A Highly Accreting Low-Mass Black Hole Hidden in the Dust: Suzaku and NuSTAR observations of the NLS1 Mrk 1239

We present torus modelling for the X-ray spectra of a nearby narrow-line Seyfert 1 galaxy Mrk 1239 ($z=0.0199$), based on archival Suzaku, NuSTAR and Swift observations. Our model suggests very soft intrinsic power-law continuum emission of $\Gamma\approx2.57$ in 2019 and $\Gamma\approx2.98$ in 2007. By applying a correction factor to the unabsorbed X-ray luminosity, we find that Mrk 1239 is accreting near or around the Eddington limit. Our best-fit spectral model also suggests a torus with a column density of $\log(N_{\rm H, ave}/$cm$^{-2})=25.0\pm0.2$ and a high covering factor of $0.90$ in Mrk 1239, indicating that this source is most likely to be viewed almost face-on with $i\approx26^{\circ}$. Our line of sight might cross the edge of the torus with $N_{\rm H, los}=2-5\times10^{23}$cm$^{-2}$. The high Eddington ratio and the high line-of-sight column density makes Mrk 1239 one of the AGNs that are close to the limit where wind may form near the edge of the torus due to high radiation pressure.


INTRODUCTION
Mrk 1239 is a narrow-line Seyfert 1 galaxy (NLS1) that shows broad Balmer components of 1000 km s −1 and strong Fe [II] emission in the optical band (Osterbrock & Pogge 1985;Véron-Cetty et al. 2001). Ryan et al. (2007) estimated the mass of the supermassive black hole (BH) in the center of Mrk 1239 to be 7.8 × 10 5 by using the size of the broad line region based on the FWHM(H )-5100 relation (Kaspi et al. 2005) and 1.3 × 10 6 on the basis of the FWHM(H )-H relation (Greene et al. 2006). By studying host bulge properties, Graham & Driver (2007) obtained a similar BH mass of 5 × 10 5 − 7 × 10 6 . In addition to a low BH mass, Mrk 1239 also shows interesting properties across multiple wavelengths: in the radio band, relatively stronger radio emission than typical NLS1s along with evidence of a kiloparsec-scale radio jet was found (Doi et al. 2015), although Mrk 1239 is identified as a radio-quiet source (e.g. 50 mJy at 20 cm, Ulvestad et al. 1995). These features were previously seen only in radio-loud NLS1s (e.g. Antón et al. 2008;Doi et al. 2012). In the near-infrared band, strong blackbody-like emission of ≈ 1200 K peaking at 2.2 µm was found (Rodríguez-Ardila & Mazzalay 2006). ★ E-mail: jcjiang@mail.tsinghua.edu.cn Such strong thermal emission might be related to the heated dusty torus located between the narrow line region and the broad line region of this system with a temperature close to the sublimation limit (Rodríguez-Ardila & Mazzalay 2006;Riffel et al. 2006). In the optical band, polarisation studies by Smith et al. (2004) suggest that the optical band of Mrk 1239 is dominated by polar-scattered emission. Our line of sight (LOS) towards the nuclei should pass through the upper layer of the torus.
In the X-ray band, Mrk 1239 shows particularly soft continuum emission. Rush et al. (1996) analysed the ROSAT spectrum of Mrk 1239 between 0.1-2.4 keV and found the soft X-ray spectrum was consistent with an absorbed power law with Γ = 2.94 ± 0.04. A later study by Grupe et al. (2004) suggested a similar conclusion of Γ ≈ 3 by analysing the XMM-Newton observation of this source in the 0.3-10 keV band. Such soft X-ray continuum emission indicates a high accretion rate in the central disc (e.g. Brightman et al. 2013). An Eddington ratio of Edd ≈ 2 was obtained by modelling a multiwavelength SED of Mrk 1239 and assuming BH = 5 × 10 6 (Grupe et al. 2004). Grupe et al. (2004) discovered an emission line feature at 0.9 keV with an equivalent width of approximately 110 eV in the same XMM-Newton observation, which was interpreted as Ne emission line in their work. A super-solar Ne/O abundance might be required to explain this line feature.
Broad band X-ray spectral analyses suggest that there are two light paths from Mrk 1239: one of the light paths is absorbed direct emission while the other is less absorbed (Grupe et al. 2004), which is consistent with the discovery of wavelength-dependent polarisation degree in the optical emission of Mrk 1239 (Goodrich 1989).
In this work, we present a broad-band spectral model for the Suzaku and NuSTAR observations of Mrk 1239. In Section 2, we introduce our data reduction processes; in Section 3, we introduce a torus-based X-ray spectral model for Mrk 1239; in Section 4, we discuss and conclude our results.

