Detection of Asymmetry in the Narrow Fe Kα Emission Line in MCG-5-23-16 with Chandra

Iron Kα (Fe Kα) emission is observed ubiquitously in active galactic nuclei (AGN), and it is a powerful probe of their circumnuclear environment. Examinations of the emission line play a pivotal role in understanding the disk geometry surrounding black holes. It has been suggested that the torus and the broad-line region (BLR) are the origins of emission. However, there is no universal location for the emitting region relative to the BLR. Here, we present an analysis of the narrow component of the Fe Kα line in the Seyfert AGN MCG-5-23-16, one of the brightest AGN in X-rays and in Fe Kα emission, to localize the emitting region. Spectra derived from Chandra/HETGS observations show asymmetry in the narrow Fe Kα line, which has only been confirmed before in the AGN NGC 4151. Models including relativistic Doppler broadening and gravitational redshifts are preferred over simple Gaussians and measure radii consistent with R ≃ 200–650 r g . These results are consistent with those of NGC 4151 and indicate that the narrow Fe Kα line in MCG-5-23-16 is primarily excited in the innermost part of the optical BLR, or X-ray BLR. Characterizing the properties of the narrow Fe Kα line is essential for studying the disk geometries of the AGN population and mapping their innermost regions.


INTRODUCTION
Iron Kα emission lines (Fe Kα) at 6.4 keV are observed ubiquitously in AGN (Nandra & Pounds 1994;Weaver et al. 2001;Shu et al. 2010).The line is produced in dense and cold material when illuminated by an X-ray source, making it a powerful probe of circumnuclear environments in AGN.A significant portion of the information we can learn about the inner regions of AGNs and their geometries comes from modeling the line and understanding what emission processes may be responsible for it.
However, the location where the Fe Kα line is emitted is not well established.It is thought that the dust sublimation radius forms an outer envelope to the emitting region (Netzer & Laor 1993;Czerny & Hryniewicz 2011;Gandhi et al. 2015;Baskin & Laor 2017).Although Nandra (2006) found no correlation between the width of the Fe Kα and the optical Hβ line in a sample of sources, implying that the Fe Kα line originates in the torus and not the optical BLR, other observations suggest otherwise (e.g Bianchi et al. (2008)).Shu et al. (2010) presented an extensive analysis of the Chandra grating spectra of 36 sources, finding that there is no universal location for the Fe Kα line-emitting region relative to the optical BLR.They found that the Fe Kα line may have contributions from parsec-scale distances from the black hole (i.e. the torus), down to matter on the optical BLR scale or smaller.
Direct measurements of the location of the line emitting regions was recently obtained for NGC 4151, the brightest Seyfert galaxy in X-rays, and hence the AGN with the brightest narrow Fe Kα line.Using Chandra grating spectra, Miller et al. (2018) found that the narrow Fe Kα line in NGC 4151 is asymmetric, with the data suggesting that the emission could originate in a region as small as 50-130 r g during high flux intervals1 .On a parallel front, Zoghbi et al. (2019) measured the time delay between the narrow Fe Kα line and the hard X-ray continuum in NGC 4151 to constrain the location of the narrow Fe Kα region-a technique called reverberation mapping-and found a time delay of 3.4 +2.5  −0.8 days.Bentz et al. (2006) conducted a similar reverberation mapping study 13 years prior, but of the optical BLR in NGC 4151, and found a time delay of 6.6 +1.1 −0.8 days between the optical Hβ line and the continuum.These results indicate that the narrow Fe Kα line in NGC 4151 is primarily excited in either the innermost part of the optical broad line region (BLR), the X-ray BLR, or closer.The central question is now whether other AGN exhibit asymmetry and reverberation in their narrow Fe Kα lines, which will crucially inform us about where the line is emitted more broadly in the population of all AGN.
Before proceeding further, we wish to make a distinction between the narrow Fe Kα line and the broad Fe Kα line.The narrow component of the line originates at large distances from the black hole (∼ 100's to 1000's of r g ) while the broad component originates less than 10 r g in the inner disk (Zoghbi et al. 2013;Uttley et al. 2014).While asymmetry and reverberation delays have been seen before in several sources' broad Fe Kα lines, in this study, we are focusing on only the narrow Fe Kα line.
Among the next brightest sources in the Fe Kα band is MCG-5-23-16 (z = 0.00849).MCG-5-23-16 is a Seyfert 1.9-AGN (Veron et al. 1980) with a central black hole mass of 10 7.9 M ⊙ (Ponti et al. 2013).There have been many studies of the narrow and broad components of the Fe Kα line in MCG-5-23-16 (e.g. Reeves et al. (2006, 2007); Zoghbi et al. (2014Zoghbi et al. ( , 2017))), but none using the high-energy resolution afforded by Chandra, which is crucial in order to accurately model the narrow Fe Kα line.In this paper, we use 12 Chandra observations of MCG-5-23-16, nine of which were recently taken in 2020, to model the narrow Fe Kα line and address the question of whether MCG-5-23-16 also has asymmetry in its narrow Fe Kα line like NGC 4151.We find that MCG-5-23-16 does indeed have line asymmetry, and that the strength of this line asymmetry varies with time.
Here, we report on the narrow Fe Kα line and its model fit parameters in MCG-5-23-16 using observations with Chandra.

