WorksPhysics and Astronomy
Unveiling Hidden X-ray Broad Line Region in Centaurus A: Broadband Analysis with XRISM and NuSTAR
Abstract
We present a broadband X-ray spectral analysis of the radio galaxy Centaurus~A, utilizing simultaneous XRISM (Resolve and Xtend) and NuSTAR observations. Unlike the previous phenomenological analysis of the Fe K line (Bogensberger et al. 2025), we employ physical reflection models developed with the radiative transfer code SKIRT. They self-consistently predict the continuum and fluorescent lines from a clumpy torus and an X-ray broad line region (xBLR). The Fe K complex is well reproduced with a three velocity component model. We interpret this as two narrow torus components (FWHM and ~km~s) and an xBLR component (FWHM ~km~s). The inner torus radius of 0.13~pc is 10 times the dust sublimation radius (0.013~pc), implying that the torus in a low-luminosity AGN is not determined by dust-related physics. The xBLR has a substantial covering factor of 0.11–0.36, a hydrogen column density (~cm), a radius of 0.014~pc from the SMBH, and a mass of . Its mass supply rate is comparable to the accretion rate inferred from the bolometric luminosity, implying efficient xBLR formation even at low Eddington ratios. The xBLR line width also provides a novel estimate of the black-hole mass, , via the optical BLR mass-scaling relation to the Fe~K. This agrees with dynamical measurements within a factor of , consistent with the scatter of the optical relation. We also infer the abundance ratios of and solar.
Email: [email protected], [email protected]
Introduction
A tight correlation exists between the mass of the central supermassive black hole (SMBH) and the properties of the host galaxy’s bulge component, suggesting a fundamental link between these two components (e.g. J. Magorrian et al. [49]; J. Kormendy & L. C. Ho [42]). This implies a co-evolutionary scenario driven by feeding and feedback processes, bridging spatial scales that differ by more than nine orders of magnitude (see A. C. Fabian [21] for a review). In the context of feedback from the SMBH to the host galaxy, the Active Galactic Nucleus (AGN) phase, during which the SMBH actively accretes mass, plays a crucial role. Observationally, AGNs are classified into two types: type-1 AGNs, which exhibit broad emission lines with a Full Width at Half Maximum (FWHM) of typically a few thousand to ten thousand km s in the UV, optical, or infrared bands and type-2 AGNs, which lack these broad features. The unified model, in which the difference between type-1 and type-2 AGNs is attributed to the viewing angle of the observer relative to a dusty torus surrounding the central engine, has been successful in explaining this dichotomy (R. Antonucci [4]). However, a complete view of the matter distribution and kinematics of the central engine (accretion disk, broad line region (BLR), and torus) has remained elusive due to observational biases.
AGN feedback mainly operates through jets, winds, and radiation (e.g. A. King & K. Pounds [41]). Radio galaxies (RGs), characterized by powerful relativistic jets, are essential for studying AGN feedback by jets. These jets propagate at relativistic speeds and interact with the interstellar medium (ISM), thereby regulating star formation in the host galaxy. Although jet launching and feedback efficiency are thought to be linked to the central accretion flow (R. D. Blandford & R. L. Znajek [11]), the detailed mechanisms and their connection to the subparsec-scale nuclear structure remain poorly understood.
X-ray observations are essential for probing the nuclear structure of heavily obscured AGNs. X-rays penetrate the dense absorbing material (the column density along the line of sight with ) to reveal the nuclear structure and the circumnuclear gas and dust across diverse thermal phases. Historically, detailed spectral decomposition has been hindered by the limited energy resolution of X-ray instruments (e.g., for XMM-Newton/EPIC-pn, and – for Chandra/HETG at 6 keV). XRISM (M. Tashiro et al. [79, 80]), launched in 2023 by JAXA, carries a microcalorimeter (Resolve) with superior energy resolution (FWHM 4–5 eV, or –1500 at 6 keV). Resolve allows us to spectroscopically resolve the velocity fields and physical conditions of the gas on spatial scales that are not accessible to imaging. For example, observations of the Circinus Galaxy (type-2) successfully constrained the chemical abundances in the torus, revealing its material origin (XRISM Collaboration et al. [92]). Furthermore, in the complex type-1.5 AGN NGC 4151, which exhibits broad optical lines but significant X-ray absorption, identified a distinct X-ray broad line region (xBLR), providing new constraints on the nuclear geometry (Xrism Collaboration et al. [90]). Subsequent studies of NGC 4151 have revealed stratified disk winds and the reverberation of the narrow Fe K line (e.g., X. Xiang et al. [89]; J. M. Miller et al. [56]).
Our target, Centaurus A (hereafter Cen A), is one of the nearest RGs located at a distance of Mpc (G. L. H. Harris et al. [30]), classified as a heavily obscured () type-2 AGN (V. Beckmann et al. [9]). Due to its strong X-ray and radio flux, Cen A has been extensively studied as a key object for understanding AGN physics. The mass of the central SMBH () has been measured with several kinematic methods. Using VLT/SINFONI, M. Cappellari et al. [15] and N. Neumayer et al. [67] derived and (errors are ) from modeling of stellar kinematics and of molecular gas, respectively, for a distance of 3.5 Mpc. Since the two measurements based on different tracers agree with each other, we adopt (their logarithmic mean) in our paper, which is also consistent with the recent result by A. Gernucci et al. (2011). We note earlier kinematic studies reported somewhat larger masses: based on stellar kinematics, J. D. Silge et al. [75]; from [S III], Pa and [Fe II] rotation curves, A. Marconi et al. [52]; from the two-dimensional Pa velocity field modeling, D. Krajnović et al. [45], which range from to 8.38 for a distance of 3.5 Mpc. The adopted mass () corresponds to a gravitational radius of about cm. The bolometric luminosity is estimated to be [44], resulting in a low Eddington ratio of .
The classification of Cen A as an RG does not limit the generality of the results presented here. Y. Nakatani et al. [63] analyzed the broad-band Suzaku and NuSTAR spectra of seven obscured narrow-line radio galaxies, including Cen A [69]. They used the XCLUMPY model [77], which describes the reflection from a clumpy torus based on Monte Carlo simulations. They found that the torus covering fractions of these radio galaxies are similar to those of radio-quiet AGNs at the same Eddington ratios. This indicates that the geometric structure of the torus is primarily determined by the Eddington ratio, regardless of the presence of a radio jet. Cen A, despite being an RG, can thus be regarded as a representative obscured AGN. Despite extensive searches in optical, infrared, and polarized light, no broad emission lines have been found [2, 53, 26]. For example, D. M. Alexander et al. [2] obtained near-infrared (K-band) spectropolarimetry of the nuclear region and searched for broad Br. They detected none in either total or polarized flux. Their result was inconclusive, in the sense that a completely hidden BLR could not be distinguished from the intrinsic absence of one. More recently, JWST observations of the nucleus have also shown no broad Pa component [26]. The visual extinction toward the near-infrared nucleus is estimated to be – mag from the 9.7 m silicate absorption feature [28, 82]. Whether the BLR in Cen A is deeply hidden or intrinsically absent has thus remained an open question. On larger scales, ALMA has resolved the circumnuclear molecular gas of Cen A. D. Espada et al. [19] imaged the circumnuclear disk (200 pc) in CO(3–2) at a resolution of pc. They found an inner molecular ring with a deprojected radius of 50–100 pc, along with filamentary structures interpreted as nuclear shocks. These structures may facilitate accretion onto the SMBH. On sub-parsec scales, mid-infrared interferometry with VLTI/MIDI resolved a dusty disk with a diameter of pc [54]. However, none of these observations constrain the kinematics of the obscuring and line-emitting material on sub-parsec scales. Such kinematic information is accessible only through high-resolution X-ray spectroscopy. In the X-ray band, Cen A has been observed by various X-ray satellites, revealing its detailed spectral characteristics. Its X-ray spectrum is characterized by strong absorption and a prominent Fe K fluorescence line (intensity of the 6.4 keV line photons cm s; [23]). The reflection features of Cen A have been described by several torus models, including MYTorus [58, 24], BNTorus [14, 24], and XCLUMPY [77, 69]. These studies suggest that the obscuring torus is characterized by an angular width of 50–60 deg (corresponding to a covering factor of ) and an inclination angle of 50–75 deg. We note that previous estimates of the system’s inclination range from edge-on (50–80 deg; [24, 69]) to face-on ( deg; [29, 12] (C. Müller et al. 2014)).
Recent high-resolution spectroscopy with XRISM has revealed a broad iron line feature in addition to the dominant narrow Fe K line core, suggesting the presence of a hidden BLR in Cen A [13]. However, a definitive physical interpretation is challenged by the spectral degeneracy between the line profile and the underlying reflection continuum. To overcome this, we perform a joint analysis using simultaneous NuSTAR data, which covers the hard band ( keV) essential for constraining the Compton hump. We employ physical reflection models based on the SKIRT radiative transfer code [6, 84] to distinguish between BLR and torus contributions. In the latest SKIRT code, the realistic line profiles measured in ground experiments [32] have been incorporated [85], which are known to well reproduce the neutral Fe K profile in a type-2 AGN [92].
The organization of this paper is as follows. In § 2 we describe the observations of XRISM and NuSTAR and data reduction. § 3 presents the details of our spectral analysis and results. The nuclear structure of Cen A is discussed in § 4, including the physical origin of the xBLR. Conclusions are summarized in § 5. Throughout the paper, errors in spectral parameters are quoted at the 90% confidence limit for a single parameter. The solar abundances refer to the protosolar table of K. Lodders et al. [48].
Observation and Data Reduction
The X-ray data of Cen A used in this study were obtained from two observatories, XRISM and NuSTAR, which provide complementary capabilities in terms of energy coverage and spectral resolution. XRISM/Resolve offers high energy resolution in the soft X-ray band (2–12 keV). XRISM/Xtend is an X-ray CCD camera with a larger effective area in the soft X-ray band (0.3–12 keV). NuSTAR provides broad energy coverage up to 79 keV with moderate energy resolution. The details of the observations and data reduction processes for each instrument are described in the following subsections. The observation logs are summarized in Table 1.
| Observatory | ObsID | Start Date (UT) | End Date (UT) | Exposure (ks) | Energy Range (keV) |
| XRISM | 300019010 | 2024 Aug 04 19:17:04 | 2024 Aug 09 16:24:04 | 225 | 3–10 |
| NuSTAR | 60901034002 | 2024 Aug 05 07:01:09 | 2024 Aug 05 19:26:09 | 25 | 8–75 |
| 60901034004 | 2024 Aug 09 02:16:09 | 2024 Aug 10 19:21:09 | 77 | 8–75 |
Table 1. Summary of observations. Exposures are based on the good time intervals of Resolve for XRISM and those of FPMA for NuSTAR.
