A wide range of data products are created for each galaxy. We have assembled a selection of these for each galaxy, which are presented in the appendix. See the appendix for a detailed listing of each panel. In summary, the 12 panels show:
(a) and (b): the Stokes I maps for 2 C-band weightings (robust 0 and a
(c): Stokes I map for one L-band weighting (robust 0 only),
(d): an optical image with contours and apparent B-vectors, both from the uv-tapered weighting of C-band
(e) and (f): polarization intensity maps with apparent B-vectors overlaid for C and L-bands, respectively
(g): a composite image of different weightings/bands (see 4.2),
(h), (i), (k) and (l) spectral index maps (h, i) with corresponding error maps (k, l) in C and L-bands respectively.
(j) a wide view of the L-band field.
Tables 4 and 5 list beam sizes, rms noises and dynamic ranges for both bands and weightings.
4.1. Rms noise
All rms noise values mentioned throughout this paper have been measured from non-primary beam corrected maps. For C-band, the rms noise is measured as an average of the regions throughout the imaged field of view which do not contain detectable background sources.
For L-band, however, there are so many background sources near the galaxy that it is difficult to find sufficiently large regions within which the rms can be consistently measured. Residual cleaning artifacts also tend to be larger at L-band, especially close to the galaxy where the primary beam response is high. Therefore the rms tends to be variable when measured near the source, but declines to more consistent values with distance from galaxy as the primary beam response declines. For consistency, we quote rms values far from the galaxies where there are smaller variations in rms between measurement regions. Rms levels near the galaxy could be up to a factor of two higher than the quoted values of Tables 4 and 5 but it is apparent from the panel c) of the appendix figures, that our choice of 3σ as the lowest contour generally well represents the faintest believable emission.
4.2. Displaying the halo
Panel (g) of the figures in the appendix provides an exploratory image intended to help the viewer discover information about individual radio halos. For example, one may like to know if point sources in the disk (anti-)correlate with very extended halo emission. The CHANG-ES survey is data rich, with two (or more) weightings provided per band and overlaying contour plots of these for both bands can result in confusing diagrams. Therefore we overlay colourized "transparent" images of the weightings so that the viewer may relatively quickly apprehend the relationship between structures in the two observing bands. Additionally, artifacts, confusion structures, and non-random noise are difficult to remove mathematically from radio data. Our visualization approach uses masking in order to mitigate "background" artifacts.
First the fits data are stretched using the KARMA 11 visualization package’s kvis task. On intensity-inverted, logarithmically scaled data, we use the "greyscale 3" option in the pseudo-colour tool for adjustments. The resulting images are saved in eps format and used as input into the Gnu Image Manipulation Program (GIMP) 12. This package allows the user to stack images in "layers" that can be combined, as if they are transparent, using a variety of blending mode algorithms. Usually we stack four images, i.e. two weightings of each observed band. We colourize the C band data (blue for the robust 0 weighted data and green for the uv-tapered weighting) while the L band data are left as greyscale images. The order in which the images are combined, and which algorithms are used, are described in Fig. 2. In the resultant qualitative image, often both point sources in the disk and diffuse halo structure are evident simultaneously.
Figure 2. Construction of combined weightings and bands in panel (g) of the figures in Appendix. The layering procedure in the Gnu Image Manipulation Program (GIMP) is used to combine the available weightings of the C and L band data. ("rob0" stands for the robust 0 weighting and "uvtap" is the uv-tapered weighting.) The algorithmic mode applied to the top layer is listed under the upper image in each column. The resulting combination is displayed in the bottom row. The left hand column represents the first 2 layers that are combined, i.e. inverted-intensity and colourized versions of the two weightings of the C band data. The second step, represented in the middle column, combines the inverted-intensity higher resolution L band data with the result of Step 1. The right hand column shows how the inverted-intensity, lower resolution L band data are applied to the result of Step 2 in order to mask out confusing structures in the off-target "background". The bottom image in the right hand column is presented as panel (g).
4.3. Comments on individual galaxies
A few galaxies warrant some extra comments, either because of differences in the data reduction procedures, or because the results were interesting or unusual and therefore caught our attention. In this section, we list these galaxies with comments. Note that this is not meant to be a thorough discussion of each galaxy.
4.3.1. NGC 660
The data of polar ring galaxy NGC 660 have the highest dynamic range in both bands, due to a strong central source. Consequently, the resulting images have higher than expected rms values, particularly in L-band. Nevertheless, the achieved signal to noise of the C-band map with robust weighting reached 70000, on the higher end of what best can be expected (see Section 3.3.2).
In spite of careful cleaning, including peeling 13 of the nearest of the two strongly interfering field sources in L-band (see Section 3.3.6), artifacts are still present. This could potentially be affecting the spectral index results, where regions of flat spectral index are crossing the disk. However, these regions could also be an effect of the possible AGN (see 5.2.1).
We note that the polarization intensity is low, with its fraction of polarization over Stokes I intensity just a 10th of the 0.5% which is considered a believable signal (see Section 3.3.3).
4.3.2. NGC 3556 and NGC 5775
Both of these galaxies are well-known to have significant halos, which we can also see in our data. Moreover, they both show very flat spectral indices at C-band throughout the disk, where the errors in spectral indices are the lowest. Neither data set showed any particular problems and were single pointings. Bearing in mind the summary regarding spectral index in Section 3.4.3, this suggests that thermal emission may be more strongly dominant for these galaxies in their disks.
