THE RELATIONSHIP BETWEEN THE DENSE NEUTRAL AND DIFFUSE IONIZED GAS IN THE THICK DISKS OF TWO EDGE-ON SPIRAL GALAXIES^*

The feedback of energy and matter from stars to the interstellar medium (ISM) plays a central role in the evolution of galaxies. It impacts the physical/thermal state of the ISM, affecting the ability of the gas to form stars and affects the galactic-scale star formation by ejecting material from the thin disk or altogether from a galaxy (e.g., Dekel & Silk 1986). Most disk galaxies with sufficient star formation rates (SFRs; or SFR surface densities) show extended "thick disks" of warm interstellar matter extending ∼2 kpc away from the galaxy midplane (Rossa & Dettmar 2003a, 2003b). The presence of a thick disk ISM supported against gravity is a direct result of energetic feedback processes (although some additional material may be contributed through infalling matter; see Sancisi et al. 2008). Thus, the study of the statistical properties (e.g., Rossa & Dettmar 2003a, 2003b; Howk & Savage 1999) and detailed physics (e.g., Haffner et al. 2009) of the thick disk ISM can help us understand the role that feedback plays in disk galaxies.

The energetic processes that expel gas from the thin disk also act on the solid phase of the ISM, interstellar dust grains. This has been utilized to study the physical structure as well as the phase structure of the thick disk ISM (Howk & Savage 1997, 1999, 2000; Alton et al. 2000; Thompson et al. 2004). While the energetic expulsion of matter from the disk can impact the dust-to-gas ratio through dust destruction, the fact that prevalent extraplanar dust is seen in ∼50% of normal spiral galaxies (Rossa & Dettmar 2003a; Howk & Savage 1999) and even in galactic superwinds such as seen in M82 (Kaneda et al. 2010; Roussel et al. 2010; Engelbracht et al. 2006; Hoopes et al. 2005; Scarrott et al. 1991) demonstrates that dust can survive the ejection from the thin disk. In fact, the detection of thick disk dust via the extinction it produces against background starlight in direct broadband images of edge-on galaxies is the most observationally efficient method for detecting extraplanar interstellar matter (Howk & Savage 2000).

The first detailed discussion of the implications of extraplanar dust was in the Howk & Savage (1997, hereafter HS97) study of the nearby edge-on spiral galaxy NGC 891, although the presence of high-z dust clouds in this galaxy was noted earlier (Keppel et al. 1991; Sandage 1961). HS97 showed that the thick disk of NGC 891 was threaded with a network of dusty clouds and filaments observed to heights up to z ∼ 2 kpc. The most remarkable implication of this work was that the individual dust-bearing clouds have very large masses, often in excess of 10⁵ M_☉. Howk & Savage (1999, hereafter HS99) subsequently surveyed a group of nine nearby edge-on spiral galaxies, demonstrating that high-z dust is found in many galaxies and is correlated with the presence of extraplanar diffuse ionized gas (DIG) seen in Hα imaging. Subsequently, Rossa & Dettmar (2003a, 2003b) showed that ∼40% of a much larger sample of spiral galaxies exhibit extraplanar dust and verified the correlation between the presence of extraplanar dust and DIG emission. HS99 and Howk & Savage (2000, hereafter HS00) argued that the gas traced by extraplanar extinction features represented a much denser medium than that seen in Hα emission, with densities an order of magnitude or more higher than DIG material (see also Keppel et al. 1991). The extraplanar dust clouds and filaments, which are not spatially correlated with the DIG, are only visible because they have significantly higher extinction—and likely higher column densities and particle densities—than their surroundings. HS00 hypothesized that the dust structures trace a cold neutral medium (CNM) in the thick disk based on the large column densities, particle densities, and masses of the clouds.

This conclusion was motivated in part by a detailed comparison of the extraplanar dust filaments and DIG emission by HS00, which demonstrated that they were tracing independent structures in the thick disk ISM of NGC 891.⁵ Following the study of NGC 891, Thompson et al. (2004, hereafter THS04) used the Hubble Space Telescope (HST) to investigate the extraplanar dust in the edge-on galaxy NGC 4217. These authors indicate that in addition to the smoother distribution of the Hα emission morphology compared to the extraplanar dust structures, they find little spatial correlation between areas of enhanced Hα emission around regions of concentrated OB associations and high-z dust structures. Rossa et al. (2004) similarly used images from HST to show that the ISM traced by extraplanar dust was distinct in morphology from the DIG in NGC 891, while Rossa & Dettmar (2003a, 2003b) argued for a similar arrangement in their larger sample of galaxies. The dusty extraplanar filaments are physically distinct from the DIG, even though the presence of these two tracers of the extraplanar ISM is strongly correlated (Rossa & Dettmar 2003b; HS99). The small-scale structures in the HS00 and subsequent Rossa et al. (2004) imaging, the high column densities (N_H ≳ 10²¹ in many cases), large cloud masses, and high particle densities (n_H ≳ 1–25 cm⁻³; HS00; Howk 2005) are all features in great contrast to the other observed phases of the thick disk ISM and point to a denser, cooler medium than the DIG. Thus, HS00 concluded that all of the thermal phases of the ISM found in the thin disks of spiral galaxies were also present in their thick disks.

The only galaxy for which a detailed comparison of the Hα (DIG) and extraplanar dust morphologies has been undertaken is NGC 891. While the structures traced by these two probes of the extraplanar ISM were almost completely distinct in NGC 891, the universality of the result is not settled. In particular, Hα images of the DIG are often taken at relatively low spatial resolution, in part due to the faintness of the emitting gas, while the dust structures are lost to confusion at similar resolution. NGC 891 has one of the brightest known Hα-emitting DIG layers (e.g., Rand 1996; Miller & Veilleux 2003) and has more prominent filaments within the DIG than many galaxies (Heald et al. 2007). For galaxies with lower SFRs, where the brightness of the DIG and the prominence of filaments diminish, are there corresponding changes in the extraplanar dust? In this work, we present high-resolution images of extraplanar dust and DIG in the edge-on galaxies NGC 4013 and NGC 4302. Our broadband images include observations from HST that elucidate the smallest scale structures in the dusty thick disk, while our Hα images are among the highest spatial resolution acquired from the ground and have good continuum subtraction. We use these images to extend the detailed comparison of extraplanar dust and DIG properties to these two galaxies, which have fainter DIG emission and a smaller fraction of the emission arising in filaments (i.e., a larger proportion coming from the smooth DIG component). In the end, we find no strong physical connection between the DIG and the structures traced by high-contrast dust absorption in these galaxies, similar to what is seen in NGC 891. The DIG and dust-laden clouds trace fundamentally different material in the thick disk.

Our work is organized as follows. We discuss the observations and data reduction in Section 2. In Section 3 we discuss the properties of the extraplanar dust and associated gas, while in Section 4 we describe the properties of the extraplanar Hα emission (DIG). In Section 5 we provide a direct comparison of the extraplanar dust morphologies and the DIG detected in Hα. We discuss in Section 6 the origins, nature, and properties of the multiphase ISM in both NGC 4013 and NGC 4302 and summarize the major results of our work in Section 7.

The work in this paper is based on optical imaging of two galaxies—NGC 4013 and NGC 4302—whose properties are listed in Table 1. We use broadband and narrowband imaging from several sources: (1) HST/Wide Field Planetary Camera 2 (WFPC2) images of both galaxies in the WFPC2 equivalents of the BVI bands to study extraplanar dust absorption at the highest resolution; (2) ground-based broadband images of NGC 4013 and NGC 4302 taken with the WIYN 3.5 m telescope and the 2 × 8.4 m Large Binocular Telescope (LBT), respectively, to study extraplanar dust at larger radial and vertical distances from the galaxy centers than allowed by HST; and (3) narrowband Hα imaging of the extraplanar DIG in these galaxies, taken with the WIYN 3.5 m telescope, to study the connection (or lack thereof) between the ionized gas and material traced by dust absorption in the thick disks of these galaxies. We discuss the processing of these data sets in this section.