Our specific aims are:
1. To determine whether the narrow Fe Kα line in MCG-5-23-16 exhibits asymmetry, just like it does in NGC 4151.2. To examine any potential variability in the narrow Fe Kα line.3. To constrain the emission processes at play in the disk and the geometry of the AGN by trying out different models for the line and examining their best-fit parameters.

OBSERVATIONS AND REDUCTION
All archival Chandra/HETG observations of MCG-5-23-16 were downloaded directly from the Chandra archive.In addition to these archival observations, we observed MCG-5-23-16 nine times from October to November 2020 with the Chandra X-ray Observatory, Chandra (Weisskopf et al. 2000) using the Advanced CCD Imaging Spectrometer (ACIS) optimized for spectroscopy (ACIS-S) in faint mode using the High Energy Grating Transmission Spectrometer (HETGS).The full dataset used in this paper is available for download at DOI: 10.25574/cdc.186.The observation identification number (ObsID), start date, and duration of each exposure is given in Table 1.
HETGS observations typically provide data from both the high-energy gratings (HEG) and the medium-energy gratings (MEG).However, the MEG has less effective area in the Fe K band and lower resolution, and is therefore less suited to our analysis.As a result, we limited our analysis to only the HEG.
The standard CIAO tools (version 4.15.1) were used to reduce the Chandra observations (Fruscione et al. 2006).For each observation, with the exception of ObsID 2121, we ran the tool "chandra repro" to produce the necessary "evt2", "pha2", RMF and ARF files.For ObsID 2121, "chandra repro" produced evt2 and pha2 files with erroneous exposure times, so we manually generated the spectrum and filtered for bad grades and for a "clean" status column using the tools "tgdetect2", "tg create mask", "tg resolve events", "dmcopy", "dmappend", and "tgextract" as according to the CIAO HETG/ACIS-S Grating Spectra thread. 2 2 The RMF and ARF files were then generated using "mktgresp".
For each exposure, we added the first-order (+1 and -1) HEG spectra, RMF files, and ARF files using "combine grating spectra".We found that individual spectra lacked sufficient data to constrain spectral features and fit complex physical models.When fit to a continuum using the model zphabs*zpowerlw, individual spectra had a consistent continuum, with each spectra having a photon index value between 1.64 and 1.70.Therefore, we further combined the spectra using "combine spectra" into a "Total" spectrum, "2000,2005" spectrum and "2020" spectrum, corresponding to all 12 observations, the three observations in 2000 and 2005, and the nine observations taken recently in 2020 respectively.This grouping by time allows us to investigate how the Fe Kα line in MCG-5-23-16 has changed from the earlier observations in 2000 and 2005 to the recent observations in 2020.From henceforth, "the spectra" will refer to the "Total", "2000,2005", and "2020" groups.
The spectra were then binned to achieve a minimum signal-to-noise ratio (SNR) of 5.0 for each energy bin using the tool "ftgrouppha" within the standard HEASFOFT suite, version 6.31.1.We set the "grouptype" parameter within "ftgrouppha" to be "optsnmin", which uses the optimal binning algorithm designed by Kaastra & Bleeker (2016).
All spectra fits were performed using PyXspec, version 2.1.2(Gordon & Arnaud 2021), on the HEAsoft environment (Nasa HEAsarc 2014).Unless otherwise noted, the errors quoted in this work reflect the value of the parameter at its 1σ confidence interval.Errors were derived using the standard PyXspec "Fit.error()"command.The script used to perform the spectral fits and generate Figures 1 and 2 in this paper is available on GitHub3 and version v1 is archived in Zenodo (Liu 2023).

