Using J-K Color as a Star-Galaxy Discriminent

I want to open dialogue on the subject of using a color criterion to discriminate between stars and galaxies. In particular, I would like to apply some kind of criterion in the 2MASS pipeline (within GALWORKS) in order to speed up the runtime and save disk space. I do not advocate using such a criterion for low stellar density fields where the level-1 specs apply (to avoid a color bias). Instead, I believe it will be most powerful (and with minimal damage to completeness) when used in high stellar density fields where the extinction is non-trivial in the infrared.

As I will demonstrate below, galaxies are distinctly redder than stars and double/triple stars for these types of regions (to a lesser extent, but still true on average, gals are redder than stars for low density fields as well). For galaxies, their red colors are easy to understand. In the near-infrared the older (red) population of stars in the bulge and core tends to dominate the light. Secondly, the redshift color change (e.g., K correction) generally pushes more light into the near-IR bands so the galaxies appear redder (e.g., 2MASS Galaxy Colors : Hercules Cluster ). Finally, for fields near the galactic plane (glat < 8), the extinction due to dust in the Milky Way works at its most optimum for galaxies (because they are subject to the entire column of gas & dust). For stars, their colors will also be reddened by the galactic dust layer, but to a lesser extent since they are for the most part located (due to the sensitivity of the 2MASS survey) within 1 kpc of the Sun. Indeed, as the plots show below, the J-K colors for stars in low and high stellar density fields are not all that different in comparison to galaxies.

I have performed a detailed classification analysis of scans taken on the night of 971205n (one of the IPAC RTB fields). About 2400 sources belong to low stellar density fields (toward the Perseus-Pisces supercluster of galaxies) and another 720 belong to higher stellar density fields (near the anticenter, at glats between 3 and 8 degrees; the number density ranges between 3.3 and 3.6 in the log for K < 14). The plots below show the colors for the two sets of data. Galaxies are denoted with white filled circles, stars and double stars with red triangles and triple stars with blue crosses (also, artifacts are denoted by red crosses).

Color vs. Mag for Srcs in Low Stellar Density Fields (Click on graph for bigger and better image) note: white circles == galaxies; red triagles == stars/doubles; blue crosses == triple stars; red crosses = artifacts

Color vs. Mag for Srcs in High Stellar Density (>3.5) Fields (Click on graph for bigger and better image) note: white circles == galaxies; red triagles == stars/doubles; blue crosses == triple stars; red crosses = artifacts

A J-K color criterion of 1.0 would eliminate over 80% of the false sources from the high density field, while leaving 99% of the galaxies in tact. Furthermore, this would be done at a point in GALWORKS that would significantly speed up the runtime. It is clear that a color criterion is a positive tool that should be used in GALWORKS. I will continue to perform classification on new 2MASS data in order to improve the statistics.

It is also interesting to note that a color criterion would work with the low stellar density fields (where the level-1 specs apply). The criterion would have to be lower, say J-K = 0.5, to avoid knocking out any blue galaxies. Even with a J-K cutoff of 0.5, most of the --bright-- false sources (double stars dominate) would be eliminated. Question: do we expect to see any galaxies with J-K colors bluer than 0.5? In other words, is there any possibility of imparting a color bias upon our sample of galaxies, or do we instead miss the --rare-- super blue galaxy ? We can probably best answer this question by continuing to look at the colors as we build our statistics.