Many mechanisms have been proposed for creating the holes and shells that have been observed in several nearby galaxies. Although there still are major difficulties, most studies put star formation forward as the most likely explanation. In some of the nearest galaxies (M 31, M 33, SMC), where individual OB associations can be detected, a correlation is found between these associations and small (0 200–300 pc) HIholes (Brinks & Bajaja 1986; Deul & Den Hartog 1990; Hatzidimitriou et al. 2005).
For large holes and shells no correlation is seen and the holes appear empty. Also the HIholes of NGC 6946 generally show no bright stellar components. Furthermore, in the 21-cm radio continuum, several holes are seen coinciding with those in HI. The same is true for the FIR, observed by ISO (Contursi et al. 2002). This suggests that the HIholes are really devoid of gas and dust.
During their formation, the interiors of the bubbles are thought to contain hot gas, heated by the SNe and stellar winds, that should be observable as X-ray emission.
Chandra observations (Schlegel, Holte & Petre 2003), indeed, show diffuse X-ray emission toward the regions with the most massive-star formation (traced by the largest HII complexes, see their Fig. 3). Only a few regions with X-ray emission clearly coincide with HIholes in our sample. Those holes are also the few cases that coincide with bright star clusters.
A prototype hole is no. 51. It is a spherical hole in HIwith a diameter of about 800 pc (Fig. 4.12, top left). In the same direction an H bubble is detected, which seems to fill the HIcavity (Fig. 4.12, top right). At the western rim, one or more OB associations are seen surrounded by a bright HIIcomplex. The HIkinematics show a two sided spur which seems to be centred on the HIIcomplex (Fig. 4.12, bottom panels). This can be interpreted as an outflow powered by stellar winds and SNe in the OB associations. The X-ray emission from Schlegel et al. coincides with the H bubble, which indicates a hot interior. The presence of the H bubble, the X-ray emission, as well as the small size and spherical shape indicate that the HIhole is still relatively young. Our estimate is 1.9 107yr.
The most extended region where stellar activity seems to coincide with a group of HI holes is the north-eastern spiral arm, about 3 from the nucleus. Especially in B-band this spiral arm seems thicker than the others, which was reason for Arp to include NGC 6946 in the Atlas of Peculiar Galaxies (Arp 1966). Also in HI the spiral arm is massive, but appears split by a group of HIholes (see Fig. 4.2). Inside the holes, large clusters of blue stars are seen. The high contrast and colour differ-ence with its surroundings could, however, be the effect of extinction. Nevertheless, stellar activity is present, which can be related to the formation of these holes. Dif-fuse soft X-rays are detected over the entire spiral arm, as is difDif-fuse H emission (Ferguson, Gallagher & Wyse 1998). This means that hot gas is still present. It is unclear, however, whether this gas is still in the disk inside the holes, or in the halo.
Figure 4.19 shows that there is high-velocity gas in the direction of the holes. The velocity structure of thisanomalousHIcan be seen in Fig. 3.22 (bottom panels). The high-velocity gas is only seen at the low-Vrotside, which is not direct support for ver-tical motions. However, deviating velocities are observed up to 200 km s 1, which areforbiddenvelocities at this position. These cannot be explained by HIrotating in a lagging halo. Furthermore, the region withanomalousHI(Fig. 3.22, bottom right) extends as far as the holes and HIIregions in this spiral arm (Fig. 3.22, bottom left).
Therefore, it is likely that we observe here HIthat is being blown into the halo. The given examples show that for some holes there seems to be a connection with stellar activity.
We could also calculate quantitatively if it is possible to create the holes by super-novae and stellar winds alone. Estimated from the velocity dispersion of the ISM, holes can last for several 107years. If they are, as suggested, formed by SNe then the holes can be interpreted as tracers for the star formation of the last few 107yr. The age histogram in Fig. 4.4c would then suggest that there was a burst about 2–3 107yr ago. However, as we already pointed out, this distribution is strongly affected by selection effects. Taking this into account, the star formation rate could as well have been constant for these the last few 107yr. This is what we assume in the following
DISCUSSION 143
Figure 4.12– Example of a hole coinciding with an H bubble and HIoutflow next to it. The top left panel shows the channel (v< 8 km s 1) in which the hole is best seen (greyscale). The contours show the location of the H emission, at levels 2, 4, 8, 16, 32, and 64 10 16erg s 1 cm 2arcsec 2. the beam is indicated by the shaded ellipse. The top right panel shows the H emission in greyscale. The bottom panels are xv-diagrams along the dashed lines in the top left panel. The crosses indicate the resolution.
calculations.