Spectral Fitting Range and Setup
The spectra in this work are restricted to the 1.5-8.0keV range due to the low SNR below and above these energies.Additional bad data points were ignored with the command "AllData.ignore('bad')".
We fit the spectra using three Gaussian models and three reflection models.Of the Gaussian models, the first model is Model A: zphabs*(zpowerlw+zgauss) with fixed σ=0, a narrow Gaussian model that probes whether the Fe Kα line originates at the torus or outer BLR and tests whether the line is resolved by Chandra/HETGS.The next model is Model B: zphabs*(zpowerlw+zgauss) with free σ, a broad Gaussian model that models emission from the intermediate regions of the AGN slightly closer to the cen- tral black hole.These first two models are both simple Gaussians and assume the Fe Kα line is symmetric.The next models we use are more physically motivated and account for potential asymmetry in the line.We next try Model C: zphabs*(zpowerlw+rdblur*zgauss), which fits to asymmetry in the line by taking into account relativistic Doppler broadening near the black hole with the "rdblur" component ("rdblur" is a convolution model extracted from the diskline model described in Fabian et al. (1989)).
We also attempt variants of Models A, B, and C, where instead of fitting a single Gaussian to the Fe Kα line, we fit a double Gaussian.Fitting a double Gaussian accounts for the fact that the Fe Kα has two line energies, one at 6.404 keV (Fe Kα1) and another at 6.391 keV (Fe Kα2).However, when we attempt these double Gaussian models, we find that the fits worsen compared to the single Gaussian models, likely due to the difference between these two line energies being unresolved in our spectra.
The "mytorus" component models Compton-scattering off of a toroidal structure around the disk.Finally, we use Model F: zphabs*(zpowerlw+rdblur*mytorus) to determine whether including relativistic Doppler broadening makes a difference to the fit.
For all models, we obtained the Fe Kα line flux by multiplying the component cflux by the last additive component (e.g.cflux*zgauss, cflux*rdblur*zgauss, cflux*xillver, cflux*mytorus and cflux*rdblur*mytorus).The E min and E max parameters for cflux were set to 6.2 and 6.45 keV, respectively, which generally bounded the entirety of the Fe Kα emission line.In all fits, the emission line energy was frozen at 6.4 keV, the intrinsic line energy.Within Model C, σ was fixed to 0 since the R in parameter from rdblur was responsible for capturing the width of the line instead of zgauss, and the outer radius of the disk R out was fixed at 10 6 r g to give the biggest range for R in to vary.When fitting to the "2000,2005" spectrum, letting inclination range from 0 • ≤ i ≤ 90 • produced a physically implausible lower bound of less than 0 • , so we restricted the inclination to range from 3 • ≤ i ≤ 87 • for the "2000,2005" spectrum.
Within Model D, the photon index of zpowerlw and xillver were linked to get a consistent photon index for the continuum; the ionization index logξ was fixed at 0 since the Fe Kα line is a neutral line; and the high energy cutoff E cut was fixed to 300 keV.In Model E and Model F, the hydrogen column density n H of mytorus was tied to the n H of zphabs, and the photon index of mytorus was linked to the photon index of zpowerlw.In Model E, the inclination was fixed to 60 • and in Model F, the inclination and line emissivity was free to vary.

Model A
Model A is zphabs*(zpowerlw+zgauss) with σ fixed to 0. In addition to testing whether the Fe Kα line emitting region originates at the torus or outer BLR, fixing sigma to 0 also examines whether the line width is below the instrumental resolution of Chandra/HETGS.Fixing σ = 0 eV assumes that the line is not resolved and that its width is due to instrumental broadening and not any physical effects from the AGN itself.Figure 1 shows that the line is clearly broader than the fit for the "2000,2005", "2020", and "Total" spectra, proving that the line is resolved.It also shows that the origin for the line is not located in the torus or outer BLR.These fits also illustrate the asymmetry of the Fe Kα line in MCG-5-23-16, as there is a clear red wing.

Model B
Model B is zphabs*(zpowerlw+zgauss) with free σ.In Model B, we let σ vary in order to fit a broad Gaussian to the line.For the "2000,2005" spectrum, this only marginally improves the fits, as the red wing of the line is still not captured (Figure 1).For the "2020" and "Total" spectra, however, letting σ vary vastly improves the fits, since most of the red wing is now captured.These results are supported by the F-tests in Table 3: the change from Model A to Model B for the "2000,2005" spectrum has a 1.95σ level of confidence, and a 5.33σ and 5.41σ level of confidence for the "2020" and "Total" spectra, respectively.
In the "2020" and "Total" spectra, the width is consistent with σ ≃ 22 eV, corresponding to a FWHM of 52 eV and a projected velocity of v ≃ 2427 km/s.These values are very similar to those obtained by Miller et al. (2018) for NGC 4151 -the width and FWHM for the broad Gaussian fits in their work is 23 eV and 54 eV, respectively, emphasizing the similarity between the Fe Kα line in  As stated in Miller et al. (2018), it's important to emphasize that the neutral iron Kα line is actually a composite of two distinct lines at laboratory energies of 6.391 keV and 6.404 keV (Bambynek et al. 1972).The difference between these lab energies constitutes just 25% of the measured FWHM = 52 eV, so it is unnecessary to model for both lines and modeling just a single broad Gaussian is sufficient.Noteworthy is the fact that the separation between these lines falls also below the predicted resolution of the first-order High Energy Grating (HEG) in the iron K band, which is around 45 eV, making modeling both lines infeasible.This decision is also followed in previous work on the narrow Fe Kα line in AGN (e.g.Shu et al. (2010) andMiller et al. (2018)).