For NGC 5775 at L-band, some broadscale polarization features have been found in the field that are most likely from foreground Galactic emission (this is more obvious when a larger field than shown in the appendix is displayed). The Galactic coordinates of this source place it almost directly over the Galactic centre, although at reasonably high Galactic latitude. Indeed, the Galactic coordinates of NGC 5775 (l = 359.4∘, b = 52.4∘) indicate that we view it through the northernmost tip of the Fermi bubbles that extend 55 degrees above and below the Galactic centre (Ackermann et al. 2014). Substantial polarized emission at 2.3 and 23 GHz has been found to coincide with the Fermi bubbles, including ridge-like filamentary structures crossing through this particular location (Carretti et al. 2013). Our polarization images are likely picking up this same extended foreground structure at lower frequency.
4.3.3. Three Virgo galaxies: NGC 4192, NGC 4388 and NGC 4438
These galaxies are highly affected by contamination from strong sources in the field, such as M87, rendering higher than normal rms noise and artifacts hard to clean out. This may have an effect on both spectral index results and polarization.
We note that the C-band spectral index is steep in the inner disk for NGC 4192 (a rather weak source at the low end of our flux density cutoff for the survey), and that the uncertainties are high.
NGC 4388 was particularly affected by M87, and we made an attempt at peeling the source in L-band, as well as self calibrating on M84 (the other disturbing source in the field), which rendered some improvement. The resulting rms at L-band was too high for us to detect significant polarized flux (panel f).
Also NGC 4438 is strongly affected by residual side lobes from M87. Despite attempts of self calibration and peeling, the rms could not be brought down to lower than almost three times the expected value. L-band observations of NGC 4438 differ from the other galaxies, in that only the upper half of the frequency range (spectral windows 16-31) was used for imaging, with a central frequency of 1.77 GHz. Q and U images were made over the same upper half of the frequency range in order to match the total intensity image, as well as the spectral index image.
4.3.4. NGC 4594
The strong centre of NGC 4594 (the Sombrero galaxy) resulted in residual side lobes which made it difficult to detect the weak east-west disk for the two-pointing data set in C-band. Careful selection of self calibration inputs eventually revealed the disk, and subsequent merging of the two pointings further helped to cancel out artifacts from the centre.
In L-band, a plume is seen to the north of the core and is roughly in the direction of the jet observed at much higher resolution by Hada et al. (2013). An unusual polarization structure and a flat spectral index associated with the core and towards the north are also observed. A more conservative spectral index cut-off could be beneficial for this data set (for example, a 10σ cut-off was adopted for the in-depth study of NGC 4845, see Irwin et al. 2015 (accepted by ApJ).
4.3.5. NGC 4631
The C-band spectral index shows an unexpected flattening from the central to the outer disk for this two-pointing data set. This gradual deviation has not been detected in previous observations of this galaxy (Hummel & Dettmar 1990), but is within the errors pointed out in point d) of Sec. 3.4.3.
4.3.6. NGC 4666
Despite its small angular size well within the symmetric primary beam, the C-band spectral index displays an asymmetry between the two halves of the disk, such that it is flatter on the south-west side compared to the north-east. This asymmetry corresponds to an asymmetry in the polarization at the same frequency. Interaction with a nearby companion, NGC 4668 (to the south-east) could be a factor, and/or motion through an intergalactic medium. No other technical issues (antenna flexure, solar influence) are problematic.
4.3.7. NGC 4845
It is worthwhile to note that this galaxy displays an interesting variation in flux density level and further exploration, being published in CHANG-ES V (Irwin et al., 2015, accepted by ApJ), indicates with little doubt that the source is indeed variable. CHANG-ES V also explores the spectral index and polarization results in more detail.
4.3.8. NGC 5084
Due to the low declination of this galaxy, the C-band observations were divided up into three scheduling blocks in order to a) avoid shadowing and b) increase uv-coverage. The scheduling block in which the first pointing was observed did not have a polarization leakage calibrator scan, and instead the secondary calibrator used for NGC 4192 was used to calibrate polarization leakage (the secondary calibrator of NGC 5084 did not have a sufficiently large parallactic angle span). Part of the second pointing was strongly affected by shadowing and only 21 antennas were used for these scans.
The C-band images show emission extending east and west of the point-like centre, as well as southward from the eastern extension – these are likely real, since their intensity is greater than 10σ (and they roughly follow the disk). However the extensions seen above and below the centre may be spurious since they seem to align with weak residual side lobes. In L-band, we find no apparent polarization from the source. The spectral index is flat in L-band and flat to slightly positive in C-band (but the positive trend coincides with higher error values).
4.3.9. UGC 10288
UGC 10288 fell within the intensity cutoff of the CHANG-ES sample by piggybacking on the previously unresolved strong background source to the west of the disk. We refer to Paper III for full details on the observations and analysis of this galaxy.
11 See http://www.atnf.csiro.au/computing/software/karma/ Back.
12 See gimp.org Back.
13 Only the Stokes I and spectral index images were produced from the peeled data, while polarization images were not, since the off-centre confusing sources are not as problematic in Q and U as for total I. Back.