eROSITA detection of a cloud obscuration event in the Seyfert AGN EC 04570 − 5206

Context. Recent years have seen broad observational support for the presence of a clumpy component within the circumnuclear gas around supermassive black holes (SMBHs). In the X-ray band, individual clouds can manifest themselves when they transit the line of sight to the X-ray corona, temporarily obscuring the X-ray continuum and thereby indicating the characteristics and location of these clouds. Aims. X-ray flux monitoring with Spectrum Roentgen Gamma extended ROentgen Survey with an Imaging Telescope Array ( SRG / eROSITA) has revealed that in the Seyfert 1 active galactic nucleus (AGN) EC 04570 − 5206, the soft X-ray flux dipped abruptly for about 10–18 months over 2020–2021, only to recover and then drop a second time by early 2022. Here, we investigate whether these flux dips and recoveries could be associated with cloud occultation events. Methods. We complemented the eROSITA scans with multiwavelength follow-up observations, including X-ray / UV observations with Swift , XMM-Newton , and NICER, along with ground-based optical photometric and spectroscopic observations to investigate the spectral and flux variability. Results. XMM-Newton spectra confirm that the soft X-ray flux dips were caused by partial-covering obscuration by two separate clouds. The 2020-2021 event was caused by a cloud with column density near 1 × 10 22 cm − 2 and a covering fraction of roughly 60 percent. The cloud in the 2022 event had a column density near 3 × 10 23 cm − 2 and a covering fraction near 80 percent. The optical / UV continuum flux varied minimally and the optical emission line spectra showed no variability in Balmer profiles or intensity.


Introduction
Active galactic nuclei (AGNs) are generally thought to be powered by the accretion of gas onto a supermassive (10 6−9 M ⊙ ) black hole.However, open questions remain regarding the precise morphology and mechanics of some of the various accreting and outflowing components.
Circumnuclear X-ray-obscuring gas, colloquially called the "torus," likely supplies gas for the accretion disk that feeds the supermassive black hole, while obscuring certain lines of sight to the central engine.Though its exact morphology and nature remain unclear, the torus likely plays a role in radiatively driven outflows and AGN feedback (e.g., Murray, Quataert, & Thompson 2005;Hönig 2019).Its morphology governs the fractions of unobscured, Compton thin-obscured, and Compton thickobscured AGNs, thereby impacting their relative fractions in comprising the cosmic X-ray background (e.g., Comastri et al. 1995;Gilli et al. 2007).The morphology of the X-ray-obscuring torus is generally accepted to be preferentially distributed toward the equatorial plane, but morphological parameters such as the global covering factor may depend on luminosity (Burlon et al. 2011;Ricci et al. 2017;Ananna et al. 2022).
In nearby Seyferts, mid-IR and sub-mm interferometry, with the Atacama Large Millimeter Array (ALMA) or Very Large Telescope (VLT) MIDI (Kishimoto et al. 2009;Tristram et al. 2009;Pott et al. 2010;García-Burillo et al. 2016;Tristram et al. 2022), for instance, and reverberation mapping of the thermal emission from warm dust (Suganuma et al. 2006) have been revealing the presence of dusty components to the torus, residing at radial scales of parsecs down to tenths of a parsec.This dusty component has been invoked in orientation-dependent "unification" schemes for Seyferts, in which "type 1" or "type 2" objects either display or lack (respectively) highly Doppler-broadened (FWHM on the order of thousands km s −1 ) emission lines from the compact optical broad line region (BLR).As noted by Netzer & Laor (1993), Elitzur (2007), and Gaskell et al. (2008), BLR structures and dusty-torus structures may comprise one continuous, radially extended structure straddling both inside and outside the dust sublimation zone, as dust embedded in dense clouds gas can suppresses optical/UV line emission.
There is mounting observational evidence that dusty and non-dusty circumnuclear gas in nearby AGNs contain components comprised of clumps or filaments; consequently, the AGN community tends to use the term "torus" to denote circumnuclear gas, although it might not necessarily form a simple axisymmetric "donut" shape.In X-rays, the community has been accumulating observations of variations in the line-of-sight column density N H on timescales of days to years in more than roughly 20 Seyferts (e.g., Mushotzky et al. 1978;Reichert et al. 1986;Risaliti et al. 2002Risaliti et al. , 2009Risaliti et al. , 2011;;Markowitz et al. 2014;Ricci et al. 2016;Zaino et al. 2020); a review is given in Ricci & Trakhtenbrot (2023).Column densities are typically observed to vary by factors of a few to ten or more; many variations are associated with structures that are neutral or lowly-ionized, and are Compton-thin (on the order of 10 22−23 cm −2 ) or Compton-thick (on the order of 10 24 cm −2 ).Column density variations have been observed both in Seyferts that usually lack significant amounts of neutral or lowly-ionized gas along their line of sight, as well as in perpetually obscured Seyferts (e.g., Risaliti et al. 2002;Bianchi et al. 2009;Rivers et al. 2011;Laha et al. 2020), with some objects in this latter group transitioning between Compton-thin and Compton-thick obscured states (e.g., Bianchi et al. 2009;Ricci et al. 2016;Marchesi et al. 2022).Variations in N H have been attributed to both full-covering and partial-covering obscurers, with covering fractions of partial-covering obscurers varying in some cases (e.g., Puccetti et al. 2007;Turner et al. 2008;Sanfrutos et al. 2013).
Such observations support the notion that discrete, dynamic structures temporarily transit the line of sight to the X-ray corona.Typically, eclipse events that occur on timescales of roughly a day and shorter are inferred to be associated with structures commensurate with the optical BLR (e.g., Risaliti et al. 2011;Sanfrutos et al. 2013;Gallo et al. 2021), based on constraints from ionization and event duration, inferred number densities, and/or from obscuring structures' being partial-covering as opposed to full-covering.Meanwhile, eclipses or transitions that take months to years to complete are typically inferred to be commensurate with the outer BLR or inner dusty torus (Markowitz et al. 2014;Beuchert et al. 2015).
In parallel, high-spatial resolution IR/sub-mm imaging (Raban et al. 2009;Izumi et al. 2018;García-Burillo et al. 2019) and IR SED modeling of thermal dust emission (Ramos Almeida et al. 2011, 2014) dust also resolve, or support, clumpy or filamentary structures.All these observations have motivated the development of models wherein the obscuring gas is composed of discrete clouds (Hoenig 2013).In such models, the total observed line-of-sight obscuration is a probability that depends on the number of clouds lying along the line of sight, which, in turn, depends on both viewing angle and the angular and radial extents of the cloud distribution; however, in most models (e.g., Nenkova et al. 2008 andLiu &Li 2014 for IR and X-ray spectra, respectively), clouds are preferentially distributed toward the equatorial plane.
However, outflowing clumpy winds, likely launched from the inner accretion disk, can also produce temporary X-ray obscuration events, as observed in a handful of nearby Seyferts.Such obscurers tend to be partial covering in the X-rays, moderately ionized, and have total columns on the order of 10 22−23 cm −2 .Such winds are often (but not always) accompanied by UV absorption lines that are blueshifted by an order of a few thousand km s −1 .Events can last up to a month or longer (e.g., NGC3783, Mehdipour et al. 2017;Mkn 817, Kara et al. 2021), and even up to a decade (NGC5548 Kaastra et al. 2014;Mehdipour et al. 2022).Correlations between the properties of the X-ray-and UV-obscuring components (Mehdipour et al. 2022) hint at X-ray obscuration being due to dense clumps embedded in UV-absorbing, outflowing gas.Wind locations are generally inferred to be on the order light-days to light-weeks from the black hole, and such winds may act as a "filter," obscuring and modifying the ionization continuum that is intercepted by the BLR (Dehghanian et al. 2019).
Accumulating statistics on the properties of individual X-ray-obscuring structures -locations, dust content, ionization structure, connections to both relativistically and subrelativistically outflowing ionized winds -is needed to constrain their morphology and possible origins.However, the detection of new obscuration events poses a challenge.With some exceptions such as NGC 1365, in which discrete eclipses or rapid transitions between Compton-thin and -thick states are frequently observed (Ricci & Trakhtenbrot 2023), obscuration events in Seyferts tend to occur very rarely on a per-object basis.Markowitz et al. (2014) and Torricelli-Ciamponi et al. (2014) demonstrated the power of long-term X-ray monitoring of a large starting sample of Seyferts to aggregate a sample of obscuration events.However, those studies were archival and had assessed events that had occurred prior.An ideal situation is to monitor a sample of Seyferts and catch obscuration events as they are occurring and to immediately apply follow-up observations to obtain optimal constraints on the properties of obscurers.
The extended ROentgen Survey with an Imaging Telescope Array (eROSITA; Predehl et al. 2021) is the soft X-ray instrument aboard Spectrum Roentgen Gamma (SRG; Sunyaev et al. 2021).Starting in December 2019, eROSITA began conducting deep 0.2-10 keV all-sky surveys (eRASS), scanning the entire sky once every six months.Its repeated scans enable time domain astronomy studies on a variety of variable-emission sources, including AGNs.By monitoring a starting sample on the order of 10 6 AGNs and visiting each source every six months, eROSITA can amplify small numbers of rare transient AGN events.
Our team has been monitoring AGNs for major changes in soft X-ray flux between successive eRASS scans.We identified a Seyfert 1 at redshift z=0.276 whose soft X-ray flux drop from eRASS1 to eRASS2 ranked it among the most significant (11σ) drops among all targets monitored during eRASS1 and 2. Continued tracking of this object revealed that this object's 0.5-2 keV flux varied drastically (factors ≳10) on timescales of months and longer.Specifically, its 0.5-2 keV flux dropped by ∼11 from eRASS1 (February 2020) to eRASS2 (August 2020), then recovered, increasing by a factor of ∼15 from eRASS3 (February 2021) to eRASS4 (August 2021) and then dropped a second time, by a factor of ∼9, by eRASS5 (February 2022).
Immediately after we detected each of these major X-ray flux variations, we triggered target-of-opportunity observations encompassing X-ray spectroscopy and photometry, space-based UV/optical photometry, and ground-based optical photometry and spectroscopy.In total, the eRASS scans and these multiwavelength follow-up observations spanned almost three years.We conclude that the two low soft X-ray states -one lasting from before August 2020 until early 2021, with the second starting by February 2022 and lasting through at least mid-2023are associated with two discrete structures that temporarily transited the line of sight to (and partially covered) the X-ray corona.Meanwhile, the high soft-X-ray flux states were associated with lack of obscuration.Our results thus represent the first two major X-ray obscuration events to be detected with eROSITA.Moreover, this object is the only candidate identified by comparing eRASS1 to eRASS2 flux changes where follow-up observations supported a changing-obscuration event.
The remainder of this paper is organized as follows: In Sect.2, we discuss the source's counterpart, the observing campaign, and the data reduction, and we summarize the sources' multiband continuum variability characteristics.In Sect.3, we discuss the spectral fits to the X-ray data to characterize the obscurers.In Sect. 4 and Sect.5, we fit the optical/UV SED and the optical spectra, respectively, to check for signs of reddening by dust.Section 6 presents a discussion, where we infer the locations of the obscuring structures and discuss their nature.A summary of our results is given in Sect.7.