Model C
Model C is zphabs*(zpowerlw+rdblur*zgauss).In Model C, we multiply zgauss with the rdblur convolution model, which introduces a relativistic Doppler broadening factor to the Gaussian (Fabian et al. 1989).This broadening factor creates a red tail in the Gaussian and thus fits to the asymmetry in the line.We first try fixing the line emissitivity at q = 3 (Model C.1), which models an isotropically radiative and flat accretion disk.
We find that the fit for the "2000,2005" spectrum significantly improves when using Model C.1.Visually, the red wing of the Fe Kα line is better captured by Model C.1 than all the previous models (top panel of Fig. 1) and the F-test between Model B and Model C.1 indicates an improvement in the fit at the 2.74σ level of confidence (Table 2).With this evidence, we take Model C.1 as the benchmark fit for the "2000,2005" spectrum.The fit for "2020" also improves when using Model C.1: visually, the red wing is captured fully by Model C.1 (middle panel of Fig. 1).However, since the amplitude of the red wing is much smaller in the "2020" spectrum, the fit with Model C.1 is not significantly better than the fit with Model B. More convincingly, the F-test between Model B and Model C.1 gives only a 1.97σ level of confidence (Table 2), statistically indicating that Model B sufficiently fits the line and that the asymmetry in the line for the "2020" spectrum is weak.The "2020" spectrum also gives much larger uncertainties in the inclination and inner radius R in than the "2000,2005" spectrum, further indicating that the asymmetry in the line for "2020" is much weaker than in the line for "2000,2005".The collected counts of the 2020 spectrum also are approximately 1.3 times higher than those in the 2000-2005 spectrum, lending further support to the weaker asymmetry of the Fe Kα line in "2020" compared to "2000,2005".Therefore, we consider Model B instead of Model C.1 as the benchmark fit for the "2020" spectrum since further fits do not significantly improve the fit.
Most importantly, the F-test between the Model B and Model C.1 for the "Total" spectrum gives a 3.05σ level of confidence (Table 3), indicating that when the data from the observations in 2020 is added to the "2000,2005" group, there is just enough statistical significance to reach the standard minimum 3σ level of confidence to support a conclusion.Based off of the 3.05σ level of confidence, we can conclude that the Fe Kα line in MCG-5-23-16 as a whole is asymmetric.
The difference between the Model B to Model C.1 Ftests and the best-fit inner disk radius R in values in the "2000,2005" group and the "2020" group demonstrate that the Fe Kα region in this source is variable, and that the emitting region has moved farther from the center of the AGN between 2000+2005 and 2020.
Afterwards, we also try letting the line emissitivity q vary within the range 2.0 ≤ q ≤ 4.0 (Model D.2). Comparing the reduced χ 2 results listed in Table 2, only marginally better fits are achieved when the emissivity varies for all spectra.In fact, evolving from Model B to Model C.2 is slightly worse than evolving from Model B to Model C.1, as seen by how the F-test values are slightly lower in the evolution from Model B to Model C.2 as compared to Model B to Model C.1 (Table 3).Relative to Model C.1, where q is fixed, smaller radii are found, but the small improvement in χ 2 /υ signals that smaller radii are not required.Therefore, we take Model C.1 as our benchmark fit, and all conclusions made about the inclination and inner disk radius R in will reference the values in Model C.1 as seen in Table 2.

