AGN Flare Refined Catalog (AGNFRC) Overview

• AGN Flare Refined Catalog (AGNFRC) is a collection of multi-timescale and multi-wavelength datasets on variable and flaring AGN, defined by rigorous statistical and filtering methods.
• It integrates 12-year gamma-ray and optical survey data using techniques like Bayesian Blocks and Gaussian processes to isolate robust flare events across various timescales.
• The catalog supports practical applications such as population studies, real-time follow-ups, and multi-wavelength cross-correlations to explore AGN jet physics and variability patterns.

Searching arXiv for the cited AGNFRC-related papers and closely related variability-catalog work. AGN Flare Refined Catalog (AGNFRC) is an acronym applied to several catalog products that isolate or characterize variable and flaring active galactic nuclei across $\gamma$-ray and optical surveys. In the current literature, it denotes a preliminary 12-year Fermi-LAT catalogue of variable AGN based on 3-day, 7-day, and 30-day light curves (Passos-Reis et al., 2 Feb 2026), a ZTF DR23 high-confidence optical flare catalog containing 1,984 events (He et al., 27 Jul 2025), an early Fermi-LAT record of 91 extreme daily $\gamma$-ray flaring AGNs (Buson et al., 2013), and a CAZ-based optical flare blueprint built on CRTS, ATLAS, and ZTF light curves for 7,918 blazar-selected AGN (Kouch et al., 18 Oct 2025). In each case, the catalog product is obtained after explicit variability modeling, thresholding, or flare-shape filtering.

1. Terminological scope and catalog variants

The term AGNFRC is attached to distinct products rather than a single canonical database. The literature explicitly uses the acronym for catalog constructions that differ in waveband, parent sample, flare definition, and unit of tabulation—sometimes one row per source, sometimes one row per flare episode (Passos-Reis et al., 2 Feb 2026).

AGNFRC usage	Survey/domain	Core sample
Multi-timescale Fermi-LAT catalogue	$0.1$–$300\,\mathrm{GeV}$; 3-, 7-, 30-day light curves	1,240 sources
ZTF AGNFRC-V1.0	ZTF DR23 $g$- and $r$-band flare search	1,984 events
Early Fermi flaring-source catalogue	Daily intervals with $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$	91 AGNs
CAZ AGNFRC blueprint	CRTS, ATLAS, ZTF nightly optical light curves	7,918 sources; $\sim 75{,}000$ flares, with $>25{,}000$ prominent

This multiplicity is important for interpretation. In the Fermi-LAT 12-year product, AGNFRC is primarily a variability catalogue with source-level flags and timescale-dependent metrics. In the ZTF and CAZ optical usages, AGNFRC is an event catalogue centered on localized flares. In the early Fermi-LAT usage, the catalogue is defined by a daily flux threshold rather than by a baseline-subtracted flare model (Buson et al., 2013).

2. The 12-year Fermi-LAT multi-timescale AGNFRC

The 2026 Fermi-LAT implementation uses 12-year mission data in the energy range $0.1$–$\gamma$0, binned into 3-day, 7-day, and 30-day intervals, with all light curves extracted from the public Fermi-LAT Light Curve Repository, LCR, of Abdollahi et al. 2023 (Passos-Reis et al., 2 Feb 2026). Its parent sample is the 4LAC-DR3 catalog of Abdollahi et al. 2020, which lists 2,825 blazars, specifically 1,699 LSP, 536 ISP, and 590 HSP sources. Cross-matching 4LAC-DR3 with the LCR yields 1,429 AGN light curves with at least 90% data coverage in all three cadences.

After quality filtering, including outlier removal following Passos Reis et al. 2025 and a requirement of detection threshold $\gamma$1 in at least 75% of bins, the final sample contains 1,240 sources (Passos-Reis et al., 2 Feb 2026). Among these, 407 sources yield a statistically significant positive normalized excess variance on 3-day scales, 1,121 on 7-day scales, and 1,198 on 30-day scales. A source is flagged variable on a given timescale if $\gamma$2 at $\gamma$3 significance. A source is flagged flaring if it exhibits at least one time bin with $\gamma$4 and $\gamma$5 on that timescale.

