betaFeedback

WorksPhysics and Astronomy

Extreme Multiwavelength Variability in the Recurrent Intermediate-type Changing-look Seyfert Galaxy SDSS J234728.73+242745.7

Abstract

We report the discovery of the Seyfert galaxy J234728.73+242745.7 as a recurrent intermediate-type CLAGN with extreme optical/UV variability. Optical spectra spanning nearly three decades reveal Type 1.0→\rightarrow1.5→\rightarrow1.2 transitions, accompanied by the disappearance and reappearance of broad He II λ4686\lambda4686 and a broad-line Balmer decrement evolving from 4.5 to 7.2 and then 2.4. Monitoring reveals coordinated variability from X-rays to the MIR. SED decomposition yields a host-corrected g-band high-to-low-state flux ratio of 5.4±0.45.4\pm0.4, exceeding the Swift/XRT ratio of 2.7±0.82.7\pm0.8 across the changing-look transition. This discrepancy may arise from reduced dust extinction, as correcting for the low-state continuum reddening (AVcont=0.66±0.23A_V^{\rm cont}=0.66\pm0.23 mag) brings the optical/UV ratios close to the X-ray value, although the correction is model-dependent and intrinsically weaker X-ray variability may also contribute. However, X-ray spectra show no compelling evidence for intrinsic gas absorption or its variability (NHintr≲5×1020N_{\rm H}^{\rm intr}\lesssim5\times10^{20} cm−2^{-2}). In contrast, EP/WXT monitoring during the recent high state reveals order-of-magnitude X-ray fluctuations on monthly-to-sub-year timescales, exceeding contemporaneous optical variations. Meanwhile, the radio flux density remains stable at ∼2\sim2 mJy across all ASKAP observations since mid-2020. Together with optical/UV and MIR variability and the broad He II transition, these results favor intrinsic accretion-state changes as the primary driver, while allowing a secondary contribution from variable obscuration. The source displays a loop-like MIR color–magnitude evolution, likely driven by the delayed response of an extended torus. This recurrent CLAGN offers a laboratory for probing the dynamic response of the broad-line region and circumnuclear dust to accretion-state transitions.

Corresponding author: Liming Dou; Jianguo Wang

[email protected]; [email protected]

keywords: Galaxies (573); Seyfert galaxies (1447); Accretion (14); Supermassive black holes (1663); High Energy astrophysics (739)

Introduction

Active galactic nuclei (AGN; C. K. Seyfert [87]; M. Schmidt [84]) are powered by accretion onto supermassive black holes (SMBHs). In the standard paradigm, an accretion disk produces the optical/ultraviolet (UV) continuum, a hot corona generates X-rays, and a circumnuclear dust torus reprocesses the central radiation into the infrared (IR). AGN are traditionally classified as Type 1 or Type 2 based on their optical spectra. Within the orientation-based unification model, Type 1 AGN show both broad and narrow emission lines from the broad-line region (BLR) and narrow-line region (NLR), respectively, whereas in Type 2 AGN, the BLR and nuclear continuum are obscured by the dust torus, leaving only narrow emission lines visible (R. Antonucci [4]).

Although all Type 1 AGN exhibit broad Hα\alpha and/or Hβ\beta emission lines, the visibility and relative strength of these broad components span a continuous range, giving rise to the spectral subclasses Types 1.0, 1.2, 1.5, 1.8, and 1.9. These intermediate subtypes were traditionally distinguished through visual inspection (D. E. Osterbrock [69]), and later quantified using empirical flux ratios involving broad Hβ\beta and [O III] λ\lambda5007 (hereafter Hβ\beta/[O III]; M. Whittle [105]; H. Winkler [107]; R. Maiolino & G. H. Rieke [63]). These subtypes are often interpreted as an obscuration sequence, representing progressively increasing levels of nuclear obscuration toward Type 2 AGN (R. Maiolino & G. H. Rieke [63]). More recent studies suggest that, at least among Type 1 subtypes, the observed spectral sequence can be explained by an accretion sequence, reflecting different accretion states rather than different levels of obscuration (L. Burtscher et al. [13]; A. H. T. Tan et al. [98]).

Contrary to this static picture, an increasing number of AGN have been found to undergo dramatic spectral transitions on human timescales. The hallmark of these changing-look AGN (CLAGN; S. M. LaMassa et al. [53]; C. L. MacLeod et al. [61, 62]; C. Ricci & B. Trakhtenbrot [77]) is the complete emergence or disappearance of broad Hα\alpha and/or Hβ\beta emission lines, accompanied by large continuum variations, driving transitions between Type 1.0–1.8 and Type 1.9/2. Such objects are commonly referred to as turn-on/off CLAGN.

Recently, the changing-look concept has been extended to transitions among intermediate subtypes (e.g., W. J. Hon et al. [46]; M. R. Senarath et al. [85]), even when broad emission lines never disappear completely, motivated by the growing number of sources showing subtype (or Hβ\beta/[O III]) variations on year-to-decade timescales (N. Amrutha et al. [3]).

Changing-look behavior has now been identified over a broad range of AGN luminosities and across diverse AGN subclasses, from nearby Seyfert galaxies to luminous quasars and LINERs (e.g., S. Frederick et al. [35]). Even blazars, whose optical spectra can exhibit a wide range of emission-line properties or featureless continua (G. Chen et al. [22]), have been observed to undergo dramatic spectral-state transitions (e.g., Ton 599; G. Chen et al. [20]). More broadly defined changing-state phenomena have also been reported beyond optical emission-line changes, such as the emergence of a radio jet in the narrow-line Seyfert 1 galaxy J1105+1452 (L. Dou et al. 2026; S. Komossa et al. [52]). These observations suggest that changing-state behavior is not restricted to a particular AGN subclass or luminosity regime, although the underlying physical mechanisms may differ among individual systems.

The physical origin of changing-look behavior remains an active topic of investigation. Growing observational evidence suggests that, for many optical CLAGN, the dramatic continuum and broad-line changes are primarily driven by intrinsic variations in the central engine rather than by variable line-of-sight (LoS) obscuration (e.g., S. M. LaMassa et al. [53]; C. L. MacLeod et al. [61]; Z. Sheng et al. [89]; C. Ricci & B. Trakhtenbrot [77]). Such variations may arise from substantial changes in the accretion rate and/or the structure of the inner accretion flow, including disk instabilities and transitions between different accretion states (e.g., H. Noda & C. Done [68]; Z. Sheng et al. [90]; J. Wang et al. [103]; Q. Dong et al. [28]; S.-L. Li & X. Cao 2026). These changes can substantially modify the ionizing continuum and consequently alter the ionization state of the BLR, leading to the weakening, disappearance, or reappearance of broad emission lines. Recent studies of recurrent and rapidly transitioning CLAGN further suggest that such changes in the inner accretion flow can occur on timescales shorter than the viscous timescale expected for a standard thin disk (e.g., J. Wang et al. [103]; S. Wang et al. [104]; Q. Dong et al. [28, 27]; S.-L. Li & X. Cao 2026; G. Chen et al. [21]). Other mechanisms may also contribute in individual sources. Transient accretion events, such as tidal disruption events (TDEs), can produce dramatic and often transient luminosity changes (e.g., A. Merloni et al. [66]; R. Li et al. 2022, 2024a,b; S. Son et al. [93]), while variable obscuration may account for a subset of changing-look events (e.g., R. Gilli et al. [38]; C. Ricci & B. Trakhtenbrot [77]). X-ray observations provide a particularly powerful diagnostic because they directly probe the innermost disk-corona emission while also constraining variable gas absorption along the LoS, thereby helping distinguish intrinsic accretion changes from obscuration-driven variability (e.g., Y. Ai et al. (2020); C. Ricci & B. Trakhtenbrot [77]).

Mid-infrared (MIR) observations provide an important complementary probe of AGN variability. The MIR emission arises primarily from hot dust heated by the accretion disk. Its variability directly reflects dust responses, producing a “dust echo” that encodes both the dust temperature evolution and the geometry of the obscuring structure. The Wide-field Infrared Survey Explorer (WISE; E. L. Wright et al. [108]) enabled decade-long, all-sky MIR monitoring, revealing stronger MIR variability in AGN than in inactive galaxies due to dust reprocessing. The MIR color criterion CMIR=W1−W2≳0.8C_{\mathrm{MIR}} = W1 - W2 \gtrsim0.8 efficiently separates AGN from stars and normal galaxies (R. J. Assef et al. [6]), where W1W1 and W2W2 denote the 3.4 and 4.6 µm WISE bands.

In J. Wu et al. (2026a) (hereafter Paper I) and J. Wu et al. (2026b) (hereafter Paper II), we systematically analyzed MIR color variations (MCVs) in broad-line AGN samples drawn from V. S. Paliya et al. [70]. We showed that a variable-temperature hot dust component diluted by invariant host starlight can reproduce different MCV trends; the host fraction determines whether the MCV is redder-when-brighter (RWB; host dominated), bluer-when-brighter (BWB; dominated by AGN-heated dust), or weak/no MCV (comparable contributions). This framework further predicts that RWB sources should exhibit bluer minimum MIR colors CMIR,minC_{\mathrm{MIR,min}} and larger MCV amplitudes ΔCMIR\Delta C_{\mathrm{MIR}} than BWB sources (see Figure 5 and Figure 6 of Papers I and II, respectively). Interestingly, previously known turn-on/off CLAGN (e.g., S. Frederick et al. [35]; B. Lyu et al. [60]) predominantly overlap with the RWB population in the CMIR,minC_{\mathrm{MIR,min}}–ΔCMIR\Delta C_{\mathrm{MIR}} plane, indicating that the changing-look phenomenon is characterized by distinctive MIR color evolution.

Figure 5. Multi-wavelength flux ratio R=⟨Fhigh⟩/⟨Flow⟩R = \langle F_{\mathrm{high}}\rangle/ \langle F_{\mathrm{low}}\rangle of J2347+2427, derived from WISE (W1 and W2), ZTF (gg, rr, and ii), Swift/UVOT (UVW1, UVM2, and UVW2), and Swift/XRT observations. Here, ⟨Flow⟩\langle F_{\mathrm{low}}\rangle and ⟨Fhigh⟩\langle F_{\mathrm{high}}\rangle represent the characteristic flux levels during SlowS_{\mathrm{low}} and ShighS_{\mathrm{high}}, respectively.

Figure 6. Enlarged view of the multi-wavelength LCs during ShighS_{\mathrm{high}}. The symbols are the same as those in Figure 1 and 3. Top: X-ray LCs in the 0.5–8 keV band, including Swift/XRT, EP/FXT, and EP/WXT data. Bottom: Optical grigri LCs from ZTF together with the W1-band data from WISE and SPHEREx.

SDSS J234728.73+242745.7 (hereafter J2347+2427), classified as a radio-quiet broad-line Seyfert 1 with a redshift of z=0.1571z = 0.1571 by V. S. Paliya et al. [70], attracted our attention during the analysis of Paper II because of its unusual MCV behavior. Although formally classified as a weak MCV source, its MIR color–magnitude diagram exhibits a prominent hysteresis loop rather than the reversible single-path evolution predicted by the host-dilution framework under the assumption of single-temperature, synchronous blackbody emission (see Figure 6 in Paper I), indicating a more complex physical process. In addition, it exhibits exceptionally large MIR variability, with amplitudes (σm\sigma_m[W1] = 0.45 mag) exceeding those of ∼99.7% of the Seyfert 1s with significant MIR variability in Paper II.

The source also displays rapid and strong optical variability: it brightened by ∼ 1 mag in the GG band within one year around 2020 December (MJD 59200.08), triggering a Gaia Science Alert (AT2020ac zq; S. T. Hodgkin et al. [45]). This alert prompted several months of Swift/XRT monitoring. Our subsequent spectroscopic monitoring reveals that the source is a recurrent intermediate-type CLAGN, exhibiting a Type 1.0→1.5→1.2 transition over the past three decades. J2347+2427 provides an excellent opportunity to investigate the physical mechanisms underlying the changing-look phenomenon, as well as the response of the accretion flow, BLR, and circumnuclear dust to dramatic changes in the central engine.

In this manuscript, we present a comprehensive multi-wavelength analysis of J2347+2427 spanning X-ray to MIR. We adopt a flat ΛCDM cosmology with H0=67.4H_0 = 67.4 km s−1^{-1} Mpc−1^{-1}, Ωm=0.315\Omega_m = 0.315, and ΩΛ=0.685\Omega_\Lambda= 0.685 (Planck Collaboration et al. [71]). For J2347+2427 at z=0.1571z = 0.1571, this corresponds to a luminosity distance of DL=778D_L = 778 Mpc.

Observations and Data Reduction

Optical and Infrared Photometric Monitoring

Optical light curves (LCs) were assembled from multiple surveys. We retrieved Gaia GG-band photometry from the Gaia Science Alerts archive (Gaia Collaboration et al. [36]), which provides multi-year monitoring covering the pre- and post-alert epochs. Early-time coverage is provided by the Catalina Sky Survey (CSS; A. J. Drake et al. [31]), which monitored the source in an unfiltered VV-band analogue before 2016 January. We restricted the CSS photometry to blend=0\mathrm{blend} = 0 to avoid contamination from neighboring sources. The Asteroid Terrestrial-impact Last Alert System (ATLAS; J. L. Tonry et al. [99]) provides cc- and oo-band forced photometry (L. Shingles et al. [91]); we selected measurements with χ2/N<5\chi^2/N < 5 and above the 5σ5\sigma detection limit. Together, the Gaia, CSS, and ATLAS data provide continuous optical coverage from 2005 July to 2026 February, and are presented in Figure 1a.