Identification and counterpart
The primary goals of eROSITA are to conduct all-sky X-ray scans and to map hot gas in ∼10 5 galaxy clusters and intercluster filaments out to redshifts ∼1.3, thus tracing evolution of large-scale structures across cosmic time, and to detect on order of a million AGNs.SRG orbits the Earth-Sun L2 point.During eROSITA's all-sky scan mode, SRG rotates once every four hours such that eROSITA, with its ∼1 • field of view, traces a narrow ring on the sky, following a great circle.SRG's orbit precesses by roughly one degree per day, tracking the Earth-Sun line, so that the entire sky is mapped every six months (each eRASS).
There were no archival optical spectra available and, thus, its redshift was not known previously.Our first follow-up optical spectra, discussed in further detail below, yielded a redshift of 0.276.In this paper, we assume a flat cosmology, H 0 = 70 km s −1 Mpc −1 , Ω M = 0.3, and Ω vac = 0.7, which yields a luminosity distance of 1400 Mpc (Wright 2006) 2 for this bestfit redshift value.Its infrared color as measured by WISE in the AllWISE survey (Cutri et al. 2013) is W1 − W2 = 1.0 mag, suggesting the presence of an AGN following Stern et al. (2012) and Assef et al. (2018).

X-ray observations and data reduction
Our follow-up campaign included X-ray data from eROSITA, XMM-Newton, Swift, and NICER.In Table 1, we present a log of the X-ray observations.

SRG/eROSITA
eROSITA completed four full scans and part of a fifth scan (eRASS5) before the instrument was placed into safe mode in February 2022; J0458-5202 was visited in each of eRASS1-5, and observations are referred to as eR1, ..., eR5, henceforth.Within each eRASS, objects at the orbital equator receive six epochs, each with a ∼40 s exposure, separated by four hours, in each eRASS scan.However, objects closer to the orbital poles receive a larger number of successive epochs.J0458-5202 lies very roughly 15 • from SRG's orbital poles within each eRASS scan.Within each of the five eRASS scans, eROSITA scanned the target 24-30 passes, once every four hours, over the course of 92-112 hours (the number of passes varied slightly due to small orbital changes between eRASS scans).
Data were extracted using event processing version c020.We used eSASS version 21121_0_4 (Brunner et al. 2022) and HEA-SOFT version 6.30.1.We combined data from all seven Telescope Modules.We extracted the source using a circular extraction region with the radius scaled to the 0.2-2.3keV maximum likelihood (ML) count rate from the eRASS source catalog; the extraction regions therefore differ somewhat from one eRASS to the next, with larger radii corresponding to higher count rates.ML count rates take into account the time when the source was in the field of view, and with corrections for vignetting effects applied.Similarly, background regions were extracted using annuli whose inner and outer radii depend on ML count rate.Extraction radii and ML count rates are listed in Table 2. Point sources detected in the background extraction regions were excised, again using circular regions whose radii depended on ML count rate.Good exposure times after screening, also listed in Table 2, were in the range 724-964 s.Images of J0458-5202 from eR1-5 are displayed in Fig. 1 to help visualize the strong soft X-ray variability.

XMM-Newton EPIC
The four XMM-Newton (Jansen et al. 2001) observations occurred on 26 December 2020, 28 January 2021, 8 September 2021, and 25 April 2022, henceforth referred to as XM1, XM2, XM3, and XM4, respectively.All observations used both the EPIC (Strüder et al. 2001) pn and MOS cameras.XM1 and XM2 Notes.All MJD dates refer to the midpoint of the observation.Exposure refers to good time after screening.For XMM-Newton, the three exposure values refer to pn, MOS1, and MOS2, respectively.Notes.Exposure refers to good time after screening.The ML count rate (Merloni et al. 2024) denotes the maximum likelihood count rate taking into account the time when the source was in the field of view and with corrections applied for vignetting effects.
each used the large-window mode for all three cameras, while XM3 and XM4 each used full-frame mode for all three cameras.The medium optical blocking filter was used for all EPIC cameras in all observations.
We reduced the data using XMM Science Analysis Software (XMMSAS) version 19.1.0and HEASOFT version 6.28, fol-lowing standard extraction procedures for point sources.Source spectra were extracted from circles 40 ′′ in radius; background spectra were extracted from source-free regions with the same size located a few arcminutes away and on the same CCD chip.We screened data against strong, time-localized background flares due to proton flux by visually inspecting >5 keV back-ground light curves.For the pn, we selected data from pattern 0 and pattern 1-4 separately (henceforth pn0 and pn14).We checked for pileup using the XMMSAS task epatplot but found no evidence for any pileup.Good exposure times after screening are listed in Table 1.

Swift X-ray Telescope (XRT)
The Neil Gehrels Swift Observatory (Swift; Gehrels et al. 2004) observed J0458-5202 eight times between October 2020 and February 2023 (Sw1, ..., Sw8).Each XRT (Burrows et al. 2005) observation was in photon counting (PC) mode.Raw event files were reprocessed using xrtpipeline version 0.13.5 in HEA-SOFT version 6.28.and the latest XRT calibration files.We extracted source spectra using circular regions of radius 20 pixels (47 ′′ ); background spectra were extracted from annular regions of inner radius 60 pixels (141 ′′ ) and outer radius 65 pixels (153 ′′ ), and confirmed to be free of background sources.We generated ancillary response files using xrtmkarf, and we selected the PC mode response files from the calibration database.Good exposure times after screening are listed in Table 1.

NICER
The Neutron Star Interior Composition Explorer Mission (NICER; Gendreau et al. 2016), aboard the International Space Station (ISS), observed J0458-5202 six times between 30 October 2020 and 4 November 2020, as listed in Table 1.We used NICERDAS version 10 software and followed standard procedures to screen data, produce cleaned event files, and extract spectra.We discarded data from detectors 14 and 34, which are prone to excessive noise.
We rejected time intervals when the detector undershoot rate3 exceeded 150 ct s −1 per module.We also screened out time intervals when the detector overshoot rate (caused when high-energy particles deposit excess charge) exceeded 1.5 ct s −1 .Given the source faintness, we discarded data taken during the ISS' passage through the "SAA" South Atlantic Anomaly boundary, which is defined to be more conservative and cover a slightly larger area than for the standard "NICERSAA" boundary.We used the 3C50 background estimation method, screening out times where the background rate in the hard band (13-15 keV) exceeded 0.5 ct s −1 in the hard band.
Good exposure times after screening are listed in Table 1.Given the danger of underestimated optical loading impacting the softest energies, and given the faintness of the source, we discarded data below 0.4 keV and above 10 keV.
However, in each observation, as well as in spectrum summed from all six observations, the source was not detected, as we obtained negative net count rates after background subtraction.For the summed spectrum, we estimate upper limits of ∼3 ×10 −13 and ∼2 ×10 −13 erg cm −2 s −1 for the 0.2-5 and 0.5-2 keV bands, respectively.

XMM-Newton Optical Monitor
The XMM-Newton Optical Monitor (OM; Mason et al. 2001) observed J0458-5202 with the UVM2 filter (effective wavelength: 231 nm) simultaneously to each of the four EPIC observations; XM4 additionally used the B filter (450 nm).Dates, start-stop times, and total good exposure times (sum of all images) are listed in Table 3.The numbers and lengths of individual exposures are listed in Table 4.We reduced the data using the XMM_SAS routines omichain and omfchain for the image and fast modes, respectively.These routines apply flat-fielding, source detection, and aperture photometry for each exposure, and they combine all exposure images into a mosaiced image, and perform source detection and aperture photometry on the mosaiced image.The source extraction radius was 12 pixels = 5 ′′ .7. These routines also correct fluxes for detector dead time.
We verified that the source was well detected within each exposure, that there were no obvious imaging artifacts in any exposure, and that the source was not too close to the edge of the window in fast mode.AB magnitudes were converted to Vega magnitudes following the zeropoints listed in Sect.3. Aperture photometry for each UVOT filter was performed using the ASI Space Science Data Center Multi-Mission Interactive Archive online data analysis tool5 .It sums all exposure fractions and performs source extraction using uvotdetect and the latest CALDB.Source counts were extracted from a circle of radius 5 ′′ ; background was extracted from an annulus of inner and outer radii 27 ′′ .5 and 35 ′′ , respectively.Aperture-corrected, background-subtracted, and Galactic extinction-corrected magnitudes and flux densities were derived; they are listed in Table A.1.The uncertainties on these magnitudes are statistical only.As detailed in Appendix B, we also extracted the magnitudes for two stars nearby in the field of view to estimate systematic uncertainties.However, statistical uncertainties dominate, so we neglect these systematic uncertainties in all plots below.
In Fig. 2, we display the V, B, U, W1, M2, and W2-band light curves obtained with Swift UVOT during observations Sw3-8.Through Sw5-7, as the source flux decreases, the relatively higher energy bands become fainter faster, although the drop is only on the order of 0.3 mag in the far-UV.By Sw8, the source flux recovers, becoming slightly brighter than during Sw3-4.As we demonstrate in Sect. 4 from broadband spectral fitting, such spectral variability is consistent with an intrinsic bluer-whenbrighter trend, and not consistent with being due to a significant degree of variable extinction.  a) The last exposure lasted 3.3 ks. (b) The first exposure lasted 4.3 ks.Fig. 2. Optical/UV light curves of J0458-5202 for observations Sw3-8, in which Swift UVOT observed with all six filters.For this overplot, the V, B, W1, M2, and W2-band light curves have been offset by the magnitude values indicated in the top left corner such that their mean magnitudes across all six observations match the mean of the U-band light curve.This action helps to illustrate the optical/UV spectral variability as a function of flux; as demonstrated in Sect.4, such spectral variability is consistent with bluer-when-brighter behavior, and not a product of variable extinction.

5.92×10
Finally, we used the ftools uvotimsum and uvot2pha to generate single-channel spectral files for each filter for the purpose of SED fitting.We used standard UVOT response files from the calibration database.The time-resolved SED fits to the UVOT spectra will be discussed in Sect. 4.

Ground-based optical photometry
We obtained 40 observations of J0458-5202 at the 0.4-m PROMPT6 telescope at Cerro Tololo Inter-American Observatory, operated as part of the Skynet Robotic Telescope Network.The first 33 observations (MJD 59164-59308) used the Johnson B filter; the final seven (MJD 59389-59541) used Johnson V. We reduced all images in a standard manner, including bias correction and flat-fielding.We performed aperture photometry by extracting source-centered circles and background annuli.Observed (not corrected for Galactic reddening) Vega magnitudes are listed in Table A.2.

Optical spectroscopic observations
We obtained ten longslit spectra of J0458-5202 between September 2020 and October 2022, as listed in Table 5.These observations included the South African Large Telescope (SALT) longslit Robert Stobie Spectrograph (RSS; Burgh et al. 2003;Kobulnicky et al. 2003), the FORS2 spectrograph (Appenzeller et al. 1998) on the 8.2 m Very Large Telescope Array's (VLT) UT1 at Cerro Paranal, and the SpUpNIC spectrograph (Crause et al. 2019) at the South African Astronomical Observatory (SAAO) 1.9 m telescope.Spectra #1-7 were taken between September 2020 and April 2021, during the first soft X-ray flux dip; spectrum #8 was coincident with the high soft X-ray flux state; spectra #9 and 10 were taken in April 2022 and October 2022, during the second soft X-ray flux dip.
For the SAAO/SpUpNIC spectrum, we used a 2.7 ′′ slitwidth, and a low-resolution grating to cover the entire optical range.
For the FORS2 observation on 27 December 2020 (#4), we made use of three gratings: G1400V (1000 s exposure), G300V + GG435+81 (450 s exposure), and G300I + OG590+32 (450 s exposure).The 300V and 300I exposures provide the full wave- Notes.For spectra #5, 6, and 7 at SALT RSS, exposures were taken with two RSS setups, hence, two exposure times are listed.For spectrum #4 at VLT FORS2, the three values listed refer to exposures using the G300V, G300I, and G1400V gratings, respectively.For spectrum #9 at VLT FORS2, the two values listed refer to exposures using the G300V and G300I gratings.
length coverage and the G1400V exposure garnered additional S/N near the Hβ-[O iii] region.We combined all exposures into a single high-quality spectrum.For the FORS2 observation of 27 April 2022 (#9), we used the G300V and G300I gratings, with exposures of 1000 s each, and again combined the results into a single spectrum covering 3520-9980 Å.
All CCD data were reduced using standard bias corrections and flat-fielding.Wavelength calibration used arc-lamp spectra taken on the night.Spectrophotometric calibrations were performed using data for a standard star taken the same night in the case of the VLT-FORS2 and SAAO 1.9m spectra, and a standard star taken within the past year in the case of the SALT-RSS spectra.

