Ultraviolet Halos around Spiral Galaxies. I. Morphology

Edmund Hodges-Kluck, Julian Cafmeyer & Joel N. Bregman hodgeskl@umich.edu

We examine ultraviolet halos around a sample of highly inclined galaxies within 25 Mpc to measure their morphology and luminosity. Despite contamination from galactic light scattered into the wings of the point-spread function, we find that UV halos occur around each galaxy in our sample. Around most galaxies the halos form a thick, diffuse disk-like structure, but starburst galaxies with galactic superwinds have qualitatively different halos that are more extensive and have filamentary structure. The spatial coincidence of the UV halos above star-forming regions, the lack of consistent association with outflows or extraplanar ionized gas, and the strong correlation between the halo and galaxy UV luminosity suggest that the UV light is an extragalactic reflection nebula. UV halos may thus represent of dust within 2–10 kpc of the disk, whose properties may change with height in starburst galaxies.

Subject headings:
galaxies: halos — ISM: dust, extinction — ultraviolet: galaxies
slugcomment: accepted by ApJ11affiliationtext: Department of Astronomy, University of Michigan, Ann Arbor, MI 48109

1. Introduction

Dust plays an important role in galaxies as a catalyst for instellar chemistry, a repository of metals, and a heat source for interstellar gas. The discovery that there is about as much dust outside of galaxies as within (Ménard et al., 2010) suggests that dust also plays an important role in the disk-halo cycle (however, see Smith et al., 2016, who argue for a smaller amount of halo dust). Dust is primarily formed in the atmospheres of evolved stars or supernova remnants, so extragalactic dust was likely transported out of the disk by gaseous flows (e.g., Ferrara et al., 1991). Indeed, dusty outflows have been detected around edge-on galaxies, including filaments seen in extinction (Howk & Savage, 1997, 1999) and dust emission at several kpc above the midplane (McCormick et al., 2013; Meléndez et al., 2015). Still, it remains unclear how and when the dust gets into the halo, how long it remains there, and whether it is altered in the process. These issues motivate further studies.

Extragalactic dust can be studied through extinction, emission, or reflection (scattering). In the first case, extraplanar dust lanes can be seen against the thick disk of stellar light (Howk & Savage, 1997), but above a few kpc from the galaxy midplane the extinction must be measured toward background continuum sources, such as quasars. Since there are few quasars behind any one galaxy, Ménard et al. (2010) stacked many sources to measure the extinction to beyond 1 Mpc. In contrast, thermal emission or reflection will only be seen near the galaxy, since both processes need a nearby light source. This makes edge-on systems the ideal laboratories to isolate extraplanar dust emission and scattered light.

The possibility of extragalactic reflection nebulae (eRN) around edge-on galaxies was raised by Ferrara et al. (1996), and Hoopes et al. (2005) reported the discovery of diffuse ultraviolet (UV) light consistent with an eRN above the disks of the edge-on starburst galaxies M82 and NGC 253. This conclusion was reinforced by Hutton et al. (2014), and Seon et al. (2014) detected a UV halo around NGC 891 that they argue is also an eRN. In Hodges-Kluck & Bregman (2014, hereafter, HKB14) we reported that UV halos are ubiquitous around highly inclined galaxies, but Sandin (2015) pointed out that the light scattered into the wings of the point-spread function (PSF) can masquerade as a physical halo. Shinn & Seon (2015) modeled several of the galaxies in HKB14 including this effect, and found that some of the UV halos we had reported are astrophysical while others are artificial. In principle, eRN are visible at other wavelengths, and some of the H attributed to extraplanar diffuse ionized gas (eDIG) may actually be reflected light (Ferrara et al., 1996; Seon et al., 2014). However, UV halos are especially promising because the scattering cross-section is high in the UV for typical dust compositions and the sky is relatively dark.

In HKB14 we examined galaxies with both Swift UVOT (Roming et al., 2005) and GALEX (Martin et al., 2005) observations, regardless of physical properties. In a series of three papers we now take a more systematic approach to measuring the physical properties of the UV halos around highly inclined spiral galaxies. In this paper (Paper I) we focus on galaxies within about 25 Mpc to determine the frequency of physical UV halos, their morphology, and whether they are eRN. The subsequent papers will present a catalog of total halo fluxes for a wider sample where morphology cannot necessarily be measured (Paper II), and models of the spectral energy distributions (SEDs) for those galaxies with the highest quality data (Paper III).

Name Type SFR(IR)
(deg) (mag) (Mpc) (km s) (mag) (mag) (mag) () erg s) ( yr)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
NGC0253 SABc 5.1 90 -21.23 3.25 189.8 0.019 0.69 3.772 4.36 9.59 3.97
M82 Scd 7.5 76.9 -20.13 3.93 65.6 0.138 0.68 4.665 1.83 15.0 9.42
NGC4631 SBcd 6.6 90 -22.42 6.02 138.4 0.015 0.39 6.465 0.98 15.0 1.03
NGC3628 Sb 3.1 79.3 -21.54 10.89 215.4 0.024 0.68 6.074 4.47 4.59 3.30
NGC4666 SABc 5.1 69.6 -21.10 17.28 192.9 0.022 0.64 7.055 3.10 16.8 4.83
NGC3079 SBcd 6.6 90 -21.56 19.28 208.4 0.01 0.53 7.262 2.91 16.9 8.17
NGC5775 SBc 5.2 83.2 -21.09 20.34 187.2 0.037 0.66 7.763 2.45 0.014 3.97
NGC4388 Sb 2.7 90 -22.13 20.5 171.2 0.029 0.57 8.004 1.55 2.35 3.16
Normal Spirals
NGC0055 SBm 8.8 90 -20.09 1.94 58.7 0.012 0.33 6.249 0.09 3.42 0.06
NGC0891 Sb 3.1 90 -20.37 9.96 212.1 0.058 0.70 5.938 3.99 5.30 2.43
NGC2683 Sb 3.0 82.8 -20.42 10.08 202.6 0.029 0.75 6.328 2.98 5.56 0.41
NGC4517 Sc 6.0 90 -21.46 10.56 139.6 0.021 0.53 7.329 0.73 0.34
NGC4565 Sb 3.3 90 -22.55 12.18 243.6 0.014 0.68 6.060 5.65 2.20 0.80
NGC4096 SABc 5.3 80.5 -20.39 12.68 144.8 0.016 0.50 7.806 0.088 5.18 0.56
NGC4313 Sab 2.1 90 -20.16 14.62 117.6 0.033 8.468 0.37
NGC3623 Sa 1.0 90 -21.02 12.77 231.2 0.022 0.78 6.066 7.17 5.76 0.38
NGC5907 SABc 5.3 90 -21.08 16.37 226.6 0.009 0.62 6.757 5.03 14.1 2.04
NGC4216 SABb 3.0 90 -20.80 16.78 244 0.028 0.83 6.524 6.06 0.44
NGC4607 SBbc 4.0 90 -20.18 17.78 98.9 0.028 0.75 9.584 0.60 1.17 0.68
NGC4522 SBc 5.9 79.2 -20.91 18.29 96.4 0.018 9.8 1.67 0.40
NGC0134 SABb 4.0 90 -21.49 18.71 220.2 0.016 0.69 6.844 5.95 4.51
NGC4157 SABb 3.3 90 -19.88 18.7 188.9 0.019 0.64 7.363 3.03 8.11 2.71
ESO358-063 Scd 6.9 75.6 -20.34 18.98 135 0.005 0.7 9.144 0.61 0.87
NGC4217 Sb 3.1 81 -20.08 19.37 187.6 0.015 0.75 7.582 3.67 3.08
NGC4330 Sc 6.0 78.9 -20.02 19.61 115.7 0.021 9.51 0.36
NGC3044 SBc 5.5 90 -20.32 22.48 152.6 0.022 0.6 8.982 1.01 15.1 2.77
NGC5170 Sc 4.9 90 -21.13 26.8 244.7 0.07 0.7 7.628 7.46 0.79

Note. – Cols. (1) Name (2) Morphological Type (3) Morphological type code (4) Inclination angle (5) magnitude (6) Distance (7) Circular rotational velocity (8) Foreground Galactic extinction (9) color (10) -band magnitude from the 2MASS Extended Source Catalog Schlafly & Finkbeiner (2011) (11) Stellar mass using relation from Bell & de Jong (2001) (12) H Luminosity (13) Star-formation rate estimated from the Kennicutt (1998) relation  yr. we measure as defined by Rice et al. (1988) , where the fluxes at 12, 25, 60, and 100 m are in Jy from the IRAS catalog. References. – Values from the NASA/IPAC Extragalactic Database, (http://ned.ipac.caltech.edu/), HyperLeda Makarov et al. (2014) and the IRAS catalog (http://irsa.ipac.caltech.edu/Missions/iras.html).

Table 1Basic Galaxy Parameters

The remainder of this paper is organized as follows: Section 2 describes our sample and data sources, and Section 3 is focused on correcting for galactic light scattered into the PSF wings. In Section 4 we describe the morphology of the UV halos and compare them to diffuse halos at other wavelengths, and in Section 5 we use information from the halo fluxes in each filter to characterize the UV halos. We interpret the results in Section 6 to argue that the UV halos are likely eRN and connect the UV measurements to extinction and emission measurements. Section 7 summarizes the paper and presents our main conclusions.

2. Data

Name Exposure Time (s) Sensitivity (AB mag) H Radio 21cm
FUV NUV uvw2 uvm2 uvw1 FUV NUV uvw2 uvm2 uvw1
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14)
NGC0253 3281 15043 25.0 26.1 5 13
M82 14707 29422 91603 147713 92788 25.9 26.4 26.8 25.7 26.5 10 21 24
NGC4631 3147 3147 1377 5474 1123 24.9 25.1 24.0 24.7 23.6 10 21 19
NGC3628 5812 17076 8095 5530 3546 25.4 26.0 25.2 24.8 24.3 21 6
NGC4666 5940 5940 19561 19610 9858 25.4 25.3 24.7 25.7 25.1 21 18
NGC3079 16108 16108 8512 7252 273 26.3 26.1 25.4 25.1 22.5 23 21 7
NGC5775 2587 5770 14978 23677 4180 24.6 25.4 25.6 25.8 24.5 4 21 8
NGC4388 2538 4993 10331 7334 5331 24.6 25.3 25.4 25.0 24.7 16 21 3
Normal Spirals
NGC0055 29347 30691 25.5 26.5 20
NGC0891 6047 6283 15353 15200 9787 25.0 25.2 25.7 25.5 25.2 16 21 14
NGC2683 1600 1600 6407 5644 4824 24.3 24.5 25.0 24.8 25.3 16 21 17
NGC4517 1906 6403 23.8 25.5
NGC4565 12050 25.9 21 25
NGC4096 1650 1650 2670 3575 24.3 24.7 24.6 24.5 10
NGC4313 3862 3862 24.2 25.2
NGC3623 1656 1656 5300 7122 3193 24.3 24.6 25.2 25.1 24.5 10 5 6
NGC5907 1543 5423 31888 45601 26482 24.2 25.5 26.3 26.3 26.0 15 21 1
NGC4216 1672 2604 24.3 24.9 16 5 3
NGC4607 2701 6181 24.7 25.5 11 3
NGC4522 2496 2496 24.7 24.9 11 9
NGC0134 5998 6336 4032 25.1 24.9 24.6
NGC4157 2494 1288 24.7 24.5 23 21 22
ESO358 2456 3197 24.7 25.2
NGC4217 11774 11774 25.8 25.9 15 21 1
NGC4330 3862 3862 2429 2429 2279 25.1 25.2 24.3 24.1 24.0 3
NGC3044 1676 1676 2910 2910 2585 24.3 24.6 4 21 12
NGC5170 1606 3307 24.0 24.9 2

References. – H and radio references: (1) Allaert et al. (2015) (2) Bottema et al. (1987) (3) Chung et al. (2009) (4) Collins et al. (2000) (5) Condon (1987) (6) Haynes et al. (1979) (7) Irwin et al. (1987) (8) Irwin (1994) (9) Kenney & Koopmann (2001) (10) Kennicutt et al. (2008) (11) Koopmann et al. (2001) (12) Lee & Irwin (1997) (13) Lucero et al. (2015) (14) Oosterloo et al. (2007) (15) Rand (1996) (16) Rossa & Dettmar (2003a) (17) Vollmer et al. (2016) (18) Walter et al. (2004) (19) Weliachew et al. (1978) (20) Westmeier et al. (2013) (21) Wiegert et al. (2015) (22) Yim et al. (2014) (23) Young et al. (1996) (24) Yun et al. (1994) (25) Zschaechner et al. (2012) Note. – Cols. (2-6) Exposure times (7-11) 3 point-source sensitivity (12-14) References for non-UV data used here.

Table 2Observations

Our goal is to measure the structure of the diffuse extraplanar UV light, so we limit the sample to nearby, highly inclined, spiral galaxies. The initial sample (drawn from the HyperLeda database111http://leda.univ-lyon1.fr/; Makarov et al., 2014) includes galaxies within 25 Mpc with an inclination angle of , a Hubble type of Sa or later, and an absolute -band magnitude of  mag. The distance limit is based on the need for high in small regions. We excluded some galaxies for which the inclination angles or types were obviously incorrect, and for our distance cut we used the redshift-independent distances from NASA/IPAC Extragalactic Database222http://ned.ipac.caltech.edu/ instead of the HyperLeda values. The initial sample includes 78 galaxies.

