BCD) galaxy. It might be that IC 2574 and Ho II have undergone a recent burst, perhaps triggered by a close passage with another member of the M 81 group, and are now in their post-starburst phase and that this is why we don't see many young holes. Our data alone don't allow us to choose between the two possibilities.

5. Conclusions

H I observations of sufficient angular and velocity resolution as well as sensitivity of dwarf irregular galaxies are now becoming available in the literature. Because dwarfs have several advantages over spiral galaxies (the absence of density waves, the absence of differential rotation and hence shear) we can expect a lot of progress to be made in our understanding of the structure of the ISM and how it is shaped by massive star formation and its aftermath. The following results seem to hold:
  1. Dwarf galaxies show a stunning amount of detail in the form of shells and holes in their interstellar medium. These features are similar to those found in large systems like our Galaxy, M 31, M 33, M 101 and NGC 6946. Often, the H I shells completely dominate the morphology of a dwarf galaxy, such as for IC 2574 or Ho II.
  2. Using two independent approaches, the thickness of the H I layer of dwarf galaxies is shown to be substantially larger than in large spiral galaxies. This facilitates the growth of H I shells. And as the H I layer is thicker, shells are contained within the H I layer up to much larger dimensions (say 1200 pc for dwarfs versus 600 pc for spirals).
  3. The radial expansion of the holes in dwarf galaxies, their indicative ages and energy requirements for their formation suggest that they have been created by the combined effects of stellar winds and multiple supernova explosions. Despite their large dimensions, the energy requirements are not extreme and infall of material such as clouds is not needed to explain the observed features.
  4. The shells can grow to these large dimensions because of several conditions which are fulfilled in dwarf galaxies. The volume density in the plane is low, which facilitates expansion. In the direction perpendicular to the disk, the gravitational pull is smaller than in a massive spiral. Also, because of the thick H I layer, shells are easily contained and unlikely to break out into the halo. Lastly, solid body rotation and a lack of spiral density waves prevent that holes are rapidly destroyed.
  5. A comparison with other galaxies with a range of Hubble types and studied in similar detail so far (M 31, M 33 and Ho II) shows that the size distribution of H I holes found in a galaxy is related to its Hubble type in the following way. The size of the largest H I shells is inversely proportional to the global gravitational potential (and hence mass surface density). The shells in dwarf galaxies show lower expansion velocities, most likely due to the fact that the shells are preserved much longer due to the absence of shear (differential rotation). The energies needed to create these structures, though, are found to be the same for all types of galaxies. The energy output of a typical starforming region is therefore not related to the host galaxy, at least to first order.
  6. If the conclusion that the energy output of a SF region is independent of the host galaxy holds then it follows that there is a lower mass limit for the formation of dwarf galaxies. In other words, if the mass of a galaxy falls below a certain threshold, the first burst of star formation will tear the whole galaxy apart (see also, e.g., Puche & Westpfahl 1994). Mac Low & Ferrara (1998) predict that this lower mass limit will be near 106 Msun, which is well below the mass of the objects dealt with here.
Acknowledgments. We should like to thank our colleagues, in particular You-Hua Chu, Deirdre Hunter, Mordecai-Marc Mac Low, Sally Oey, Jan Palous and Eric Wilcots for stimulating discussions and permission to use material in advance of publication. Both authors are indebted to the `Deutsche Forschungsgemeinschaft (DFG)' for financial support in the framework of the Graduiertenkolleg on ''The Magellanic Clouds and other Dwarf Galaxies'' without which none of their work would have been possible. In addition, EB acknowledges a grant awarded by CONACyT (grant number 0460P-E). Finally we thank Uli Klein for organising such a wonderful meeting and the editors for their patience.

References


Links (back/forward) to:

First version: 08thJuly,1998
Last update: 25thSeptember,1998

Jochen M. Braun   &   Tom Richtler
 (E-Mail: jbraun|richtler@astro.uni-bonn.de)