Abstract

Poster - Splinter EScience   (MW-2235)

Modelling AGN Optical Lightcurves via Inference of AD-BLR delay (MAGNOLIA)

Iliana Cortés, Nikolaos Gianniotis, and Kai Lars Polsterer
Heidelberg Institute for Theoretical Studies (HITS)

MAGNOLIA is a machine-learning framework based on Gaussian Processes (GPs) to address the challenge of disentangling multiple spectral components and observational effects in time domain observations of Active Galactic Nuclei (AGN). AGN are among the most luminous sources in the Universe, fuelled by the supermassive black holes (SMBH) at their centres. Their UV-optical emission originates primarily from two regions: the accretion disk (AD) and the broad-line region (BLR). The BLR reprocesses radiation from the AD, producing a delayed response whose timescale depends on the SMBH mass. We construct a spectral model that explicitly accounts for both AD and BLR contributions, while a latent GP describes their temporal variability. We demonstrate that our framework can recover the AD-BLR time delay using only photometric observations and a single-epoch spectrum, opening a path toward reverberation mapping with limited spectroscopic data.