XRISM
XRISM carries two focal plane instruments, each coupled with an X-ray Mirror Assembly (T. Hayashi et al. [31]): the high-energy-resolution X-ray microcalorimeter Resolve (Y. Ishisaki et al. [34]; R. L. Kelley et al. [40]) and the X-ray CCD camera Xtend (H. Noda et al. [68]). XRISM observed Cen A in Aug 2024, for 225 ks (Table 1). We use both the Resolve and Xtend data in this work. The Resolve data provide the line profiles at high energy resolution. The Xtend data provide an independent measurement of the continuum shape in the soft X-ray band. Cross-calibration uncertainties among the instruments are taken into account by free multiplicative constants (§ 3.2). The Resolve and Xtend data were reduced following the XRISM Quick-Start Guide version 3.2 2. The same observation was first analyzed by D. Bogensberger et al. [13]. We utilized the XRISM calibration database (CalDB) version 20260315 for both Resolve and Xtend. For this observation, the gain uncertainty is estimated to be less than 1 eV at 6.4 keV. The Resolve spectrum was binned according to the optimal binning scheme by J. S. Kaastra & J. A. M. Bleeker [37] through the ftgrouppha task, ensuring an appropriate sampling of the instrumental response. In the 3–10 keV band, the source count rate ( counts s) exceeds the estimated non-X-ray background (NXB) rate ( counts s) by a factor of (i.e., the NXB contributes of the total). Therefore, following T. Kayanoki et al. [39], the Resolve background spectrum was not subtracted. The light curve of Resolve in the 2–10 keV band is plotted in the upper panel of Figure 1. The Xtend data reduction was performed using the xapipeline task. The Xtend data were taken with CCD1 and CCD2 in the 1/8 window mode. We extracted the Xtend source spectrum from a rectangular region of 126 arcseconds 360 arcseconds centered on the source position. The source extraction region is contained within a single CCD chip. The Xtend background spectrum was extracted from a source-free rectangular region of 90 arcseconds 432 arcseconds on the other chip. The redistribution matrix file was generated with the xtrdmf task. The exposure map was generated with the xaexpmap task. The ancillary response file for a point source was generated with the xaarfgen task. The number of photons per raytraced energy in xaarfgen (numphotons) was set to . The Xtend spectrum was binned to a minimum of 50 counts per bin using the grppha tool. The effective exposure time of Xtend is 183 ks. The light curve of Xtend in the 0.5–10 keV band is plotted in the middle panel of Figure 1.
Figure 1. (Upper panel) XRISM/Resolve light curve in the 2–10 keV band. (Middle panel) XRISM/Xtend light curve in the 0.5–10 keV band. (Lower panel) NuSTAR/FPMA light curves in the 3–78 keV band for the two observations, ObsID 60901034002 (blue) and ObsID 60901034004 (red). The time bin is 1200 s for XRISM/Resolve and NuSTAR, and 1280 s for XRISM/Xtend. The light curves for the FPMB detector are consistent with those of FPMA but are omitted for clarity.
NuSTAR
To constrain the high-energy X-ray continuum and reflection components, we also utilized a set of two NuSTAR observations performed simultaneously with the XRISM observation (25 and 77 ks, respectively; Table 1). The data were processed using HEASoft version 6.37.1 with the CalDB version 20260923. We extracted the source spectra from circular regions with radii of 120 arcseconds centered on the source position for both the FPMA and FPMB detectors. The background spectra were extracted from source-free circular regions with radii of 120 arcseconds on the same detectors. We generated the source and background spectra, along with the corresponding response files, using the nuproducts script. In order to improve the signal-to-noise ratio, the spectra from the two observations were combined separately for FPMA and FPMB using the addascaspec tool. The combined NuSTAR spectrum was binned to have a minimum of 600 counts per bin using the grppha tool. We first extended the NuSTAR energy range up to 79 keV. However, the bin containing 79 keV does not have enough statistics (a minimum of 600 counts per bin), and thus it is not used. As a result, we use the NuSTAR data up to keV, where the source is still detected above the background. Conservatively, we do not use the NuSTAR data below 8 keV, given the possible cross-calibration uncertainties in the energy-dependent effective area relative to XRISM/Resolve at these energies. The light curve of NuSTAR/FPMA in the 3–78 keV band for each observation is plotted in the lower panel of Figure 1, illustrating the overlap with the XRISM observing window.
X-ray Spectral Analysis
In this section, we present our spectral analysis of the simultaneously observed data of XRISM/Resolve, XRISM/Xtend, and NuSTAR. To model the spectra, we consider two distinct structures in the circumnuclear region: a clumpy torus and a uniform disk-like xBLR in the equatorial plane, where the inner xBLR extends up to the torus. First, we describe the simulations using the SKIRT code to create a torus reflection model tailored for Cen A (§ 3.1.1), and a general xBLR reflection model (§ 3.1.2). Next, we explain the overall spectral model used in our analysis (§ 3.2). Finally, we present the fitting results with a primary focus on the modeling of the Fe K line profile (§ 3.3). All fits were performed by minimizing the Cash statistic (C-stat: [16]) using the XSPEC package ([5]) version 13.0.1. We simultaneously fitted the XRISM/Resolve spectrum in the 3–10 keV band, XRISM/Xtend spectrum in the 3–6.0 keV and 7.7–10 keV bands, and the NuSTAR/FPMA and NuSTAR/FPMB spectra in the 8–75 keV band. For NuSTAR data, we set the upper limit of the energy range to 75 keV because the bin containing 79 keV does not have enough statistics. Here, we conservatively exclude the 6.0–7.7 keV band of the Xtend spectrum, which contains the Fe K, Fe K, and Ni K lines. The line profiles in this band are determined by the Resolve data. The Xtend data are used to constrain the continuum shape. To account for possible cross-calibration uncertainties among the instruments, we introduce multiplicative constant factors for the XRISM/Xtend and NuSTAR/FPMA and FPMB spectra relative to the XRISM/Resolve spectrum, as detailed in § 3.2.
Before presenting the detailed spectral model, we first illustrate its basic features by fitting the broadband spectrum with a simple absorbed cutoff power-law model. Figure 2 shows the observed broadband spectrum and the ratio of the observed spectrum to this simple model. Two features stand out: a prominent Fe K emission line, and a weak Compton reflection hump around 20–30 keV. The soft excess below ∼ 4 keV is attributable to the too simplified modeling, which will be reduced in the final model presented below (Figure 4); the slight difference in the ratios between Resolve and Xtend is due to the cross calibration uncertainties at the lowest energies ([91]). The strong Fe K line and the weak Compton reflection hump, which have long been known for Cen A, motivate the detailed reflection modeling described in the following subsections.
Figure 2. (Upper panel) Observed X-ray spectra folded with the energy responses (but corrected for effective area). Black, blue, red, and orange crosses represent the XRISM/Resolve, XRISM/Xtend, NuSTAR/FPMA, and NuSTAR/FPMB spectra, respectively. The model spectrum is shown as a green solid line. (Middle panel) Ratio of the Resolve spectrum to a simple absorbed cutoff power-law model. (Lower panel) Ratio of the Xtend and NuSTAR spectra to the same model. The model is fitted simultaneously to all the spectra. The Xtend data in the 6–7.7 keV band are not used for the fit. The data points are rebinned for display. In all panels, the vertical blue lines mark the redshifted () energies of the Ca K, Fe K, Fe K, and Ni K fluorescence lines.
Figure 4. (Upper panel) Observed X-ray spectra folded with the energy responses (but corrected for effective area): Black, red, blue, and orange crosses represent the XRISM/Resolve, XRISM/Xtend, NuSTAR/FPMA, and NuSTAR/FPMB spectra, respectively. The Xtend data in the 6–7.7 keV band are not used for the fit. The dashed lime line represents the best-fit “two torus” model for the broadband fitting. The other colored solid lines show the contributions of each model component, in the same way as in the right panel of Figure 3. The model curves are shown only for Resolve and NuSTAR. (Middle panel) Fitting residuals of the XRISM/Resolve spectrum in units of 1σ error. (Lower panel) Same as the middle panel, but for the XRISM/Xtend and NuSTAR (FPMA and FPMB) spectra. For display purposes, the data points in all panels are rebinned.
Physical Modeling of Reflection Components
We developed two radiative transfer based numerical models representing the reflection components from the torus and the xBLR. Since XRISM/Resolve enables us to resolve the detailed line profiles of fluorescent lines with unprecedented accuracy, it is necessary to calculate models including realistic line profiles with sufficiently fine energy grids that are compatible with the observed data. For physically self-consistent interpretation, it is critical to include the reflection continuum as well. To calculate the reflection spectra, we performed Monte Carlo simulations using the 3D radiative transfer code SKIRT ([6, 84]). In our simulations, we assumed that all material was neutral (i.e., not ionized). The physical processes considered in the simulations include photoelectric absorption, Compton scattering by bound electrons, and fluorescent line emission.
The two models were computed independently to minimize computational cost. This means that we neglect any potential interactions between the torus and the xBLR, such as corrections of the incident spectrum to the torus due to the absorption by the xBLR and multiple scattering between the torus and xBLR regions. We have confirmed that these effects are not significant in our case as the angular width of the xBLR is much smaller than that of the torus (see § 3.3). Because both the torus and xBLR models are scale-invariant (i.e., the results depend only on the relative geometry of the reflector, not on its absolute scale), we can set their sizes independently when we apply the models to the observed spectra. In the radiative transfer simulations, any velocity structure is ignored for simplicity. The Doppler broadening effects will be taken into account by convolving the resulting spectral models with Gaussian kernels.
Torus
We adopt a clumpy torus geometry of the CLUMPY model [65 b]. We follow its formulation in the XCLUMPY model [77 Section 2]. The clumpy structure is modeled by distributing spherical clumps randomly. Each clump has the same fixed size ( pc) and the same hydrogen number density (). The clumps are distributed according to a power-law with a radial density profile index (), and a Gaussian distribution with an angular width (). Therefore, in the spherical coordinate system centered at the SMBH, the number density function of clumps is given by [77 Equation (1]
where is the normalization factor and is the inner radius of the torus. The normalization is set by the mean number of clumps crossed by a straight path running outward from to in the equatorial plane (), which is given by
where is the outer radius of the torus. Thus, can be expressed as [77 Equation (2]
The hydrogen column density along the equatorial plane is related to as [77 Equation (5]
The term represents the average column density of a randomly located single clump along a line of sight through it.
We fix , pc, pc, and , following the XCLUMPY model [77]. These values correspond to the mean values obtained by K. Ichikawa et al. [33], who applied the CLUMPY model to the infrared spectral energy distributions of 21 nearby AGNs. Thus, the free geometrical parameters are and (or equivalently ). We treated the abundances of calcium (Ca), and nickel (Ni) as variable parameters in the table model. The abundances of the other elements were all fixed at the solar values [48], because they could not be constrained by our data due to the limitation of band coverage or insufficient photon statistics.
