WorksPhysics and Astronomy
Origin of the Fe K red excess in IC 4329A with XRISM and multi-wavelength spectroscopy
Abstract
We present high-resolution X-ray spectroscopy of the bright local Seyfert 1 galaxy IC 4329A obtained with XRISM/Resolve, supported by simultaneous NuSTAR data and quasi-simultaneous optical/near-infrared (NIR) spectroscopy. Resolve separates the Fe K line into a narrow core and a broad component exhibiting a red excess. We examined three scenarios for this red excess: (1) a redshifted X-ray broad line region (xBLR), (2) relativistic disk reflection, and (3) an unshifted xBLR partially absorbed by a mildly ionized absorber. All three models reproduce the X-ray spectra equally well. Our optical (H) and NIR (Pa) spectroscopy shows that the broad lines are redshifted by , close to the solution of the redshifted xBLR model (). Motivated by this, we constructed a hybrid model in which the xBLR shares the velocity structure of the optical/NIR BLR and the remaining red excess is attributed to a relativistic disk line, which also provides an acceptable fit. The inferred xBLR radius is ( light-days). From the narrow Fe K core in the hybrid model (), we estimate a torus distance of pc ( light-days), close to the dust sublimation radius. Furthermore, two ultra-fast outflow (UFO) absorption features are identified with and , both confirmed at confidence.
Introduction
Active Galactic Nuclei (AGNs) comprise a complex, multi-scale architecture including an accretion disk, a broad line region (BLR), and a dusty molecular torus, as conceptualized in the unified model (C. M. Urry & P. Padovani [54]).
In the X-ray regime, the Fe Kα fluorescent line at 6.4 keV serves as an indispensable diagnostic of the geometry and physical conditions of the circumnuclear material surrounding the central supermassive black hole (SMBH; e.g., I. M. George & A. C. Fabian [14]; K. Nandra et al. [43]). In particular, the narrow component (FWHM ) is generally attributed to fluorescence from distant, cold gas (e.g., Xrism Collaboration et al. [58]. Accurately determining the spatial location and physical nature of this Fe-K reprocessor is fundamental to mapping the circumnuclear structure of AGNs.
To constrain the reprocessor scale, spectroscopic line-profile analysis provides a direct and powerful method. The Resolve microcalorimeter (Y. Ishisaki et al. [22]; R. L. Kelley et al. [27]) onboard the X-Ray Imaging and Spectroscopic Mission (XRISM; M. Tashiro et al. [50]) has achieved an energy resolution better than 5 eV across the Fe K band, corresponding to a velocity resolution of a few hundred km s. This unprecedented spectral precision allows us to dissect the complex profile of the Fe K emission line and unambiguously disentangle distinct physical components, such as cold reflection from the distant torus and broader fluorescence arising from the BLR or photoionized winds (e.g., Xrism Collaboration et al. [58]; D. Bogensberger et al. [6]; E. Kammoun et al. [26]; C. Li et al. 2026; K. Fujiwara et al. [11]). Consequently, XRISM/Resolve enables simultaneous and self-consistent mapping of both the outer circumnuclear structure and the inner sub-parsec accretion dynamics.
IC 4329A (; C. N. A. Willmer et al. [57]) is one of the X-ray brightest Seyfert 1 galaxies in the local Universe. Due to its minimal line-of-sight intrinsic absorption and prominent, narrow Fe K emission, it offers an ideal laboratory for probing the unabsorbed continuum and its reprocessed radiation from the surrounding structures. Based on fundamental plane studies (M. J. Hudson et al. [20]; M. Bernardi et al. [4]), its distance is estimated to be Mpc (R. B. Tully et al. [53]). Optical reverberation mapping constraints yield a black hole mass of , a BLR inclination angle of , and an opening angle of (M. C. Bentz et al. [3]). Its bolometric luminosity is about erg s (M. J. Koss et al. [30]), which corresponds to of the Eddington luminosity.
Despite extensive study, the precise radial location of the Fe K reprocessor in IC 4329A remains debated. X-ray reverberation mapping attempts have yielded conflicting results, largely due to differences in temporal sampling and observation epochs. Y. Fukazawa et al. [12] reported significant Fe K flux variations (–) on timescales of days, suggesting a reprocessor location at several light-years. Conversely, C. Andonie et al. [1] identified short-term line variability on timescales of days, placing an upper limit of light-days on the reprocessor radius, which points toward the outer BLR or the inner edge of the torus. High-resolution spectroscopy with XRISM/Resolve is therefore essential to resolve this long-standing ambiguity.
In this paper, we present high-resolution X-ray spectroscopy of IC 4329A obtained with XRISM/Resolve. To securely constrain the broadband continuum and high-energy reflection features (e.g., the Compton hump), simultaneous hard X-ray observations were conducted with NuSTAR. Furthermore, to investigate the dynamical link between X-ray and other bands, we performed quasi-simultaneous optical and near-infrared (NIR) high-resolution spectroscopy using Seimei/KOOLS-IFU (M. Kurita et al. [31]; K. Matsubayashi et al. [35, 36]) and Magellan/WINERED (Y. Ikeda et al. [21]; S. Otsubo et al. [46]).
The structure of this paper is organized as follows. In Section 2, we detail the X-ray, optical, and NIR observations and their respective data reduction procedures. Section 3 presents the spectral modeling of the X-ray Fe K complex alongside the optical/NIR emission line profiles. In Section 4, we discuss the physical origins of the redshifted xBLR, the narrow Fe K line, and the detected ultra-fast outflow (UFO) features. Finally, our conclusions are summarized in Section 5. Throughout this paper, statistical uncertainties are quoted at the 90% confidence level unless stated otherwise.
Observations and data reduction
XRISM and NuSTAR
XRISM observations were carried out from 2025-08-07 to 11, accumulating a net exposure time of 146 ks (OBSID: 201044010). During this observation, the Resolve instrument (R. L. Kelley et al. [27]; Y. Ishisaki et al. [22]) was operated in the OPEN filter wheel position with the gate valve closed. A portion of the XRISM observation was conducted simultaneously with NuSTAR (F. A. Harrison et al. [18]), yielding a net exposure time of 26 ks (OBSID: 91101617002).
