T. Jarrett, IPAC
This memo describes the results from a "second pass" at the M51 data. We have modified GALWORKS to operate on galaxies as large as M51 -- if for any reason to see what happens when such a large object is fully processed by GALWORKS. An initial effort to process M51 was described in the memo M51 -- What Happens within GALWORKS? . Again, we will discuss the background determination, including detailed J and K images of intermediate steps, and of course we describe the results of standard GALWORKS production operations on M51 and its large companion. Also, see the ADDENDUM at the end of the memo for a brief discussion of what can be expected for the super large galaxies (e.g., M31).
The whirlpool galaxy, M51, is a large nearby spiral galaxy with a barred companion. It fills about half of one coadd. In the GALWORKS known-galaxy catalog, it has a recorded major axis size of 11.2 arcmin, while its companion has size 5.8 arc min.
The data were acquired with the 2MASS 3-channel camera and 2MASS telescope atop Mt. Hopkins. Thus, this data represents our first efforts at galaxy analysis with real data. M51 is the largest galaxy (to date) to pass through the GALWORKS end of the pipeline. Since M51 is so large, it surely affects the coadd background, blanking and source extraction, so it behooves us to check what actually happens during production. This memo provides some details into this matter.
Search Galaxy Database for Large Galaxies
The first step that GALWORKS performs is to check the database catalog for any large galaxies that may be within the coadds or partially covered by the coadds (very large galaxies). For scan 006 of the night of 970418, it finds two galaxies, M51 and its companion, within coadd "184". The galaxies have a recorded size of 11.2 arcmin and 5.8 arcmin, respectively. More importantly, both galaxies are nearly fully contained on one coadd, thus we should be able to process them. Additional postage stamp images were "cut" and extracted from the coadd since these galaxies are known cataloged objects. This information was recorded in the file "bright006.gals" under the "extd" directory of scan 006.
Background Determination
Before the background can be properly determined, all detected sources (including galaxies) must be blanked from the coadd. With sources blanked from the coadd, the resulting 'cleaned' image should contain only "background" light and whatever diffuse light from large galaxies and bright stars not blanked in the coadd. The background is computed by passing the cleaned image through a median filter and iteratively fitting the residuals with a cubic polynomial function (further details on this operation can be found in the GALWORKS SDS).
Since M51 and its companion are so large, nearly half of the coadd is therefore blanked away (all of it in the lower half of the coadd) rendering background subtraction a bit less than routine. The following images show some of the key intermediate steps in this operation. Note: the images are grey-stretched from about -1 to 3 sigma to bring out the background details.
Left panel: Raw coadd image with M51. Image size = 8 X 16 arcmin
center panel: Coadd image after stars, M51 and companion blanked.
right panel: After Median filter: 16 X 16 filter window
Notice from the images that even after the galaxies are blanked, there does remain a slight diffuse component (due to M51 and companion) in the cleaned image. The effect still remains in the median filter image, which means that it probably will be present in the final background solution since we fit a cubic polynomial to a median-filtered cleaned image, any diffuse component will be easily detected if it has a size scale greater than 1-2 arcmin.
The diffuse component is amplified because there is little or no information in the southern half of the coadd (because of source blanking) to offset this emission. This case only occurs when a large object (e.g., nearby galaxy) fills the coadd thus rendering background determination less certain. Fortunately this case will happen very infrequently.
After the median filter step, each column is fit by a cubic polynomial. The resulting image can be seen below in the middle panel. Blanked pixels are replaced by the mean value of the pixel distribution in the same column that the blanked pixels are located. The "streak" effect seen in the column solution (see middle panel) is due to this pixel replacement. Note that this effect is completely smoothed away when a row solution is applied to the column solution (see right panel).
Left panel: After Median filter: 16 X 16 filter window
center panel: Column solution
right panel: Row Solution to Column fit (i.e., 2-D solution)
The final background solution and background subtracted coadd are shown in the final set of images.
Left panel: M51 raw coadd
center panel: background solution
right panel: background subtracted coadd
The background appears to be reasonably fit to the data given the difficulty of this case. Of course we realize that some of the M51 galaxy has been subtracted out since we did not fully blank the galaxy (which is impossible because it fills the entire coadd if you include the diffuse component and its companion galaxy). Nevertheless the background subtracted image appears to be clean and free of any artifact structures from the fitting solution itself. This is in fact our goal, since measuring the various properties of M51 to a high degree of accuracy is not our goal here (2MASS is not designed to measure the properties of giant nearby galaxies), but rather we hope to find fainter galaxies and extended objects within the same coadd as M51. The last section of this memo describes what was found other than M51.
GALWORKS Production of M51
Any large known galaxies are processed first before star-galaxy discrimination operations are performed on the detected sources. One of the first steps is to construct a "super" coadd image from the individual JHK images and compute the elliptical parameters. The following images are centered on the M51 field, but show only the area that it used by GALWORKS in its analysis of this object (the maximum size that GALWORKS operates within is about 80 pixels in radius).
