7. THE OVERALL OBSERVATIONAL PICTURE

Summarizing the observational results collected in the previous chapters, we can now draw a much more detailed picture of the halo ISM in spiral galaxies than only a few years ago and thereby put tighter constraints on competing theoretical models of the galactic ISM when checking their validity.

7.1 Which spiral galaxies have halos?

7.1.1 Search for energy sources

Currently, no sources of energy are known that might reside outside the galaxy disks and account for our observational data. As stated above, the decay of dark matter theory by Sciama (1990a, b) faces several problems explaining observations (Dettmar & Schulz 1992). The energy sources driving disk-halo interactions reside in the galaxy disks and in this context it is important to note that in a few cases (see Section 6) there is kinematical evidence that the direction of the bulk flow is indeed outward, ruling out a general accretion of matter from intergalactic space. This means that the halo gas, driven out or heated by these energy sources, must originally also come from the disk. There are two possible kinds of heating sources: AGNs (including galactic radio cores) and high-level SF (i.e., winds and SNe of high-mass stars).

AGNs: Many spirals host AGNs and in part of them outflow cones have been found. The base of some of these outflows in the central region is so narrow that they are clearly associated with the nuclear engine and not with possibly existing more extended starbursts (e.g., Wilson & Zvetanov 1994; Colbert et al. 1996). However, the galaxies listed in Table 1 do not have luminous AGNs. Therefore, AGNs cannot be the common cause for the observed disk-halo interactions, in particular in cases where disk-halo interactions are observed over large parts of the disks.

Radio cores: Radio cores play practically no role in the late-type spirals investigated here. Radio galaxies are normally elliptical and S0 galaxies (e.g., Zirbel 1996, and references therein).

High-level star formation: As stated in Section 6, SF can occur in different locations within galaxy disks, namely in circumnuclear starbursts and in giant H II regions further out in the disks. Individual H II regions can be very powerful, for example 30 Dor in the Large Magellanic Cloud, with L(H) 6.7 x 10³⁹ ergs s^-1 (Kennicutt et al. 1989). Therefore, both circumnuclear starbursts and widespread SF activity must be considered as potential energy sources for disk-halo interactions. In the following I will investigate in more detail the role of different sources of energy and the processes related to massive SF that are involved in heating the ISM.

In the optical, bright SF regions were detected near the nuclei of starburst galaxies by Sérsic & Pastoriza (1965), who dubbed them ``hot spots''. However, an observational problem in edge-on galaxies is that individual SF regions are often invisible in the optical due to the high optical depth for both blue continuum and H line emission (both of which are good tracers of massive stars). Kronberg et al. (1981, 1985) discovered radio continuum emission from a population of compact sources in the central region of M82, so-called ``radio knots''. The radio spectral indices of these sources indicate that many of them must be young compact SN remnants (SNRs), while others are H II regions. Subsequent high-resolution radio imaging resolved individual sources and verified the earlier result (Muxlow et al. 1994). Similar compact sources where subsequently discovered in other starburst galaxies, e.g., NGC 253 (Antonucci & Ulvestad 1988), NGC 1808 (Saikia et al. 1990; Collison et al. 1994), NGC 3628 (Carral et al. 1990), NGC 4736 (Duric & Dittmar 1988), and the central 4 kpc in NGC 4631 (Golla; in prep.). The surprising result is that many of the radio knots have flat spectral indices, indicating that the thermal emission of H II regions - and not that of the synchrotron radiation of SNRs - dominates (e.g., Golla; in prep.) or that synchrotron self-absorption must be important. M82, with mostly steep-spectrum radio sources, might be an exception rather than a ``typical'' starburst. In all galaxies mentioned above, the SF regions form (often ring-like) structures located near the turnover of galactic rotation (Lesch et al. 1990; see also Lehnert & Heckman 1996b). In these galaxies, this turnover is often close to or at the inner Lindblad resonance, if present (Combes 1987). The search for such energy sources in normal edge-on galaxies has only now begun. Dahlem et al. (in prep.) found about 10 compact ( 20 pc) radio sources in NGC 891.

As mentioned above, OB stars with their winds and subsequent SN explosions are the main sources of energy. The total energy output by stellar winds is as large as that of SNe, however the former is released over a longer timespan. In both cases, for an instantaneous starburst, Solar metallicity, a 1-100 M Salpeter initial mass function, and a total mass of the stellar cluster of 10⁶ M, the total energy lies in the 10⁵⁵ ergs range (Leitherer et al. 1992).

The observed dependence of halo morphology on the level of disk emissivity (DLG95; Rand 1996) establishes a link between these energy sources and the detected halo emission. This implies that at least part of the halo gas and the energy heating it are coming from SF processes in the disks of these galaxies. Another strong argument favoring SF-related processes as heating sources of the ISM in galaxy disks (and thereby the energy sources driving disk-halo interactions) is the detection of a hot gas phase with temperatures of up to ~ 10⁷ K. The cooling time for such gas is very short, requiring continuous heating in the disk, causing buoyancy of the disk gas and thereby leading to upward motions (Norman & Ikeuchi 1989, hereafter NI89, and references therein). Stellar winds and SNe were identified by Weaver et al. (1977) and Smith (1977) as the energy sources driving this process.