Data reduction and spectral analysis were performed using HEASoft v6.36 (NASA High Energy Astrophysics Science Archive Research Center (HEASARC) [44]; J. K. Blackburn et al. [5]). Although data processing generally followed standard pipelines, standard processing partially failed due to telemetry anomalies encountered during the observation. We therefore performed customized data reduction and reprocessed the Resolve event data following instructions provided by the XRISM Science Data Center1. The data were processed using pre-pipeline version 005_004.28Feb2025_V002.015, pipeline script version 03.02.014.010, and CALDB version 20260315. Only High-resolution Primary (Hp) events were selected in the Resolve spectroscopy. The Response Matrix File (RMF) was generated using the extra-large (“X”) option to fully incorporate instrumental effects. The response files were created based on the cleaned event file with Low-resolution Secondary (Ls) events excluded, in order to avoid potential contamination from pseudo-events2. The 2.4–12 keV energy band was used for the spectral fitting. Throughout this paper, spectral fitting was performed using optimal binning scheme (J. S. Kaastra & J. A. M. Bleeker [24]) alongside -statistics (W. Cash [7]), although the visual binning in certain figures has been modified for illustrative clarity. The Non-X-ray Background (NXB) for Resolve was estimated using the NXB database and modeled accordingly3. Both the source and NXB models were fitted simultaneously to the observed spectrum. For visual clarity, NXB-subtracted spectra are presented in all spectral figures in this work.
For the NuSTAR data analysis, source photons were extracted from a circular region with a radius of 50″, while the background was extracted from a surrounding annulus with inner and outer radii of 150″ and 250″, respectively. The CALDB version 20250812 was used. Above 35 keV, the NuSTAR spectrum becomes dominated by background; hence, higher-energy data were omitted from our analysis.
To assess the goodness of fit, we computed the expected value () and variance () of the -statistic for each best-fit model following J. S. Kaastra [23]. Instead of the approximate formulae given there, we evaluated and exactly by summing the Poisson distribution of the predicted model counts in each spectral bin. A fit is regarded acceptable when the observed -statistic is within a few () of . Strictly, this method applies to Poisson data without background subtraction, which holds for the Resolve spectrum because its NXB is modeled rather than subtracted. For the NuSTAR spectra, for which background spectra are used (i.e., effectively the -statistic), we applied the same calculation as an approximation; this is justified because the background contributes only of the total counts in the 3–35 keV band.
Seimei/KOOLS-IFU
Optical integral-field spectroscopy was carried out on 2025 April 26, approximately three months prior to the XRISM campaign, using the 3.8 m Seimei Telescope (M. Kurita et al. [31]) equipped with KOOLS-IFU (K. Matsubayashi et al. [35, 36]) under program IDs 25A-N-CT01 and 25A-K-0012. KOOLS-IFU utilizes a 110-fiber array providing an 8.4″ × 8.0″ field of view, corresponding to a physical scale of approximately 2.4 × 2.3 kpc at Mpc. We employed the VPH495 grism covering 4300–5900 Å with a spectral resolving power of , corresponding to a velocity resolution (FWHM) of ∼ 200 km s. The total on-source exposure time for IC 4329A was 900 s (300 s × 3).
Data reduction was conducted using IRAF (D. Tody [51, 52]) and the standard KOOLS-IFU pipeline. Basic calibration included bias subtraction, flat-fielding, cosmic-ray removal, and spatial fiber-distortion correction using arc lamp and flat-field frames. Sky background was subtracted using adjacent sky exposures of equal duration. One-dimensional spectra were extracted by co-adding the spectra from fibers within a 6″-diameter aperture centered on the centroid, using equal weights. Absolute flux calibration was established using the spectrophotometric standard star HR 5501 (observed with an exposure of 5 s × 10). For visual clarity in presentation, the spectrum was lightly smoothed to suppress high-frequency noise, although all spectral fitting was performed on the unsmoothed data.
Magellan/WINERED
Near-infrared (NIR) high-resolution spectroscopy was obtained using the Warm INfrared Echelle spectrograph to Realize Extreme Dispersion and sensitivity (WINERED; Y. Ikeda et al. [21]; S. Otsubo et al. [46]) mounted on the 6.5 m Clay/Magellan II telescope at Las Campanas Observatory. WINERED covers 9100–13500 Å. We adopted the WIDE mode with a 0.60″ slit width (4.0 pixels), yielding a resolving power of ( km s). Observations were carried out on 2025 August 14, three days after the XRISM campaign. Target exposures for IC 4329A consisted of 60 s × 4 and 120 s × 2 (total exposure of 480 s). Sky background was subtracted using off-source sky frames taken with identical exposure times. Data reduction was executed using the automated WARP pipeline (S. Hamano et al. [17]). Our analysis focuses on the 12000–13200 Å region (orders 12 to 46), encompassing the Pa line.
Telluric absorption features in orders 43–46 were corrected using TerraPCA4 (N. Matsunaga et al. [37]), a Principal Component Analysis framework that models complex telluric variations using standard star observations with intrinsic stellar lines masked out. Because order 42 lies outside the spectral coverage of TerraPCA, its telluric features were corrected using the telluric task in IRAF with the A-type standard star HD 160365 (exposure 2 s 4). The spectra were normalized per order, telluric-corrected, restored to their flux scale, stitched together, and normalized across orders 42–46 with a linear continuum fit. The stitched spectrum was smoothed for presentation purposes only (Figure 1b).

Figure 1. (a) Optical H spectral region observed with Seimei/KOOLS-IFU. Black data points show the observed spectrum. Red, blue, and orange lines correspond to the total best-fit model, the narrow H component, and the red-shifted broad H component, respectively. Gray lines represent other decomposed components (e.g., host galaxy, AGN continuum, and [O III] lines). (b) High-resolution NIR spectrum around Pa obtained with Magellan/WINERED. Colored data points display individual echelle orders. The solid black line represents the stitched and smoothed spectrum after subtracting narrow emission lines (Pa, [Fe II] , and [S IX] ), which are indicated by dashed lines.