Multiband variability overview
As an overview of J0458-5202's X-ray variability, we list 0.2-5.0 and 0.5-2.0keV fluxes in Table 6.For XM1-4, eR1-5, and Sw3, fluxes were derived directly from the best-fitting spectral models, as detailed in Sect.3.All other fluxes are based on count rates, and conversions using the best-fitting models, also discussed in Sect.3.
Meanwhile, the middle and lower panels of Fig. 3 display, respectively, the M2 and B-band photometry light curves.As discussed in Appendix B, combining the XMM-Newton OM and Swift UVOT M2 observed magnitudes into one light curve required some magnitude corrections, given that the energy peaks of the two instruments' effective areas differ by ∼10 percent.We estimated spectral slopes from the SED and derived corrections of −0.070 (XM1-3) and −0.031 (XM4) to apply to the OM data points, as detailed in Appendix B, with the final, corrected light curves plotted in Fig. 3. Overall, the B and M2 light curves dis-play much lower levels of variability compared to the X-rays.The combined OM and UVOT M2 band light curve shows a decrease in flux of only 30 percent over ∼400-450 days through Sw7, followed by a recovery by Sw8.The B-band continuum is overall characterized by mild variations on the order of ∼30 percent over the two-year campaign, although there is additionally a sharp, temporary drop in flux by ∼40 percent over ∼40 days, starting after roughly MJD 59210.Overall, the results of Figs. 2 and 3 rule out any major (e.g., order of magnitude of more) drop in optical/UV luminosity or accretion rate occurring during the campaign.Notes.Observed 0.2-5.0 and 0.5-2.0keV fluxes, determined from model fits to all spectra except for Sw1, 2, 4, 5, 6, 7, and 8, which were based on measured count rates.N1-6 denotes the estimated upper limit associated with the non-detection in the summed NICER data.

X-ray spectral fits
Our spectral fitting strategy is to start by modeling the XMM-Newton spectra (Sect.3.1), as they had the highest S/N.We use those results to inform fitting of the other (lower S/N) Xray spectra; eRASS and Swift XRT spectra are described in Sects.3.2 and 3.3, respectively.
All X-ray spectral fits were done in Xspec version 12.13.0c.All parameter uncertainties are for one interesting parameter, and were derived using a Markov chain Monte Carlo (MCMC) algorithm via the chain routine in Xspec.We used the Goodman-Weare sampler (Goodman & Weare 2010), chains of length 25000, 20 walkers, and a burn length of 5000.Parameter errors are at the 90 percent confidence level, and are taken from the 5th and 95th percentile values of the parameter distribution.In all models, we included a TBabs component to account for Galactic absorption by H i and H 2 totaling 1.04 × 10 20 cm −2 (Willingale et al. 2013).We assumed the abundances of Anders & Grevesse (1989).

Spectral fits to XMM-Newton data
Our strategy was to start with XM3, which sampled a high soft X-ray flux state and had the highest total counts (19200; summing pn0+pn14+MOS1+MOS2), and then fit XM1, XM2, and XM4, which sampled low soft X-ray flux states and had fewer total counts (16600, 6700, and 4000, respectively).For each observation, we fit pn0 (0.25-10 keV) + pn14 (0.5-10 keV) + MOS1 + MOS2 (both 0.2-10 keV) jointly.We applied instrumental constant components for cross-calibration purposes, keeping the constant for pn0 fixed at unity; constants for the other spectra were usually within a few per cent of unity for best-fitting models.All spectra were grouped to 20 counts per bin to ensure use of χ 2 statistics.

XM3
We used a baseline spectral model consisting of the following components, and with an eye toward developing a physically self-consistent model to apply to all four XMM-Newton spectra: -A hard X-ray power law to model emission from the hot (T e ∼ 10 9 K), optically thin corona.
Warm comptonization of optical/UV thermal photons emitted by the accretion disk by plasma with electron temperature T e ∼ 0.1-1.0keV and optical depth τ ∼ 10-40 has been successful in modeling the soft X-ray excesses of many nearby Seyferts, (e.g., Mehdipour et al. 2011;Di Gesu et al. 2014;Porquet et al. 2018;Petrucci et al. 2018).We assumed a sphere geometry, and fixed the seed photon temperature T seed at 20 eV, though the fit results were insensitive to the value of T seed .In all four XMM-Newton observations, the value of T e pegged at a lower limit of 0.1 keV, so we froze T e to this value.6-A Compton reflection component (Compton hump and Fe K emission), assuming a clumpy medium lying out of the line of sight.We used UxClumpy (Buchner et al. 2019), which follows Nenkova et al. (2008) in assuming a distribution of clouds that is preferentially concentrated toward the equatorial plane, with a Gaussian angular height distribution and a uniform radial distribution.There is also a Compton-thick inner ring with variable covering factor.However, data above 10 keV (observed frame) does not exist for J0458-5202, and the insensitivity of each XMM-Newton spectrum to the exact shape of the Compton hump means we could not attain reliable constraints on any UxClumpy geometry or viewing parameters.We froze the system inclination to 30 • , the cloud angular Gaussian distribution σ TOR to 20 • and the Comptonthick inner ring's covering fraction to 0. We froze the input power-law normalization to 3.5 × 10 −4 ph cm −2 s −1 keV −1 , the best-fitting value obtained from a joint fit to all XM1-4 spectra -likely constrained via the marginal detection of the Fe K line in XM4, as discussed below in Sect.3.1.5-and representing the average flux response from a cloud distribution extending up to a few light years or more away.The input photon index was frozen to 2.
We obtained a good fit, with χ 2 /do f = 704.26/650= 1.083.The warm Comptonization optical depth was τ warm = 45 +9 −6 .The hard X-ray power-law photon index was Γ HX = 1.90 ± 0.04.The data and model residuals are plotted in Fig. 4. Other best-fitting parameters as listed in Table 7.We refer to this model as M1 or "SXCOM+HXPL+UXCL." As an alternate model, we replaced the phenomenological hard power law with a more physical, second CompTT component.We again used a spherical geometry and fixed T seed to 20 eV.We kept T e fixed at 100 keV (e.g., Tortosa et al. 2018), due to the lack of constraints on any high-energy power-law cutoff in J0458-5202.The fit remained virtually unchanged, with χ 2 /do f = 703.65/650= 1.083, and with parameters for the warm Comptonization component unchanged.The hot Comptonization depth was τ hot = 0.27 ± 0.03.Other best-fitting parameters as listed in Table 7.We refer to this model as M1-alt or "SXCOM+HXCOM+UXCL."In XM3 as well as to fits to all other spectra, replacing the hard X-ray power law with a hot Comptonized component made negligible impact on data/model residuals or fit statistics.Next, we tested if adding a full-covering neutral obscuration component, modeled with zTBabs (Wilms et al. 2000), improved the fit; it did not, with an upper limit to column density N H of 8.6 × 10 20 cm −2 .We thus do not include this component further.
Finally, we tested if the soft excess could be modeled via reflection off an ionized, relativistic accretion disk illuminated by a hard X-ray power law.We used relxill v.1.4.3 (García et al. 2014;Dauser et al. 2014), keeping the outer radius fixed at 400 R g , the Fe abundance fixed to the solar value, and powerlaw cutoff energy fixed at 100 keV (though the fit was insensitive to thawing these parameters).We initially set the inner radius to the inner stable circular orbit, but left this parameter free.We kept the black hole spin parameter, disk inclination, disk ionization parameter, and power-law emissivity index all free.We again include Compton-scattered components as in the previous fits to model emission from a distant, neutral torus.We refer to this model as "RelXill+HXPL+UXCL."However, the best fit had χ 2 /do f = 712.33/646= 1.103, with data/model residuals near 0.6-0.8keV a bit worse compared to the warm Comptonization model, as plotted in Fig. 4. In addition, the Akaike information criterion (AIC; Akaike 1973) with finite sample correction by Sugiura (1978) yields that ∆AIC going from the "SXCOM+HXCOM+UXCL" model to the "RelX-ill+HXPL+UXCL" model is positive (+16.74,driven by both the difference in fit statistic and the four additional free parameters), indicating a preference for the warm Comptonization model.However, we caution that the choice of soft excess model does not significantly impact our conclusions in this paper concerning the obscuring components.

XM1
We now turn our attention to fitting XM1, which was obtained in December 2020, during the first low soft X-ray flux state.We first tried the unobscured phenomenological M1 ("SX-COM+HXPL+UXCL") model.We obtain a moderately reasonable fit, with χ 2 /do f = 671.29/646= 1.039, and with datamodel residuals plotted in Fig. 5.However, the hard X-ray power-law photon index Γ HX is quite flat, at 1.32 ± 0.04.
We then tested if the flat hard X-ray spectral slope could be attributed to partial-covering obscuration masking a steeper power-law slope.We applied a partial-covering component using zpcfabs. 7We applied the partial-covering component only to the SXCOM and HXPL components, and not the UxClumpy component.Our best-fitting "M2" model, or "PC*(SXCOM+HXPL)+UXCL" model, had χ 2 /do f = 625.28/644= 0.973, with improved data/model residuals as plotted in Fig. 5, and a steeper value for Γ HX , 1.68 +0.11  −0.10 .The partial-covering obscurer had column density N H,PC = 9.9 +6.2 −3.3 × 10 21 cm −2 and covering fraction CF = 62 +11 −10 percent.Other bestfitting model parameters are listed in Table 8.A set of 500 Monte Carlo simulations using the simftest command in Xspec indicates that the addition of the partial-covering obscurer improves the fit at the >99.8 percent confidence level.In addition, ∆AIC when comparing the model lacking the obscurer to the model containing the obscurer is −41.9, confirming the necessity of the obscurer in the model.

XM2
Observation XM2 occurred 33 days after XM1 and we found similar soft and hard X-ray flux levels and a very similar X-ray spectral shape to XM1.Spectral fits to XM2 thus proceeded in a manner identical to those for XM1, although the lower exposure time for XM2 yielded total spectral counts a factor of 2.5 lower than for XM1.Again, we first fit an unobscured M1 ("SX-COM+HXPL+UXCL") model, whose best fit again required a very flat power-law photon index (Γ HX = 1.44 +0.08 −0.05 ); χ 2 /do f was 305.07/294 = 1.038.
When we added a partial-covering obscurer (M2; "PC*(SXCOM+HXPL)+UXCL"), the improvement in fit was negligible, with χ 2 /do f falling by only 2.94/2 and with Γ HX increasing to 1.64 +0.11  −0.07 .In addition, Monte Carlo simulations indicate that the addition of the obscurer improves the fit at merely the 80 percent confidence level.Other best-fitting model parameters are listed in Table 8, and data/model residuals are plotted in Fig. 6.The similarity in flux levels and spectral shape to XM1 suggest that XM2 was similarly obscured.However, we caution that detection of the partial-covering obscurer in XM2 is not statistically robust.Under the assumption that during both XM1 and XM2, the continuum is obscured by the same cloud of gas, XM1 provides better insight into the cloud properties.Finally, the parameters for our best-fitting M2-alt ("PC*(SXCOM+HXCOM)+UXCL") model, which yielded data-model residuals that are virtually identical to those for M2, are listed in Table 8.