In our SKIRT simulations for the torus, we configured the spatial domain as a cube with sides of 40 pc. The computational grid utilizes an octree-based adaptive mesh, with subdivision levels of 7 to 15. This setup ensures that the smallest cell size is pc, allowing for sufficient spatial resolution to resolve each clump structure. The spectral energy grid spans 1.5–500 keV and was constructed as follows: from 1.5 keV to 10.0 keV, the grid is linearly divided with a step size of 0.5 eV, and from 10 keV to 500 keV, the grid is divided into 500 bins on a logarithmic scale. The energy resolution of the grid for the soft X-ray band is set to be higher than that of XRISM/Resolve to accurately reproduce the line profiles (matching the 0.5 eV grid of the XRISM/Resolve response matrix). The incident spectrum energy distribution is modeled with a power-law with an exponential cutoff: , where is the photon index and is the cutoff energy. For each simulation, photon packets were launched in total across the full energy range, with the energy of each packet sampled from this incident spectrum. The parameter grid for the torus model is summarized in Table 2 (note that the model is scale-free and absolute distances do not enter the results).
| Note | Parameter | Grid Value | Units |
| Torus (XCLUMPY) | |||
| (1) | 30, 35, 40, 45, 50, 55, 60, 65, 70 | deg | |
| (2) | 23.20, 23.30, 23.40, 23.50, 23.60, 23.70, 23.80, 23.90 | ||
| (3) | 1.750, 1.775, 1.800, 1.825, 1.850 | ||
| (4) | 0.70, 0.90, 1.10, 1.30, 1.50, 1.70, 1.90 | ||
| (5) | 0.90, 1.00, 1.05, 1.20, 1.35, 1.50, 1.65 | ||
| BLR (XBLR) | |||
| (6) | 6, 12, 18, 24, 30, 37, 45 | deg | |
| (7) | 1, 5, 10, 30, 50, 80, 100, 130, 150, 200, 500, 1000 | ||
| (8) | 1.750, 1.800, 1.850 | ||
| (9) | 0.70, 1.10, 1.90 | ||
| (10) | 0.90, 1.20, 1.65 | ||
| Common Parameters | |||
| (11) | 50 | deg | |
| (12) | 370 | keV | |
| (13) | 1.0 |
Table 2. Summary of Parameters. (1) Torus angular width. (2) Hydrogen column density along the equatorial plane. (3) Photon index. (4) Abundance of calcium relative to hydrogen. (5) Abundance of nickel relative to hydrogen. (6) BLR angular width. (7) Hydrogen column density in the radial direction. (8) Photon index. (9) Abundance of calcium relative to hydrogen. (10) Abundance of nickel relative to hydrogen. (11) Inclination angle. (12) Cutoff energy. (13) Abundance of iron relative to hydrogen.
xBLR
For the xBLR model, we assume a partially spherical geometry with two polar cuts between inner and outer radii, as adopted by M. Stalevski et al. [76] and G. L. Granato & L. Danese [27]. We consider a small angular width (i.e., small covering factor), assuming that the mass distribution of the xBLR is concentrated close to the equatorial plane [17]. The density distribution is expressed in spherical coordinates centered on the SMBH. Here, is the polar angle measured from the symmetry axis of the xBLR, and the density is axisymmetric. The equatorial plane corresponds to . The density is given by
where is the normalization factor, is the radial density profile index, is the angular density profile index, and are the inner and outer radii of the xBLR, respectively, and is the xBLR angular width (i.e., the half-opening angle of the xBLR from the equatorial plane). We normalize the density using the hydrogen column density in the radial direction, . This is the column density along the equatorial plane, measured on one side of the SMBH from to . We set , and assume the outer radius to be ten times the inner radius (). For simplicity, we assume a uniform density within the xBLR by setting and . Then, Equation (5) reduces to within the xBLR. We regard as the hydrogen number density , and hence . Because is constant, is given by
The normalization factor is therefore given by
The effects of clumpiness will be evaluated in Appendix § 7. The xBLR model has two free geometrical parameters: the hydrogen column density in the radial direction (), and the xBLR angular width (). The solid angle of the region is . For clarity, which is the fraction of the sky covered by the xBLR as seen from the central SMBH, is therefore . We note that this xBLR model is also a scale-free model. As in the torus model, the abundances of Ca, and Ni relative to H are set as variable parameters.
In our SKIRT simulations for the xBLR, we configured the spatial domain as a cylinder with a radius of 0.11 pc and a height of 0.22 pc, providing a small margin outside the physical outer radius () to avoid computational boundary artifacts. The spatial grid is a 2D cylindrical grid with 1000 radial bins and 1000 vertical bins. This setup ensures that the edge of the xBLR is well resolved. The number of photon packets launched in each simulation is . We used the same approach as with the torus model (§ 3.1.1) for the gas composition, the spectral energy grid, and the incident spectrum energy distribution. The parameter grid for the xBLR model is summarized in Table 2.
Spectral Model
Our model represents a typical X-ray spectrum of an obscured AGN, including the contribution from the xBLR, which has been mostly ignored in previous X-ray studies. It consists of a primary power-law component with an exponential cutoff from the central hot corona, its reflection component from the xBLR, the emission component from photoionized plasmas, and the reflection component from the torus. The first three components are subject to the absorption by the dusty torus along the line of sight.
We find that the most prominent observed spectral feature, Fe K line, is modelled with three components (see § 3.3): (1) the reflection from the xBLR that produces a broad line, (2) and (3) two narrow line components from the torus with different velocity widths (torus2 and torus1, respectively) To account for the kinematic broadening caused by the macroscopic motion of the gas (e.g., Keplerian rotation), we adopted a Gaussian smoothing kernel (gsmooth) to broaden the components, instead of applying a disk-line convolution kernel (such as rdblur). The use of a disk line kernel introduces degeneracy among the geometric parameters; specifically, the inner radius () is strongly coupled with the emissivity index and the outer radius (). Fixing either or both of these parameters would impose artificial biases in estimating , rendering its physical meaning ambiguous.
In the XSPEC terminology, our model is described as follows:
(1) The constant1 term represents an instrumental cross-calibration constant among different detectors. We set constant1 of XRISM/Resolve to unity as a reference. We allow those of XRISM/Xtend, NuSTAR/FPMA, and NuSTAR/FPMB to vary independently. The phabs term represents the Galactic absorption. We fixed its hydrogen column density at , the total value of and based on https://www.swift.ac.uk/analysis/nhtot/ ([88]).
(2) The zvphabs*cabs term represents the photoelectric absorption and the extinction due to Compton scattering by the line-of-sight material. Its hydrogen column density is calculated via Equation (9) (see item (6) below) assuming that it originates only from the clumpy torus. As described in § 3.1.2, the xBLR is assumed to be concentrated near the equatorial plane with an angular width smaller than , and therefore does not intercept the line of sight.
(3) The zcutoffpl term represents the primary power-law component from the hot corona. Since our fits do not find the cutoff energy to be below 79 keV, and due to a lack of data above this energy, we fixed it to a typical value of 370 keV ( keV: [72]). The free parameters of this component are the photon index () and the normalization ().
(4) The atable{PION_CenA.fits} terms represent the photoionized emission components. We adopted the same model as used in [39], which consists of three components with different ionization parameters, by converting their SPEX models into XSPEC table files. All parameters (hydrogen column density, ionization parameter, outflow velocity, and turbulent velocity) in each component and relative normalizations among the three components are fixed at the best-fit values reported in [39].
(5) The gsmooth1 * atable{xBLR_CenA_Resolve.fits} terms represent the reflection component from the xBLR (§ 3.1.2). The gsmooth1 term denotes the Doppler broadening effect. Its width at 6.0 keV () is set to be a free parameter, which is assumed to be proportional to the line center energy. The xBLR model parameters are: the input power-law photon index, the hydrogen column density in the radial direction (), the xBLR angular width (), the abundances of calcium (), and nickel (), the redshift, and the normalization. We limited the xBLR angular width to deg. For the assumed inclination of deg, the line of sight is deg above the equatorial plane. This limit therefore ensures that the line of sight does not cross the xBLR in our model. We note that, in this setting, is distinct from the line-of-sight column density () in item (2) above.
(6) The XCLUMPY table, atable{xclumpy_CenA_Resolve_Continuum.fits}, represents the reflection continuum without fluorescent lines from the torus (§ 3.1.1). The parameters of XCLUMPY are the input power-law photon index, the hydrogen column density along the equatorial plane (), the torus angular width (), the abundances of Ca () and Ni (), the redshift (), and the normalization. The hydrogen column density along the line of sight () in item (2) is related to the torus parameters as follows:
In practice, the number of clumps intersecting the line of sight can fluctuate stochastically. Equation (9) gives the expectation value of , which we adopt in this work.
(7) These two XCLUMPY tables, constant2*gsmooth2*atable{xclumpy_CenA_Resolve_Lines.fits} and (1−constant2)*gsmooth3*atable{xclumpy_CenA_Resolve_Lines.fits} represent the reflection fluorescent line components from the torus. As we will explain in § 3.3, we found that the velocity profile of Fe K is much better explained by two broadened components rather than a single one. The constant2 term, which we denote as (), represents the fraction of the most narrow component (torus1) among the total torus component. The gsmooth2 and gsmooth3 terms represent the Doppler broadening effects for the two components. Their widths () at 6.0 keV are allowed to vary with the physical constraint that , both proportional to the line center energies. The other parameters of these two XCLUMPY line components are the same as those of the XCLUMPY continuum component described in (6) above.
Here we summarize the details of our parameter settings. We fixed the inclination angle at 50 deg in all reflection models (items (5), (6), and (7)) based on broadband X-ray studies using the XCLUMPY model [69], in order to avoid degeneracy among torus geometrical parameters. Varying within the range reported by S. Ogawa et al. [69] does not significantly affect our results. Since the data do not well constrain the abundance ratio of iron with respect to hydrogen (, see § 4.3), we fixed the Fe abundance at the solar value. Fixing ties the absolute normalization of the reflection continuum to the fluorescent line fluxes at the solar iron abundance. Consequently, the Ca and Ni abundances relative to Fe are determined directly by the Ca K/Fe K and Ni K/Fe K line intensity ratios. These ratios are largely insensitive to the poorly constrained overall strength of the reflection continuum. Since is fixed at the solar value, the grid parameters and (defined relative to hydrogen in solar units; see Table 2) are numerically identical to the abundance ratios and discussed throughout the paper. The redshift and the abundances of Ca and Ni are set to be free but linked among the torus line-of-sight column density (item (2)) and all reflection models (items (5), (6), and (7)). The incident photon index, cutoff energy, and normalization in the reflection models (items (5), (6), and (7)) are linked to those of the direct component (zcutoffpl) in item (3).
