- The paper presents the first AGN evidence that characteristic damping timescales coevolve with X-ray time lags, using damped-random-walk models of 17 reliable XMM-Newton observations of RE J1034+396.
- Hard-band inverse damping timescales and lags favor a counterclockwise closed loop in 90.5% of Monte Carlo trials, while soft-band and QPO loops show stronger support at 99.7% and 100.0%, respectively.
- The results support a changing hot inner flow or corona as the common origin of QPO and stochastic variability, with inferred characteristic radii broadly overlapping independent coronal-size estimates of roughly 5–12 gravitational radii.
This Letter by Zhang and Zhang (2608.19610) reports the first observational evidence among active galactic nuclei (AGNs) that the characteristic damping timescale (CDT) of stochastic X-ray variability coevolves with the soft–hard X-ray time lag in a manner analogous to the previously established coevolution between quasi-periodic oscillation (QPO) frequency and time lag. The analysis targets the narrow-line Seyfert 1 galaxy RE J1034+396, the only AGN with a robust, recurrent, multi-epoch X-ray QPO, using 18 archival XMM-Newton EPIC-pn observations. The central claim is that changes in the spatial extent of the hot inner accretion flow/corona simultaneously shape both the QPO and the stochastic variability properties of the source.
Motivation and physical framework
In black hole X-ray binaries (XRBs), QPO frequency and X-ray time lag are known to evolve together, with lag reversals between hard- and soft-lag modes observed in sources such as GRS 1915+105. Among AGNs, such coevolution had been reported only for RE J1034+396 itself, where Xia et al. showed that the QPO frequency and time lag trace a closed loop in parameter space across the 2020–2021 observing campaign [2024ApJ...961L..32X]. What remained untested was whether the stochastic variability participates in the same structural evolution.
The theoretical anchor is the truncated-disk/hot-inner-flow framework of Ingram & Done [2011MNRAS.415.2323I], in which the PSD bending frequency is tied to the viscous timescale at the outer edge of the hot inner flow, while the QPO frequency is set by Lense–Thirring precession of the same flow. If this picture holds, the CDT — defined as τdamping=(2πfbend)−1 — should evolve systematically alongside the QPO frequency as the flow's outer radius changes. Testing this prediction in an AGN is the paper's stated objective.
DRW modeling of multi-epoch light curves
The authors modeled the soft-band (0.3–1 keV) and hard-band (1–4 keV) light curves separately with a damped random walk (DRW; equivalently an Ornstein–Uhlenbeck process), fitted via Gaussian-process likelihood maximization with celerite and MCMC sampling with emcee. The DRW approach avoids Fourier-domain pathologies such as red-noise leakage and aliasing, and naturally incorporates per-point measurement errors. Reliability was enforced through standard criteria: the upper confidence limit on the CDT must be shorter than one-tenth of the observational baseline, the lower limit must exceed the mean cadence, and the standardized residuals must be consistent with white noise in Lomb–Scargle periodograms.
Of the 18 observations analyzed, 17 yielded reliable CDTs in both bands. Two results stand out:
- Soft-band CDTs exceed hard-band CDTs in all 17 valid measurements, with hard-band posterior distributions broader due to a count rate roughly an order of magnitude lower than in the soft band.
- One observation (0675440301) failed the reliability criteria and was excluded.
The measured logarithmic hard-band CDTs span lnτ≈6.02–$6.61$ (in seconds) across the loop-forming observations, corresponding to rest-frame timescales of roughly $410$–$740$ s after correcting for the source redshift of z=0.043.
Closed-loop coevolution and its statistical robustness
Adopting the hypothetical evolutionary sequence proposed by Xia et al. [2025ApJ...983...13X] — ordering the observations as Obs-b, Obs-5 through Obs-9, Obs-1, Obs-10, Obs-2 through Obs-4, and Obs-a — the authors show that when reciprocal CDT is plotted against X-ray time lag, the measurements trace counterclockwise closed loops resembling the QPO-frequency–time-lag loop. By visual inspection, the hard-band loop more closely reproduces the QPO loop, whereas the soft-band loop exhibits additional internal structure.