2.1. HST/WFPC2 Observations and Processing

We obtained broadband imaging observations of NGC 4013 and NGC 4302 with WFPC2 on board HST. WFPC2, described fully in Biretta et al. (2002), consists of four cameras. The three Wide Field Cameras (WFCs) together cover an "L"-shaped region 150'' × 150'' with intrinsic spatial sampling of 01 per pixel (f/12.9). The Planetary Camera (PC) covers a 34'' × 34'' square field with 0046 per pixel (f/28.3).

Our observations, summarized in Table 2, were acquired as part of GO program 8242 (PI: Savage). We imaged the majority of the galaxy NGC 4302 in one pointing, while we used two separate pointings to cover a significant fraction of NGC 4013. For NGC 4302 and one of the NGC 4013 pointings, we acquired eight exposures in the F450W filter and four exposures in each of the F555W and F814W filters. We obtained four images in each of the three filters for the secondary NGC 4013 pointing. The F450W, F555W, and F814W filters roughly correspond to the Johnson–Cousins B, V, and I bands.⁶

Table 2. Log of HST/WFPC2 Observations^a

Filter	Total Exp. Time (s)
	NGC 4013b	NGC 4302
F450W (B band)	4000	4000
F555W (V band)	3000	2000
F814W (I band)	3000	2000

Notes. ^aThe angular and physical resolutions are 01 and 8 pc, respectively. ^bThe NGC 4013 images have non-uniform exposure times per pixel. Two separate visits with HST had distinct pointing centers and orientations. The central and northern sections of the galaxy, respectively, have three and two times the exposure time of the southernmost region.

Download table as:  ASCIITypeset image

The images acquired for each pointing were split equally between two dithered sub-positions, with a linear dither step of 3.5 WFC CCD pixels between the positions. We used the Dither II package and its drizzle algorithms (see Fruchter et al. 1998; Fruchter & Hook 1997) to combine the dithered images onto a finer grid, following THS04. The resulting cosmic-ray-cleaned images have an angular sampling of 005 per pixel.

Figures 1 and 2 show composite color images of NGC 4013 and NGC 4302, respectively, created from the WFPC2 BVI images. The WFPC2 images have been combined with ground-based observations in these images. The ground-based images are more sensitive to faint, extended structures (i.e., the faint thick disk stellar emission) and probe a larger field of view. For NGC 4013, the ground-based data are from the WIYN 3.5 m telescope and are discussed in Section 2.2. The ground-based observations of NGC 4302 were obtained with the 2 × 8.4 m LBT and are described in Section 2.3.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

2.2. WIYN Observations and Processing

2.3. LBT Observations and Processing

Visual inspection of Figures 1 and 2 reveals the presence of strong, patchy extinction in the thick disks of these two galaxies. Filaments and clouds of absorbing material are seen to nearly the radial edge of the galaxies' optical disks and to z-heights approaching the largest distances to which stellar light can be seen in the relatively shallow WFPC2 images, although the ground-based LBT and WIYN images reveal that the dust extinction extends to lesser heights than the stellar thick disk. This extraplanar extinction is a direct tracer of the thick disk ISM in these galaxies (e.g., Howk 2005). In order to be seen, these dusty regions must have higher column densities than their surroundings with these filaments having excess extinction compared with nearby sight lines, providing significant contrast between these clouds and their local environment in our images. Smoothly distributed dust does not produce such contrast; thus, a diffusely distributed component of dust may exist undetected in the thick disks of these galaxies. That the column densities of the observed dust-bearing clouds are higher than their surroundings almost certainly implies that the dust features seen in our images have significantly higher volume densities than their surroundings, as well.

Figure 3 shows the HST and WIYN V-band images of NGC 4013. The top panel shows the HST V-band data and the middle panel shows the unsharp masked version of the top panel. The display of the galaxy in the top panel is not able to show the absorbing clouds against the bulge and further out in a single display, making the unsharp masking necessary. The unsharp masked versions of our images remove the large-scale gradients in the underlying stellar light, revealing structure on scales smaller than the smoothing kernel (FWHM ≲ 350 pc; see HS00 and THS04 for detailed discussions of this approach). We only use these images for display and have taken care to assess and minimize masking-related artifacts. Both HST images include insets showing the central bulge region with a stretch that shows the dust absorption more clearly. The bottom panel shows the unsharp masked display of our final WIYN data. The WIYN image gives almost full radial coverage of the galaxy, revealing extraplanar dust structures extending along nearly the entire radial length of the disk of NGC 4013. Extraplanar dust clouds can be identified against the background stellar light z  ∼  2.0 kpc from the midplane. Those dust clouds at the largest heights show less prominent substructure and generally have lower a_V values, but they are still well detected in our WIYN images. We see no evidence for disparities in the amount of extraplanar dust comparing opposite sides of the plane.

Zoom In Zoom Out Reset image size

Similarly, Figure 4 shows the HST and LBT V-band images for NGC 4302. The top panel shows the HST V-band data and the middle panel shows the unsharp masked version of the top panel. The HST images cover ∼6 kpc projected radial distance from the center on either side, while the LBT images show ∼13 kpc radial span of the optical disk and more. The dashed box in the LBT V-band indicates the extent of the HST image shown in the upper panels. Extraplanar dust clouds can be identified against the background stellar light z  ∼  2.5 kpc from the midplane, detected in the LBT images. The FWHM of the Gaussian used to produce the unsharp masked images for NGC 4302 were the same as in NGC 4013.

Zoom In Zoom Out Reset image size

3.1. Quantifying Dust Extinction

Using the HST V-band data for our measurements (top panels of Figures 3 and 4) and following HS00 and THS04, we provide spatially averaged estimates of the apparent opacities of a small number of extraplanar dust structures. From these we derive the "apparent" extinction, column densities, particle density, and masses of these dust-bearing structures. We define the "apparent" extinction, a_λ, for a given waveband λ as

$$\begin{equation} a_\lambda =-2.5 \log (S_{{\rm dc},\lambda } / S_{{\rm bg},\lambda }); \end{equation} \tag{ 1 }$$

here, S_dc,λ is the average surface brightness measured toward a dust cloud (dc) and S_bg,λ is the surface brightness of the local background (bg). Defined in this way, the apparent extinction is a lower limit to the true extinction, A_λ, because S_dc,λ includes both extincted starlight from behind as well as unextincted light emitted in front of the feature. One can, in theory, use a_λ for multiple wavelengths with an assumed extinction curve to estimate the fraction of light originating in front of the dust feature (e.g., Gallagher & Hunter 1981; Knapen et al. 1991; HS97). In this case, the fraction of S_dc,λ arising in front of the clouds is x and the true extinction is a function of a_V and x. However, HS00 found that this approach does not always lead to self-consistent results, likely due to the effects of spatial averaging of probing differing depths into clouds at different wavebands and of scattering. HS97 (who used a narrower range of wavelengths) provided evidence for a significant amount of stellar light arising in front of the dust clouds in NGC 891.

Here we follow HS00 and THS04 in assuming A_V ≳ a_V (equivalent to assuming x ⩾ 0). This has the advantage that our calculations of the physical properties of the structures seen in our images will provide lower limits to the true properties. HS00 discuss the degree to which A_V departs from this assumption for several values of x. For example, for x = 0.25, the V-band extinction A_V ≈ 1.5 a_v. We do not report the extinction in the other bands.