Fe K Line Modeling
Firstly, we fitted the spectrum with a single velocity width for the torus component. We fixed constant2 = 1.0. In this modeling, we assumed that two reflection components, the torus and the xBLR, have different velocity widths. Their best-fit FWHMs were found to be km s and km s, for the torus and the xBLR components, respectively. These widths are consistent with those obtained by D. Bogensperger et al. (2025) from the same XRISM data. They found a FWHM of km s for the narrow cores and km s for the broad component. The fit statistic was for 3906 degrees of freedom (d.o.f.). However, the single velocity torus model failed to reproduce the resolved shallow peak feature of Fe K line, leaving residuals in the 6.4 keV region, as shown in Figure 3. To account for this, we introduced a second component for the torus reflection with a different velocity width.

Figure 3. (Left panel) Zoom-in on the Fe Kα line with the “one torus” model (constant2 = 1.0). Black crosses represent the XRISM/Resolve observed spectrum folded with the energy responses (but corrected for effective area). The dashed red line represents the best-fit model, and the colored solid lines show the contributions of each model component: direct component (orange), photoionized component (purple), reflection component from the xBLR (blue), reflection continuum from the torus (magenta), reflection emission lines from torus1 (green). (Right panel) The same plot but for the “two torus” model (constant2 is free). The solid red line represents the contribution of the second torus (torus2) component, which has a broader velocity width than the torus1 component. The bottom sub-panels for both models show the fitting residuals in units of , plotted on the same vertical scale and with the same binning. The dashed gray lines indicate levels.
Secondly, we set constant2 as a free parameter, allowing both constant2gsmooth2atable{xclumpy_CenA_Resolve_Lines.fits} and (1 − constant2) * gsmooth3 * atable{xclumpy_CenA_Resolve_Lines.fits} components to contribute to the torus reflection. Here, we note that the total intensity from the torus reflection is still conserved. This treatment of splitting the torus reflection into two velocity components is for testing whether a non-uniform torus provides a better description of the data. With this modeling, we obtained the best-fit FWHMs of , , and , for the two torus components and the xBLR component, respectively. The fit statistic improved from (3906 d.o.f.) to (3904 d.o.f.), corresponding to for two additional free parameters (). To evaluate the statistical significance of adding the second torus component, we calculated the Akaike Information Criterion (AIC; [1]) and the Bayesian Information Criterion (BIC). For the single-torus model, we obtained AIC = 4209.0 and BIC = 4303.1. With the two-torus model, these values improved to AIC = 4160.3 and BIC = 4267.0. The substantial reductions of AIC = −48.7 and BIC = −36.1 support the necessity of the dual velocity components. Furthermore, the systematic residuals seen around the Fe K line in the single-velocity model (Figure 3) are fully resolved by including this second torus component. To check the goodness of fit, we binned the Resolve and Xtend spectra to have at least 300 counts per bin and performed the fitting. The reduced chi-square value was , indicating a reasonable fit without significant systematic residual structures, given the high statistical quality of the data.
Thus, our final model includes three reflection components: two velocity components for the torus reflection component and one for the xBLR. The best-fit parameters are summarized in Table 3, and the fit results are shown in Figure 4.
| Model | Units | ||
| (1) | Abundance | ||
| (2) | zcutoffpl | ||
| ; see caption | |||
| (3) | xBLR | ; | |
| ; deg | |||
| ; eV at 6 keV | |||
| (4) | XCLUMPY | ; | |
| ; deg | |||
| ; | |||
| ; eV at 6 keV | |||
| ; eV at 6 keV | |||
| (5) | PION | ; see caption | |
| (6) | ; | ||
| (7) | constant | ||
| constant | |||
| constant | |||
| (8) | C-stat | 4126.31 | |
| d.o.f. | 3904 |
Table 3. Table 3. Best Fit Parameters. (1) Abundance ratios of calcium and nickel to iron. (2) Photon index of the direct component and the normalization in photons keV cm s at 1 keV. (3) Hydrogen column density of the xBLR in cm, the angular width of the xBLR in degree, and the velocity dispersion of the xBLR in eV. (4) Hydrogen column density along the equatorial plane in cm, torus angular width in deg, the fraction between the torus1 in the total torus component, hydrogen column density along the line of sight in cm, and the Doppler broadening sigma at 6.0 keV in eV for torus1 and torus2 components, respectively. (5) Intrinsic (absorption-corrected) photon flux of the PION component integrated over 6.0–8.0 keV, in units of photons cm s. (6) Redshift. (7) Instrumental cross-calibration constants, relative normalization of XRISM/Xtend to XRISM/Resolve, NuSTAR/FPMA to XRISM/Resolve, and NuSTAR/FPMB to XRISM/Resolve. (8) C-stat and degrees of freedom (d.o.f.) values of the best-fit models.
Discussion
Torus Structure of Cen A
We investigate the torus structure of Cen A, on the basis of our fitting results of the torus component modeled via XCLUMPY. Through the broad-band analysis utilizing the simultaneous XRISM/Resolve, XRISM/Xtend, and NuSTAR data, we obtained the torus angular width of and the hydrogen column density in the equatorial plane of for an assumed inclination of 50 deg. These values are consistent with the previous results obtained by S. Ogawa et al. [69] based on the simultaneous analysis of a NuSTAR spectrum and Suzaku/XIS ones.
Thus, the broadband continuum and high-resolution X-ray spectra point to a picture in which the nucleus in Cen A is surrounded by a torus that is geometrically extended (having a large solid angle) but globally Compton-thin. The estimated covering fraction of the torus with is close to unity, in agreement with the relation between the Eddington ratio and torus covering fraction reported by C. Ricci et al. [70]. The absence of broad line components in either the direct or polarized optical spectra of Cen A may be explained by this “deeply buried” nature of the nucleus. Crucially, such a highly developed solid angle is also independently required to produce the prominent Fe Kα emission lines observed by XRISM/Resolve, under the condition of the globally low column density of the reflecting matter.
This globally Compton-thin nature is strictly constrained by the NuSTAR data. As shown in Figure 2, there is no significant excess observed around 20–30 keV, and we obtain an upper limit of ≈ 5% on the hump strength relative to the continuum, confirming quantitatively the long-known weakness of the reflection in Cen A. Under our assumed model, the observed data are well explained by the relevant Thomson optical depths of along the line of sight, 0.2 along the equatorial plane of the torus, and only 0.02 through a single torus clump.
The necessity of evaluating the global gas distribution, rather than relying solely on the line-of-sight absorption, is well demonstrated by Mrk 3. T. Yaqoob et al. [94] showed that self-consistent modeling of the Fe Kα line and the reflection continuum instead implies a global (equatorial) column density well below the Compton-thick threshold, illustrating that the line-of-sight absorption alone does not necessarily reflect the properties of the reprocessing matter as a whole. The high energy resolution of XRISM/Resolve provides further support for this insight. Furthermore, Mrk 3 operates at a low accretion rate (; M. J. Koss et al. [44]). Given that Cen A is also a low-Eddington-ratio AGN, such globally Compton-thin but geometrically extended tori could potentially represent a common feature among this population, raising the possibility of a wider diversity in torus properties than commonly assumed.
We note that whereas D. Bogensberger et al. [13] obtained somewhat lower inclinations of degrees by fitting the Fe K profile with the MYTORUS model (K. D. Murphy & T. Yaqoob [58]) broadened by an rdblur kernel, the profile is still consistent with a high inclination up to 90 degrees within the 90% confidence limit. However, when incorporating constraints from reverberation mapping, high inclinations are ruled out (D. Bogensberger et al. [13]). The low inclination is also consistent with the jet proper motion measurements ( degrees as a upper limit; D. Bogensberger et al. [12]), where slightly higher angles (e.g., degrees) are only mildly disfavoured, and any minor discrepancy can be naturally accommodated by a disk-torus misalignment.
We have found that the profile of Fe K component from the torus is well characterized by two broadened components with different velocity widths, eV and eV, which are converted to velocity dispersions in FWHM of km s and km s, respectively. The line width reported in D. Bogensberger et al. [13] based on a single Gaussian fit, km s, falls between these values. It is noteworthy that the two-velocity component models are also found to well reproduce the Fe K profiles observed with XRISM in the Seyfert galaxies NGC 4388 (K. Fujiwara et al. [22]) and NGC 4507 (Aiso et al. in prep.), which have Eddington ratios of 0.1 (K. Fujiwara et al. [22]) and 0.07 (M. J. Koss et al. [44]), respectively. Thus, the two-velocity component models (i.e., three component model including the xBLR component) may give a common description of the line profile in obscured AGNs over a wide range of Eddington ratio. Cen A is a radio galaxy, while NGC 4388 and NGC 4507 are radio-quiet. Y. Nakatani et al. [63] found that the geometric structure of the torus depends on the Eddington ratio, not on the presence of a jet. The comparison above is therefore not affected by the different radio properties of these objects. Following these works, we hereafter refer to the torus1 and torus2 components as “main torus” and “inner torus” regions, respectively.
Assuming Keplerian motion, we estimate representative distances of the main and inner torus regions from the SMBH to be pc and pc ( is the gravitational radius where is the gravitational constant, is the black-hole mass, and is the light speed) from the observed line widths, respectively. Here we used the relation
where is the inclination angle of the system. We adopted deg, which is the value fixed in our reflection models (§ 3.2). The estimated distances scale as . These distances are consistent with those of two iron-line emitting regions suggested by T. Iwata et al. [35] based on the reverberation mapping of the Fe K line using the NuSTAR, Suzaku, XMM-Newton, and Swift/XRT data with the Swift/BAT light curve. We note that, in general, the two-velocity component model is a phenomenological description of the data and does not necessarily indicate that the emitter actually consists of two “distinct” regions. If a single smooth region is located over a wide range of radius with comparable intensities, the observed profile should be better represented by multiple velocity components. It is very difficult to distinguish which (distinct or smooth matter distribution) is the case, solely based on the line profile. In fact, we find that the narrow line profile can be also reasonably fit with a rdblur kernel, yielding the inner and outer radii of and for a fixed emissivity index of .
A notable difference in torus structure between Cen A and the higher-Eddington-ratio Seyfert 2s NGC 4388 and NGC 4507 lies in the relation between the dust sublimation radius () and the torus inner radius. In Cen A, is estimated to be 0.013 pc from equations (10) and (11) of Y. Kudoh et al. [46], one order of magnitude smaller than the torus inner radius. In NGC 4388 and NGC 4507, by contrast, and the torus inner radius are comparable to each other. This gap between and the torus inner radius in Cen A implies that dust can exist in its inner region, but the inner edge of the dense torus itself is not set by dust sublimation physics, such as radiation pressure on dust (e.g., K. Wada et al. [87]). Instead, in such low-luminosity (low Eddington-ratio) systems, this torus inner edge is likely determined by other (unknown) large-scale processes, such as the total gas mass in the circumuclear region. Further studies using a large sample covering a wide range of Eddington ratio are crucial to reach a unified view on the structure of AGN tori and their evolutionary scenario.