Figure 1. Multi-wavelength light curves of J2347+2427 across the optical/UV, MIR, and radio bands. The vertical black dashed line marks the Gaia Alert date. The green (620 days) and diagonally hatched blue (1300 days) shaded regions together mark the extended low-flux period (SlowS_{\mathrm{low}}): the optical emission becomes quiescent first in the green region, while the WISE flux continues to decline, with both settling into quiescence only in the blue region. The two gray shaded regions denote the 150-day SED windows for the declining (SdeclS_{\mathrm{decl}}) and rising (SriseS_{\mathrm{rise}}) phases (§ 4.1). Vertical red and cyan dashed lines indicate the spectroscopic observation epochs from the SDSS and Lijiang 2.4 m telescope, respectively.

High-cadence multi-band photometry was obtained from the Zwicky Transient Facility (ZTF; F. J. Masci et al. [64]) in the gri bands, retaining only high-quality detections (catflags = 0). We also retrieved grizy aperture photometry from Pan-STARRS1 [34], requiring airmass ≤1.4\le1.4 and a reliable aperture filling factor apFillFactor >0.8> 0.8. These data are shown in Figure 1b.

For each optical survey, outliers were removed by iteratively clipping points deviating by >3σ> 3\sigma from a 15-day sliding-window median. The cleaned LCs were binned into 5-day intervals to reduce scatter while preserving the intrinsic variability. All optical magnitudes were corrected for Galactic extinction using the dust maps of E. F. Schlafly & D. P. Finkbeiner (2011) and the Milky Way extinction curve of J. A. Cardelli et al. [16] with RVMW=3.1R_{V}^{\mathrm{MW}} = 3.1.

For the MIR, we used multi-epoch photometry from the AllWISE and NEOWISE missions [108], following the data reduction procedure of Paper II to construct W1 and W2 LCs spanning 2010–2024 at a typical six-month cadence. We also included data from the Spectro-Photometer for the History of the Universe, Epoch of Reionization and Ices Explorer (SPHEREx; launched 2025 March; J. J. Bock et al. 2026), which provides all-sky 0.75–5 μ\mum spectroscopy. From the SPHEREx spectra, we synthesized W1 and W2 fluxes and transformed them to the WISE photometric system. The MIR LCs are shown in Figure 1c.

The CSS, ATLAS, and Gaia LCs together span nearly two decades, revealing a broad transition from an early high state to a prolonged low-flux phase. After several years of quiescence, the source brightened rapidly around the Gaia alert (MJD 59200.08; black dashed line) and entered a renewed high state lasting ∼\sim4.5 yr (2021.8–2026.2; MJD 59427–61069). The MIR emission broadly follows the optical evolution, fading into the low state and brightening after the Gaia alert.

We defined a low state SlowS_{\mathrm{low}} and a recent high state ShighS_{\mathrm{high}}. ShighS_{\mathrm{high}} corresponds to the period from the Gaia alert to the present. SlowS_{\mathrm{low}} spans ∼5\sim5 yr, during which the optical emission stays at a low flux level. In Figure 1 we mark SlowS_{\mathrm{low}} with two shaded intervals: a green region of 620 days and a blue region of 1300 days with diagonal hatching. The blue hatched region indicates the stage where both the optical and MIR emission have fully entered the quiescent state. During the first ∼1.5\sim1.5 yr of SlowS_{\mathrm{low}} (i.e., green shading), the WISE fluxes are still declining significantly and continuously.

Archival SDSS and GALEX Photometry

We also searched the SDSS imaging archive and the Galaxy Evolution Explorer (GALEX) archive for complementary early-time photometry.1

The primary SDSS detection (ObjID 1237678579827474520) was obtained on 2008 October 1 (MJD 54740.394) and has clean=1. We adopt the modelMag photometry,

The primary SDSS detection (ObjID 1237678579827474520) was obtained on 2008 October 1 (MJD 54740.394) and has clean=1. We adopt the modelMag photometry, obtaining u=16.834±0.009u = 16.834 \pm0.009, g=16.744±0.004g = 16.744 \pm0.004, r=16.513±0.004r = 16.513 \pm0.004, i=16.037±0.004i = 16.037 \pm0.004, and z=16.157±0.008z = 16.157 \pm0.008 mag. An earlier secondary detection on 2004 September 17 (MJD 53265.235; ObjID 1237666244691886261) has clean=0; we therefore do not use it for quantitative variability measurements. GALEX detected a counterpart within 1.1′′1.1^{\prime\prime} of the optical position in the AIS field AIS_145_1_71, observed on 2006 October 26–29 (MJD 54034.416–54037.361), with a total exposure of 204 s in each UV band. The magnitudes are NUV=17.86±0.04NUV = 17.86 \pm0.04 and FUV=18.21±0.07FUV = 18.21 \pm0.07 mag, and both artifact flags are zero. The quoted values are AB magnitudes.

Although differences in filter response limit a direct comparison, the 2008 SDSS photometric measurements are broadly consistent with a relatively high optical flux state before SlowS_{\mathrm{low}}. Similarly, the 2006 GALEX measurements indicate a relatively high early UV flux level compared with the subsequent Swift/UVOT observations (see § 2.4.2).

Spectroscopic Observations

Archival spectra include a low-resolution objective-prism spectrum from the Hamburg Quasar Survey (HS 2344+2411; 1997 August), obtained with the 80 cm Hamburg Schmidt telescope at Calar Alto Observatory as part of a wide-field slitless spectroscopic survey (H.-J. Hagen et al. [41]), and an SDSS/eBOSS spectrum (2016 July, MJD 57580) obtained during the low state SlowS_{\mathrm{low}}.

We acquired two recent high-state spectra (2026 Jan. 10 and Feb. 7, hereafter LJ110 and LJ207) with the YFOSC spectrograph on the Lijiang 2.4 m telescope (LJT). Both cover 3858–7330 Å with a resolution of R≈280R \approx280 and were reduced using standard IRAF procedures.

The four spectra are shown in Figure 2a, with their corresponding epochs marked in Figure 1. All spectra were corrected for Galactic extinction and flux-calibrated by scaling the [O III] λ5007\lambda5007 to the SDSS value during the emission-line modeling described in § 3.1.

Figure 2. Panel (a): Optical spectra of J2347+2427: HS, SDSS, and the two LJT spectra, all scaled to the [O III] λ\lambda5007 core and vertically offset for clarity. The shaded regions mark the Hβ\beta (4500–5100 Å) and Hα\alpha (6400–6800 Å) regions. Panels (b)–(e): Zoom-in of the continuum-subtracted Hβ\beta region for each spectrum, shown in chronological order from left to right. The spectra include broad He II λ\lambda4686 (orange), broad Balmer lines (purple), and narrow lines (green). The [O III] wing in the SDSS spectrum is shown in cyan, while in the other spectra the [O III] wing and excess emission are jointly modeled with an additional broad Gaussian (dashed black). Panels (f)–(i): same as (b)–(e), but for the Hα\alpha region.

Swift Observatory

X-ray Telescope

J2347+2427 was observed multiple times by the Neil Gehrels Swift Observatory (N. Gehrels et al. [37]). We retrieved all archival X-ray Telescope (XRT; D. N. Burrows et al. [12]) observations with exposure times >150> 150 s, yielding 18 usable epochs. Three observations were obtained during SlowS_{\mathrm{low}}, while the remaining fifteen were taken after the Gaia alert, including a 2.7 ks observation that we requested in 2026 January.

servation that we requested in 2026 January. The individual exposure times range from 0.157 to 2.7 ks. All XRT observations were performed in photon counting mode (J. E. Hill et al. [44]). The data were reduced using HEASOFT (v6.35) and the xrtpipeline task. Source events were extracted from a circular aperture with a radius of 50″ centered on the optical position, while the background was estimated from a nearby source-free region.

UV–Optical Telescope

During the XRT observations, the Swift UV–Optical Telescope (UVOT; P. W. A. Roming et al. [79]) obtained simultaneous photometry in the optical and UV. Observations were taken in six filters: VV (5468 Å), BB (4392 Å), UU (3465 Å), UVW1UVW1 (2600 Å), UVM2UVM2 (2246 Å), and UVW2UVW2 (1928 Å). The full set of six filters was used only for our requested 2026 Jan. observation; all other epochs provided a subset.

Source counts were extracted within a 10″ radius aperture centered on the target, with the background estimated from a nearby 20″ source-free region. After background subtraction, AB magnitudes were derived with uvotsource using the most recent calibrations ([72, 9]) and corrected for Galactic extinction. Only measurements with detection significance >3σ> 3\sigma were retained. The UVOT LCs are shown in Figure 1d and are consistent with the optical variability.

Einstein Probe

Wide-field X-ray Telescope

The Wide-field X-ray Telescope (WXT) on board the Einstein Probe (EP; W. Y. Yuan et al. [117]) is a wide-field X-ray monitor employing novel lobster-eye micro-pore optics, providing a field of view (FOV) of ∼3600\sim3600 deg2^{2}. Operating in the 0.5–4 keV band, WXT reaches a sensitivity of (2(2–3)×10−113) \times10^{-11} erg cm−2^{-2} s−1^{-1} for a typical 1 ks exposure. The energy resolution is ∼170\sim170 eV at 1 keV, and the angular resolution is ∼5′\sim5^{\prime} (FWHM).

The position of J2347+2427 was covered repeatedly by EP/WXT between 2024 July and 2026 July. Photon events were extracted from a circular aperture with a radius of 3′3^{\prime}. Because the source is too faint to be significantly detected in individual WXT snapshots, the observations were grouped into 14-day bins to improve the signal-to-noise ratio.

To assess whether the apparent WXT variability could be contaminated by nearby X-ray sources, we visually inspected the most recent 2.7 ks Swift/XRT image, which provides substantially higher angular resolution than WXT. No significant X-ray source other than

J2347+2427 is visible within the WXT aperture. We further cross-matched the region with the Living Swift XRT Point Source (LSXPS; P. A. Evans et al. [32]) catalog and found no additional X-ray sources within the aperture.

Follow-up X-ray Telescope

The Follow-up X-ray Telescope (FXT; Y. Chen et al. [23]) on board EP is a Wolter-I telescope operating in the 0.3–10 keV band. EP/FXT consists of two identical modules, FXT-A and FXT-B. FXT observed J2347+2427 simultaneously with both modules on 2026 July 5 in the Full Frame (FF) mode, with an effective exposure of 2.4 ks for each module. The data were reduced using the Follow-up X-ray Telescope Data Analysis Software (FXTDAS2, v1.30). Source events were extracted from a circular aperture with a radius of 100″ centered on the target, while the background was estimated from a nearby source-free circular region.

ROSAT Archival Data

J2347+2427 was detected in the ROSAT All-Sky Survey (RASS; W. Voges et al. [102]) from sequence rs931361n00. We retrieved the archival Position Sensitive Proportional Counter (PSPC) event data and extracted source photons from a circular aperture of radius 510\farcs6 centered on the source position. The background was estimated from a nearby source-free region. Inspection of the archival event data confirmed that the source was observed on 1990 July 11–13 and December 20–21. Because the ONTIME keyword records the total accepted time rather than the local exposure in RASS scanning-mode data, we adopted an exposure time of 694 s from the exposure map.

ASKAP Radio Observations

We collected radio observations of J2347+2427 from the Australian SKA Pathfinder (ASKAP; A. W. Hotan et al. [47]) telescope. The source is detected as an unresolved radio source in multiple ASKAP observations from the Rapid ASKAP Continuum Survey (RACS; D. McConnell et al. [65]; C. L. Hale et al. [42]) and the First Large Absorption Survey in H i (FLASH; J. R. Allison et al. [2]; H. Yoon et al. [116]), spanning MJD 58972–61050. The radio continuum images show a compact source at the position of J2347+2427, with no obvious extended emission. We measure the peak flux density, SνpeakS_{\nu}^{\mathrm{peak}}, at the source position from each ASKAP image, yielding values of approximately 1.3–2.4 mJy across the central frequencies of 0.85–1.65 GHz (see Figure 1e for the ASKAP LC). These ASKAP images are averaged across the full 15 min or 2 hr observation, for RACS and FLASH respectively, and a bandwidth of 288 MHz.

Analysis and Results

Optical Spectral Analysis

Continuum and Emission-line Fitting

We use the SDSS and LJT spectra to study the spectral evolution of J2347+2427 before and after the 2020 outburst, while the HS spectrum provides an additional view of its earlier state. We first perform a spectral decomposition of the highest-resolution SDSS spectrum and then use the resulting narrow lines to constrain other spectra.

We fit the SDSS spectrum over the rest-frame wavelength range of 3500–8000 Å using the Bayesian AGN Decomposition Analysis for SDSS Spectra (BADASS3; R. O. Sexton et al. [86]). The continuum is modeled as the sum of host starlight and AGN emission. The host component is represented by four single stellar population templates with ages of 0.1, 1, 5, and 10 Gyr from the EMILES library (A. Vazdekis et al. [100]), while the AGN component consists of a smoothly broken power law plus the optical Fe II template of M.-P. Véron-Cetty et al. [101].