For each dust feature we measure the ratio of surface brightnesses S_dc,λ/S_bg,λ using intensity distributions measured vertically or horizontally, with respect to the disk, averaging the results for a given cloud. We estimated the background surface brightness appropriate for each feature by fitting a spline to the intensity distribution of the starlight in each cut. This technique is different than that used in our previous works where we were more concerned with the impact of the smearing due to the seeing. The current measurements will average the a_V values over more of the clouds than those earlier measurements, leading to somewhat smaller extinction values.

3.2. Physical Properties of Individual Dust Structures

Table 5 summarizes the physical properties of several cloud complexes in the thick disks of the two target galaxies from the HST V-band data. Figures 5 and 6 identify the clouds discussed in Table 5; they are displayed in the unsharp masked WFPC2 images to better show the small-scale features of the selected clouds. Several measured quantities are given, including their positions, physical dimensions, height above the midplane, and a_V values. Estimates of the column densities, densities, and masses associated with the clouds are also given, as outlined below. In this work we have favored small, relatively isolated structures in order to study the densities implied by the small-scale structure observed in the WFPC2 imaging. The choice to work with isolated clouds tends to favor lower extinction clouds at relatively large z. The smaller sizes (to emphasize high densities) tend to favor lower masses.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 5. Properties of Individual High-z Dust Features

Clouda	Dimensions	|z|b	aVc	NHd	nHe	Massf
(J2000)	(pc×pc)	(pc)	(mag)	(cm−2)	(cm−3)	(M☉)
NGC 4013
115835.5+435702.0	90 × 60	590	0.20	>4 × 1020	2.1	>3 × 104
115834.4+435713.5	50 × 45	640	0.17	>3 × 1020	2.2	>1 × 104
115833.8+435713.8	70 × 70	890	0.15	>3 × 1020	1.3	>2 × 104
115834.4+435651.9	170 × 150	930	0.22	>4 × 1020	0.9	>7 × 104
115832.9+435706.8	40 × 30	690	0.22	>4 × 1020	4.5	>5 × 103
115833.5+435648.8	90 × 55	850	0.12	>2 × 1020	1.4	>1 × 104
115833.3+435648.5	55 × 25	820	0.18	>3 × 1020	4.4	>1 × 104
115833.2+435648.4	65 × 60	810	0.13	>3 × 1020	1.3	>9 × 103
115833.1+435647.9	110 × 70	650	0.21	>4 × 1020	1.9	>3 × 104
115832.2+435644.6	30 × 30	690	0.20	>4 × 1020	4.4	>4 × 103
115832.0+435644.1	60 × 50	660	0.19	>4 × 1020	2.3	>1 × 104
115831.1+435658.2	55 × 40	740	0.17	>3 × 1020	2.7	>1 × 104
115830.3+435651.5	45 × 40	510	0.18	>4 × 1020	2.8	>4 × 103
115830.8+435640.2	130 × 70	500	0.23	>4 × 1020	2.0	>3 × 104
115832.8+435638.6	70 × 60	600	0.16	>3 × 1020	1.8	>1 × 104
115830.0+435648.9	55 × 55	430	0.25	>5 × 1020	2.9	>2 × 104
NGC 4302
122141.9+143506.8	90 × 80	590	0.14	>3 × 1020	1.1	>3 × 104
122143.1+143512.4	110 × 90	900	0.19	>4 × 1020	1.3	>3 × 104
122143.0+143518.9	110 × 100	820	0.22	>4 × 1020	1.4	>6 × 104
122141.9+143545.4	90 × 65	560	0.33	>6 × 1020	3.2	>3 × 104
122143.1+143550.6	60 × 50	860	0.26	>5 × 1020	3.3	>2 × 104
122141.8+143601.9	130 × 90	660	0.21	>4 × 1020	1.5	>3 × 104
122143.1+143620.4	90 × 70	950	0.17	>3 × 1020	1.5	>4 × 104
122142.8+143627.3	95 × 60	600	0.19	>4 × 1020	1.9	>5 × 104
122141.8+143633.9	95 × 65	610	0.31	>6 × 1020	2.0	>7 × 104
122141.6+143649.3	105 × 95	820	0.24	>5 × 1020	1.5	>4 × 104
122143.2+143658.4	95 × 90	1040	0.12	>2 × 1020	0.9	>2 × 104

Notes. ^aCloud ID based on R.A. and decl. ^bProjected vertical distance from the midplane. ^cAverage apparent V-band extinction. ^dColumn density derived using measured a_V and the Milky Way value for N_H/A_V ∼1.9 × 10²¹ cm⁻² mag⁻¹. These are lower limits since a_V < A_V. ^eMean density derived from N_H and the minor axis width. ^fApproximate mass based upon the estimated column density and projected area. Includes a factor of 1.37 correction for He.

Download table as:  ASCIITypeset image

The values of a_V in Table 5 are in the range 0.1–0.3 mag. These values are more representative of the outer regions of the clouds than the cores of the structures. Given that the values of a_V are also lower than the true extinction, realistic values of A_V for much of the clouds are likely significantly larger than those values given in Table 5.

Transforming the apparent extinctions discussed above into physical properties of the clouds requires an assumption about the gas-to-dust relationship in the thick disk. Following earlier works, we assume that the clouds follow dust-to-gas relationships appropriate for the disk of the Milky Way (HS97; HS99; HS00). This is a necessary step for estimating the physical conditions of the dust features from the measured apparent extinctions, one whose consequences are discussed at length in HS00. We adopt the average relationship between total hydrogen and the color excess E(B − V) of diffuse clouds in the Galactic disk as estimated by Bohlin et al. (1978): N(H i + H₂)/E(B − V) = 5.8 × 10²¹ atoms cm⁻² mag⁻¹. To relate this to the extinction, we assume R_V ≡ A_V/E(B − V) ≈ 3.1, roughly the mean for diffuse clouds in the Galactic disk (e.g., Cardelli et al. 1989). This gives the total hydrogen (H i + H₂) column density from the apparent extinction:

$$\begin{equation} N_{{\rm H}}({\rm cm}^{-2}) > 1.9 \times 10^{21} a_{V}. \end{equation} \tag{ 2 }$$

The inequality arises since a_V < A_V. Since both NGC 4013 and NGC 4302 are relatively massive spiral galaxies, the assumption of a Milky Way gas-to-dust ratio will probably yield a reliable estimate of the associated gas column density. However, if significant dust destruction has occurred in any particular feature, the derived N_H will be an underestimate, while if the dust has been separated from the gas then N_H will be an overestimate. We favor scenarios in which little dust destruction occurs in moving the dust from the thin to thick disk or where some destruction has occurred in which the dust is separated from the gas. We thus quote the column densities as lower limits.

We give lower limits to the masses for the features in Table 5, calculated as M ∼ μm_HN_HA_dc, where A_dc is the projected area of the dusty clouds. We assume μ = 1.37 to correct for the contribution to the mass from helium and the heavy elements. The definition of the cloud areas is a critical and highly subjective component of the mass estimates. In the present work, we have favored small clouds and have been conservative in our derivation of the projected areas. The implied masses should be treated as lower limits.

In addition to the columns and masses, we also include a very crude density estimate in Table 5. This estimate assumes that the clouds we see are largely cylindrically symmetric such that n_H ∼ N_H/Δx, where Δx is the width of the minor axis of the clouds. The only support for this assumption is the general prevalence of filamentary absorbing structures seen in our images, although this impression could be affected by the general confusion caused by overlapping clouds along the line of sight. In general the morphologies of the absorbing clouds are complex, but we move forward using this simplistic assumption for estimating the densities.