Suzaku
Suzaku observed Mrk 1239 in 2007 for 63 ks (obs ID: 702031010). We produce cleaned event files for all operating XIS detectors (0, 1, 3) using AEPIPELINE v1.1.0 and the latest CALDB as of 2019 November. Source extraction regions are chosen to be circles with radii of 120 arcsec and background are taken from nearby regions with the same shape. We use tasks XISRMFGEN and XISSIMAR-FGEN to create response files for each detector. The spectra and response files of the front-illuminated instruments (XIS 0 and 3) spectra are combined by using the ADDSPEC tool. The combined spectrum is called FI spectrum hereafter. The spectrum of the backilluminated instrument (XIS 1) is called BI spectrum hereafter. The spectra are grouped to have a minimum number of 20 counts per bin. During the spectral modelling, we ignore the energy band below 0.5 keV and the 1.7-2.5 keV band due to calibration uncertainty.

NuSTAR
Mrk 1239 was observed by the NuSTAR satellite in 2019 for ≈ 21 ks (observation ID 60360006002). The NuSTAR data are reduced using the standard pipeline NUPIPELINE v1.9.0 and instrumental responses from NuSTAR caldb V20200510. We extract the source spectra from circular regions with radii of 70 arcsec, and the background spectra from nearby circular regions of 110 arcsec on the same chip. The tool NUPRODUCTS is used for this purpose. The 3-40 keV band is considered for both FPMA and FPMB spectra. The energy band above 40 keV is dominated by background. The FPM spectra are grouped to have a minimum number of 20 counts per bin.

Swift
A Swift observation of Mrk 1239 with a length of 6 ks that was taken simultaneously with our NuSTAR observation is also considered (observation ID: 00081986001). The calibration file version used for XRT data reduction is 20190412. The standard pipeline XRTPRODUCTS v0.4.2 is used for data processing. The source spectrum is extracted from a circular region with a radius of 40 arcsec and the background spectrum is extracted from a circular region with a radius of 100 arcsec nearby. The spectrum is binned to have a minimum count of 20 per bin. We consider the 0.5-6 keV band of the XRT spectrum.
All the spectral analysis is processed by XSPEC v12.11.01 (Arnaud 1996)  In the rest of this section, we first present an overlook of the spectral properties of Mrk 1239, and then we present a torus-based X-ray spectral model for the data.

Iron Line and Compton Hump
Previous studies suggest that the emission of Mrk 1239 consists of two parts, one direct absorbed emission and one less absorbed scattered emission (Goodrich 1989;Grupe et al. 2004). Therefore, we first model the Suzaku spectra of Mrk 1239 in the 2.5-10 keV band with an absorbed power-law model plus a scattered power-law model. The full model is constant1 * tbnew * zmshift * ( vphabs * cabs * cutoffpl1 + constant2 * cutoffpl2 ). The first constant1 is used to account for calibration uncertainty of different instruments. The second constant2 is the scaling factor ( S ) of the scattered power-law component (cutoffpl2) relative to the direct obscured power-law component (cutoffpl1). The normalisation parameters and the photon index (Γ) of these two components are linked. The zmshift model is used to account for the source redshift ( =0.0199). The tbnew model accounts for Galactic absorption. The vphabs and cabs models are used to account for additional line-of-sight column density at the source's redshift.
Such a model can mostly describe the continuum emission in the hard X-ray band very well. A line-of-sight column density of 5 × 10 23 cm −2 is required. See the top panel of Fig. 2 for the corresponding data/model ratio plot. A narrow emission line feature was found peaking at the rest frame 6.4 keV, which is the Fe K emission from a cold emitter. We apply the same model to the 3-10 keV band spectra of the NuSTAR observation. The data/model ratio plots are shown in the lower panel of Fig. 2. NuSTAR spectra also show narrow Fe K emission and a strong Compton hump above 10 keV.
The existence of narrow Fe K emission and Compton hump in the X-ray spectra suggest a cold, neutral emitter in this obscured system. Based on the evidence of an obscured continuum emission, we propose a torus interpretation (Antonucci 1993;Urry & Padovani 1995) for the X-ray reflector in Mrk 1239.

