ATLAS22kjn (AT 2022fpx): A Coronal Line Emitter with an Early Light Curve Bump and Mid-Infrared Dust Echo

This paper presents a comprehensive multi-wavelength analysis of the tidal disruption event (TDE) ATLAS22kjn (AT 2022fpx), which exhibits both high-ionisation coronal lines (CLs) and a pre-peak light curve bump. The object was discovered on 31 March 2022 (MJD = 59669.5) by the Asteroid Terrestrial-impact Last Alert System (ATLAS) and is located in the nucleus of the galaxy SDSS J153103.70+532419.3 at redshift z = 0.073, corresponding to a luminosity distance of 332.0 Mpc.

The host galaxy shows no strong AGN activity, with an AGN fraction of LAGN/Lstellar = (1.1 ± 0.2) × 10−2, consistent with a low-luminosity AGN (LLAGN). The AllWISE colour (W1 − W2) = 0.17 ± 0.06 mag is too blue for significant AGN emission. Using the M BH–M stellar scaling relation from Reines & Volonteri (2015), the authors estimate a SMBH mass of log(M BH/M⊙) = 6.44 ± 0.13, with additional intrinsic scatter of 0.24 dex. The host galaxy has a stellar mass of (1.1 ± 0.1) × 1010 M⊙, a star formation rate of 0.12 ± 0.02 M⊙/yr, and an age of 3.8 ± 0.1 Gyr.

The UV/optical light curve shows a long-lived flare of 900 days that began at MJD 59670. Using a curved power-law model (Vallely et al. 2020), the authors determine a rise start time of t1 = 59670.3+0·3−0·9 (MJD), implying ATLAS22kjn was discovered 1 day before it began to rise. The initial rise slope a1 3 is consistent across different bands, suggesting a relatively constant temperature during the rise to UV/optical peak.

From blackbody fitting of the Swift UVOT photometry, the authors measure a peak bolometric luminosity of (5.4 ± 0.2) × 1043 erg s−1 at MJD = 59805+5−3. The rise time is 125.4+0·3−0·9 rest-frame days. At peak, the effective temperature is 16,000 K, which remains relatively constant over time. This temperature is low compared to typical TDEs but similar to the 18,000 K temperature found for the CLE AT 2017gge. The blackbody radius peaks at the bolometric luminosity peak and decreases with luminosity. The peak luminosity lies at the low end of TDE peak luminosities, and the decline rate places ATLAS22kjn amongst ambiguous nuclear transients (ANTs) rather than typical TDEs.

A prominent pre-peak bump is observed in the light curves, first identified by Koljonen et al. (2024) as a 'precursor' from ZTF data. With the addition of TESS data, the authors resolve both the rise and decline of this feature. Using a Gaussian Process (GP) model on the unbinned TESS light curve (after subtracting the main rise model), they find:

• The bump peaks 134+5−3 observed days (125+5−3 rest-frame days) before the main UV/optical peak

• Peak luminosity of (3.0+0·2−0·4) × 1042 erg s−1 after subtracting the rise model

• Bump duration of 9+4−2 days (defined as when the GP fit exceeds 20% of peak flux)

• Radiated energy in the TESS band of (1.5+0·9−0·5) × 1048 erg

• Peak bump colour g − r = 0.5 ± 0.1 mag, corresponding to a blackbody temperature of 7800 ± 500 K

• Blackbody radius of (1.06 ± 0.12) × 1015 cm

• Rise time of 3.7+0·9−0·7 days, implying a velocity of (3.3+0·7−0·9) × 104 km s−1 (0.1c)

The authors examine six possible mechanisms for the bump: nozzle shock, cooling of unbound stellar debris, stream-disk collision in a repeating partial TDE, stream-stream collision, double TDE, and wind-stream collision. They conclude that the observed characteristics are most consistent with stream-stream collision and wind-stream collision scenarios. The bump luminosity is a 1000% increase relative to the rise model, contrasting with the ≲20% increase predicted by Calderón et al. (2024) for wind-stream collision, though the authors note their rise model is not a one-to-one comparison. For stream-stream collision, the peak luminosity and duration are consistent with predictions from Huang et al. (2023b, 2024b), and reproducing the bump energy with 5% kinetic energy of a stream requires a stream velocity of 2000 km s−1, similar to the lowest injected velocities in simulations.

