Modern ΛCDM theory predicts more dwarfs in the Local Volume than what our telescopes have found. Most of these unseen dwarfs are predicted to be ultra-faint galaxies that would be detected by their individual member stars. Thus, the most reasonable explanation for the current number of observed galaxies is the limited sensitivity of existing galaxy surveys. One possible way to detect more distant ultra-faint galaxies is to combine deep optical and near-infrared imaging to produce a more complete dataset of the sky. Hence, we used python codes as well as simulated data from the Vera Rubin Observatory and Nancy Roman Space Telescope to study the star-galaxy separation techniques that are critical for the detection of more ultra-faint galaxies; specifically, we used DESC DC2 and Roman DC2 Simulations as our main catalogs. Results show that the classification accuracy for Rubin Observatory data alone is very good for bright objects, but becomes inaccurate for dimmer ones due to the abundance of faint unresolved galaxies. We identified distinct patterns in the color-color and color-magnitude distributions of stars and galaxies. We further demonstrated a method for star-galaxy separation based on the distance from the color-color stellar locus distribution in addition to a morphological classifier, providing an adjustable classifier to have a more complete or more pure stellar sample in the Rubin data. This research demonstrated that it is possible to increase the stellar efficiency using alternative methods in addition to extendedness. Further research is needed to optimize star-galaxy classification with the combination of ground-based optical and space-based near-IR imaging.
Acknowledgement: This work was supported by the National Science Foundation’s REU program in Astrophysics through NSF award AST-1852136.