Torus Modelling
In this section, we model the full band spectra of Mrk 1239 using the borus02 model introduced in Baloković et al. (2018).
The borus02 model represents reprocessed X-ray radiation from an approximately toroidal geometry originally proposed by Brightman & Nandra (2011). It self-consistently accounts for the continuum and the fluorescent emission line components. The main parameters relevant for the spectral shape of the reprocessed component in the borus02 model are the torus covering factor, its average column density, inclination and the relative abundance of iron. Despite the simplicity of the assumed geometry, these parameters form a complex and partially degenerate parameter space. The model is similar to, and broader than, the torus model of Brightman & Nandra (2011), which has been shown to be incorrect (Liu & Li 2015;Baloković et al. 2018). In this work we make use of the table model 'borus02_v200623sa.fits', which was calculated specifically to extend the photon index parameter space to accommodate sources with very steep intrinsic continua, such as Mrk 1239.
In Appendix A, we present a spectral model for Mrk 1239 using the mytorus model, an alternative model for torus emission . Data/model ratio plots of XIS spectra using model M0-M4. FI and BI spectra show significant evidence for an emission line feature at 0.9 keV and tentative evidence for a second line at 1.0 keV (marked by black arrows). They correspond to Ne and Ne lines respectively. (Murphy & Yaqoob 2009). The spectra of Mrk 1239 require a very soft power-law continuum, which is beyond the allowed parameter range of the public version of mytorus. So, we only present the analysis using the borus02 model in this section. The full model is constant1 * tbnew * zmshift * ( borus02 + vphabs * cabs * cutoffpl1 + constant2 * cutoffpl2 ) (MO0) in the XSPEC format.
During our spectral fitting process, the photon index (Γ), the high-energy cut-off ( cut ) and the normalisation parameters of the borus02 model are linked to the corresponding parameters of cut-offpl1 and cutoffpl2. Previous optical polarisation studies suggest our LOS might cross the upper layer of the torus in Mrk 1239 (e.g. Smith et al. 2004). We thus tie the half-opening angle and the inclination angle of the torus in our analysis. Similar approach was taken in the X-ray data analysis of other obscured AGNs (e.g. Kamraj et al. 2019). Later in Section 3.4.1, we will discuss the values of these two parameters in detail. The abundance parameters of vphabs model is linked to the iron abundance of borus02 by using the solar abundances calculated from Anders & Grevesse (1989).
MO0 is able to provide a reasonable fit for the FI and BI spectra above 2.5 keV. But the fit is less satisfying in the soft X-ray band with 2 / = 527.23/297. See the first panel of Fig. 3 for the corresponding data/model ratio plot.

