We are interested in GALFA-HI fibers. They tell us about the B-field. But how do they relate the microphysics? Two important unanswered questions are: do they have a characteristic width? There are arguments for a particular width, but GALFA-HI does not have the resolution to resolve them. And what is their dust content? Small grains control ISM heating, as they dominate the surface area, large grains control ISM metal content, as they dominate the grain volume. Both are important indicators of the content of the fibers. To check these we use Planck 353 emission data and WISE emission maps from (Meisner 2013) and compare to the GALFA-HI DR2 data cubes
We snooped around for a region of sky in the (Meisner 2013) maps that fulfilled the following criteria
Was in the Arecibo field of view and thus DR2
Was not contaminated by a lot of junk.
Had clear fiber-like features
Didn’t have much “dark” H2 or CO.
Since the dust maps, both 12 micron and 300 micron, are much less sensitive than the HI maps, we had to look at intermediate latitudes but avoid (Magnani 1985)-like clouds and other high latitude molecular clouds. We found the region WISE-306, 12 degrees on a side, centered at RA = 11.5\(^\circ\), Dec=30\(^\circ\). We then remapped the WISE data to somewhat bigger pixels, and then sampled the GALFA-HI DR2 data cube on those pixels, as well as the Planck extinction map.
First thing we did is run the RHT algorithm (some reasonable parameters) on each of the HI data slices (binned at 4x 0.736 = 3 km/s) relevant to the fibers. Then we made masks from each of those RHT results. In particular, we chose the angle range that corresponded to the HI fiber direction, and integrated the RHT, and set a threshold of 0.1 to make a bitmask. This gives a reasonable mask that “finds” the fibers. Then, using dialation, we make an “off” mask for each velocity slice. We then concatenate these masks. On masks are shown in red, off in blue below.
We then take the WISE map and smooth it to the Planck resolution. We try to be very careful to get it just right, as it the results are sensitive to this. We then find the average flux in the WISE map in the fibers / flux in the off mask. Same for Planck, and the total HI column. We exclude the top of the region where there seems to be some dark H2.
We find that the total HI is enhanced a little in this region, as expected; about 1.5%. Planck dust emission is somewhat more enhanced than this (1.9%) and wise is more than twice as enhanced (3.8%). This indicates that the fibers are more dusty than the surrounding gas, and that they are especially enhanced in small grain / PAH 12 micron emission.
CAVEATS: haven’t smoothed the GALFA-HI data. Not totally sure how sensitive this is to RHT parameters. Also not sure how sensitive it is to getting WISE smoothing dead right.
Furthermore, we want to know how narrow the fibers are. WISE clearly shows the same fibers as GALFA-HI, from the above experiment, so we try to use the RHT on the WISE data to get a guess as to a characteristic, or minimum width of the fibers.
We use the unsharp mask smoothing radius as a way to measure fiber width. If the fibers are very wide, then if we set the smoothing radius to be very small we should destroy all the features, and see no signal. Since there is more power at larger angular scales, the smoothing radius, acting as a high pass filter, essentially sets the size scale of the fibers we are most sensitive to.
We run a series of tests to measure how the RHT responds to fiber widths. First we run the RHT on the WISE data over a range of smoothing radii (1 - 10) with other parameters fixed. Then we generate a phase-scrambled version of the data, keeping the 2D power spectrum fixed. As we suspect this completely destroys the fibers, but does retain some sense of the orientation of the striation. We then run the same RHT, as a normalization routine. We take the ratio of the total of their backprojections as a function of smr as some measure of power in fibers at different size scales.
We also generate fake fibers to test this method. We inject straight lines, of fixed length, and fixed integrated flux, with a range of widths. We then perform the same experiment, with changing RHT widths, and normalize by dividing by the RHT values for the same SMRs for the phase scrambled image, as with the original image. The figure below shows the effects of these three different faked data, and where the real data lives between them. By eye it would seem that the WISE data have fibers no narrower than about 4’ FWHM.
CAVEATS: I have normalized, so far, not by the phase scrambled versions of the tests, but of the original image. I doubt it will matter lots, but who knows. I hold fixed the total flux in the fake lines at a fraction of the overall flux. Not sure if that’s reasonable. There may be some sensitivity to the scale of the the RHT, or the threshold. One needs to do a quick parameter study. The 4’ number seems a little fishy to me – I think there are narrower fibers than we see in GALFA-HI; it may be I have gotten something wrong in all the conversions? I should probably be doing this on slightly higher resolution; I don’t like working so close to the resolution limit, but it will be tons slower to work with larger windows, as Susan is well aware.