Recent JWST spectroscopic observations of the Cen A nucleus have provided new constraints on the line profiles in the near-infrared (NIR) band. Given that the torus is composed of molecular gas and dust, the Fe K reflection in X-rays and the molecular emission lines in the NIR should trace the kinematics of the same material. K. Goold et al. [26] reported multiple vibrational-rotational transitions of H emission lines. The (1-0)O(5) line at 3.235 m and the (1-0)O(6) line at 3.501 m have FWHM km s and 160 km s, respectively (both detected at ). These measurements agree, within the uncertainties, with the velocity widths of the inner torus component (torus2, FWHM km s) and the main torus component (torus1, FWHM km s) obtained in this study, respectively. We note that H emission is strongly affected by dust extinction, and thus it is not obvious whether the inner torus region can be probed as in the X-ray band. To estimate the extinction, we use the line-of-sight column density of cm obtained in this X-ray observation. Assuming the standard Galactic gas-to-dust ratio, this corresponds to mag at 3.235 m, i.e., a flux suppression by a factor of . The observed H line fluxes may therefore represent only a fraction of the total emission from the torus, and deeper regions could remain unprobed by these NIR lines. We note, however, that the signal-to-noise ratios of the H detections are low (3.8–4.0), and further confirmation with higher-sensitivity infrared spectroscopy is needed to establish this correspondence more robustly.
Broad Line Component of Fe K: X-ray BLR
In this subsection, we discuss the structure and physical origin of the broad line component of Fe K in Cen A on the basis of our spectral analysis results employing the xBLR reflection model (see Table 3 for the best-fit parameters). The velocity width of the xBLR component was found to be eV, which is converted to km s. We obtained deg and cm. In addition to characterizing the xBLR itself, we also demonstrate how the precise measurement of its velocity width can be applied to estimate the SMBH mass in obscured AGNs more generally.
Physical Structure of the X-ray BLR
The xBLR angular width (6.3–21 degrees) corresponds to a covering factor of (see § 3.1.2). Similarly to the case of the torus reflection, the lack of a strong reflection hump feature around 20–30 keV suggests that the xBLR also has a low column density. Figure 5 shows the confidence contours of the xBLR angular width and the radial column density. As noticed, to reproduce the observed intensity of the broad Fe K line without a strong reflection hump, the xBLR must have a large covering factor and a low column density.
Figure 5. Confidence contours of the xBLR angular width and radial column density. The color map shows , the -statistic relative to its best-fit value. The white curves show the 1, 2, and contours. The yellow dashed curve shows the joint 90% confidence region for two parameters (). The red cross marks the best-fit position. The solid blue, orange, green, red, and purple curves represent constant xBLR mass (0.1, 0.5, 0.9, 3.0, and 5.0 , respectively). These curves are calculated for the shell geometry described in § 4.2.2, with and fixed at their best-fit and derived values.
Assuming that the xBLR gas is in Keplerian rotation around the SMBH with the kinematic mass of (§ 1), we estimate the distance of the xBLR from the SMBH to be , according to Equation (10). This value is close to the dust sublimation radius , implying that dust-related physics (such as dust opacity) is relevant to produce this structure. Note that dust can survive at the surface of the accretion disk even at smaller radii than , and hence common mechanisms may also work in higher luminosity AGNs than Cen A.
The center energy of the Fe K line is consistent with 6.4 keV, indicating that the xBLR is mainly composed of neutral material. The ionization parameter is defined as ([78]), where is the luminosity, the density, and the distance from the irradiating source. For the material to remain neutral, must be smaller than ([38]). Adopting the 14–195 keV luminosity as ([43]) and using , we estimate the lower limit of the gas density as . Then, from the radial column density, the net thickness in the radial direction is estimated as , which is 10 times smaller than . Thus, the xBLR must be located in a spatially limited region if the gas distribution is smooth, whereas it can have a larger radial extent if it is clumpy.
In general, the absence of broad emission lines in the scattered (i.e., polarized) lights is explained by (1) the intrinsic absence of the BLR (known as “true type-2”) and/or (2) the lack of the scattering region that allows the BLR polarized light to be observed. Our results, confirming the presence of the xBLR in Cen A, provide important insights into this issue. We propose a new explanation that the xBLR clouds in Cen A are located close to the inner wall of the torus () and the geometrically thick torus () obstructs light traveling from the BLR to the scattering region. Figure 6 summarizes the nuclear structure we infer from our results including the torus and xBLR.

Figure 6. Schematic diagram of the geometry we infer for the sub-parsec scale nuclear structure. Around the dust sublimation radius, xBLR clouds originate or accumulate. The ordinary torus structure consists of two components, which can be interpreted as the ”inner wall” of the torus and the ”mean torus body”, respectively. Estimated distances in each component are labelled in the figure, not to scale.
Mass of the xBLR
Since X-rays are an unbiased tracer of all matter with a wide range of physical states and are little affected by dust physics, they give us the most complete measurement of the total mass of the reflector, once its geometry is known.
In our (smooth) xBLR model, the total mass of the xBLR is estimated as
where is the proton mass. The quoted uncertainty range is derived from the map in the plane of and (Figure 5). We adopt the joint 90% confidence region of these two parameters, defined by for two degrees of freedom (the yellow dashed curve in Figure 5). The minimum and maximum values of within this region are taken as the uncertainty range. In this evaluation, and are fixed at their best-fit and derived values, respectively. Their uncertainties are therefore not included. Note that this value should be taken as a lower limit, because the geometry is assumed to be smooth. If the xBLR is clumpy and (instead of ), the estimate increases by a factor of . Although the mass of the BLR has been estimated in various ways, there remain large uncertainties due to assumptions on the covering factor, column density, and physical state of the gas, and to possible dust extinction effects. J. A. Baldwin et al. [7] estimated the mass of the BLR to be for luminous quasars with 1450 Å luminosities of . According to J. A. Baldwin et al. [7] and R. Maiolino et al. [50], the mass of the BLR inferred from photoionization models scales with the AGN luminosity as . Assuming ([95, 73]), this formula yields for Cen A, which is lower than our estimate by more than an order of magnitude. Note that this relation is derived from luminous quasars, and may not be directly applicable to low-luminosity AGNs like Cen A. In X-ray absorption studies, the mass of the BLR has been estimated from the time variability of partial covering absorbers, such as in NGC 1365 (R. Maiolino et al. [50]). They derived the mass of the BLR to be two orders of magnitude lower than the value inferred from photoionization models. This absorption method relies on the assumption that the absorbers belong to the BLR, and only captures clouds along the line of sight, leading to large uncertainties in the total number of clouds.
Assuming a BLR cloud lifetime of 100 years as derived below from the Failed Radiatively Accelerated Dusty Outflow (FRADO) model (see below), the xBLR formation rate can be estimated as –. The mass accretion rate onto the black hole of Cen A is estimated to be (where is the radiative efficiency and we assume for a standard disk). This model requires that the majority (76–91%) of the inflowing mass around the dust sublimation radius be converted to xBLR clouds and retained. These xBLR clouds are thought to be bound gravitationally, to be a failed wind, and to provide feedback to the outer dusty torus [46, 86]. This points to a robust xBLR formation mechanism that remains efficient even in the low-Eddington regime. We note, however, that given the low Eddington ratio of Cen A, the innermost accretion flow may become radiatively inefficient. In such a case, the radiative efficiency could be smaller than the standard value of 0.1 assumed above (e.g., [64] (T. Manmoto et al. 1997)), and the actual mass accretion rate could be correspondingly larger than our estimate.
The estimated mass of the xBLR and its location provide a further constraint on how the xBLR gas is distributed. If we take the xBLR mass to be – and set a shell volume of (here we take pc for the most conservative case), the smooth density model with a filling factor would give a number density –. This sits one to two orders of magnitude below the lower limit of the density estimated from the ionization parameter, . This exposes a tension between the smooth density model and the ionization parameter constraint, suggesting that the xBLR is likely to be clumpy rather than smoothly distributed and requiring a filling factor of – to reconcile the two estimates. In Appendix § 7 we quantify the effect of such clumpiness on the reflection spectrum, showing that it changes the spectral shape by at fixed column density, so that the parameters derived above are unaffected, while remaining too small to be detected with the present data.
Implications for BLR Formation Models
In this subsection, we compare our results with one of the leading BLR formation models, called the Failed Radiatively Accelerated Dusty Outflow (FRADO) model ([17, 18, 60, 61, 59, 62]). The FRADO model operates on the principle of a dust-driven outflow triggered at the disk surface where temperatures are lower than the sublimation limit ( K). The dust clouds form at these cooler regions on the accretion disk, and radiation pressure from the disk lifts these dusty clouds off the disk surface, forming the BLR. The dynamics of each cloud are characterized by a recurring cycle: (1) the cloud is lifted off the disk by radiation pressure on the dust, (2) as it is exposed to radiation from the central SMBH corona, the dust within the cloud sublimates, (3) deprived of the driving force provided by radiation pressure, these clouds eventually follow a “failed” trajectory, falling back to the disk surface. This mechanism establishes a self-consistent lifecycle for the BLR material. The lifetime of each BLR cloud in this cycle is estimated to be on the order of 100 years based on nonhydrodynamical calculations.
Assuming a Shakura–Sunyaev disk ([74]), the dust formation radius, where the disk temperature drops below the dust sublimation temperature ( K), is estimated to be around . At the projected location of the xBLR (), the disk surface temperature is lower, estimated at K. Thus, the xBLR is located outside the dust formation radius. Given this separation, while dust may play a role in cloud formation and inflate the BLR, it is necessary to consider additional physical processes to transport the xBLR clouds outward and accumulate them at such a specific location.
For a black-hole mass of and an Eddington ratio of 0.01, M.-H. Naddaf et al. [60] calculated a BLR opening angle ratio of approximately 0.03, which corresponds to an opening angle of deg. In contrast, we find an xBLR opening angle of deg. This larger angle aligns with the findings of M.-H. Naddaf et al. [60] for high-Eddington-ratio () AGNs. As noted by B. Czerny et al. [18], the insufficiency of radiation pressure in standard dust-driven winds implies the presence of optically thick clouds or additional mechanisms. Our results provide the first X-ray-based constraints that can be used to further refine and update the standard BLR formation scenario.
The inflow required by the xBLR formation rate estimated above is indeed observed on larger scales. L. Evangelista et al. [20] found that shocks excite the H gas, regulate its temperature, and prevent cooling below 100 K. These shocks may drive angular momentum loss and help fuel the nucleus. As evidence for such an inflow, they identified a coherent, low-dispersion (70 km s) streamer spiralling inward in the central 170 100 pc region. O. C. Jones et al. [36] further showed that the warped central dust disk of Cen A was formed in a gas-rich merger, and that star formation there is primarily regulated by merger-accreted gas rather than by the jet. These results provide the large-scale counterpart of the mass budget required by our estimate.