Ne and Ne Emissions
In order to fit the soft X-ray emission better, we include a diffuse hot plasma model vmekal (Liedahl et al. 1995) by following the same approach for typical obscured Seyfert AGNs (e.g. Itoh et al. 2008;Hernández-García et al. 2017). In the beginning, we fix the abundances of vmekal at solar values (Anders & Grevesse 1989). The full model is constant1 * tbnew * zmshift * ( borus02 + vmekal + vphabs * cabs * cutoffpl1 + constant2 * cutoffpl2 ) (MO1). MO1 significantly improves our fit of the Suzaku spectra of Mrk 1239 in the soft X-ray band with 2 = 432.89/298. See the second panel of Fig. 3.
However, two emission line features are still seen around 1 keV. The first emission line feature is at the rest frame 0.9 keV, which is similar to the Ne line identified in the XMM-Newton observation (Grupe et al. 2004). A second emission line is found around 1 keV. We first model these emission lines by using two simple Gaussian line models. The full model is constant1 * tbnew * zmshift * ( gauss1 + gauss2 + borus02 + vmekal + vphabs * cabs * cutoffpl1 + constant2 * cutoffpl2 ) (MO2).
By fitting the first line around 0.9 keV with gauss1, the fit is improved by Δ 2 = 104 and 3 more free parameters. The best-fit parameters are shown in Table 1. This line is at 0.884 ± 0.007 keV, and lies at the energy of Ne emission. The equivalent width of the line is 117 +23 −12 eV. The best-fit values are consistent with previous XMM-Newton measurements (Grupe et al. 2004).
By fitting the second line around 1 keV with gauss2, we are able to improve the fit by Δ 2 = 17 and 3 more free parameters. The line is at 1.024 +0.010 −0.012 keV in the rest frame and the equivalent width is 52 +14 −23 eV. We only obtain an upper limit of its line width ( < 0.02 keV). This narrower emission line at 1.03 keV can be interpreted as Ne emission line. A quick F-test based on the 2 improvement suggests an F statistic value of 7.7, which is much less significant than the Ne line. We conclude that we find tentative evidence for Ne emission line in addition to strong evidence of Ne line as in previous analyses.

Hot Diffuse Plasma with Super-solar Ne Abundances
The two narrow emission lines lie well with the energy of Ne and Ne lines. As Grupe et al. (2004) argued, super Ne abundances relative to oxygen might be needed to explain these lines. Instead of modelling the lines with simple gauss models, we propose a physical model with a super-solar Ne abundance for the data. To do so, the Ne abundance parameter ( Ne ) of the vmekal component is allowed to be free during spectral fitting (MO3). Other abundances are fixed at solar (Anders & Grevesse 1989).
MO3 is able to provide a better fit than MO1 with Δ 2 = 84.3 and one more parameter. Corresponding data/model ratio plots are shown in the fourth panel of Fig. 3. Best-fit parameters are shown in Table 1. MO3 offers a reasonable fit to the emission line at 0.88 keV with a super-solar Ne abundance, although some residuals still remain around 0.86 keV. The second emission line at 1.02 keV cannot be modelled by MO3. Therefore, we add a second vmekal component (MO4). The Ne abundances of these two vmekal components are linked. MO4 provides a very good fit of the data with 2 / = 317.42/292. The best-fit parameters are shown in Table 1, and corresponding data/model ratio plots are shown in the last panel of Fig. 3.
In comparison with MO2 where gauss models are used, the fit using MO4 has a slightly higher 2 (Δ 2 = 5.1) but 3 fewer parameters. In the end, we decide to choose MO4 as our best-fit model instead of MO2. Because MO4 provides a more physical interpretation to the narrow emission lines.
In conclusion, two diffuse plasma components are required to fit the soft X-ray spectrum of Mrk 1239: one has a temperature of ≈ 0.22 keV and the other one has a temperature of ≈ 0.64 keV. A super-solar Ne abundance of Ne = 4.9 +1.2 −1.3 is needed. Buhariwalla et al. (2020) suggested that these hot plasma components may be associated with the star-burst region in the host galaxy of Mrk 1239. This component has been often found in the soft X-ray emission of many obscured sources (e.g. Franceschini et al. 2003). When the AGN emission is high and not heavily obscured, this component still exists, but overwhelmed by the central AGN emission.