This catalogue emphasizes population-level variability behavior. FSRQs, predominantly LSP objects, systematically exhibit larger $\gamma$6 at all cadences, with median $\gamma$7, whereas BL Lacs have median $\gamma$8 (Passos-Reis et al., 2 Feb 2026). The dispersion of $\gamma$9 values is broader for FSRQs, indicating more extreme flaring behavior. For $0.1$0 of sources, the ordering $0.1$1 holds, demonstrating a mild increase of variability amplitude with observation duration. Highly luminous sources with $0.1$2 tend to have higher $0.1$3 and more frequent flares, consistent with compact emission regions and efficient particle acceleration.

The catalogue also defines “prime flare candidates” by ranking sources on each timescale according to the product $0.1$4 and selecting the top 20; all have at least one bin with $0.1$5 (Passos-Reis et al., 2 Feb 2026). Example 3-day entries include 4FGL J2254.0+1608 (3C 454.3; $0.1$6) with $0.1$7, $0.1$8, $0.1$9, and $300\,\mathrm{GeV}$0; 4FGL J1229.1+0202 (3C 273; $300\,\mathrm{GeV}$1) with $300\,\mathrm{GeV}$2, $300\,\mathrm{GeV}$3, $300\,\mathrm{GeV}$4, and $300\,\mathrm{GeV}$5; and 4FGL J1512.8−0906 (PKS 1510–089; $300\,\mathrm{GeV}$6) with $300\,\mathrm{GeV}$7, $300\,\mathrm{GeV}$8, $300\,\mathrm{GeV}$9, and $g$0.

3. The early Fermi-LAT daily flaring-source AGNFRC

A distinct Fermi-LAT AGNFRC was defined from the first 3.5 years of sky-survey observations, spanning MJD 54682 (2008 Aug 4) through MJD 55934 (2012 Feb 4), with a flare criterion based on daily-binned intervals satisfying $g$1 (Buson et al., 2013). This threshold matches the criterion used by the LAT Flare Advocacy, FA–GSW, to issue Astronomer’s Telegrams.

The sample contains 91 distinct AGNs that exhibited at least one daily-binned interval above threshold (Buson et al., 2013). Its class composition is 68 FSRQs, 14 BL Lacs, 4 AGN of uncertain type, 2 other AGN, 2 narrow-line Seyfert 1 galaxies, and 1 radio galaxy, NGC 1275. Among the BL Lacs, 6 are LSP, 6 ISP, and 2 HSP. Of the 91 sources, 83 have measured redshifts, 7 are not associated with any 2FGL counterpart, and 9 are detected at TeV energies.

The redshift distribution is broad for FSRQs, with $g$2 and a peak near $g$3, whereas BL Lacs are generally at $g$4 (Buson et al., 2013). Assuming a power-law spectrum, the apparent $g$5-ray luminosity is

$g$6

with typical peak luminosities in the range $g$7–$g$8. The flux-increase ratio $g$9 has mean $r$0, and several sources reached $r$1; FSRQs tend to have the largest $r$2 values, while BL Lacs show more modest brightening.

This catalogue isolates the most extreme daily-scale states rather than the broader variability continuum. Only $r$3 of 2FGL AGNs reached daily fluxes above the threshold, and typical flare durations are $r$4–$r$5 days, although complex multi-peak structures occur (Buson et al., 2013). During flares, FSRQs exhibit spectral hardening, with $r$6–$r$7, whereas BL Lacs show $r$8. The paper further notes preliminary cross-correlation studies showing lags of hours to days between $r$9-ray and lower-energy bands.

4. The ZTF DR23 AGNFRC-V1.0

He et al. (2025) define AGNFRC-V1.0 as the high-confidence subset of extreme brightness excursions uncovered through a two-stage search of the six-year ZTF DR23 light-curve database (He et al., 27 Jul 2025). Starting from approximately $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$0 AGNs with at least 30 high-quality $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$1- and $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$2-band observations, the initial AGN Flare Coarse Catalog, AGNFCC, contains 28,504 candidate flares, corresponding to 1.5% of the parent sample. After stringent filtering, the AGN Flare Refined Catalog retains 1,984 events, 0.1% of the parent sample and approximately 7% of AGNFCC, each with $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$3 in both bands.

