4 Results

   
4.1 Prevalence of dust chimneys

Our primary objectives in this investigation are to ascertain whether dust chimneys are commonplace within quiescent galaxies and how, if at all, they might relate to recent star-formation within the disk. We begin by addressing the first of these issues.

Unfortunately, there are only 5 galaxies, out of the original sample, which are orientated sufficiently close to edge-on ( $i>87^{\circ}$) to carry out an unambiguous search for extraplanar structures. This subset is too small to result in any unequivocal statements about the overall occurence of dust chimneys. Of the 5 galaxies depicted in Figs. 1 to 6, 3 harbour very chaotic dust lanes (NGC 891, NGC 4013, NGC 4302) with numerous curvi-linear structures extending up to 2 kpc from the midplane. Many of the extinction features extend to a kpc or so above the midplane before arching back towards the disk. Others show no sign of reconnecting with the main dust layer and appear to fragment at maximum z-height. More spherically-shaped dust clouds are also recognizable, usually at $z\simeq2$ kpc, and appear completely isolated from the disk. The maximum z-height at which extraplanar dust is discernible is about the same in all 3 galaxies (z=1.7 kpc). (However, we might expect that diminishing bulge light would preclude a detection of extinction features at much greater z-heights). Of the 2 remaining galaxies in Figs. 1 to 6, UGC 4278 appears to contain only small amounts of extraplanar dust and NGC 4762 none at all. In the HS99 study, 5 galaxies are labelled as manifesting chimneys (out of a total of 7). Three of these objects appear in our final sample (NGC 891, NGC 4302, NGC 4013) but the other two (NGC 4217 and NGC 5907) we have already rejected as not being sufficiently edge-on to make any unambiguous statements about extraplanar dust. It is interesting to note that NGC 4217 (and possibly NGC 5746 from our original sample) both exhibit some filament-like features which extend to larger (apparent) z-heights than can probably be accounted for by projection of disk structure onto the plane of the sky. However, in the strictest sense (using the $i>87^{\circ}$ criterion), the HS99 study and the present work indicate collectively that 5 out of 10 spirals possess dust chimneys. Whilst the current statistics are small, we concur with HS99 that the "chaotic" dust lane in NGC 891 can no longer be considered anomalous and, indeed, high-lattitude dust appears to occur in half of all quiescent spiral galaxies.

Figure 6: Unsharp-mask image of NGC 4762 in a blue filter

The detection of structures both below and above the midplane, argues against warps being the cause of the chimney phenomenon. However, it is less easy to dismiss the possibility of flaring in the outer regions of the disk being a contributory factor. Optical images of spirals more inclined to the line-of-sight, such as NGC 253 ( $i=78^{\circ}$), do exhibit numerous protuberances from the main dust lane and, moreover, at various distances from the nucleus (Sofue et al. 1994). In Fig. 7, we highlight similar features evident in two other slightly-inclined spirals that have recently come to our attention (NGC 2976 and NGC 4928). Although line-of-sight ambiguities would inhibit us from labelling these particular features as definitely extraplanar, their similarity to the vertical structures depicted in Figs. 1 to 6 is striking. The fact that such structures are not concentrated at the edge of the disk, appears to confirm that flaring cannot be the mechanism producing dust chimneys. HS97 have also presented cogent arguments, based on velocity information for the gas in NGC 891 and its relationship to the dust chimneys, as to why gas-flaring cannot be the reason why chimneys are observed in this galaxy.

In Table 4, we present the mean optical depths of extraplanar dust features found in our final galaxy sample. We also include total masses for dust and gas residing outside the main disk. The gas mass is calculated assuming a gas-to-dust ratio of 150 i.e. appropriate to gas with a solar-type metal abundance (Whittet 1992). The typical optical depth for the extinction features identified in Figs. 1 to 6 is estimated as $\tau_{B}=0.1-0.2$. Our poor resolution of the features, combined with the neglect of scattering effects, means that we could have underestimated the true column density of dust by at least a factor of 5. Thus an optical depth closer to unity is probably more likely. The extraplanar gas associated with the chimneys is $\sim$ 1% of the total HI+H₂ mass, assuming a canonical mixture of gas and dust in these structures. We note, however, that if grains are expelled from the disk by radiation pressure rather than convective flow (Sect. 6.4), dust and gas might be effectively decoupled from one another. Under such circumstances the amount of gas in the extraplanar layer may be considerably different to our estimate.

