Introduction
An analysis of single-band sources by John Carpenter uncovered clusters of detection artifacts at positions corresponding to the diffraction spikes from very bright stars ( > Ks = -2.00 ). To gauge the extent of these artifacts, a systematic review of all catalog sources around very bright stars was performed. These very bright stars were first grouped into bins of 0.5 or 1.0 magnitude and then all catalog sources within two degrees were plotted (one degree for parents fainter than -2 mag). We examined the residual spike artifact sources to attempt to identify them in parameter space. Our analysis focused on examining the rd_flg, ph_qual, and the ?_psfchi flags of the sources.
Note: throughout this page and all pages linked to this page, some plots will have "skip=n" in the title. This notation is used to indicate when sources were randomly selected from the input file and plotted. Due to the sheer number of sources in some of these files, it was sometimes necessary to plot only every nth entry; otherwise, the plot was too dense to read. Also note the labeling of the axes using delx and dely in all the plots. The bright star was fixed to coordinate (0, 0) and delx is the difference in position between a source and the parent in RA; dely is the difference in position between a source and the parent in Dec. Both delx and dely are measured in arcsec.
In the plots below, catalog sources were plotted around stars in the specified magnitude bin. The residual diffraction spike sources become most obvious where the parents are brighter than -2 mag, though some faint spike remnants can be seen in fainter mag bins in the southern hemisphere. MAPCOR and DB_MAPCOR failed to remove all the unreliable sources for several reasons:
- Since all of the brightest parents are variable stars (see the tables below immediately following each set of plots), it is impossible to tell the actual magnitudes of the parents when neighboring scans were observed on different nights.
- Also, variable stars were removed from the parent list used for tuning the version 3 MAPCOR and DB_MAPCOR parameters, so these brightest parents were not even considered. In fact, all parents brighter than 0 mag in that parent list had to be lumped together because there were so few of them; thus the derived parameters were known to be uncertain below 0 mag.
- The weak residual spikes from parents fainter than -2 mag do not statistically stand out very much against the random backgrounds. Since the spike lengths were determined by statistical source excesses in similar delx vs. dely plots, these faint spikes probably would not have been detected during the parameter tuning in any case.
 |  |  |
15 parents | 35 parents | 43 parents |
 |  |  |
8 parents | 17 parents | 30 parents |
 |  |  |
2 parents | 7 parents | 9 parents |
 |  |  |
3 parents | 9 parents | 11 parents |
 |  | |
4 parents | 5 parents | |
The following table is a listing of the brightest stars in the
northern hemisphere
that are generating the most obvious residual
diffraction spike sourcs in the table above:
| alf Tau, HD 29139 -- Variable Star | ||||||
| alf Her, 64 Her A & B, HD 156014 -- Semi-regular pulsating Star | ||||||
| g Her, HD 148783 -- Semi-regular pulsating Star | ||||||
| alf Boo, SV* ZI 1054, HD 124897 -- Variable Star | ||||||
| bet Peg, HD 217906 -- Pulsating variable Star |
 |  |  |
21 parents | 50 parents | 57 parents |
 |  |  |
8 parents | 19 parents | 39 parents |
 |  |  |
8 parents | 17 parents | 15 parents |
 |  |  |
3 parents | 10 parents | 13 parents |
 |  | |
4 parents | 8 parents | |
 |  | |
3 parents | 6 parents | |
The following table is a listing of the brightest stars in the southern hemisphere
that are generating the most obvious residual diffraction spike sources:
| R Dor, HD 29712 -- Semi-regular pulsating Star | ||||||
| bet Gru, HD 214952 -- Variable Star | ||||||
| alf Sco, HD 148478 -- Semi-regular pulsating Star | ||||||
| alf Ori, HD 39801 -- Semi-regular pulsating Star | ||||||
| gam Cru, BV 839, HD 108903 -- Variable Star | ||||||
| R Hya, HD 117287 -- Variable Star of Mira Cet type | ||||||
| W Hya, HD 120285 -- Semi-regular pulsating Star |
Analysis
The pages containing the plots of rd_flg, ph_qual, and ?_psfchi values for the sources in the brightest mag bins are below:
- North, J band
- North, H band
- North, K band
- South, J band
- South, H band
- South, K band
The spike parameters in MAPCOR and DB_MAPCOR assume that the spikes are wedges which are symmetric about delx = 0 or dely = 0. However, the spikes are not actually symmetric; a close examination of the delx vs. dely plots will show that in the northern hemisphere, the northern half of the western spike, the southern half of the eastern spike, and the eastern half of the southern spike are longer and/or wider, to varying degrees. For the southern hemisphere, it is the southern half of the western spike, the southern half of the eastern spike, and the eastern half of the northern spike.