Each galaxy must have moderately deep data in at least one of the GALEX or Swift UVOT filters: GALEX FUV (1516Å) and NUV (2267Å), or the UVOT uvw1 (2600Å), uvm2 (2246Å), and uvw2 (1928Å) filters. Most of the data come from the GALEX or Swift archives, but we also obtained new UVOT data for several targets through the Swift program GO 1013198. Based on our prior work, we expect a UV halo flux that is about 1% of the total galaxy flux, so to accumulate enough we require exposure times greater than 2-4 ks in either GALEX filter or 5-6 ks in any UVOT filter (this excludes the 600 s exposures from the GALEX all-sky imaging survey). We excluded galaxies that are in regions with strongly variable Galactic cirrus or are in very crowded fields. Swift targets must also fit within the  arcmin field of view for accurate background subtraction. The 26 galaxies that meet this criteria form our working sample, and the basic properties of these objects are given in Table 1. The UV exposure times and point-source sensitivities are given in Table 2. The observation IDs used in this paper are given in the Appendix.

Our reliance on archival data means that the sensitivity and filter coverage vary widely among the galaxies. We also note that a disproportionate number of galaxies are in the Virgo Cluster, which has been observed extensively. Some of the galaxies are included in published work on UV halos (Hoopes et al., 2005; Hutton et al., 2014; Seon et al., 2014; Hodges-Kluck & Bregman, 2014; Shinn & Seon, 2015; Baes & Viaene, 2016).

We processed the data for each galaxy as described in HKB14, including cleaning the UVOT and GALEX images of diffuse artifacts in each exposure. We combined the Swift exposures for each filter into a single, exposure-corrected image. We masked all the point sources outside the galaxy that were detected at 3 or greater, and we also masked each source detected at 4 or greater in any one filter in all the remaining filters. This helps to exclude stars that are close to the galaxy but are not formally detected at different wavelengths or in shorter observations.

3. Extended PSF Wings and Halo Light

Figure 1.— Measured, radially-averaged PSF profiles for each filter (filled circles) and the best-fit profile (black solid line). The fit is based on points between the vertical dashed lines, and the profile is extrapolated from the data starting halfway between these lines. The statistical 1 error bars on the fit are shown as dashed lines. For comparison, the profile is shown as a red line. The last panel shows the extrapolated profiles to a radius of 300 arcsec for each filter.

Sandin (2015) showed that the diffuse light around edge-on galaxies consists in part of light from the galaxy image scattered into the wings of the PSF (the Airy pattern), since the sum of many overlapping Airy patterns leads to a diffuse glow around the galaxy. To measure and subtract this component, a PSF must be measured to at least twice the angular size of the galaxy. Hereafter, we refer to this light as “PSF-wing contamination.” Shinn & Seon (2015) were the first to address this issue in the GALEX FUV filter, and here we extend their analysis by measuring and extrapolating the PSFs for each filter in order to determine the PSF-wing contamination in the measured halo fluxes. We want to obtain accurate 2D maps of the halo rather than only a profile along the minor axis, so the typical approach of convolving a model with the PSF is not sufficient.

3.1. Measured PSFs

The GALEX and Swift calibration teams have provided PSFs measured to radii of about 60 and 30 arcsec, respectively333see http://www.galex.caltech.edu/researcher/techdoc-ch5.html and http://heasarc.gsfc.nasa.gov/docs/heasarc/caldb/swift/docs/uvot/, but for the galaxies in this work the PSFs need to extend to 200-600 arcsec. Extrapolating the wings from the available PSFs may not be sufficient, so we measured PSFs from bright point sources to better characterize the wings.

For GALEX, we used deep observations of the quasars 3C 273 (30 ks, tile GI4_012003_3C273) and PKS  (32 ks, tile PKS2155m304), which are relatively isolated on the sky and not bright enough to produce strong ghost images or diffraction spikes. These quasars are too bright to measure extended PSFs with Swift, so we used a combined image of Mrk 501, which has an effective exposure of 129 ks in uvw1, 131 ks in uvm2, and 170 ks in uvw2. The data were processed as described above and in HKB14, except that we used at least fifteen sources to compute the astrometric solution for each image and we rejected images where the source was outside a few arcminutes from the chip center. The observation IDs we used are given in the Appendix.

Figure 2.— Left: The FUV image of NGC 3079 before correcting for galactic light scattered into the PSF wings. The contours are 3, 4, and 5 above background. Right: The same image after correction for PSF-wing contamination. The galaxy image used to compute the correction has been masked.

We then measured the PSF by constructing an azimuthally symmetric radial profile to  arcsec. The actual PSF is not azimuthally symmetric, but simulations show that unless it is very asymmetric (to an extent that would be noticeable on the image), this approximation does not noticeably change the measured PSF-wing contamination for any galaxy with a projected size of about an arcminute or more (even at smaller sizes, it will not usually make a difference). The background level is crucial to measuring the small amount of light scattered by the PSF to large radii, so we masked all point sources detected using the Swift task uvotdetect with a 3 detection threshold (based on SExtractor; Bertin & Arnouts, 1996) other than the target. Since these sources are much fainter than the target, we assume that masking to the nominal 99% encircled energy radius is sufficient.

The measured profiles are shown in Figure 1, along with fitted profiles of the form , where , , and are determined using least-squares fits to the data between the dashed vertical lines in Figure 1 (in each plot, has been subtracted). For comparison, we also plot a fitted profile where . The extrapolation based on the best-fit model begins at the midpoint between the dashed vertical lines, but our results are insensitive to where we match the extrapolated profile to the measured one. The FUV is close to at all radii, but the other filters concentrate light within about 30 arcsec, with the subsequent decay being steeper than .

The best-fit exponents are given in Table 3. The fit depends strongly on , so we verified that the background level in the fitting zone was consistent with the background measured outside the fitting zone. We also repeated the analysis using a 2 threshold for point source removal and measured values that are consistent with the earlier measurements. Finally, we verified that the PSFs successfully clean the outer regions in the source images and other images of bright point sources (within 10 arcsec an azimuthally symmetric PSF is a poor match to the data).

There are two caveats to this analysis. First, the PSF becomes increasingly distorted with increasing distance from the optical axis, so for images “far” from the optical axis the assumption of radial symmetry is badly wrong (where the critical distance depends on the instrument). The PSF is naturally hard to measure to large radii far from the optical axis, and there are few bright sources that are both near the chip edge and have very deep GALEX or Swift data. Thus, for galaxies near the chip edge the measured PSFs may not accurately account for the PSF-wing contamination.

Filter Exponent Filter Exponent
FUV uvm2
NUV uvw1

Note. – Best-fit exponents ( in ) and statistical errors from the PSF profile fitting.

Table 3PSF Wing Exponents

Second, the PSF may vary with time (Sandin, 2015, although the variability is probably much larger for ground-based instruments). We could not make useful tests with GALEX data because the bright sources we used for comparison typically had different PSF cores that were obviously due to the different positions on the chip. For Swift we selected all exposures of Mrk 501 in a given filter where the source was within 1 arcmin of the chip center, and searched for variability in the best-fit between combined images taken in different months. We found no clear variability, but some low level of variability beyond 50 arcsec is possible. We note that the Swift PSF core does vary with time in the sense that the full-width at half maximum depends on the temperature of the UVOT focusing rods444https://heasarc.gsfc.nasa.gov/ftools/caldb/help/uvotapercorr.html. However, we did not find any variation in the best-fit exponent for the PSF wings.

Our results are consistent with the best-fit exponents found in other sources where the PSF cannot be measured to such large distances, so we believe that the fits are reasonably accurate.

3.2. Estimating the PSF-wing Contamination

We estimated the PSF-wing contamination for each image by creating a model using the input image and the extrapolated PSF models, using the best-fit exponent for each filter (Table 3). We first clipped the input image to the region that just contains the galaxy. We did not use an objectively defined region because of the varied disk morphologies and inclinations in our sample, but the galaxy regions are conservative in that they extend to at least 2 kpc above the midplane. We then convolved the clipped image with a PSF model that was extrapolated to at least twice the angular diameter of the galaxy.

The resulting image shows the amplitude of the PSF-wing light at each position around the galaxy, but the total flux in the model image is too low and the light is too smeared out because we do not deconvolve the galaxy image before convolving it with the extrapolated PSF (the galaxy is too complex for accurate deconvolution). The former effect is corrected by multiplying the fake image by the ratio of the flux in the galaxy in the original and fake images (this is not an exact correction, but simulations of exponential disk models show that it underestimates the PSF-wing flux by much less than the statistical uncertainties for our sample). The latter effect cannot be easily corrected, but we find (again using simulations of exponential disk models) that the most severe impact is within about 10 pixels of bright regions in each galaxy, and in practice the galaxy regions we define extend at least this far from bright clumps. The combination of these issues leads to an overestimate of the PSF-wing contamination by about 10% immediately adjacent to the galaxy, and which quickly declines with height. We also investigated the error due to using a symmetric PSF in the FUV, where the PSF is measured to over 100 arcsec. We convolved model images with angular sizes smaller than this with our model PSFs and the measured PSF. This leads to differences of 0-3% in the PSF-wing light measured around the galaxy. On the scale of the whole halo, these effects are very modest.

The model image can then be used in conjunction with the original image to measure the PSF-wing contamination at any point in the halo, and we also subtract the model from the original to create corrected halo images. The PSF-wing contamination differs for each galaxy and filter, but it ranges from 1-90% of the total extraplanar flux in our sample. The contamination is generally larger in the FUV than the other filters. The importance of the correction is illustrated in Figure 2 for NGC 3079, where we show FUV images before and after subtracting the model image. We verified the Sandin (2015) proposition that convolving the galaxy image with a PSF extrapolated to more than twice the angular size of the galaxy does not significantly alter the PSF-wing contamination. In Section 5 we assess the reliability of our PSF-wing subtraction.

Figure 3.— UV halos (orange) around normal galaxies in the GALEX and UVOT filters, shown with increasing wavelength from left to right. The final column shows the combined data, referenced to the GALEX resolution. The orange maps have been corrected for the PSF-wing contamination from the galaxy (image shown in blue) and point sources have been masked. For each filter, the contours are the 2, 4, 8, and 16 contours above the background. A 2 arcmin line is shown at the bottom left of each row, and the 3 point-source sensitivity for the map is shown at the bottom right of each panel. The composite image only shows the 3 contour.
Figure 3.— continued
Figure 3.— continued

4. Extraplanar Ultraviolet Morphology

In this section we describe the qualitative structure of the diffuse extraplanar UV light in our sample and compare it to the diffuse structure seen at other wavelengths where the source of the light is basically understood. The initial work on UV halos (Hoopes et al., 2005) focused on galaxies with galactic superwinds, and in this section we divide our sample into galaxies with known winds (starburst galaxies) and those without (normal galaxies).

4.1. Existence of UV Halos

We use UV images where the PSF-wing contamination has been subtracted. Figures 3 and 4 show the images in each filter for the normal and starburst galaxies, respectively. In these images, the galaxy region used to estimate the PSF-wing contamination is masked, so we superimpose the unsmoothed image of the galaxy for that filter in this region (in blue). Outside of this region, point sources have been masked, the image has been smoothed with a Gaussian kernel (with  pixels, depending on the angular size of the galaxy), and the image has been clipped to 1 above the background. The contours show the 2, 4, 8, and 16 level above the background. For each image we also list the 3 point-source sensitivity in AB magnitudes (Oke & Gunn, 1983), which shows the differences in depth between filters.

Although the UVOT and GALEX images have different PSFs, artifacts, and systematic uncertainties, the morphology of the residual extraplanar UV light is similar between filters for most galaxies, so UV halos appear to be physical phenomena. Thus, in the final columns of Figures 3 and 4 we show a composite image for each galaxy made by summing each filter (referenced to the effective area of the uvm2 filter). These images have roughly twice the of the images in each filter, so they highlight the extent of the halo. Remarkably, every galaxy in the sample has a UV halo extending at least to 6 kpc above the midplane. Considering that our selection criteria were distance, luminosity, and inclination, this suggests that UV halos are a common feature of luminous spiral galaxies.

There are some differences between filters for several galaxies that may be artificial. For example, in the uvw2 images of NGC 4666 and NGC 5775 (Figure 4), the halo appears to be more prominent than in the uvm2 or FUV images, even though the uvm2 has a higher sensitivity. This may indicate residual light in the central 2 arcmin halo ring in the large-scale uvw2 scattered-light artifact (HKB14), which could mimic a halo in an image centered on the galaxy. The smaller field of view of the UVOT can also lead to differences: in the relatively shallow UVOT images of NGC 4631 (Figure 4), the halo is both less prominent and more symmetric than in the GALEX images. This probably results from the “background” regions in the UVOT being inside the UV halo itself, a scenario that is supported by the GALEX images. In other cases, differences between the morphology in each filter likely indicate physical differences. For example, in M82 and NGC 253 (Figure 4), the NUV halo appears more like a thick disk than the FUV halo, which is more filamentary in nature. This morphology cannot be explained by PSF-wing contamination or filter artifacts.

Figure 4.— UV halos (orange) around starburst galaxies. The format is the same as in Figure 3.