SMBH Mass Estimation Using High Energy-Resolution X-ray Spectra
The detection of this hidden xBLR has an impact extending well beyond Cen A, providing a powerful new tool to estimate SMBH masses in obscured AGNs. In this section, we discuss this method, enabled by XRISM’s capability to spectroscopically resolve broad X-ray line components. While standard techniques using the line width of the (optical) BLR and continuum luminosity in the UV to optical bands have been successful in estimating their SMBH masses, these cannot be applied to type 2 AGNs. Thanks to the strong penetrating power of X-rays, a BLR hidden in the optical band can still be observed in X-rays, as demonstrated in our observations. Since a major fraction of AGNs in the universe are obscured [83], establishing a method of SMBH mass estimation for obscured AGNs is crucial to understanding the growth history of whole SMBHs.
Our purpose is to extend the established relation in the optical band using the FWHM of the H line and (e.g., [10, 81]) to the X-ray regime. Here we assume that the optical BLR is a part of the xBLR and their locations are the same. In fact, quasi-simultaneous observations of the Seyfert 1.5 galaxy NGC 4151 with XRISM and Seimei telescope have shown that the widths of the H line and the Fe K line are similar to each other (Noda et al., in prep.), supporting this interpretation. We can estimate the intrinsic luminosity at 5100 Å of an obscured AGN from the 14–195 keV luminosity by the relation
where the luminosities are in units of erg s and the value in parentheses is the 1 scatters in dex. We have derived this relation from 327 broad-line AGNs of the BAT AGN Spectroscopic Survey (BASS) utilizing the most up-to-date estimates of (see Appendix § 8).
The SMBH mass estimation formula based on optical reverberation mapping is ([81])
where 8.02 = . This relation assumes a virial factor of . The index of 0.65 is that of the radius-luminosity (–) relation of the H BLR. We generalize Equation (13) for the Fe K line and hard X-ray luminosity as
Here is a dimensionless coefficient that scales the normalization of the H relation, , and is the power-law index of the luminosity dependence. Both must be calibrated for the Fe K line. Empirically, the index of the optical – relations lies in the range of 0.5–0.7 ([81]), which can differ among emission lines (e.g., Mg II and H).
We first apply Equation (14) to Cen A with the optical values, and . We use FWHM(Fe K) = 6000 km s and ([43]). This gives and . The uncertainty is quoted at 1, combined in quadrature from two independent sources: the statistical error in the Fe K line width and the scatter in the conversion to multiplied by . The best estimate is lower than the mass adopted in § 1 () and the difference is 0.70 dex (a factor of about 5). For with , we obtain . This is consistent with 7.7 within 1. These results suggest that this method of SMBH mass estimation using the broad Fe K line is promising.
To make the relation exactly consistent with the adopted mass, and should satisfy (e.g., for ). An earlier study reported a larger mass of ([75]; see § 1). For this mass, and would be required. Determining these coefficients requires a sample of AGNs with broad Fe K widths and independent mass measurements covering a wide range of X-ray luminosity.
However, this method is still in its early stages and several caveats remain. First, adopting the optical values for and relies on the assumption that the xBLR is co-spatial with the optical BLR, a hypothesis that remains to be confirmed ([53]). This needs to be addressed through simultaneous observations of xBLR with XRISM and optical BLR using a large sample of unobscured AGNs.
Second, geometry-induced bias may exist. The optical - relations are biased toward face-on systems; thus, applying them to edge-on systems (obscured AGNs) may introduce systematic errors due to the anisotropic velocity structure of the xBLR (see discussion in [22] for NGC 4388). It is therefore important to perform reverberation mapping of Fe K (e.g., [56]) to directly determine the relation for obscured AGNs. Third, the conversion from the 14–195 keV luminosity to the 5100 Å luminosity has a 1 scatter of 0.50 dex (Appendix § 8). This is the dominant term in the uncertainty of our estimate. This could be improved by considering the dependence of the spectral energy distribution on AGN luminosity and/or Eddington ratio.
A black-hole mass estimation method for obscured AGNs was recently proposed by S. LaMassa et al. [47]. Their method is based on the virial relation of D. Baron & B. Ménard [8]. Therefore, the basic framework is similar to ours, but the method to derive each parameter is different. Instead of measuring the velocity width kinematically, S. LaMassa et al. [47] adopt an empirical estimate from the optical narrow-line ratio . Our method has the advantage of directly measuring the velocity width from the Fe K line. S. LaMassa et al. [47] report a luminosity scatter of 0.68 dex in derived from the intrinsic 2–10 keV luminosity. This is comparable to the 0.52 dex scatter that we obtain for the same band (Appendix § 8). Our adopted conversion from the 14–195 keV luminosity has a scatter of 0.50 dex. However, a direct comparison for Cen A is difficult because the strong nuclear extinction makes it hard to reliably measure the narrow-line ratio, and it does not satisfy their selection criteria. Applying both methods to a common sample of nearby obscured AGNs with high-resolution X-ray data is left as future work.
Recent JWST/MIRI observations have characterized the circumnuclear region of Cen A in detail. A. Alonso Herrero et al. [3] studied the kinematics of the ionized gas and revealed a fast outflow along the radio jet. L. Evangelista et al. [20] traced the feeding and feedback of the warm molecular gas, as described in § 4.2.3. O. C. Jones et al. [36] characterized the embedded star formation in the warped dust disk. None of these studies, however, constrains the torus or the BLR. A direct test of the co-spatiality would be near-infrared spectroscopic observations of the core, targeting hydrogen recombination lines such as Pa and Br (e.g., with JWST/NIRSpec). Existing near-infrared spectroscopy shows no broad component in Pa or Br ([26]), which is consistent with the strong extinction toward the nucleus. A search in the longer-wavelength lines, where the extinction is smaller, is therefore the natural next step.
Chemical Abundances
In this subsection, we briefly discuss the abundance ratios of Ca and Ni to Fe in the torus and xBLR of Cen A, based on the spectral fitting results using the XCLUMPY and xBLR reflection models. The high-energy resolution spectroscopy with XRISM/Resolve allows us to detect and precisely measure the intensities of these fluorescent lines for the first time in a radio galaxy. Figure 7 shows the zoomed-in spectra around the Ca K (3.69 keV in rest frame) and Ni K (7.48 keV) lines. In our best-fit model, the abundance ratios of Ca and Ni to Fe are found to be and .

Figure 7. Zoom-in around the fluorescence lines, Ca Kα and Ni Kα. The XRISM/Resolve data points and the best-fit model are shown in the same way as in the right panel of Figure 3.
As an independent check, we estimated from the observed Fe K-to-Ni K photon flux ratio () using the empirical relation of T. Yaqoob & K. D. Murphy [93]. In the optically thin limit, this relation is independent of the continuum shape and the covering factor. When we adjust their normalization to the K. Lodders et al. [48] solar abundance scale adopted in this paper, we obtain . This value is qualitatively consistent with our best-fit value. However, we do not expect exact quantitative agreement because the relation of T. Yaqoob & K. D. Murphy [93] strictly holds only for . This limit is lower than the column density inferred for the reflecting material in Cen A.
However, we note some limitations in our analysis that may affect the accuracy of these abundance ratio measurements besides the statistical uncertainties. The data taken with NuSTAR do not show strong evidence of a Compton hump in the hard X-ray band, offering us only upper limits on the reflection continuum strength. It leads to uncertainties in the ratio of fluorescent line intensities to that of the reflection continuum, which is a key quantity that determines the elemental abundance relative to hydrogen. Our current estimate of the reflection continuum strength depends on the iron to hydrogen abundance ratio, which is assumed to be the solar value. In addition, the uncertainty on the reflection continuum weakens the constraints on the torus geometric structure and affects the evaluation of self-absorption. It becomes more important for modeling of the reflection continuum at soft energies, thereby impacting the evaluation of the abundance ratio of .
Bearing these caveats in mind, the abundance ratio patterns we obtained for the reflecting materials in Cen A provide insights into the possible origin of the mass reservoirs around the AGN. For the Circinus Galaxy, XRISM Collaboration et al. [92] reported that the abundance pattern of the reflecting materials shows sub-solar and ratios, while the ratio is super-solar. They concluded that the central SMBH is fed by gas enriched predominantly by recent core-collapse supernovae from progenitors less massive than 20 , reflecting the active nuclear star formation in the spiral host. In contrast, our results for Cen A show solar value for and nominally super-solar value for , implying a different enrichment history. Given the limited number of elements and large statistical and systematic uncertainties, we leave detailed modeling of the abundance pattern with supernova nucleosynthesis models for future work. Future deeper observations that enable us to better determine the abundance ratios including other elements (particularly Ar and Mn) would be required to exploit the abundance-pattern diagnostics.
Conclusions
We summarize our main findings from the broadband spectral analysis of Cen A using simultaneous XRISM (Resolve and Xtend) and NuSTAR observations. For our analysis, we developed radiative transfer-based numerical models representing the reflection continuum and line components from the torus and the xBLR, employing the SKIRT framework.
The velocity profile of the Fe Kα line is well reproduced with three Gaussian kernels, similarly to the case of NGC 4388. This feature may be common in AGNs when sufficient photon statistics of Fe Kα are obtained. We interpret that the two narrower components (with FWHMs of km s and km s) are from the inner and main torus regions, whereas the broad one ( km s) is from the xBLR. This enables physical decomposition of the broadband spectra, consisting of the primary emission, ionized emission, and reflection components from the torus and xBLR. We demonstrate that high-energy-resolution X-ray spectroscopy can detect a deeply buried BLR that remains undetected in the optical-to-infrared bands due to severe extinction (– mag).
Our results confirm a picture that the torus in Cen A has a large covering factor ( deg) with moderate hydrogen column densities ( cm). This “buried” nature may explain the absence of broad emission lines even in the polarized optical spectra.
The representative distances of the inner and main torus regions from the SMBH are pc and pc, respectively. This is about 10 times larger than the estimated dust sublimation radius, implying that the innermost radius of the torus is not determined by dust-related physics in a low-luminosity AGN.
We suggest that the xBLR has a substantial covering factor – with a moderate hydrogen column density ( cm) and is located at 0.014 pc from the SMBH. The estimated total mass in the xBLR is , which is larger than those in other AGNs based on Balmer lines or X-ray absorption variability. The xBLR mass relative to the accretion rate suggests a robust xBLR formation mechanism that remains highly efficient even in low-Eddington ratio regimes, providing a new constraint on BLR production mechanisms.
We propose a novel method to estimate the SMBH mass of obscured AGNs from the xBLR line width and the X-ray luminosity applicable to obscured AGNs. The X-ray luminosity is converted to with a relation derived from BASS broad-line AGNs. With the constants of the optical H relation, this method yields from the observed 14–195 keV luminosity. This is smaller than the kinematic mass by a factor of
about 5. With a flatter luminosity dependence (), the method yields 7.60 0.26. This is consistent with the kinematic mass within . Determining the two constants requires a sample of AGNs with independent mass measurements.
