Reducing the Dimensions of Active Galactic Nuclei Light-curve Manifolds

February 2026 • 2026ApJ...998..130H

Authors • Hemmati, Shoubaneh • Krick, Jessica • Stern, Daniel • Desai, Vandana • Faisst, Andreas • Martín-García, Lucas • Gorjian, Varoujan • Haghjoo, Aryana • Nikakhtar, Farnik • Raen, Troy • Sanjaripour, Sogol • Sipőcz, Brigitta M. • Shupe, David

Abstract • The active galactic nuclei (AGNs) glossary is vast and complex. Depending on the selection method, observing wavelength, and brightness, AGNs are assigned distinct labels, yet the relationship between different selection methods and the diversity of time-domain behavior within and across classes remains difficult to characterize in a unified framework. Changing-look AGNs (CLAGNs), which transition between classifications over time, further complicate this picture. In this work, we learn a data-driven, low-dimensional representation of multiwavelength photometric light curves of AGNs, in which the structure of the projected manifold correlates with AGN class and independent spectroscopic properties. Using the NASA Fornax Science Platform, we assemble light curves from Zwicky Transient Facility, Pan-STARRS, Gaia, and Wide-field Infrared Survey Explorer/NEOWISE for two samples: (1) a heterogeneous set of ∼2000 AGNs spanning z ≲ 1, including Sloan Digital Sky Survey quasars, variability-selected sources, and CLAGNs; and (2) a homogeneous sample of ∼65,000 narrow-line AGNs at z ≍ 0.1 with well-characterized optical emission-line measurements. Without using class labels during training, the learned manifolds organize variability-selected AGNs into coherent regions of the low-dimensional space, distinguish between turn-on and turn-off CLAGNs, and place tidal disruption events in distinct regions. Manifold coordinates correlate with key spectroscopic and host-galaxy properties—including stellar mass, [O III] luminosity, and D_n(4000)—demonstrating that heterogeneous multiband variability can be combined in a purely data-driven manner to recover correlations with independent physical diagnostics, without requiring explicit physical modeling. These results show that manifold learning offers a practical, assumption-light approach for integrating time-domain surveys and prioritizing spectroscopic follow-up.

Links