We investigated the contribution of stellar clusters or background sources to the diffuse light in those galaxies with HST observations in the band or bluer. For these objects, we identified point-like sources in the HST maps that were not already identified and removed in our maps and predicted a Swift or GALEX flux for each, assuming that each source has the spectrum of a B star. We then compared the flux to the total diffuse UV flux, and found that it can explain at most a few percent of the light. The exception that proves the rule is NGC 4522, where the Virgo cluster is stripping the ISM (Kenney & Koopmann, 1999) and there are known stellar clusters in the halo. In this case, most of the flux in the FUV comes from these clusters. Even NGC 4522 has a diffuse component, however, so we conclude that all of these galaxies have true diffuse UV halos.

4.2. Ultraviolet Morphology

There are three qualitative morphological metrics we use to compare the UV halos in our sample: filamentary structure, concentration of the UV light (both in the vertical direction and along the disk), and symmetry across the midplane. In general, these are consistent among filters for a given galaxy.


Filaments are most easily visible in the FUV, and they are seen in most starburst galaxies (M82, NGC 253, NGC 3079, NGC 3628, NGC 4666, and NGC 5775), as well as two normal galaxies (NGC 4522, NGC 4330). The filaments in the latter two are associated with gas stripped by ram pressure in the Virgo cluster. We cannot rule out the presence of filaments in non-stripped normal galaxies, but those galaxies with high images do not have filaments.


The UV halos differ in extent, both in galactocentric radius and projected height. For example, in NGC 891, NGC 2683, and NGC 3044 most of the light is concentrated near the center of the galaxy (Figure 3), whereas in NGC 4216 or NGC 5907 the surface brightness is comparable across most of the disk. We examined the degree of concentration by dividing the UV halo at the midplane and projected galactic center into four quadrants, which were summed. Figure 5 shows the resulting images. Each panel has dimensions of , where we use the HyperLeda values for . The final two panels show the sum of all of the normal and starburst galaxies, respectively, where the images were reprojected and stacked on a common axis relative to . We excluded the stripped galaxies NGC 4330 and NGC 4522 from this stack.

Figure 5.— Combined images from each filter stacked in the first quadrant and shown relative to the optical radius, . The contour represents the combined 3 threshold above background, and the dimensions of each image are . The dashed lines show . Starburst galaxies are labeled with red names. The final two panels show the combined normal and starburst galaxies, respectively, with scale heights measured as a function of galactocentric radius shown as points. The “normal” stack excludes NGC 4330 and NGC 4522, which are being stripped by ram pressure. The dip at the center of NGC 4631 is due to masking a satellite galaxy, but it is accounted for in the starburst stack and does not cause the central dip in the scale height there.

There are two aspects to the concentration: the extent of the light (as measured by the 3 contour in Figure 5) and its distribution. Among the UV halos of normal galaxies, 6/18 extend in the radial direction to and only one (NGC 4522) extends beyond in the vertical direction. In contrast, all of the starburst halos (8/8) extend to in the radial direction, and 6/8 extend to in the vertical direction. In terms of distribution, for most normal galaxies the UV light is brightest and most vertically extended above the galactic center, with a modest decline with increasing galactocentric radius. This is consistent with a disk-like structure. The starburst galaxies other than NGC 4631 are also less concentrated and filamentary structure at larger heights is superimposed. The similarity in sensitivity and distance between several of the starburst and normal galaxies (e.g., NGC 3079 and NGC 5907) indicates that UV halos are intrinsically brighter and more morphologically complex around starburst galaxies.

When compared to near-IR images, the radial extent of the UV halo along the major axis (relative to ) is unrelated to the prominence of the bulge, so it is unlikely that the radially concentrated UV halos are merely light from the bulge outskirts. The radial or vertical concentration is also unrelated (within each subsample) to the total halo luminosity. The radial concentration also appears to be unrelated to the vertical concentration, and the normal galaxies appear to span a narrow range in vertical extent below .

Bilateral Symmetry

We determined the degree of bilateral symmetry by folding images across the midplane. Most galaxies show some asymmetry, although in most cases the visible structure is symmetric while the flux is not, which could be a function of the inclination angle. Among the normal galaxies the bilateral asymmetry is only pronounced for NGC 134, NGC 4330, and NGC 4522, which are experiencing ram-pressure stripping (Kenney & Koopmann, 1999; Abramson et al., 2011). Among the starburst galaxies there is asymmetry in the filamentary structures, but overall the halos are nearly symmetric. An exception is NGC 4631, in which the northern part of the halo is somewhat more extended. NGC 3628 is not bilaterally symmetric, but its halo appears to be warped; if we invert the image along the axis of the midplane, it becomes nearly symmetric. A hint of a similar warp can be seen in the combined image of NGC 4388.

4.3. FuvNUV Color

In addition to structures visible in each band, we looked for spectral structure in the UV halos. In regions where the halo is detected in all five filters we can construct SEDs, but most galaxies in the sample do not have enough data. Thus, we adopt the FUVNUV color as a proxy. Figures 6 and 7 show the FUVNUV color in regions where the flux is detected at 2 above background in each filter for normal and starburst galaxies, respectively. In these images, the smoothing kernel is usually larger than in Figures 3 and 4. The colorbar in the plots is such that blue regions correspond to a blue spectrum (small FUV-NUV) and red regions correspond to a red spectrum (large FUV-NUV). However, values at the edges are not reliable as they are sensitive to the background subtraction (e.g., the red boundaries on the halo of NGC 55 in Figure 3).

Figure 6.— The FUVNUV color (per square arcsec) for normal galaxies. The FUV and NUV images used to create these maps were smoothed and clipped at the mutual 2 level (i.e., where emission is detected at 2 above background in each image). The GALEX pixel size is 1.5 arcsec.
Figure 7.— The FUVNUV color for starburst galaxies. The plot style is the same as in Figure 6.

There is no obvious structure in the FUVNUV color maps for most normal galaxies. There appears to be a tendency for the Sc-type galaxies (NGC 5907, NGC 4522, NGC 4330, and NGC 3044) to have bluer halos than the rest (which are Sb, except for NGC 55), but the strength of this connection cannot be assessed with the current sample. This is especially true since NGC 4522 and NGC 4330 are being stripped of their gas by ram pressure from the Virgo intracluster medium. For galaxies where the is reasonably high across the disk (NGC 55, NGC 891, NGC 4217, and NGC 4330), the color does not change much across with projected galactocentric radius. However, in the starburst galaxies the filaments and regions associated with galactic winds are bluer than the rest of the halo. The underlying FUV-NUV color differs between galaxies, with NGC 3628 and NGC 4388 having generally redder halos; both of these galaxies are classified as Sb type, whereas the others are Sc or Sd. Combined with the normal galaxies, this suggests that bluer galaxies tend to have bluer halos, but the sample size is small. In M82 the FUV-NUV color tends to decrease with height, but the filamentary structures within the wind that are visible in both the FUV and NUV images do not appear in the FUV-NUV color map. In other words, small UV structures in the wind appear to have the same color as the rest of it. It is also of interest that the bluer regions in the halo of NGC 4631 are asymmetric across the midplane, although both sides are blue in the central region where the wind occurs.

4.4. Multiwavelength Comparison

We compare the diffuse UV halos to diffuse extraplanar light in other bands, as well as the disk emission in the same bands. These include H, soft X-rays, 160 m emission, radio continuum (at 1.4 or 4.8 GHz) and 21 cm emission from neutral hydrogen. The H traces the diffuse ionized gas in and above the disk, and the presence of extraplanar diffuse ionized gas (eDIG) indicates disk porosity and ongoing star formation (it may also produce UV emission nebulae through ionized helium). We use H images from several atlases (Lehnert & Heckman, 1996; Rand, 1996; Collins et al., 2000; Rossa & Dettmar, 2003b). Diffuse soft X-rays trace hot gas, and X-ray halos seen near the disk are primarily from outflows connected to star formation (Li & Wang, 2013). Within the disk, some of the X-rays also come from X-ray binaries. We use X-ray data from the Chandra and XMM-Newton archives, and the presence of X-ray halos is determined from the literature (Strickland et al., 2004; Tüllmann et al., 2006; Li & Wang, 2013). The 160 m images trace thermal emission from dust, and we obtain them from the Herschel archive. We use 160m because it is near the expected peak of the modified blackbody curve for thermal emission, and the spatial resolution at longer wavelengths is considerably worse. The 160m images are dominated by the disk and the PSF is about 6 arcsec, so we adopt the same approach to clean the overlapping Airy patterns and isolate extraplanar emission (except in NGC 4522). We use the PSF from Aniano et al. (2011), and in some cases the galaxy box is smaller than for the UV. We also refer to McCormick et al. (2013) regarding extraplanar polycyclic aromatic hydrocarbon (PAH) emission and Howk & Savage (1997) and Howk & Savage (1999) for the presence of extraplanar dust seen in absorption as filaments in the optical band. The radio continuum emission comes from cosmic rays produced in supernova remnants, which diffuse out of the disk. We primarily refer to data taken as part of the CHANG-ES program (Wiegert et al., 2015), but some images come from prior atlases (especially Condon, 1987). Finally, we use or refer to high resolution 21 cm maps from a variety of papers (Weliachew et al., 1978; Haynes et al., 1979; Irwin et al., 1987; Irwin, 1994; Yun et al., 1994; Lee & Irwin, 1997; Kenney & Koopmann, 2001; Walter et al., 2004; Oosterloo et al., 2007; Chung et al., 2009; Zschaechner et al., 2012; Westmeier et al., 2013; Yim et al., 2014; Allaert et al., 2015; Lucero et al., 2015; Vollmer et al., 2016). These maps often show filaments of cold gas or warped disks.

Figure 8.— A comparison of the UV halos to diffuse emission in other bands for normal galaxies. Empty panels indicate where we have no image file, but in some cases we can compare the morphology from published images (see Table 2). The orange maps show the FUV (or uvw2) contours from Figure 3 while the blue maps show images at the other wavebands. The thick black contour is the 3 contour above background for each of these images. The thin black box shows our galaxy region for PSF-wing contamination. To emphasize the extraplanar emission at 160 m we also corrected these images for PSF-wing contamination, and the dotted lines show the galaxy regions used. In some images the map edges are visible as regions of heightened noise, and the contours there should be ignored.
Figure 8.— continued
Figure 8.— continued

H, soft X-rays, thermal dust emission, and the radio continuum are associated with ongoing star formation, whereas H i need not be. Figures 8 and 9 show the UV contours from above overlaid on H, X-ray, 160 m, radio continuum, and 21 cm images for normal and starburst galaxies, respectively. In some cases there is no publicly available data, but we can usually rely on images in the literature (this is especially true for 21 cm data). Galaxies are not included in these figures if they do not have publicly available FITS files in at least two bands. As with the UV images, the is variable among the galaxies and bands, which limits the power of this comparison.

Extraplanar Diffuse Emission: Frequency

About 40% of non-starburst galaxies have eDIG (Rossa & Dettmar, 2003a), and a similar number of galaxies in our sample with data from these bands (Table 6) have eDIG (38%), radio (45%), or soft X-ray halos (40%). However, 67% of galaxies with 160 m images have extraplanar emission (sometimes at lower heights than the UV halos; Figure 8). The sample overlap with McCormick et al. (2013) is not large enough to determine the fraction with extraplanar PAH emission. In contrast, UV halos are seen around every normal galaxy. Several of the galaxies in our sample were also investigated by Howk & Savage (1997) and Howk & Savage (1999), who used unsharp mask images in the optical to identify high latitude dust absorption. These include NGC 891, NGC 4517, NGC 4565, NGC 5907, NGC 4217, and NGC 4157, of which NGC 891 and NGC 4217 have high latitude dust. The starburst galaxies all have eDIG, soft X-ray halos, extraplanar thermal dust emission (except for NGC 3628), and radio halos. H i emission is also seen around all galaxies in the sample with high resolution maps, although in some cases the beamsize makes it unclear how much of the emission is truly extraplanar.

Extraplanar Diffuse Emission: Morphology

The extraplanar diffuse emission in normal galaxies shows some of the same features as the UV halo morphology, although sometimes it is only detected at lower heights where we cannot measure the UV halo morphology directly. The eDIG tends to have a similar morphology as the UV in terms of filaments, concentration, and bilateral symmetry, but it is usually not detected at larger heights where the UV halo remains bright. The soft X-ray halos correspond reasonably well to the UV halo structure in NGC 891, NGC 2683, and NGC 4217, but not in NGC 5170. There are fewer normal galaxies with X-ray data than with H images, and only four with X-ray halos, so it is unclear how well the morphology matches the UV. The extraplanar 160 m emission coincides with the UV, but it is not always as extensive. NGC 55, NGC 891, and NGC 4096 have the most convincing 160 m halos, but their prominence does not appear to be connected to that in H or the radio continuum. Likewise, the prominence of the radio halos seems unrelated to the prominence of the UV halo or extraplanar emission in other bands. The only normal galaxies in McCormick et al. (2013) with extraplanar PAH emission are NGC 55 and NGC 891. In the former case, the PAHs do not follow the UV halo, whereas in the latter they do. NGC 891 and NGC 4217 have high latitude dust seen by Howk & Savage (1999). Although these two galaxies do have brighter UV halos than others in the Howk & Savage (1999) sample, the UV halos in the four galaxies without high latitude dust absorption have about the same vertical extent and radial concentration.