Model D
Model D is zphabs*(zpowerlw+xillver). Model D replaces the zgauss component of the previous three models with the reflection model xillver (Garcia & Kallman 2010;Garcia et al. 2013).From Figure 2, it is clear there is asymmetry in the line in the "2000,2005" spectrum.xillver models the asymmetry in the line by taking into account Compton scattering of line photons due to reflection off the disk.The xillver model parameters include the photon index of the illuminating power law spectrum, the iron abundance, the ionization of the disk, the high-energy cutoff of the power law, the inclination at which the emitting region is viewed, and the redshift of the source.The specific table used is "xillver-a-Ec5.fits",the ionization logξ is frozen to 0, and the high-energy cutoff for the power law E cut is frozen to 300 keV.
When the line was fitted to Model D, the fits worsened substantially compared to the Gaussian models (see Figure 1 and Figure 2).The reduced χ 2 values increased across all three spectra compared to Model B and is even higher than the reduced χ 2 value for Model A in most of the spectra (Table 2 and Table 2).Because Model D fits the data so poorly, using a F-test to compare it to Models A and B is meaningless, so we did not Model D in our comparisons of models in Table 3.This also means that the parameter values in Table 2 for Model D should not be quoted as physically real.
The disk reflection model pexmon gave a similarly bad fit (Nandra et al. 2007).Therefore, we can conclude that the line asymmetry in MCG-5-23-16 line cannot be from a disk with only intrinsic reflection broadening.
Model E uses the toroidal reflection model mytorus (Murphy & Yaqoob 2009) instead of the disk reflection models xillver and pexmon.When fitting to the narrow Fe Kα line, we found that the fit is insensitive to the inclination angle.Therefore, for each spectrum, we fixed the inclination of Model E at 60 • .For consistent hydrogen column density and photon index values, we also tied the nH and PhoIndx parameters of the mytorus component to those of the zphabs and zpowerlw components.
We found that the Model E does not capture the red wing of the Fe Kα line in all three spectra and also doesn't capture the full blue wing in the "2020" and "Total" spectra (Fig. 2).The reduced χ 2 values for Model E are also higher than those of both Model C.1 and Model C.2 (Table 2), indicating that Model E is a worse fit than Model C.This demonstrates that the relativistic Doppler broadening introduced by rdblur is responsible for the asymmetry we observe in the narrow Fe Kα line, and not the Compton-scattering off of a toroidal structure as modeled by mytorus.
We did not include the F-tests involving Model E in Table 3 because the number of free parameters in Model E is less than or equal to the number of free parameters in all other models, and the F-test is used only when a new model has more free parameters than previous models (e.g. when the new model constitutes an increase in complexity).

Model F
Model F is zphabs*(zpowerlw+rdblur*mytorus).Since introducing relativistic Doppler broadening improved the fits for the Gaussian models (from Model B to Model C), we convolved the mytorus component by rdblur in Model F to see if the Doppler broadening will similarly improve the mytorus fits.Unlike the Model E fit, the Model F fit was not insensitive to the inclination so we let the inclination freely vary in Model F. We also tried both fixing the line emissivity q to 3.0 and letting q freely vary, and found that when q was able to freely vary, the fit was better able to capture the red wing.Similarly to Model F, we tied the nH and PhoIndx parameters of the mytorus component to those of the zphabs and zpowerlw components.
We found that adding the rdblur component to the mytorus model in Model F significantly improved the fit to the asymmetry in the Fe Kα line compared to Model E (see Fig. 2).This is supported by the F-tests between Model E and Model F, which have a level of confidence of 3.75σ for the "2000,2005" spectrum and 4.19σ for the "Total" spectrum (Table 3).The Doppler broadening component introduced by rdblur was able to fully capture the red wing of the line, while the mytorus model alone was not.This strongly indicates that the asymmetry in the Fe Kα line is due to Doppler broadening and not to Compton scattering, as both Model D and Model E were unable to fit the red wing in the line, but when introducing the rdblur component in Model F, the model was able to fully capture the asymmetry.We have analyzed deep Chandra/HETG spectra of the Type 1.9 Seyfert AGN, MCG-5-23-16.We find that the narrow Fe Kα emission line is asymmetric, likely due to relativistic Doppler broadening.Models A and B are simple Gaussians, are physically unmotivated, and do not fully account for the red wing in all three spectra.Model D gives bad fits to the spectra, indicating that the asymmetry in the line is physical and not due to Compton scattering of the emission line photons off the disk.Model E improves the fits but still does not fully capture the red wing.Model F includes relativistic Doppler broadening and fits the spectra much better.Model C.1 seems to best fit the spectra, as it captures the red wing while preserving the simple disk geometry assumed by fixing the line emissivity to q = 3. Model C.2 only improves the fits marginally but complicates the disk model by letting q vary freely, so we consider Model C.1 as our benchmark fit rather than Model C.2.The best fit Model C.1 suggests R ≃ 200-650 r g , where r g = GM/c 2 , which suggests that the line originates from the innermost extents of the optical BLR or closer.
The Gaussian and reflection models with relativistic Doppler broadening fully fit the asymmetry in the Fe Kα line.Without the Doppler broadening component, the Gaussian models and even the reflection models were unable to fully capture the asymmetry in the line (e.g. they underfit the red wing).This result indicates that the asymmetry in the Fe Kα line in MCG-5-23-16 is due to solely relativistic Doppler broadening near the black hole and not due to Compton scattering.
Our results also indicate the asymmetry in the Fe Kα line in MCG-5-23-16 is likely to be variable over time, as the "2000,2005" spectrum is best fit by an asymmetric line but the "2020" spectrum is sufficiently fit by a broad Gaussian.The background-subtracted net count rate for the "2000,2005" spectrum is 0.4079±0.0028counts s −1 and for the "2020" spectrum is 0.2903±0.0015counts s −1 ; the "2000,2005" spectrum has higher flux than the "2020" spectrum, so this variability in the line could also depend on source flux level as well.This line variability would be interesting to follow-up in a future observing campaign of MCG-5-23-16, e.g. with the recent launch of the XRISM satellite, an X-ray telescope with even higher energy resolution than Chandra and thus more detailed study of the line.
This narrow Fe Kα line asymmetry has only been confirmed previously in NGC 4151, the brightest Seyfert galaxy in X-rays.Using Chandra grating spectra, Miller et al. (2018) found that the line asymmetry depends on source flux level, and that the emitting region could be located as close as 50-130 r g during the high flux intervals.Additional study of the Fe Kα emitting   .Fits to the Fe Kα line using three Gaussian models: a narrow Gaussian (Model A, column 1), a broad Gaussian (Model B, column 2), and a broad Gaussian with relativistic Doppler broadening with line emissivity q fixed to 3 (Model C.1, column 3).Model C.2, a broad Gaussian with relativistic broadening with line emissivity q free to vary, is not shown.Row 1 is the "2000,2005" spectrum, row 2 is the "2020" spectrum, and row 3 is the "Total" spectrum.Below each plot is a plot of delchi versus energy; delchi uses XSPEC's definition of (data -model) / error.In these plots, "data" is the continuum-subtracted spectrum.A dotted line at 6.346 keV is drawn to indicate the expected energy of the peak of the line after accounting for the source redshift z = 0.00849.The star in each row represents the benchmark model for each spectrum.The fit parameters and their errors are listed in Table 2.