Results
Optical/NIR spectra
First we study the optical H and NIR Pa broad lines. We performed spectral decomposition of the optical Seimei/KOOLS-IFU spectrum using pyQSOFit (H. Guo et al. [16]; Y. Shen et al. [48]; W. Ren et al. [47]). The rest-frame spectrum was decomposed into host-galaxy light and AGN continuum via PCA. After subtracting the host galaxy, AGN continuum, and optical Fe II empirical template, the emission-line complex around H and [O III] was fitted with multiple Gaussian components: three Gaussians for H and two Gaussians for each [O III] line.
Figure 1(a) illustrates the velocity profile, where zero velocity (0 km s) is defined by the peak of the narrow H component. The broad H emission component exhibits a redshift of +1050 460 km s, with the line broadening of FWHM = 9700 1000 km s.
Figure 1(b) presents the NIR Magellan/WINERED spectrum focusing on Pa. To measure the broad-line centroid velocity , we evaluated profile symmetry by comparing the observed profile within km s against its velocity-reversed profile . The optimal velocity shift was determined by maximizing the Pearson correlation coefficient between and , with uncertainties (90% confidence level) estimated via Monte Carlo simulations. This analysis yields a broad Pa redshift of km s. The line broadening is FWHM 6000 km s.
In summary, both the optical and NIR datasets consistently reveal a velocity offset of approximately 600–1000 km s in the BLR. While the estimated FWHM values differ between the two lines, literature
values have been reported as 6472 km s (observed on 2004-05-21; M. Koss et al. [29]) and 5906 656 km s (observed on 2012-04-16; J. E. Mejía-Restrepo et al. [39]). This comparison implies that the FWHM derived from the Seimei data is likely biased toward higher values, possibly because the relatively low S/N of the spectrum prevents the extended broad component from being tightly constrained.
X-ray spectra
Figure 2 displays the broad-band X-ray spectra overlaid with the baseline continuum model. The continuum is represented by a power law fitted outside the Fe-K band (6–7 keV). The best-fit photon index is , and the 2–10 keV unabsorbed flux is erg cm s, consistent with historical values for this source. Focusing on the Fe-K energy band, we clearly detect a narrow Fe K core accompanied by a broader component that is slightly redshifted. Furthermore, the Fe XXV forbidden line () at 6.64 keV, the Fe XXVI Ly lines at 6.95–6.97 keV, and the Fe K emission line at 7.06 keV are resolved. Outside the Fe-K band, the spectrum is well described by a power-law continuum; notably, the Compton hump in the NuSTAR band is remarkably weak.
We performed detailed spectral fitting within the XRISM/Resolve bandpass (2.4–12 keV). The narrow Fe K and Fe K emission lines from the distant cold material are modeled using the xclumpy model convolved with a Gaussian velocity profile (gsmooth). The xclumpy model was originally developed by A. Tanimoto et al. [49] and subsequently extended to cover high-resolution Resolve spectra (Xrism Collaboration et al. [59]). The Fe K line is modeled by seven Lorentzians (G. Hölzer et al. [19]). We adopted the solar abundance table from K. Lodders et al. [33]. The parameters are equatorial column density , torus angular width , inclination angle , and photon index . Here we assume that the torus and BLR share the same inclination angle (40 deg; M. C. Bentz et al. [3]) and the torus angular width is 22.1 deg based on the X-ray broad band fitting (S. Ogawa et al. [45]). Next, to account for the highly ionized Fe XXV and Fe XXVI lines, positive Gaussian profiles taking into account cosmological redshift were included. For the Fe XXV triplet, only the forbidden line () was incorporated, as adding the intercombination () and resonance () lines yielded normalizations consistent with zero within the confidence level. For Fe XXVI Ly, individual Gaussian components were introduced, with the normalization of Ly tied to half that of Ly, and their line widths linked together. The line widths of Fe XXV and Fe XXVI were fitted independently. Additionally, two ultra-fast outflow (UFO) zones manifest as absorption features at keV and keV. It should be noted that A. Markowitz et al. [34] also reported a UFO absorption line but different energy, at 7.68 keV (rest frame), with and , using the 2003 XMM-Newton observation. Because the continuum of IC 4329A is well approximated by a power law, we employed the photoionized plasma absorption table generated by Y. Mochizuki et al. [41] using XSTAR v2.58e (T. Kallman & M. Bautista [25]), assuming a power-law SED with . In this model, the column density (), ionization parameter (), outflow velocity (), and velocity dispersion () were treated as free parameters. While we can see a hint of absorption lines at keV, the residuals do not show

Figure 2. Broadband X-ray spectra of IC 4329A overlaid with the baseline power-law continuum model. Black, red, and blue points represent the XRISM/Resolve, NuSTAR/FPMA, and NuSTAR/FPMB data, respectively. The lower panels show the residuals. The lower figure shows the inset around the Fe-K band.
highly significant deviations, with (see also § 4.4).
The central objective of this work is the physical interpretation of the broad, redshifted Fe K feature. We evaluated three candidate scenarios: (1) a redshifted xBLR, (2) relativistic accretion disk reflection, and (3) an unshifted xBLR combined with a mildly ionized absorber.
For all models considered below, the expected -statistic is , which is essentially determined by the number of spectral bins and is insensitive to the model details.
Redshifted xBLR model
First, we modeled the broad Fe K emission using an xBLR model convolved with a Gaussian kernel. The xBLR model was originally developed by K. Fujiwara et al. [11] and Y. Nakatani et al. submitted, and was tailored to match the parameters of IC 4329A. In this model, a smooth, partially spherical geometry with two polar cuts is assumed. Based on optical reverberation mapping constraints (M. C. Bentz et al. [3]), the BLR inclination angle was fixed to and the half opening angle was fixed to (corresponding to a BLR angular width of ). To accommodate the redshifted broad feature, a velocity shift was introduced. The equatorial column density , velocity shift , and velocity dispersion were treated as free parameters for the xBLR model. The model function is
The best-fit parameters are summarized in Table 1 and shown in Figure 3. The width of the narrow Fe K core corresponds to a velocity dispersion of (FWHM ). The broad Fe K component exhibits a significant redshift of , which is slightly larger than the one in the optical data. Its velocity dispersion is (FWHM ), which is consistent with or slightly smaller than the one in the optical data. The fit quality is good: the observed -statistic of 3074.05 for 3089 degrees of freedom is fully consistent with the expected value of , deviating by only .