For the emission lines, we model the Hβ\beta (4500–5100 Å) and Hα\alpha (6400–6800 Å) regions. The Hβ\beta region includes narrow He II λ4686\lambda4686, [O III] λλ4959,5007\lambda\lambda4959,5007, and Hβ\beta, together with broad Hβ\beta and He II λ4686\lambda4686. We also consider a broad [O III] wing component. The Hα\alpha region includes narrow Hα\alpha, [N II] λλ6548,6583\lambda\lambda6548,6583, and [S II] λλ6716,6731\lambda\lambda6716,6731, together with broad Hα\alpha. Each component is modeled with a Gaussian profile. In the Hβ\beta region, the widths and velocity offsets of all narrow emission lines are tied to those of the [O III] λ5007\lambda5007 core, while those in the Hα\alpha region are tied to [S II] λ6731\lambda6731. The flux ratios of the [O III] and [N II] doublets are fixed to 3:1.

For the HS spectrum, we follow the same fitting procedure as for the SDSS spectrum, except that the relative fluxes of the narrow emission lines are fixed to the SDSS values. The initial fitting reveals significant excess emission on the red side of Hβ\beta, which cannot be adequately reproduced by the [O III] wing alone. Because of the lower spectral resolution, however, the excess emission and the [O III] wing cannot be reliably separated. We therefore use a single broad Gaussian to jointly account for the [O III] wing and the red-side excess. As the relative narrow-line fluxes are fixed, the [O III] core fluxes are well constrained. A similar component is also included on the red side of Hα\alpha to improve the fit.

For the LJT spectra, we further simplify the fitting procedure by restricting the fitting range to 4200–7000 Å and modeling the continuum with a smoothly broken power law and the optical Fe II emission, while neglecting the host contribution because of the dominant blue continuum. The relative fluxes of the narrow emission lines are again fixed to the SDSS values. Similar excess emission is seen on the red sides of both Hβ\beta and Hα\alpha, as in the HS spectrum, and the Hβ\beta excess cannot be fully reproduced by the [O III] wing alone. We therefore adopt the same emission-line treatment as for the HS spectrum.

All spectra were flux-calibrated by scaling the modeled [O III] λ5007\lambda5007 core flux to the SDSS value. The fitting results, including the Hβ\beta and Hα\alpha decompositions, are shown in Figure 2, with the broad-line parameters and λFλ5100\lambda F_{\lambda5100} listed in Table 1. As described above, the narrow-line flux ratios are fixed to the SDSS values across all epochs, with (Hα/Hβ)n=5.2(\mathrm{H}\alpha/\mathrm{H}\beta)_{\mathrm{n}} = 5.2. The additional broad component in the HS and LJT spectra has FWHM values of ≳6000\gtrsim6000–10,000 km s−1^{-1} and fluxes of (0.4(0.4–1.1)×10−131.1) \times10^{-13} erg s−1^{-1} cm−2^{-2} in Hα\alpha and (0.3(0.3–0.6)×10−130.6) \times10^{-13} erg s−1^{-1} cm−2^{-2} in Hβ\beta. Broad He II λ4686\lambda4686 is detected in the HS and LJT spectra but not in the SDSS spectrum. Given the absence of prominent Fe II features around 4500–4700 Å and 5100–5400 Å, the broad He II emission is unlikely to arise from Fe II contamination.

epochFluxFWHMλF5100\lambda F_{5100}log⁡MBH\log M_{\mathrm{BH}}log⁡Lbol\log L_{\mathrm{bol}}log⁡λEdd\log\lambda_{\mathrm{Edd}}
(1)Hα\alpha (2)Hβ\beta (3)He II (4)Hα\alpha (5)Hβ\beta (6)He II (7)(8)(9)(10)(11)
HS5.08±0.205.08\pm0.201.14±0.071.14\pm0.070.27±0.040.27\pm0.043.50±0.123.50\pm0.124.00±0.514.00\pm0.516.24±0.596.24\pm0.5935.65±8.3135.65\pm8.318.4945.31-1.29
SDSS1.15±0.011.15\pm0.010.16±0.010.16\pm0.01<0.02<0.025.40±0.035.40\pm0.035.61±0.035.61\pm0.03—6.37±0.146.37\pm0.148.6144.62-2.09
LJ1101.09±0.041.09\pm0.040.45±0.030.45\pm0.030.28±0.030.28\pm0.033.77±0.073.77\pm0.074.43±0.134.43\pm0.1312.33±0.6012.33\pm0.6052.10±4.8652.10\pm4.868.2845.46-0.92
LJ2071.00±0.061.00\pm0.060.44±0.070.44\pm0.070.18±0.030.18\pm0.033.77±0.163.77\pm0.164.10±0.174.10\pm0.1712.00±0.6012.00\pm0.6052.42±5.0152.42\pm5.018.2645.46-0.90

Table 1. Broad line and continuum properties of optical spectra. Note—Columns: (1) observation epoch; (2)–(4) broad line fluxes for Hα\alpha, Hβ\beta, and He II, respectively, in 10−1310^{-13} erg s−1^{-1} cm−2^{-2}; (5)–(7) corresponding FWHMs in 10310^{3} km s−1^{-1}; (8) continuum λFλ\lambda F_{\lambda} at 5100 Å in 10−1310^{-13} erg s−1^{-1} cm−2^{-2}; (9) black hole mass in units of M⊙M_{\odot}; (10) bolometric luminosity in erg s−1^{-1}; (11) Eddington ratio.

Seyfert Type Transition and Spectral Evolution

According to the classification scheme of R. Maiolino & G. H. Rieke [63], the measured Hβ\beta/[O III] (≃4.1\simeq4.1, 0.7, and 1.6) classify the HS, SDSS, and LJT spectra as Seyfert Types 1.0, 1.5, and 1.2, respectively. The observed subtype transition occurs on a decadal timescale, similar to the Hβ\beta/[O III] variability reported by N. Amrutha et al. [3], confirming J2347+2427 as a recurrent intermediate-type CLAGN.

Across all epochs, the dominant behavior is a strong BWB continuum trend. During the brightest state, the two LJT spectra exhibit consistent luminosity, continuum shape, and broad-line properties over a one-month interval, with the rest-frame 4200 Å power-law flux density nearly 15×15 \times that of SDSS. The earlier HS spectrum exhibits an intermediate continuum level, higher than that of SDSS but lower than those of LJT.

Broad Balmer lines are present in all spectra. The broad Hα\alpha flux drops from 5.08 to 1.15×10−131.15 \times10^{-13} erg s−1^{-1} cm−2^{-2} between HS and SDSS but is comparable between the SDSS and LJT epochs (1.09×10−131.09 \times10^{-13} erg s−1^{-1} cm−2^{-2}), despite the much brighter continuum. In contrast, broad Hβ\beta and Hγ\gamma increase significantly from SDSS to LJT. Consequently, the broad-line Balmer decrement (Hα\alpha/Hβ\beta) evolves from 4.5 (HS) to 7.2 (SDSS) and then drops sharply to 2.4 (LJT), anticorrelated with the AGN luminosity.

Broad He II λ4686\lambda4686 also shows dramatic behavior (orange lines; Figure 2b–Figure 2e): it is detected with ≳7σ\gtrsim7\sigma significance in high states (HS and LJT), but is undetected (<1σ< 1\sigma) in the SDSS spectrum, indicating changing-look behavior in the high-ionization emission.

Black Hole Parameters

The bolometric luminosity LbolL_{\mathrm{bol}} was derived from the 5100 Å continuum using the quasar correction of J. C. Runnoe et al. [82],

log⁡(Lbolerg s−1)=4.89+0.91log⁡(λL5100erg s−1),(1)\log\left(\frac{L_{\mathrm{bol}}}{\mathrm{erg\ s}^{-1}}\right) = 4.89 + 0.91\log\left(\frac{\lambda L_{5100}}{\mathrm{erg\ s}^{-1}}\right), \tag*{(1)}

and the virial black hole mass MBHM_{\mathrm{BH}} from the broad Hα\alpha line using Y. Shen et al. [88],

log⁡(MBHM⊙)=0.379+0.43log⁡(LHα1042 erg s−1)+2.1log⁡(FWHMHαkm s−1).(2)\log\left(\frac{M_{\mathrm{BH}}}{M_{\odot}}\right) = 0.379 + 0.43\log\left(\frac{L_{\mathrm{H}\alpha}}{10^{42}\ \mathrm{erg\ s}^{-1}}\right) + 2.1\log\left(\frac{\mathrm{FWHM}_{\mathrm{H}\alpha}}{\mathrm{km\ s}^{-1}}\right). \tag*{(2)}

Table 1 lists the derived parameters. log⁡MBH\log M_{\mathrm{BH}} remains consistent across epochs (≈8.5\approx8.5), while log⁡Lbol\log L_{\mathrm{bol}} ranges from 44.62 to 45.46 and log⁡λEdd\log\lambda_{\mathrm{Edd}} from −2.09-2.09 to −0.90-0.90. Both decline from HS to SDSS and then rise to their highest values in the LJT epoch.

X-ray Spectral Analysis

X-ray Flux and Spectral Measurements

We analyzed the X-ray spectra (ROSAT, Swift/XRT, and EP/FXT) of J2347+2427 for each observation. Spectral fitting was performed in XSPEC (v12.15.0; K. A. Arnaud [5]) using C-statistics (J. S. Kaastra [50]). All spectra were grouped to a minimum of two counts per bin, and uncertainties are reported at the 90% confidence level (ΔC=2.706\Delta C = 2.706). Each spectrum was modeled with a simple absorbed power law (phabs×\timeszpowerlw in XSPEC), with the absorption column density fixed to the Galactic value (NH=3.32×1020N_{\mathrm{H}} = 3.32 \times10^{20} cm−2^{-2}; HI4PI Collaboration et al. [43]). The 0.5–8.0 keV fluxes F0.5−8 keVF_{0.5-8\ \mathrm{keV}} were derived using the cflux convolution model. As FXT-A and -B observed our target simultaneously, their spectra were fitted jointly, allowing only a cross-normalization constant between the two modules.

This model provided acceptable fits (C/ν=0.37C/\nu= 0.37–1.39) to all XRT spectra except the shortest exposure (0.157 ks, taken during SlowS_{\mathrm{low}}). Owing to low S/N, this exposure yielded a poorly constrained fit (C/ν=9.4/4C/\nu= 9.4/4) with Γ=3.2−1.5+2.0\Gamma= 3.2^{+2.0}_{-1.5} and F0.5−8 keV=1.4−0.8+2.1×10−12F_{0.5-8\ \mathrm{keV}} = 1.4^{+2.1}_{-0.8} \times10^{-12} erg s−1^{-1} cm−2^{-2}. For the well-fitted epochs, F0.5−8 keVF_{0.5-8\ \mathrm{keV}} ranges from 2.1×10−122.1 \times10^{-12} to 8.3×10−128.3 \times10^{-12} erg s−1^{-1} cm−2^{-2} and the photon indices span Γ=1.2±0.6\Gamma= 1.2 \pm0.6 to 2.2±0.42.2 \pm0.4. The ROSAT and FXT spectra were likewise well described by the same model. The ROSAT spectrum yields F0.5−8 keV=7.93−1.24+1.47×10−12F_{0.5-8\ \mathrm{keV}} = 7.93^{+1.47}_{-1.24} \times10^{-12} erg cm−2^{-2} s−1^{-1} and Γ=1.38−0.28+0.25\Gamma= 1.38^{+0.25}_{-0.28} (C/ν=83.86/103C/\nu= 83.86/103), while the FXT spectrum gives F0.5−8 keV=5.47−0.40+0.43×10−12F_{0.5-8\ \mathrm{keV}} = 5.47^{+0.43}_{-0.40} \times10^{-12} erg cm−2^{-2} s−1^{-1} and Γ=1.96±0.08\Gamma= 1.96 \pm0.08 (C/ν=437.72/433C/\nu= 437.72/433).

To improve S/N, we stacked three low-state XRT spectra and all 15 high-state XRT spectra, respectively, using the addspec tool. The stacked low-state spectrum yielded a best-fit Γ=1.50±0.31\Gamma= 1.50 \pm0.31 and F0.5−8 keV=2.05−0.49+0.63×10−12F_{0.5-8\ \mathrm{keV}} = 2.05^{+0.63}_{-0.49} \times10^{-12} erg s−1^{-1} cm−2^{-2} (C/ν=22.96/30C/\nu= 22.96/30). The stacked high-state spectrum was softer, with Γ=1.71±0.05\Gamma= 1.71 \pm0.05 and F0.5−8 keV=(5.58±0.24)×10−12F_{0.5-8\ \mathrm{keV}} = (5.58 \pm0.24) \times10^{-12} erg s−1^{-1} cm−2^{-2} (C/ν=370.07/390C/\nu= 370.07/390). The resulting X-ray LC is shown in Figure 3, including the stacked results. The X-ray flux follows the optical evolution, with the post-alert flux significantly exceeding that during SlowS_{\mathrm{low}}. The ROSAT flux is comparable to the recent bright XRT and FXT measurements, recording an earlier bright X-ray phase in 1990, consistent with a high optical flux state before SlowS_{\mathrm{low}}.

Long-term X-ray monitoring of J2347+2427

Figure 3. Long-term X-ray monitoring of J2347+2427 from 1990 to 2026, combining observations from ROSAT (grey point), Swift/XRT (black points), EP/FXT (orange point), and EP/WXT (blue symbols). The red point and black horizontal line show the stacked low- and high-state XRT results. All error bars correspond to the 90% confidence level. Top: Intrinsic 0.5–8.0 keV flux evolution. WXT bins (bin size = 14 days) are shown as blue filled points (S/N > 5), open circles (2 < S/N ≤ 5), and downward triangles (90% upper limits for S/N ≤ 2). The shaded region indicates the SlowS_{\mathrm{low}} interval, as in Figure 1. Bottom: Evolution of the X-ray photon index (Γ\Gamma).

