October 1
AGN photometric redshifts in the time-domain era
Speaker: Dr. Sarath Satheesh Sheeba, UNAB/UDP, Santiago, Chile
Abstract: Quasars exhibit stochastic intrinsic variability across a wide range of time scales, providing a powerful probe of accretion physics and a potential tool for redshift estimation. With upcoming large time-domain surveys such as the Vera C. Rubin LSST expected to discover millions of AGN, spectroscopic follow-up will be limited, making robust photometric redshift estimation a central challenge.
In this talk, I present my work developing VAR-PZ, a variability-constrained photometric redshift framework that links cosmological time dilation with intrinsic AGN variability. Using Gaussian process modelling of optical light curves, we constrain the structure function to derive variability-based redshift priors. We combine these variability priors with traditional SED-based priors to improve redshift estimation. Complementing this physics-driven approach, I will present VAR-PZnn, a neural-network extension designed to scale to Rubin-era datasets. VAR-PZnn leverages variability features derived from the ALeRCE light-curve classifier together with multi-band photometric information to predict AGN photometric redshifts directly. This empirical framework provides a data-driven counterpart to VAR-PZ and enables fast inference for large AGN samples. Together, VAR-PZ and VAR-PZnn demonstrate how combining physically motivated variability modelling with machine-learning approaches can significantly improve AGN photometric redshift estimation in next-generation time-domain surveys.
Host: Jillian Rastinejad & Robert Stein