7.1.2 The morphology of gaseous halos

Observations so far show a clear trend that the shapes of outflows - and thereby the morphologies of gaseous halos - depend on the distribution of the energy sources in the underlying galaxy disks.

Based on halo morphology, one can distinguish three types of outflows driven by SF processes: central outflow cones, extended halos, and mixtures of both. These three correlate with the distribution of high-level SF in the galaxy disks. Cones are observed above circumnuclear starbursts (e.g., M82); the width of the base of the cone depends on the diameter of the starburst region. Extended halos are found above widespread SF in galaxy disks (e.g., NGC 891), and mixed activity leads to a superposition of the two types of outflows (e.g., NGC 4666). The outflow cones of starbursts are driven by the coordinated energy input of many SF regions in a small area, leading to so-called galactic ``superwinds'' (HAM90). Extended halos are fed by a multitude of individual outflows above H II regions that are far apart from each other (DLG95). Far above the disk of NGC 891 there are signs for the existence of a galactic ``wind'', which is fed by the individual small-scale outflows (Breitschwerdt et al. 1993; Dahlem et al. 1994b). Another piece of evidence for a direct dependence of halos on the SF activity in the disks is the result by Beuermann et al. (1985) based on radio continuum emission in our own Galaxy, which clearly indicates such a trend (Section 4.4).

The common denominator between the different outflow scenarios is that there is a minimum energy requirement for matter to escape from the thin disks, i.e., to reach the so-called ``break-out'' condition as defined by Ikeuchi (1988) and NI89. A certain minimum level of energy input into the ISM per unit surface area and unit time must be exceeded for disk-halo interactions to start. DLG95 gave first - although yet poorly constrained - limits for this energy threshold. Rand (1996) came to the same qualitiative conclusion based on H imagery (see Section 8.4 for more details).

This correspondence of the distribution of SF regions in the disks and the shapes of the halos, together with kinematic evidence in a few cases for outflowing gas, is the strongest argument in favor of the interpretation that energy sources in SF regions in the galaxy disks are driving the observed disk-halo interactions.

7.1.3 Why only late-type spirals?

The question arises which types of spiral galaxies have disk-halo interactions and in which respect these galaxies differ from others.

Most early-type spirals (Hubble types S0 to Sab) have little gas (e.g., Eder et al. 1991) and/or are dynamically ``hot'', meaning that the velocity dispersion of their ISM is high. Thus, gas condensation into cold dense clouds is suppressed (e.g., Lees et al. 1991). The FIR-radio continuum correlation in elliptical and S0 galaxies is similar to that observed in late-type spirals (Walsh et al. 1989), suggesting that dust in early-type galaxies is also heated by SF processes. However, the SF levels deduced from FIR data are much lower than in spirals (their Figures 5 and 9; see also Soifer et al. 1987a; Dwek & Arendt 1992).

The kinematics and dynamics of late-type spirals (Sb - Sdm) are different. The gas disks are more relaxed, enabling the condensation of gas into giant molecular clouds (GMCs; Kennicutt & Chu 1988), which are the birthplaces of massive stars. Thus, only galaxies of type Sb or later host giant extragalactic H II regions (GEHRs; Kennicutt 1984; Kennicutt et al. 1989). As noted above, the distribution of these GEHRs can vary (starburst vs. widespread SF). However, individual (complexes of) GEHRs appear to be able to drive disk-halo interactions, as e.g., suggested by the correlation of the sites of radio knots with gaps and filaments in the radio halo of M82 (Reuter et al. 1994), the association of the most powerful SF region and the brightest radio continuum spur in the halo of NGC 4631 (Golla & Hummel 1994), and the detection of a local outflow above a very bright GEHR in the outer disk of NGC 4666 (Dahlem et al. 1997). This suggests that high-level SF in GEHRs creates overpressured hot ionized gas through the conversion of kinetic energy (from stellar winds and SNe) into heat (via shocks). The gas then convects out of the thin disk along the steepest pressure gradient (i.e., perpendicular to the disk).

The above results indicate that not every galaxy, and not every H II region within a galaxy, produces enough energy to initiate disk-halo interactions. This leads to a whole complex of questions, which will be addressed in the following sections:

(1)	How much energy is required to initiate an outflow from the thin disk of a galaxy?
(2)	Can the sources detected in galaxy disks provide sufficient energy?
(3)	Via which mechanism(s) is the heating facilitated?
(4)	Is the gas heated in the disk and then transported into the halo or is gas in the halo heated by energy sources in the underlying disk?