Luis Suelves

Narodowe Centrum Badań Jądrowych

Sesja VII: Kosmologia i ewolucja galaktyk

Czwartek 14.09.2023 17:12 – 17:24

abstrakt:
Galaxy merger identification is a key step in the contemporary studies of galactic evolution. Their abundance and physical properties can answer many questions about their interaction processes and life cycle. In this talk, we will explain the methodology for merger identification that we developed through a Neural Network (NN) for galaxy mergers classification, using as input only photometric information from SDSS DR6. For training, we built a class-balanced set, with merging and non-merging galaxies from Galaxy Zoo DR1 visual classification in redshifts between 0.01 and 0.1. The mergers are visually confirmed galaxy pairs from Darg et al (2010). We discovered that the error in the 5-band sky background estimation allowed the NN to achieve a 92.64 ± 0.15 % of training accuracy, and a 92.36 ± 0.21 % in test. Moreover, we found out that this sky error is enough for the classification: simply drawing a decision boundary in the g – r bands plane, one can get a 91.59 % using all our data. We consider that this sky background error is sensitive to the stripped material surrounding the merging sources. Currently, we are extending this decision boundary to all SDSS DR6 within Galaxy Zoo DR1, addressing the stars and galaxies forming visual pairs that contaminate the results. Our plan is to build a pipeline to discard sources whose sky background is affected by the contamination, rather than by merging related low surface brightness material. The proper understanding and extension to deeper images of this methodology could support LSST’s science very effectively.