X-ray Reverberation in Black Hole Environments

• X-ray reverberation is the delayed echo of variable X-ray emission reprocessed by matter near black holes, revealing inner disc structures and coronal geometry.
• It employs Fourier spectral timing techniques to measure energy-dependent lags that map the geometry, mass, and spin of the accreting black hole.
• Observations in AGN and X-ray binaries show reverberation lags correlate with black-hole mass and disc truncation, offering constraints on inner accretion physics.

X-ray reverberation is the light-travel-time delayed echo of variable coronal X-ray emission after reprocessing by matter in the immediate black-hole environment, most commonly the inner accretion disc and, in some regimes, a wind funnel or warm Comptonizing atmosphere. Because the direct continuum and the reprocessed spectrum have distinct energy dependence, the delay can be measured as a function of Fourier frequency and photon energy, thereby mapping the inner accretion flow on scales of a few gravitational radii. In accreting supermassive and stellar-mass black holes, reverberation has become a central spectral–timing probe of coronal geometry, disc truncation, black-hole mass and spin, and the coupling between rapid variability and strong-gravity transfer effects (Uttley et al., 2014).

1. Physical basis and characteristic scales

The standard picture begins with a compact, hot corona that produces a variable hard X-ray power-law continuum, usually attributed to Comptonization of seed disc photons. A fraction of these photons reaches the observer directly, while another fraction irradiates optically thick matter and is reprocessed into a reflection spectrum containing a soft excess below about $1\ \mathrm{keV}$, a broad Fe K$\alpha$ feature around $6$–$7\ \mathrm{keV}$, and a Compton hump at $\sim 20$–$30\ \mathrm{keV}$. The reflected signal arrives later because the photon path is longer and because general-relativistic effects alter both travel times and observed energies (Marco et al., 2018).

The natural scales are the gravitational radius and light-crossing time,

$r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$

If the relevant geometry is expressed in units of $r_g$, then observed delays in seconds scale linearly with $M$. In a lamppost-like configuration, with the corona at height $h$ above the black hole, a characteristic reverberation lag obeys schematically $\alpha$0, up to corrections from light bending, Shapiro delay, and disc response smearing (Alston et al., 2020).

The inner disc radius is often associated with the innermost stable circular orbit, $\alpha$1, so spin enters timing through geometry as well as spectroscopy. A rapidly spinning black hole pushes the disc inward to smaller radii, enhancing Doppler and gravitational shifts, strengthening the red wing of the Fe line, and shortening intrinsic reverberation delays for a fixed coronal height (Caballero-Garcia et al., 2017).

A related but distinct regime appears in super-Eddington accretion. There, GRRMHD-based reverberation models predict that hard X-rays are reflected predominantly by the optically thick wind funnel rather than by a thin disc, producing a compact delay distribution, strong blueshifts, and lag spectra that differ qualitatively from thin-disc lamppost expectations (Thomsen et al., 2021).

2. Timing observables and spectral–timing formalism

The central object is the transfer function or impulse response $\alpha$2, which gives the energy- and time-dependent reflected response to a coronal flash. In its convolution form,

$\alpha$3

where $\alpha$4 is the driving continuum. $\alpha$5 encodes geometry, relativistic energy shifts, path-length delays, and the spatial distribution of ionization across the reflector (Wilkins, 2019).

In Fourier space, reverberation is measured with the cross-spectrum,

$\alpha$6

from which one defines the phase lag

$\alpha$7

and time lag

$\alpha$8

The coherence,

$\alpha$9

quantifies the linearly correlated fraction of variability between bands after appropriate Poisson-noise treatment (Uttley et al., 2014).

Two derived observables dominate the field. Lag–frequency spectra isolate reverberation at high frequencies, where light-travel delays dominate, and distinguish it from low-frequency “hard lags” usually attributed to inward propagation of accretion-rate fluctuations through a spectrally stratified flow. Lag–energy spectra, computed in a selected frequency interval, reveal which spectral components lag: the continuum-dominated $6$0–$6$1 region generally responds earliest, while the soft excess and Fe K bands respond later. Within the Fe K line itself, the red wing can respond before the line core, reflecting smaller emitting radii (Uttley et al., 2014).