XM4
Observation XM4 occurred in April 2022, during the second major soft X-ray flux drop.The counts spectra are plotted in the top panel of Fig. 7, and indicate a hard X-ray spectrum that is both very flat and has a strong convex shape.Again, we first tested an unobscured phenomenological M1 ("SXCOM+HXPL+UXCL") 7 We also tried a full-covering component, but obtained only an upper limit to N H , in this case N H < 5 × 10 20 cm −2 .Observed Energy (keV) Fig. 5. Same as Fig. 4, but for XM1.In panel a), crosses denote the counts spectra, and the histograms denote the best-fitting M2 ("PC*(SXCOM+HXPL)+UXCL") model.In panels b) and c), respectively, we plot the χ residuals for the best-fitting M1 ("SX-COM+HXPL+UXCL") model (no obscuration), and for our preferred model, M2 ("PC*(SXCOM+HXPL)+UXCL"; with obscuration).
model, which yielded χ 2 /do f =196.06/177=1.108,Γ HX = 1.69 ± 0.16, and moderate data-model residuals as plotted in Fig. 7. Adding a full-covering neutral obscurer (zTBabs) to the model failed to bring any improvement (upper limit to N H of 8 × 10 20 cm −2 ), as model corrections impacting the hard band were instead needed.Again, adding a partial-covering component (M2, or "PC*(SXCOM+HXPL)+UXCL") yielded a good fit, with χ 2 /do f = 177.05/175= 1.012 and good data/model residuals, as plotted in Fig. 7.A set of 500 Monte Carlo simulations indicates that the addition of the partialcovering obscurer improves the fit to XM4 at the >99.8 percent confidence level.In addition, ∆AIC going from a model lacking the obscurer to the model containing the obscurer is −14.62,confirming the necessity of the obscurer in the fit to XM4.The partial-coverer had N H,PC = 2.84 +3.98  −1.29 × 10 23 cm −2 and CF = 0.79 +0.07 −0.13 , with other best-fitting model parameters listed in Table 8.For completeness, we once again fit an M2alt ("PC*(SXCOM+HXCOM)+UXCL") model, achieving a fit nearly identical to that for M2, with χ 2 /do f = 176.50/175= 1.009 and best-fitting parameters listed in Table 8.

Comparison of XMM-Newton spectral fit results
In Fig. 8, we plot the results of MCMC analysis for column density N H,PC and covering fraction CF as a function of the unab-  2022).However, given that we lack energy coverage above 10 keV, and given that we have observed only two obscuration events, such a conclusion for J0458-5202 must be treated as tentative at best.Finally, in Fig. 10, as a summary of our best-fit models, we present the spectra and best-fit models for XM3, XM1, and XM4 in "model space" (i.e., "unfolded" and deconvolved from the instrument responses).One can see the hardening of the hard X-ray spectrum as the source transitions between the unobscured state (XM3), the first obscured state (XM1), and the second, more strongly obscured state (XM4).
One can also see the increasing relative prominence of the Fe Kα line in the hard X-ray spectrum as the transmitted continuum drops to the level in XM4, suggesting that the Fe K line originates in distant material.A set of 500 Monte Carlo simulations for a model lacking the UxClumpy component and instead containing a Gaussian with rest-frame energy centroid fixed at 6.4 keV and width fixed at 1 eV indicate that inclusion of the Gaussian improves the fit at the 93.6 percent confidence levels for XM3, 97.2 (90.0) percent for XM1 (XM2), and 99.4 percent (2.8σ) for XM4.In addition, in XM4, the 90 percent confidence uncertainty on the equivalent width relative to the local continuum is quite high -339 +147 −133 eV.We thus consider detection of the Fe K line in J0458-5202 to be tentative at best and only in XM4.

Spectral fits to eRASS data
We first fit the spectrum for eR1, which had 531 counts in 0.2-5.0keV.We binned it to a minimum of 15 counts per bin, and fit using the C-statistic.When we fit a simple power law, we get a modest fit, with C/do f = 43.39/29= 1.50, and with large datamodel residuals are plotted in Fig. 11.We then applied a "SX-COM+HXPL+UXCL" (M1)8 model.τ warm was not constrained when left free, so we froze it at 45, the best-fitting value from Fig. 8. MCMC uncertainty analysis for column density N H,PC (top panels) and covering fraction CF (bottom panels) as a function of F 1−10 , the unabsorbed 1-10 keV power-law best-fitting M1 (unobscured) model for XM3 (gray), and M2 (obscured) models for XM1 (magenta), and for XM4 (blue).Black crosses represent the best-fitting parameter values; dashed lines indicate the 90 percent confidence limits.F 1−10 is plotted in units of 10 −12 erg cm −2 s −1 .N H,PC is in units of cm −2 .XM2 is omitted for clarity, given its relatively poor model constraints.XM3 did not have a partial-covering obscurer.However, to be able to include it in these panels for the purpose of comparing F 1−10 to XM1 and XM4, we include XM3 here at an arbitrary limit of log(N H,PC )<20.9 to represent the zero value.
XM3.We obtained an improved fit, with C/do f = 35.06/28= 1.25, Γ HX = 2.03 +0.40  −0.36 , power-law F 1−10 = 7.65 +0.37 −0.24 × 10 −13 erg cm −2 s −1 , and observed, absorbed flux values as listed in Table 6.Data-model residuals are plotted in Fig. 11.A set of 500 MC simulations indicate that adding the soft component to the model improves the fit at the >99.8 percent confidence level; the model with the soft component corresponds to a value of T seed = 20 eV* Γ HX = 1.90 ± 0.04 F 0.5−2 = 4.28 +0.17  AIC that is lower by 5.93 compared to the simple power-law model.The addition of a full-covering component does not improve the fit; the upper limit is 3 × 10 20 cm −2 .A partial-covering component does not improve the fit either.Upper limits to N H,PC are (3 − 5) × 10 20 cm −2 when values of CF above 0.75 are assumed, and peg at 1 × 10 24 cm −2 for values of CF 0.7 or lower.These limits, combined with similarity in broadband flux level to XM3, support the notion that eR1 sampled J0458-5202 in an unobscured state.
The spectrum of the other relatively high-flux eRASS observation, eR4, had 297 counts in 0.2-5.0keV.Again, we binned to 15 counts per bin, and fit using the C-statistic.A simple powerlaw yielded an adequate fit, with C/do f = 8.41/15 = 0.56, and unobscured flux F 1−10 = 2.31 +0.85 −0.66 × 10 −13 erg cm −2 s −1 .In order to avoid overparametrizing the data given the limited number of observed counts, we do not explore more complex models9 .The resulting 0.2-5.0 and 0.5-2.0keV observed, absorbed fluxes are listed in Table 6; these flux values are closer to that of eR1 and XM3 (unobscured state) than to those of XM1, XM2, or XM4.In addition, eR4 occurred only 24 days before XM3.We thus infer that eR4 has also sampled J0458-5202 in its unobscured state.Notes.
An asterisk (*) denotes a fixed parameter. (a) Unobscured 1-10 keV observed-frame flux of the hard power-law component. (b)

N
HC and N SC denote, respectively, the normalizations of the hard and soft X-ray CompTT components.Finally, we turn our attention to eR2, eR3, and eR5, which had only 36, 19, and 31 counts in 0.2-5.0keV, respectively.We binned to five counts per bin, and could only fit up to ∼2 keV.All fits used the C-statistic.The data quality was low enough that a simple power law with Γ frozen to 2, and obscured only by the Galactic column, yielded satisfactory fits, with no significant improvement to the fits when thawing Γ.The best fits had C/do f = 6.5/5, 1.2/1, and 14.9/4, respectively.The resulting Xray fluxes for all eRASS spectra are listed in Table 6; they are close to fluxes for the obscured XMM-Newton spectra.

Spectral fits to Swift XRT data
We performed spectral fitting only for Sw3 (90 total counts in 0.2-10 keV); all other spectra had too few counts (30 or fewer) for broadband spectral fitting.
Sw3 occurred 50 and 74 days, respectively, before eR4 and XM3, which were consistent with lack of obscuration.We thus started with the hypothesis that the source was also unobscured during Sw3.We binned Sw3 to 10 counts per bin, and fit the 0.4-4.5 keV band using the C-statistic.A simple power law yielded a poor fit (C/do f = 17.74/7 = 2.52); We thus added a soft excess via warm Comptonization, and fit using a "SX-COM+HXPL+UXCL" model following the M1 fit to XM3, keeping τ warm fixed to 45. Formally, the best fit was good, with C/do f = 6.07/6 = 1.01, and good data/model residuals.A set of 500 MC simulations indicate that adding the soft component to the model improves the fit at the >99.8 percent confidence level; the model with the soft component corresponds to a value of AIC that is lower (by 6.87) compared to the simple power-law model.However, it is suspicious that the best-fitting value of Γ HX was very flat, at 0.7 ± 0.5.The addition of a partial-covering obscurer yields C/do f = 5.55/4 = 1.39;Γ HX remains very poorly constrained, and there is strong degeneracy in both the N H,PC -N WC and CF-N WC planes.Moreover, even with Γ HX frozen in this model, the value of AIC has actually increased from 16.87 in the model lacking the obscurer to 23.55 here, as there is a heavy penalty for the correction for the small sample size.We thus cannot draw any firm conclusions about whether the first obscuration event is still ongoing during Sw3.However, the closeness in time to eR4 and XM3 as well as the similarity in soft X-ray flux values (see Table 6) would argue that it is more likely that the event had ended (or was ending) by this time.
For Sw1, 2, 4, 5, 6, 7, and 8, we used the UKSSDC Swift-XRT products website10 to derive count rates and errors via the Bayesian method of Kraft et al. (1991).We converted to fluxes assuming a power-law model with a photon index of 2.0.Fluxes are listed in Table 6.