The EP/WXT all-sky survey substantially fills the temporal gaps over the past two years. We converted the WXT count rates into approximate 0.5–4.0 keV fluxes using an empirical energy conversion factor of 2×10−92 \times10^{-9} erg cm−2^{-2} cts−1^{-1}, adopted solely for order-of-magnitude characterization under the current instrumental calibration (H. Cheng et al. [24, 25]). Assuming a power-law spectrum with Galactic absorption and a photon index fixed at the stacked high-state XRT value (Γ=1.71\Gamma= 1.71), we estimated the intrinsic 0.5–8.0 keV fluxes and include them as blue symbols in Figure 3 (with S/N ≤ 2 bins shown as 90% upper limits). The estimated fluxes agree well with the nearly contemporaneous FXT observation and are broadly consistent with the XRT measurements, although they are systematically lower than the high-state XRT level. Given the approximate nature of the conversion, we do not investigate this offset further. We also note that, the first WXT bin, despite its low S/N (< 1), has a substantial exposure of ∼86\sim86 ks and yields F0.5−8 keV≲4×10−13F_{0.5-8\ \mathrm{keV}} \lesssim4 \times10^{-13} erg cm−2^{-2} s−1^{-1}, indicating that the source was indeed in a relatively faint X-ray state during ShighS_{\mathrm{high}}.

Spectral Variability and Absorption Constraints

The two stacked Swift/XRT spectra suggest a mild softer-when-brighter (SWB) trend, although their Γ\Gamma uncertainties overlap, with recent EP/FXT observation providing further support for this trend (Figure 4). The ROSAT best-fit spectrum is harder than the stacked high-state XRT spectrum at a comparable flux. However, the large ROSAT uncertainty in Γ\Gamma and the extrapolation of its flux beyond the PSPC bandpass (0.1–2.4 keV) limit this comparison.

Figure 4. Photon index Γ\Gamma vs. 0.5–8 keV X-ray flux, with 90% error bars. The stacked low- and high-state Swift/XRT spectra exhibit a tentative SWB trend, further reinforced by the recent EP/FXT observation.

Since the SWB trend can be associated with changes in LoS absorption, we added an intrinsic absorption component (zphabs) to the Galactic-absorbed power-law model and fitted FXT, ROSAT, and the two stacked XRT spectra independently. All four spectra yielded only 90% upper limits on the intrinsic absorption column, with NHintr<5.10×1020N_{\mathrm{H}}^{\mathrm{intr}} < 5.10 \times10^{20} cm−2^{-2} (FXT), NHintr<5.12×1020N_{\mathrm{H}}^{\mathrm{intr}} < 5.12 \times10^{20} cm−2^{-2} (ROSAT), 5.06×10205.06 \times10^{20} cm−2^{-2} (low-state XRT), and 4.38×10204.38 \times10^{20} cm−2^{-2} (high-state XRT).

As a complementary test, we jointly fitted the two stacked XRT spectra to examine whether variable absorption alone could account for the observed X-ray variability, allowing NHintrN_{\mathrm{H}}^{\mathrm{intr}} to vary independently among epochs. To avoid the high-state spectrum dominating the fit, we rebinned these spectra to comparable numbers of bins (22–25 each). We first tied both Γ\Gamma and the power-law flux across epochs, obtaining a poor fit of C/ν=108.69/43C/\nu= 108.69/43. We then allowed the power-law flux to vary independently while keeping Γ\Gamma tied, which significantly improved the fit to C/ν=44.01/42C/\nu= 44.01/42, with Γ=1.74−0.07+0.09\Gamma= 1.74^{+0.09}_{-0.07}.

Overall, the X-ray spectra provide no significant evidence for intrinsic absorption, although a weak absorber (<1021 cm−2< 10^{21}\ \mathrm{cm}^{-2}) cannot be excluded.

Multi-wavelength Variability

Overview of the Multi-wavelength Variability

The multi-wavelength LCs of J2347+2427 reveal dramatic variability from X-rays to the MIR (Figure 1 and Figure 3). In the optical, the source evolved from an early bright state into a prolonged faint phase, remained quiescent for ∼5\sim5 yr (SlowS_{\mathrm{low}}), and then underwent a rapid rebrightening following the Gaia alert. The optical rise associated with the 2020 outburst lasted ∼1.7\sim1.7 yr (2019.11–2021.8), during which the source brightened by ∼1.7\sim1.7 mag in the gg band. Despite returning to a bright phase, the optical emission continued to vary significantly over the subsequent several years rather than settling into a steady phase. Noticeable variability is also present during SlowS_{\mathrm{low}} in both the optical and MIR, indicating that the central engine remains active during its faintest phase. Over the nearly two decades, the source exhibits a peak-to-peak variability amplitude of Δg=2.29±0.03\Delta g = 2.29 \pm0.03 mag (a factor of ∼8.2\sim8.2 in flux). Beyond the overall flux changes, the ZTF grigri LCs display a clear BWB trend (i.e., Δg>Δr>Δi\Delta g > \Delta r > \Delta i; Figure 1b), with several short-duration optical flares superimposed on the long-term evolution. These behaviors are widely observed in quasar variability (e.g., [80, 39]).

The UV, MIR, and X-ray emissions all broadly track the long-term optical evolution. The UV emission exhibits an even larger peak-to-peak amplitude of ΔUVM2=2.87±0.13\Delta UV M2 = 2.87 \pm0.13 mag (a factor of ∼14.1\sim14.1 in flux). The extreme and highly correlated optical/UV variability establishes J2347+2427 as an extreme variability Seyfert galaxy, analogous to the extreme variability quasars (EVQs; [81]). The Swift/XRT flux increased by ∼3×\sim3\times between SlowS_{\mathrm{low}} and ShighS_{\mathrm{high}}, while the ROSAT flux measured in 1990 was comparable to the recent bright X-ray measurements. The MIR LC exhibits clear delays during both the declining and rising phases and lacks the short-timescale structures seen in the optical, as expected from dust reprocessing in an extended torus.

In contrast to the strong variability observed at shorter wavelengths, the radio emission remains relatively stable, with a flux density of ∼2\sim2 mJy throughout the ASKAP monitoring period from 2020 May to 2026 January. A χ2\chi^2 test against a non-variable hypothesis yields χ2=10.45\chi^2 = 10.45 for 11 degrees of freedom (p=0.49p = 0.49), suggesting that the observed radio fluctuations are consistent with measurement uncertainties. Notably, these observations cover both sides of the 2020 outburst, yet no corresponding change in the radio flux density is observed.

Flux Changes Across the 2020 Outburst

The 2020 outburst marks a clear turning point in the long-term flux variability of J2347+2427. To compare the multi-wavelength fluxes between the pre- and post-outburst epochs, we define the flux ratio R≡⟨Fhigh⟩/⟨Flow⟩R \equiv\langle F_{\mathrm{high}}\rangle/ \langle F_{\mathrm{low}}\rangle, where ⟨Flow⟩\langle F_{\mathrm{low}}\rangle and ⟨Fhigh⟩\langle F_{\mathrm{high}}\rangle are the median fluxes during SlowS_{\mathrm{low}} and ShighS_{\mathrm{high}}, respectively. Uncertainties are estimated from Monte Carlo simulations that account for photometric errors and finite sampling. Figure 5 shows that RR increases systematically from the MIR to the UV. For the X-ray regime, using the 0.5–8 keV flux ratio from the stacked Swift/XRT spectra (§ 3.2), we obtain RX=2.7±0.8R_{\mathrm{X}} = 2.7 \pm0.8, comparable to or slightly below the gg-band value Rg=4.1±0.4R_g = 4.1 \pm0.4 and significantly smaller than the UV amplitudes.

Variability in the High-state

As noted above, the optical emission remained highly variable throughout ShighS_{\mathrm{high}}. The gg band exhibited a peak-to-peak variability of ∼1.3\sim1.3 mag (a factor of ∼3.3\sim3.3 in flux) during ShighS_{\mathrm{high}}. Figure 6 zooms in on ShighS_{\mathrm{high}}, showing the contemporaneous X-ray, optical, and MIR LCs. A pronounced local minimum occurred around 2024 September, after which the optical flux showed an overall rising trend, despite several small month-scale fluctuations. The MIR LC follows the same overall evolution as the optical emission, but with smoother variations.

The EP/WXT monitoring covers the period during which the optical emission gradually recovered from its local minimum. Notably, the first WXT bin with a faint upper limit was obtained near the optical local minimum. Although the X-ray sampling is much sparser than that of the optical monitoring, the available data suggest that the X-ray evolution broadly follows the optical trend even during ShighS_{\mathrm{high}}. Unlike the optical LC, the WXT LC exhibits much stronger fluctuations on monthly to sub-year timescales, with the inferred X-ray flux varying by a factor of ≳18\gtrsim18 within the WXT data alone, based on the ratio between the upper limit of the first WXT bin and the maximum flux among the high-significance detections (S/N > 5). This inferred variability is subject to uncertainties in the flux conversion, including those associated with the current WXT effective-area calibration (systematic uncertainty ≲10%\lesssim10\%; [25]) and the assumed fixed Γ\Gamma. However, the relatively stable spectral shape in ShighS_{\mathrm{high}} (Figure 3), together with the limited calibration uncertainty, suggests that these effects are unlikely to erase the large-amplitude variability, which remains close to an order of magnitude.

Color Variations

Bluer-When-Brighter Trend in the Optical

As noted in § 3.1 and § 3.3, J2347+2427 exhibits a BWB trend in the optical. To examine the color behavior in more detail, we rebinned the ZTF gg and rr photometry onto a common time grid: each gg-band epoch was paired with the closest rr-band observation within 10 days; unpaired points were discarded. The resulting g−rg-r vs. gg color–magnitude diagram (CMD) is shown in Figure 7 (left). The ZTF CMD confirms the BWB trend again. At magnitudes brighter than g=17.6g = 17.6 mag, g−rg-r follows a tight linear relation with Pearson r=0.93r = 0.93 (p=1.5×10−17p = 1.5 \times10^{-17}) and best-fit g−r=0.24g−3.87g-r = 0.24g - 3.87 (i.e., Δg/Δr=1.32\Delta g/\Delta r = 1.32). For g>17.6g > 17.6 mag, the colors lie systematically above (i.e., redder than) this linear relation, with a maximum excess of Δ(g−r)=0.22±0.03\Delta(g-r) = 0.22 \pm0.03 mag.

Observed optical and MIR color–magnitude diagrams

Figure 7. Observed optical and MIR color–magnitude diagrams. Left: ZTF g−rg-r vs. gg. The reddest points at faint magnitudes (red) lie systematically above the linear relation defined by the brighter data (black). Right: WISE W1−W2W1-W2 vs. W1W1. The dashed line traces the temporal evolution of the source (arrows indicate the direction of evolution), and the blue cross marks the SPHEREx measurement. The source is formally classified as a weak MCV object (ρMIR=0.47\rho_{\mathrm{MIR}} = 0.47), while exhibiting a prominent MIR color loop.

Long-term Mid-infrared Color Variation

We further investigated the long-term MCV of J2347+2427 using the MIR LCs. The Pearson correlation coefficient between MIR color CMIRC_{\mathrm{MIR}} and W1 magnitude is ρMIR=0.47\rho_{\mathrm{MIR}} = 0.47 (p=0.02p = 0.02, N=25N = 25), indicating a moderate but statistically significant correlation. Following the Papers I/II classification, J2347+2427 falls into the “weak MCV” category (∣ρMIR∣<0.6|\rho_{\mathrm{MIR}}| < 0.6). Additionally, we note that the MIR CMD (CMIRC_{\mathrm{MIR}} vs. W1; Figure 7, right panel) reveals an intriguing loop. During the initial MIR fading phase, CMIRC_{\mathrm{MIR}} remained nearly constant; the source then brightened rapidly with a pronounced BWB trend, before appearing to return toward the loop’s starting point.

As shown in Papers I/II, RWB Seyfert 1s generally exhibit bluer minimum MIR colors (CMIR,minC_{\mathrm{MIR,min}}) and larger MCV amplitudes (ΔCMIR=CMIR,max−CMIR,min\Delta C_{\mathrm{MIR}} = C_{\mathrm{MIR,max}} - C_{\mathrm{MIR,min}}) than BWB Seyfert 1s, with CMIR,min=0.7C_{\mathrm{MIR,min}} = 0.7 marking the approximate boundary between the two populations. Previously known turn-on/off CLAGN predominantly overlap with the RWB population. For J2347+2427, we measure (CMIR,min,ΔCMIR)=(0.71,0.35)(C_{\mathrm{MIR,min}}, \Delta C_{\mathrm{MIR}}) = (0.71, 0.35). Its CMIR,minC_{\mathrm{MIR,min}} lies near the RWB/BWB boundary, while its large ΔCMIR\Delta C_{\mathrm{MIR}} is >5σ> 5\sigma above the BWB population but close to the RWB Seyfert 1 median (=0.31= 0.31). In the CMIR,minC_{\mathrm{MIR,min}}–ΔCMIR\Delta C_{\mathrm{MIR}} plane, it lies within the 80.2% RWB and 94.4% BWB highest-density contours, slightly favoring the RWB regime.

