A Systematic Search for AGN Flares in ZTF Data Release 23

TL;DR

This study addresses the challenge of identifying extreme AGN flares within large time-domain surveys. By modeling intrinsic AGN variability with Gaussian Processes and leveraging Bayesian-block segmentation for candidate detection, the authors build two catalogs from ZTF DR23: a comprehensive AGNFCC with 28,504 flares and a high-confidence AGNFRC with 1,984 flares. They characterize the demographics, redshift distribution, and potential origins of these events, finding associations with SNe, TDEs, blazars, microlensing, and possible EM counterparts to BBH mergers, while noting many flares remain unexplained by known mechanisms. The catalogs provide a valuable resource for probing accretion physics and transient phenomena in AGNs, though spectroscopic and multi-wavelength follow-up will be essential to disentangle intrinsic variability from true transient events.

Abstract

Active galactic nuclei (AGNs) are known to exhibit stochastic variability across a wide range of timescales and wavelengths. AGN flares are extreme outbursts that deviate from this typical behavior and may trace a range of energetic physical processes. Using six years of data from Zwicky Transient Facility (ZTF) Data Release 23, we conduct a systematic search for AGN flares among a sample of well-sampled AGNs and AGN candidates. We construct two catalogs: the AGN Flare Coarse Catalog (AGNFCC), containing 28,504 flares identified via Bayesian blocks and Gaussian Processes, and the AGN Flare Refined Catalog (AGNFRC), comprising 1,984 high-confidence flares selected using stricter criteria. We analyze their spatial distribution, temporal characteristics, host AGN type and potential origins. Some flares can be associated with known supernovae, tidal disruption events, or blazars, and a few may be linked to binary black hole mergers or microlensing events. These catalogs provide a valuable resource for studying transient phenomena in AGNs and are publicly available at https://github.com/Lyle0831/AGN-Flares.

Figures (18)

• Figure 1: Mollweide projection of the sky positions in equatorial coordinates of the objects used in this work. The resolution is that of a HealPix map with NSIDE=64.
• Figure 2: Redshift distributions of AGNs from different catalogs, binned with a bin size of $\Delta z=0.05$.
• Figure 3: ZTF detection counts distribution for all AGNs.
• Figure 4: Distribution of the highest $p_{\mathrm{flare}}$ values for each AGN. The vertical dashed line represents the threshold of $p_{\mathrm{flare}}=0.998$.
• Figure 5: The detection processing flow for a real AGN light curve. Panel (a) shows the light curve of a real AGN (blue), with the red curve representing the fit obtained using Bayesian block representation. The green shaded regions indicate the time intervals where potential flares are detected. Panel (b) displays the simulated light curves, while panel (c) presents the GP hyperparameters for all the light curves. Panel (d) shows the $\lambda$ values computed for all light curves in the potential flare regions. One point corresponding to the real AGN exceeds the threshold derived from simulations, which corresponds to a flare significance of $p_{\mathrm{flare}} > 0.998$, indicating that this event is likely a genuine flare.