ATLAS22kjn exhibits numerous high-ionisation (≳100 eV) coronal lines in both optical and NIR spectra, including [Ne V] λ3347, [Ne V] λ3427, [Fe VII] λ3586, [Fe VII] λ3760, [Ar XIV] λ4414, [Fe VII] λ5160, [Fe XIV] λ5304, [Fe VII] λ5722, [Fe VII] λ6088, [Fe X] λ6376, [S XII] λ7612, and [Fe XI] λ7894. In the NIR, [Si VI] λ19620 is detected at +186d, and [Si X] λ14300 at +596d. The CLs appear before UV/optical light curve peak, preceding the detection of X-rays by 300 days and persisting after X-rays are no longer detected.

The CL luminosities are comparable to the [O III] λ5007 line luminosity, with L[Fe XIV]/L[O III] = 1.38+0·06−0·04 at +8d, indicating the CL emission is powered by the TDE flare rather than an AGN. The luminosity evolution of the CLs shows an ionisation-potential dependent trend: lines with higher ionisation potentials decline more rapidly. Using the CL luminosities, the authors constrain the temperature evolution of the ionising source, finding a decrease of ≲10% over 500 days relative to the weighted-average X-ray temperature of T̄ X ≃ 1.2 × 106 K.

The X-ray emission of ATLAS22kjn is soft, with little significant emission above 1 keV. The hardness ratio is negative throughout (HR = −0.8 to −0.4 for detections), with no clear evolution. The X-ray emission begins 250 observed days after the UV/optical peak, peaking 20 days later. This peak aligns with a small rise in the UV/optical light curves, possibly evidence for reprocessing. The X-ray blackbody temperatures are 0.09-0.11 keV, similar to typical TDEs. The authors note that the X-ray radii obtained are less than the gravitational radius and therefore unphysical, suggesting that a blackbody is a simplification and detailed disk modelling would provide a better fit.

The NEOWISE MIR light curves show a prominent dust echo, with no significant variability prior to the UV/optical flare. From blackbody fitting, the authors estimate a dust covering fraction of f c = 0.40 ± 0.03, which is 1-2 orders of magnitude higher than typical TDEs (0.001 ≲ f c ≲ 0.02) and similar to expectations for hot dust in AGNs. This suggests the presence of a dusty torus in the host galaxy.

The rest-frame lag between MIR and UV/optical peaks is 200 ± 34 days, implying the MIR-emitting dust is at a radius of 0.17 ± 0.03 pc from the SMBH (assuming light-travel time). This is similar to the 177-day lag for the CLE AT 2017gge and the 200-day lag for PTF-09ge, both with similar SMBH masses.

The authors construct a picture of the emission regions surrounding the central SMBH (Figure 17). From smallest to largest: the X-ray emitting region (though unphysically small under blackbody assumption), the UV/optical blackbody radius, the broad line region (virial estimates), the CL region, and the MIR-emitting dust at 0.17 ± 0.03 pc. The CL region lies between the broad line region and the MIR-emitting dust, with the far edge of the CL emission coinciding with the inner edge of the dusty torus. The CL widths decrease over time, suggesting the CLs probe greater radii at later times under the assumption of virialised gas. The presence of CLs at early and late times when X-rays are not detected suggests the X-ray emission is obscured, either by the CLR itself or by the dusty torus.

ATLAS22kjn is compared with other coronal line emitters including ASASSN-18jd, AT 2019qiz, AT 2021dms, AT 2022upj, AT 2017gge, and ASASSN-18ap. These objects share characteristics including blue continuum emission, strong Balmer lines, [Fe VII] emission, MIR dust echoes with high covering fractions, and similar X-ray blackbody luminosities (1042-1043 erg s−1). ATLAS22kjn is unique in showing [Ar XIV] λ4414, and its early-time detection of [Fe VII] (as early as +8d) differs from predictions by Mummery et al. (2025), who expected lower-ionisation lines to switch on 1 year post-flare.

The authors conclude that ATLAS22kjn is a TDE occurring in a galaxy nucleus with a high density of gas and a dusty torus. The paper demonstrates the importance of multi-wavelength and early-time observations, and the utility of CLEs in characterising the otherwise unobservable EUV/ultrasoft X-ray emission of TDEs. The authors encourage future theoretical work making observational predictions for TDE CL emission, and note that upcoming UV missions like ULTRASAT and UVEX will provide independent constraints on TDE SEDs further into the UV.

Improvements to AI Systems Based on This Paper:

1. Multi-wavelength transient classification and anomaly detection: The AI can be trained to identify tidal disruption events (TDEs) with ambiguous decline rates (ANTs) and pre-peak bumps by integrating UV/optical, X-ray, and MIR light curve features simultaneously, rather than relying on single-band classification. It can flag objects like ATLAS22kjn that show both coronal lines and early bumps, which current surveys may misclassify.