Starburst UV halos have filamentary structure that is also seen at other wavelengths. This is especially true in H, although a similar pattern is seen in X-ray and radio halos. There are some disagreements in extraplanar morphology among these bands (such as in M82, NGC 4631, or NGC 4388), but there is a clear connection between the UV filaments and those seen at other wavelengths. The same is true for the extraplanar dust emission, although not every filament is associated with 160 m emission. However, unlike at other wavelengths the UV halos span the disks of the starburst galaxies, and this non-filamentary structure is not present in H, X-rays, thermal dust emission, or radio continuum. In general the PAHs do not follow the UV halos (McCormick et al., 2013). Some of the same filaments are visible in PAHs, but the coincidence of UV and PAH filaments is worse than at other wavelengths, and the diffuse components above the disk have different shapes.

The connection between the 21-cm morphology and the UV halos is unclear. We note that the sensitivities and resolutions of the H i images in Figure 9 differ, so the  cm contour (assuming optically thin hydrogen) is shown as a shorthand for the transition from disk to halo gas. The H i and UV morphology are similar for the galaxies undergoing ram-pressure stripping (NGC 134, NGC 4522, NGC 4330, and NGC 4388). The galaxies in the Virgo cluster also tend to have less extensive extraplanar H i, but there is no clear difference between the UV halos of the Virgo galaxies (NGC 4313, NGC 4216, NGC 4330, NGC 4607, NGC 4522, NGC 4217, NGC 4388, and NGC 5775) and the others. There is also no apparent connection between the UV halo morphology and H i warps, or between the vertical concentration in the H i total column maps and the diffuse UV images.

Figure 9.— A comparison of the UV halos to diffuse emission in other bands for starburst galaxies. The plot style is the same as in Figure 8.
Figure 9.— continued

The UV and H i filaments do coincide near the disk in the starburst galaxies NGC 253, NGC 3628, NGC 4666, NGC 5775, and NGC 3079 (not all shown in Figure 9). The UV halo is also brighter and more vertically extended on the side of the galaxy with the large H i filament in NGC 891, and a similar phenomenon is seen in NGC 3044. However, the large-scale hydrogen filaments seen around galaxies in our sample (such as NGC 55, NGC 3628, NGC 4565, NGC 4631, or M82) are not visible in the UV.

Disk Emission

There is good agreement in most normal galaxies between the projected radial extent (along the disk) of the bright H disk emission and the UV halo above it, which mostly explains the difference in UV halo concentration among normal galaxies. The same is true for soft X-rays, radio continuum, and thermal emission from dust in the disk, except for NGC 4565 and NGC 4522 (Figure 8). In the starburst galaxies, the disk emission is less connected to the UV halo and is instead brightest in the regions where the galactic winds occur.

4.5. Summary

UV halos around luminous spiral galaxies are ubiquitous. In normal galaxies, the UV halos tend to form a thick disk. Their presence is independent of diffuse extraplanar material seen in other wavebands, but in those cases where a diffuse extraplanar layer does exist, it has a similar morphology to the UV halo and occurs in about the same place. This suggests that the UV halos arise from some gaseous medium, which is supported by the asymmetric halos in NGC 4522 and NGC 4330 that match the morphology of the stripped gas, but not the morphology of the stellar disk. The FUVNUV colors differ between galaxies, but there are no clear structures in the FUVNUV colors, except perhaps in stripped galaxies. The UV halos do appear to be connected to star formation in the disk, as they are brightest over the parts of the disk with bright H, X-ray, or radio emission, but not in a way that requires ionizing photons or winds to escape the disk.

UV halos around starburst galaxies are more extended and have filamentary structures coincident with the galactic winds. Apart from these filaments, there is little agreement between the extraplanar or disk emission in other wavebands. The winds are visible in the FUVNUV maps as bluer regions in the UV halo, and may be superimposed on a disk structure like that in the normal galaxies, but the origin of the color difference is not clear.

5. Ultraviolet Halo Fluxes

In the prior section we focused on qualitative morphological trends. Here we compare the average halo fluxes and FUVNUV colors, as well as the flux as a function of height above the disk, to various metrics for each galaxy. It is especially important to understand the connection between the structure visible in the UV maps and the average properties because it will not be possible to observe this structure with reasonable exposure times for more distant galaxies.

We measured halo UV fluxes in long boxes parallel to the midplane of the galaxy, as in HKB14. The length of each box is equal to the length of the region containing the galaxy image, and the width of each box is a projected 2 kpc at the distance of the galaxy. We also reference the height of the box to the projected midplane of the galaxy. For galaxies with lower inclinations, such as NGC 4666, the projected heights are too large. The halo fluxes are all measured starting from a projected height of 2-5 kpc from the midplane in our sample. This conservative threshold excludes galaxy disk light, but likely misses halo light at lower latitudes. The galaxy fluxes measured in the boxes used for PSF-wing correction are reported in Table 4.

The halo fluxes are corrected for PSF-wing contamination, and the uvw2 fluxes are corrected for the red leak using a galaxy-type correction from HKB14 to isolate the true UV flux (shortward of 3000Å). We used a correction factor of 0.93, except for M82 (0.95) and NGC 3623 (0.85). The uvw1 filter has a more severe red leak for which this correction is unreliable, so we report the uncorrected fluxes. The total halo fluxes, amount of PSF-wing contamination, and uvw2 correction factors are listed in Table 5.

In the remainder of this section, we assess the reliability of the measured fluxes, measure scale heights for the UV halos, and examine the change in color as a function of height. We then search for correlations between the average UV halo fluxes and galactic properties, and compare our results to HKB14.

5.1. Reliability of Halo Fluxes

The similarity in the halo morphology between different UV filters suggests that our PSF-wing subtraction is accurate, but it is worth considering the reliability of the measured fluxes when the PSF-wing contamination can be a large fraction of the total flux. We performed several tests, and conclude that the fluxes are indeed reliable in the sense that the systematic error from PSF-wing subtraction is smaller than the statistical error.

Name FUV NUV uvw2 uvm2 uvw1 uvw2 Factor
(1) (2) (3) (4) (5) (6) (7)
Starburst Galaxies
NGC0253 0.93
M82 0.95
NGC4631 0.93
NGC3628 0.93
NGC4666 0.93
NGC3079 0.93
NGC5775 0.93
NGC4388 0.93
Normal Galaxies
NGC0055 0.93
NGC0891 0.93
NGC2683 0.93
NGC4517 0.93
NGC4565 0.93
NGC4096 0.93
NGC4313 0.93
NGC3623 0.85
NGC5907 0.93
NGC4216 0.93
NGC4607 0.93
NGC4522 0.93
NGC0134 0.93
NGC4157 0.93
ESO358 0.93
NGC4217 0.93
NGC4330 0.93
NGC3044 0.93
NGC5170 0.93

Note. – The galaxy fluxes are measured in boxes as described in Section 5. The uvw2 scale factor is based on the galaxy type as described in HKB14.

Table 4Galaxy UV Fluxes

First, the PSF-wing contamination is not very sensitive to the best-fit exponent because most of the PSF-wing contamination comes from the region where the profile is measured rather than extrapolated. If we extrapolate the PSFs with an exponent of in each filter, which is unacceptably shallow (Table 3) and inconsistent with the background, the fractional increase in the PSF-wing contamination is only about 2%, except in the uvw1 filter where the change is about 8%.

Second, the NUV filter covers approximately the same wavelength range as the combined uvm2 and uvw1 filters. Since the PSF-wing contamination is smaller in the NUV, comparing the NUV fluxes to the average of the uvm2 and uvw1 can tell us if the PSF-wing contamination is obviously too high or low. For galaxies with fluxes in each filter we measured the flux with and without accounting for PSF-wing contamination, and averaged the uvm2 and uvw1 values. Without the PSF-wing correction, the mean UVOT/GALEX flux ratio is about 1.6 across our sample, whereas after correction it is about 1.05. We cannot exactly synthesize the NUV filter from the UVOT filters, so a ratio of 1.0 is not expected and this comparison cannot tell us if there are small systematic uncertainties. Nonetheless, this exercise suggests that the reported fluxes are close to the true flux.

Finally, there are HST observations of the halo of M82 in several UV bands (using the Wide Field Camera 3 UVIS detector), including the F225W filter that overlaps with the uvw2, NUV, uvm2, and uvw1 filters. The sub-arcsecond angular resolution and small field of view mean that PSF-wing contamination is negligible. The wind is clearly visible in the HST image, and the flux measured in this region is consistent with that measured with Swift and GALEX after PSF-wing subtraction but not before. For example, in one region (a circle of radius 15 arcsec centered on 09:55:52.22, 69:40:02.39), we measure an AB magnitude of  mag in the HST image,  mag in the uvm2 image before PSF-wing correction, and  mag after. Similar values are found in the other filters, indicating that the PSF-wing subtraction recovers the true flux to within the statistical uncertainty.

Halo Flux (mJy) PSF-Wing Contamination (mJy) uvw2 Factor
Name FUV NUV uvw2 uvm2 uvw1 FUV NUV uvw2 uvm2 uvw1
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
Starburst Galaxies
NGC0253 4.89 0.60 0.93
M82 0.29 0.53 0.69 0.54 2.94 0.95
NGC4631 5.75 0.99 3.16 3.01 6.39 0.93
NGC3628 0.20 0.11 0.21 0.19 0.80 0.93
NGC4666 0.16 0.06 0.13 0.11 0.39 0.93
NGC3079 0.53 0.23 0.43 0.41 1.15 0.93
NGC5775 0.13 0.11 0.17 0.14 0.45 0.93
NGC4388 0.14 0.08 0.16 0.14 0.49 0.93
Normal Galaxies
NGC0055 13.99 0.51 0.93
NGC0891 0.15 0.05 0.15 0.10 0.47 0.93
NGC2683 0.29 0.11 0.28 0.22 0.94 0.93
NGC4517 0.33 0.12 0.93
NGC4565 0.17 0.93
NGC4096 0.34 0.18 0.31 0.27 0.93
NGC4313 0.01 0.03 0.93
NGC3623 0.06 0.01 0.07 0.06 0.37 0.85
NGC5907 0.39 0.10 0.29 0.25 0.77 0.93
NGC4216 0.07 0.08 0.93
NGC4607 0.01 0.03 0.93
NGC4522 0.11 0.09 0.93
NGC0134 0.26 0.21 0.75 0.93
NGC4157 0.13 0.08 0.93
ESO358 0.06 0.03 0.93
NGC4217 0.04 0.06 0.93
NGC4330 0.08 0.16 0.12 0.11 0.26 0.93
NGC3044 0.31 0.32 0.29 0.28 0.69 0.93
NGC5170 0.08 0.09 0.93

Note. – The halo fluxes are measured as described in Section 5. The uvw2 scale factor is based on the galaxy type as described in HKB14. The total flux measured on the CCD is the sum of the inferred astrophysical flux (“halo flux”) and the PSF-wing contamination.

Table 5Halo UV Fluxes

5.2. Scale Heights

UV halos appear to have a morphology like a thick disk, so we measured scale heights () based on fitting a function of the form , where is the flux, is a normalization, and is the projected height above the midplane. We used the fluxes measured in long boxes because the is only high enough in a few galaxies to measure scale heights at different radii. To see how the aggregate behaves, we fit scale heights at several galactocentric radii in the stacked quadrant images. Figure 5 shows the scale height as a function of galactocentric radius, along with the 1 error bars, for the normal and starburst galaxies. In the starburst galaxies the source mask near for NGC 4631 is taken into account. In the normal galaxy stack the scale height within is consistent with being constant, with a possible increase with larger radii that may be due to a more complex background that results from the stacking. The starburst stack shows that the scale height changes with galactocentric radius, which can be explained by filaments. The scale heights can also differ between filters, as is the case for M82 and NGC 253, but in most galaxies the scale heights measured in different filters agree within the uncertainty. Here we report the value of measured from the average radial profile and combining each filter (we allowed to vary between filters, but fit for a common ). Table 6 contains the best-fit and the associated reduced .

Scale heights range from  kpc. The mean value of for starburst galaxies is  kpc ( kpc when excluding NGC 4666, which has a lower inclination), and it is  kpc for normal galaxies. There is no dependence on H luminosity, the presence of eDIG, Hubble type, UV halo luminosity, or the in the image. Galaxies with stripped gas (NGC 4522, NGC 134, and NGC 4330) have small scale heights ( kpc), but NGC 891 and NGC 55 have similar values even without stripped gas. Likewise, there is no connection between and the fraction of UV light from filaments. The galactic or halo properties that seem to be connected to other morphological indicators of UV halos do not predict the scale height.

In several galaxies, varies between filters. For example, in NGC 253 the joint fit to all filters results in but with a reduced . The FUV profile is marginally fit for  kpc (reduced ), but the NUV profile cannot be fit with a single exponential profile ( kpc with reduced ). A double exponential profile is a good fit with an inner component ( kpc) and an outer component ( kpc), so it is possible that the NUV light has a compact component and a component similar to the FUV halo. Figure 4 shows that the NUV light follows the disk much better than the FUV light, suggesting a possible difference in physical origin. A similar difference is found in M82, where the joint fit produces  kpc () and the best-fit values of in each filter are: , 3.0, 2.9, 3.2, and 2.5 kpc for the FUV, uvw2, uvm2, NUV, and uvw1 filters, respectively (with corresponding , 4.4, 0.8, 2.9, and 12). As in NGC 253, the FUV halo is more extended, and the redder filters cannot be adequately described by single exponential models. On the other hand, in NGC 55 (joint  kpc and ) the FUV halo has a smaller scale height than the NUV ( kpc compared to 1.3 kpc). It is tempting to ascribe the difference in behavior to the presence of galactic winds, and indeed NGC 3079, NGC 4631, and NGC 4666 do have shallower FUV than NUV profiles, but the is generally too low in the normal galaxies to determine if this is a clear difference between them.