Checking for reddening in the optical/UV continuum SED
We now discuss fits to the time-variable optical/UV SED, using the six-filter Swift UVOT observations Sw3-8.As demonstrated by the light curves in Fig. 2, the optical/UV SED becomes relatively redder as overall flux decreases from observations Sw3 and Sw4 to Sw5-7, and bluer as flux increases to Sw8.As Sw4-8 occur during the second X-ray obscuration event, we thus fit their six-channel SEDs to test if there is associated reddening impacting the optical/UV SED.We perform all fits in Xspec.
To model extinction associated with the Galactic column, we include a redden component with E(B − V) fixed to 0.009 mag in all fits, based on the dust maps of Schlegel et al. (1998).We also include a Sb host galaxy template from the SWIRE galaxy template library (Polletta et al. 2007) in all fits.However, due to the vast difference in spectral shape between the Seyfert component and the galaxy component, the galaxy component was not required in the fits; the AGN continuum clearly dominates.We obtained small upper limits to the host template's normalization: 2.0-7.5 eV (1653-6199 Å) flux < 1.34 × 10 −14 erg cm −2 s −1 , compared to 1.22 × 10 −11 erg cm −2 s −1 for J0458-5202 and for the same bandpass during Sw7.
We first attempted to model the optical/UV SEDs with simple power laws, with no reddening beyond Galactic, and with all photon indices and normalizations untied.This is a relatively poor fit, with the sum of χ 2 /do f across all six spectra equal to 203.30/23, and with strong convex curvature across the UVOT band.However, the fits did support a general trend of steeper spectral slope at lower 2.0-7.5 eV flux, with Γ = 1.30 ± 0.04 and 1.39 ± 0.04 for Sw3 and Sw4, steepening to 1.66 ± 0.07 and 1.58 +0.04 −0.05 by Sw6 and Sw7, respectively, and flattening to 1.33 ± 0.05 by Sw8, and consistent with the picture of spectral variability in Fig. 2. Applying a reddening component at the system redshift with zredden did not yield any improvement in fit, with upper limits to E(B − V) ranging from 0.017 to 0.056.Data/model residuals are plotted in Fig. 12.
We substituted the power-law components in both the reddened and unreddened fits with multitemperature blackbody disk components, modeled with diskpbb.The index describing the power-law dependence of temperature was tied across all fits.We allowed the temperature at the inner radius, T 1 , to be untied across all SEDs, since tying these parameters together in a joint fit caused χ 2 /do f to increase by over 70/6.Without reddening, the fit was still poor, with χ 2 /do f = 77.24/23.
Allowing for reddening improved the fit considerably, to χ 2 /do f = 13.56/17.The data-model residuals for both cases are plotted in Fig. 12, and best-fitting parameters for the reddened case are displayed in Table 9.Compared to the unreddened case, most of the improvement is in the W1 band bin.The W1 bandpass's effective wavelength is 2630 Å, which for J0458-5202, probes roughly 2060 Å in the rest frame.This wavelength is very close to 2175 Å, the wavelength at which there is a strong absorption bump associated with absorption in many spiral galaxies, including the Milky Way, for instance as seen in A λ (λ) plots of Galactic absorption (e.g., Cardelli et al. 1989).That is, the zredden component is impacting the W1 band significantly in this manner, and such reddening is required for each UVOT spectrum: the resulting increase in χ 2 /do f when E(B − V) is frozen at zero for any one spectrum (when reddening is turned off) is always ≥ 5/1, with a corresponding increase in the Akaike information criterion value (∆AIC) of at least +7.3.We emphasize that the best-fit values of E(B − V) are always non-zero, but always less than 0.11, and that they do not increase as optical/UV flux -nor X-ray flux -decreases.These observations suggest that the decrease in observed optical/UV flux from Sw3-4 to Sw6-7 is not attributable to an increase in reddening.
Assuming that the dust/gas ratio in the host galaxy of J0458-5202 is identical to that for the Galaxy, N H = A V × 2.69 × 10 21 cm −2 mag −1 (Nowak et al. 2012), and assuming R V ≡ A V /E(B−V) = 3.1, then the average value of E(B−V) from the diskpbb fits, 0.08, corresponds to a column of 6.7×10 20 cm −2 .This value is typical for line-of-sight galactic absorption and smaller than the upper limits from fitting a full-covering column to the XMM-Newton spectra.

Fits to optical emission-line spectra
All optical spectra were corrected for Galactic extinction.We used E(B − V) = 0.009 from the dust maps of Schlegel et al. (1998), obtained via the sfdmap module in python.We assumed R ≡ A V /E(B − V) = 3.1, and used the reddening curve of Fitzpatrick (1999), via the extinction module in python.Absolute flux calibration was not available for the "red" setups for three SALT spectra, #5, 6 and 7, as well as for both SALT setups in spectrum #10.For spectra #4, 5, 6, 7, and 9, fluxes were manually adjusted (gray scaling) to match the Hβ-[O iii] region from the simultaneous "blue" setups.Both segments for spectrum #10 were adjusted to match the Hβ-[O iii] region from spectrum #9.
The spectra were taken across different instruments, spanning a range of wavelength resolutions, slit widths, air masses, and seeing conditions.However, the integrated line flux of the narrow [O iii] emission line is a constant that can be used to cross-calibrate the spectra (van Groningen & Wanders 1992; Fausnaugh 2017) to account for differences in instrument resolutions, seeing conditions, wavelength offsets, etc.
We flux-scaled all spectra using spectrum #1 (SALT) as a reference spectrum, and following Saha et al. (submitted;their Sect. 4).To summarize our method, for each spectrum (in #2-10), we defined the continuum via bins on either side of the [O iii] λ5007 emission line, interpolated, subtracted the continuum, and used a Goodman-Weare MCMC sampler to fit the emission line to determine wavelength and flux offsets.Wavelength shifts for the other spectra were typically ∼ +1.6 Å for the FORS2 spectra, ∼ −0.6 Å for the SAAO spectrum, and ∼ −0.2 to −0.6 Å for the other SALT spectra.Flux scaling values (applied as grayshifting) were in the range 0.95-1.71.
Prior to this campaign, the redshift of J0458-5202 was not known.From application of our best-fit model to the [O iii] λ5007 emission line in spectrum #1 (see below for fit details), we obtain z = 0.276.
The resulting spectra are presented in Figs. 13 and 14, and they are all typical for a Sy 1.The most obvious features by eye present in all spectra are broad emission lines associated with Hα λ6563, Hβ λ4861, Hγ λ4341, Hδ λ4102, and He i λ5876; broad Fe ii emission peaking near 4550 Å; and narrow [O iii]λλ5007,4959.
Other narrow emission lines due to [N ii] λ6584 (the accompanying 6548 Å line is blended with broad Hα and not visually obvious, but modeled in fits below), and [S ii] λλ6731,6716 are also noticeable in most, if not all, spectra.Figs. 13 and 14 illustrate that there are no obvious, major changes (for example, disappearance of broad lines) between the spectra.The continuum levels across the spectra seem to vary, but such variability is likely not directly representative of intrinsic continuum variability.There can be artifacts associated with the [O iii] re-scaling process, or variations in seeing and sky transparency during the observations.
First, we approximated the continuum, fitting by masking out all line features.For the Hβ fitting, our continuum included a power-law continuum; for the Hα fitting, we required a broken power-law continuum to obtain satisfactory fits.We also included an Fe ii template, convolved with a Gaussian to include the effects of velocity broadening.For the Hβ fitting, we used the templates of Kovačević et al. (2010) 11 , which provide modeling for the F, S, G, P, and I Zw 1 groups of Fe ii separately in the 4000-5500 Å range.The F group, whose peaks blend to form a broad emission feature near 4550 Å, was especially prominent in J0458-5202.For the Hα fitting, we used the template of Bruhweiler & Verner (2008), which is based on observations of I Zw 1.
We also included a host galaxy template from the SWIRE library (Polletta et al. 2007).There are no host galaxy morphology classifications available on NED for J0458-5202.We used an Sb template, and such a choice is admittedly arbitrary.However, our fits were insensitive to the sub-type (Sa-d) because a host galaxy component was not required in the fits: fits for both the Hα and Hβ regions preferred the power-law component to strongly dominate the continuum and to fully describe the total continuum.Finally, we added Gaussians as needed for the emission lines.The final model for the Hβ region fitting contains: -A continuum dominated by a power-law with best-fitting values of spectral index α spanning 1.51-1.95-Narrow [O iii] emission lines at both 5007 and 4959 Å (rest frame).Each line profile was modeled with the sum of 1) a narrow Gaussian with rest-frame width σ ∼ 3-5 Å (values between the two lines were kept tied) and energy centroid within 1-2 eV of the expected rest frame energy, plus 2) a wider Gaussian with rest-frame σ ∼ 10-12 Å (again, widths between the two lines were kept tied), and energy centroid 9-13 Å blue (5007 Å line) or 6-9 Å blue (4959 Å line) of the expected rest frame energy.We tried fitting with the amplitude of the 4959 Å broad component tied to 1/3 that of the 5007 Å component, but residuals became poor, so we had to leave the amplitudes untied.In the best fits, the mean and standard deviation of the ratio of amplitudes All: p > 1.1 (pegged at 1.5) χ 2 /do f = 13.92/17 Notes.The model used was redden(zredden(cflux * diskpbb)).p is the diskpbb power-law index describing the radial temperature relation, T (r).Reddened Disk PBB Fig. 12. Data-model residuals to fits to the optical/UV SED using the Swift UVOT data from observations Sw3-8.Data for Sw3, 4, 5, 6, 7, and 8 are plotted in black, blue, purple, red, orange, and green, respectively.In the top panels, the continuum is modeled as a simple power law.In the bottom panels, it is modeled as a multitemperature disk blackbody.In the panels in the left column, the models contain no reddening, whereas in the right column, the models do contain reddening.was 0.28 ± 0.04.NLR outflows in AGN, manifested by blueasymmetric [O iii] profiles, are not uncommon in both nearby and high-redshift AGNs (e.g., Schmidt et al. 2018;Leung et al. 2019;Mahoro et al. 2023).In J0458-5202, the energy centroid offsets correspond to line-of-sight velocities on the order of 400-800 km s −1 relative to systemic, and the line widths correspond to FWHM velocity dispersions of ∼600-700 km s −1 , similar to typical values obtained by Schmidt et al. (2018) and Leung et al. (2019).
-Broad emission components for Hβ, Hγ, and Hδ; velocity widths were on the order of 4650-5050 km s −1 , 4900-6500 km s −1 , and 3600-5200 km s −1 , respectively.-Narrow emission components for Hβ, Hγ, and Hδ.Their velocity widths were tied to that for the [O iii]λ5007 line.It was necessary to keep their energy centroids frozen to expected rest-frame values, since the Balmer profiles were dominated by the broad components.-There was no significant emission from broad or narrow He ii λ4686 lines.
For the Hα region fits, the final model for the contains: -A continuum modeled as a broken power law (an unbroken power law did not yield satisfactory fits).residuals in all fits.(As an aside, we subsequently re-fit the Hβ profile to test for presence of this "ultra-broad" component in Hβ, but it was not required in those fits.Forcing an "ultra-broad" component with amplitude and σ set to values 0.65 and 2.5 times those of the broad component caused fits to worse considerably, with χ 2 typically increasing by +4 and AIC increasing by up to 100.) -Broad emission from He i, with σ typically ∼ 30-45 Å.
The other was centered at 6450 Å (rest frame) = 8224 Å (observed frame), and was narrower, with σ = 3 Å (observed frame).These features were strongest in spectra #5-7, moderately strong in #1-4, and weak/absent in #8-10.Their strengths correlate well with the strengths of atmospheric absorption features masked out above and we conclude they are also atmospheric in nature.
Gaussian components are adequate to fit all broad and narrow emission lines.There is no strong evidence for asymmetry in any of the broad Balmer components.For brevity, all fit results are presented in Appendix C, with final best-fitting parameters listed in Tables C.4 and C.5 for Hβ and Hα, respectively.Two typical spectral decompositions are plotted in Figs. 15 and 16, for Hβ and Hα, respectively.
One main takeaway from the spectral fits is that there are no major changes in continuum shape or in broad line profiles or heights across spectra.If either of the X-ray obscurers contained substantial dust, and if they occulted substantial fractions of the BLR in addition to occulting the X-ray corona, then we might expect the Hβ and Hα broad component fluxes to drop appreciably due to the extinction, accompanied by an increase in the observed ratio of Hα/Hβ broad component fluxes, in spectra #1-7 (event 1) and/or #9-10 (event 2) compared to spectrum #8 (unobscured).However, the average Hβ integrated broad component fluxes for spectra #1-7, #8, and #9-10 are 4.44 ± 0.06 × 10 −12 , 4.13 ± 0.06 × 10 −12 , and 4.36 ± 0.06 × 10 −12 ph cm −2 s −1 , respectively.Similarly, the ratio of Hα/Hβ broad component fluxes remains roughly constant: Given the necessity of the "ultra-broad" component to the Hα profile, we consider the ratio of the maximum heights 12 of the wide components.The average Hα/Hβ ratio across spectra #1-7 (event 1), spectrum #8 (unobscured), and spectra #9-10 (event 2) are 2.25 ± 0.27, 2.22 ± 0.04, and 2.67 ± 0.03, respectively.
We conclude that: 1) neither X-ray-obscuring cloud contains substantial dust and we can rule out values of E(B − V) above ∼0.1-2(corresponding to values of A V ∼ 0.3-0.7) in both clouds or 2) neither cloud significantly occulted the line of sight to the BLR.
Furthermore, spectrum #5 coincides with the dip in optical photometric flux near MJD ∼ 59210-59250.However, there is no change in the broad Balmer component fluxes or ratio here either.We conclude that the optical continuum dip is a luminosity variation intrinsic to the disk, and not a product of dust obscuration within cloud #1.