Overall, the weak MCV and its intermediate position in the CMIR,minC_{\mathrm{MIR,min}}–ΔCMIR\Delta C_{\mathrm{MIR}} plane indicate that J2347+2427 cannot be unambiguously associated with either the BWB or RWB population. Nevertheless, as an intermediate-type CLAGN, its large ΔCMIR\Delta C_{\mathrm{MIR}} is similar to that of turn-on/off CLAGN. Moreover, most turn-on/off CLAGN (e.g., Q. Yang et al. [114, 113]) exhibit an overall negative trend in the CMIRC_{\mathrm{MIR}}–W1W1 diagram, in contrast to the positive ρMIR\rho_{\mathrm{MIR}} observed in J2347+2427. We discuss these unusual MCV properties in § 5.2.

Spectral Energy Distribution and Host-Galaxy Decomposition

If the host contribution is significant in SlowS_{\mathrm{low}}, host contamination would dilute the observed color variations (Figure 7) and suppress the flux ratios RR (Figure 5) in the nucleus. Furthermore, the presence of obscuring material could provide an explanation for the redder optical continuum and higher Balmer decrement in SlowS_{\mathrm{low}}. The main objective of this section is therefore to assess the host contribution and search for evidence of obscuration through spectral energy distribution (SED) analysis. Given the dramatic variability of J2347+2427, we first examine the evolution of its SED across different epochs (§ 4.1) and then perform SED decomposition (§ 4.2). We finally examine the influence of host on color variations (§ 4.3) and flux variations (§ 4.4).

Multi-epoch SEDs

To examine the SED evolution, we constructed multi-epoch SEDs by combining photometry from ZTF, PS1, WISE, and Swift/UVOT. Two 150-day windows, marked by gray shaded regions in Figure 1, sample the declining (SdeclS_{\mathrm{decl}}) and rising (SriseS_{\mathrm{rise}}) phases. The low state SlowS_{\mathrm{low}} is also included, with UV coverage supplied by three Swift/UVOT observations obtained during the early part of SlowS_{\mathrm{low}} (green shaded region in Figure 1). For each window, average flux densities were computed in all available bands, and the resulting SEDs are presented in Figure 8, together with the most recent Swift/UVOT photometry and the SPHEREx spectrum (cyan symbols).

Figure 8. Multi-epoch SEDs of J2347+2427. Average flux densities for SlowS_{\mathrm{low}} (black), SdeclS_{\mathrm{decl}} (red), and SriseS_{\mathrm{rise}} (blue). Cyan symbols denote the most recent Swift/UVOT photometry and SPHEREx spectrum. MIR flux-density variations are not fully synchronized with the optical/UV variability, showing an apparent time delay.

The multi-epoch SEDs further illustrate the dust-delayed response: from SdeclS_{\mathrm{decl}} to SlowS_{\mathrm{low}}, the optical flux decreased only mildly while the MIR flux dropped dramatically; from SlowS_{\mathrm{low}} to SriseS_{\mathrm{rise}}, the optical/UV flux rose rapidly, but the MIR brightened more gradually and did not recover to the SdeclS_{\mathrm{decl}} level. The most recent Swift/UVOT and SPHEREx data show a further mild increase in the optical/UV continuum. At the same time, the MIR emission now exceeds the SdeclS_{\mathrm{decl}} flux, consistent with the dust needing sufficient time to respond to the luminosity changes.

SED Decomposition for the Low State

Based on the analysis in § 4.1, we consider SlowS_{\mathrm{low}} to be the period least affected by dust-heating delays. We decomposed the SlowS_{\mathrm{low}} photometry using CIGALE (v2025.0; M. B. Bogue et al. 2019), which enforces energy balance between dust-absorbed stellar light and its re-emission. The AGN component (disk + torus) was modeled with the SKIRTOR UV-to-IR templates (M. Stalevski et al. [95, 96]). To account for potential LoS extinction in the SED modeling, we activated the “polar dust” module in CIGALE, which acts as a dust screen in front of the accretion disk, without specifying its radial location, and is characterized by various empirical extinction curves (G. Yang et al. [112]). The principal parameter ranges are:

  • Star formation history: delayed model with e-folding time τmain∈[500,8000]\tau_{\mathrm{main}} \in[500, 8000] Myr, age = 5000 Myr, no burst.

  • Stellar populations ([11]): Chabrier IMF ([17]), Z=0.02Z = 0.02.

  • Dust attenuation ([18]): AVISM∈[0.01,1.0]A_{V}^{\mathrm{ISM}} \in[0.01, 1.0] mag, slope_ISM =−0.7= -0.7, slope_BC =−1.3= -1.3, mu =0.44= 0.44.

  • AGN torus: Θ∈[20∘,60∘]\Theta\in[20^\circ, 60^\circ], τ9.7∈{3,5,9}\tau_{9.7} \in\{3, 5, 9\}, p,q∈{0.0,1.0,1.5}p, q \in\{0.0, 1.0, 1.5\}, Rout/Rin=20R_{\mathrm{out}}/R_{\mathrm{in}} = 20, i∈[0∘,50∘]i \in[0^\circ, 50^\circ].

  • Polar dust: SMC law, E(B−V)∈[0,0.7]E(B - V) \in[0, 0.7] mag.

We kept all other parameters at the CIGALE defaults. The best fit yields a reduced χ2=0.94\chi^2 = 0.94 (Figure 9), with a torus half-opening angle Θ≈37∘±10∘\Theta\approx37^\circ\pm10^\circ and a low inclination i≈15∘±11∘i \approx15^\circ\pm11^\circ, consistent with the Type 1 nature. The fit requires a moderate polar dust extinction E(B−V)SED=0.26±0.08E(B-V)_{\mathrm{SED}} = 0.26 \pm0.08 mag, providing evidence for LoS obscuration during the low state. Disabling the polar dust component severely degrades the fit, yielding a reduced χ2=5.5\chi^2 = 5.5. During SlowS_{\mathrm{low}}, the host contributes significantly to the observed flux: 27.1% in gg, 48.3% in rr, 23.6% in W1, and 13.8% in W2. Such contamination can significantly dilute the observed nuclear variability, particularly in the optical bands. The BADASS decomposition of the SDSS/eBOSS spectrum yields host fractions of 25.9% in gg and 36.7% in rr. The gg-band estimate

Figure 9. CIGALE SED fitting results for SlowS_{\mathrm{low}}. Individual model components are shown as solid lines. The green dashed line indicates the intrinsic (unattenuated) accretion disk SED adopted in CIGALE based on A. Feltre et al. [33].

agrees closely with CIGALE, while the rr-band fraction is lower, although a direct comparison is limited by differences in aperture coverage (2′′2^{\prime\prime}-diameter) and decomposition methods.

Host-subtracted Color Variations

The SED decomposition (§ 4.2) shows that the host contributes appreciable flux in SlowS_{\mathrm{low}}. To isolate the intrinsic nuclear colors, we subtracted the host component and reconstructed the optical and MIR CMDs (Figure 10). After host subtraction, the optical colors become bluer (g−rg-r decreases) and the MIR colors redder (W1−W2W1-W2 increases), reflecting the fact that starlight is redder than the disk continuum in the optical but bluer than the hot-dust emission in the MIR. For the brighter ZTF data, the host-subtracted CMD yields a slope of Δg/Δr=1.14\Delta g/\Delta r = 1.14. The faint points remain systematically offset from the linear relation, with a maximum offset of Δ(g−r)=0.24±0.03\Delta(g-r) = 0.24 \pm0.03 mag. In the MIR, the host-subtracted ρMIR\rho_{\mathrm{MIR}} increases to 0.67, moving J2347+2427 from the weak MCV (ρMIR<0.6\rho_{\mathrm{MIR}} < 0.6) to the BWB category (ρMIR>0.6\rho_{\mathrm{MIR}} > 0.6). Nevertheless, the prominent MIR color loop persists.

Figure 10. Host-subtracted color–magnitude diagrams. The symbols are identical to those in Figure 7. Left: Host-subtracted ZTF g−rg-r vs. gg. The points at faint magnitudes (red) remain systematically above the linear relation defined by the brighter data (black). Right: Host-subtracted WISE W1−W2W1-W2 vs. W1W1. After host subtraction, the source exhibits a BWB trend (ρMIR=0.67\rho_{\mathrm{MIR}}=0.67), while the MIR color loop persists.

Host-subtracted Flux Ratios

A non-negligible host contribution dilutes the observed optical and MIR variability. After subtracting the host, the peak-to-peak gg-band variability factor increases from ∼8.2×\sim8.2\times to 12×12\times, consistent with the large continuum variation between the SDSS and LJT spectra derived from the spectral decomposition. We further correct the flux ratios RR between the characteristic flux levels during SlowS_{\mathrm{low}} and ShighS_{\mathrm{high}} (purple diamonds in Figure 11). The gg-band flux ratio increases to Rgsub−host=5.4±0.4R^{\mathrm{sub-host}}_{g}=5.4\pm0.4, making the contrast with RXR_X even more pronounced.

Figure 11. Multi-wavelength flux variation factor of J2347+2427. Black symbols show the observed values (same as Figure 5), purple symbols show the host-corrected values, and red symbols show the intrinsic values after further correcting for continuum extinction.

Discussion

A Recurrent Intermediate-type CLAGN Driven by Intrinsic Accretion-State Changes

The optical spectra of J2347+2427 over the past three decades reveal a Type 1.0→\rightarrow1.5→\rightarrow1.2 transition (§ 3.1): the broad Balmer lines weakened from the earliest HS spectrum to the SDSS spectrum and re-strengthened in the recent LJT spectra following the 2020 outburst. This behavior is characteristic of the broader CLAGN population. In tandem with the Balmer-line changes, the X-ray-to-MIR fluxes exhibit coordinated dramatic recurrent variability.

The dramatic state transitions naturally raise the question of whether the observed variability is driven primarily by changes in the LoS obscuration or by intrinsic changes in the accretion flow. Several observational characteristics favor the latter. The optical LCs exhibit short-timescale stochastic flares superposed on the long-term brightening trend, together with a clear BWB behavior, both of which are characteristic signatures of thermal fluctuations in an inhomogeneous accretion disk (J. Dexter & E. Agol [26]; Z.-Y. Cai et al. [14, 15]). More importantly, the MIR emission varies coherently with the optical/UV continuum and exhibits a large amplitude (ΔW1=1.14\Delta W1=1.14 mag). Explaining such a large MIR brightening solely by a reduction in obscuration would require an extreme extinction of AV≳15A_V\gtrsim15 mag (assuming an SMC extinction law; RVSMC=2.7R^{\mathrm{SMC}}_V=2.7; M. L. Prevot et al. [73]), which is inconsistent with the optical data. Furthermore, the joint modeling of stacked Swift/XRT spectra requires intrinsic X-ray luminosity variations (§ 3.2). At the same time, the X-ray spectra provide no compelling evidence for variable gas absorption, yielding only a relatively low upper limit on the absorbing column density.

Although some CLAGN have been suggested to be triggered by TDEs, this scenario appears unlikely for J2347+2427. The optical LC remained in a high state following the 2020 outburst rather than exhibiting the monotonic t−5/3t^{-5/3} decline expected for a TDE (M. J. Rees [74]). Moreover, J2347+2427 has undergone recurrent changing-look transitions over nearly three decades, making it difficult to reconcile the observed behavior with a single stellar disruption. In the MIR, Y. Yao et al. [115] showed that TDEs typically exhibit much more rapid color evolution than CLAGN, with CMIRC_{\mathrm{MIR}} variation rates (CVRs) generally >0.2>0.2 mag yr−1^{-1}. By comparison, the 2020 MIR outburst of J2347+2427 has CVR ≈0.06\approx0.06 mag yr−1^{-1}, substantially slower than expected for TDEs.

Recently, A. Jana et al. [49] found that spectroscopic changing-look transitions are universally accompanied by substantial changes in Eddington ratio and identified a characteristic transition value of log⁡λEdd∼−2\log\lambda_{\mathrm{Edd}}\sim-2 for Type 1.0 ↔\leftrightarrow 1.8/1.9/2 transitions. The evolution of J2347+2427 is broadly consistent with this picture: its Eddington ratio increased from $\log\lambda_{\mathrm{Edd}}=-2.09 to −0.90, crossing the proposed transition regime. In particular, the low-state Eddington ratio lies close to the proposed transition value, where the inner optically thick accretion disk may become truncated and replaced by a hot radiatively inefficient flow. Such a structural change could naturally reduce the contribution from the hottest part of the accretion disk, producing additional reddening that contributes to the offset of the low-state data from the linear optical color–magnitude relation (red points in Figure 7 and Figure 10).

Nevertheless, the substantial evolution of the broad-line Balmer decrement across different epochs may also reflect a contribution from variable obscuration (§ 3.1). X. Dong et al. [29] found that (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} in unobscured AGN are clustered around 3. The much larger value of 7.2 observed during SlowS_{\mathrm{low}} may therefore indicate additional dust reddening. We further discuss this possibility in § 5.4.

The radio emission is a notable exception to the coordinated variability: it remained steady at ∼2\sim2 mJy across the 2020 outburst (§ 3.3). This may indicate that the radio emission arises from an extended, optically thin region (e.g., shocked gas at the terminus of a jet or outflow) that does not respond promptly to changes in the central engine, and/or that the jet power is governed primarily by the black hole spin rather than by the instantaneous accretion rate.

Long-term MCV of the Changing-look Transition

Our interest in J2347+2427 was initially motivated by its unusually large MIR variability amplitude and distinctive MCV, as introduced in § 1 (also see the details in § 3.4). Such a large amplitude is naturally expected for an AGN undergoing an accretion-state transition. Here we focus on the MCV behavior of J2347+2427.