Within this framework, an AGN flare is defined as a localized, significant flux enhancement that cannot be accounted for by a Gaussian-process-modeled stochastic baseline and is identified in both ZTF $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$4 and $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$5 light curves with high significance, again requiring $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$6 (He et al., 27 Jul 2025). The detection pipeline first applies Bayesian Blocks change-point segmentation following Scargle et al. 2013, then models the entire light curve with a Gaussian process using a Matern-1/2 covariance kernel implemented with celerite following Foreman-Mackey et al. 2017.

Refinement from AGNFCC to AGNFRC uses eight criteria, and a flare is retained only if all eight flags are false (He et al., 27 Jul 2025). The listed cuts are: max_gap > 500 d; min(peak_σ_g, peak_σ_r) < 3; more than one non-flare point above $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$7; $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$8 for the flare versus flat GP comparison; incomplete coverage of $$F(E>100\,\mathrm{MeV})>1\times10^{-6}\,\mathrm{ph\,cm^{-2}\,s^{-1}}$$9; $\sim 75{,}000$0; $\sim 75{,}000$1; and source confusion within $\sim 75{,}000$2.

The refined sample has well-defined statistical properties. Rise times $\sim 75{,}000$3 range from approximately 3 to 200 days, with median $\sim 75{,}000$4 days and $\sim 75{,}000$5 days; decay times $\sim 75{,}000$6 span approximately 5 to 300 days, with median $\sim 75{,}000$7 days and $\sim 75{,}000$8 days (He et al., 27 Jul 2025). The amplitude in magnitudes is

$\sim 75{,}000$9

with $>25{,}000$0 by selection; the median $>25{,}000$1 is approximately $>25{,}000$2 with $>25{,}000$3. Flare duration $>25{,}000$4 spans approximately 10 to 500 days, with median approximately 100 days.

Cross-matching to Million Quasars Catalog types yields an AGNFRC composition of approximately 58% QSOs, 30% generic AGN, 10% blazars, and 2% narrow-line AGNs (He et al., 27 Jul 2025). Blazars exhibit systematically shorter $>25{,}000$5 and larger $>25{,}000$6, consistent with jet-driven variability, while NLAGNs tend to have longer $>25{,}000$7 owing to reprocessing in obscured environments. The redshift distribution peaks at $>25{,}000$8 with a long tail to $>25{,}000$9, and compared to AGNFCC the refined catalog is more concentrated at low redshift, with median $0.1$0 versus $0.1$1.

The catalogue is also explicitly heterogeneous in physical origin. Cross-matching assigns likely associations with blazars, supernovae, tidal disruption events, BBH merger candidates, microlensing, and intrinsic AGN variability (He et al., 27 Jul 2025). Specifically, the paper reports 830 blazar matches to Roma-BZCAT and Milliquas blazar flags, 71 matches to TNS-reported supernovae, 2 TDE matches, 2 candidate electromagnetic counterparts to LVK O3 BBH triggers surviving the refined cuts, and a handful of microlensing-like cases for which $0.1$2. The remaining approximately 50% lack clear counterparts and may represent extreme red-noise or changing-look behavior. ALeRCE alert-stream classifications match approximately 70% of AGNFRC flares to “QSO,” “AGN,” “SN,” or “blazar.”

5. The CAZ-based optical AGNFRC blueprint

Kouch et al. (2025) present a self-contained AGNFRC blueprint built directly on the CAZ catalog and optical light curves of 7,918 blazar-selected AGN (Kouch et al., 18 Oct 2025). The input data consist of nightly-binned optical flux densities between 2007 and 2023 from CRTS, ATLAS, and ZTF, cleaned of outliers as described in their Section 3.2. The stated goal is a homogeneous optical flare catalog based on Bayesian blocks segmentation of merged light curves.

The flare definition is explicitly block-based. A flare is one contiguous valley–peak–valley sequence of Bayesian blocks, identified with the BBHOP procedure of Wagner et al. 2022 (Kouch et al., 18 Oct 2025). A valley block is one whose two immediate neighbors are both strictly higher in flux; a peak block is one whose two neighbors are both strictly lower in flux. The procedure then hops from a valley through ascending blocks to a single peak and descends to the next valley. Adjacent BBHOP flares are merged if the gap between them is shorter than 2 days or if the rise/decay significance at that boundary is less than $0.1$3.