Keeping in mind the uncertainties discussed above, the properties of the clouds summarized in Table 5 are nonetheless intriguing. The features have estimated columns N_H ≳ 2 × 10²⁰ cm⁻², simply as a matter of selection (it is difficult to identify clouds with a_v < 0.1). We infer cloud masses ≳10⁴ M_☉ for all but the smallest two clouds in Table 5, and the densities implied for those clouds included in the table are of order n_H ∼ 1–4 cm⁻³. Additionally, we examined some of the darkest narrow filaments in the WFPC2 images located in regions too confused to show well in Figures 5 and 6, so they are not included in Table 5. Several of these clouds have column densities approaching or exceeding 10²¹ cm⁻², giving mean densities n_H ≳ 8–10 cm⁻³, which is extraordinary for thick disk clouds.

The column densities and densities measured for the thick disk clouds in NGC 4013 and NGC 4302 are similar to those seen in NGC 891 and NGC 4217 (HS97; HS00; THS04; Howk 2005). The derived densities likely depend on the resolution with which one observes the structures, as higher resolution images not only smear less background light into the structure but also allow the identification of smaller scale filaments. The masses given in Table 5 are smaller than have been reported in earlier works (HS99; HS00; THS04), but that is largely a function of our newly adopted measurement technique and the choices we have made for emphasizing the smaller subcomponents rather than the largest complexes. Complexes of absorption with implied total masses well in excess of 10⁵ M_☉ certainly exist in these galaxies. Overall, we do not see any clear evidence that the dust-bearing thick disk clouds in these galaxies are significantly different in character than those seen in NGC 891 (HS00) or NGC 4217 (THS04).

3.3. Apparent Morphologies of Dust-bearing Gas Clouds

Our images allow us to identify large-scale dust-bearing clouds in the thick disks of these galaxies and study their internal structure. Our high angular resolution HST images reveal substructure within the cloud complexes, down to the limit of our resolution, but only for the inner sections of both galaxies. Our LBT and WIYN data allow us to identify dust clouds at larger radial and vertical distances due to their greater field of view and depth compared with the HST images. Similar to clouds in the Milky Way, in which structure is seen on many scales, the thick disks of NGC 4013 and NGC 4302 contain narrow filamentary structures, large complexes of clouds (specifically in NGC 4302), as well as small, very dense clouds seen to our resolution limit.

3.3.1. NGC 4013

The extraplanar dust-bearing clouds in NGC 4013 are generally highly structured filaments, although we see a broad range of morphologies in both the HST and WIYN images in Figure 3. The typical dust filaments are ∼100 pc by a few × 100 pc in size (see Table 5 and Figure 3). We note that the dust filaments are unlikely to be associated with a flare in the outer galaxy (see detailed discussion in HS99; HS00) or as a result of the warp in NGC 4013, which is perpendicular to our line of sight (Bottema 1996) and therefore would not give rise to the symmetrical distribution of clouds seen in our images.

There is a strong preference for the highest-z absorbing structures (z ⩾ 2 kpc) in NGC 4013 to appear bent toward the northeast, especially on the eastern side of the galaxy (see the bottom panel of Figure 3). Such an arrangement could occur if there is a rotational lag that increases with height (e.g., Heald et al. 2007). This is consistent with the sense of rotation of NGC 4013, with the southwest side receding from us (Bottema 1996). NGC 4013 is a member of the spiral-rich Ursa Major group of galaxies. Ursa Major is considered to be a loose group with no central concentration and no sign of strong H i deficiencies, suggesting that this group lacks a strong intracluster medium (Angiras et al. 2007). As a member of the Ursa Major group, NGC 4013 is relatively isolated. Thus, the bending of filaments is not likely the result of movement through an intracluster medium in this case (Martínez-Delgado et al. 2009). We find no evidence for obvious supershell structures in NGC 4013, although identifying such structures within the complex distribution of clouds might be difficult.

We see a set of very large scale complexes of dust above and below the bulge of the galaxy in Figure 3. These large dust complexes, traced out to z ∼ 1.8 kpc on either side of the bulge, show significant small-scale structure within them, including dense clouds similar to those cometary-shaped clouds identified in HS00. Figure 7 shows one of these complexes to the southeast of the bulge. This large complex (∼900 pc × 200 pc) extends from within ∼500 pc of the plane, bending toward the east with increasing height above the plane. It shows several cometary-like substructures at heights z ∼ 600–850 pc. Figure 7(b) provides a close-up image of this large complex, which has a mean extinction value of a_V = 0.04 and total mass of order ≳10⁵ M_☉. Figure 7(c) shows two of the cometary-like clouds, clouds 115832.2 + 435644.6 and 115832.0 + 435644.1 from Table 5. Contours are included to guide the eye (but are based on the unsharp masked image and are not quantitatively useful). Cloud 115832.2 + 435644.6 has a head–tail-type structure; the head of the cloud has a_V = 0.36 and implied density n_H ⩾ 7 cm⁻³. The association of these cometary structures with the larger complex suggests that the complex is interacting with an ambient medium as it falls toward the disk through a corona or outflowing wind from the disk.

Zoom In Zoom Out Reset image size

All of the head–tail clouds we see in NGC 4013 seem to be associated with larger complexes, like that shown in Figure 7. The head–tail clouds do not seem to be common, although the clouds may still be interacting with an ambient medium. Such morphologies are reminiscent of those seen in some high-velocity clouds (HVCs) passing through the diffuse halo of the Milky Way. Putman et al. (2011) investigated the H i 21 cm emission-line morphologies of isolated HVCs, finding that only ∼35% of them show head–tail morphology, suggesting such morphologies are short lived. If the extraplanar clouds in our images are analogous to the HVCs, those with cometary morphologies may be moving at larger velocities and interacting more strongly with the thick disk ISM. While the HVCs of the Milky Way do not appear to contain dust with the exception of the Magellanic Stream (Lu et al. 1998; Sembach et al. 2001; Richter et al. 2001; Fox et al. 2010), and their origins therefore are different from the thick disk dusty clouds studied here, the physics that dictates the head–tail morphology is likely quite similar. That is, the magnetohydrodynamical processes that shape those cometary structures likely do not strongly depend on the dust content of the clouds.

Our HST images reveal a vast array of small-scale absorbing features with dimensions approaching our resolution limits. The smallest scale dust structures are smaller than a few  ×  10 pc in size. We have measured the sizes and opacities of a number of these very small structures and estimate that they have a_V ∼ 0.2–0.4 mag and n_H ⩾ 8 cm⁻³. Indeed, each complex exhibits significant substructure. To demonstrate this we show in Figure 8 an expanded view of NGC 4013 cloud 115834.4 + 435651.9, located at z  ∼  930 pc from the midplane. Prominent substructures are clearly seen. The contours in Figure 8 range from a_V ∼ 0.2 to 0.6. For comparison, our spatially averaged value for this cloud is a_V ∼ 0.22 (Table 5). This demonstrates the impact of our averaging technique when making our measurements. Two particular subclouds, which are similar in size and opacity, have a_V ≳ 0.5 mag, implying column densities of N_H ⩾ 10²¹ cm⁻² and particle densities of n_H>7 cm⁻³. These are lower limits for the column densities and imply that these dust structures have properties similar to Milky Way diffuse clouds, yet are located ∼1 kpc away from the midplane.

Zoom In Zoom Out Reset image size

While the dust in highly structured clouds is most readily apparent in our images, one can find very low extinction, more diffuse material in the thick disk in our images (particularly the WIYN and LBT images, which have greater sensitivity to low surface brightness fluctuations). The lowest extinction clouds visible in our images of NGC 4013 have a_V ∼ 0.03 mag. If we couple that with a rough measure of the area occupied by these low column density clouds (with very low contrast), the total mass implied in this galaxy is of order ∼10⁸ M_☉. This is likely similar to the total mass in the population of denser, more readily apparent complexes discussed above (see HS00).