We derived abundance ratios of , and in the reflecting material of Cen A.
Software and third party data repository citations
Facilities: XRISM, NuSTAR
Software: XSPEC [5], SKIRT [6, 84]
APPENDIX
Effects of Clumpiness in xBLR Gas on the Model Spectra
The xBLR model in the main text assumes a smooth geometry, yet observational and theoretical studies suggest that the BLR is not smooth but rather inhomogeneous. If the BLR has a clumpy density distribution, the self-absorption within individual clouds can differ from that of a smooth model with the same average column density along the line of sight. This difference could alter the ratio of the Fe K line intensity to the Compton hump intensity, thereby introducing a systematic error into the abundance ratio measurements derived in the main text. This appendix quantifies the systematic impact of BLR clumpiness on the X-ray reflection spectrum using SKIRT simulations and evaluates its implications for the main-text results and conclusions. We also assess whether the high-energy-resolution data from XRISM/Resolve can place quantitative constraints on clumpiness.
For the simulations, we adopt the same xBLR geometry as in the main analysis, but replace the smooth density distribution with a clumpy one composed of spherical clumps randomly distributed with a uniform probability density within the BLR volume. We parameterize clumpiness by varying the number of clumps along the line of sight, . At the model is most clumpy; increasing reduces individual clump sizes and column densities, and the model converges toward a smooth distribution. We assume that each clump has a radius of . In this setting, the volume filling factor (the total clump volume relative to the BLR volume) remains constant across all models. In order to maintain a constant across all models, the hydrogen number density of each clump is adjusted accordingly. The computational grid is refined so that each clump is resolved by a minimum of cells. All models share the following fixed parameters: pc, pc, , deg, , keV, and solar abundances. We compare these clumpy models with the smooth model used in the main analysis in terms of their reflection spectra.
The ratios to the smooth reference model are presented in Figure 8(a). Spectral differences between the clumpy and smooth models are –20% at , at , and below 2% at . In order to quantify the spectral differences, we focus on the flux ratio of the Fe K line (6.0–6.8 keV) to the Compton hump (15–40 keV). As increases, increases and converges toward the smooth reference value from below (Figure 8(c)). This trend can be understood as follows: in low models, each clump has a large column density and enhances self-absorption, leading to a more suppressed Fe K line intensity, while high energy photons of the Compton hump are less susceptible to self-absorption. To assess the relative importance of clumpiness, we compare this effect with the spectral variation caused by changes in the inclination angle. Figure 8(b) shows the spectral variation computed for each model with inclination angles of 30, 50, 75, and 85 deg. A change in inclination angle alters the -direction column density and hence the degree of self-absorption, producing changes of the spectral shape, mainly on hardness ratios. This variation exceeds the effect due to clumpiness, so the systematic error arising from clumpiness is negligible. Consequently, quantitative constraints on clumpiness remain beyond the reach of current observations, even with the XRISM/Resolve data.
Figure 8. Dependence of the spectral shape produced by a clumpy reflector on the line-of-sight clump number and the inclination angle . All spectra shown here are smoothed with a Gaussian of 50 eV (at 6 keV) for display purposes. (a) Ratio of the reflection spectrum (reflection continuum plus line components) at relative to the smooth reflector (Smooth), shown over the 2–30 keV energy range. The blue, red, green, and orange lines represent the ratios for the clump models with , respectively. The black line is for the reference Smooth model. (b) Spectra of the smooth reflector model at normalized by the spectrum. The black, blue, red, and green lines represent the ratios of the Smooth model at , and , respectively. (c) Dependence of on . Here, is defined as the ratio between the iron-line intensity (6.0–6.8 keV) and the Compton-hump intensity (15–40 keV). Solid lines with circles represent the clump models for each inclination, and dashed lines represent the Smooth model at the same inclination. The black, blue, red, and green colors indicate , and , respectively.
Relation between the 5100 Å and X-ray Luminosities
The relation between the 14–195 keV and luminosities was reported by M. Koss et al. [43] based on the BAT AGN Spectroscopic Survey Data Release 1. Here we rederive the relation by utilizing more recent results of (J. E. Mejía-Restrepo et al. [55]) based on a larger sample of broad-H AGNs in BASS DR2 (M. J. Koss et al.
2022). The final sample consists of 327 AGNs after excluding 34 blazars and 3 outliers3. The observed rest-frame 14–195 keV luminosities (not corrected for absorption) are taken from tbl:13 of C. Ricci et al. [71].
Figure 9 plots the relation between the and of this sample. We fit these data with , where and are the logarithms of the 14–195 keV and 5100 Å luminosities, respectively. We obtain with a scatter of 0.50 dex, which we adopt in Equation (12)4. There are caveats, however. First, the optical and X-ray data are not simultaneous and could be affected by time variability. Second, the values in J. E. Mejía-Restrepo et al. [55] include the host-galaxy light, which may therefore make overestimated, especially at low luminosities.
Figure 9. Relation between and 14–195 keV luminosity for the BASS broad-line AGNs (blue circles). Blazars, which are not included in the fit, are marked in light-blue stars. The solid line is the best-fit relation , where is the logarithm of the X-ray luminosity and . The dashed lines show the region within scatter.
Acknowledgments
Y.N. acknowledges financial support from Grant-in-Aid for the Japan Society for the Promotion of Science (JSPS) Fellows (24KJ1507). This work was partially supported by JSPS KAKENHI Grant Numbers JP25H00671 (Y.U.), JP26K17197, JP21K13958 (M.M.), and JP25KJ1868 (T.K.). D.B. acknowledges financial support from the Centre national d’études spatiales (CNES), France (ROR: https://ror.org/04h1h0y33) within the framework of the XMM-Newton mission. L.G. acknowledges support from the Canadian Space Agency grant 25EXPRSM1 and 18XARM-STMA. T.Y. acknowledges support from CRESST II under NASA award number 80GSFC24M0006. B.V. acknowledges support through the European Space Agency (ESA) Research Fellowship in Space Science. M.M. acknowledges support from Yamada Science Foundation.
References
- [1]Akaike, H. 1974, IEEE Transactions on Automatic Control, 19, 716, doi: 10.1109/TAC.1974.1100705DOI
- [2]Alexander, D. M., Efstathiou, A., Hough, J. H., et al. 1999, MNRAS, 310, 78, doi: 10.1046/j.1365-8711.1999.02908.xDOI
- [3]Alonso Herrero, A., Hermosa Muñoz, L., Labiano, A., et al. 2025, A&A, 699, A334, doi: 10.1051/0004-6361/202554823DOI
- [4]Antonucci, R. 1993, ARA&A, 31, 473, doi: 10.1146/annurev.aa.31.090193.002353DOI
- [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
- [6]Baes, M., Davies, J. I., Dejonghe, H., et al. 2003, MNRAS, 343, 1081, doi: 10.1046/j.1365-8711.2003.06770.xDOI
- [7]Baldwin, J. A., Ferland, G. J., Korista, K. T., Hamann, F., & Dietrich, M. 2003, ApJ, 582, 590, doi: 10.1086/344788DOI
- [8]Baron, D., & Ménard, B. 2019, MNRAS, 487, 3404, doi: 10.1093/mnras/stz1546DOI
- [9]Beckmann, V., Jean, P., Lubinski, P., Soldi, S., & Terrier, R. 2011, A&A, 531, A70, doi: 10.1051/0004-6361/201016020DOI
- [10]Bentz, M. C., Peterson, B. M., Netzer, H., Pogge, R. W., & Vestergaard, M. 2009, ApJ, 697, 160, doi: 10.1088/0004-637X/697/1/160DOI
- [11]Blandford, R. D., & Znajek, R. L. 1977, MNRAS, 179, 433, doi: 10.1093/mnras/179.433
- [12]Bogensberger, D., Miller, J. M., Mushotzky, R., et al. 2024, ApJ, 974, 307, doi: 10.3847/1538-4357/ad73a1DOI
- [13]Bogensberger, D., Nakatani, Y., Yaqoob, T., et al. 2025, PASJ, 77, S209, doi: 10.1093/pasj/psaf084DOI
- [14]Brightman, M., & Nandra, K. 2011, MNRAS, 413, 1206, doi: 10.1111/j.1365-2966.2011.18207.xDOI
- [15]Cappellari, M., Neumayer, N., Reunanen, J., et al. 2009, MNRAS, 394, 660, doi: 10.1111/j.1365-2966.2008.14377.xDOI
- [16]Cash, W. 1979, ApJ, 228, 939, doi: 10.1086/156922DOI