| Component | Parameter | Best-fit Value |
| Continuum | ||
| norm | ||
| Torus (XCLUMPY) | ||
| (fixed) | ||
| (fixed) | ||
| (fixed) | ||
| xBLR | ||
| (fixed) | ||
| (fixed) | ||
| Fe XXV He | (fixed) | |
| (fixed) | ||
| norm | ||
| Fe XXVI Ly | (fixed) | |
| (fixed) | ||
| norm | ||
| Fe XXVI Ly | (fixed) | |
| Same as Ly | ||
| (fixed) | ||
| norm | ||
| UFO 1 | ||
| UFO 2 | ||
| Const. | ||
| -statistic / d.o.f. | / | |
| Expected -statistic |
Table 1. Best-fit spectral parameters for the redshifted xBLR model.

Figure 3. X-ray spectra of IC 4329A fitted with the redshifted xBLR model. (Top) XRISM/Resolve spectrum in the Fe K band. Green, magenta, and blue lines represent the torus reflection, redshifted xBLR emission, and ionized Fe emission lines, respectively, added to the power-law continuum. The red line indicates the total model. (Middle) Broadband fit extended to the NuSTAR FPMA (red) and FPMB (blue) data. (Bottom) Zoomed-in view of the UFO absorption features in the 7.5–10 keV range. The blue solid line shows the best-fit model including UFO absorbers, while the blue dashed line illustrates the continuum model without UFO absorption. The residuals against the continuum model is shown in the lower panel.
The lower panel in Figure 3 shows the UFO features. To account for the broad negative residuals spanning 8.5–9.5 keV, it was necessary to introduce at least two distinct velocity components of relatively broad UFO absorbers. Both components have ionization parameters of , where absorption features from Fe XXV predominate. Compared to UFOs previously detected in local Seyfert 1 galaxies (S. Yamada et al. [60]), these absorbers exhibit relatively low and modest . Their statistical significance is discussed in § 4.4.
Disk line model
Next, we replaced the redshifted xBLR component with a relativistic disk reflection model, RELXILL v2.9.0 (T. Dauser et al. [8]; J. García et al. [13]). The emissivity index, black hole spin, outer disk radius, and iron abundance were fixed to , , , and , respectively. The disk inclination angle , inner disk radius , ionization parameter , and reflection fraction were allowed to vary. The parameters for the highly ionized emission lines and UFO absorbers were fixed to the values obtained in the redshifted xBLR model. The model function is
The inner disk radius is primarily constrained by the extent of the red wing, whereas the blue wing cutoff () is jointly determined by and . The fit favors a low inclination angle of (Table 2 and Figure 4). This value differs substantially from the BLR inclination angle of derived from optical reverberation mapping, implying a warped disk structure if this scenario holds. Furthermore, because does not extend down to the innermost stable circular orbit, the inner accretion disk may either be truncated or fully ionized, suppressing iron fluorescence in the innermost region (e.g., [42]). The fit quality is also good ( for 3088 d.o.f., compared with ; ). We note that fixing the spin parameter at (maximum spin Kerr BH) does not significantly change the results, including .
| Component | Parameter | Best-fit Value |
| Continuum | ||
| norm | ||
| Torus (xclumpy) | (cm) | |
| (deg) | 22.1 (fixed) | |
| (deg) | 40 (fixed) | |
| (km s) | 0 (fixed) | |
| (km s) | ||
| Disk Reflection (relxill) | (deg) | |
| () | ||
| 1 (fixed) | ||
| reflection fraction | ||
| Const. | ||
| -statistic / d.o.f. | ||
| Expected -statistic |
Table 2. Best-fit spectral parameters for the relativistic disk reflection model.

Figure 4. Same as Figure 3 (top), but for the relativistic disk reflection model (RELXILL). The orange line represents the disk line reflection profile.
Absorbed xBLR model
Third, we considered a scenario where the blue wing of an unshifted broad Fe K line is partially suppressed by a mildly ionized absorber. A similar feature was detected in NGC 4388 by K. Fujiwara et al. [11], where an absorption feature near 6.5 keV distorted the xBLR profile. Following this result, we applied a mildly ionized absorber to an unshifted xBLR component.
This fitting also yields a statistically good fit ( for 3086 d.o.f., compared with ; ; Table 3 and Figure 5). The velocity of the mildly ionized absorber is km s; it has no radial motion or is slightly inflowing. The column density and ionization parameter are consistent with those of typical warm absorbers (e.g., [60]).

Figure 5. Same as Figure 3 (top), but for the absorbed BLR model.
| Component | Parameter | Best-fit Value |
| Continuum | ||
| norm | ||
| Torus (XCLUMPY) | (cm) | |
| (deg) | 22.1 (fixed) | |
| (deg) | 40 (fixed) | |
| (km s) | 0 (fixed) | |
| (km s) | ||
| xBLR | (cm) | |
| (deg) | 21 (fixed) | |
| (deg) | 40 (fixed) | |
| (km s) | 0 (fixed) | |
| (km s) | ||
| Mildly Ionized Absorber | (cm) | |
| (km s) | ||
| (km s) | ||
| Const. | ||
| -statistic / d.o.f. | 3083.35 / 3086 | |
| Expected -statistic |
Table 3. Best-fit spectral parameters for the absorbed xBLR model.
Hybrid model, or optically constrained xBLR and disk line model
Finally, we assumed that the optical/NIR BLR and xBLR share identical velocity widths and velocity shifts. Based on § 3.1, the FWHM was fixed at 6000 km s, while the velocity shift () was allowed to vary within the range of 600–1000 km s during the spectral fitting. Under this assumption, the Fe-K line
profile cannot be fully reproduced; thus, we also introduced the relativistic disk reflection component. The model function corresponds to Equation (1) with the addition of the RELXILL component.