 

  Table 4: Properties of the extraplanar dust structures identified in the final, edge-on sample. $<\tau _{B}>$ is the mean B-band opacity of the individual features, whilst $M_{\rm d}$ and $M_{\rm g}$ represent the total dust and gas masses found outside the main disk. $M_{\rm d}$ and $M_{\rm g}$ have been calculated assuming Galactic-type grains are present according to a solar gas-to-dust ratio. $M_{\rm G}$ is the total gas mass of the disk (HI+H₂). Except for NGC 891, where atomic and molecular gas masses are available in the literature (Scoville et al. 1993), we use 21 cm line fluxes in the RC3 (de Vaucouleurs et al. 1991) to obtain the HI mass and assume that an equal amount of gas is present in molecular form (Devereux & Young 1990). $<\tau _{B}>$ severely underestimates the true chimney opacity (probably by a factor of 5) because the width of the features are comparable to the PSF of the observations. The distances listed in Table 1 have been used in order to obtain the dust and gas masses

Galaxy	$<\tau _{B}>$	$M_{\rm d}$	$M_{\rm g}$	$M_{\rm G}$	$M_{\rm g}$/$M_{\rm G}$
		(105 $M_{\odot}$)	(107 $M_{\odot}$)	(109 $M_{\odot}$)	($\times$ 100%)
NGC 891	0.16	2.1	3.1	9.8	0.3
NGC 4013	0.11	0.93	1.4	4.8	0.3
NGC 4302	0.19	2.6	3.9	4.0	1.0
UGC 4278	0.071	0.047	0.071	2.8	0.025
NGC 4762	0.00	0.0	0.0	0.17	0.0

HS99 do not appear to list total masses for the extraplanar structures identified in their larger edge-on galaxy survey (although they give masses for individual, selected features). For NGC 891 (HS97), an estimate of $\sim$ 2% of the neutral ISM is cited, which is significantly higher than the value of 0.3% we give in Table 4. The HS97 estimate relies on 12 well-studied features epitomizing all 120 structure identified in their investigation. Thus some of the discrepancy may arise from an extrapolation of 12 to 120 objects. However, it is also likely that the current work is less sensitive than the HS97 study - our larger PSF means that we ultimately overlook both smaller and less optically-thick high-z clouds. We view the extraplanar masses listed in Table 4 very much as lower limits. In 5.1 we analyse the thermal emission from NGC 891 in the submm regime in order to obtain an upper limit on the amount of dust and gas situated outside the disk.

   
4.2 Potential energy of chimneys

We derived the potential energy of the extraplanar structures, identified in the previous section, using a numerical simulation based on an exponential disk of stars and a spherical dark matter halo. In the first instance we concentrate on NGC 891 for which the vertical features are best resolved. Using the dust free photometry from the Xilouris radiation transfer model, we infer a central luminosity density ( $5.5 \ 10^{9}$$L_{\odot}$ kpc^-3), a radial scale-length (3.9 kpc) and an exponential scale-height (0.35 kpc) for the stars. The stellar luminosity density is converted to a mass density assuming a mass-to-light ratio of unity. Although the rotation curve of NGC 891 has been investigated quite closely (Bahcall 1983), the dark matter halo can be described by a variety of models and is not well constrained in this case. Thus we refer to a fit by van Albada et al. (1985) to the rotation curve of NGC 3198 - a galaxy similar in type, size and total magnitude to NGC 891. Following their inference of how dark matter scales with luminous matter, the dark matter radial mass density was prescribed as follows:

$$\rho(r) = \frac{ 5.7\ 10^{7} } { 1+ (r/11.4)^{2.1} }$$ (3)

where r is the galactic-centric radius, in kpc, and $\rho(r)$ constitutes the mass density in $M_{\odot}$ kpc^-3.