Discussion
In this paper, we explore the first major obscuration events detected with eROSITA and studied with triggered multiwavelength followups.We have tracked the durations and properties of the obscurers through a series of X-ray flux and spectral monitoring observations.We summarize our main observations: -eROSITA's all-sky X-ray surveys revealed in August 2020 (eR2) that J0458-5202's 0.5-2.0keV X-ray flux had dropped by a factor of ∼11 compared to February 2020 (eR1).Follow-up XMM-Newton EPIC spectra (XM1 and XM2) revealed obscuration by a Compton-thin cloud with N H,PC ∼ 1 × 10 22 cm −2 and covering fraction CF ∼ 60 percent (using values from XM1).The soft X-ray flux remained low through eR3 (February 2021), though it had recovered by Sw3 (June 2021).Assuming that eR1 represented an unobscured state, this obscuration event lasted somewhere between 309 and 539 days (observed frame).-By August 2021 (eR4), J0458-5202 had returned to an unobscured state, in which it remained through at least September 2021 (XM3).It displayed relatively higher periods of soft X-ray flux (F 0.5−2 ∼ (3.0-4.3)×10 −13 erg cm −2 s −1 ) during this state, and the XMM-Newton EPIC spectrum taken during XM3 was very typical for unobscured Seyferts.-By February 2022 (eR5), the soft X-ray flux had fallen back to F 0.5−2 = 3.4 +1.1 −1.0 × 10 −14 erg cm −2 s −1 , and remained near this level through at least February 2023 (Sw4, XM4, Sw5-8).As revealed by the XMM-Newton EPIC spectrum taken during XM4, the source was partially obscured by a new transiting cloud, with N H,PC ∼ 2.8 ×10 23 cm −2 and CF ∼ 79 percent.The duration of obscuration is a minimum of 527 days in the observed frame; the eclipse had not ended as of the most recent Swift observation (Sw8, in February 2023).
-Optical and UV continuum photometry obtained via monitoring with ground-based telescopes, Swift UVOT and XMM-Newton OM revealed minimal variations in optical/UV accretion disk luminosity, with long-term variations being less than ∼ 0.3 mag in both the B and M2 bands.An exception is a temporary drop in B-band flux by ∼0.4 mag over 40 days in early 2021.-We obtained a series of optical spectra of J0458-5202 at SALT, VLT, and SAAO to establish the source's redshift and to monitor Balmer profiles.We deconvolved the Hα, Hβ, and Hγ profiles into narrow and broad components, with the broad Hβ component correspond to an average FWHM velocity of 4870 km s −1 .We observed no significant changes in profile or intensity, and no significant changes in Balmer decrement.In addition, the [O iii] λ5007 and λ4959 emission lines were each modeled by a narrow component (FWHM velocity width ∼ 400-700 km s −1 ) and a blue-skewed broad component (offset by ∼ 400-800 km s −1 ), consistent with a bulk outflow.
These observations argue strongly against the soft X-ray dimming being due to temporary drops in accretion luminosity.We observed no major variations in the Balmer intensities or profiles, suggesting that the >13.6 eV continuum held steady.The fact that we did not observe any major decrease in the optical/UV thermal emission emanating from the inner accretion disk (< 18 lt-dy, or < 500R g , as estimated in Sect.6.1) strongly argues against any major decreases in accretion rate or in disk temperature, for instance, due to propagating cold fronts (Ross et al. 2018).Importantly, the minor variations in optical/UV flux we did observe were not correlated with the onset of X-ray obscuration.Finally, we note the improvement in X-ray spectral fits to XM1 and XM4 when obscuration components were added.
As noted above, the high soft X-ray flux states, probed by eR1, eR4, and XM3, support the absence of eclipsing by clouds along the line of sight.The limited data quantity means we cannot distinguish between a scenario wherein obscuration in J0458-5202 is due to discrete X-ray-obscuring clouds versus a scenario wherein obscuration is caused by a contiguous but patchy structure containing some "holes."In this scenario, the ratio of obscured/unobscured lines of sight can be estimated very crudely via the fraction of time having observed the source in obscured/unobscured states.During X-ray monitoring spanning 1090 days, J0458-5202 spent at least 846 days, or 78 percent, in an obscured state.However, we must note the limited number of observed transitions between obscured and unobscured states during our campaign, so this value is to be treated with extreme caution pending significant amounts of additional longterm monitoring.
In addition, the limited data quantity, particularly with regard to obtaining a large number of high-quality spectra, means we are not highly sensitive to tracking the transverse column density profile of the obscuring matter, as has been done by, for example, Maiolino et al. (2010), Rivers et al. (2011Rivers et al. ( , 2015)), and Markowitz et al. (2014).For the purposes of this Discussion, we assume for simplicity that the two obscuration events correspond to two separate X-ray-obscuring clouds transiting the line of sight, and henceforth referred to as "cloud 1" and "cloud 2." A host galaxy template was included in the fit, but the best-fit value of its amplitude is zero.

Basic system parameters
We first obtain virial estimates for the radial location of the BLR and for the black hole mass, M BH , using the average value of Hβ FWHM velocity across all spectral fits, 4870 km s −1 (rest frame), and the empirical relation between optical luminosity and BLR radius R BLR in Seyferts (Bentz et al. 2009(Bentz et al. , 2013)).From the optical spectral fits, we find an average flux density of λF 5100Å to be 7.2 × 10 −16 erg cm −2 s −1 Å −1 , which for a luminosity distance of 1400 Mpc, corresponds to a monochromatic luminosity of λL 5100Å = 1.1×10 45 erg s −1 .We apply the λL 5100Å −R BLR relation of Bentz et al. (2013), log(R BLR , lt-dy) = K + αlog(λL 5100Å / (10 44 erg cm −2 s −1 )), with their best-fitting values of K = 1.56 and α = 0.55.We obtain R BLR = 2.8 × 10 17 cm = 130 ltdays.We do not have information on the inclination of the system, so we assume an arbitrary virial factor f of 1.0; we obtain The corresponding Eddington luminosity is thus L Edd = 7.8 × 10 46 erg s −1 .Considering the scatter in the λL 5100Å − R BLR relation of Bentz et al. (2013), we assign an uncertainty of ∼ 0.13 dex to R BLR , M BH , and L Edd .
We now estimate the bolometric luminosity, L Bol , via two methods: First, we estimate L Bol from the optical luminosity, using the bolometric correction of Duras et al. (2020).From the Swift UVOT observations, the observed B-band flux density is typically 1.8−2.1×10−15 erg cm −2 s −1 Å −1 .These values include contamination from the host galaxy, but the spectral decomposition of the optical spectra suggest that by 4400Å, such contamination is negligible, likely less than 5 percent.The corresponding monochromatic luminosity, λL 4400Å , is 2 × 10 45 erg cm −2 s −1 .From Eq. 4.2 of Duras et al. (2020), the correction factor is 5, and we obtain L Bol = 1 × 10 46 erg cm −2 s −1 , and L Bol /L Edd = 0.13.Given uncertainties on flux density, correction factor, and on L Edd of 0.013 dex, 0.27 dex, and 0.13 dex, respectively, we assign a total uncertainty of 0.41 dex to this estimate of L Bol .
Using this estimate of L Bol , we estimate the radial location of the dust sublimation region, assuming isotropic luminosity.We follow Sect.2.1 of Nenkova et al. (2008, also discussed in Barvainis 1987): R dust = 0.4 (T sub /(1500 K)) −2.6 (L Bol /10 45 egs) 0.5 .For a sublimation temperature T sub of 1500 K, our estimate of R dust is 1.3 pc (1500 lt-day).However, as discussed in Nenkova et al. (2008), the transition between dusty and non-dusty regions is likely gradual, as relatively larger grains are likely able to survive at radii factors of up to ∼2-3 smaller than predicted from this equation.In addition, different dust species sublimate at different temperatures, e.g., sublimation of graphite grains and silicate grains at T = 1500 and T = 1000 K, respectively (Schart-mann et al. 2005).We therefore treat this estimate of R dust as an estimate of the dust sublimation radial "zone." Finally, we estimate the radial location of the optical and UV thermal emission in the inner accretion disk.We follow Edelson et al. (2019, their Eq. 1) for a standard optically thick, geometrically thin disk. 13We set the multiplicative scaling factor in that equation to X = 2.49 for the flux-weighted radius.For the M2 band (2246 Å observed frame ∼ 1760 Å rest frame), peak emission is at 4.6 +2.9 −1.8 lt-dy, which corresponds to 130 +90 −50 gravitational radii R g for a 6.2 × 10 8 M ⊙ black hole14 .For the B band (4400 Å observed frame = 3448 Å rest frame), we obtain radii of 11.3 +6.9 −4.4 lt-dy, or 310 +190 −120 R g ).