To try to separate the possible spike sources in the combined source lists from the sources not likely to be affected by residual spikes, we used tight horizontal and vertical rectangles which were not symmetric about 0 -- basically the tightest criteria that could be defined in a simple awk script. The resulting counts of sources inside and outside these rectangles for the brightest mag bins are in the following table; these sets of sources are also the input data for the PSF chi2 histograms found on the individual band plot pages linked above.
Band # of 1-band srcs # of 2-band srcs in approx.
spike rectanglenot in approx.
spike rectangle% in rectangle in approx.
spike rectanglenot in approx.
spike rectangle% in rectangle North J 591 4115 12.6% 231 16045 1.4% H 872 698 55.5% 333 17881 1.8% Ks 804 1299 38.2% 252 3959 6.0% South J 2915 15772 15.6% 1289 73400 1.7% H 3044 2894 51.3% 1570 72105 2.1% Ks 5029 2818 64.1% 1694 11112 13.2%
We also counted the number of sources inside and outside of these rectangles that were in scans taken on different nights than the parent, and those taken on the same night as the parent. The triple-band sources were also added to the single- and double- band sources located outside the spike rectangles. Neighboring scans taken on different nights than the scan containing the parent presumably are more likely to have residual spikes, since all of these parents are variable stars and thus we don't know their actual magnitudes on those different nights.
However, the fraction of unreliable sources from different nights than the parent is not always higher than the fraction of reliable sources from different nights than the parent (see the table below). Even when the sources belonging to parents that had no residual spikes were removed from the unreliable population (one parent in each band in the north and one parent in the south K band), the fractions didn't change very much. Possible explanations are that the "unreliable" data sets contain real sources, but we don't know how many; also, the differences between the parent magnitudes on the different nights are completely unknown. We believe the parent variability is still an important effect, especially since we can see spikes disappearing and reappearing on the plots approximately at integral scan-width distances, but this analysis was unable to prove it.
Band # of 1- & 2-band srcs in approximated spike rectangles total # not in approximated spike rectangles in different night in same night % in diff night in different night in same night % not in diff night North J 298 352 42% 89868 64803 58% H 641 456 58% 95293 102225 48% Ks 575 455 56% 110539 98902 53% South J 2552 1652 61% 346912 73049 83% H 2398 2216 52% 334050 71738 82% Ks 2982 3160 49% 496848 269697 65%
Some of the delx vs. dely plots of single- and double-band sources and specific ph_qual sources show increased densities of sources in circles around the parents' halos. At first glance, these might be interpreted as residual confusion artifacts. But since these increased densities are not found in any of the delx vs. dely plots on this page, which contain all rd_flg and ph_qual sources, it is more likely that these single- and double-band sources are actually real sources with one or two bands' detections removed by the varying artifact confusion radii in MAPCOR.
Conclusion
The analyses confirm the vast majority of these possible residual artifact sources are single- and double-band sources. Traces of artifacts show up in some triple-band sources in the southern hemisphere K band, but these all have low photometric quality. The PSF chi2 values tend to be higher for the population of residual artifact sources, but the extended high-chi2 tails of both populations makes it impossible to use this quantity alone to separate the populations.
The approximate symmetric boundaries (in arcsec) of the residual spikes are found in the table below. The "width" parameter refers to the spike half-width at its widest point; usually this is at the end of residual spikes farthest from the parent, especially for the east and west spikes. However, for the north and south spikes it may be located near the middle of the residual spikes because of their rope-like, twisted-strand structure.
Band north spike west spike south spike east spike width y1 y2 x1 x2 width width y1 y2 x1 x2 width North J -- -- -- 500 3000 110 -- -- -- -2000 -500 50 H -- -- -- 400 5200 160 25 -1100 -700 -2000 -400 50 Ks -- -- -- 700 5200 200 40 -5000 -3000 -2000 -500 50 South J 50 1000 5000 800 2000 40 -- -- -- -4500 -800 160 H 50 800 5000 800 2000 40 -- -- -- -4500 -800 160 Ks 50 800 5100 400 2000 40 -- -- -- -4800 -400 170
R. Tam, T. Evans - IPAC
Last Update - 8 Nov 2002