Within each merged BBHOP flare, the blueprint records $0.1$4, $0.1$5, and $0.1$6 in MJD, and defines $0.1$7 and $0.1$8 (Kouch et al., 18 Oct 2025). The baseline flux is the median of all block fluxes in the lowest 50% of light-curve blocks; the peak flux is $0.1$9; and $\gamma$00. A time-integrated flare energy proxy is also tabulated:

$\gamma$01

Uncertainties on start, peak, and end times, rise and decay times, fluxes, and $\gamma$02 are propagated from nightly-bin errors through the block-averaging and derived-quantity formulas.

The column schema is extensive. Each flare row includes source_id, J2000 coordinates, redshift, $\gamma$03, radio variability Doppler factor $\gamma$04, $\gamma$05, $\gamma$06, $\gamma$07, $\gamma$08, $\gamma$09, $\gamma$10, $\gamma$11, $\gamma$12, the corresponding uncertainties, and an integer bit-mask of flags (Kouch et al., 18 Oct 2025). For the subset of 25,194 prominent flares, typical ranges are given as $\gamma$13–$\gamma$14 days, $\gamma$15–$\gamma$16 days, $\gamma$17–$\gamma$18, $\gamma$19–$\gamma$20 with median approximately $\gamma$21, $\gamma$22–$\gamma$23, and duty cycle per source of 0.01–0.9.

The population-level conclusions are organized around synchrotron peak frequency and Doppler factor (Kouch et al., 18 Oct 2025). Optical flares generally have a faster rise than decay. Optical brightness and variability are strongly dependent on synchrotron peak frequency. Flat spectrum radio quasars and BL Lac objects have comparable optical variability and flare characteristics at the same synchrotron peak frequency. Optical flare times tend to decrease and amplitudes increase with increasing radio variability Doppler factor. In the prominent-flare subset, the temporal-asymmetry statistic

$\gamma$24

is narrowly centered on $\gamma$25, indicating that flares rise slightly faster than they fall.

6. Statistical formalisms and catalog data models

Although the AGNFRC label spans several surveys, the underlying methodologies are unusually explicit about their statistics. The 12-year Fermi-LAT catalogue quantifies intrinsic variability with the normalized excess variance of Vaughan et al. 2003,

$\gamma$26

computed separately for 3-day, 7-day, and 30-day light curves, and defines the fractional variability amplitude as $\gamma$27 (Passos-Reis et al., 2 Feb 2026). Its catalog entries contain source name, position, redshift when available, SED class, mean and peak fluxes for all three cadences, average TS values, $\gamma$28 and $\gamma$29 for each cadence, variable and flaring flags, and quality flags.

The ZTF AGNFRC-V1.0 formalism combines Bayesian Blocks with GP significance tests (He et al., 27 Jul 2025). For a block spanning data points $\gamma$30, the constant-flux block fitness is

$\gamma$31

where $\gamma$32 is the weighted block mean. Candidate flare intervals are defined as contiguous regions where

$\gamma$33

The GP baseline uses the Matern-1/2 kernel

$\gamma$34

with covariance matrix $\gamma$35 and log marginal likelihood

$\gamma$36

For a flare interval, the deviation statistic is

$\gamma$37

and simulations from the best-fit GP define $\gamma$38 by the fraction of $\gamma$39. Criterion 4 in the refined selection uses $\gamma$40 and retains only flares with $\gamma$41.

The early Fermi-LAT flaring-source catalogue is centered on likelihood-based source detection and spectral fitting (Buson et al., 2013). Its test statistic is

$\gamma$42

with $\gamma$43 implying removal of the added source from the RoI model and $\gamma$44 implying a non-detection with a 95%-confidence upper limit. The variability index adopts the 2FGL-style definition

$\gamma$45

where $\gamma$46 for 23 degrees of freedom indicates less than 1% chance of steady emission. Flare spectra are fitted with the power law

$\gamma$47

The CAZ blueprint uses Bayesian Blocks in the Astropy v5 implementation and introduces a survey-calibrated change-point prior (Kouch et al., 18 Oct 2025). In the ideal-Gaussian limit, the prior is written as

$\gamma$48

but Kouch et al. calibrate it against 408 non-variable white-dwarf light curves and adopt upward adjustments of $\gamma$49 for CRTS and $\gamma$50 for ATLAS+ZTF segments. The total fitness is