From the H observations and calculations by Degioia-Eastwood et al. (1984), we estimate the star-formation rate for the inner disk of NGC 6946 to be 7 M yr 1, although this number depends strongly on assumptions about foreground extinction and absorption in NGC 6946 itself. Assuming a Salpeter IMF and that all stars with a mass above 8 M end their lives in a core-collapsed supernova explosion, the SN type II rate in NGC 6946 is about 0.05 SN yr 1. The rate of SN type Ia is about the same as that of the type II (Tammann 1977), giving a total rate of about 0.1 SN yr 1. Interestingly, this estimate is consistent with the present, observed SN rate. Last century, 8 SNe have been recorded in NGC 6946, which is the highest number ever
observed in a galaxy.
The energy input from a single supernova is generally taken to be 5 1050erg for a SN type Ia and 1051erg for a core-collapsed SN (Chevalier 1977). In the formation time of 4 107yrs, 4 106 SNe would have occurred involving a total energy of 3 1057erg. The ratio in energy input from supernovae to that from stellar winds is about 3:1 (Abbott 1982; van Buren 1985; Rosen & Bregman 1995), which gives a total energy release from the massive stars in NGC 6946 of 4 1057 erg. On the other side of the balance, our estimated total energy needed to create the holes is of the order of 1.8 1056erg (using formula 4.1). Even if we take into account the large number of small holes that we have missed, the HIholes could well have been formed by the energy input from constant high massive-star formation in the disk.
The time scales and the energy budget seem right for the stellar feedback to pro-duce the holes. However, it is still a puzzle that we observe many HI holes with-out progenitor remnants. If a hole was formed by 1000 SNe, an over-density of the lower mass stars that formed together with the massive SNe-progenitors would be expected: 6000 upper main sequence stars (late B, A, and F) should remain after 108 yr (Rhode et al. 1999). After that time, the clusters will not have dispersed signifi-cantly and they should be observable as blue point sources inside the holes. Some holes in the inner disk, such as those in the north-eastern arm, do seem to show an over-density of stars in B and V (for the latter see Fig 4.3), but this could also well be an extinction effect. As already noted, many of the holes are also clearly devoid of the radio continuum and infrared emissions. If there is no dust in those holes, the extinction in those regions is expected to be much lower compared to the surround-ings.
Evolution of spiral structure
In NGC 6946 most of the holes are seen in the spiral arms or regions with high HI
column densities. This could be a selection effect, since one needs high enough HI
contrast to identify a hole. Furthermore, in the inner disk the spiral structure is fila-mentary and poorly defined, which makes it difficult to trace the arm and interarm regions. Here the situation is confusing: The holes themselves appear to disrupt the spiral structure. Many spiral arms appear split along their ridge by a chain of holes.
The formation of holes, together with the differential galactic rotation may have a strong effect on the evolution of the gaseous spiral pattern. Simulations (e.g. Gerrit-sen & Icke 1997; Bottema 2003; Pelupessy 2005; Cox et al. 2006) of galaxy disks show that high SN feedback can destroy the coherent spiral structure of the gas. The cor-responding picture in Fig. 5 of Bottema (2003) looks very similar to the filamentary spiral pattern in NGC 6946. This is another indication that SNe may be important for the structure and evolution of the disk of NGC 6946.
Self propagating star formation
Often, HIIcomplexes are seen at the rims of the HIholes in NGC 6946 (Figs 4.13 and 4.14). These may have formed due to the expansion of the shells, which compresses its surroundings. Such a causal connection is, however, hard to prove, because the HIIcomplexes in NGC 6946 are always found in regions of high HIcolumn density
DISCUSSION 145
Figure 4.13– The HIholes plotted on top of an H image from Ferguson, Gallagher & Wyse (1998).
as are the shells of HIholes. Nevertheless, cases such as the arc of HIIregions south of hole no. 107 (Fig 4.14, top panels) are very suggestive for this causal interpretation.
If such a hole-driven star formation (e.g. Elmegreen & Lada 1977) is indeed oc-curing in NGC 6946, the hole formation may be a self-propagating process.
Holes and star formation in other galaxies
How does the population of holes found in NGC 6946 compare to those in other galaxies? First, the hole catalogues of other galaxies consist of many holes with sizes smaller than a few 100 kpc. Only a few are larger than a few kpc. The smallest hole we could detect in NGC 6946 is 770 pc in diameter, and we estimate that we may have missed about 250 HIholes with sizes down to our spatial resolution. All the galaxies with small holes are at smaller distances than NGC 6946. On the other
Figure 4.14– HIIregions at the rims of HIholes. The left panels show the total HIdensity distribution around some holes. The ellipses show the derived size and orientation of the HI
holes. The beam is shown in the bottom left corner. The right panels show the same regions in H , with the dashed ellipses outlining the HIholes.