Table 4 and Figure 6 show the fitting results. The EW of the broad component is predominantly accounted for by the xBLR component, with the remaining residuals being filled by a strongly redshift-skewed disk line component. To accommodate the required stronger gravitational redshift, the inner radius () shifts to a smaller value compared to that in the disk-line model. The inclination angle remains consistent with a nearly face-on geometry. The torus component has FWHM = km s.
| Component | Parameter | Best-fit Value |
| Continuum | ||
| norm | ||
| Torus (xCLUMPY) | (cm) | |
| (deg) | (fixed) | |
| (deg) | (fixed) | |
| (km s) | (fixed) | |
| (km s) | ||
| xBLR | (cm) | |
| (deg) | (fixed) | |
| (deg) | (fixed) | |
| (km s) | ||
| (km s) | (fixed) | |
| Disk Reflection (relxill) | (deg) | |
| () | ||
| (fixed) | ||
| reflection fraction | ||
| Const. | ||
| -statistic/d.o.f. | ||
| Expected -statistic |
Table 4. Best-fit spectral parameters for the hybrid model.

Figure 6. Same as Figure 3 (top), but for the hybrid model. The magenta and orange lines represent the xBLR and disk reflection components, respectively.
The hybrid model yields for 3086 d.o.f., which is again consistent with (). Although the -statistic improves by –40 relative to the other models, the observed -statistics of all four models lie within of the expected value (Table 1–4). Consequently, distinguishing between these models is not feasible based solely on the current X-ray spectral fitting. We note that these models are not nested, so the values cannot be directly converted into formal significances.
Discussion
Plausible origin of the red wing in the Fe K line
We evaluated four candidate scenarios: (1) a redshifted xBLR, (2) relativistic accretion disk reflection, (3) an unshifted xBLR combined with a mildly ionized absorber, and (4) a hybrid model. While all the models provide statistically acceptable fits to the X-ray spectrum alone, the redshifted xBLR scenario or the hybrid model offer the most physically self-consistent picture when integrated with our optical and NIR spectroscopic findings. The primary support for the redshifted xBLR origin is the consistent redshift detected across multiple wavelength regimes. Both optical (H) and NIR (Pa) broad emission lines show velocity redshifts of – km s (§ 3.1). It is natural to infer that the xBLR gas originates from the same dynamical structure responsible for the optical/NIR broad lines. Recent comparative analyses combining XRISM data with ground-based optical/NIR spectroscopy have similarly revealed matching velocity profiles between xBLR and optical/NIR BLRs in other AGN (e.g., F. Zaidouni et al. [62], H. Noda et al. in prep.). While the other scenarios cannot be ruled out, one would be forced to invoke an ad hoc explanation for why the centroid velocity of the broad Fe K line differs markedly from its optical/NIR counterparts.
Asymmetric and shifted broad-line profiles are increasingly recognized as a ubiquitous feature in AGNs, particularly in hard X-ray-selected sources operating at moderate-to-low accretion rates (C. Ward et al. [56]). Such asymmetries are commonly attributed to Keplerian motion within non-axisymmetric accretion structures, such as spiral density waves, localized hot spots, or asymmetric wind outflows. Furthermore, optical and NIR interferometric studies by GRAVITY Collaboration et al. [15] have resolved distinct spatial photocenter shifts and velocity offsets within the BLR in IC 4329A, pointing toward strong dynamical influences from radiation pressure and non-axisymmetric gas distributions.
Size scale of xBLR
To quantitatively evaluate the scale of the xBLR, we estimate its characteristic radius () from the broad Fe K line width. The physical radius is expressed as:
where is a fudge factor that depends on the assumed BLR model (R. Decarli et al. [9]). Here, denoting as the one-dimensional (line-of-sight) turbulent velocity dispersion, the net observable velocity width can be written as
In a disk with finite vertical scale height , hydrostatic equilibrium relates the vertical (one-dimensional) turbulent velocity to the rotational velocity via (R. Decarli et al. [9]). This yields
Identifying the observed velocity with , the fudge factor is calculated to be
in agreement with Equation (4) of R. Decarli et al. [9]. With deg and , . Following the redshifted xBLR model, the line width is FWHM = km s. With , light-days pc. Based on the hybrid model, in which the line width is fixed at FWHM km s (§ 3.2.4), light-days pc.
Narrow Fe Kα line
The line width of the Fe Kα emission was constrained using the Chandra/HETG data (see e.g., [61, 38]. We have revisited the HETG spectrum of IC 4329A, which combines six observations conducted between 2002 and 2018, yielding a total net exposure time of 230 ks. The spectrum in the 2–7.5 keV energy range was analyzed using a baseline model consisting of a power-law continuum and zbfeklor, which is a seven Lorentzian approximation model to the line profile of Fe Kα from neutral material [19] with line broadening and redshift. The best-fit results yielded an Fe Kα line width of FWHM km s. This value lies intermediate between the narrow and broad components resolved by XRISM. This indicates that due to the spectral resolution limits of the HETG, the two components were indistinguishable, resulting in a single line fit over an unresolved, blended profile.
Next, we discuss the radius of the torus. Assuming Equation (5) holds for the torus region as well, X-ray spectroscopy can probe not only the dense torus body itself but also sub-structures such as an inner fountain flow or a dusty wind (polar dust). According to the radiation fountain flow model, the outflow velocity reaches up to 500 km s at smaller radii and decelerates to approximately 100 km s toward the outer region [55]. Determining the observable line width of the Fe Kα emission in a dynamical torus model requires detailed radiative transfer calculations. For simplicity, here we consider isotropic motion in the torus and adopt in Equation (5). Based on the hybrid model with FWHM km s, the torus radius is estimated to be light-days pc.
The inner radius of the dust torus () is determined by the dust sublimation temperature, K [2]. Dust evaporates when the irradiation flux reaches , where is the Stefan-Boltzmann constant. Therefore, . It is calculated to be light-days or 0.4 pc in IC 4329A [1]. When we simply assume a circular cross-section torus, it has a self-shielding radius (), as , where is the torus angular width [40]. In IC 4329A, pc. Our estimate of is comparable to or slightly larger than (consistent within the uncertainties), which provides a consistent picture if the narrow Fe Kα line broadly traces the torus structure.