We also searched for different scale heights across the midplane. In many cases, one filter has a different profile across the midplane while the others do not, but these can usually be explained by a small background gradient across the region of the galaxy (e.g., due to Galactic cirrus, which differs in strength between filters). The exception is M82, where the FUV markedly differs across the midplane in a way that other filters do not. The most likely explanation is that the wind is brightest in the FUV, as seen in Figure 7. Most galaxies have approximately symmetric flux profiles, despite differences in the visible morphology across the midplane (e.g., Figure 4). The galaxies where there is a notable difference in multiple filters include NGC 4666, NGC 4522, NGC 4330, and NGC 134. The latter three galaxies have an obvious asymmetry due to stripped gas (Figure 3). NGC 4666 is the least inclined galaxy in the sample, and the asymmetry in scale heights may pertain to inclination and projection effects, as mentioned in HKB14.

5.3. FUV-NUV Color with Height

We measured the correlation between the FUVNUV color in each flux measurement bin and the projected height of the bin for each galaxy with data in both filters. Most galaxies have no significant correlation (for a threshold of ), and the exceptions are all starbursts. In NGC 253, M82, NGC 4631, NGC 3628, NGC 3079, and NGC 4388, the FUVNUV color decreases with height. This is consistent with the FUVNUV color maps (Figure 7) and suggests that the UV halos comprise two components.

To improve the signal, we measured the FUVNUV color as a function of height in composites of the normal, starburst, and stripped galaxies (of which NGC 4522 and NGC 4330 have GALEX data). The composite measurements are shown in Figure 10. The data were binned in boxes 2 kpc wide, with the first box having a central height of 3 kpc from the midplane. The projected distances from the midplane for each measurement were used to assign each measurement to the appropriate bin, and the galaxies were projected to a common distance of 20 Mpc for the purpose of computing the magnitudes. Figure 10 reinforces the conclusions from the individual galaxies: the starburst halos become bluer with height, the normal galaxies have an approximately constant color, and the stripped galaxies are strikingly blue near the disk and similar to the normal galaxies at larger heights. However, some variation in the color with height is seen in a few individual normal galaxies. More data are necessary to determine if this is similar to what is seen in the starbursts.

5.4. Correlation Analysis for Average Properties

Name FUVNUV Scale Height Extraplanar Diffuse?
(mag) (kpc) (H) (X-ray) (160m) (GHz) (mag) (mag) (FUV) (uvm2) (FUV) (uvm2) (FUV) (uvm2)
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) (14) (15) (16) (17)
NGC0253 1.260.02 2.20.1 38 Y Y Y Y 67.6 3.6 6.9 1448.2 0.005
M82 0.870.02 2.70.1 32 Y Y Y Y 86.8 7.5 6.2 13.0 24.6 2621.6 3151.3 0.005 0.008
NGC4631 0.730.03 1.80.1 1.3 Y Y Y Y 6.4 1.5 1.2 11.9 21.8 612.7 648.2 0.019 0.035
NGC3628 1.550.08 3.90.2 1.8 Y Y N N 5.7 4.1 3.0 6.5 17.2 1347.4 1128.5 0.005 0.016
NGC4666 0.900.08 7.60.4 0.3 Y Y Y Y 3.9 3.7 3.0 8.2 12.9 2333.3 1932.9 0.003 0.007
NGC3079 0.350.01 3.40.1 1.6 Y Y Y Y 4.9 3.3 2.6 18.7 28.0 3772.2 3214.7 0.005 0.009
NGC5775 0.590.04 4.50.3 2.6 Y Y Y 2.3 4.0 3.2 11.4 19.8 1887.0 1563.7 0.006 0.013
NGC4388 1.020.06 2.70.1 3.5 Y Y Y Y 1.1 3.0 2.1 6.0 14.6 945.9 878.7 0.006 0.017
Normal Spirals
NGC0055 2.80.6 1.10.1 4.2 Y N Y 4.0 0.5 0.5 65.7 0.007
NGC0891 0.330.09 1.50.2 0.8 Y Y Y Y 6.8 5.0 4.3 6.9 11.4 1088.3 1043.1 0.006 0.010
NGC2683 2.00.4 2.70.2 0.9 N N N 1.1 2.3 1.7 3.6 5.8 203.6 219.7 0.018 0.025
NGC4517 1.50.2 3.40.2 1.9 Y 0.7 1.9 3.1 152.2 0.020
NGC4565 2.00.3 0.7 N N N N 1.2
NGC4096 1.80.3 4.30.3 0.7 N Y N 0.9 2.0 1.5 1.9 7.9 277.6 312.3 0.007 0.024
NGC4313 1.00.3 6.63.3 2.7 0.2 3.9 1.3 73.7 0.021
NGC3623 3.3. 4.1. 3 N N N 0.6 2.3 1.4 0.4 3.3 169.9 209.4 0.002 0.015
NGC5907 0.740.09 6.40.2 1.3 N N Y N 1.7 2.5 2.0 7.1 13.8 816.6 829.1 0.009 0.016
NGC4216 1.60.3 3.40.3 0.9 N N N 0.3 2.6 2.7 160.1 0.017
NGC4607 1.20.4 3.71.3 1.0 N 0.4 4.7 0.8 202.1 0.004
NGC4522 0.550.04 1.90.1 1.8 Y Y 0.2 1.6 6.0 139.4 0.043
NGC0134 1.90.3 0.7 2.8 2.5 10.1 1548.1
NGC4157 0.80.1 4.70.5 2.1 N Y 2.0 3.6 6.3 1171.6 0.005
ESO358 1.20.1 5.30.9 0.1 0.5 2.8 2.6 296.9 0.009
NGC4217 1.20.2 3.10.4 0.1 Y Y Y Y 7.6
NGC4330 0.510.03 1.10.1 1.6 0.1 1.7 1.4 5.1 6.9 101.7 116.5 0.050 0.055
NGC3044 0.20.1 3.80.7 0.8 Y Y Y 1.1 2.4 1.9 6.7 6.0 998.1 1011.2 0.007 0.006
NGC5170 3.2. 7.90.5 0.8 N N 0.2 1.5 3.6 269.1 0.013

Note. – Cols. (1) Name (2) FUVNUV color averaged over the halo (3-4) Best-fit scale height to combined data from all wavebands and reduced (5-8) Is there extraplanar diffuse emission? (9) FIR flux used to determine galactic extinction in units of  erg s cm (10-11) Estimated extinction in the FUV and uvm2 bands from Buat et al. (1999) (12-13) Halo luminosity density in units of  erg s Hz (13-14) De-reddened galaxy luminosity density in units of  erg s Hz (15-16) Halo-to-galaxy luminosity ratio using de-reddened galaxy luminosity. References. – The presence of extraplanar diffuse emission was determined from the literature. H: Rossa & Dettmar (2003a), X-ray: Strickland et al. (2004); Tüllmann et al. (2006); Li & Wang (2013), Radio: Wiegert et al. (2015)

Table 6Diffuse Halos

We compared the total halo fluxes, FUVNUV colors, and the scale heights to the following galaxy parameters: the UV luminosity of the galaxy (, corrected for internal extinction), the star-formation rate (SFR) and specific SFR (sSFR), the H luminosity of the galaxy (), the stellar mass (), the rotation velocity (), the morphological type code (; de Vaucouleurs et al., 1991), the inclination (), and the distance from Earth ().

Figure 10.— FUVNUV color as a function of height for composites of the normal, starburst, and stripped galaxies in our sample.

, , , , and come from NED and HyperLeda values. Some of the H fluxes include some flux from the neighboring [Nii] line. We compute from the -band luminosity from the 2MASS Extended Source Catalog and the mass-to-light ratio from Bell & de Jong (2001), for which we use colors from HyperLeda. The SFR is based on the relation from Kennicutt (1998), which uses far-infrared (FIR) fluxes measured with the IRAS satellite. Fluxes exist for all but one galaxy (NGC 4217). The values we use are reported in Table 1.

We estimated in each filter by “de-reddening” the measured flux following Buat et al. (1999), who found a relation for the UV extinction based on the far-infrared (FIR) and measured UV fluxes. The relation is best calibrated for the FUV and uvm2 filters, so we focus on these filters. We further correct for the light that would escape the galaxy assuming an extinction out of the disk of 0.5 mag in the band (Calzetti, 2001). We convert this to the extinction in the FUV and uvm2 filters using a modified Calzetti et al. (2000) law from Kriek & Conroy (2013), who derived extinction curves for composite galaxy SEDs. Following their scheme, we adopt an attenuation slope of for the starbursts and for the normal galaxies, and a 2175Å bump strength of or for starbursts and normal galaxies, respectively. These are both more similar to the Calzetti et al. (2000) law than the traditional Milky Way extinction law towards (Weingartner & Draine, 2001). This leads to and for the starburst galaxies, whereas and for the normal galaxies. These parameters are appropriate for star-forming galaxies, but the analysis below is insensitive to the exact or , or whether includes attenuation along the minor axis. The measured , the estimated extinction, and the corrected values are given in Table 6.

We used Spearman’s rank correlation with a significance threshold of to search for correlations between galaxy and halo properties. A summary of the findings is given in Table 7. Here we describe the results and give the correlation coefficient () and value for correlations with . As we tested for correlations between a range of measurements, we adopted the Benjamini & Hochberg (1995) procedure to control for false positives, also using a threshold of . There were 61 total comparisons. Correlations with -values that remain significant for the revised threshold are noted in Table 7 and discussed below.

The halo luminosity is strongly correlated with , with and in FUV and and in uvm2. It is also correlated with the SFR in both ( and for the FUV and with in uvm2), and with the specific SFR ( and in FUV and and in uvm2). is closely related to the SFR, but whereas there is a significant correlation between and in the FUV (, ) there is no significant correlation for uvm2 (). is correlated with morphological type for uvm2 (, ) but not for the FUV (). Overall, the -values are smaller for the FUV, possibly because there are 24 FUV measurements and 15 uvm2 measurements. There is no other significant correlation between and the other parameters in either filter.

is also correlated with the FUVNUV color for the FUV band (, ), but not uvm2 (). The FUV-NUV color is significantly correlated with the specific SFR (, ) but not the SFR (), nor with in either FUV or uvm2 ( and , respectively). There are no other significant correlations between FUVNUV and other galaxy properties.

In addition to the FUV , the FUVNUV color is correlated with the FUV (, ). It is not correlated with other halo luminosities. FUVNUV is also marginally correlated with sSFR (, ), but the correlation with SFR is not significant. There are no other significant correlations between FUVNUV and the galaxy properties.

We also compared to the galaxy properties for the FUV and uvm2 filters. is correlated with the SFR ( and for the FUV, and with for uvm2). There are no other significant correlations with other properties.

The scale height is not correlated with any galaxy or halo property except , where and . This is almost entirely due to NGC 5170 ( kpc), which is near the distance cutoff and has an unusually large value.

Applying the Benjamani-Hochberg procedure for and 61 tests, we find that the following correlations are significant: in both FUV and uvm2 is correlated with and with the SFR, and is correlated with FUVNUV, specific SFR, and in the FUV only. The other correlations with described above are not significant at this threshold.

Quantity FUVNUV
(FUV) (uvm2) (FUV) (uvm2) (kpc) (mag)
(1) (2) (3) (4) (5) (6) (7)
0.87aaThese correlations survive the Benjamini-Hochberg test with for 61 tests. - N - N 0.41aaThese correlations survive the Benjamini-Hochberg test with for 61 tests.
- 0.79aaThese correlations survive the Benjamini-Hochberg test with for 61 tests. - N N N
N N N N - N
0.62aaThese correlations survive the Benjamini-Hochberg test with for 61 tests. N N N N N
SFR(IR) 0.82aaThese correlations survive the Benjamini-Hochberg test with for 61 tests. 0.77aaThese correlations survive the Benjamini-Hochberg test with for 61 tests. 0.42 0.59 N N
SFR(IR)/ 0.64aaThese correlations survive the Benjamini-Hochberg test with for 61 tests. 0.59 N N N 0.48
N 0.58 N N N N
N N N N 0.47 N

Note. – Correlations between halo quantities and galaxy and halo properties. For significant correlations () the Spearman ranked correlation coefficient is given. Non-significant correlations are marked by an ‘N’. A dash (‘-’) indicates no measurement. Cols. (1) Quantity (2-3) Halo luminosity (4-5) Ratio of halo luminosity to de-reddened galaxy luminosity (6) Scale height of the UV halo (7) Average FUVNUV color in the halo. See text for more a more detailed description.

Table 7UV Halo Correlations

In summary, is related to and some the metrics of star formation, as well as the halo color. However, the scale height has no (strong) dependence on the halo or intrinsic galaxy properties.

5.5. Average Properties and Morphology

For most galaxies in the GALEX and Swift archive it is not possible to examine the morphology in detail because of low or large distance. Thus, we compare the average properties , FUVNUV, and to the UV maps in Figures 3 and 4. Overall, we find that averaging obscures the presence of halo components such as winds, filaments, or bright emission near the disk, so that these properties are not useful metrics of the halo morphology.

