ML-SMBH · Multifrequency and Machine Learning methods to Search for Early Super Massive Black Holes

Multifrequency and Machine Learning methods to Search for Early Super Massive Black Holes

This project aimed to leverage multifrequency astrophysical data to identify high-redshift blazars and quasars, essential for exploring the early Universe and the Epoch of Reionization (EoR). The goal was to uncover new high-redshift jetted supermassive black holes (SMBHs), guiding observational campaigns, directly contributing to understanding galaxy evolution and SMBH growth during the EoR. The PI developed a multifrequency data frame representative of blazars detected from radio up to gamma-rays, creating the First Cosmic Gamma-ray Horizon (1CGH) catalogue. The catalogue nearly doubled the number of sources detected above 10 GeV and is now accepted for publication in MNRAS. The 1CGH catalogue allows precise measurements of Extragalactic Background Light (EBL) density up to z ~ 3.2, enabling accurate estimates of star formation rates (SFR) at high redshifts (z ~ 6-7), within the EoR. A critical gap was addressed, revealing that 72% of the sources lack robust redshift characterization, thus strategically guiding future observational efforts. The resulting multifrequency dataframe will be publicly available on Vizier. Additionally, a sample of high-z jetted quasars was selected using data from the radio Rapid ASKAP Continuum Survey (RACS) combined with deep wide-area optical/near-infrared surveys. This resulted in selecting 45 new high-z radio quasar candidates, 24 spectroscopically confirmed, including 11 at z >5. Results published in "High-z radio Quasars in RACS I: Selection, identification, and multi-wavelength properties” (accepted in A&A) significantly update the density estimate of jetted SMBH at high redshift. Throughout this project, the PI actively focused on machine learning (ML) methods and astrophysical data handling, conducting educational initiatives at the Institute of Astrophysics (IA) in Lisbon, teaching multifrequency data analysis and ML applied to astrophysics to PhD, MSc, and internship students. These efforts notably resulted in developing an ML photometric model capable of predicting quasar redshifts up to z~7.5. This model can efficiently identify high-redshift quasar candidates and inform observational campaigns. Furthermore, leveraging expertise in gamma-ray analysis, the PI explored heliophysics, discovering unexpected anisotropy and temporal variability in gamma-ray emissions from the solar disk during the 2014 solar maximum. This groundbreaking result led to a highly impactful publication in the Astrophysical Journal (ApJ) and extensive media coverage, significantly exceeding the project's original expectations.

Data: CORDIS, © European Union