Multi-Epoch Spectral Analysis
In this section, we apply MO4 to all the spectra of Mrk 1239 simultaneously. As shown in Fig. 1, the soft X-ray band of Mrk 1239, which is dominated by distant diffuse plasma emission, shows a consistent flux level in two epochs. The signal-to-noise (S/N) of Swift XRT data does not allow us to investigate the soft X-ray emission during the second epoch in detail. Therefore, we link the parameters of diffuse plasma emission components for two epochs. Other parameters, such as the torus inclination angle, the average column density and the iron abundance of the torus, are expected to be consistent on observable timescales. They are thus linked during our spectral fitting too.
MO4 provides a good fit to the data of both epochs. The best-fit parameters are shown in Table 2, and the best-fit data/model ratio plots are shown in Fig. 4.

The Inclination Angle and the Opening Angle of the Torus
The best-fit average column density of the torus log( H,ave is approximately 10 25 cm −2 . The line-of-sight column density H,los is 2-5×10 23 cm −2 , which is 2 orders of magnitudes lower than H,ave . In MO4, the inclination angle ( ) and the half-opening angle 1 of the torus (Cfact) are linked in our analysis as in Kamraj et al. (2019) our LOS might cross the edge of the torus where the column density is only 2-5% of the average column density of the whole torus.
In this part of the section, we investigate how valid the Cfact≈ cos( ) relation is by allowing these two parameters to be free in our analysis. All the spectra are used for this test. Fig. 5 shows the 2 distribution on the Cfact and parameter plane. The XSPEC tool STEPPAR is used to do so. Only lower limit has been obtained: a 3 uncertainty range of Cfact is >0.815 and that of cos( ) is >0.6. They suggest a half-opening angle smaller than 35 • and an inclination angle smaller than 53 • . The model is more sensitive to Cfact/the half-opening angle of the torus, because this parameter directly modifies the continuum emission. The constraint of is relatively weaker as the model is less sensitive to this parameter. We show the Cfact=cos( ) reference line in Fig. 5. Our assumption of Cfact=cos( ) lines well within the 1 uncertainty range, indicating that these two parameters approach a similar value in our analysis-our LOS might cross the edge of a torus which has a small half-opening angle. Similar conclusions were found in the optical polarisation studies: Mrk 1239 turns out to be a polar-scattered NLS1 in which our LOS passes the upper layer of its torus (Smith et al. 2004).
By linking these two parameters, we obtain a better constraint of both parameters Cfact=cos( ) = 0.90 +0.04 −0.03 (90% confidence range) as shown in Table 2. This best-fit value corresponds to a small half-opening and inclination angle of approximately 26 • . A covering factor that is as high as ≈ 90% in Mrk 1239 has been also seen in NuSTAR observations of other obscured Sy1 galaxies but with lower accretion rates than in Mrk 1239 (e.g. Baloković et al. 2018;Kamraj et al. 2019).

Ultra-soft X-ray Continuum Emission
The photon index of the power-law continuum emission is 2.98 ± 0.02 for the Suzaku epoch and 2.57 +0.03 −0.02 for the NuSTAR and Swift epoch. Despite of the obscured nature, the very soft X-ray continuum emission from the hot corona of Mrk 1239 is very similar to that of other NLS1s (e.g. Gallo 2018).
In particular, X-ray studies of a sample of unobscured, extreme ultra-soft NLS1s suggest similar continuum emission with Γ > 2.5 (Jiang et al. 2020). Detailed modelling of their multi-wavelength SEDs suggest an Eddington ratio that is around or a few times higher than the Eddington limit (e.g. Jin et al. 2009;Jiang et al. 2020). Strong soft excess emission is shown in their data, and can be interpreted as part of reflection from a highly ionised inner disc region as well broad Fe K emissions (Jiang et al. 2020). Unfortunately, due to the obscuration along our LOS towards Mrk 1239, we are unable to constrain the soft excess emission from the center of the AGN in Mrk 1239.
In Section B, instead of investigating in the soft X-ray band, we discuss possible contribution of a disc reflection component in the hard X-ray band, particularly in the observed back-scattering Compton hump.