A catalogue was created of the 60 or so extraplanar dust features in NGC 891, containing the relevant optical depth, mass, height above the plane and radial distance. We allow, as we did previously, for the likelihood that features seen towards the nucleus are situated on the nearside of the disk. Thus, we set $r=0.5 \times R_{25}$ for features viewed towards the centre, r=R₂₅ for structures identified at the maximum distance along the major axis and interpolated values of r for the features between these extremes. The catalogue records are fed into the numerical simulation which calculates, stepwise, the amount of energy that is required to lift any particular feature above the relevant part of the plane. In Table 5, we show entries from the catalogue, including the derived potential energies, for the 4 extraplanar features highlighted in Fig. 1.

 

  Table 5: An extract from a catalogue of extraplanar dust structures in NGC 891. The features refer to those structures annotated in Fig. 1. We have computed the potential energy of the features using a numerical model to describe the stellar disk and dark matter halo

Feature	Mean z-height	Radius	Mass	P.E.
	(kpc)	(kpc)	$10^{5}M_{\odot}$	(1051ergs)
A	0.85	9.6	8.2	87
B	0.73	9.9	4.8	39
C	1.5	10.5	1.6	31
D	0.92	9.9	1.1	13

Figure 7: B-band images of the inclined galaxies NGC 4928 (top) and NGC 2976 (bottom). The arrows indicate possible chimney structures. Such features occur at a range of distances from the nucleus and do not appear to be associated with flaring towards the edge of the gas disk. CCD frames for these objects were obtained from the ING telescope archive (1200 s JKT exposure for NGC 2976 and 1000 s INT exposure for NGC 4928)

Figure 8: Unsharp-mask images of the galaxies shown in Fig. 7

We derive a total potential energy of $3.9\ 10^{54}$ ergs for all the extraplanar structures in NGC 891 with a mean value of $6.4 \ 10^{52}$ ergs per structure. The energies are similar to the values found by HS97 but this is coincidental because these authors use a plane-parallel model with a much lower mass density in the midplane (0.185 10⁹ $M_{\odot}$ kpc^-3) but a considerably larger stellar scale-height (0.7 kpc). Given that one of our key objectives will be to relate star-formation in the disk with the occurence of dust chimneys, we compare briefly now the derived potential energy with the anticipated injection of energy from both supernovae and the HII radiation field. As we show in Sect. 6.4, the timescale for the formation of dust chimneys is at least $5 \ 10^{6}$ yr. Thus, for a supernova rate of 0.01 yr^-1 and an injected energy of 10⁵¹ ergs per supernova (Leitherer et al. 1992), only 8% at most of the supernova energy needs to do work in lifting the chimneys out of the plane. We estimate the power in the HII radiation field from the FIR luminosity, $L_{\rm FIR}$, which we assume to be reprocessed optical and ultraviolet energy. Devereux & Young (1993) have shown that $L_{\rm FIR}$ correlates spatially with H${\alpha }$ emission and sites of massive star formation in nearby disks (although perhaps 50% may arise from the diffuse interstellar medium in some spiral types; Walterbos & Greenawalt 1996). Thus, using $L_{\rm FIR} \sim 10^{10}$ $L_{\odot}$ to gauge the radiation energy impinging on grains situated in HII regions, we estimate that only 0.1% of this energy must be transferred into lifting dust out of the plane. We have assumed here that the photolevitated grains can successfully transfer their upward momentum to the gas during collisions - a situation which is reached relatively quickly in numerical simulations of radiation pressure (Davies et al. 1998). We can conclude, therefore, that both supernovae and radiation pressure provide viable means for the creation of dust chimneys above spiral disks. As pointed out by the referee, however, supernova explosions provide a more natural explanation for dust chimneys because their energy is focussed directly into the base of the chimney structure.