A crucial correction is dilution. Every band typically mixes prompt continuum and delayed reflection, so the observed lag underestimates the intrinsic light-travel delay. A common approximation is

$6$2

or, in terms of reflected-to-direct flux ratio $6$3, $6$4 (Uttley et al., 2014).

Methodologically, Fourier analysis has been extended in two notable directions. Gaussian-process spectral timing reconstructs gapped light curves by sampling continuous realizations consistent with a stationary covariance kernel, enabling low-frequency reverberation studies even for low-Earth-orbit sampling. Wavelet spectral timing relaxes stationarity by localizing coherence and lag in both time and scale, making transient reverberation and time-variable contamination directly observable (Wilkins, 2019).

3. Geometrical models and transfer-function families

The dominant modeling baseline is the relativistic lamppost: a point-like corona on the spin axis above a thin, ionized, Keplerian disc. In this geometry, smaller $6$5 increases light bending, steepens the emissivity profile, enhances the reflection fraction, and shifts reverberation to shorter delays and higher Fourier frequencies. Fully relativistic fitting packages implementing this framework include KYNREFREV, KYNREVERB, and RELTRANS, which combine Kerr ray tracing with ionized reflection tables and predict lag–frequency, lag–energy, and cross-spectral products (Caballero-Garcia et al., 2017).

These models also predict timing diagnostics beyond lags. In PSD-based reverberation theory, the observed power spectrum is the intrinsic PSD multiplied by the squared modulus of the transfer function. A high-frequency dip followed by damped oscillations is expected, with dip frequency set mainly by $6$6 and $6$7, largely independent of energy, while dip depth increases with reflection fraction and can be enhanced by high spin and inclination for $6$8 (Papadakis et al., 2016).

Several departures from the razor-thin, point-source lamppost are now important. Finite disc thickness introduces self-shielding: the convex inner disc can suppress late-time outer-disc response, truncate the transfer-function “blue wing,” reduce reverberation amplitudes, and bias inferred heights downward if ignored. These effects are strongest for low $6$9, larger $7\ \mathrm{keV}$0, higher inclination, and lower spin (Taylor et al., 2018).

Extended-corona models were developed partly because simple lamppost fits leave high-frequency “wavy” residuals in some AGN lag spectra. Radially extended, low-height coronae with inward viscous propagation reproduce low-frequency hard lags, while vertically extended central structures with slow upward propagation can generate the pronounced $7\ \mathrm{keV}$1 dip seen in high-frequency lag–energy spectra. A two-axial-point-source approximation can reproduce oscillatory lag–frequency structure that a single source cannot (Wilkins et al., 2016).

In super-Eddington flows, the relevant geometry changes more radically. Reverberation off the wind funnel wall yields a step-function-like decline in lag–frequency spectra near the first zero crossing, and lag–energy shapes that remain almost independent of frequency band over a broad parameter range. Applied to Swift J1644+57, this framework favors $7\ \mathrm{keV}$2–$7\ \mathrm{keV}$3 and $7\ \mathrm{keV}$4–$7\ \mathrm{keV}$5, and is slightly preferred over a thin-disc lamppost (Thomsen et al., 2021).

At still smaller radii, reverberation has been proposed as a probe of the plunging region between the ISCO and the horizon. In these calculations, plunging-region signatures are weak in time-averaged spectra but appear in the highest Fourier frequencies, above $7\ \mathrm{keV}$6, in the $7\ \mathrm{keV}$7–$7\ \mathrm{keV}$8 band for a non-spinning black hole or $7\ \mathrm{keV}$9–$\sim 20$0 for near-maximal spin. Detecting them would require lag accuracies of order $\sim 20$1, with $\sim 20$2 needed to distinguish plunging from artificially extended circular motion (Wilkins et al., 2020).