The Accretion Rate of Mrk 1239
Our best-fit model suggests a very soft intrinsic power-law emission of Γ = 2.6 − 3. The softness of the continuum emission suggests a high mass accretion rate in the disc (e.g. Brightman et al. 2013, and references therein). We estimate the Eddington ratio of Mrk 1239 by using its 2-10 keV luminosity. The absorption-corrected flux of Mrk 1239 is 1.0 × 10 −11 erg cm −2 s −1 in 2007 and 8.0 × 10 −12 erg cm −2 s −1 in 2019 calculated by our best-fit model. Assuming a BH mass of 1 × 10 6 , they correspond to X = 7.1 − 8.8 × 10 42 erg s −1 = 0.06 − 0.07 Edd . When considering a typical correction factor of = 10 − 20 for this luminosity (Vasudevan & Fabian 2007), we estimate Mrk 1239 has a bolometric luminosity of 0.6-1.4 of the Eddington limit assuming a BH mass of 10 6 . Previous independent measurements of the mass of the SMBH in Mrk 1239 all agree with a relatively low value of ≈ 1 × 10 6 (see Section 1). We consider the largest measurement uncertainty in the literature: Graham & Driver (2007) estimated BH = 5 × 10 5 − 7 × 10 6 by using the host bulge properties in Mrk 1239. After taking into account the uncertainty of BH mass measurements, our estimation of Edd is 0.1-2.8 for Mrk 1239. Similar conclusions were found in Grupe et al. (2004) where a multi-wavelength SED was used for the calculation of the Eddington ratio: Edd ≈ 2 assuming BH = 5 × 10 6 . Buhariwalla et al. (2020) found an Eddington ratio of Edd = 1 − 1.5. Yao et al. (2018) applied a correction factor to the 5100 of Mrk 1239 and obtained Edd = 1.12. The uncertainties of their estimations were not mentioned in their work. But they are all consistent with our measurement.
In conclusion, Mrk 1239 is one of the most extreme AGNs that are accreting near or around the Eddington limit.

Stability of the Dusty Torus in Mrk 1239
The high accretion rate in the central engine of Mrk 1239 may explain the high temperature blackbody emission in the near-infrared band (Rodríguez-Ardila & Mazzalay 2006). The dusty torus is heated by the radiation from the inner accretion region close to the sublimation limit. In this section, we discuss the stability of such a heavy torus and the radiation pressure from the luminous nucleus onto the torus.
Laor & Draine (1993); Scoville & Norman (1995); Murray et al. (2005) calculate the effective Eddington ratio for cool dusty gas, which is found to be much lower than the Eddington ratio for ionised dust-free gas. As an example, the black solid line in Fig. 6 shows the effective Eddington limit for different column densities. The line is adopted from Fabian et al. (2009). The radiation pressure will play an important role for systems that correspond to the right of the solid line. The obscuring dusty materials will be blown away due to high radiation pressure when luminosity exceeds the effective Eddington limit. The dashed line shows the typical column density of galactic dust lane that may make contribution to the X-ray spectrum.
Large X-ray surveys, e.g. the Swift BAT catalog, indeed find Obscuring dusty materials will be blown away due to radiation pressure when luminosity exceeds the effective Eddington limit (white forbidden region in the figure). Most Swift BAT AGN lie within the green shaded region (Fabian et al. 2009;Ricci et al. 2017b). Some exceptional dusty quasars at higher redshifts, e.g. > 0.2, are found to be located in the forbidden region (Lansbury et al. 2020). The black circle shows the column density of the torus in Mrk 1239. The red and blue diamonds shows the line-of-sight column density respectively in 2007 and 2019. The luminosity of the nuclei in Mrk 1239 may exceed the effective Eddington limit in the upper layer of the torus. Wind might be forming on the edge due to high radiation pressure. The left and right grey dashed lines show the values Edd given in Yao et al. (2018) and Grupe et al. (2004) respectively. The shaded region shows the range of Edd estimated by Buhariwalla et al. (2020).
that most of the known AGNs tend to avoid the forbidden region of this diagram (Fabian et al. 2009;Ricci et al. 2017b) and lie within the green shaded region in Fig. 6. Some exceptional cases are, however, found in very luminous dusty quasars at higher redshifts (e.g. > 0.2, Lansbury et al. 2020). We show Mrk 1239 in the H − Edd diagram in Fig. 6, and it is located on the edge of the 'long-lived cloud' region. The lineof-sight column density of Mrk 1239 is estimated to be 2 − 5 × 10 23 cm −2 . The luminosity of Mrk 1239 can exceed the effective Eddington limit at this low value of column density. Therefore, radiation pressure-driven wind might be forming near the edge of the torus along our LOS.
Last but not least, it is important to note that the H emission of Mrk 1239 shows an asymmetric profile with a minimum polarisation degree in the blue wing and a maximum degree in the red wing, which might be related to a radial outflow (Smith et al. 2004).
Future high-S/N, TES-based (Transition Edge Sensor) observations, such as from XRISM with a spectral resolution of 2.5 eV, might enable us to look for any evidence of wind absorption features in the X-ray spectra, such as blueshifted Fe K edges.