$\gamma$51

The data products are similarly explicit. The ZTF AGNFRC-V1.0 is available as CSV and FITS tables at https://github.com/Lyle0831/AGN-Flares (He et al., 27 Jul 2025). The 12-year Fermi-LAT AGNFRC hosts FITS light curves for all 1,240 sources at https://fermi.gsfc.nasa.gov/ssc/data/access/lat/AGNFRC, provides Fermipy-based Python notebooks and outlier-filter code under MIT license at https://github.com/agn-pop-ctao/AGNFRC, and supplies additional machine-readable supplemental tables (Passos-Reis et al., 2 Feb 2026). The early Fermi-LAT AGNFRC is delivered as FITS and ASCII tables (Buson et al., 2013). The CAZ blueprint specifies AGNFRC_Sources.fits, AGNFRC_Flares.fits, and CSV counterparts, with TUNIT and TFORM headers, hosted at Zenodo and mirrored on VizieR (Kouch et al., 18 Oct 2025).

7. Scientific interpretation, misconceptions, and observational use

The scientific content of AGNFRC products is strongly modality-dependent. In $\gamma$52 rays, both Fermi-LAT catalogues emphasize the dominance of FSRQs in the most extreme variability regimes (Passos-Reis et al., 2 Feb 2026). In the 12-year catalogue, FSRQs have larger median normalized excess variance and broader dispersion than BL Lacs. In the early daily-threshold catalogue, FSRQs show the largest flux-increase ratios and measurable spectral hardening during flares, whereas BL Lacs show more modest brightening and little spectral-index change (Buson et al., 2013). This suggests that, within these catalog definitions, the strongest short-timescale $\gamma$53-ray activity is preferentially associated with the FSRQ population.

In the optical domain, the refined ZTF catalogue and the CAZ blueprint show that “flare” need not mean a single physical mechanism. The ZTF AGNFRC contains likely blazar flares, supernova impostors or true supernovae in AGN hosts, tidal disruption events, BBH-merger counterpart candidates, microlensing-like events, and a residual population consistent with extreme red-noise or changing-look behavior (He et al., 27 Jul 2025). By contrast, the CAZ blueprint is explicitly framed around blazar-selected AGN and ties flare statistics to synchrotron peak frequency and radio variability Doppler factor (Kouch et al., 18 Oct 2025). A common misconception is therefore that AGNFRC always denotes a pure intrinsic-AGN flare sample; the documented catalog usages do not support that simplification.

A second misconception is that AGNFRC always tabulates discrete flare episodes. The 12-year Fermi-LAT implementation is a source catalogue of multi-timescale variability, with flare flags and variability metrics attached to sources rather than one-row-per-outburst events (Passos-Reis et al., 2 Feb 2026). The early Fermi-LAT catalogue and the optical AGNFRCs are event-centered, but even there the operative definitions differ: a daily flux threshold in one case, GP-significant baseline deviations in another, and valley–peak–valley Bayesian-block structures in the CAZ blueprint (Buson et al., 2013).

The practical use cases are correspondingly varied. The 12-year Fermi-LAT catalogue is designed to optimize follow-up with CTAO, ASTRI Mini-Array, and SWGO, recommending short exposures of approximately 30 minutes on 3- to 7-day-timescale flare candidates, weekly monitoring of the top 50 AGNFRC sources for coordinated multi-wavelength campaigns, and scheduled deep exposures of 2–5 hours for sources with high $\gamma$54 (Passos-Reis et al., 2 Feb 2026). It further recommends a 3-day revisit cadence for FSRQs with $\gamma$55, a 7-day cadence for BL Lacs with intermediate variability, and exposures of 0.5–1 hour per pointing to achieve at least $\gamma$56 detections on flare peaks, assuming CTAO sensitivity.

The optical AGNFRC products support different workflows. He et al. emphasize population studies of rare, energetic events in AGN disks, while Kouch et al. explicitly recommend population studies of duty cycle versus SED class or redshift, multiwavelength cross-correlations using well-defined flare epochs, selection of flaring versus quiescent states for SED modeling or reverberation mapping campaigns, machine-learning classification of blazar variability classes, and searches for neutrino or cosmic-ray coincidences with optical flares on well-defined time windows (He et al., 27 Jul 2025). Taken together, these catalogues show that AGNFRC has become a recurring label for high-purity, strongly filtered AGN variability resources rather than a single standardized reference object.