4. Empirical landscape across source classes

In radio-quiet Seyferts, high-frequency soft and Fe K reverberation lags are now a common observational outcome. A systematic XMM-Newton archive study of 43 Seyferts found iron K reverberation in 20 sources, while low-frequency hard lags were present in 29 of the 35 sources where reverberation did not dominate the lowest frequencies probed. The iron K lag amplitude correlates positively with black-hole mass, and the characteristic reverberation frequency decreases with mass, implying compact coronae and reflection within $\sim 20$3 before dilution correction (Kara et al., 2016).

The source 1H 0707−495 provided the first robust soft reverberation lag, with a $\sim 20$4 delay at $\sim 20$5, while NGC 4151 provided the first clear Fe K reverberation lag, with the red wing responding before the near-rest-energy core, directly supporting a radius-ordered response across the line profile (Uttley et al., 2014).

In black-hole X-ray binaries, the same phenomenology is compressed by mass scaling. High-frequency thermal reverberation lags of a few milliseconds have been detected in hard-state systems such as GX 339−4, while low-frequency hard lags remain associated with propagating fluctuations. The lag amplitudes correspond to distances of tens of $\sim 20$6, suggesting a geometry different from that inferred in Seyferts, often framed as disc truncation plus an inner hot flow (Marco et al., 2018).

Cygnus X-1 has become the benchmark stellar-mass case for reverberation parameter estimation. Joint fits of the time-averaged spectrum and the complex cross-spectrum with RELTRANS yielded the first timing-based stellar-mass black-hole mass measurement, $\sim 20$7, consistent with the dynamical value when physically plausible ionization constraints are imposed. Updated distance-sensitive reverberation fits produced $\sim 20$8, $\sim 20$9 in one emissivity prescription and $30\ \mathrm{keV}$0, $30\ \mathrm{keV}$1 in another, demonstrating that current uncertainties are dominated by coronal angular-emissivity assumptions rather than by counting statistics alone (Mastroserio et al., 2019).

Reverberation is not confined to radio-quiet AGN. In the radio-loud broad-line radio galaxy 3C 120, a Gaussian-process reconstruction of gapped XMM-Newton light curves revealed low-frequency reverberation in the $30\ \mathrm{keV}$2–$30\ \mathrm{keV}$3 range. The Fe K core lagged the continuum by $30\ \mathrm{keV}$4, implying a lamppost height of $30\ \mathrm{keV}$5 for $30\ \mathrm{keV}$6 (Wilkins, 2019).

5. Dynamic reverberation and non-stationary inner flows

A major recent development is the treatment of reverberation as time dependent rather than stationary. In IRAS 13224−3809, 2 Ms of XMM-Newton data split into 16 epochs were fitted simultaneously with a fully relativistic transfer-function model, tying global parameters while allowing the coronal height to vary. This breaks the usual $30\ \mathrm{keV}$7–$30\ \mathrm{keV}$8–$30\ \mathrm{keV}$9 degeneracy present in single-epoch fits and yields $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$0, $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$1, $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$2, and $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$3. The corona rises from $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$4 at low flux to $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$5 at high flux, with a linear height–luminosity correlation inconsistent with constancy at $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$6 (Alston et al., 2020).

Wavelet spectral timing has shown that reverberation can be intermittent. In IRAS 13224−3809, within the reverberation frequency range, the iron K band lags the continuum only 49% of the time, leads it 32% of the time, and is incoherent 19% of the time; in MCG−6−30−15 the corresponding fractions are 23%, 34%, and 43%. Time-filtered covariance spectra indicate that ultrafast outflow variability can dominate the iron band and temporarily erase the reverberation signature, explaining previous non-detections in time-averaged Fourier analyses (Wilkins, 2023).