J2347+2427 is initially classified as a weak MCV source (∣ρMIR∣=0.47<0.6|\rho_{\mathrm{MIR}}| = 0.47 < 0.6). Within the framework of Papers I/II, this indicates that the variable nucleus does not fully dominate the MIR emission and that host contamination contributes to the weak MCV, as seen in SlowS_{\mathrm{low}}. After subtracting the host contribution, the MCV strengthens to ρMIR=0.67\rho_{\mathrm{MIR}} = 0.67, only slightly above the BWB threshold, but still supports the proposed host-dilution scenario.

In the CMIR,minC_{\mathrm{MIR,min}}–ΔCMIR\Delta C_{\mathrm{MIR}} plane, turn-on/off CLAGN predominantly overlap with the RWB population (Paper I). This can be naturally understood as a consequence of the dramatic transition between the AGN-dominated and host-dominated states. For example, Z. Sheng et al. [90] showed that turn-off CLAGN evolve from AGN-like MIR colors (CMIR≈0.8C_{\mathrm{MIR}} \approx0.8) to galaxy-like colors (CMIR≲0.5C_{\mathrm{MIR}} \lesssim0.5), resulting in both bluer CMIR,minC_{\mathrm{MIR,min}}, measured during the host-dominated state, and larger ΔCMIR\Delta C_{\mathrm{MIR}}.

In contrast, J2347+2427 does not enter a host-dominated MIR state. Although the host contributes non-negligibly in SlowS_{\mathrm{low}}, it accounts for only 23.6% of the W1W1 emission according to the CIGALE result. Combined with the fact that its broad Balmer lines never completely vanish, this indicates that the nucleus remains active throughout the changing-look cycle. Thus, its CMIR,minC_{\mathrm{MIR,min}} is instead measured during an active nuclear state, when the surrounding dust is hotter. This leads to a more moderate MIR color evolution (CMIR=0.71C_{\mathrm{MIR}} = 0.71–1.06, remaining close to AGN-like color) compared to extreme CLAGN, while still exceeding that of ordinary Seyfert 1s, placing J2347+2427 between the RWB and BWB populations in the CMIR,minC_{\mathrm{MIR,min}}–ΔCMIR\Delta C_{\mathrm{MIR}} plane. The persistent nuclear activity also explains the positive ρMIR\rho_{\mathrm{MIR}} of J2347+2427, as its MIR variability is less affected by host dilution than in host-dominated systems, where negative trends are more common, including RWB Seyfert 1s (Papers I/II) and turn-on/off CLAGN.

Intermediate-type CLAGN may similarly retain substantial nuclear activity in their low states without becoming host-dominated, and may therefore exhibit MCV properties more similar to J2347+2427 than to turn-on/off CLAGN.

MIR Color–Magnitude Hysteresis

Another remarkable property of J2347+2427 is its MIR color–magnitude loop (Figure 7 and Figure 10, right panels), which a single-temperature, synchronous dust model in Papers I/II cannot reproduce. This loop contributes substantially to the weak MCV, even after host subtraction.

One possibility is a change in the dust distribution, such as widespread grain sublimation during the brightening phase. To test this, we estimated the dust temperature TdT_{\mathrm{d}} for each WISE epoch by fitting the host-subtracted W1W1 and W2W2 flux densities with a modified blackbody Fν∝Qabs(λ;a)Bλ(Td)F_{\nu} \propto Q_{\mathrm{abs}}(\lambda; a) B_{\lambda}(T_{\mathrm{d}}), where absorption efficiencies QabsQ_{\mathrm{abs}} are taken from A. Laor & B. T. Draine [54]. For typically a=0.1a = 0.1 μ\mum silicate, silicon carbide, and graphite grains, we obtain temperatures in the ranges 705–928 K, 882–1287 K, and 575–711 K, respectively, all well below sublimation thresholds. Extensive dust sublimation is therefore unlikely to account for the loop.

An extended dust torus with a radial temperature gradient naturally explains the loop. Since dust at smaller radii responds more quickly to changes in the central engine, a brightening event heats the innermost dust first, shifting the MIR color blueward while the cooler outer dust has yet to respond. Conversely, when the central luminosity declines, the inner dust cools promptly and the color reddens, whereas the outer dust still retains its previous elevated temperature. Consequently, at a given observed W1W1 flux, the source appears bluer when it is brightening and redder when it is fading, tracing a loop in the CMD. Indeed, a MIR hysteresis loop has also been observed in the nucleus of UGC 11487 following a TDE, likely arising from a similar multi-zone dust response (I. O. Izviekova et al. [48]).

For J2347+2427, the recurrent changing-look behavior provides particularly favorable conditions for revealing the MIR loop. The circumnuclear dust has undergone repeated cooling and reheating over several years, allowing the delayed response of dust at different radii to become visible in the CMD. A single changing-look episode might reveal only part of the loop, whereas shorter-timescale stochastic AGN variability could partially wash out this hysteresis signature. More extreme changing-look events or other outbursts (e.g., TDEs) may additionally involve substantial dust sublimation, producing a more complex morphology.

Broad-line Balmer Decrement and Possible Contribution from Variable Obscuration

As discussed in § 5.1, the dramatic variability of J2347+2427 is primarily driven by an accretion-state transition, potentially accompanied by intrinsic continuum reddening in SlowS_{\mathrm{low}} (e.g., a truncated accretion disk). Nevertheless, we cannot exclude a secondary contribution from variable obscuration. The substantial variation in the broad-line Balmer decrement provides one line of evidence for variable obscuration. At the same time, dust obscuration could also produce continuum reddening similar to that caused by intrinsic changes in the accretion flow. Consistent with this interpretation, the SED decomposition of SlowS_{\mathrm{low}} requires a polar dust component with E(B−V)SED=0.26±0.08E(B - V)_{\mathrm{SED}} = 0.26 \pm0.08 mag. In contrast, ShighS_{\mathrm{high}} exhibits a very blue continuum, consistent with the removal of the obscurer. The obscuration during SlowS_{\mathrm{low}} may have arisen from a dusty cloud in the clumpy torus crossing the LoS occasionally, or from dusty material on sub-pc scales that was lifted into the LoS by radiation-driven flows and/or formed in cooled BLR gas (e.g., M. Kokubo & T. Minezaki [51]). For an obscurer located on sub-pc scales, dust sublimation could remove it in the 2020 outburst. Radiation pressure may have further aided the clearing by displacing dusty material out of the LoS (C. Ricci et al. [78]).

We quantify the dust extinction from the host-subtracted ZTF g−rg-r color. The maximum deviation from the bright-state relation reported in § 4.3 corresponds to a dust color excess E(g−r)ZTF=Δ(g−r)=0.24±0.03E(g-r)_{\mathrm{ZTF}} = \Delta(g-r) = 0.24 \pm0.03 mag. Adopting an SMC extinction law with kg−kr=1.10k_g-k_r=1.10 gives

E(B−V)ZTF=E(g−r)ZTFkg−kr=0.22±0.03 mag,(3)E(B-V)_{\mathrm{ZTF}} = \frac{E(g-r)_{\mathrm{ZTF}}}{k_g-k_r} = 0.22 \pm0.03\ {\rm mag}, \tag*{(3)}

consistent with the SED-derived value E(B−V)SEDE(B-V)_{\mathrm{SED}}. We adopt the average E(B−V)cont=0.24±0.08E(B-V)_{\mathrm{cont}} = 0.24 \pm0.08 mag as the extinction toward the accretion disk continuum in SlowS_{\mathrm{low}}, corresponding to AVcont=RVSMC⋅E(B−V)cont=0.66±0.23A_V^{\mathrm{cont}} = R_V^{\mathrm{SMC}} \cdot E(B-V)_{\mathrm{cont}} = 0.66 \pm0.23 mag. This value is consistent with the median AV=0.60−0.19+0.32A_V = 0.60^{+0.32}_{-0.19} mag reported for reddened Type 1 AGN by M. A. Montalvo Hernandez et al. [67], suggesting a moderate and reasonable level of extinction for a Type 1 AGN. The extinction estimates above implicitly assume that the observed reddening of the continuum is entirely caused by dust extinction. Given that the changing-look phenomenon in J2347+2427 is primarily driven by intrinsic accretion-state changes, the extinction value derived here should therefore be regarded as an upper limit.

The broad-line Balmer decrement provides an additional probe of the obscuration toward the BLR, evolving from (Hα/Hβ)br≈4.5(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} \approx4.5 in the earliest HS spectrum to 7.2 during the SDSS epoch and 2.4 in the recent LJT spectra. If interpreted solely as dust extinction, reproducing the SDSS decrement from the LJT value requires AVBLR≈2.95A_V^{\mathrm{BLR}} \approx2.95 mag, implying a dereddened broad Hα\alpha flux ∼9×\sim9\times higher than that measured in the LJT spectra. Such a large intrinsic flux in the low state is difficult to reconcile with the observed accretion-state evolution.

In reality, the broad-line Balmer decrement is a complex diagnostic that can be affected by factors beyond dust extinction. The intrinsic (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} depends on the physical conditions of the dense BLR, with radiative-transfer effects and changes in gas density and ionization state potentially producing substantial variations. Photoionization calculations by Z. Yan et al. [111] show that, at high BLR densities (nH≳109 cm−3n_{\mathrm{H}} \gtrsim10^9\ \mathrm{cm}^{-3}), the intrinsic (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} can exceed 10 at lower densities and decreases with increasing gas density. Observationally, X. Dong et al. [29] found (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} of 2.3–4.2 in blue AGNs, spanning both below and above the canonical Case B value of ≈2.86\approx2.86. More recently, multi-epoch spectroscopy by S. Son et al. [94] revealed systematic variations in the Balmer decrement with the AGN continuum luminosity that cannot be fully explained by standard dust-extinction curves, suggesting that optical-depth effects, collisional excitation in the high-density BLR may play important roles. Similarly, C.-H. Chen et al. (2026) found that the elevated broad-line decrements in little red dots are primarily associated with high-density gas rather than dust extinction. The substantial change in (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} for J2347+2427 therefore likely includes a contribution from intrinsic BLR evolution.

A delayed response of the BLR to the changing ionizing continuum may provide another contribution. The “changing-look sequence” proposed by H. Guo et al. [40] describes a progressive response of the broad emission lines to changes in the ionizing continuum, with Hβ\beta responding more rapidly than Hα\alpha. Such a sequence could naturally produce a temporarily enhanced (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} during the fading phase and a lower ratio as the BLR re-equilibrates following the subsequent brightening. This scenario is qualitatively consistent with the observed evolution of the broad-line decrement in J2347+2427. This effect may be relevant to the 1997 HS spectrum, which exhibits relatively strong broad Hα\alpha emission despite its intermediate continuum level. The earlier ROSAT observation in 1990 revealed an X-ray flux comparable to recent measurements, suggesting that J2347+2427 has experienced a previous high-activity state. If the ionizing continuum subsequently declined before the HS observation, the delayed BLR response could have maintained a relatively high broad Hα\alpha flux.

Other factors, such as obscurer kinematics, may also affect the observed decrement. An obscuring structure moving relative to the BLR can modify the observed decrement and produce a large (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} with less extinction than expected for a static foreground screen ([92]). In the present case, however, reconciling the SDSS Balmer decrement with the much smaller continuum extinction of AVcontA_{V}^{\mathrm{cont}} would require a relative velocity of ≳1.5×105 km s−1\gtrsim1.5 \times10^{5}\ \mathrm{km\ s^{-1}}, which is implausibly large. Thus, the motion of the obscuring material is unlikely to be a significant factor.

Relatively Modest X-ray Variability across the Changing-look Transition

The X-ray and optical emission exhibit broadly consistent long-term behavior, suggesting that the two bands remain coupled across the changing-look transition. However, as shown in Figure 5, the X-ray variability (RX=2.7±0.8R_{\mathrm{X}} = 2.7 \pm0.8) is smaller than the gg-band variability, and this difference becomes even more pronounced after subtracting the host contribution (Figure 11, Rgsub−host=5.4±0.4R_{g}^{\mathrm{sub-host}} = 5.4 \pm0.4).

The presence of obscuration in SlowS_{\mathrm{low}} may help explain the relatively modest X-ray variability. We therefore correct the host-subtracted flux ratios for the extinction in SlowS_{\mathrm{low}} derived in § 5.4. Because the UV correction is highly sensitive to the adopted extinction curve, we conservatively apply the gg-band extinction correction to the UV bands, such that the resulting UV flux ratios should be regarded as upper limits. After correcting for both host dilution and extinction (red crosses in Figure 11), the extinction-corrected optical/UV ratios become comparable to the observed RXR_{\mathrm{X}}.

Nevertheless, the relatively modest X-ray variability may also be intrinsic. If the changing-look transition is primarily driven by long-term variations in the accretion flow, the optical/UV emission from the accretion disk is expected to respond directly to the changing accretion power, whereas the X-ray corona may exhibit a weaker response.

In contrast to its relatively modest long-term variability, EP/WXT monitoring reveals much stronger X-ray fluctuations on sub-year timescales, substantially exceeding those of the contemporaneous optical emission. Such behavior is consistent with X-ray emission originating from the compact inner corona, whereas the optical emission is produced over the much larger accretion disk.