Because visual inspection alone is not decisive given the measurement uncertainties, the authors quantified loop orientation using the signed area enclosed by each trajectory (shoelace formula), propagated through 105 Monte Carlo realizations:
| Relation |
Normalized signed area |
Fraction counterclockwise |
| QPO frequency – lag |
3.28−0.93+0.96 |
99.975% |
| Hard-band 1/τ – lag |
4.67−3.71+5.79 |
90.506% |
| Soft-band lnτ≈6.020 – lag |
lnτ≈6.021 |
99.746% |
All three relations favor the same counterclockwise evolution, though the hard-band distribution is considerably broader — a caveat the authors state plainly rather than obscure. A further notable result is the ordered displacement of extrema along the sequence: the minima occur successively at Obs-9 for lnτ≈6.022, Obs-1 for lnτ≈6.023, Obs-2 for the QPO frequency, and Obs-3 for the time lag, indicating that all four timing observables trace a common evolutionary pattern with distinct relative phase offsets. This phase-offset structure is what accounts for the deviations of the CDT–lag loops from the QPO-frequency–lag loop, particularly in the soft band.
Physical interpretation: radial scales of the inner flow
Interpreting the hard-band CDT under dynamical, thermal (lnτ≈6.024), or viscous prescriptions, and adopting lnτ≈6.025–lnτ≈6.026, lnτ≈6.027, and lnτ≈6.028–lnτ≈6.029, the authors infer characteristic radii summarized below:
| Timescale prescription |
Parameters |
Inferred radius ($6.61$0) |
| Dynamical |
— |
4.01–27.56 |
| Thermal |
$6.61$1 |
0.86–5.94 |
| Thermal |
$6.61$2 |
2.52–17.36 |
| Viscous |
$6.61$3, $6.61$4 |
0.34–2.36 |
| Viscous |
$6.61$5, $6.61$6 |
1.00–6.89 |
Imposing an upper limit of $6.61$7 on coronal outer radii constrains the black hole mass under the dynamical interpretation and excludes part of the low-mass end under the thermal interpretation with $6.61$8. Critically, these inferred radii overlap with the independent constraints from Taylor et al. [2025ApJ...987..135T], who argued on spectral-timing grounds that the QPO originates in the hot corona, requiring an outer hot-flow radius of approximately 5–12 $6.61$9 (periastron-precession interpretation) or less than about 6 $410$0 (Lense–Thirring interpretation). This overlap, combined with the loop-like coevolution, constitutes the paper's central evidence that the hard-band CDT and the QPO probe the same evolving structure.
For the soft band, the authors attribute the systematically longer CDTs to a mixture of effects: intrinsic variability of an extended warm Comptonizing region responsible for the soft X-ray excess, delayed reprocessing/irradiation, and direct contamination by the low-energy extension of the coronal power law. They explicitly caution that the soft-band CDT should be regarded as an effective timescale rather than a uniquely local physical one — an honest limitation on any geometric inference drawn from the soft band alone.
Limitations and open questions
Several caveats bear directly on the strength of the conclusions. First, the evolutionary sequence itself is hypothetical, adopted from prior work rather than independently established here; the loop morphology depends on this assumed ordering. Second, the hard-band CDT uncertainties are large enough that only 90.5% of Monte Carlo realizations preserve the counterclockwise orientation, so the hard-band loop result, while preferred, is the least secure of the three statistically. Third, the radial-scale inference depends on the assumed viscosity parameter, scale height, and black hole mass, none of which are tightly constrained for this source; different prescriptions yield radii spanning nearly two orders of magnitude. Fourth, the identification of the soft-band CDT with any single physical radius is precluded by its composite origin. Finally, the physical mechanism producing the relative phase offsets among the four observables — dynamical, stochastic, and radiative responses to a common structural cycle — remains qualitatively described but not modeled quantitatively. Whether comparable CDT–lag loops exist in other AGNs with recurrent QPO candidates is left unaddressed, as is a quantitative propagation-model fit to the multidimensional evolutionary cycle.
Conclusion
By fitting DRW models to 17 multi-epoch XMM-Newton observations of RE J1034+396, this work demonstrates that the hard-band characteristic damping timescale coevolves with the X-ray time lag in a counterclockwise closed loop closely resembling the known QPO-frequency–time-lag loop, with all three relations shown to be robust against measurement uncertainties via Monte Carlo signed-area analysis. Combined with the consistency between CDT-inferred radii and independent coronal-size constraints, the result provides the first direct observational support among AGNs for the Ingram & Done framework, in which changes in the spatial extent of the hot inner flow/corona simultaneously govern QPO and stochastic variability. The extension of this coupled-evolution picture from stellar-mass to supermassive black hole systems reinforces the scale-invariance of inner accretion-flow physics, while leaving open the quantitative modeling of the phase offsets among the dynamical, stochastic, and radiative observables.