3.3.2. NGC 4302

The extraplanar dust-bearing clouds in NGC 4302 are filamentary with some large dust complexes also notable in the thick disk in both the HST and LBT images in Figure 4. The LBT unsharp masked display (the bottom panel of Figure 4) shows extensive extraplanar dust-bearing clouds along the entire length of the plane detectable to z ∼ 2.5 kpc. While large complexes of extraplanar dust are seen on both sides of the planes, the intensity of starlight as a function of z shows a clear asymmetry across the plane in the bulge region in NGC 4302. The apparent deficit of starlight at z  ∼  1 kpc on the west side of the plane corresponds to a large diffuse complex of dust on this side of the disk. This diffuse complex is not easily seen in the HST unsharp masked images given its size and relatively little substructure, but is visible in the unsharp masked LBT image (the bottom panel of Figure 4). This extinction complex on the western side of the bulge area may be the cause for a measured asymmetry in Hα emission image as well (see Section 4).

A common characteristic in the majority of extraplanar clouds in NGC 4302 is the apparent bending of the large complexes toward the north at large z-height, out to z  ∼  2.3 kpc. This may be connected with the motion of NGC 4302 through the Virgo Cluster, as discussed by Chung et al. (2007). These authors postulate that the extensive tail of the H i disk extending along the northern edge of NGC 4302 is due to the galaxy's motion toward the south as it falls into the Virgo Cluster. Our images may be showing a physical manifestation of this movement through the intracluster medium in the shaping of the largest z dust features. We note that NGC 4298, projected close on the sky to NGC 4302, is also located in the Virgo Cluster. NGC 4302 shows no sign of a strong interaction with NGC 4298.

The ratio of thick-disk to thin-disk stellar scale heights is smaller in NGC 4302 than many other edge-on galaxies (as it is in the Hα; Rand 1996). Combined with the line-of-sight overlap of structures, this makes it difficult to cleanly identify large-scale extraplanar dust structures. Our images do not provide strong evidence for any obvious supershells or galactic chimney features extending from the midplane of NGC 4302. Figure 6 shows the unsharp masked HST images of this galaxy with the clouds from Table 5 marked. As in NGC 4013, we see small-scale structures in the dusty clouds in NGC 4302 to the resolution limit of our HST images (∼10–20 pc). The implied densities of the small-scale clouds range from n_H ⩾ 2 to 14 cm⁻³ for the narrowest, darkest thick disk clouds.

Similar to the case of NGC 4013, there are diffuse, lower column density structures seen over a significant amount of the thick disk region of NGC 4302. These tend to have minimum extinctions of a_V ∼ 0.02 mag, implying a total thick disk mass in this more diffuse matter of ∼10⁸ M_☉.

Extraplanar DIG has been previously studied in both NGC 4013 (Rand 1996; Miller & Veilleux 2003) and NGC 4302 (Pildis et al. 1994; Rand 1996; Rossa & Dettmar 2000; Collins & Rand 2001; Heald et al. 2007). We have revisited these galaxies in an effort to achieve higher resolution, greater depth, and/or more consistent continuum subtraction in previous works. Our overall goal is to compare the Hα and dust morphologies to study their connection in these galaxies, which requires high-resolution imagery. We detect thick disk Hα emission, with 08 resolution, in both galaxies.

4.1. NGC 4013

Extraplanar DIG at large z-heights can be seen over nearly the whole of the star-forming disk (to R ∼ 9 kpc) in NGC 4013 (see the bottom-right panel of Figure 9, which emphasizes the faintest Hα emission). The extraplanar Hα emission in NGC 4013 is considerably fainter and shows much less prominent filamentary emission than in NGC 891 (Rand 1996). Rand (1996) gives Hα emission scale heights of h_z ∼ 0.7–0.9 kpc (for the eastern side); however, he also notes that significant continuum subtraction difficulties affect his images. Miller & Veilleux (2003) fit a scale height of h_z ∼ 1.5 kpc (for the "southeastern" side). Figure 9 shows our Hα images of NGC 4013 in four panels; the top-left panel is an unsharp masked WIYN V-band image to show the dust absorption at the same resolution as the Hα emission, while the bottom-left panel is an unsharp masked Hα image to show the small-scale features (as before, the smoothing kernel has FWHM  ≲  350). Both right panels in Figure 9 are Hα emission, stretched to emphasize the brightest emission (top-right panel) and the faintest emission (bottom-right panel). Our images are consistent with the description of the DIG morphology for NGC 4013 given by Rand (1996): diffuse with few filaments. Over the majority of the disk, the DIG morphology is smooth and lacks bright, narrow filamentary structures like those seen in the Hα emission of NGC 891 and NGC 5775 (Rand 1996; HS00). What filaments exist in Figure 9 are relatively subtle, the brightest of which have peak intensities of ∼10%–20% of the local smooth DIG. Even the brightest filaments in NGC 4013 are comparatively fainter than those in NGC 891, which have 30%–100% of the brightness of the local diffuse DIG emission (HS00). Given the faintness of the smooth DIG layer in NGC 4013, the filaments have a much lower absolute intensity as well.

Zoom In Zoom Out Reset image size

One of the most notable morphological features of the DIG in NGC 4013 is the large "H"-shaped structure emanating from both sides of the bulge region (see Rand 1996). This "H" structure consists of four large-scale filaments present at similar positions on either side of the bulge, at projected radial distances R ∼ 200–300 pc and visible in our images to z  ∼  2.2 kpc. Rand (1996) argued that this feature was a relic nuclear (bipolar) outflow from an earlier nuclear starburstor active galactic nucleus. García-Burillo et al. (1999) mapped the "H" structure in CO emission (see their Figure 3) and argued for a close association between the extraplanar DIG filaments and the extraplanar CO emission. Schinnerer et al. (2004) also mapped CO emission from the plane of NGC 4013, connecting the "H" structure in a high-density CO ring in the disk. These works both invoke a superwind outflow as the origin of this structure, in agreement with Rand's conclusion.

4.2. NGC 4302

Figure 10 shows the continuum-subtracted Hα emission in NGC 4302. The images of this galaxy are not as deep as for NGC 4013, but they have better resolution than most previous Hα images of this galaxy (Pildis et al. 1994; Rand 1996; Rossa & Dettmar 2000, 2003a, 2003b). NGC 4302 has one of the faintest detectable DIG layers (Rand 1996; Rossa & Dettmar 2000). Collins & Rand (2001) measure the emission scale height h_z ∼ 0.55–0.7 kpc. Rand (1996) characterizes the extraplanar DIG morphology of NGC 4302 as "faint and diffuse"; our images are consistent with both of those characterizations. In our images the thick disk DIG is smoothly distributed within the first z  ∼  2 kpc of the plane, but only within R  ∼  7 kpc and 8.5 kpc radially from the bulge in northern and southern halves of the galaxy, respectively. The Hα emission abruptly decreases to undetectable levels at larger radii. We detect slightly brighter extraplanar DIG on the eastern side compared to the western side of the galaxy, most notably around the bulge region. This asymmetry is also seen in continuum starlight and is due to enhanced extinction from large extraplanar dust complexes on the western side of the plane at z  ∼  1 kpc (discussed in Section 3.3). We find no bright plumes or filamentary structures. There are some very weak DIG filaments, the brightest of which measure only ∼20%–25% above the local (very faint) diffuse background. Thus, the DIG in NGC 4302 shows little to no structure or filaments, and what substructure is detected seems to be the result of dust extinction effects.

Zoom In Zoom Out Reset image size

Dettmar (1995) and Heald et al. (2007) found that the presence of Hα emission in NGC 4302 was weak and diffuse, especially in comparison to NGC 891, although both galaxies have prominent extraplanar dust. Our Hα images are the highest resolution available for both NGC 4013 and NGC 4302 and show both have smoothly distributed DIG layers with much less prominent filaments than the DIG of NGC 891 and even less than NGC 5775 (Rand 1996; HS00). Both NGC 4013 and NGC 4302 also have comparably weaker Hα emission than NGC 891 and NGC 5775, which are known for their bright Hα emission and have larger DIG emission scale heights (Collins & Rand 2001). Thus, the galaxies can be ranked in order of decreasing DIG scale height: NGC 5775, NGC 891, NGC 4013, NGC 4302. This is the same order one would have if ranking by decreasing Hα intensity and decreasing prominence of DIG filaments.