   
Comparison with H${\alpha }$ images

Within several nearby, starburst galaxies, such as M 82, an association has been observed between kpc-scale dust outflows and concentrations of recent star-formation in the central disk. Hot gas, pushed out along the minor axis by massive stars and supernova explosions, generates a diffuse nebula of H${\alpha }$ shells and streamers as it collides with ambient gas in the halo. At the same time, dust and neutral gas from the disk are entrained at the working surface of the expansion so that a network of optically-thick filaments becomes established around the H${\alpha }$ superbubble (Ichikawa et al. 1994; Phillips 1993). In this section, we search for an analogous effect in quiescent disks. We examine the distribution of H${\alpha }$ emission from our sample and attempt to relate it to the occurence of dust chimneys. R96 has already presented a detailed description of the H${\alpha }$ emission from our edge-on galaxies. In addition to bright emission from HII regions, a layer of diffuse ionized gas (DIG) is found in several of the objects. When present, the DIG often extends to several kpc above the disk and possesses an emission measure, perpendicular to the disk, which is comparable to the Reynolds layer in the Galaxy (Reynolds 1992; Kulkarni & Heiles 1988). Brightest DIG emission tends to occur above regions undergoing enhanced star-formation but, as we discuss later, the mechanism by which the emission-line nebula becomes established above quiescent disks may differ in some ways from the starburst superwind phenomenon.

To compare the distribution of extraplanar dust with H${\alpha }$ emission from both the disk and the DIG layer, we employ two methods. Firstly, we superimpose the unsharp-mask images in Figs. 1 to 6 onto the relevant H${\alpha }$ image so that a simple visual inspection can be made. Secondly, we produce a profile by integrating, in z-height, both the extraplanar dust mass and the H${\alpha }$ emission at each position along the major axis. An overlay of the unsharp-mask image over the emission-line frame reveals no obvious correspondance between either (i) the distribution of extraplanar dust and HII regions in the disk, or (ii) the presence of extraplanar dust and an increased DIG brightness. Figures 9 and 10 illustrate the case for NGC 891 and NGC 4302. We find that H${\alpha }$ emission from the extraplanar layer exhibits a far more diffuse distribution than that of the dust chimneys. Overall, the extraplanar ionized gas is far less likely to form discrete structures which might be matched to individual extinction features. Certainly, no parallel can be found with starburst galaxies where dust appears to collect at the working surface of hot gas expelled from the disk.

Figure 9: At the top, H${\alpha }$ emission superimposed as contours onto unsharp-mask B-band image of NGC 891. Isophotes begin at 23 EM and increase in steps of 0.5 mag (1 EM = $2 \ 10^{-18}$ ergs cm-2 s-1). The same contours are overlayed onto a greyscale H${\alpha }$ image at the bottom. Both images subtend $2' \times 3'$ and are orientated with North at the top and East to the left

Our more quantitative approach, of profiling along the major axis, employs a rectangular mask to separate H${\alpha }$ radiation from the disk with that emitted by the DIG layer. The width of the mask is fixed at $4\times l$ where:

Figure 10: Left, H${\alpha }$ emission has been superimposed as contours onto unsharp-mask B-band image of NGC 4302. Isophotes begin at 10 EM and increase in steps of 0.75 mag (1 EM = $2 \ 10^{-18}$ ergs cm-2 s-1). The same contours are overlayed onto a greyscale H${\alpha }$ image shown on the right. Both images subtend $0.75' \times 1.5'$ and are orientated with North at the top and East to the left

$$l = h_{\rm S} {\rm Cos}(i) + z_{\rm S}.$$ (4)

Here, i, $h_{\rm S}$ and $z_{\rm S}$ are the inclination, stellar scale-length and stellar scale-height, respectively. These are taken from the radiation transfer parameters given in Table 3 (cf. Eq. (1) used to define the extraplanar layer for the dust). The H${\alpha }$ flux is integrated, in z-height, up to the box boundary for emission from the disk and beyond the box boundary for radiation from the DIG. In Fig. 11, we plot the amount of extraplanar dust at each point along the major axis against the corresponding H${\alpha }$ emission from HII regions in the disk. In Fig. 12, the extraplanar dust is compared with the corresponding DIG emission. A sampling interval, along the major axis, of 1 kpc has been used for all the galaxies depicted. Of the 4 galaxies plotted (NGC 4762 shows no evidence of either extraplanar dust or ionized gas and is therefore not included) only NGC 4013 appears to show any correlation between the distribution of extraplanar dust and the brightness of H${\alpha }$ emission. We checked for stronger correlations over smaller and larger scale-sizes but found that the behaviour of the plotted data remained more or less unaltered for different sampling intervals.