0.7429
Note-Best fit properties of fitting the spectra from 2000+2005 (top), from 2020 (middle), and from all observations (bottom) using PyXspec.
Model A represents a Gaussian with sigma fixed to 0, zphabs*(zpow + zgauss); Model B is the same as Model A, but with sigma able to vary freely, zphabs*(zpow + zgauss); Model C.1 is the same as Model B but with the Gaussian multiplied by "rdblur", zphabs*(zpow + rdblur*zgauss) and line emissitivity q fixed to 3. Model C.2 is the same as Model C.1, but with line emissitivity free to vary between 2 ≤ q ≤ 4, zphabs*(zpow + rdblur*zgauss).Model D represents the disk reflection model xillver, which accounts for Compton scattering of the Fe Kα line off a disk, zphabs*(zpow + xillver); Model E represents the reflection model mytorus, which accounts for Compton scattering off of a torus, zphabs*(zpow + mytorus).Model F is the same as Model E but with "mytorus" multiplied by "rdblur" and line emissivity and inclination free to vary.In all cases, the energy of the Fe Kα line is fixed at 6.40 keV, and redshifts of all components are set to 0.00849, the redshift of MCG-5-23-16 according to https://ned.ipac.caltech.edu.The inner radius Rin is given in units of gravitational radii rg = GM/c 2 , the power law normalization NP L in units of 10 −2 photons/keV/cm 2 /s, and line flux F l in units of 10 −13 ergs cm −2 s −1 .The minimum and maximum energy ranges of "cflux" are 6.2 keV and 6.45 keV, respectively.1σ errors are given in superscripts and subscripts.Greater than and less than signs indicate the 1σ upper bound and lower bound of the parameter, respectively.Parameters marked with an asterisk were fixed at the value given.region by Zoghbi et al. (2019) using reverberation mapping demonstrated that the inner disk radius is smaller than the optical BLR radius, supporting the conclusion that in NGC 4151, the line originates in the inner regions.
Our results provide the second-ever evidence of Fe Kα line asymmetry.Asymmetry in the Fe Kα line in both NGC 4151 and in MCG-5-23-16 are dependent on flux and the inner radius are both constrained to be on the order of 100s to 1000 r g .The emitting region in MCG-5-23-16 is thus likely to be smaller than the optical BLR as well, allowing it to be able to be used as a probe for the inner AGN geometry.
The identification of asymmetry in the narrow Fe Kα line only in MCG-5-23-16 and NGC 4151 thus far raises intriguing questions regarding the prevalence of the line asymmetry across the broader population of X-ray AGN.It is possible that we have only detected this line asymmetry in MCG-5-23-16 and NGC 4151 thus far because of a selection bias: these two AGN were intentionally selected because of they are among the brightest AGN in X-rays and thus made natural choices for analyzing the narrow Fe Kα line.As long as these two sources are not drastically different from the rest of the X-ray AGN population in metrics such as inclination or Eddington fraction, it is not out of the question that less X-ray bright AGN would also exhibited a skewed Fe Kα line.The absence of such observations may also be attributed to low flux.Notably, in the case of NGC 4151, the narrow Fe Kα line displayed a more pronounced red wing during high flux states compared to low flux states, emphasizing the potential correlation between the line profile and X-ray flux levels (Miller et al. 2018).With higher spectral resolution from XRISM, it is possible that asymmetric narrow Fe Kα lines could be unveiled even in dimmer AGN.
Another explanation could be that both MCG-5-23-16 and NGC 4151 are observed to have X-ray absorption, albeit not reaching Compton-thick extremes.This intermediate absorption level may suggest an intermediate inclination for these two AGN where the asymmetry in the narrow Fe Kα line is most discernible.High inclinations, where the BLR obstructs the Fe Kα emitting region, or low inclinations, where the line-of-sight does not pass through enough of the redshifted material might hinder clear detection of the Fe Kα asymmetry.It is possible that other X-ray AGN possess inclinations closer to either extreme, and thus we have not been able to clearly detect the line asymmetry in other sources.