NGC 4013 and NGC 4302 show both dust and DIG in the thick disk. Is there any relationship between the physical structures probed by these tracers of the thick disk ISM? NGC 891 is the only galaxy for which a detailed comparison between the dust-bearing cloud structures and the DIG in the thick disk has been made at high resolution, although several works have noted the distinct morphologies of thick disk dust and DIG in NGC 891 (Dettmar 1990; HS97; HS00; Rossa & Dettmar 2003a, 2003b; Rossa et al. 2004; THS04). Howk & Savage (HS00) argued that the dusty clouds in the thick disk of NGC 891 trace a dense medium, in part because of these differing morphologies. They also estimated typical densities for the extraplanar dust-bearing clouds of n_H ⩾ 2–10 cm⁻³, significantly higher than DIG structures at any distance from the midplane. Furthermore, HS00 found very few cases where structures traced in Hα emission and dust absorption could be even loosely associated. They found no direct physical connection between the extraplanar dust-bearing structures and the DIG seen in Hα emission. Rossa et al. (2004) reached the same conclusion after comparing the thick disk dust extinction and Hα emission in high-resolution HST imaging of NGC 891. Keppel et al. (1991) also noted that the emission measure from clouds as dense as those probed by the extraplanar dust should be much higher than that observed from the DIG in NGC 891, suggesting the dusty clouds trace a different medium.

Our observations allow us to explore the relationship between extraplanar dust extinction and Hα emission at high resolution in two additional galaxies. Extraplanar dust filaments are very prominent in the thick disk of both galaxies, whereas the extraplanar DIG layer lacks prominent filaments or structures in both. Figures 9 and 10 show the Hα emission images on the same physical scales as the broadband ground-based images, allowing for a direct comparison of the dense thick disk dust cloud structures and the Hα emitting DIG structures.

In NGC 4013 we see filamentary dust structures in the thick disk in our broadband HST and WIYN images along the entire radial length of the galaxy, with extensive and highly structured dust filaments seen near the bulge region. Our images show absorbing dust structures out to heights z  ∼  2.0 kpc on both sides of the plane of NGC 4013. On the largest scales, the extraplanar dust and DIG show distinct distributions. The DIG is relatively smooth with weak and few filaments. The DIG filaments that do exist are not reflected in the dust. The large extraplanar dust complexes themselves are not directly identifiable in the Hα emission. The lack of a morphological connection between extraplanar dust clouds and the DIG is consistent between the largest and smallest scales in NGC 4013.

In NGC 4302 we come to the same conclusion: the morphologies of the absorbing dust structures and Hα emitting structures show no obvious physical relationship. In NGC 4302 our broadband images show filamentary dust-bearing clouds extending to z ∼2.5 kpc along the length of the galaxy. The thick disk DIG layer has very few filamentary structures and is smoothly distributed along the inner R(north) ∼7 kpc and R(south) ∼8.5 kpc; at larger R the extraplanar Hα emission quickly disappears. This drop-off in the DIG emission contrasts with the behavior of the extraplanar dust filaments, which are seen extending to large z-heights to nearly the edge of the optical disk. The Hα morphology we see in NGC 4302 is a result of large dust complexes extinguishing emission, yet no physical DIG structures are traced by any dust filaments seen in our HST and LBT images. Overall, the dust clouds are structured on very small scales and strongly filamentary, while the extraplanar ionized gas is primarily diffuse and smooth.

When comparing the morphologies of the dust and DIG in the thick disk, neither NGC 4013 nor NGC 4302 show evidence for a direct physical relationship between the thick disk dust seen in the broadband images and the extraplanar ionized gas seen in Hα emission. NGC 4302 shows the strongest contrast in the physical appearance of the dust and DIG as it has one of the weakest extraplanar DIG layers yet has a very dusty thick disk. In a sense, we expect very little connection between the dust structures and DIG given the importance of the smooth components of the DIG in these two galaxies (Rand 1996; Rossa & Dettmar 2000; Heald et al. 2007). The smaller scale structures in the dust-bearing clouds, their larger column densities, and even the differences in their large-scale distribution emphasize that the dust traces a phase of the thick disk ISM distinct from the DIG.

NGC 4013 and NGC 4302 are two edge-on galaxies that display a lack of a direct physical correlation between thick disk dust and ionized gas. This physical difference between the morphology of the extraplanar DIG and dust is not unique to NGC 891. At the same time, there is a correlation between the presence of extraplanar dust and DIG disks of galaxies. As first noted by HS99 in a small sample of galaxies and subsequently in the much larger survey of Rossa & Dettmar (2003a, 2003b), when DIG is detected in a galaxy, extraplanar dust is also visible ∼90% of the time. Thus, although there is no evidence for a morphological relationship between the dust and DIG, there is a strong correlation between the presence of extraplanar dust and DIG in galaxies, a result reflected in NGC 4013 and NGC 4302.

We have presented deep broadband and Hα images of the edge-on spiral galaxies NGC 4013 and NGC 4302. Our images reveal significant numbers of dusty extraplanar clouds via the extinction they produce against the background thick disk and bulge starlight in both galaxies. These galaxies were chosen in part because the morphology, brightness, and scale height of the DIG emission in these galaxies were significantly different than that in the other galaxy for which a detailed comparison of the extraplanar dust and DIG is available, NGC 891 (HS00; Rossa et al. 2004). Thus, our program is designed to understand whether the changes in the morphology and properties of the DIG are reflected in any way in the extraplanar dust. While various studies have argued that the thick disk dust absorption is tracing a different medium or phase than the DIG (Keppel et al. 1991; HS99; HS00; Rossa et al. 2004) with distinct morphologies, we are testing whether this dense medium responds at all to the processes that shape the DIG properties.

We find no evidence that the morphologies and properties (quantity, sizes, masses, etc.) of the extraplanar dust-bearing clouds are subject to the strong changes like those seen in the DIG morphologies. Table 6 presents L_FIR, SFR, and star formation surface density ($\dot{\Sigma }_{*}$) for NGC 4013 and NGC 4302 in comparison to previously studied edge-on galaxies NGC 5775, NGC 891, and NGC 4217. The star formation surface density (SFR per unit area) is $\dot{\Sigma }_{*} \equiv$ SFR/A₂₅, where A₂₅ = πD²₂₅/4. Ordering these galaxies by SFR (derived from the L_FIR), Table 6 shows a connection between the morphology of the extraplanar DIG and the SFR. Thus, as the importance (prominence, brightness, etc.) of the DIG filaments and scale height decrease along the sequence of NGC 5775, NGC 891, NGC 4217, NGC 4013, and NGC 4302, the morphological and physical properties of the observable dust seem to vary little. This is in part a selection effect, since the detection of the extraplanar dust in our images relies on the small-scale structure of the absorbing structures. The thick disk dust is only visible because it is more opaque than its surroundings, and this likely implies the clouds in our images have significantly higher particle densities than their surroundings. However, this also emphasizes our point: the properties of the dust are disconnected from that of the DIG in the thick disk.