Apart from the clear divide between normal and starburst galaxies, appears unrelated to the halo morphology, except in stripped galaxies. For example, NGC 5907 has a very patchy, low halo in Figure 3, but its is among the highest of the normal galaxies. Since we know that is uncorrelated with it does not provide information about the halo structure. Likewise, the most spectacular UV halos around starburst galaxies (M82, NGC 253, NGC 4631, and NGC 3079) have similar . The stripped galaxies (NGC 134, NGC 4522, and NGC 4330) have unusually high (5-10% instead of the typical 0.5-2%). As discussed earlier, this is largely attributable to extraplanar star formation.

We also find no connection between the halo morphology and the average FUVNUV color, except insofar as bluer halos are more structured in our sample because they predominantly belong to starburst galaxies. Likewise, the scale height appears to be independent of radial concentration, although in starburst or stripped galaxies the scale height can change with galactocentric radius. The stripped galaxies cannot be distinguished from the rest of the sample by their average FUVNUV color and are only barely distinguishable by scale height; it is not likely that these metrics could identify stripped galaxies in a wider sample.

5.6. Differences with HKB14

In HKB14 we measured fluxes without the PSF-wing correction and used these fluxes to find scale heights for several galaxies and measure correlations between some of the same quantities as above. The PSF-wing contamination is strongest near the disk, so we might expect to increase when the spurious flux is removed, but this is not generally the case. However, the correlations that we find between and , the H luminosity, and the FIR SFR are stronger after is corrected for the PSF-wing contamination. For example, in the FUV the correlation between and is whereas between and it is . The lower fluxes also change the SEDs, but the only trend reported in HKB14 that is truly artificial is the uniform rise in the FUVNUV color with height; we find in most galaxies no change with height.

Shinn & Seon (2015) investigated the role of PSF-wing contamination in several galaxies included in HKB14, and found that in two cases (NGC 24 and IC 5249) the UV halo that we reported appears to be entirely artificial. They used the GALEX images exclusively, and extrapolated the PSF wings from the PSF provided by the GALEX calibration team. Using the PSFs described in Section 3, we agree with their findings for the GALEX images. For the considerably deeper Swift images, we find a faint UV halo around NGC 24 (which is not edge-on) and no UV halo above 2 kpc around IC 5249. IC 5249 is seen edge-on has a very thin optical disk, so it is possible that any astrophysical UV halo exists primarily below this. Based on these results and the detections in our present sample, UV halos are ubiquitous but may not be universal (or universally detectable with current instruments).

6. Discussion

6.1. The Diffuse UV Light is Probably a Reflection Nebula

Prior studies have argued that the UV halos are eRN on the basis that the light is too bright to come from shock-heated or photoionized gas (Hoopes et al., 2005), that line emission from starburst halos is polarized in a way that is consistent with dust scattering (Yoshida et al., 2011), that it can be successfully modeled by Monte Carlo radiative transfer scattering models (Seon et al., 2014; Shinn & Seon, 2015), that it is too blue to originate in the stellar halo (HKB14), and that they are coincident with eDIG, winds, and other outflow tracers.

Our results further support this scenario:

  1. The UV halos are truly diffuse, as determined by comparison with existing HST data, and they trace filamentary morphology seen at other wavelengths.

  2. UV halos are a broadband phenomenon, with similar morphology seen from 1500-2600Å.

  3. is strongly correlated with after de-reddening the galaxy flux, and it is strongly correlated with the SFR.

  4. Extraplanar UV light is most visible above regions of active star formation as seen in H, X-ray, and radio continuum data, but the presence or prominence of the UV halo does not depend on the presence of extraplanar H, X-rays, or radio continuum. This rules out an emission nebula as the source.

  5. Starburst galaxies have more luminous UV halos, but not higher values or scale heights.

  6. Starburst winds are visible through their smaller FUVNUV colors, which appear to be superimposed on a thick disk similar to (although brighter than) that around normal galaxies.

  7. The UV halo flux, color, and scale height are not significantly correlated with galaxy properties that are not closely connected to star formation.

The strong connection between and and the weak connection between and the presence of outflows strongly suggests that UV halos are eRN. This also explains their broadband visibility and ubiquity, as non-ionizing UV photons can escape even if the disk is not porous. If UV halos are eRN, then they indicate that dust is widespread in the halos of spiral galaxies of all types. This implies that dust is long-lived in galaxy halos to the extent that it exists around galaxies without strong outflows and has a similar scale height as the dust around starburst galaxies. In the remainder of this section, we assume that the UV halos are eRN and examine some basic properties (however, we defer a detailed analysis of the UV halo SEDs to Paper III). We then discuss the results in the context of other work on extraplanar dust.

6.2. Dust Mass

We estimate the dust mass using Monte Carlo radiative transfer (MCRT) models based on the model described in Wood et al. (2001) and Whitney (2011), which incorporates the Henyey-Greenstein functions for the angular dependence of dust scattering and guarantees that each photon scatters once. These models are meant to be first-order estimates and are not as carefully constructed as other MCRT models of eRN such as in Shinn & Seon (2015) or Baes & Viaene (2016), but they are nonetheless useful.

Thick Disk Milky Way Dust SMC Dust
(kpc) (kpc) () () ()
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
NGC 891 12 1.9 1.0 0.9 3.2
NGC 4631 8 2.3 4.6 4.1 0.7
NGC 5775 12 3.0 9.9 9.3 -
NGC 5907 17 3.0 1.1 1.4 6.1

Note. – Cols. (1) Name (2-3) Scale lengths for the thick exponential disk from matching H i profiles (4-6) Best-fit , reduced , and implied extraplanar dust mass above 2 kpc for models with Milky Way dust. The error bars are statistical from the MCRT fitting only and underestimate the true uncertainty. (7-9) Best-fit , reduced , and extraplanar dust mass for SMC dust. The fits were performed using a dust opacity and scattering albedo at 2000Å and uvw2 data. See text for details. (10) Lower bound to dust mass above 2 kpc from 160 m fluxes. See text for details.

Table 8MCRT Models & 160 m Masses

In principle, the dust mass can be measured directly from the ratio, since


where is the scattering albedo and is the optical depth. is the extinction cross-section and the column. The extinction through the halo is low (), so a single-scattering approximation is reasonable. As described in HKB14, and can be determined from the shape of the SED, so the normalization is related to a characteristic column density from the disk through the halo. However, the measured flux at a given height above the disk cannot be straightforwardly interpreted as a dust column because the light source is not behind the dust. Thus, one must also adopt a geometric model for the halo dust and the disk emission, for which we use MCRT.

The MCRT model needs input distributions for the emission and the scattering/absorbing medium, as well as the viewing angle and the scattering cross-section ().

We assume that the dust is embedded in the neutral medium and model the gas density using several components: a thin exponential disk, simple logarithmic spiral arms, and a thick exponential disk that represents the halo. Each exponential disk takes the form , where is the galactocentric radius in the plane and is vertical height above the plane. We constrain , and by using high resolution H i maps, so we only model galaxies where we have such maps in hand: NGC 891, NGC 4631, NGC 5775, and NGC 5907. in the halo is unconstrained, so we tie it to the value of in the disk. The spiral arms are described by a polar equation , where is the normalization and the rate of growth (the parametric equations for a Cartesian grid are and ). We choose and and compute and for with a minimum radius  kpc. To add thickness, these spiral arms are convolved with a 3D Gaussian kernel with  pixels and an amplitude of , which is added to the underlying exponential disk. We fixed the parameters based on the extent of typical spiral arms, but for an edge-on viewing angle our results are insensitive to a wide range of , , or spiral arm width. The lower panels of Figure 11 show the model for NGC 4631; we do not attempt to reproduce large-scale filaments or warps.

The light source consists of a very thin exponential disk with the same scale length as the thin gas disk and light from the midplane in the spiral arms, since we assume that UV light comes from young stars near the midplane and in the arms. Our results are not sensitive to the shape or the location of the spiral arms because we measure the halo fluxes by summing over (described below).

Paper III will be devoted to modeling the SED and determining the dust size and composition, which will yield and . Here we adopt a simpler approach by using the values from two dust models from Weingartner & Draine (2001): dust in the bar of the Small Magellanic Cloud (SMC), and dust in the Milky Way for sightlines where the extinction law . The choice of these models is motivated by the observation that the halo dust at larger radii appears to have an SMC-like extinction curve (Ménard et al., 2010), and that many galaxy disks have extinction curves similar to the Milky Way. As we shall see, the inferred dust masses for each model are similar.

The density and emission model is gridded on a cube with a pixel scale of 2.5 pixels kpc, and we model the galaxy at 2000Å only (where the corrected is most reliable). The viewing angle is matched to the inclination of each galaxy, and we apply the dust model by assuming that the extinction within the disk follows the Kriek & Conroy (2013) prescription with and , with a gas-to-dust ratio () of 100, which is typical for spiral disks. Then, for a given halo we simulate the light scattered into the line of sight above 2 kpc from the midplane and compare to the observations. We repeated this procedure for a grid of with spacing of 10 between to 100, and a spacing of 50 thereafter. The 1 acceptable range shown in Table 8 is based on on this grid. The best-fit halo then implies a dust mass. For each simulation, we use scattered photons (we found for several test cases that results do not differ when using photons).

For each we simulated an image and measured the halo light using long boxes parallel to the midplane that are analogous to those used to measure real UV halos. This binning smooths over differences between spiral arm models that are not well constrained by the data. We normalized the real and simulated fluxes to the projected galaxy luminosity, and found the best-fit using the statistic. The best fit is often not a good fit, but the simulated halos reasonably reproduce the observed scale heights and .

As an example, we show the fit to NGC 4631 in Figure 11. The top panels show the outcome of the MCRT simulation. In the upper left panel, noise comparable to that in the uvw2 image was added to show how far the simulated halo could be detected in the UVOT images (3 and 6 contours are shown). The morphology of the observed halos is more complex than the symmetric, disky morphology that we obtain by construction, but since and the scale height are not morphological indicators (Section 5), the MCRT models are reasonable proxies for obtaining (Seon et al., 2014). The top right panel shows the best-fit model compared to the measurements. The best-fit values are for Milky Way dust and for SMC dust, which implies a dust mass above 2 kpc of for Milky Way dust and for SMC dust. We emphasize that we do not know whether SMC or Milky Way dust is closer to the true halo composition, but the similarity in masses suggests that is insensitive to the dust model for a plausible range of models. The bottom panels of Figure 11 show the H i model compared to the data.

In Table 8 we give the best-fit , , halo component parameters, and dust mass above  kpc for the four galaxies we considered. We expect the best-fit to be similar between filters, but with the current data this is not a useful test because the true extinction curve is not known, and changes in are degenerate with changes in the incident spectrum. We caution that there could be systematic shifts in due to our choice of geometric model, the assumption that the dust is hosted by the neutral medium, the adopted in the disk, and the scattering recipe in the MCRT code. Even if all of these are correct or unimportant, we expect the MCRT values to be lower limits to the halo dust in this region because shallower profiles are allowed by the H i data and could not be constrained by our UV measurements.

Bearing these caveats in mind, it is nonetheless worth noting that the MCRT masses are consistent with extrapolating the spherically symmetric Ménard et al. (2010) radial profile (measured between 20-1000 kpc) to 2-10 kpc, when we scale their average profile based on the ratio of the mass of each of the four galaxies considered here to their average galaxy. This suggests that there is a smooth density transition from the disk to the halo, at least in projection.

6.3. 160 m Dust Mass

The thermal emission from extraplanar dust provides a more straightforward measurement of because the flux is directly related to the column density. Here we use the 160 m images (as opposed to the other FIR bands) because 160 m is close to the peak of the FIR SED for cool dust and because there is more coverage of our sample with Herschel/PACS at this wavelength.

In most cases, the morphology of the 160 m emission does follow that of the UV in the sense that it tends to be brightest where the UV halo is also the brightest. However, the FIR emission is often not as extended vertically or radially as the UV halo (the exceptions being NGC 891 and NGC 4096). This could be a function of the much lower UV background or because most of the FIR maps are shallow snapshots obtained to measure emission from the disk, not the halo, as part of Herschel surveys (such as the Herschel Reference Survey; Boselli et al., 2010).

The dust mass can be obtained from the flux, provided one knows the temperature and grain emissivity law . The number of grains is obtained from the flux and these quantities:


where is the dust emissivity and the grain size. The mass implied is


where is the intrinsic density of the material. We assume that m and that  g cm. There is actually a distribution of grain sizes that can be constrained with a dust model, which we defer to Paper III. However, we do not know the temperature or . is measured by fitting a modified blackbody to the far-infrared and sub-mm SED, but almost all of the galaxies in the sample lack the data to measure in the halo.

For an order-of-magnitude estimate for comparison with the MCRT models, we assume that (in the literature it varies from 1-2 in nearby galaxies; e.g., Bendo et al., 2003; Casey, 2012), and use the temperature of the cold dust measured in the disk for each galaxy. If the dust is embedded in neutral gas, its temperature will likely decline with height, so these values will be lower limits.

We measured 160 m fluxes above a projected height of 2 kpc for NGC 891, NGC 4631, and NGC 5907 from Herschel maps (NGC 5775 does not have Herschel data, and the angular resolution of Spitzer/MIPS is too poor to isolate halo flux). These fluxes are  Jy for NGC 891,  Jy for NGC 4631, and  Jy for NGC 5907. For temperatures we use 23 K for NGC 891 (Hughes et al., 2014), 22 K for NGC 4631 (Meléndez et al., 2015), and 18 K for NGC 5907 (Dumke et al., 1997). Under these assumptions, the dust masses are , , and for NGC 891, NGC 4631, and NGC 5907, respectively (Table 8).