The Variability of the Line-of-Sight Column Density of Mrk 1239
In Section 3.4.1, we estimate the half-opening angle and the inclination angle of the torus in Mrk 1239 by allowing them to vary during spectral fitting. The results suggest an inclination angle smaller than 53 • and a half-opening angle smaller than 35 • . When we link these two parameters during spectral fitting, the best-fit value corresponds to a small half-opening angle of approximately 26 • .
In this scenario, our LOS crosses the upper edge of the torus in Mrk 1239. Therefore, measured H,los corresponds to the column density near the edge of the torus, which is only a few per cent of the average column density of the torus: H,los = 5.2 +0.7 −0.6 × 10 23 cm −2 in 2007 and 2.0 +0.4 −0.3 ×10 23 cm −2 in 2019. A similar H,los was found in the XMM-Newton observation of Mrk 1239 in 2001 ( H,los ≈ 3× 10 23 cm −2 , Grupe et al. 2004). These observations were separated by 18 years and suggest that some variability of H,los with a small amplitude may exist in Mrk 1239 during this period.
As shown in the previous section, radiation pressure-driven dusty wind may be forming near the edge of the torus when H < 10 24 cm −2 . If true, our LOS intercepts with the wind and some variability of H,los is expected. In comparison, the average column density of the torus is as high as 10 25 cm −2 . The radiation pressure from the central region of the AGN in Mrk 1239 is not high enough to 'blow' wind from the equatorial plane of torus.
Alternatively, a clumpy torus model may also explain the variability in H,los . In the unification paradigm of AGNs (Antonucci 1993), Seyfert 1 AGNs (Sy1s) and Seyfert 2 AGNs (Sy2s) are distinguished by their inclination angles, i.e. whether our LOS crosses the torus. This standard picture assumes a homogeneous dusty torus. However, observations suggest that the LOS obscuration is determined not only by inclination angle but also by the probability of absorption clouds intercepting our LOS (Nenkova et al. 2008). Evidence for the clumpiness of the torus includes a large and rapid variability of the absorber observed in several sources (e.g. Risaliti et al. 2002;Bianchi et al. 2012;Laha et al. 2020). Markowitz et al. (2014) estimated the probability of an absorption event regardless of constant absorption due to non-clumpy material to be 0.003-0.16 for Sy1s. Dramatic changes of H,los up to a few orders of magnitude have indeed been seen in Sy1s on observable timescales (e.g. Simm et al. 2018).
Unfortunately, archival observations are not sufficient enough to distinguish these two models for the small variability of H,los in Mrk 1239. As concluded in the previous section, a search for absorption features, e.g. blueshifted absorption edge due to dusty wind, is required when TES-based observations are available.