Disk Geometry
In the best fits to the spectra, a low inclination is strongly preferred in all cases: the best fit Model C.1 to the "2000,2005" spectrum gives i = 6.4 +1.5 −1.2 degrees, i = 9.2 +2.3 −1.8 degrees for the "2020" spectrum, and i = 7.8 +2.0 −1.4 degrees for the "Total" spectrum (Table 2).These inclinations are consistent with those of the emitting region in NGC 4151, the only other source confirmed to have an asymmetric narrow Fe Kα line: the best fit inclinations for NGC 4151 are 10 • (Miller et al. 2018).
We would like to highlight that these inclinations for the narrow component of the Fe Kα line are much less than the inclinations of the broad component of the line in MCG-5-23-16: for the broad component, Reeves et al. (2007) and Zoghbi et al. (2017) cite an inclination of i ∼ 50 • .The low inclination is also not consistent with the classification of the source as a Seyfert 1.9-AGN, since in the classical Seyfert picture, Seyfert 1.9-AGN are high inclination sources.One possible explanation for the discrepancy is since the narrow component originates from the intermediate disk or outer disk and the broad component from the inner disk, the difference in inclination hints that this source contains a possible warp or a disk whose scale height increases from the inner regions to the outer regions (e.g. a flared disk), with the viewing angle towards the source at roughly 50 • .
We also note that the Fe Kα line flux is consistent between the "2000,2005" spectrum and the "2020" spectrum across all models A-F, which indicates a stable AGN environment around the emitting region between 2000 and 2020, such as a a stable X-ray source and/or consistent geometry surrounding the emitting region (Table 2).

CONCLUSIONS
Our results provide the second-ever evidence of asymmetry in the narrow Fe Kα line.The only previous confirmed detection of asymmetry in the narrow component of this line is in the source NGC 4151 (Miller et al. 2018), the brightest Seyfert galaxy in X-rays.As one of the next brightest Seyfert galaxies in X-rays, MCG-5-23-16 serves as a prime candidate for verifying that the asymmetry in the line in NGC 4151 may be able to be generalized to other X-ray bright AGNs.In the best fit models to the Fe Kα line, the inner radius of the line emitting region in both NGC 4151 and in MCG-5-23-16 is constrained to be on the order of R ≃ 100s to 1000 r g , making the narrow Fe Kα line originate in the innermost extents of the optical BLR, the X-ray BLR.This paper in tandem with Miller et al. (2018) illustrates a novel method for studying the circumnuclear environment around AGN in an unprecedented manner and that the narrow Fe Kα line can be asymmetric in AGN.
Interestingly, we find that the asymmetry in the line is due to relativistic Doppler broadening and not Compton scattering, as both the Gaussian models and reflection models only fully fit the Fe Kα line for all three spectra when Doppler broadening was included in the models.We also see evidence that the line emitting region in MCG-5-23-16 is variable over time.We detect asymmetry on the 2000 and 2005 observations but not in the 2020 data, where the flux dropped by a factor of ∼1.4.This indicates that the line and emitting region are changing over time.Follow-up observations of MCG-5-23-16 are necessary to examine how the line may change on different timescales.
It also does not escape our attention that there may be an additional emission line at ∼6.48 keV in the "2000,2005" and "Total" spectra and a weak absorption feature at ∼6.68 keV in the "2020" and "Total" spectra (Fig. 1).
Our main results are as follows: 1.In the AGN MCG-5-23-16, we have found the second-ever evidence of asymmetry in the narrow Fe Kα line.The only previously confirmed detection of asymmetry in the narrow component of the line is in the source NGC 4151.2. Interestingly, this asymmetry is due to solely relativistic Doppler broadening near the black hole and not Compton scattering.3.With a radius of R ≃ 200-650 r g , the narrow Fe Kα line emitting region in MCG-5-23-16 is located in the innermost extents of the optical BLR, or closer.4. The asymmetry in the line is variable over time, indicating that the distance of line emitting region from the black hole may be changing.
5. The narrow component of the line strongly prefers low inclinations of 10 • , while the broad component strong prefers inclinations of ∼ 50 • .This suggests that the AGN has a warp or a flared disk.

