On the search for rare AGNs with machine learning

AGNs are among the most energetic phenomena in the universe, detectable across an extensive redshift range extending beyond z>7. Thus, expanding the known AGN population is essential for understanding the accretion physics of SMBHs and galaxy evolution. Variability is a key feature of AGNs, yet the mechanisms behind it are not fully understood, and studying AGNs with anomalous variability can help uncover the processes driving their emission. In this study, we employ machine learning methods on data from the Zwicky Transient Facility (ZTF) to identify AGNs exhibiting unusual variability (e.g. changing-look AGN). We combine high-cadence optical observations from ZTF with 12-band photometry from the Southern Photometric Local Universe Survey (S-PLUS), consisting of 5 broad and 7 narrow bands covering the southern sky, expected to improve detection of sources with subtle spectral features and high-z AGNs. We also introduce QuCatS, a catalog containing over 600,000 quasar candidates with photometric redshifts derived from S-PLUS. Given the significant decrease in known quasars beyond z~3.5, we also target high-z quasars as potential anomalous candidates. Promising candidates will receive spectroscopic follow-up observations; notably, we have already identified 4 luminous quasars around z~3 using Gemini Telescope data. This work demonstrates the effectiveness of integrating variability analyses and multi-band photometry to deepen our understanding of AGN emission and variability.