2. EXTRAPLANAR DUST IN SPIRAL GALAXIES

It is now established that at least ~ 40% of spiral galaxies in the local Universe show some form of extraplanar (thick disk) dust and, hence, gas (Rossa & Dettmar 2003; Howk & Savage 1999). The most common technique currently employed for identifying extraplanar dust is to image directly edge-on spiral galaxies in the optical, searching for obvious evidence of shadowing of the thick disk, halo, and bulge stars by optically-thick foreground dust. Figure 1 gives an example of this approach, showing two views of the V-band image of NGC 891 from Howk & Savage (2000). The presence of dust in the thick disk of this galaxy is obvious from these images.

Figure 1. Two versions of a broad-band optical (V-band) image of the edge-on galaxy NGC 891 from Howk & Savage (2000). The top panel shows the direct V-band image. The bottom panel shows an unsharp-masked version of the V-band image. The latter is produced by smoothing the original image and dividing the original by this smoothed version. The purpose is to remove large-scale surface brightness variations, e.g., due to the vertically-decaying light from the stellar disk or from the bulge of the galaxy.

From the stand-point of total exposure time required, such broadband imaging is by far the most efficient manner of finding extraplanar material in galaxies (although it does require high-resolution - 1" - imaging). The presence of extraplanar dust is likely a flag that a significant amount of extraplanar gas is present, as well. Not only is there likely to be a significant amount of gas associated with the dust structures detected through direct optical imaging, but the surveys of Rossa & Dettmar (2003) and Howk & Savage (1999) have shown that ~ 90% of galaxies exhibiting extraplanar dust also have extraplanar diffuse ionized gas (DIG) detectable through H emission.

While it is straightforward to note the presence of extraplanar dust from images such as that shown in Figure 1, there are several biases inherent to this approach. As discussed in Howk & Savage (2000) and more recently in Thompson et al. (2004), we are only able to detect dusty regions through direct optical imaging because of they have significantly lower surface brightness than their surroundings. This implies that the dust-bearing clouds seen in our images must have a higher column density of dust (and gas) than their surroundings: a smooth distribution of dust would produce no contrast and would be undetectable in these images. In practice this likely implies the dusty clouds are more dense than their surroundings (assuming the thick disk gas has a relatively uniform dust content). A number of other, less scientifically interesting, biases should be considered when looking at images such as those in Figure 1. The requirement of large contrast tends to bias us toward detecting dust on the near side of galaxies, and, due to signal-to-noise constraints, we are more likely to detect high-z dust in regions of intrinsically higher surface brightness (e.g., against the bright light of a galactic bulge or at lower heights above the midplane).

Using estimates of the "apparent extinctions" produced by individual thick disk dust clouds, Howk & Savage (1997, 1999, 2000) and Thompson et al. (2004) have estimated physical properties of these clouds, albeit crudely. Because the apparent extinctions will always underestimate the true extinctions through the clouds (see Howk & Savage 1997), all of the physical quantities derived using the apparent extinctions are lower limits. Assuming the dust-to-gas ratio in these clouds is similar to that found in the disk of the Milky Way (which is probably not too bad an assumption; Thompson et al. 2004), the dusty cloud complexes detected in our images must have N(HI) 10²⁰ cm^-2. Furthermore, the densities in these clouds must be quite high. Examining the apparent extinctions and sizes of the smallest structures in the extraplanar cloud complexes in NGC 891 as seen in new images from the Advanced Camera for Surveys on board the Hubble Space Telescope suggests the densities in these clouds may be n_H 25 cm^-3.

The combination of these estimated column densities and the projected sizes of the cloud complexes in our images imply total masses of ~ 10⁴ to 10⁵ M or higher in each complex. Such masses are consistent with those of the individual giant molecular clouds in the Milky Way. In galaxies with detectable extraplanar dust, we typically find hundreds of absorbing structures at z 2 kpc all along the central regions of the disk (within R 8 kpc, similar to the radial extent of detectable DIG). Howk & Savage (1997) estimated the total mass of the ensemble of clouds in NGC 891 to be ~ 10⁸ M, comparable to the total mass of extraplanar DIG material (Dettmar 1990).

Morphologically, the extraplanar dust structures seen in direct optical images are quite complex and varied. Much of the complexity is likely due to the observed clouds residing at different depths through a galaxy. The identification of cloud complexes which may represent coherent structures is extremely difficult at heights z 1 kpc from the midplanes of spirals, particularly toward the centers of galaxies where more structures may be present along a given sight line. In fact, we believe that sight lines through an edge-on galaxy with extraplanar dust are typically optically thick for heights z 1 kpc. At such heights, every sight line intercepts at least one dust-bearing cloud, each of which we believe to have A_V > 1. At larger heights, where the confusion is significantly lessened, it is possible to identify what appear to be individual, sometimes isolated clouds or structures. As an example of this, Figure 2 shows a portion of our HST image of NGC 4217 (Thompson et al. 2004). Thompson et al. note the presence of a structure that appears to be a large loop (marked in Figure 2) with diameter ~ 800 pc centered at z ~ 1300 pc from the midplane of this galaxy. Such a structure would be confused at lower heights from this or other galaxies. Even at the height of this structure, though, one worries that what appears to be a loop may be caused by several overlapping, but unrelated, cloud complexes.

Figure 2. Two versions of a broad-band optical (B-band) image of the edge-on galaxy NGC 4217 from Thompson et al. (2004) taken with the Wide Field and Planetary Camera 2 on board HST. The top panel shows the direct B-band image. The bottom panel shows an unsharp-masked version of the B-band image. The large loop discussed in the text is marked with an arrow.