Future Steps
We now have two confirmed sources of asymmetry in the narrow Fe Kα line in two of the brightest AGN in X-rays, NGC 4151 and MCG-5-23-16.Future steps to study the asymmetry in the line include, but are not limited to: 1. Studying the line shape in other X-ray bright sources.2. Studying the variability of the line using NICER.

1 .
Second-Ever Evidence for Asymmetry in the Narrow Fe Kα Line

Figure 1 .
Figure 1.Spectral Fits with Gaussian Models Figure1.Fits to the Fe Kα line using three Gaussian models: a narrow Gaussian (Model A, column 1), a broad Gaussian (Model B, column 2), and a broad Gaussian with relativistic Doppler broadening with line emissivity q fixed to 3 (Model C.1, column 3).Model C.2, a broad Gaussian with relativistic broadening with line emissivity q free to vary, is not shown.Row 1 is the "2000,2005" spectrum, row 2 is the "2020" spectrum, and row 3 is the "Total" spectrum.Below each plot is a plot of delchi versus energy; delchi uses XSPEC's definition of (data -model) / error.In these plots, "data" is the continuum-subtracted spectrum.A dotted line at 6.346 keV is drawn to indicate the expected energy of the peak of the line after accounting for the source redshift z = 0.00849.The star in each row represents the benchmark model for each spectrum.The fit parameters and their errors are listed in Table2.

Figure 2 .
Figure 2. Spectral Fits with Reflection Models

Figure 2 .
Figure 2. Fits to the Fe Kα line using three reflection models: xillver (Model D, column 1), mytorus (Model E, column 2), and mytorus with relativistic Doppler broadening (Model F, column 3).See the caption of Figure 1 for more figure details.Model D and Model E do not fully capture the red wing.Only after introducing Doppler broadening with Model F, we have fits that adequately capture the red wing.The fits with Model F compared to Models D and E demonstrate that the asymmetry in the line must be due to Doppler broadening and not Compton scattering.The star in each row represents the benchmark model for each spectrum.The fit parameters and their errors are listed in Table 2.
3. Studying the detailed shape of the line in MCG-5-23-16 and other sources using XRISM.This work has been supported by the CRESST II program at the NASA Goddard Space Flight Center.The material is based upon work supported by NASA under award numbers 80GSFC21M0002 and GO0-21086A.The research in this article has made use of data obtained from the Chandra Data Archive, and the software CIAO provided by the Chandra X-ray Center (CXC) (Fruscione et al. 2006) and Python package PyXspec (Gordon & Arnaud 2021) and HEAsoft environment (Nasa HEAsarc 2014) developed by the HEASARC Software Development team at NASA/GSFC.

Table 3 .
F-tests between ModelsNote-F-tests were performed in order to determine the statistical significance of evolving from one model to the next.See caption of Table1for a detailed description of each model and spectrum.F-tests with Model D were not performed because Model D gave significantly worse fits to the data than all the other models, and thus the F-test would've given an negative value for the F-statistic.Evolving from Model B to Model C.1 in the "Total" group gives a 3.05σ confidence level, while the same model change for the "2000,2005" group gives a 2.74σ level.This indicates that the addition of the data from the observations in 2020 to the "2000,2005" group just gave enough statistical significance to reach the minimum 3σ level needed to conclude that the Fe Kα line in MCG-5-23-16 is asymmetric.