The same wavelet framework tracks short-timescale coronal evolution. In I Zw 1, the soft lag during bright flares rose from an upper limit of $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$7 at flare onset to $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$8, which, after dilution correction and GR height conversion, corresponds to coronal height increasing from $r_g = \frac{GM}{c^2}, \qquad t_g = \frac{GM}{c^3}.$9 to $r_g$0. In IRAS 13224−3809, soft lags exhibit counter-clockwise hysteresis with count rate, and the coronal expansion lags the luminosity rise by $r_g$1–$r_g$2, consistent with viscous propagation through the inner disc (Wilkins, 2023).

Ark 564 adds a further layer: multi-epoch XMM-Newton and NuSTAR analysis shows that the soft-band reverberation lag grows with flux, whereas Fe K lags are detected only intermittently and do not correlate with the soft lag amplitude. Low-flux epochs are broadly described by a lamppost plus thin-disc reflection model, but high-flux epochs require additional reverberation from a warm Comptonizing atmosphere. The preferred interpretation invokes a puffed-up inner disc and a vertically extended corona whose higher-flux configuration illuminates the warm atmosphere more effectively, thereby enhancing soft reverberation without a corresponding Fe K increase (Yu et al., 22 Jul 2025).

6. Interpretive debates, limitations, and future directions

Despite the success of inner-disc reflection models, the interpretation of X-ray lags has not been entirely uncontested. Analyses of 1H 0707−495 and other NLS1s argued that the full lag spectrum, including sign changes, can be explained by reverberation from partially covering circumnuclear scattering material at tens to hundreds of $r_g$3, rather than by reflection from $r_g$4. In that view, negative lags arise from Fourier phase wrapping in mixed direct-plus-delayed signals, and the data favor a high-covering-fraction scattering medium rather than an extreme inner-disc reflector (Miller et al., 2010). A related NLS1 study reported that the reverberating region lies a few tens to a few hundreds of $r_g$5 from the black hole and that the energy dependence disfavors the $r_g$6 inner-reflection interpretation for 1H 0707−495 (Miller et al., 2011). These results remain part of the field’s methodological and physical debate.

Even within the relativistic reflection framework, several assumptions are recurrent sources of systematics. Common reverberation models idealize the corona as a point-like, stationary lamp-post; assume a thin disc; neglect returning radiation or multiple scattering in winds; and approximate low-frequency intrinsic lags phenomenologically. Finite disc thickness, radial density structure, warm-absorber variability, ultrafast outflows, and non-stationary coronal geometry can all reshape transfer functions and bias inferred $r_g$7, $r_g$8, $r_g$9, or $M$0 if omitted (Taylor et al., 2018). This suggests that reverberation-derived black-hole parameters are only as robust as the assumed geometry and radiative transfer.

Future progress is expected along both instrumental and modeling fronts. A dedicated XMM-Newton legacy program with long exposures of order $M$1–$M$2 was proposed specifically to recover detailed transfer functions, emissivity profiles, and coronal dynamics rather than single effective lags; the proposed scale was about $M$3 of XMM-Newton time, or $M$4 over five years (Fabian et al., 2016). Wavelet-localized timing, higher-throughput missions, and extended-corona transfer-function models are expected to sharpen measurements of transient reverberation, absorber contamination, and disc–corona hysteresis (Wilkins, 2023).

The scope of reverberation is also broadening beyond X-ray reflection lags in Seyferts. In luminous quasars, analytical X-ray thermal reverberation models for UV/optical continuum lags can reproduce the mean lag–wavelength relation provided the corona lies at $M$5, though the inferred height depends on the assumption of a flat disc (Langis et al., 2024). This suggests that “X-ray reverberation” now names a family of causally related transfer problems spanning inner-disc reflection, thermal reprocessing, wind-funnel reflection, and possibly emission from the plunging region.

In that wider sense, X-ray reverberation is no longer merely the detection of a delayed echo. It is a relativistic transfer-function framework for inferring the geometry, dynamics, and radiative coupling of the innermost accretion flow—provided that dilution, propagation lags, absorption, and source non-stationarity are modeled with commensurate care.