Narrow-line Balmer Decrement

The narrow-line Balmer decrement, (Hα/Hβ)na=5.2(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{na}} = 5.2, provides an additional constraint on the dust distribution. Unlike the BLR, the NLR is a relatively low-density environment where the extreme high-density effects discussed above are not expected to dominate. The deviation (Hα/Hβ)na(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{na}} from the Case B value can therefore be more naturally attributed to dust extinction. Assuming an intrinsic ratio of 2.86, this corresponds to AVNLR≈1.6 magA_{V}^{\mathrm{NLR}} \approx1.6\ \mathrm{mag}.

Given the much larger spatial extent of the NLR, if its extinction were produced by an isotropic obscuring screen, the nuclear continuum and broad-line emission would be expected to experience comparable or stronger extinction than AVNLRA_{V}^{\mathrm{NLR}}. Interestingly, the inferred continuum extinction is only AVcont≈0.66 magA_{V}^{\mathrm{cont}} \approx0.66\ \mathrm{mag} even in the reddest low state, and (Hα/Hβ)na(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{na}} is also higher than (Hα/Hβ)br(\mathrm{H}\alpha/\mathrm{H}\beta)^{\mathrm{br}} in high states. This suggests that the dust distribution across the extended NLR is spatially inhomogeneous and anisotropic.

HeII 4686 Changing-look

Broad He II λ4686\lambda4686 in J2347+2427 exhibits a clear changing-look behavior (§ 3.1). One possible explanation is variable obscuration. Because He II originates from the innermost, high-ionization BLR, its emission region should experience dust extinction compa rable to that affecting the optical continuum. Under this assumption, the expected low-state He II flux (∼0.1×10−13 erg s−1 cm−2\sim0.1 \times10^{-13}~\mathrm{erg~s^{-1}~cm^{-2}}) remains well above the observational upper limit. Therefore, obscuration alone cannot account for the disappearance of broad He II, implying that intrinsic changes in the ionizing source and/or BLR are required.

Broad He II requires photons above the He+^{+} ionization threshold and is therefore a sensitive tracer of the EUV ionizing continuum. Its disappearance in the SDSS spectrum indicates that the EUV continuum was substantially suppressed during Slow. Such behavior is expected if the inner accretion flow underwent intrinsic structural changes, for example through a truncated accretion disk at low Eddington ratio discussed in § 5.1. Consequently, the He II changing-look behavior suggests that the low-state continuum is intrinsically redder, consistent with the multi-epoch SEDs, although a contribution from variable obscuration cannot be fully excluded.

This result is consistent with the He II and Hβ\beta changing-look quasars studied by W.-T. Lu & J.-X. Wang (2025), who found that both classes exhibit large-amplitude optical variability and that broad He II generally shows stronger variability than the Balmer lines, a trend also seen in J2347+2427. Such behavior is naturally understood within the changing-look sequence discussed above. Owing to its significantly higher ionization potential, broad He II responds more rapidly to changes in the ionizing continuum than the Balmer lines. Consequently, the highest-ionization region of the BLR appears to respond first to both the fading and renewed ionizing continuum, turning off before and recovering ahead of the Balmer-emitting region.

Comparison with Extreme Variability Quasars

With its large-amplitude optical/UV variability on a timescale of only a few years, J2347+2427 exhibits EVQ-like behavior despite its Seyfert luminosity. Recent statistical studies of EVQs have shown that they exhibit systematically stronger very broad emission-line components (VBCs) than ordinary quasars, suggesting enhanced activity in the innermost BLR (W. Ren et al. [75]). This behavior is consistent with a two-component BLR consisting of an intermediate line region and a very broad line region located much closer to the SMBH (e.g., B. J. Wills et al. [106]; M. S. Brotherton et al. 1994; J. W. Sulentic et al. [97]).

Our multi-epoch spectral decomposition requires an additional very broad Gaussian (FWHM≳6000–10,000 km s−1\mathrm{FWHM} \gtrsim6000\text{--}10{,}000~\mathrm{km\,s^{-1}}) to reproduce the extended line wings in both Hβ\beta and Hα\alpha regions. Although broad wings of [O III] are commonly observed in the Hβ\beta region and can contribute to the profile asymmetry, their typical widths are substantially smaller than those required here and cannot account for the additional component detected in both Hβ\beta and Hα\alpha. The VBC varies coherently with the optical continuum and He II. Notably, the He II line turned off during the low state and re-emerges in the high state, closely mirroring the VBC behavior. This tight coupling implies that the VBC originates in the innermost BLR, making it highly sensitive to variations in the ionizing continuum.

The coexistence of pronounced optical variability, an intermediate-type changing-look transition, and a strongly variable VBC in J2347+2427 is broadly consistent with the emerging view that many CLAGN represent a selection-biased subpopulation of extreme variability objects rather than a fundamentally distinct class (W. Ren et al. [76]).

Conclusions

We present a comprehensive multi-wavelength study of the extreme variability Seyfert galaxy J2347+2427, combining three decades of optical spectroscopy with optical/UV, MIR, radio, and X-ray monitoring. The source exhibits recurrent intermediate-type changing-look behavior, with Seyfert-type transitions from Type 1.0 to 1.5 and then to 1.2, accompanied by the disappearance and reappearance of broad He II λ4686\lambda4686 and strong evolution of the broad-line Balmer decrement. Long-term monitoring reveals coordinated variability from the optical/UV to the X-ray and MIR, including a loop-like MIR color variation (MCV) and no significant intrinsic X-ray absorption from X-ray spectral analysis. Our main conclusions are summarized as follows.

  1. J2347+2427 is identified as a recurrent intermediate-type CLAGN, exhibiting repeated Seyfert-type transitions (Type 1.0→\rightarrow1.5→\rightarrow1.2) together with coordinated variability from the X-ray to the MIR. The broad He II line disappears during the low state and re-emerges in the high state, indicating dramatic changes in the ionizing continuum.

  1. The coordinated multi-wavelength variability, the evolution of the Eddington ratio across the proposed transition regime, and the lack of compelling evidence for variable X-ray absorption favor intrinsic changes in the accretion state as the primary driver of the changing-look behavior. Nevertheless, we cannot completely exclude a secondary contribution from variable obscuration.

  1. Host-galaxy subtraction moves J2347+2427 from the weak MCV to the bluer-when-brighter (BWB) regime, providing direct support for the host-dilution scenario proposed in Papers I/II. Compared with classical turn-on/off CLAGN, J2347+2427 exhibits a more moderate MCV, occupying an intermediate position between the BWB and redder-when-brighter (RWB) populations while retaining AGN-like MIR colors throughout the changing-look cycle, consistent with the more moderate state transition of this recurrent intermediate-type CLAGN.

  1. The prominent MIR color loop reveals a delayed response of the torus to the varying accretion disk, indicating that the MCVs cannot be attributed solely to the host-dilution effect proposed in Papers I/II, particularly for sources exhibiting non-linear color–magnitude evolution (i.e., weak/no MCV).

With its extreme variability and repeated state transitions, J2347+2427 is an excellent laboratory for future time-domain studies, including spectroscopic reverberation mapping campaigns to resolve the BLR structure and trace the accretion–obscuration interplay.

Acknowledgments

This work was supported by the National Science Foundation of China (NSFC-12573110, 12333002, 12233001, 12221003, 12133001, 11833007 and 12433004) and the China Manned Space Program (CMS-CSST-2025-A07, CMS-CSST-2025-A09). YLA was supported by the Natural Science Foundation of Top Talent of SZTU (GDRC202208) and the Shenzhen Science and Technology program (JCYJ20230807113910021). LMD also acknowledges the support from the Key Laboratory for Astronomical Observation and Technology of Guangzhou, the Astronomy Science and Technology Research Laboratory of the education department of Guangdong Province.

This work is based on data obtained by Einstein Probe (also known as “Tianguan” in Chinese), a space mission supported by the Strategic Priority Program on Space Science of Chinese Academy of Sciences, in collaboration with ESA, MPE, and CNES (Grant No. XDA15310000, No. XDA15052100). This research makes use of data products from the Wide-field Infrared Survey Explorer (WISE) and the Near-Earth Object Wide-field Infrared Survey Explorer (NEOWISE). WISE is a joint project of the University of California, Los Angeles, and the Jet Propulsion Laboratory/California Institute of Technology; NEOWISE is a project of the Jet Propulsion Laboratory/California Institute of Technology. WISE and NEOWISE are funded by the National Aeronautics and Space Administration. This scientific work uses data obtained from Inyarrimanha Ilgari Bundara / the Murchison Radio-astronomy Observatory. We acknowledge the Wajarri Yamaji People as the Traditional Owners and Native Title Holders of the Observatory site. CSIRO’s ASKAP radio telescope is part of the Australia Telescope National Facility (https://ror.org/05qajvd42). Operation of ASKAP is funded by the Australian Government with support from the National Collaborative Research Infrastructure Strategy. ASKAP uses the resources of the Pawsey Supercomputing Research Centre. Establishment of ASKAP, Inyarrimanha Ilgari Bundara, the CSIRO Murchison Radio-astronomy Observatory and the Pawsey Supercomputing Research Centre are initiatives of the Australian Government, with support from the Government of Western Australia and the Science and Industry Endowment Fund. This paper includes archive data obtained through the CSIRO ASKAP Science Data Archive, CASDA (https://data.csiro.au). DeepSeek was used to assist with language editing and polishing of the manuscript. All AI-assisted text was subsequently reviewed, verified, and edited by the authors.