2. Early precursor/bump detection and physical mechanism inference: The AI can use Gaussian Process models on unbinned, high-cadence photometry (e.g., TESS) to automatically detect and characterize pre-peak bumps (rise time, duration, peak luminosity, color temperature). It can then compare observed bump properties against a library of physical models (nozzle shock, stream-stream collision, wind-stream collision, partial TDE) to output the most probable mechanism and associated physical parameters (e.g., stream velocity 2000 km/s).

3. Coronal line spectral analysis for EUV/X-ray proxy: The AI can be trained to use optical/NIR coronal line luminosities (e.g., [Fe VII], [Fe X], [Fe XIV], [Si X]) to reconstruct the unobservable EUV/soft X-ray ionizing continuum. It can predict the ionizing source temperature evolution (e.g., 1.2 × 106 K with <10% decline over 500 days) and detect changes in ionizing flux even when X-ray observations are absent or obscured.

4. Dust echo and covering fraction estimation: The AI can automatically fit MIR light curves (e.g., NEOWISE) to derive dust covering fractions and lags, then flag outliers like ATLAS22kjn with f c = 0.40 (1-2 orders of magnitude above typical TDEs) to identify host galaxies with dusty tori. It can also predict the inner dust radius (0.17 pc) from the lag and SMBH mass.

5. Physical consistency checking for emission region geometry: The AI can integrate multi-wavelength data to construct a self-consistent radial structure (X-ray, UV/optical, BLR, CLR, dust) and flag unphysical results—e.g., X-ray blackbody radii smaller than the gravitational radius—prompting alternative models (e.g., disk emission) rather than accepting a blackbody fit.

6. Early-time detection and rapid follow-up prioritization: The AI can be trained to recognize that coronal lines appear before X-rays and persist after X-ray disappearance, and that early UV/optical bumps precede the main peak by 125 days. This enables automated triggering of high-cadence UV/optical and NIR spectroscopy within days of discovery, maximizing the chance of capturing CLs and the bump.

7. Cross-survey data fusion and gap filling: The AI can combine sparse ZTF, dense TESS, Swift UVOT, and NEOWISE data to reconstruct continuous light curves, subtract main rise models, and isolate transient features (like the bump) even when individual surveys have gaps or different cadences.

8. Theoretical model comparison and prediction: The AI can be fine-tuned on simulated TDE light curves and CL emission predictions (e.g., from Mummery et al. 2025, Calderón et al. 2024) to automatically test whether observed early-time [Fe VII] detection (at +8d) contradicts current models, and to suggest revisions to the timing of lower-ionization line turn-on.

What the improved AI system can do:

• Automatically detect and classify rare TDE subtypes (e.g., CLEs with bumps) in real-time from multi-survey alert streams.

• Provide physical interpretations (mechanism, geometry, ionizing source properties) within hours of a transient's discovery, not months.

• Predict when and where to observe coronal lines and MIR echoes, optimizing telescope time.

• Flag anomalous objects for theoretical follow-up, accelerating discovery of new physical phenomena.

• Reconstruct unobservable EUV/soft X-ray emission from optical/NIR lines, enabling population studies of TDEs without X-ray telescopes.

We present an analysis of ATLAS22kjn (AT 2022fpx), whose high-ionisation coronal lines (CLs) and pre-peak light curve bump provide distinctive opportunities to investigate the physical mechanisms powering tidal disruption events (TDEs). In addition to CLs, the optical spectra show common TDE features, including a strong, blue continuum and broad Balmer and He II lines. The CLs appear before UV/optical light curve peak, preceding the detection of X-rays by about 300 days and persisting after X-rays are no longer detected, suggesting the X-ray emission is obscured at both early and late times. Using the CL luminosities, we constrain the temperature evolution of the ionising source, finding a decrease of 10 % over 500 days. In the UV/optical light curve, we observe a 9 day bump that peaks 125 rest-frame days before the peak of the main flare. Although we cannot definitively determine the origins of the bump, we find that its timescale and luminosity are most consistent with theoretical predictions for a precursor feature produced by a stream-stream collision or a wind-stream collision. ATLAS22kjn also shows a prominent dust echo in its mid-infrared (MIR) light curves, indicating a high dust covering fraction f c 0.40 plus or minus 0.03, similar to the covering fractions of other CL-emitting TDEs. From the multi-wavelength observations of ATLAS22kjn, we estimate the size and relative radii of the emission regions in its nuclear environment and determine that the CL region lies between the broad line region and the MIR-emitting dust. ATLAS22kjn demonstrates the importance of multi-wavelength and early-time observations, and the utility of CLEs in characterising the otherwise unobservable EUV/ultrasoft X-ray emission of TDEs.