Table 6. Comparison of Extraplanar Dust and DIG Properties

Galaxy	LFIRa	SFRb	$\dot{\Sigma }_{*}$c	Dust	DIG	DIG
	(109 L☉)	(M☉ yr−1)	(10−3 M☉ yr−1 kpc−2)	Morphology	Morphology	Reference
NGC 5775	32.5	9.9	12.0	Highly structured filamentsd	Many bright filaments	1
NGC 891	12.4	3.8	3.4	Highly structured filaments	Bright filaments, faint + diffuse	2,3,4
NGC 4217	8.1	2.5	4.6	Highly structured filaments	Two faint patches	5
NGC 4013	4.8	1.5	2.8	Highly structured filaments	Faint filaments; faint + diffuse	6
NGC 4302	2.6e	0.9	1.4	Large complexes + filaments	Faint + diffuse	5,6

Notes. ^aThe L_FIR values were calculated using IRAS 60 and 100 μm fluxes from Sanders et al. (2003). ^bThe SFRs were calculated following Kewley et al. (2002) using the calculated L_FIR and the Kennicutt (1998) star formation calibration. For NGC 5775, NGC 891, and NGC 4217, we adopt distances of 26.7, 9.6, and 17 Mpc, respectively. ^cThe star formation surface density, $\dot{\Sigma }_{*} \equiv$ SFR/A₂₅, where A₂₅ = πD²₂₅/4. ^dDescription based on images from the Hubble Legacy Archive. ^eThe IRAS images do not resolve NGC 4302 and NGC 4298. We estimate the fractional contribution to both IRAS bands using the relative 100 μm fluxes for these two galaxies from more recent Herschel images (Corbelli et al. 2012). References. (1) Collins et al. 2000; (2) HS00; (3) Rand et al. 1990; (4) Dettmar 1990; (5) Rand 1996; (6) this work.

Download table as:  ASCIITypeset image

In fact, one can see morphological and structural changes in the DIG that are not mirrored in the extraplanar dust within individual galaxies. In NGC 891 the southwest half of the disk has a smaller DIG scale height, with many fewer and less prominent filaments than the northeastern half of the disk (HS00). However, there is relatively little change in the properties and incidence of dust between these regions, save that there are a very few DIG filaments in the northeastern half that are physically related to extraplanar dust structures. (Such an arrangement is only seen in this portion of NGC 891 and in some portions of the strongly star-forming galaxy NGC 4631; see Wang et al. 2001.) Similarly, the present observations show the DIG layer in NGC 4302 cuts off sharply at a projected radial distance from the nucleus of R ∼ 7–8.5 kpc, while the extraplanar dust can be seen well beyond these projected radii. Additionally, the thick disk near the bulge region of NGC 4013 shows highly structured filaments whereas the extraplanar DIG morphology is smooth even in the brightest Hα emission.

Apparently the factors that shape the morphology and quantity of the extraplanar DIG do not affect the dusty extraplanar clouds seen in our images. While there is a strong correlation between the existence of extraplanar DIG and extraplanar dust within a galaxy (HS99; Rossa & Dettmar 2003b), the ionized gas traced by Hα emission and dense clouds traced in optical continuum absorption have very little to do with one another. What drives this distinct behavior of the DIG and dense, dusty clouds? One possibility is that the SFR or star formation surface density (SFR per unit area) plays a role in shaping the morphology of the DIG layer. Rossa & Dettmar (2003a) found that DIG morphology was connected to star formation activity in a galaxy, with the transition from higher to lower $\dot{\Sigma }_{*}$ or SFR being associated with a progression from DIG with prominent filamentary structures through fainter, more diffuse DIG. The sequence then of NGC 5775, NGC 891, NGC 4217, NGC 4013, to NGC 4302 is one of decreasing SFR as well as decreasing DIG scale height and Hα filament prominence. While this is a limited sample, we also see changes in star formation affect the DIG morphology on smaller scales. Rossa & Dettmar (2003a, 2003b) note that star formation on the local and global scales were factors in the presence of extraplanar DIG in their survey. For example, the southwestern half of NGC 891 has a smaller DIG scale height and fewer prominent filaments than the northeastern half (HS00) while also having a lower SFR as evidenced by radio continuum and far-infrared measurements (Dahlem et al. 1994; Wainscoat et al. 1987). The edges of NGC 4302 beyond the end of the DIG layer are also representative of this trend.

By what mechanism could the star formation or its surface density affect the morphology of the DIG in this way? The answer is unclear. It could be that the filaments seen in the DIG are tracing ionized fragments of supershells whose presence and brightness require very recent star formation in clusters to maintain the ionization, while the smoother DIG component is ionized by longer-lived field OB stars (e.g., Hoopes & Walterbos 2000). However, Rand (1998) found that the bright filaments in NGC 891 did not show unusual line ratios (e.g., [N ii]/Hα, [S ii]/Hα) for the DIG. In fact, the filaments seemed to show the same overall behavior as the more diffuse DIG, with lower forbidden line to Hα ratios associated with brighter emission. Thus, if the filaments are associated with recent superbubbles, they still share common ionization and excitation characteristics with the general DIG. At this point this assignment of cause is largely speculation. It seems likely that SFR and/or $\dot{\Sigma }_{*}$ likely play a significant role in not only determining whether a galaxy has significant DIG emission (Rossa & Dettmar 2003a, 2003b; Rand 1996), but also in shaping the DIG structure (Rossa & Dettmar 2003b).

As with the extraplanar DIG, a minimum SFR or $\dot{\Sigma }_{*}$ seems to be required before extraplanar dust is observable (Rossa & Dettmar 2003a, 2003b), but above that threshold there is little discernable effect on the morphology (Table 6). Furthermore, there is no evidence that the positional changes in $\dot{\Sigma }_{*}$ that modify the DIG morphology affect the dense extraplanar clouds. In comparing the DIG morphology sequence outlined in Table 6 to the dust morphologies, the DIG changes in connection with the SFR, while the extraplanar dust is both present and consistently similar between all of the galaxies. That the dust clouds have little associated Hα emission likely means they are not directly illuminated by ionizing radiation (Keppel et al. 1991), since they are optically thick to such radiation. Perhaps they have a different evolutionary timescale for production than the DIG (e.g., Alton et al. 2000), and the stars that would have ionized the clouds have already died off. It may also be that they have a distinct radial distribution compared with the DIG. We see only the foreground dust-bearing clouds with the direct imaging techniques, but perhaps there are few such clouds at the smaller radial distances at which the DIG dominates. It is unlikely we could show this without velocity information for the extraplanar dust clouds, which is not likely unless they can be measured through CO emission (e.g., García-Burillo et al. 1999).

Discussions of the origin of the matter making up the dense clouds seen in our images and the phase structure of these clouds have been presented elsewhere (HS97; HS99; HS00; Howk 2005). Here we expand on those discussions briefly to reflect new thinking (both our own and in the literature). The interstellar thick disk clouds seen via their dust absorption within z ∼ 2 kpc of the midplanes of a number of spiral galaxies are almost certainly dominated by material that has been expelled from the thin disk via stellar feedback, perhaps both mechanical and radiative. In this case, the signpost used to trace the matter—the dust—identifies the clouds as arising from metal-rich material. This has two important implications. First, the great quantities of dust in these clouds implies that they are dominated by metal-rich thin disk matter that has been expelled rather than by infalling (metal-poor) matter as invoked to explain the change in DIG rotation curves with z (e.g., Fraternali & Binney 2006; Heald et al. 2007).