Needless to say GALWORKS could not estimate the elliptical parameters for the "super" image because the 3-sigma contour extends well beyond the maximum window area that GALWORKS considers. The same was true for the individual JHK images. The elliptical parameters were set to the defaults: axil ratio = 1.0, position angle = 0.0, semi-major axis radius = maximum window size = 80 pix. The companion galaxy was also too large to estimate elliptical parameters at the 3-sigma level.
Star Subtraction
Five "stars" were detected near M51 which were subtracted from the coadd (in this case, they were blanked). These sources are, of course, not stars but part of M51 itself (probably H II regions). This demonstrates the danger of subtracting anything near large galaxies since we easily resolve H II regions and other structures in the spiral arms of M51. However, for most other galaxies we do not resolve structures associated with the galaxy itself and thus indescriminate star subtraction is the best best way to minimize contamination from foreground stars.
After star subtraction, blanked pixels are "cleaned" from the subimages by applying isophotal substitution. Remember that GALWORKS assumes that M51 and its companion are circular. The following images show the M51 subimages and the companion to M51 subimages. Since the companion was processed after M51 (because M51 is larger), a large fraction of the area near the companion was blanked after the completion of the M51 processing (once an object is completed, the area encorporating the object is blanked from the coadd).
Photometry
At this point the only remaining operations are the numerous photometry calculations, including elliptical and circular adaptive aperture photometry, isophotal photometry, fiducial photometry, petrosian photometry and fixed circular photometry.
For the adaptive apertures (i.e., flux growth), isophotal apertures (i.e., 20 and 21 mag per sq. arc sec isophot) and the petrosian apetures, the resultant radii are maxed-out at the window size of 80 pixels. So the photometry simply reflects the total flux within 80 pixels radius. The JHK mags in this case are about 7.2, 6.5, 6.2, respectively for M51, and 7.3, 6.6 and 6.3 for the companion. Remember that only a fraction of the galaxy is actually integrated (the inner 80 arcsec).
The fixed circular mags for M51 ranged from 9.3 mag at K for a radius of 5 pixels to 6.2 mag for a radius of 80 pixels. The companion was about 0.2 mag fainter.
The good news is that GALWORKS did not crash when presented with a galaxy that fully fills the maximum working window area.
Galaxy Candidates
GALWORKS found 4 galaxies in the coadd containing M51, two of which are m51 and its companion. The postage stamp images are given below:
Two faint galaxies were also found in the M51 coadd. The first, ID = 5004, is bright enough to estimate the elliptical parameters. The image shows the 20 mag per sq. arc sec isophot aperture and the total elliptical aperture. The second galaxy is too faint (K > 14) to compute the elliptical parameters, so the default values are assumed. The same is also true for the "super" coadd of this object. By inspection we see that the galaxy is a faint edge-on system.
Finally, we note that several "lcsb" candidates were detected in this
coadd. It turns out that all of the candidates are associated with the
M51 galaxy not blanked from the image (i.e., diffuse emission). This
result should be expected any time we have a very large object in view,
including nearby galaxies and galactic nebulae (e.g., Orion). The
database manager should be able to clean these false objects from the
database simply by their association with an "extraordinary" known
object.
Conclusion
We have demonstrated that GALWORKS suitably computes the background solution for the 2MASS coadd image that contains M51 and its companion galaxy. Although M51 and companion comprise over one-half of the coadd itself, the background is well determined for the region of the coadd where the galaxy pair was not subtracted (in this case, the upper 50% of the coadd). So we conclude that for galaxies as big as the M51 system (~10 arcmin) and smaller, GALWORKS is able to determine the background with minimal effect from the galaxy(s) itself. For larger galaxies, this conclusion obviously may not hold. Further testing of larger galaxies is required (e.g., M31).
M51 and its companion passed smoothly through the main processing modules of GALWORKS without any technical hitches. It was observed, however, that GALWORKS was limited at what information (e.g., elliptical parameters, photometry) it could derive from the galaxy image. This will always be the case for galaxies larger than 80 arcsec in radius. In order to do a better job at computing the properties of M51 and galaxies like it, this will require detailed examination of a large mosaic comprised of several coadd images.
Steve Schneider specifically asked me to run a case in which we ignore the known galaxy catalog and see what happens when GALWORKS encounters an "previously unknown" large galaxy -- M51. The results are shown below, but needless to say, GALWORKS breaks M51 into several pieces. The cores of m51 and its companion are detected and extracted, along with pieces of the spiral arms and diffuse emission thereabout. These results hint at what can be expected for M31, the LMC/SMC system and a few other large galaxies. The current plan is to blank and extract known galaxies out to a maximum radius (set at 60 arc min). For those galaxies that extend beyond the maximum exclusion radius, galworks should detect and extract pieces thereof (bright HII regions, e.g.). Until we take a look at M31 or the LMC, we cannot know for sure what will happen.
