The presence of dust in the thick disks of galaxies has important implications for understanding the nature of extraplanar gas in spiral galaxies. Perhaps the most fundamental implication of significant amounts of dust in the thick disks of spiral galaxies is that much of the interstellar material in the thick disk in these systems has been expelled from the underlying thin disk rather than accreted from a reservoir of primordial material. This is strongly suggested by the large amounts of dust directly visible in our images. Furthermore, while not generally true, there are indeed structures connecting the extraplanar dust to the underlying thin disk (Figure 3). The precise mechanism for expelling gas and dust from the thin disk is not constrained, but some aspect of feedback from massive stars likely drives the expulsion.
Figure 3. Sections of the unsharp-masked V-band image of NGC 891 from Howk & Savage (2000); these regions are centered to the NE (left) and SW (right) of the galaxy center. The high-z structures marked in these images are clearly connected to the disk of the galaxy. The source of thick disk dust observed in this and other galaxies is expulsion from the thin disk. The connection to the thin disk is not unique to NGC 891, but is also seen in other galaxies (e.g., Thompson et al. 2004; Sofue et al. 1994).
The statement that much of the thick disk ISM must have been expelled from the thin disk does not necessarily constrain the nature of "halo" gas, material at very large distances from the plane. The maximum extent of dusty material observable through its optical absorption against background starlight is z ~ 2 kpc. This varies slightly among the observed galaxies and potentially with position in an individual galaxy. However, in the best studied case of NGC 891, there is sufficient light from the stellar bulge and thick disk that clumpy dust structures could have been seen to much larger heights in the central regions (Howk & Savage 2000).
The lack of detectable dust at larger heights has a few potential causes. Having argued the strongly-clumped dust at z 2 kpc is associated with a CNM, we have suggested that the interstellar pressures at larger heights may be insufficient to support a CNM (Howk & Savage 1999, 2000). In which case, the dust at these heights is associated with a diffuse medium with little in the way of small scale, high column density structures that would be detectable via direct optical imaging. Alternatively, there could simply be a lack of dust at high-z, either due to the smaller amount of gas at such heights or a changing gas-to-dust ratio.
While the current observations do not allow us to determine which of these scenarios is more likely, up-coming observations with the Spitzer Space Telescope and the Galaxy Evolution Explorer may help by revealing a smooth component of dust at z 2 kpc in galaxies. The distinction is potentially important, as it could bear on the amount of gas contributed to modern galaxy halos by on-going infall of primordial material and on the possibility that dust (and potentially gas) may escape a galaxy's potential altogether.
It is worth noting, also, that the presence of dust grains in the thick disks of galaxies implies that the mechanisms that transport material from the thin to thick disks are not sufficiently violent to completely destroy the dust. Thus, if feedback processes are responsible for expelling the dust and gas into the thick disks of galaxies, the characteristic velocities with which much of the present thick disk matter is expelled was likely too small to produce much dust destruction. This suggests that much of the extraplanar material may be simply displaced disk gas rather than material that started as shock-heated gas and metal-enriched supernova ejecta. The lifting of material through magnetically-driven mechanisms (e.g,. the Parker instability) or other more quiescent mechanisms may be important, although they are not necessary.
Acknowledgments. Thanks to my collaborators T.W.J. Thompson (UCSD) and B.D. Savage (Wisconsin) for their contributions to this work.