Significance of the UFO features
To assess the statistical significance (or local null-hypothesis probability) of the candidate ultra-fast outflows (UFOs) while properly accounting for trial factors, we implemented a Look-Elsewhere Effect (LEE) correction framework (e.g., [28]). The trial factor was empirically calibrated by generating a synthetic ensemble of 1000 Resolve spectra under the baseline continuum model devoid of any UFO absorption features. Subsequently, each simulated dataset was fitted with an additional xstar grid component, treating the column density, ionization parameter, and line velocity dispersion as free variables. We performed a systematic red-shift scan across the interval in 350 uniform steps, tracking the corresponding improvement in the fit statistic (). Figure 7 displays the resulting empirical cumulative probability distribution. Based on this benchmark, a feature is considered detected at the 99% significance level if the fit improvement satisfies . The inclusion of UFO1 and UFO2 yields and , respectively. Thus, both features comfortably exceed the threshold, establishing their detection at a confidence level exceeding 99% under the LEE criteria.

Figure 7. Empirical cumulative probability distribution of the improvement in fit statistic () obtained from 1000 simulated XRISM/Resolve spectra under the null hypothesis (no UFO absorption). Vertical dashed lines indicate significance thresholds corresponding to different local null probabilities.
Conclusion
We have performed multi-wavelength spectral analyses of the X-ray bright Seyfert 1 galaxy IC 4329A using XRISM/Resolve, NuSTAR, Seimei/KOOLS-IFU, and Magellan/WINERED data. Figure 8 shows the schematic picture of IC 4329A.

Figure 8. Schematic picture of the circumnuclear environment of IC 4329A, based on the hybrid model.
Based on the X-ray spectrum from XRISM/Resolve and NuSTAR, four physical scenarios were tested: (1) a redshifted xBLR model, (2) a relativistic disk reflection model, (3) an unshifted xBLR model affected by a mildly ionized absorber, and (4) a hybrid model. All provide statistically acceptable and comparable fits.
As a result of quasi-simultaneous optical (H) and NIR (Pa) spectroscopy, the optical/NIR broad line shows redshifts of km s. Thus, it is reasonable to expect that the xBLR exhibits a similar redshift.
The physical radius derived for the xBLR is –– light-days (FWHM – km s). Meanwhile, the line width of the narrow Fe K core in the hybrid model (FWHM km s) places the torus reflection radius at light-days pc, tracing fluorescence across the torus structure beyond the dust sublimation ( pc) and self-shielding ( pc) boundaries.
Two robust UFO absorption components are resolved at keV and keV, corresponding to ultra-fast outflow velocities of and . Both absorption features exceed a 99% confidence level threshold after accounting for trial factors via look-elsewhere effect Monte Carlo simulations.
Acknowledgments
This research has made use of the NuSTAR Data Analysis Software (NuSTARDAS) jointly developed by the ASI Space Science Data Center (SSDC, Italy) and the California Institute of Technology (Caltech, USA). WINERED was developed by the University of Tokyo and the Laboratory of Infrared High-resolution spectroscopy (LiH), Kyoto Sangyo University. The authors thank the staff of Las Campanas Observatory and Okayama Observatory for operations of the Seimei telescope and the Magellan telescope, respectively. The authors also thank Scarlet Elgueta and Matilde Coello-Guzmán for WINERED operation and the data collection at the Magellan telescope.
Part of this work was supported by JSPS KAKENHI Grant Number JP26K17197 (MM), JP26K17187 (SY), JP23K13153, JP24K00673 (TK) and Yamada Science Foundation (MM). Y.N. acknowledges financial support from Grant-in-Aid for JSPS Fellows (24KJ1507).
Machine learning tools were used to suggest language improvements within the manuscript. Responsibility for the final manuscript text lies entirely with the authors.
The optical/NIR data used in this paper are archived in https://github.com/mskimizu/IC4329A_paper_data. The X-ray data are archived in the HEASARC web page.
Author Contributions
MM led the project and was responsible for writing and submitting the manuscript. MM, YN, SY, YU, TK, SO, YM, and KF discussed interpretation of the XRISM data. MM, YN, SY, and FI performed the Seimei/KOOLS-IFU observations and its data reduction. SO, NM, DT, and HK performed the Magellan/WINERED observations and its data reduction. All authors were responsible for the manuscript proofing.
Facilities: XRISM, NuSTAR, Seimei (KOOLS-IFU), Magellan:Clay (WINERED)
Software: HEASoft/FTOOLS (NASA High Energy Astrophysics Science Archive Research Center (Heasarc) [44]; J. K. Blackburn et al. [5]), IRAF (D. Tody [51, 52]), pyQSOFit (H. Guo et al. [16]; Y. Shen et al. [48]; W. Ren et al. [47]), WARP (S. Hamano et al. [17]), TerraPCA (N. Matsunaga et al. [37])
Comparison with the Inner Radius Estimation Based on the rdblur Model
In the main text, the physical radii of the torus and xBLR are constrained based on Equation (4). On the other hand, the relativistic convolution model rdblur (A. C. Fabian et al. [10]) is frequently employed to constrain the characteristic radius in X-ray spectral analysis (e.g., Xrism Collaboration et al. 2024, 2026). The rdblur model computes the relativistic smoothing of an emission line from an accretion disk around a Schwarzschild black hole, parametrized by the emissivity index (where emissivity ), inner radius , outer radius , and inclination angle . When considering circumnuclear structures such as the torus or the xBLR, the emissivity index is conventionally fixed to (assuming uniform illumination), and the inclination angle is fixed to an independently constrained value. Because disk emissivity declines steeply with radius, emission from larger radii contributes less to the overall line broadening, making challenging to constrain from spectral fitting alone. Consequently, is typically fixed at an arbitrarily large value (e.g., or ). Under these assumptions, remains the sole free parameter governing relativistic line broadening. The objective of this section is to establish an empirical conversion relation between the Gaussian line width (FWHM) and the inner radius derived from rdblur, thereby enabling a direct and quantitative comparison between studies employing Gaussian velocity broadening and those adopting rdblur.