- [17]Czerny, B., & Hryniewicz, K. 2011, A&A, 525, L8, doi: 10.1051/0004-6361/201016025DOI
- [18]Czerny, B., Li, Y.-R., Hryniewicz, K., et al. 2017, ApJ, 846, 154, doi: 10.3847/1538-4357/aa8810DOI
- [19]Espada, D., Matsushita, S., Miura, R. E., et al. 2017, ApJ, 843, 136, doi: 10.3847/1538-4357/aa78a9DOI
- [20]Evangelista, L., Guillard, P., Martin, J., et al. 2026, A&A, 711, A290, doi: 10.1051/0004-6361/202558362DOI
- [21]Fabian, A. C. 2012, ARA&A, 50, 455, doi: 10.1146/annurev-astro-081811-125521DOI
- [22]Fujiwara, K., Ueda, Y., Ogawa, S., et al. 2026, ApJ, 1003, 6, doi: 10.3847/1538-4357/ae5e5cDOI
- [23]Fukazawa, Y., Hiragi, K., Mizuno, M., et al. 2011, ApJ, 727, 19, doi: 10.1088/0004-637X/727/1/19DOI
- [24]Fürst, F., Müller, C., Madsen, K. K., et al. 2016, ApJ, 819, 150, doi: 10.3847/0004-637X/819/2/150DOI
- [26]Goold, K., Seth, A., Molina, M., et al. 2026, ApJ, 1000, 281, doi: 10.3847/1538-4357/ae4bddDOI
- [27]Granato, G. L., & Danese, L. 1994, MNRAS, 268, 235, doi: 10.1093/mnras/268.1.235DOI
- [28]Grasdalen, G. L., & Joyce, R. R. 1976, ApJ, 208, 317, doi: 10.1086/154610DOI
- [29]Hardcastle, M. J., Worrall, D. M., Kraft, R. P., et al. 2003, ApJ, 593, 169, doi: 10.1086/376519DOI
- [30]Harris, G. L. H., Rejkuba, M., & Harris, W. E. 2010, PASA, 27, 457, doi: 10.1071/AS09061DOI
- [31]Hayashi, T., Boissay-Malaquin, R., Tamura, K., et al. 2024, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 13093, Space Telescopes and Instrumentation 2024: Ultraviolet to Gamma Ray, ed. J.-W. A. den Herder, S. Nikzad, & K. Nakazawa, 130931L, doi: 10.1117/12.3019600DOI
- [32]Hölzer, G., Fritsch, M., Deutsch, M., Härtwig, J., & Förster, E. 1997, PhRvA, 56, 4554, doi: 10.1103/PhysRevA.56.4554DOI
- [33]Ichikawa, K., Packham, C., Ramos Almeida, C., et al. 2015, ApJ, 803, 57, doi: 10.1088/0004-637X/803/2/57DOI
- [34]Ishisaki, Y., Kelley, R. L., Awaki, H., et al. 2025, Journal of Astronomical Telescopes, Instruments, and Systems, 11, 042023, doi: 10.1117/1.JATIS.11.4.042023DOI
- [35]Iwata, T., Tanimoto, A., Odaka, H., et al. 2024, PASJ, 76, 923, doi: 10.1093/pasj/psae059DOI
- [36]Jones, O. C., Jones, M. L., Dicken, D., et al. 2026, arXiv e-prints, arXiv:2607.04942, doi: 10.48550/arXiv.2607.04942DOI
- [37]Kaastra, J. S., & Bleeker, J. A. M. 2016, A&A, 587, A151, doi: 10.1051/0004-6361/201527395DOI
- [38]Kallman, T. R., Palmeri, P., Bautista, M. A., Mendoza, C., & Krolik, J. H. 2004, ApJS, 155, 675, doi: 10.1086/424039DOI
- [39]Kayanoki, T., Fukazawa, Y., Mao, J., et al. 2026, ApJ, 1002, 91, doi: 10.3847/1538-4357/ae47e0DOI
- [40]Kelley, R. L., Ishisaki, Y., Costantini, E., et al. 2025, Journal of Astronomical Telescopes, Instruments, and Systems, 11, 042026, doi: 10.1117/1.JATIS.11.4.042026DOI
- [41]King, A., & Pounds, K. 2015, ARA&A, 53, 115, doi: 10.1146/annurev-astro-082214-122316DOI
- [42]Kormendy, J., & Ho, L. C. 2013, ARA&A, 51, 511, doi: 10.1146/annurev-astro-082708-101811DOI
- [43]Koss, M., Trakhtenbrot, B., Ricci, C., et al. 2017, ApJ, 850, 74, doi: 10.3847/1538-4357/aa8ec9DOI
- [44]Koss, M. J., Ricci, C., Trakhtenbrot, B., et al. 2022, ApJS, 261, 2, doi: 10.3847/1538-4365/ac6c05DOI
- [45]Krajnović, D., Sharp, R., & Thatte, N. 2007, MNRAS, 374, 385, doi: 10.1111/j.1365-2966.2006.11181.xDOI
- [46]Kudoh, Y., Wada, K., Kawakatu, N., & Nomura, M. 2024, ApJ, 977, 48, doi: 10.3847/1538-4357/ad8b42DOI
- [47]LaMassa, S., Farrow, I., Urry, C. M., et al. 2025, ApJ, 981, 101, doi: 10.3847/1538-4357/adb0bfDOI
- [48]Lodders, K., Palme, H., & Gail, H.-P. 2009, Landolt Börnstein, 4B, 712, doi: 10.1007/978-3-540-88055-4_34DOI
- [49]Magorrian, J., Tremaine, S., Richstone, D., et al. 1998, AJ, 115, 2285, doi: 10.1086/300353DOI
- [50]Maiolino, R., Risaliti, G., Salvati, M., et al. 2010, A&A, 517, A47, doi: 10.1051/0004-6361/200913985DOI
- [52]Marconi, A., Pastorini, G., Pacini, F., et al. 2006, A&A, 448, 921, doi: 10.1051/0004-6361:20053853DOI
- [53]Marin, F., Hutsemékers, D., Jiang, C.-Z., et al. 2025, A&A, 695, A55, doi: 10.1051/0004-6361/202452663DOI
- [54]Meisenheimer, K., Tristram, K. R. W., Jaffe, W., et al. 2007, A&A, 471, 453, doi: 10.1051/0004-6361:20066967DOI
- [55]Mejía-Restrepo, J. E., Trakhtenbrot, B., Koss, M. J., et al. 2022, ApJS, 261, 5, doi: 10.3847/1538-4365/ac6602
- [56]Miller, J. M., Zoghbi, A., Xiang, X., et al. 2026, ApJL, 1005, L48, doi: 10.3847/2041-8213/ae7e82DOI
- [58]Murphy, K. D., & Yaqoob, T. 2009, MNRAS, 397, 1549, doi: 10.1111/j.1365-2966.2009.15025.xDOI
- [59]Naddaf, M.-H., Czerny, B. 2024, Universe, 10, 29, doi: 10.3390/universe10010029DOI
- [60]Naddaf, M.-H., Czerny, B., & Szczerba, R. 2020, Frontiers in Astronomy and Space Sciences, 7, 15, doi: 10.3389/fspas.2020.00015DOI
- [61]Naddaf, M.-H., Czerny, B., & Szczerba, R. 2021, ApJ, 920, 30, doi: 10.3847/1538-4357/ac139dDOI
- [62]Naddaf, M. H., Martinez-Aldama, M. L., Hutsemékers, D., Savic, D., & Czerny, B. 2025, A&A, 702, A46, doi: 10.1051/0004-6361/202555767DOI
- [63]Nakatani, Y., Ueda, Y., Ricci, C., et al. 2023, MNRAS, 523, 6239, doi: 10.1093/mnras/stad1836DOI
- [64]Narayan, R., & Yi, I. 1995, ApJ, 452, 710, doi: 10.1086/176343DOI
- [65]Nenkova, M., Sirocky, M. M., Ivezić, Ž., & Elitzur, M. 2008a, ApJ, 685, 147, doi: 10.1086/590482DOI
- [67]Neumayer, N., Cappellari, M., Reunanen, J., et al. 2007, ApJ, 671, 1329, doi: 10.1086/523039DOI
- [68]Noda, H., Mori, K., Tomida, H., et al. 2025, PASJ, 77, S10, doi: 10.1093/pasj/psaf011DOI
- [69]Ogawa, S., Ueda, Y., Tanimoto, A., & Yamada, S. 2021, ApJ, 906, 84, doi: 10.3847/1538-4357/abccceDOI
- [70]Ricci, C., Trakhtenbrot, B., Koss, M. J., et al. 2017a, Nature, 549, 488, doi: 10.1038/nature23906DOI
- [71]Ricci, C., Trakhtenbrot, B., Koss, M. J., et al. 2017b, ApJS, 233, 17, doi: 10.3847/1538-4365/aa96adDOI
- [72]Ricci, C., Ho, L. C., Fabian, A. C., et al. 2018, MNRAS, 480, 1819, doi: 10.1093/mnras/sty1879DOI
- [73]Richards, G. T., Lacy, M., Storrie-Lombardi, L. J., et al. 2006, ApJS, 166, 470, doi: 10.1086/506525DOI
- [74]Shakura, N. I., & Sunyaev, R. A. 1973, A&A, 24, 337
- [75]Silge, J. D., Gebhardt, K., Bergmann, M., & Richstone, D. 2005, AJ, 130, 406, doi: 10.1086/431315DOI
- [76]Stalevski, M., Fritz, J., Baes, M., Nakos, T., & Popović, L. Č. 2012, MNRAS, 420, 2756, doi: 10.1111/j.1365-2966.2011.19775.xDOI
- [77]Tanimoto, A., Ueda, Y., Odaka, M., et al. 2019, ApJ, 877, 95, doi: 10.3847/1538-4357/ab1b20DOI
- [78]Tarter, C. B., Tucker, W. H., & Salpeter, E. E. 1969, ApJ, 156, 943, doi: 10.1086/150026DOI
- [79]Tashiro, M., Maejima, H., Toda, K., 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, 1144422, doi: 10.1117/12.2565812DOI
- [80]Tashiro, M., Watanabe, S., Maejima, H., et al. 2024, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 13093, Space Telescopes and Instrumentation 2024: Ultraviolet to Gamma Ray, ed. J.-W. A. den Herder, S. Nikzad, & K. Nakazawa, 130931G, doi: 10.1117/12.3019325DOI
- [81]Trakhtenbrot, B., & Netzer, H. 2012, MNRAS, 427, 3081, doi: 10.1111/j.1365-2966.2012.22056.xDOI
- [82]Turner, P. C., Forrest, W. J., Pipher, J. L., & Shure, M. A. 1992, ApJ, 393, 648, doi: 10.1086/171532DOI
- [83]Ueda, Y., Akiyama, M., Hasinger, G., Miyaji, T., & Watson, M. G. 2014, ApJ, 786, 104, doi: 10.1088/0004-637X/786/2/104DOI
- [84]Vander Meulen, B., Camps, P., Stalevski, M., & Baes, M. 2023, A&A, 674, A123, doi: 10.1051/0004-6361/202245783
- [85]Vander Meulen, B., Camps, P., Tsujimoto, M., & Wada, K. 2024, A&A, 688, L33, doi: 10.1051/0004-6361/202451370
- [86]Wada, K., Nagao, T., Shimizu, T., et al. 2026, ApJ, 998, 60, doi: 10.3847/1538-4357/ae36a7DOI
- [87]Wada, K., Schartmann, M., & Meijerink, R. 2016, ApJL, 828, L19, doi: 10.3847/2041-8205/828/2/L19DOI
- [88]Willingale, R., Starling, R. L. C., Beardmore, A. P., Tanvir, N. R., & O’Brien, P. T. 2013, MNRAS, 431, 394, doi: 10.1093/mnras/stt175DOI
- [89]Xiang, X., Miller, J. M., Behar, E., et al. 2025, ApJL, 988, L54, doi: 10.3847/2041-8213/adee9bDOI
- [90]Xrism Collaboration, Audard, M., Awaki, H., et al. 2024, ApJL, 973, L25, doi: 10.3847/2041-8213/ad7397
- [91]Xrism Collaboration, Audard, M., Awaki, H., et al. 2025, A&A, 702, A147, doi: 10.1051/0004-6361/202556000DOI
- [92]XRISM Collaboration, Audard, M., Awaki, H., et al. 2026, Nature Astronomy, 10, 880, doi: 10.1038/s41550-026-02817-6DOI
- [93]Yaqoob, T., & Murphy, K. D. 2011, MNRAS, 412, 1765, doi: 10.1111/j.1365-2966.2010.18012.xDOI
- [94]Yaqoob, T., Tatum, M. M., Scholtes, A., Gottlieb, A., & Turner, T. J. 2015, MNRAS, 454, 973, doi: 10.1093/mnras/stv2021DOI
- [95]Young, M., Elvis, M., & Risaliti, G. 2010, ApJ, 708, 1388, doi: 10.1088/0004-637X/708/2/1388DOI