DISCUSSION 147
Figure 4.15– The average size of the 5 largest HIholes plotted against the star formation rate (calculated from the IRAS far-infrared flux) per kpc 2.
hand, there may be another selection effect preventing the detection of large holes in very nearby galaxies. To test this, we have smoothed the HIdata of M 33 by Deul
& van der Hulst (1987) to the same spatial resolution as our data of NGC 6946. We have found two large holes of 0.9 and 1.5 kpc which were not included in the hole catalogue by Deul & Den Hartog (1990). At close distances the substructure inside the largest holes becomes resolved and the hole itself is difficult to identify as such.
Even so, the number of very large holes in M 33 is small compared to NGC 6946.
If the holes are indeed formed by stellar processes, the difference in hole sizes may be connected to observable differences in the star formation in each galaxy. The only galaxy known to surpass NGC 6946 in average hole sizes is M 101. Kamphuis (1993) has found more than 50 holes ranging in size from 0.8 to 5 kpc, and a median size of about 2 kpc. Even though these holes are on average twice as large as those in NGC 6946, the SFR in M 101 is of the same order as that of NGC 6946. Furthermore, if we compare the SF density by dividing the SFR by the size of the optical disk (defined by R25), the SF density in M kpc 2of NGC 6946 is ten times higher than that of M 101, while the holes in M 101 are larger (Fig. 4.15, see for calculations of
0 D, 5 and SFRIR below). Probably, the average amount of star formation is not directly related to the average hole size, as the formation of holes is a local process.
The compactness of the star formation could be a better measure. For the latter, we should look at the sizes of the HIIcomplexes. For example, M 101 is known to harbour the giant HIIcomplexes NGC 5471, NGC 5455, and NGC 5462. These star formation complexes have diameters of a few hundred pc. The number of candidate
Figure 4.16– The average size of the 5 largest HIholes plotted against the average size of the 3 largest HIIregions per galaxy.
SNe in a single HIIcomplex of that size is probably sufficient to create a kpc-size hole. Perhaps, the sizes of the largest HIIregions determine the sizes of the largest possible HIholes.
To test this, we have compared the sizes of the largest HIIregions with the di-ameters of the largest HIholes in each catalogue. For the diameters of the HII com-plexes we adopted the measurements by Sandage & Tammann (1974a,b,c), which they used to calibrate their distance scale. For a large collection of nearby galaxies they measured and averaged the sizes of the 3 largest HIIregions. In the same way, we averaged the diameters of the 5 largest HIholes (a few more than 3 for slightly better number statistics) from the published HIholes catalogues. For NGC 4559 no such catalogue exists. Therefore, we examined the HIdata from Barbieri et al. (2005) and determined the 5 largest holes ourselves. The result is shown in Fig. 4.16. At first sight, there seems to be the expected trend, where galaxies with giant HIIcomplexes also show large HIholes. M 101 clearly stands out in both sizes. The relation is, how-ever, less clear for intermediate size HIIcomplexes. The HIIcomplexes in NGC 6946 are about the same size as those in NGC 2403 and M 33, and yet the largest HIholes in NGC 6946 are about 3 times larger (this changes only slightly when including the two newly discovered HI holes in M 33). Probably, the number statistics are still poor. A more reliable, but also much more difficult, comparison would be to com-pare the complete observed size distribution of HIholes with that of a complete size distribution of HIIregions.
The size of the largest holes does seem to scale with distance as expected from
DISCUSSION 149
Figure 4.17– The average size of the 5 largest HIholes plotted against the distance of the galaxy.
the selection effects (Fig. 4.17), but also here the scatter is large. M 101 still stands out in average HIhole size.
We also compared the size of the HIIcomplexes with the global star-formation rate in this sample of galaxies. The latter, however, cannot be determined very pre-cisely. Published measurements of the SFR in NGC 6946, for example, easily differ by a factor 2–3. To have at least a reliable relative SFR for these galaxies, we used the far-infrared flux at 60 and 100 m as detected by IRAS (taken from Rice et al.
1988; Moshir et al. 1990). We used the method as described by Kewley et al. (2002) to calculate the SFR in M yr 1.
The most remarkable point in Fig. 4.18 is that of NGC 6946. In comparison with all other galaxies, NGC 6946 has HIIcomplexes that are too small for its overall SFR, if assumed a trend of increasing complex-size with SFR. Also M 101 stands out, but it still has both large HIIcomplexes and a large SFR. From this simple exercise we can conclude that the sizes of HIholes are difficult to predict from the star formation properties of a galaxy.