Assuming that the circumnuclear gas exhibits no turbulent motion, the fudge factor in Equation (4) can be parametrized as . Under this formulation, FWHM .
Equation (8) strictly holds if the emitting region is an infinitesimally thin rotating ring. However, under the realistic assumption where the material extends continuously from to , the power-law index of is expected to deviate from . Therefore, we adopt the generalized relation:
or equivalently,
We determine the normalization factor and the power-law index through numerical simulations. Spectral simulations were carried out using the fakeit command in XSPEC. Synthetic spectra were generated by convolving an intrinsically narrow (zero-width) Gaussian emission line with rdblur, and the resulting convolved profile was subsequently fitted with a standard Gaussian model to determine the velocity dispersion and the corresponding FWHM. This procedure was systematically performed over a grid of and . Figure 9 illustrates the results for and . The derived slopes are shallower than the theoretical value of . This shallower slope is attributed to the fact that, for a fixed , emission from larger radii (possessing lower orbital velocities) non-negligibly contributes to the composite line core, thereby narrowing the overall profile. This dilution effect becomes increasingly pronounced at smaller , where the dynamic range between and is larger. Consequently, we obtain and for , and and for .

Figure 9. FWHM of the rdblur-convolved line profile as a function of the inner radius for various inclinations (), simulated with (a) and (b) . The emissivity index is fixed to . Dashed lines represent the best-fit power-law relations.
In the hybrid spectral model presented in § 3.2.4, the narrow torus component exhibits a line width of FWHM . Substituting this value into Equation (10) yields an inner radius of for , or for . These values are one to two orders of magnitude smaller than the radius of the torus obtained in § 4.3 (). Indeed, represents the geometric mean of the inner and outer radii.
References
- [1]Andonie, C., Bauer, F. E., Carraro, R., et al. 2022, A&A, 664, A46, doi: 10.1051/0004-6361/202142473DOI
- [2]Barvainis, R. 1987, ApJ, 320, 537, doi: 10.1086/165571DOI
- [3]Bentz, M. C., Onken, C. A., Street, R., & Valluri, M. 2023, ApJ, 944, 29, doi: 10.3847/1538-4357/acab62DOI
- [4]Bernardi, M., Alonso, M. V., da Costa, L. N., et al. 2002, AJ, 123, 2990, doi: 10.1086/340463DOI
- [5]Blackburn, J. K., Shaw, R. A., Payne, H. E., Hayes, J. J. E., & Heasarc. 1999, FTOOLS: A general package of software to manipulate FITS files,, Astrophysics Source Code Library, record ascl:9912.002
- [6]Bogensberger, D., Nakatani, Y., Yaqoob, T., et al. 2025, PASJ, 77, S209, doi: 10.1093/pasj/psaf084DOI
- [7]Cash, W. 1979, ApJ, 228, 939, doi: 10.1086/156922DOI
- [8]Dauser, T., Garcia, J., Parker, M. L., Fabian, A. C., & Wilms, J. 2014, MNRAS, 444, L100, doi: 10.1093/mnrasl/slu125DOI
- [9]Decarli, R., Dotti, M., Fontana, M., & Haardt, F. 2008, MNRAS, 386, L15, doi: 10.1111/j.1745-3933.2008.00451.xDOI
- [10]Fabian, A. C., Rees, M. J., Stella, L., & White, N. E. 1989, MNRAS, 238, 729, doi: 10.1093/mnras/238.3.729DOI
- [11]Fujiwara, K., Ueda, Y., Ogawa, S., et al. 2026, ApJ, 1003, 6, doi: 10.3847/1538-4357/ae5e5cDOI
- [12]Fukazawa, Y., Furui, S., Hayashi, K., et al. 2016, ApJ, 821, 15, doi: 10.3847/0004-637X/821/1/15DOI
- [13]García, J., Dauser, T., Lohfink, A., et al. 2014, ApJ, 782, 76, doi: 10.1088/0004-637X/782/2/76DOI
- [14]George, I. M., & Fabian, A. C. 1991, MNRAS, 249, 352, doi: 10.1093/mnras/249.2.352DOI
- [15]GRAVITY Collaboration, Amorim, A., Bourdarot, G., et al. 2024, A&A, 684, A167, doi: 10.1051/0004-6361/202348167
- [16]Guo, H., Shen, Y., & Wang, S. 2018, PyQSOFit: Python code to fit the spectrum of quasars,, Astrophysics Source Code Library, record ascl:1809.008 http://ascl.net/1809.008
- [17]Hamano, S., Ikeda, Y., Otsubo, S., et al. 2024, PASP, 136, 014504, doi: 10.1088/1538-3873/ad1b38DOI
- [18]Harrison, F. A., Craig, W. W., Christensen, F. E., et al. 2013, ApJ, 770, 103, doi: 10.1088/0004-637X/770/2/103DOI
- [19]Hölzer, G., Fritsch, M., Deutsch, M., Härtwig, J., & Förster, E. 1997, Phys. Rev. A, 56, 4554, doi: 10.1103/PhysRevA.56.4554DOI
- [20]Hudson, M. J., Lucey, J. R., Smith, R. J., Schlegel, D. J., & Davies, R. L. 2001, MNRAS, 327, 265, doi: 10.1046/j.1365-8711.2001.04786.xDOI
- [21]Ikeda, Y., Kondo, S., Otsubo, S., et al. 2022, PASP, 134, 015004, doi: 10.1088/1538-3873/ac1c5fDOI
- [22]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