Constraints on clouds' locations and sizes
We can use constraints on ionization parameter, eclipse duration, and column density to derive the distance R cl from the X-ray continuum source to each cloud, assuming for simplicity that clouds' kinetics are dominated by Keplerian motion.We follow Lamer et al. (2003, their Sect. 4): R cl = 4 × 10 16 M 1/5 7 L 2/5 ion,42 t 2/5 d N −2/5 H,22 ξ −2/5 cm.Here, M 7 is the black hole mass in units of 10 7 M ⊙ , t d is the eclipse duration in units of days, N H,22 = N H /(10 22 cm −2 ), and L ion,42 is the ionizing (>13.6 eV) continuum luminosity in units of 10 42 erg cm −2 s −1 .To estimate L ion , we follow Vasudevan & Fabian (2009) and adopt L ion = 0.2 × L Bol = 2 × 10 45 erg cm −2 s −1 .The ionization parameter ξ is defined as L ion /(n e r 2 ), where n e is the gas electron density.To estimate ξ, we revisited the best-fit "PC*(SXCOM+HXPL)+UXCL" models for XM1 and XM4, and replaced the partial-covering neutral obscuring component with an ionized one, modeled with zxipcf in Xspec.We obtain upper limits to log(ξ, erg cm s −1 ) of +1.5 for cloud 1 and +2.5 for cloud 2. For cloud 1, the eclipse duration is constrained to be in the range 309-539 days observed frame, or 242-422 days rest frame.The upper limit on ξ translates into a lower limit on cloud location: R cl,1 > 3.0 × 10 18 cm = 1100 lt-dy, which would place it at radii well outside the Hβ-emitting region of the BLR, and approaching the dust sublimation zone.For cloud 2, which was still ongoing as of the most recent observations, the eclipse duration is a minimum of 527 days observed frame = 413 days rest frame.For this cloud, we obtain R cl,2 > 3.3 × 10 17 cm = 130 ltdy.This lower limit on radial location is commensurate with the Hβ region of the BLR.We emphasize the notion of "commensurate," as X-ray absorbers lie along the line of sight, but BLR clouds likely contain components out of the line of sight.
Still working with the simplifying assumption that the clouds are in strictly Keplerian motion, we can derive upper limits on their transverse velocity v cl = (GM BH R −1 cl ) 1/2 .For clouds 1 and 2, respectively, we obtain v cl,1 < 1700 km s −1 and v cl,2 < 5000 km s −1 .The clouds' sizes in the transverse direction, D cl can be estimated via D cl = v cl t D .For cloud 1, adopting t D = 400 days, D cl,1 < 6 × 10 15 cm ∼ 2 lt-dy ∼ 65 R g .For cloud 2, we have a situation where transverse velocity is an upper limit and duration is a lower limit.Nonetheless, adopting t D ∼ 400 days, we obtain D cl,2 ∼ 2×10 16 cm ∼ 6.5 lt-dy ∼ 180 R g .Such inferred diameters are typically an order of magnitude more than those found by Markowitz et al. (2014), and two orders of magnitude greater than clouds inferred to exist in the BLR of NGC 1365 (Maiolino et al. 2010).
We can estimate crude lower limits to the cloud density n H , via N H /D cl .For cloud 1, we obtain n H,1 > 2 × 10 6 cm −3 .For cloud 2, we obtain n H,2 > 2 × 10 7 cm −3 .Such density limits are consistent with typical values for the BLR.
As a caveat, we recall that if the clouds' true motion is strongly non-Keplerian (for instance, if the velocities contain a strong radial component, such as that associated with a wind flowing vertically upward from a disk), then the accuracy of these estimates of R cl , D cl , and n H can be compromised.
Finally, we can consider the results in the context of the clumpy torus model of Nenkova et al. (2008), wherein all obscuration is accounted for by a large number (on the order of 10 4−5 ) discrete clouds, and obscuration for any given source is a probability depending on source inclination and cloud distribution parameters.Clouds are distributed preferentially toward the equatorial plane, but with an angular distribution that is Gaussian.The average number of clouds along our line of sight to the central source is N i (σ, i, N 0 ) = N 0 exp(−((90 − i)/σ) 2 ).Here, N 0 denotes the average number of clouds along an equatorial ray, typically on the order of 5-15; σ denotes the angular width of the distribution.The inclination of the system is denoted by i: 0 • for face-on, and 90 • for edge-on.The likelihood to observe a source in an obscured state is dependent on N i and i, (e.g., Nikutta et al. 2009) P obsc = 1−exp(−N i ).The clumpy model was originally used to model the distribution of clouds residing in the dusty torus, that is, outside the dust sublimation radius, but for the purpose of this discussion, we can consider that the radial distribution of clouds may extend to inside the dust sublimation radius.
In the case of J0458-5202, we have observed transitions between zero and one cloud along the line of sight, such that N i is always either 0 or 1.If we assume for academic purposes that N 0 is, say, 5 (10), then having N i = 1 means that ((90 − i)/σ) 2 must be 1.26 (1.52).Though we lack information on the inclination of J0458-5202, we can assume for simplicity that the system is relatively face-on, with an arbitrarily chosen inclination angle of 30 • .In this case, the above value of N 0 = 5 (10) yields a value of σ = 48 • (39 • ).Such values are roughly consistent with values for the tori of type 1 Seyferts obtained by Ramos Almeida et al. (2011) in their SED model fits.

Exploring whether the clouds dusty or non-dusty
We did not observe drastic changes in optical or UV continuum luminosity between the X-ray obscured and unobscured periods.The optical/UV thermal emission emanates from a relatively compact region, roughly 5 − 11 lt-dy, as discussed in Sect.6.1.The cloud sizes are crudely estimated to be on the order of ≲2-7 lt-day (Sect.6.2).If it were the case that the clouds somehow partially obscured the X-ray-emitting region while simultaneously covering all or most of the optical/UV continuum-emitting region of the disk, then we might expect to observe evidence for extinction, particularly in the UV band, if the clouds contained significant amounts of embedded dust.
As a reminder, our best-fit values of column density were N H,PC = 1.0×10 22 cm −2 (cloud 1) and 2.8×10 23 cm −2 (cloud 2).To estimate the corresponding dust content, we assume a dust/gas ratio equivalent to the Galactic value, N H = A V × 2.69 × 10 21 cm −2 mag −1 (Nowak et al. 2012).We obtain A V ∼ 3.7 mag (cloud 1) and 105 mag (cloud 2).Following Cardelli et al. (1989), we can translate these V-band extinctions into estimates of extinction in each of the M2 and B bands; we use A M2 ∼ 2.7A V and A B ∼ 1.3A V , respectively.Subsequently, for cloud 1, we expect extinction levels of A M2 ∼ 10 mag and A B ∼ 4.8 mag.For cloud 2, we expect A M2 ∼ 280 mag and A B ∼ 135 mag.
However, we do not observe such strong variations in the light curves of these bands, with conservative upper limits on observed variations of 0.4 mag for the M2 band and 0.6 mag for B band, consistent with standard variability in persistentlyaccreting Seyferts.
We conclude that if either cloud fully covers the optical/UVemitting regions of the disk, they must contain negligible dust: we estimate a limit of A V < ∼0.15 mag, assuming R Similarly, the optical-UV continuum SED fitting (Sect.4) indicated upper limits to E(B − V) of roughly 0.1 mag.On the basis of the stability of the Hβ broad component flux and the ratio of broad Hα/Hβ emission, if the X-ray-obscuring gas covered substantial fractions of the BLR, then they must be dust-free, too.In addition, the extreme tension between the lack of observed extinction and the high expected values of extinction argue against a scenario wherein the clouds are dusty and they partially cover the optical/UV-emitting annular region of the disk with the same covering fractions impacting the X-ray band.
However, an alternative and simple scenario is based on geometry: the clouds may partially intersect the line of sight to the X-ray corona, but they do not intersect the bulk of the lines of sight to the optical/UV continuum-emitting annular region in the disk.In such a case, both dusty and dust-free clouds could be accommodated.Similarly, if the X-ray obscurers were discrete clouds residing in the BLR, then single clouds could partially obscure the X-ray corona, but not the other BLR clouds, depending on the BLR's precise geometry and the system's inclination angle.

Nature and origin of the clouds
The lower limits to radial locations of the clouds, estimated in Sect.6.2, are commensurate with the far-outer BLR or dust sublimation zone (cloud 1), or commensurate with the Hβ-emitting BLR (cloud 2).The limits on cloud density and cloud size estimated in Sect.6.2 are very crude, but these limits can accommodate typical values for BLR gas (e.g., Netzer 2013).Several previous works, such as Mushotzky et al. (1978, NGC 4151), Risaliti et al. (2009, NGC 1365), Risaliti et al. (2011, Mkn 766), and Miniutti et al. (2014, ESO323−G77), found support for their observed X-ray-eclipsing clouds to be commensurate with, and possibly associated with, the same clouds that emit optical broad lines, due to inferred radial locations, densities, and/or sizes.The two obscuration events observed in J0458-5202 could potentially also fall into this category, though this statement is likely more applicable to cloud 2, given the limit on its inferred radial location.The events observed in J0458-5202 are orders of magnitude longer in duration than the events in Risaliti et al. (2009) and others, on timescales on the order of about a day or shorter.This discrepancy is likely an artifact of J0458-5202's having a black hole mass ∼1-2 orders of magnitude larger than these objects and having size scales for structures scale at least roughly linearly with black hole mass.
The clouds' ionization levels are poorly constrained, although ionization levels above ∼ 10 2.5 erg cm s −1 are excluded; it is possible (albeit highly speculative) that these clouds' vol-umes are dominated by neutral gas and have not yet reached an ionization equilibrium.
The obscuration events in J0458-5202 are generally consistent with multiple models of cloud formation.A wind launched from the disk and accelerated outward is a strong possibility and there are multiple models for such winds.A dusty, turbulent wind launched from the disk may form the low-ionization part of the BLR (Czerny & Hryniewicz 2011;Naddaf et al. 2021).Such a wind can launch from the region of the disk where the temperature is near 1000 K, which we estimate to be 200 +130 −75 lt-dy, consistent with the radial location of cloud 2. Another possibility is magnetohydrodynamic-driven winds (e.g., Fukumura et al. 2010), wherein gas is accelerated outward from the accretion disk along magnetic field lines.The obscurers are also broadly consistent with a UV line-driven wind launched from the accretion disk (Proga et al. 2000;Proga 2005).Following Giustini & Proga (2019), the estimated values of M BH and L Bol /L Edd place J0458-5202 in the range of [M BH , L Bol /L Edd ] space where UV-driven line winds can be launched and there is sufficient UV radiation from the disk to sustain their driving.
One can consider a single stream or filament that partially covers the X-ray corona.Such a stream would be rotated out of the line of sight on dynamical timescales, which for a black hole mass of 6 × 10 8 M ⊙ and for radii of ≳4 lt-dy is ≳360 days, consistent with the event durations in J0458-5202.However, at such radial distances, it would be challenging to only partially, not fully, cover the compact corona.
Another plausible scenario is a non-homogeneous disk wind containing both localized embedded dense regions of X-rayobscuring gas (e.g., Waters et al. 2022) and regions where column densities are low enough to not obscure X-rays.Critically, if the wind is continually replenished from the disk and if the number density of clumps remains roughly constant, then such a scenario can explain the long durations of sustained partial-covering obscuration.Support for outflowing winds as the source of sustained periods of partial-covering, Compton-thin X-ray obscuration has been found in several local Seyferts (e.g., Mehdipour et al. 2017Mehdipour et al. , 2022;;Kang et al. 2023), and J0458-5202 may thus be in that same category.The similarity in covering fractions during the two events in J0458-5202 could simply indicate similar levels of clumping during each wind event.
On the other hand, it is not clear what would case a wind to launch in J0458-5202, shut off the supply of gas to end the wind, and launch a second time, given that disk luminosity was minimally variable during the entire campaign.One possible cause is that both obscuration events are part of the same continuous wind outflow, with dense X-ray-obscuring clumps being absent (or having non-detectable column densities) during the unobscured states (sampled by eR1 and eR4/XM3).
If the X-ray obscurers in J0458-5202 are outflowing winds located interior to the BLR (Dehghanian et al. 2019), and if the optical/UV continuum from the disk must pass through the wind before intersecting the BLR, and if the wind's number density is sufficiently high enough to impact the >13.6 eV continuum, then the observed lack of response in the Balmer lines to the presence or absence of the obscurer along our line of sight would suggest that the presence or absence of the wind has no immediate global impact on the BLR (even considering the >100-day light-travel time from the optical-UV-emitting disk to the BLR).In such a case, perhaps the wind's global covering factor as seen by the X-ray corona is small and we, as observers, are "lucky" that the wind crosses our line of sight to the X-ray corona.Alternately, if the wind's global covering fraction is high, then the wind's number density must be lower than an order of 5 × 10 10 cm −3 as to not impact transmission of >13.6 eV continuum photons that ultimately power the Balmer lines (Dehghanian et al. 2019) 15 .
Finally, if the clouds originate at radii larger than the dust sublimation zone, then filament-like, dusty winds driven by infrared radiation (Dorodnitsyn et al. 2012) or by X-ray heating and radiation pressure (Wada 2012) can be relevant.Scenarios incorporating fragmentation of cooling clouds into mildlyionized or neutral filaments such as those of Gaspari et al. (2013Gaspari et al. ( , 2018) ) and Sparre et al. (2019) are also relevant at such radial locations.
6.5.A brief comment on the anticorrelation between N H,PC and hard X-ray flux As noted in Sect.3.1, we find evidence from the fits to the XMM-Newton EPIC spectra for an anticorrelation between N H,PC and unabsorbed power-law flux F 1−10 .We consider this anticorrelation to be only tentative at best given that it is based on only two obscuration events and given that the lack of > 10 keV coverage induces some degeneracy for cloud 2.
Nonetheless, such anticorrelations have been reported previously, by Couto et al. (2016) in their monitoring study of the partial-covering obscurer in NGC 4151, and by Marchesi et al. (2022) in their study of NGC 1358's transition between Compton-thick-and Compton-thin-obscured states. 16As discussed in Krolik & Kriss (2001) and Couto et al. (2016), lowerflux states may see high columns of neutral or lowly-ionized gas because more material condenses from higher-ionization states.Meanwhile, at sufficiently high luminosities, an excess of ionizing radiation may overionize the wind, causing it to disappear (e.g., Giustini & Proga 2019).Alternatively, as also discussed in Couto et al. (2016), if radiation pressure is responsible for pushing gas structures into the line of sight, then it is possible that the lower continuum radiation during the lower flux states may simply be relatively less efficient at driving the gas.