Figure 11.— The best-fit MCRT model to NGC 4631 for Milky Way-like dust. Top Left: MCRT image at 2000Å with 3 and 6 contours overlaid, along with flux measurement boxes. The average noise level from the uvw2 image was added to create the contours and show the “observable” extent of the halo. Top Right: Measured uvw2 fluxes from NGC 4631 compared to values measured from the MCRT model in the boxes at left. Fluxes in both cases are normalized to the measured galaxy flux. Bottom Left: Model H i contours overlaid on the MCRT image, which is rotated to the position angle of NGC 4631. Bottom Right: Measured H i column densities overlaid on our H i model. The thick black line shows the profile along the minor axis, while the thin line shows the major axis.

These values are very sensitive to ; if , the masses are about ten times larger. This highlights the need for deeper FIR data to constrain in the extraplanar dust. Considering the large uncertainties, the masses are consistent with the MCRT values but do not provide useful constraints.

6.4. Dust Spatial Variation

Some UV halos vary in FUVNUV color with galactocentric radius (Figures 6 and 7) or height (Figure 10). If the UV halos are eRN, these differences arise either through changes in the dust or the incident spectrum, although both effects may be present. M82 provides a good example of the general behavior seen in starbursts, namely that the eRN is bluer in the wind region (Figure 7) and becomes bluer with height (Figure 10).

If the dust changes in composition with height (for example, if the fraction of silicate grains increases), the reflected spectrum could become bluer given a constant incident spectrum. Since the optical depth through the halo is small (as is clear from observing face-on galaxies), this is a good approximation. On the other hand, given that the nuclear starburst is obscured through the disk but (presumably) not along the minor axis, the incident spectrum into the halo may differ across the disk, leading to a projected change in the color. In this case, sightlines vertically above the starburst region would reflect a bluer incident spectrum than the rest of the disk. This behavior can be generalized to H ii regions in the disk.

We cannot rule out either possibility, but in M82 the data suggest that the color change in height results from a change in the dust. Let us assume that the halo dust does not change. Then, for a partially obscured nuclear starburst viewed edge-on, in which light can escape along the minor axis, we expect the eRN to form bright lobes above the starburst and near the disk. These will be visible against the rest of the eRN. The lobes are bright near the disk because both the incident flux and column density are higher there. Lobes of bluer color than the rest of the eRN will also occur, with a small height offset between the bluest and brightest region. This offset occurs because the contribution to the eRN from the rest of the disk (with a redder incident spectrum) is also greatest near the disk. Above the blue lobes, the halo becomes redder as the incident flux from the starburst declines and the reflected flux from this region competes with the flux from the remainder of the disk.

We confirmed these features through MCRT models of partially obscured starbursts, as shown in Figure 12. We tried two models: a model with the same setup as described above except with 90% of the emissivity clustered in the inner 1 kpc, and a variation of this model with a “wind cone” of denser material. To examine the color we assumed that the starburst region has a Starburst 2 spectrum and the remainder of the disk has an Sc spectrum from the Kinney-Calzetti atlas (Kinney et al., 1996). We do not simulate a full spectrum, but rather determined the proper relative luminosities for the FUV and NUV bands. The cone has an opening angle of 50 degrees, a width of 0.8 kpc, and a density that declines exponentially with height but is uniform at each height within the cone region. The vertical profile is , where is 1/10 of the peak, central density that is constrained by the Hi data and the scale height  kpc. These parameters are motivated by the appearance of limb-brightened winds in H, but are not fits to the data. We tried one additional variation in which the interior of the cone has zero density (i.e., it lacks dust).

In the case without the wind cone the main feature is the appearance of the lobes described above, whereas the remainder of the eRN looks like that for a normal disk galaxy. When the cone is present, it is brighter than the rest of the halo and its center is bluer than average, while the projected sides are relatively red. However, the presence of a cone does not change the presence of lobes in the FUVNUV color map even if the interior of the cone is dust-free because the light is scattered by the dense cone walls.

Figure 12.— Top Left: Combined UV intensity map for M82. Top Right: FUVNUV color map for M82. Blue or purple colors indicate where the halo is bluer, while redder colors indicate larger FUVNUV (in magnitudes). Bottom Left: an MCRT simulation of an obscured starburst with a wind cone and a redder disk. The brightest region of contours from the blue light are shown. A horizontal box indicates the galaxy disk. Bottom Right: A synthetic FUVNUV color map showing the halo colors with the contours from left superimposed. We expect blue “lobes”, which are not seen in M82.

M82 has bright lobes near the disk and there is indeed an offset between these lobes and the bluer region above, but there are no corresponding lobes in the FUVNUV map and the color instead continues to become bluer with height and covers an increasingly large fraction of the projected area (Figure 12). These features cannot be explained by the simple geometric model above. The optical depth may be higher than we assume, which would suppress lobes in the color map, but the halo at larger heights would also be redder as the incident spectrum would change substantially through the halo. Some of the FUV light may come from line emission, but the same morphology is seen in the very deep uvw2 image. The simplest explanation of the color map is a change in the dust with height. This could either be a physical change in a single dust component with height or an increasing contribution from one component of dust.

In other cases, such as NGC 253 or NGC 3079, there are blue regions in the halo near the bases of the winds that may indicate a different light source rather than a difference in dust (Figure 7). However, in both of these galaxies and in the composite starburst measurement, FUVNUV color increases within a few kpc of the disk and then decreases with height (Figure 10). This suggests that these filamentary structures exist alongside an overall change in the dust with height. In contrast, most of the normal galaxies do not have significant FUVNUV structure in Figure 6 and the composite FUVNUV color does not appear to change with increasing height. This suggests that the dust has similar properties between 2-12 kpc. However, NGC 891 and NGC 5907 have high data and show a modest decline with height above 5 kpc. Deeper data for more galaxies are necessary to determine whether the color is constant with height. Finally, in Figure 10 the stripped galaxies appear bluer than starburst galaxies at low heights and redder at larger heights. Since some of the FUV emission comes from extraplanar star formation in these few cases, we speculate that the difference is not explained by a change in the dust with height.

If the FUVNUV color change with height around starbursts results from a change in the dust, two plausible explanations are that galactic winds eject a different type of dust and carry it to larger heights than the processes in normal galaxies, or that winds and a comparatively intense radiation field alters the pre-existing dust at large heights. In normal galaxies, the dust may circulate through a galactic fountain (Bregman, 1980), which is consistent with the measured UV scale heights of several kpc. On larger scales, we expect the dust to change from Milky Way-like dust in the disk to SMC-like dust by a radius of 20 kpc (Ménard et al., 2010), which (for a given radiation field) will lead to smaller FUVNUV color (however, cf. Smith et al., 2016, who find that most “extragalactic” dust can be explained by an extended disk). In the composite (Figure 10) we do not detect the transition through FUVNUV color alone, but most of the galaxies do not have reliable FUVNUV colors above 10 kpc; the values there for normal galaxies come from 2-4 galaxies, depending on the height. A clean test for this transition requires deeper data for more individual galaxies. If the dust does not change around normal galaxies up to heights difficult for a galactic fountain to reach, one possibility is that starburst episodes provide most of the halo dust.

6.5. M82 and NGC 891

NGC 891 and M82 both have UV halos that have been previously studied, and here we compare our maps and results to those in the literature. NGC 891 is a normal Milky Way analog with an elevated star formation rate near 4  yr, (Popescu et al., 2004) and a bright X-ray halo (Bregman & Pildis, 1994), whereas M82 is the archetypal superwind galaxy.

The UV halo around NGC 891 was modeled by Seon et al. (2014), who found that the UV light is consistent with a scale height between 1.2-2 kpc and a dust mass above a projected height of 2 kpc of 3-5% of the total. Hughes et al. (2014) find a dust mass in the disk of , so the halo dust mass from the MCRT models in Seon et al. (2014) is between . This is consistent with the MCRT scale height and mass we measured:  kpc and (for Milky Way dust), whereas Seon et al. (2014) fit a profile along the projected minor axis, incorporating the core of the PSF but not the highly extended wings.

In HKB14, where we did not account for the extended PSF contamination of the halo, the UV intensity falls off more slowly with height on the side with the large H i filament, which also occurs in the X-rays and the 160 m image. The combined, PSF-corrected UV map shows that the part of the UV halo clearly associated with NGC 891 is brighter on that side (Figure 3), but the large-scale very diffuse emission is too extensive to be scattered light from dust around NGC 891 and may be Galactic cirrus. The increased UV intensity near the galaxy may arise from dust within the filament material adding to the total dust column.

Howk & Savage (1997) examined extraplanar dust around NGC 891 in extinction at a lower height than the UV halo, finding filamentary structures that form a thick disk and have a mass of at least a few, or about 10-50% of the total (Hughes et al., 2014). The resolution of their images is much higher than the UV images, so we cannot rule out the possibility that the eRN is also highly structured. However, the mass in the scattering component (estimated through MCRT or 160 m flux) is less than half than that inferred by Howk & Savage (1997). This is also true if we measure the diffuse dust mass between 0.5-2 kpc in our MCRT models. While they argue that it would be challenging for a galactic fountain (Bregman, 1980) or radiation pressure (Ferrara et al., 1991) to lift enough dust mass to account for the extinction features they see, either mechanism could produce the smaller amount of diffuse dust inferred at large heights. This suggests that much of the dust in the chimneys falls back to the disk, and that the diffuse medium above 2 kpc is a separate component.

M82 has one of the first UV halos reported (Hoopes et al., 2005), since its extraplanar diffuse UV light is obvious from the image even without special processing. It is also the best-studied UV halo (Coker et al., 2013; Hutton et al., 2014), and the dust content of the wind has also been investigated by several authors (e.g., Engelbracht et al., 2006; Roussel et al., 2010; Yoshida et al., 2011). The consensus that the UV halo in M82 is an eRN supports the identification of other UV halos as eRN. The FIR dust emission is seen to about the same or larger distances than in the UV, implying a mass of about (Roussel et al., 2010). It is more difficult to construct an accurate MCRT model for M82 than other disk galaxies because of its irregular H i morphology, but using the same method as described above we obtain a value of , which is consistent with the FIR measurement. M82 also allows us to test our de-reddening procedure: our reported FUV  erg s, whereas Coker et al. (2013) estimate that the UV luminosity seen by the halo is  erg s.

We suggest that the best simple explanation for the the FUVNUV color change with height (in M82 and starburst halos generally) is a change in the dust. We note that Roussel et al. (2010) show that there is little change in the 250 m/350 m ratio in the region with the UV halo, but if the change is primarily one of composition rather than size the FIR bands may not be sensitive to it. For example, Nozawa & Fukugita (2013) successfully fit a two-component (graphite and silicate) dust model to extinction along different Milky Way and SMC sight lines by changing the mixture but fixing the size distribution using the same law found by Hutton et al. (2014). Likewise, Hutton et al. (2014) found, using shallow Swift data, that the dust is consistent with a single size distribution. They focused on color-color plots where a change in composition is degenerate with a change in the incident spectrum, so neither of these studies presents strong evidence against a change in the dust type with height. However, we emphasize that we have not conclusively shown that the dust composition must change.

7. Summary and Conclusions

We have described the morphology of UV halos around nearby, edge-on galaxies with and without superwinds, and here summarize our findings:

  1. UV halos are astrophysical and can be separated from galactic light scattered into the PSF wings (Airy patterns). They are broadband phenomena and truly diffuse, with a flux of 1-20% of the apparent galaxy luminosity (a few percent or less of the de-reddened values). They are visible to beyond 10 kpc above the midplane.

  2. Around normal galaxies, UV halos tend to have a thick-disk morphology, but they differ in their radial concentration and vertical prominence. In superwind galaxies, the UV halos are visible around the whole galaxy and contain filamentary structures seen at other wavelengths and associated with the winds. Galaxies that are being stripped of their ISM by ram pressure have asymmetric UV halos and some extraplanar star formation.

  3. Among a variety of galactic parameters, we found that the UV halo luminosity is only correlated with the galaxy luminosity and the SFR.

  4. The structure, broadband visibility, and strong correlation of with for UV halos lead us to conclude that they are eRN, which supports prior arguments made in the literature.

  5. The frequency of UV halos (100% in our UV-selected sample of highly inclined galaxies within 25 Mpc) indicates that eRN are ubiquitous, but not as extensive as reported in HKB14.

  6. The dust mass of the diffuse component above 2 kpc, inferred from MCRT models that are constrained by the measured and 21-cm maps, is a few percent of the dust mass in the disk, and perhaps 10% of the dust mass seen in filamentary extraplanar structures at lower heights. Deeper FIR observations are needed to constrain in the halo dust for a direct measurement.

  7. There is tentative evidence for a change in the dust properties with height in starburst galaxies, but normal galaxies are consistent with constant dust properties between 2-12 kpc.

We expect virtually every star-forming galaxy to form an eRN, but their detectability in a given sample is a strong function of the UV luminosity of the galaxy and the inclination. Beyond about 25 Mpc it becomes difficult to isolate filamentary structure with GALEX or Swift, but within this limit the UV halos around normal galaxies appear smooth, and there are no clear features in the FUVNUV color maps. One consequence is that the height profile used to constrain MCRT models (Seon et al., 2014; Shinn & Seon, 2015) provides a reasonable estimate of the dust mass, assuming a thick disk morphology. UV observations of more distant starburst or wind galaxies may not discern the wind structure, but an unusually extensive or blue UV halo may indicate the presence of a wind.