- [23]Kaastra, J. S. 2017, A&A, 605, A51, doi: 10.1051/0004-6361/201629319DOI
- [24]Kaastra, J. S., & Bleeker, J. A. M. 2016, A&A, 587, A151, doi: 10.1051/0004-6361/201527395DOI
- [25]Kallman, T., & Bautista, M. 2001, ApJS, 133, 221, doi: 10.1086/319184DOI
- [26]Kammoun, E., Kawamuro, T., Murakami, K., et al. 2025, ApJL, 994, L13, doi: 10.3847/2041-8213/ae18c2DOI
- [27]Kelley, R. L., Ishisaki, Y., Constantini, E., et al. 2025, Journal of Astronomical Telescopes, Instruments, and Systems, 11, 042026, doi: 10.1117/1.JATIS.11.4.042026DOI
- [28]Kosec, P., Pinto, C., Fabian, A. C., & Walton, D. J. 2017, Monthly Notices of the Royal Astronomical Society, 473, 5680, doi: 10.1093/mnras/stx2695DOI
- [29]Koss, M., Trakhtenbrot, B., Ricci, C., et al. 2017, ApJ, 850, 74, doi: 10.3847/1538-4357/aaa8ec9
- [30]Koss, M. J., Ricci, C., Trakhtenbrot, B., et al. 2022, ApJS, 261, 2, doi: 10.3847/1538-4365/ac6c05DOI
- [31]Kurita, M., Kino, M., Iwamuro, F., et al. 2020, PASJ, 72, 48, doi: 10.1093/pasj/psaa036DOI
- [33]Lodders, K., Palme, H., & Gail, H.-P. 2009, Landolt Börnstein, 4B, 712, doi: 10.1007/978-3-540-88055-4_34DOI
- [34]Markowitz, A., Reeves, J. N., & Braito, V. 2006, ApJ, 646, 783, doi: 10.1086/505107DOI
- [35]Matsubayashi, K., Ohta, K., Iwamuro, F., et al. 2019, PASJ, 71, 102, doi: 10.1093/pasj/psz087DOI
- [36]Matsubayashi, K., Tsutsui, H., Nakaya, H., et al. 2025, PASJ, 77, 1065, doi: 10.1093/pasj/psaf083DOI
- [37]Matsunaga, N., Tsujimoto, T., Taniguchi, D., et al. 2026, ApJS, 284, 49, doi: 10.3847/1538-4365/ae5a9dDOI
- [38]McKernan, B., & Yaqoob, T. 2004, ApJ, 608, 157, doi: 10.1086/383261DOI
- [39]Mejía-Restrepo, J. E., Trakhtenbrot, B., Koss, M. J., et al. 2022, ApJS, 261, 5, doi: 10.3847/1538-4365/ac6602
- [40]Mizumoto, M., Done, C., Tomaru, R., & Edwards, I. 2019, MNRAS, 489, 1152, doi: 10.1093/mnras/stz2225DOI
- [41]Mochizuki, Y., Mizumoto, M., & Ebisawa, K. 2023, MNRAS, 525, 922, doi: 10.1093/mnras/stad2329DOI
- [42]Nagele, C., Krolik, J. H., Kinch, B. E., & Schnittman, J. D. 2026, ApJ, 1005, 119, doi: 10.3847/1538-4357/ae6effDOI
- [43]Nandra, K., O’Neill, P. M., George, I. M., & Reeves, J. N. 2007, MNRAS, 382, 194, doi: 10.1111/j.1365-2966.2007.12331.xDOI
- [44]NASA High Energy Astrophysics Science Archive Research Center (Heasarc). 2014, HEASoft: Unified Release of FTOOLS and XANADU,, Astrophysics Source Code Library, record ascl:1408.004
- [45]Ogawa, S., Ueda, Y., Tanimoto, A., & Yamada, S. 2021, ApJ, 906, 84, doi: 10.3847/1538-4357/abccceDOI
- [46]Otsubo, S., Sarugaku, Y., Takeuchi, T., et al. 2024, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 13096, Ground-based and Airborne Instrumentation for Astronomy X, ed. J. J. Bryant, K. Motohara, & J. R. D. Vernet, 1309631, doi: 10.1117/12.3018617
- [47]Ren, W., Guo, H., Shen, Y., et al. 2024, ApJ, 974, 153, doi: 10.3847/1538-4357/ad6e76DOI
- [48]Shen, Y., Hall, P. B., Horne, K., et al. 2019, ApJS, 241, 34, doi: 10.3847/1538-4365/ab074fDOI
- [49]Tanimoto, A., Ueda, Y., Odaka, H., et al. 2019, ApJ, 877, 95, doi: 10.3847/1538-4357/ab1b20DOI
- [50]Tashiro, M., Kelley, R., Watanabe, S., et al. 2025, PASJ, 77, S1, doi: 10.1093/pasj/psaf023DOI
- [51]Tody, D. 1986, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 627, Instrumentation in astronomy VI, ed. D. L. Crawford, 733, doi: 10.1117/12.968154
- [52]Tody, D. 1993, in Astronomical Society of the Pacific Conference Series, Vol. 52, Astronomical Data Analysis Software and Systems II, ed. R. J. Hanisch, R. J. V. Brissenden, & J. Barnes, 173
- [53]Tully, R. B., Rizzi, L., Shaya, E. J., et al. 2009, AJ, 138, 323, doi: 10.1088/0004-6256/138/2/323DOI
- [54]Urry, C. M., & Padovani, P. 1995, PASP, 107, 803, doi: 10.1086/133630DOI
- [55]Wada, K., Schartmann, M., & Meijerink, R. 2016, ApJL, 828, L19, doi: 10.3847/2041-8205/828/2/L19DOI
- [56]Ward, C., Koss, M. J., Eracleous, M., et al. 2025, ApJ, 991, 116, doi: 10.3847/1538-4357/adfa25DOI
- [57]Willmer, C. N. A., Focardi, P., Chan, R., Pellegrini, P. S., & da Costa, N. L. 1991, AJ, 101, 57, doi: 10.1086/115665DOI
- [58]Xrism Collaboration, Audard, M., Awaki, H., et al. 2024, ApJL, 973, L25, doi: 10.3847/2041-8213/ad7397
- [59]Xrism Collaboration, Audard, M., Awaki, H., et al. 2026, Nature Astronomy, 10, 880, doi: 10.1038/s41550-026-02817-6DOI
- [60]Yamada, S., Kawamuro, T., Mizumoto, M., et al. 2024, ApJS, 274, 8, doi: 10.3847/1538-4365/ad5961DOI
- [61]Yaqoob, T., & Padmanabhan, U. 2004, ApJ, 604, 63, doi: 10.1086/381731DOI
- [62]Zaidouni, F., Kara, E., Kosec, P., et al. 2026, ApJL, 997, L26, doi: 10.3847/2041-8213/ae307dDOI