References

  1. [2]Allison, J. R., Sadler, E. M., Amaral, A. D., et al. 2022, PASA, 39, e010, doi: 10.1017/pasa.2022.3DOI
  2. [3]Amrutha, N., Wolf, C., Onken, C. A., et al. 2026, Nature Communications, 17, 2385, doi: 10.1038/s41467-026-69166-wDOI
  3. [4]Antonucci, R. 1993, ARA&A, 31, 473, doi: 10.1146/annurev.aa.31.090193.002353DOI
  4. [5]Arnaud, K. A. 1996, in Astronomical Society of the Pacific Conference Series, Vol. 101, Astronomical Data Analysis Software and Systems V, ed. G. H. Jacoby & J. Barnes, 17
  5. [6]Assef, R. J., Stern, D., Kochanek, C. S., et al. 2013, ApJ, 772, 26, doi: 10.1088/0004-637X/772/1/26DOI
  6. [9]Breeveld, A. A., Curran, P. A., Hoversten, E. A., et al. 2010, MNRAS, 406, 1687, doi: 10.1111/j.1365-2966.2010.16832.xDOI
  7. [11]Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000, doi: 10.1046/j.1365-8711.2003.06897.xDOI
  8. [12]Burrows, D. N., Hill, J. E., Nousek, J. A., et al. 2005, SSRv, 120, 165, doi: 10.1007/s11214-005-5097-2DOI
  9. [13]Burtscher, L., Davies, R. I., Gracia-Carpio, J., et al. 2016, A&A, 586, A28, doi: 10.1051/0004-6361/201527575DOI
  10. [14]Cai, Z.-Y., Wang, J.-X., Gu, W.-M., et al. 2016, ApJ, 826, 7, doi: 10.3847/0004-637X/826/1/7DOI
  11. [15]Cai, Z.-Y., Wang, J.-X., Zhu, F.-F., et al. 2018, ApJ, 855, 117, doi: 10.3847/1538-4357/aab091DOI
  12. [16]Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245, doi: 10.1086/167900DOI
  13. [17]Chabrier, G. 2003, PASP, 115, 763, doi: 10.1086/376392DOI
  14. [18]Charlot, S., & Fall, S. M. 2000, ApJ, 539, 718, doi: 10.1086/309250DOI
  15. [20]Chen, G., Yang, W., Liu, Y., et al. 2025, MNRAS, 544, 1926, doi: 10.1093/mnras/staf1818DOI
  16. [21]Chen, G., Yang, W., Ye, X., et al. 2026a, arXiv e-prints, arXiv:2605.24429, doi: 10.48550/arXiv.2605.24429DOI
  17. [22]Chen, G., Yang, W., Ye, X., et al. 2026b, A&A, 707, A347, doi: 10.1051/0004-6361/202557975DOI
  18. [23]Chen, Y., Cui, W., Han, D., et al. 2021, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 11444, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, ed. J.-W. A. den Herder, S. Nikzad, & K. Nakazawa, 114445B, doi: 10.1117/12.2562311DOI
  19. [24]Cheng, H., Zhang, C., Ling, Z., et al. 2025, Experimental Astronomy, 60, 15, doi: 10.1007/s10686-025-10025-9DOI
  20. [25]Cheng, H., Pan, H.-W., Liu, Y., et al. 2026, arXiv e-prints, arXiv:2606.028101, doi: 10.48550/arXiv.2606.028101
  21. [26]Dexter, J., & Agol, E. 2011, ApJL, 727, L24, doi: 10.1088/2041-8205/727/1/L24DOI
  22. [27]Dong, Q., Zhang, Z.-X., Gu, W.-M., et al. 2026, ApJ, 1006, 133, doi: 10.3847/1538-4357/ae8095
  23. [28]Dong, Q., Zhang, Z.-X., Gu, W.-M., Sun, M., & Zheng, Y.-G. 2025, ApJ, 986, 160, doi: 10.3847/1538-4357/add331
  24. [29]Dong, X., Wang, T., Wang, J., et al. 2008, MNRAS, 383, 581, doi: 10.1111/j.1365-2966.2007.12560.xDOI
  25. [31]Drake, A. J., Djorgovski, S. G., Mahabal, A., et al. 2009, ApJ, 696, 870, doi: 10.1088/0004-637X/696/1/870DOI
  26. [32]Evans, P. A., Page, K. L., Beardmore, A. P., et al. 2023, MNRAS, 518, 174, doi: 10.1093/mnras/stac2937DOI
  27. [33]Feltre, A., Hatziminaoglou, E., Fritz, J., & Franceschini, A. 2012, MNRAS, 426, 120, doi: 10.1111/j.1365-2966.2012.21695.xDOI
  28. [34]Flewelling, H. A., Magnier, E. A., Chambers, K. C., et al. 2020, ApJS, 251, 7, doi: 10.3847/1538-4365/abb82dDOI
  29. [35]Frederick, S., Gezari, S., Graham, M. J., et al. 2019, ApJ, 883, 31, doi: 10.3847/1538-4357/ab3a38DOI
  30. [36]Gaia Collaboration, Vallenari, A., Brown, A. G. A., et al. 2023, A&A, 674, A1, doi: 10.1051/0004-6361/202243940
  31. [37]Gehrels, N., Chincarini, G., Giommi, P., et al. 2004, ApJ, 611, 1005, doi: 10.1086/422091DOI
  32. [38]Gilli, R., Maiolino, R., Marconi, A., et al. 2000, A&A, 355, 485, doi: 10.48550/arXiv.astro-ph/0001107DOI
  33. [39]Guo, H., & Gu, M. 2016, ApJ, 822, 26, doi: 10.3847/0004-637X/822/1/26DOI
  34. [40]Guo, H., Shen, Y., He, Z., et al. 2020, ApJ, 888, 58, doi: 10.3847/1538-4357/ab5db0DOI
  35. [41]Hagen, H.-J., Groote, D., Engels, D., & Reimers, D. 1995, A&AS, 111, 195
  36. [42]Hale, C. L., McConnell, D., Thomson, A. J. M., et al. 2021, PASA, 38, e058, doi: 10.1017/pasa.2021.47DOI
  37. [43]HI4PI Collaboration, Ben Bekhti, N., Flöer, L., et al. 2016, A&A, 594, A116, doi: 10.1051/0004-6361/201629178
  38. [44]Hill, J. E., Burrows, D. N., Nousek, J. A., et al. 2004, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 5165, X-Ray and Gamma-Ray Instrumentation for Astronomy XIII, ed. K. A. Flanagan & O. H. W. Siegmund, 217-231, doi: 10.1117/12.505728DOI
  39. [45]Hodgkin, S. T., Breedt, E., Delgado, A., et al. 2020, Transient Name Server Discovery Report, 2020-3830, 1
  40. [46]Hon, W. J., Webster, R., & Wolf, C. 2020, MNRAS, 497, 192, doi: 10.1093/mnras/staa1939DOI
  41. [47]Hotan, A. W., Bunton, J. D., Chippendale, A. P., et al. 2021, PASA, 38, e009, doi: 10.1017/pasa.2021.1DOI
  42. [48]Izviekova, I. O., Kompaniiets, O. V., & Vavilova, I. B. 2026, arXiv e-prints, arXiv:2607.17815, doi: 10.48550/arXiv.2607.17815DOI
  43. [49]Jana, A., Ricci, C., Temple, M. J., et al. 2025, A&A, 693, A35, doi: 10.1051/0004-6361/202451058DOI
  44. [50]Kaastra, J. S. 2017, A&A, 605, A51, doi: 10.1051/0004-6361/201629319DOI
  45. [51]Kokubo, M., & Minezaki, T. 2020, MNRAS, 491, 4615, doi: 10.1093/mnras/stz3397DOI
  46. [52]Komossa, S., Grupe, D., Kraus, A., et al. 2026, ApJ, 1003, 46, doi: 10.3847/1538-4357/ae610fDOI
  47. [53]LaMassa, S. M., Cales, S., Moran, E. C., et al. 2015, ApJ, 800, 144, doi: 10.1088/0004-637X/800/2/144DOI
  48. [54]Laor, A., & Draine, B. T. 1993, ApJ, 402, 441, doi: 10.1086/172149DOI
  49. [60]Lyu, B., Wu, Q., Yan, Z., Yu, W., & Liu, H. 2022, ApJ, 927, 227, doi: 10.3847/1538-4357/ac5256DOI
  50. [61]MacLeod, C. L., Ross, N. P., Lawrence, A., et al. 2016, MNRAS, 457, 389, doi: 10.1093/mnras/stv2997DOI
  51. [62]MacLeod, C. L., Green, P. J., Anderson, S. F., et al. 2019, ApJ, 874, 8, doi: 10.3847/1538-4357/ab05e2DOI
  52. [63]Maiolino, R., & Rieke, G. H. 1995, ApJ, 454, 95, doi: 10.1086/176468DOI
  53. [64]Masci, F. J., Laher, R. R., Rusholme, B., et al. 2019, PASP, 131, 018003, doi: 10.1088/1538-3873/aae8acDOI
  54. [65]McConnell, D., Hale, C. L., Lenc, E., et al. 2020, PASA, 37, e048, doi: 10.1017/pasa.2020.41DOI
  55. [66]Merloni, A., Dwelly, T., Salvato, M., et al. 2015, MNRAS, 452, 69, doi: 10.1093/mnras/stv1095DOI
  56. [67]Montalvo Hernandez, M. A., Goulding, A. D., & Greene, J. E. 2026, arXiv e-prints, arXiv:2604.13031, doi: 10.48550/arXiv.2604.13031
  57. [68]Noda, H., & Done, C. 2018, MNRAS, 480, 3898, doi: 10.1093/mnras/sty2032DOI
  58. [69]Osterbrock, D. E. 1977, ApJ, 215, 733, doi: 10.1086/155407DOI
  59. [70]Paliya, V. S., Stalin, C. S., Domínguez, A., & Saikia, D. J. 2024, MNRAS, 527, 7055, doi: 10.1093/mnras/stad3650DOI
  60. [71]Planck Collaboration, Aghanim, N., Akrami, Y., et al. 2020, A&A, 641, A6, doi: 10.1051/0004-6361/201833910
  61. [72]Poole, T. S., Breeveld, A. A., Page, M. J., et al. 2008, MNRAS, 383, 627, doi: 10.1111/j.1365-2966.2007.12563.xDOI
  62. [73]Prevot, M. L., Lequeux, J., Maurice, E., Prevot, L., & Rocca-Volmerange, B. 1984, A&A, 132, 389
  63. [74]Rees, M. J. 1988, Nature, 333, 523, doi: 10.1038/333523a0DOI
  64. [75]Ren, W., Wang, J., Cai, Z., & Guo, H. 2022, ApJ, 925, 50, doi: 10.3847/1538-4357/ac3828DOI
  65. [76]Ren, W., Wang, J., Cai, Z., & Hu, X. 2024, ApJ, 963, 7, doi: 10.3847/1538-4357/ad17cbDOI
  66. [77]Ricci, C., & Trakhtenbrot, B. 2023, Nature Astronomy, 7, 1282, doi: 10.1038/s41550-023-02108-4DOI
  67. [78]Ricci, C., Trakhtenbrot, B., Koss, M. J., et al. 2017, Nature, 549, 488, doi: 10.1038/nature23906DOI
  68. [79]Roming, P. W. A., Kennedy, T. E., Mason, K. O., et al. 2005, SSRv, 120, 95, doi: 10.1007/s11214-005-5095-4DOI
  69. [80]Ruan, J. J., Anderson, S. F., Dexter, J., & Agol, E. 2014, ApJ, 783, 105, doi: 10.1088/0004-637X/783/2/105DOI
  70. [81]Rumbaugh, N., Shen, Y., Morganson, E., et al. 2018, ApJ, 854, 160, doi: 10.3847/1538-4357/aaa9b6DOI
  71. [82]Runnoe, J. C., Brotherton, M. S., & Shang, Z. 2012, MNRAS, 422, 478, doi: 10.1111/j.1365-2966.2012.20620.xDOI
  72. [84]Schmidt, M. 1963, Nature, 197, 1040, doi: 10.1038/1971040a0DOI
  73. [85]Senarath, M. R., Brown, M. J. I., Cluver, M. E., et al. 2021, MNRAS, 503, 2583, doi: 10.1093/mnras/stab393DOI
  74. [86]Sexton, R. O., Matzko, W., Darden, N., Canalizo, G., & Gorjian, V. 2021, MNRAS, 500, 2871, doi: 10.1093/mnras/staa3278DOI
  75. [87]Seyfert, C. K. 1943, ApJ, 97, 28, doi: 10.1086/144488DOI
  76. [88]Shen, Y., Richards, G. T., Strauss, M. A., et al. 2011, ApJS, 194, 45, doi: 10.1088/0067-0049/194/2/45DOI
  77. [89]Sheng, Z., Wang, T., Jiang, N., et al. 2017, ApJL, 846, L7, doi: 10.3847/2041-8213/aa85deDOI
  78. [90]Sheng, Z., Wang, T., Jiang, N., et al. 2020, ApJ, 889, 46, doi: 10.3847/1538-4357/ab5af9DOI
  79. [91]Shingles, L., Smith, K. W., Young, D. R., et al. 2021, Transient Name Server AstroNote, 7, 1
  80. [92]Sitaram, M., Li, H., Zheng, Y., et al. 2026, ApJ, 1001, 87, doi: 10.3847/1538-4357/ae5222DOI
  81. [93]Son, S., Kim, M., Ho, L. C., Kim, D., & Kim, T. 2022, ApJ, 937, 3, doi: 10.3847/1538-4357/ac8a9dDOI
  82. [94]Son, S., Kim, M., Ho, L. C., & Li, R. 2025, ApJ, 995, 37, doi: 10.3847/1538-4357/aelef1
  83. [95]Stalevski, M., Fritz, J., Baes, M., Nakos, T., & Popović, L. Č. 2012, MNRAS, 420, 2756, doi: 10.1111/j.1365-2966.2011.19775.x
  84. [96]Stalevski, M., Ricci, C., Ueda, Y., et al. 2016, MNRAS, 458, 2288, doi: 10.1093/mnras/stw444DOI
  85. [97]Sulentic, J. W., Marziani, P., & Dultzin-Hacyan, D. 2000, ARA&A, 38, 521, doi: 10.1146/annurev.astro.38.1.521DOI
  86. [98]Tan, A. H. T., Wolf, C., Amrutha, N., et al. 2026, MNRAS, 549, stag1052, doi: 10.1093/mnras/stag1052DOI
  87. [99]Tonry, J. L., Denneau, L., Heinze, A. N., et al. 2018, PASP, 130, 064505, doi: 10.1088/1538-3873/aabdf
  88. [100]Vazdekis, A., Koleva, M., Ricciardelli, E., Röck, B., & Falcón-Barroso, J. 2016, MNRAS, 463, 3409, doi: 10.1093/mnras/stw2231DOI
  89. [101]Véron-Cetty, M.-P., Joly, M., & Véron, P. 2004, A&A, 417, 515, doi: 10.1051/0004-6361:20035714
  90. [102]Voges, W., Aschenbach, B., Boller, T., et al. 1999, A&A, 349, 389, doi: 10.48550/arXiv.astro-ph/9909315DOI
  91. [103]Wang, J., Xu, D. W., Cao, X., et al. 2024, ApJ, 970, 85, doi: 10.3847/1538-4357/ad4d89DOI
  92. [104]Wang, S., Woo, J.-H., Gallo, E., et al. 2025, ApJ, 981, 129, doi: 10.3847/1538-4357/adadf3DOI
  93. [105]Whittle, M. 1992, ApJS, 79, 49, doi: 10.1086/191644DOI
  94. [106]Wills, B. J., Brotherton, M. S., Fang, D., Steidel, C. C., & Sargent, W. L. W. 1993, ApJ, 415, 563, doi: 10.1086/173186DOI
  95. [107]Winkler, H. 1992, MNRAS, 257, 677, doi: 10.1093/mnras/257.4.677DOI
  96. [108]Wright, E. L., Eisenhardt, P. R. M., Mainzer, A. K., et al. 2010, AJ, 140, 1868, doi: 10.1088/0004-6256/140/6/1868DOI
  97. [111]Yan, Z., Inayoshi, K., Chen, K., & Guo, J. 2026, ApJ, 1002, 159, doi: 10.3847/1538-4357/ae5bdb
  98. [112]Yang, G., Boquien, M., Buat, V., et al. 2020, MNRAS, 491, 740, doi: 10.1093/mnras/stz3001DOI
  99. [113]Yang, Q., Green, P. J., Wu, X.-B., et al. 2025, ApJ, 980, 91, doi: 10.3847/1538-4357/ad94edDOI
  100. [114]Yang, Q., Wu, X.-B., Fan, X., et al. 2018, ApJ, 862, 109, doi: 10.3847/1538-4357/aaca3aDOI
  101. [115]Yao, Y., Ye, J., Sun, L., et al. 2025, arXiv e-prints, arXiv:2503.10053, doi: 10.48550/arXiv.2503.10053
  102. [116]Yoon, H., Sadler, E. M., Mahony, E. K., et al. 2025, PASA, 42, e088, doi: 10.1017/pasa.2025.10046DOI
  103. [117]Yuan, W., Zhang, C., Chen, Y., & Ling, Z. 2022, in Handbook of X-ray and Gamma-ray Astrophysics, ed. C. Bambi & A. Sangangelo (Springer Nature Singapore), 86, doi: 10.1007/978-981-16-4544-0_151-1

Paper details

vlnc:W_twrbxyje
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 10

downloads · 3

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

DeepSeek (multiple versions)

View statement
Looking for related papers…
Contents