Second, the presence of dust-rich matter in the thick disk implies that the processes that lift the dust from the thin disk are not so violent that they destroy a large fraction of the dust. If a significant amount of dust is destroyed, the masses calculated for our observed clouds scale upward. If only 10% of the dust survived the trip to the thick disk the total mass in these clouds would approach the total gas mass of the ISM in these galaxies (HS00), which is unrealistic. The existence of dust in the thick disk fueled by feedback-driven expulsion is clearly not a new discovery. It does not cause difficulties with existing models for galactic-fountain-type phenomena. The walls of a supershell may extend to quite large distances from the plane, but much of the mass was incorporated into those walls while the shell expanded relatively slowly (e.g., Weaver et al. 1977) destroying little of the associated dust. Furthermore, it is possible that radiation pressure plays a role in this process (Murray et al. 2011; Franco et al. 1991), which would leave all of the dust intact. A completely independent method of probing the existence of dust surviving far out of the plane is through the infrared wavelengths. NGC 891 has detectable mid-infrared radiation out to z ∼ 5 kpc, which is dominated by emission from diffuse interstellar dust (Burgdorf et al. 2007). Thus, extraplanar dust is present many kiloparsecs further from the star-forming disk than our images reveal. Its existence to z ∼ 5 kpc (and even beyond; see Ménard et al. 2010) implies that the dust is not all destroyed as it is lifted from the thin disk.

The thick disk clouds traced by the extraplanar dust absorption in our optical images are physically distinct from the DIG. They trace material with densities larger than that of the DIG by an order of magnitude or two. The large gas densities (n_H ≳ 1–25 cm⁻³; HS00; Howk 2005; this work) and high gas column densities (N_H ≳ 10²⁰ to ≳10²¹) traced by the extraplanar dust are largely consistent with expectations for gas associated with a warm neutral or cold neutral medium. Both are likely present in these clouds, given the level of substructure and evidence for core-halo-type structure in some clouds (see also HS00). Although in principle one might be able to use the distribution of apparent extinction in isolated regions of these galaxies to disentangle the relative contributions of these phases, the results are ambiguous given the fitting uncertainties in the background light distribution and the presence of clouds deeper in the galaxy with a larger fraction of unabsorbed starlight.

HS00 and HS99 have argued that the extraplanar dust may to a large degree trace a thick disk CNM based on assessments of the physical properties and a comparison of the cloud masses with Galactic molecular clouds (Blitz 1990, 1991). There is complementary evidence in our work for the presence of a thick disk CNM in the form of the extraplanar CO emission detected in NGC 4013 (García-Burillo et al. 1999) and the probable presence of thick disk H ii regions tracing recent star formation (HS00; Rossa & Dettmar 2003a). Candidate thick disk H ii regions can be seen in both NGC 4013 and NGC 4302 (point-like knots of Hα emission visible in Figures 9 and 10). In particular, Figure 9 shows a prominent candidate to the north of the bulge of NGC 4013 (spectroscopy of this nebula is focus of K. Rueff et al., in preparation). Potential areas of star formation suggest a CNM as a component of the thick disk ISM in these galaxies. While the masses of the clouds reported in Table 5 are somewhat lower than those studied in earlier works, this difference is in part due to the conscious choice to study smaller clouds in order to emphasize the small-scale structure revealed by our high-resolution images. The properties of the extraplanar dust-bearing structures in our images are consistent with a significant CNM present in the thick disks of these spiral galaxies. However, the fact that the clouds vastly outnumber the H ii regions so far identified suggests that the fraction of the former that trace molecular clouds able to support star formation (as opposed to "diffuse clouds" in the Galactic parlance) may be small.

While the thick disk gas traced by these extraplanar dust structures largely originates in the thin disk, the production or maintenance of a thick disk CNM can come about in several ways. HS00 and HS99 discussed the possibility that the CNM is consistent with standard theories for multiphase thermal equilibrium of the ISM (e.g., McKee & Ostriker 1977; Wolfire et al. 1995a, 1995b, 2003). Here we add that more dynamical origins may be even more important. As feedback-driven outflows escape from the thin disks of galaxies, they sweep up significant masses of material. Some of this material is compressed at a rate insufficient to shock the gas to high temperatures and/or at a time when the shell walls might radiatively cool. In this case, the shell serves to compress the gas, triggering the formation of a cold or even molecular medium (e.g., McCray & Kafatos 1987). This is, of course, the origin of the self-propagating star formation theories.

In a similar vein, recent work has highlighted the instabilities that can arise in converging flows of neutral material. This can serve to compress and cool the gas (e.g., Heitsch et al. 2008), and the convergence of the walls of adjacent shells may be a trigger for cloud formation (Ntormousi et al. 2011). Our observations of extraplanar dust always show numerous and very complicated structures within the first kiloparsec of the galaxy midplanes, with few readily identifiable supershells. Some of this is undoubtedly due to line of sight confusion. However, given the difficulty of distinguishing physically connected structures in our images, it is not a stretch to imagine some of the complexity we see in the dust occurs as walls of adjacent supershells or other structures converge. Converging flow-type scenarios provide a sustainable method for forming CNM clouds in the thick disk and for reshaping the structures in a less-ordered fashion. The precise mechanism of the formation of CNM clouds in the thick disk is little constrained by our data. However, we note that the associations of cool (Moss et al. 2012) or even molecular gas (Dawson et al. 2011a, 2011b) with high-latitude supershells in the Milky Way offer supporting evidence that the ejection of the gas from the thin disk may also be in part responsible for triggering the formation of a cool medium. The global characteristics and physical conditions for the material traced by the extraplanar dust in our images suggest the presence of a thick disk CNM in addition to warm neutral and ionized gas. In future works we will explore the thick disk star formation, perhaps from this CNM gas, implied by extraplanar H ii regions.

We have presented HST/WFPC2 and ground-based WIYN 3.5 m and LBT 2 × 8.4 m images of the edge-on spiral galaxies NGC 4013 and NGC 4302. These images were used to study the distribution of extraplanar dust and ionized gas in these galaxies. The major results of our work are as follows.

• 1.  
  Our new images reveal extensive distributions of extraplanar dust clouds in NGC 4302 and NGC 4013, with absorbing structures traced to heights z ∼2 kpc in both galaxies. The thick disk dust-bearing gas clouds are seen in large complexes as well as narrow filaments, with significant substructure seen to the limit of our HST resolution (<10–20 pc).
• 2.  
  We derive the properties of several extraplanar dust clouds in each galaxy (Table 5). We estimate hydrogen column densities for individual dust clouds N(H) ≳2 × 10²⁰ cm⁻² and masses ≳10⁴ M_☉, with the larger complexes having quite a bit higher masses. The implied particle densities range from n_H>1.0to4.5 cm⁻³ for the larger clouds, with substructure clouds having even larger densities, n_H>7 cm⁻³. The properties of these individual absorbing structures lead us to conclude that the dust clouds trace a dense phase of the thick disk ISM.
• 3.  
  Our high-resolution Hα images show that the extraplanar ionized gas layers of both galaxies have a diffuse, smooth distribution, with very few filamentary structures. Candidate thick disk H ii regions can be seen in both NGC 4013 and NGC 4302. In NGC 4302, the Hα emission abruptly cuts off at projected radii of R ∼ 7 kpc and 8.5 kpc in the northern and southern halves of the galaxy, respectively.
• 4.  
  Comparing the morphologies of the extraplanar dust clouds to the extraplanar DIG emission, we find little evidence for a direct physical connection between the dust and the DIG. The dust-bearing clouds are strongly filamentary while the DIG layers lack filaments and are smoothly distributed. The features and physical properties of the extraplanar dust and DIG in NGC 4013 and NGC 4302 are distinct.
• 5.  
  The thick disk ISM extending ∼2 kpc away from galactic disks is a multiphase system, including a dense, cold phase traced by dust extinction and perhaps formed through thermal instabilities and/or in converging flows. This thick disk in NGC 4013 and NGC 4302 is ultimately fueled by galactic fountain-like expulsion from the star-forming thin disk that does not destroy the dust as it raises material to high-z.

We thank the anonymous referee for helpful comments on the paper. K.M.R. and J.C.H. recognize support from NASA through grant NNX10AE87G. Support for HST program number 8242 was provided by NASA through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555.

This research has made use of the NASA/IPAC Extragalactic Database (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This research has made use of the VizieR catalog access tool, CDS, Strasbourg, France.