Conclusions
eROSITA's all-sky X-ray scans have been opening a new window into time-domain astronomy, thus providing an opportunity to search for major changes in soft X-ray flux on timescales of six months or longer in a given object.By monitoring a very large starting sample of AGNs (on the order of a million), we can thus amplify small numbers of events that occur rarely on a per-object basis.Such events include major changes in soft Xray flux due to obscuration by discrete clouds transiting the line of sight to the X-ray corona.
In this paper, we present the first X-ray obscuration events detected in a Seyfert galaxy using eROSITA, including the results of a two-year multiwavelength follow-up campaign.In J0458-5202 (EC 04570-5206, eRASSt J045815−520200), a Sy 1 AGN at z=0.276, we detected two major, separate dips in soft X-ray flux: the first in 2020, followed by a recovery in soft X-ray flux by August 2021, and a second soft X-ray dip occurring by early 2022.From October 2020 through February 2023, we complemented eRASS X-ray monitoring with X-ray flux and spectral monitoring courtesy of XMM-Newton EPIC, Swift XRT and NICER, optical and UV photometric monitoring via XMM-Newton OM, Swift UVOT, and ground-based photometry, and optical spectroscopy to track the BLR emission.These followups confirmed each flux dip as being due to partial-covering obscuration, with XMM-Newton EPIC providing the constraints on lineof-sight column density, N H,pc , and covering fraction, CF.Meanwhile, the optical/UV continuum did not drop appreciably during the campaign, with variations of less than 0.4 mag in any band.We can thus exclude temporary major decreases in accretion or disk temperature.
The obscuration events are consistent with transits by discrete, Compton-thin clouds near the black hole.The first cloud obscuration event lasted between 309 and 539 d (observed frame), and is attributed to a cloud with N H,PC ∼ 1 ×10 22 cm −2 and CF ∼ 60 percent.The second event lasted at least 527 days (observed frame), and was still in-progress as of our most recent observation in February 2023.The responsible cloud had N H,PC ∼ 3 × 10 23 cm −2 and CF ∼ 80 percent.Both clouds are neutral or lowly ionized, with log (ξ, [erg cm s −1 ]) < 1.5-2.5.
The derived limit on cloud 1's radial distance from the Xray corona, assuming Keplerian motion, is >1100 lt-dys, placing it at radii at least commensurate with the far-outer BLR or the dust sublimation zone (which is on the order of 1500 lt-dy).For cloud 2, the limit is >130 lt-dy, which means that a radial location commensurate with the Hβ-emitting region of the BLR (also 130 lt-dy) cannot be excluded.It is (speculatively) possible that cloud 2 may even be a BLR cloud itself.In this case, the second obscuration event may fall into the same category as previous events that lasted up to roughly a day or shorter, and whose inferred cloud properties place them at radii commensurate with the BLR (Risaliti et al. 2009(Risaliti et al. , 2011)).In the case of J0458-5202, perhaps owing to its huge black mass, M BH ∼ 6.2 × 10 8 M ⊙ , the resulting event timescales are on the order of several months to roughly a year.
Optical/UV continuum SED modeling did not yield any evidence for reddening due to dust, and no change in Balmer decrement was observed in the optical spectra.One possibility is that the clouds obscure the X-ray emitting region as well as the optical/UV continuum-emitting region and/or BLR, but they are inherently non-dusty, thus yielding no reddening of the optical/UV continuum or emission-line spectra.However, a second possibility is that the clouds are compact enough to partially obscure only the X-ray-emitting corona while obscuring only a small fraction of the inner accretion disk or BLR, in which case the clouds could be either dusty or dust-free.
Outflowing winds, launched from the disk and radiatively accelerated outward, are likely responsible for several long-duration (months-years) partial-covering X-ray obscuration events observed in nearby Seyfert galaxies.Such winds can contain numerous compact embedded X-ray obscuring clumps with column densities on the order of 10 22−23 cm −2 , similar to those observed in J0458-5202.Such a wind also seems to be a plausible explanation for the two long-duration obscuration events observed in J0458-5202.In this scenario, the wind may be spatially extended (to yield the long event durations).Consequently, the scenario where the wind covers large swaths of the disk or even BLR (in addition to the X-ray corona) but remains inherently dust-free may be preferred in this framework.
Although J0458-5202 has only been monitored for roughly three years, our observations have captured the source to be in an obscured state 78 percent of the time, with three transitions between obscured and unobscured states during that time (noting the obvious bias that our observations were conducted be-cause of the detection of the obscuration events).Nonetheless, it worth speculating on comparisons to NGC 1365, a source with repeated transitions between Compton-thin and -thick-obscured states, as well as with numerous observations of discrete clumpy structures transiting the line of sight to the corona (Ricci & Trakhtenbrot 2023).Thus, J0458-5202 is worth keeping an eye on.
We will continue to monitor this source to determine when the second obscuration event ends, thereby improving estimates of that obscurer's properties.In addition, it is a source worth monitoring in the coming years in the pursuit of potential additional obscuration events.Moreover, additional tracking of the source and identification of new transitions between obscured and unobscured states can solidify and quantify any anticorrelation between column density and intrinsic X-ray luminosity as in NGC 4151.Such observations can provide constraints for modeling the obscurers in the context of outflowing, radiatively driven winds.2, but for two stars nearby in the field of view, TYC 8083-1113-1, and TYC 8083-1608-1.For plotting purposes, the magnitudes for the V, B, W1, M2, and W2 bands have been offset by the magnitude values listed in the top left corner, such that their mean magnitudes match that for U-band.

Fig. 1 .
Fig.1.eRASS1-5 X-ray images of the field around J0458-5202. Gaussian smoothing is applied.The different colors correspond to the different energies of the photons (red channel: 0.2-0.6 keV, green channel: 0.6-2.3keV, blue channel: 2.3-5.0 keV).In the eRASS1 image, we show the scale and the orientation of the image; this is the same for all images.

Fig. 3 .
Fig.3.Overview of multiband light curves.The top panel shows the soft X-ray (0.5-2.0 keV) fluxes across all observations.Green, blue, and red denote eRASS, Swift XRT, and XMM-Newton EPIC, respectively; the brown upper limit denotes the combined NICER observations.The corresponding observation abbreviation is written above each point.The middle panel shows the UVM2 photometric Vega magnitudes from XMM-Newton OM (filled circles) and Swift UVOT (open circles).The bottom two panels show B-and V-band photometry from ground-based measurements (filled circles), Swift UVOT (open circles), and XMM-Newton OM (open star).The optical and UV magnitudes plotted here are not corrected for Galactic absorption.XMM-Newton M2 and B magnitudes were matched to those of Swift UVOT following the cross calibration procedure described in Appendix B. In each panel, there are some data points whose error bars are smaller than the data point symbol.Vertical dotted lines indicate the dates of optical spectroscopic measurements.

Fig. 4 .
Fig.4.Spectral data and fits for XM3.Black, red, green, and blue denote pn0, pn14, MOS1, and MOS2, respectively.Data have been rebinned by a factor of 3 for clarity.In panel a), crosses denote the counts spectra, and the histograms denote the best-fitting M1 ("SX-COM+HXPL+UXCL") model.The corresponding χ residuals are plotted in panel b).In this model, reflected emission is modeled with Ux-Clumpy.In panel c), we plot the χ residuals for the best-fitting model wherein the soft X-ray excess and Compton reflected component are modeled as blurred, ionized reflection via relxill.

Fig. 9 .
Fig. 9. MCMC uncertainty analysis for covering fraction CF as a function of column density N H,PC for the best-fitting M2 (obscured) models for XM1 (magenta) and XM4 (blue).Black crosses represent the bestfitting parameter values; dashed lines indicate the 90 percent confidence limits.N H,PC is in units of cm −2 .

Fig. 10 .
Fig. 10.Unfolded spectra for the best fits to XM3, XM1, and XM4.Data and the best-fitting M1 (unobscured) model for XM3 are shown in black.For XM1 and XM4, we plot the data and the best-fitting M2 (obscured) models in magenta and blue, respectively.

Fig. 11 .
Fig. 11.Spectral data for observation eR1.In the top panel, crosses denote the counts spectra, and the histograms denote the best-fitting M1 ("SXCOM+HXPL+UXCL") model, with data-model residuals plotted in panel c).Panel b) shows the residuals to a simple power-law fit.

Fig. 13 .Fig. 14 .
Fig.13.Overplots of all ten optical spectra of J0458-5202, with the most obvious emission features labeled, covering the region from Hδ through ∼5500 Å. Spectra have been re-scaled by their integrated [O iii] fluxes, as described in the text, but otherwise have not been flux-shifted.

Fig. 15 .Fig. 16 .
Fig.15.Sample spectral deconvolution for the Hβ fitting; shown here is spectrum #2.The data are plotted in blue.The red-dashed line is the power-law continuum.Black denotes broad Balmer components; gray denotes narrow [O iii] and Balmer components; pink denotes the broad and blueshifted [O iii] components; and navy blue, yellow, turquoise, and green denote the P-, F-, I Zw 1-, and S-groups of Fe ii emission.The total model is shown in beige.A host galaxy template was included in the fit, but the best-fit value of its amplitude is zero, as the AGN continuum dominates.

FigFig
Fig. B.1.B-band and M2-band Vega magnitude light curves for J0458-5202 and for nearby stars TYC 8083-1113-1 and TYC 8083-1608-1.XMM-Newton OM points are denoted by filled circles; Swift UVOT points are denoted by open circles.Uncertainties for J0458-5202 are statistical only.Some XMM-Newton OM M2-band data points have error bars smaller than their data point symbol.

Table 2 .
eRASS extraction regions, exposure times, and ML count rates

Table 3 .
Optical/UV photometric observations of J0458-5202Notes.MJD date refers to the midpoint of the observation (time average of all exposures).