Our results confirm the presence of diffuse dust around galaxies as a general phenomenon. The dust is a reliable tracer of material that, at one point, originated in a galaxy disk, so the amount and physical properties (grain size distribution, chemical composition, temperature, etc.) tell us about the role of stellar (or possibly AGN) feedback in the history of the galaxy and the pollution of the circumgalactic medium with metals. However, much work remains: a larger, less biased sample of UV halo properties is needed, at the cost of poorer resolution. This will be addressed in Paper II. In Paper III, we will constrain dust properties in a simple model using the UV SED. Beyond this, we need to understand the dust outflow mechanism (the roles of radiation pressure and hydrodynamic entrainment), connect the UV SED to the FIR SED, and determine if (or how) the dust within 10 kpc of the galaxy connects to a larger circumgalactic component.

We thank the anonymous referee for a helpful report that improved the quality of this paper. 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. We acknowledge the usage of the HyperLeda database (http://leda.univ-lyon1.fr). E. H.-K. and J.C. gratefully acknowledge support from NASA grant NNH13ZDA001N-SWIFT. Facilities: GALEX, Swift


  • Abramson et al. (2011) Abramson, A., Kenney, J. D. P., Crowl, H. H., et al. 2011, AJ, 141, 164
  • Allaert et al. (2015) Allaert, F., Gentile, G., Baes, M., et al. 2015, A&A, 582, A18
  • Aniano et al. (2011) Aniano, G., Draine, B. T., Gordon, K. D., & Sandstrom, K. 2011, PASP, 123, 1218
  • Baes & Viaene (2016) Baes, M., & Viaene, S. 2016, A&A, 587, A86
  • Bell & de Jong (2001) Bell, E. F., & de Jong, R. S. 2001, ApJ, 550, 212
  • Bendo et al. (2003) Bendo, G. J., Joseph, R. D., Wells, M., et al. 2003, AJ, 125, 2361
  • Benjamini & Hochberg (1995) Benjamini, Y., & Hochberg, Y. 1995, Journel of the Royal Statistical Society. Series B (Methodological), 57, 289
  • Bertin & Arnouts (1996) Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393
  • Boselli et al. (2010) Boselli, A., Eales, S., Cortese, L., et al. 2010, PASP, 122, 261
  • Bottema et al. (1987) Bottema, R., van der Kruit, P. C., & Freeman, K. C. 1987, A&A, 178, 77
  • Bregman (1980) Bregman, J. N. 1980, ApJ, 236, 577
  • Bregman & Pildis (1994) Bregman, J. N., & Pildis, R. A. 1994, ApJ, 420, 570
  • Buat et al. (1999) Buat, V., Donas, J., Milliard, B., & Xu, C. 1999, A&A, 352, 371
  • Calzetti (2001) Calzetti, D. 2001, PASP, 113, 1449
  • Calzetti et al. (2000) Calzetti, D., Armus, L., Bohlin, R. C., et al. 2000, ApJ, 533, 682
  • Casey (2012) Casey, C. M. 2012, MNRAS, 425, 3094
  • Chung et al. (2009) Chung, A., van Gorkom, J. H., Kenney, J. D. P., Crowl, H., & Vollmer, B. 2009, AJ, 138, 1741
  • Coker et al. (2013) Coker, C. T., Thompson, T. A., & Martini, P. 2013, ApJ, 778, 79
  • Collins et al. (2000) Collins, J. A., Rand, R. J., Duric, N., & Walterbos, R. A. M. 2000, ApJ, 536, 645
  • Condon (1987) Condon, J. J. 1987, ApJS, 65, 485
  • de Vaucouleurs et al. (1991) de Vaucouleurs, G., de Vaucouleurs, A., Corwin, Jr., H. G., et al. 1991, Third Reference Catalogue of Bright Galaxies, ed. Roman, N. G., de Vaucouleurs, G., de Vaucouleurs, A., Corwin, H. G., Jr., Buta, R. J., Paturel, G., & Fouqué, P.
  • Dumke et al. (1997) Dumke, M., Braine, J., Krause, M., et al. 1997, A&A, 325, 124
  • Engelbracht et al. (2006) Engelbracht, C. W., Kundurthy, P., Gordon, K. D., et al. 2006, ApJ, 642, L127
  • Ferrara et al. (1996) Ferrara, A., Bianchi, S., Dettmar, R.-J., & Giovanardi, C. 1996, ApJ, 467, L69
  • Ferrara et al. (1991) Ferrara, A., Ferrini, F., Barsella, B., & Franco, J. 1991, ApJ, 381, 137
  • Haynes et al. (1979) Haynes, M. P., Giovanelli, R., & Roberts, M. S. 1979, ApJ, 229, 83
  • Hodges-Kluck & Bregman (2014) Hodges-Kluck, E., & Bregman, J. N. 2014, ApJ, 789, 131
  • Hoopes et al. (2005) Hoopes, C. G., Heckman, T. M., Strickland, D. K., et al. 2005, ApJ, 619, L99
  • Howk & Savage (1997) Howk, J. C., & Savage, B. D. 1997, AJ, 114, 2463
  • Howk & Savage (1999) —. 1999, AJ, 117, 2077
  • Hughes et al. (2014) Hughes, T. M., Baes, M., Fritz, J., et al. 2014, A&A, 565, A4
  • Hutton et al. (2014) Hutton, S., Ferreras, I., Wu, K., et al. 2014, MNRAS, 440, 150
  • Irwin (1994) Irwin, J. A. 1994, ApJ, 429, 618
  • Irwin et al. (1987) Irwin, J. A., Seaquist, E. R., Taylor, A. R., & Duric, N. 1987, ApJ, 313, L91
  • Kenney & Koopmann (1999) Kenney, J. D. P., & Koopmann, R. A. 1999, AJ, 117, 181
  • Kenney & Koopmann (2001) Kenney, J. D. P., & Koopmann, R. A. 2001, in Astronomical Society of the Pacific Conference Series, Vol. 240, Gas and Galaxy Evolution, ed. J. E. Hibbard, M. Rupen, & J. H. van Gorkom, 577
  • Kennicutt (1998) Kennicutt, Jr., R. C. 1998, ApJ, 498, 541
  • Kennicutt et al. (2008) Kennicutt, Jr., R. C., Lee, J. C., Funes, José G., S. J., Sakai, S., & Akiyama, S. 2008, ApJS, 178, 247
  • Kinney et al. (1996) Kinney, A. L., Calzetti, D., Bohlin, R. C., et al. 1996, ApJ, 467, 38
  • Koopmann et al. (2001) Koopmann, R. A., Kenney, J. D. P., & Young, J. 2001, ApJS, 135, 125
  • Kriek & Conroy (2013) Kriek, M., & Conroy, C. 2013, ApJ, 775, L16
  • Lee & Irwin (1997) Lee, S. W., & Irwin, J. A. 1997, in Proceedings of the 21st Century Chinese Astronomy Conference, ed. K. S. Cheng & K. L. Chan, 95
  • Lehnert & Heckman (1996) Lehnert, M. D., & Heckman, T. M. 1996, ApJ, 462, 651
  • Li & Wang (2013) Li, J.-T., & Wang, Q. D. 2013, MNRAS, 428, 2085
  • Lucero et al. (2015) Lucero, D. M., Carignan, C., Elson, E. C., et al. 2015, MNRAS, 450, 3935
  • Makarov et al. (2014) Makarov, D., Prugniel, P., Terekhova, N., Courtois, H., & Vauglin, I. 2014, A&A, 570, A13
  • Martin et al. (2005) Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJ, 619, L1
  • McCormick et al. (2013) McCormick, A., Veilleux, S., & Rupke, D. S. N. 2013, ApJ, 774, 126
  • Meléndez et al. (2015) Meléndez, M., Veilleux, S., Martin, C., et al. 2015, ApJ, 804, 46
  • Ménard et al. (2010) Ménard, B., Scranton, R., Fukugita, M., & Richards, G. 2010, MNRAS, 405, 1025
  • Nozawa & Fukugita (2013) Nozawa, T., & Fukugita, M. 2013, ApJ, 770, 27
  • Oke & Gunn (1983) Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713
  • Oosterloo et al. (2007) Oosterloo, T., Fraternali, F., & Sancisi, R. 2007, AJ, 134, 1019
  • Popescu et al. (2004) Popescu, C. C., Tuffs, R. J., Kylafis, N. D., & Madore, B. F. 2004, A&A, 414, 45
  • Rand (1996) Rand, R. J. 1996, ApJ, 462, 712
  • Rice et al. (1988) Rice, W., Lonsdale, C. J., Soifer, B. T., et al. 1988, ApJS, 68, 91
  • Roming et al. (2005) Roming, P. W. A., Kennedy, T. E., Mason, K. O., et al. 2005, Space Sci. Rev., 120, 95
  • Rossa & Dettmar (2003a) Rossa, J., & Dettmar, R.-J. 2003a, A&A, 406, 493
  • Rossa & Dettmar (2003b) —. 2003b, A&A, 406, 505
  • Roussel et al. (2010) Roussel, H., Wilson, C. D., Vigroux, L., et al. 2010, A&A, 518, L66
  • Sandin (2015) Sandin, C. 2015, A&A, 577, A106
  • Schlafly & Finkbeiner (2011) Schlafly, E. F., & Finkbeiner, D. P. 2011, ApJ, 737, 103
  • Seon et al. (2014) Seon, K.-i., Witt, A. N., Shinn, J.-h., & Kim, I.-j. 2014, ApJ, 785, L18
  • Shinn & Seon (2015) Shinn, J.-H., & Seon, K.-I. 2015, ApJ, 815, 133
  • Smith et al. (2016) Smith, M. W. L., Eales, S. A., De Looze, I., et al. 2016, ArXiv e-prints, arXiv:1607.01020
  • Strickland et al. (2004) Strickland, D. K., Heckman, T. M., Colbert, E. J. M., Hoopes, C. G., & Weaver, K. A. 2004, ApJS, 151, 193
  • Tüllmann et al. (2006) Tüllmann, R., Pietsch, W., Rossa, J., Breitschwerdt, D., & Dettmar, R.-J. 2006, A&A, 448, 43
  • Vollmer et al. (2016) Vollmer, B., Nehlig, F., & Ibata, R. 2016, A&A, 586, A98
  • Walter et al. (2004) Walter, F., Dahlem, M., & Lisenfeld, U. 2004, ApJ, 606, 258
  • Weingartner & Draine (2001) Weingartner, J. C., & Draine, B. T. 2001, ApJ, 548, 296
  • Weliachew et al. (1978) Weliachew, L., Sancisi, R., & Guelin, M. 1978, A&A, 65, 37
  • Westmeier et al. (2013) Westmeier, T., Koribalski, B. S., & Braun, R. 2013, MNRAS, 434, 3511
  • Whitney (2011) Whitney, B. A. 2011, Bulletin of the Astronomical Society of India, 39, 101
  • Wiegert et al. (2015) Wiegert, T., Irwin, J., Miskolczi, A., et al. 2015, AJ, 150, 81
  • Wood et al. (2001) Wood, K., Smith, D., Whitney, B., et al. 2001, ApJ, 561, 299
  • Yim et al. (2014) Yim, K., Wong, T., Xue, R., et al. 2014, AJ, 148, 127
  • Yoshida et al. (2011) Yoshida, M., Kawabata, K., & Ohyama, Y. 2011, PASJ, 63, 493
  • Young et al. (1996) Young, J. S., Allen, L., Kenney, J. D. P., Lesser, A., & Rownd, B. 1996, AJ, 112, 1903
  • Yun et al. (1994) Yun, M. S., Ho, P. T. P., & Lo, K. Y. 1994, Nature, 372, 530
  • Zschaechner et al. (2012) Zschaechner, L. K., Rand, R. J., Heald, G. H., Gentile, G., & Józsa, G. 2012, ApJ, 760, 37

Tables 9 and 10 contain the Swift and GALEX observation IDs for the Mrk 501 PSF measurements and the galaxies used in this work, respectively. The GALEX PSF models were based on data sets for 3C 273 (obsID GI4_012003_3C273) and PKS  (obsID PKS2155m304).

uvm2 uvw1 uvw2
30793006 30793001 30793006
30793007 30793002 30793007
30793008 30793003 30793008
30793009 30793004 30793009
30793010 30793006 30793010
30793011 30793007 30793011
30793012 30793008 30793012
30793013 30793009 30793013
30793014 30793010 30793014
30793015 30793011 30793015
30793016 30793012 30793016

Note. – Swift observations of Mrk 501 used to construct the UVOT PSF models used in this paper. The full table is available online.

Table 9Swift Mrk 501 Observations
Galaxy ObsIDs
FUV NUV uvw2 uvm2 uvw1
M82 GI1_071001_M81 GI1_071001_M81 31201001 31201001 31201001
32503004 31201002 32503002
32503016 32503003 32503015
32503020 32503007 32503018
32503028 32503008 32503022
32503032 32503011 32503025
32503036 32503017 32503026
32503040 32503021 32503034
32503046 32503024 32503043
32503057 32503029 32503045

Note. – Swift and GALEX observations used in this paper. The full table is available online.

Table 10Swift Mrk 501 Observations
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description