The opening angle of the torus in Mrk 1239
As introduced above, Sy1s and Sy2s are intrinsically the same type of object but viewed from different inclination angles in the unification paradigm of AGNs (Antonucci 1993), which is supported by the detection of polarised broad lines in Sy2s (e.g. Antonucci & Miller 1985;Miller & Goodrich 1990;Young et al. 1996). Meanwhile, it is believed that two scattering regions, the equatorial plane (e.g. Goodrich & Miller 1994;Cohen et al. 1999;Cohen & Martel 2002) and the ionisation cone on the torus axis (e.g. Antonucci 1983;Smith et al. 2002), both contribute to the polarised emission in AGNs. Smith et al. (2004) argued that the inclination angle plays an important role in the interpretation of the detected polarisation in AGNs. Polar-scattered Sy1s represent a bridge between Sy1s and Sy2s. Their optical polarisation position angle is perpendicular to the projected radio source axis as in Sy2s (Smith et al. 2004), whilst the majorty of Sy1s show optical polarisation properties that are not consistent with polar scattering (e.g. Antonucci 1983;Smith et al. 2002). This is because our LOS may intercept with the edge of the torus in polar-scattered Sy1s, e.g. Mrk 1239 (Smith et al. 2004), where the equatorial emission is obscured by the dusty torus and makes little contribution to the detected optical polarisation.
Our torus modelling of the X-ray spectra of Mrk 1239 provides another supporting evidence for this model. Our results suggest that the data are consistent with the case that our LOS intercepts with the upper edge of the torus.
Assuming a toroid-shaped inhomogenous torus, we obtain an upper limit of 35 • for the half-opening angle parameter at 3 uncertainty level (see Section 3.4.1). Such a low half-opening angle corresponds to a very high covering factor for Mrk 1239, e.g. Cfact>0.81.
A similar covering factor was found in Mrk 231, another polarscattered Sy1. For instance, Piconcelli et al. (2013) found that Mrk 231 shows two partial-covering absorbers with H ≈ 10 22 and 10 24 cm −2 . A very high covering factor of more than 0.9 and 0.8 are respectively found for these two absorbers (Piconcelli et al. 2013). Interestingly, Smith et al. (2004) found that the optical polarisation spectra of Mrk 1239 and Mrk 231 are very similar too: their continuum polarisation rises at shorter wavelengths and peaks at the red wing of their broad lines.

Comparison with other polar-scattered Sy1s
Polar-scattered Sy1s, including Mrk 1239, often show significant absorption features in the X-ray band, which is in agreement with the picture that our LOS passes close to their torus opening angles (e.g. Jiménez-Bailón et al. 2008;Piconcelli et al. 2013;Laha et al. 2011;Newman et al. 2021). For instance, the polar-scattered Sy1 Mrk 704 shows not only a partial covering neutral absorber with H ≈ 10 23 cm −2 and Cfact=0.22 but also two layers of warm absorbers that are associated with the broad line region (Laha et al. 2011).
Meanwhile, polar-scattered Sy1s show a variety of X-ray spectral features. For example, the neutral absorber in Mrk 704 has a much lower covering factor than those in Mrk 1239 and Mrk 231. This indicates that the torus opening angles in polar-scattered Sy1s may not be uniform. Ricci et al. (2017a) applied a torus model to a sample of Swift BAT AGNs. For the 12 objects with constrained torus half-opening angles, they found a median value of 58 ± 3 • . Mrk 1239 and Mrk 704 are respectively located in the lower and higher ends of the global distribution of torus opening angles.
Besides, as demonstrated in this work and Piconcelli et al. (2013), Mrk 231 and Mrk 1239 show no significant evidence of reprocessing emission from the innermost region, e.g. disc reflection (see Section B). In comparison, other polar-scattered Sy1s show evidence of either broad Fe K emission (e.g. Fairall 51, Svoboda et al. 2015) or soft excess emission from the innermost accretion region (e.g. NGC 3227 and Mrk 704, Laha et al. 2011;Newman et al. 2021).

CONCLUSIONS
We present a torus model for the X-ray spectra of the NLS1 Mrk 1239 based on archival Suzaku, NuSTAR and Swift observations. The main results are as follows.
• The primary X-ray continuum of Mrk 1239 is described by a power law with slope Γ = 2.6 − 3.0. Such a soft continuum suggests that Mrk 1239 is one of the most extreme AGNs that are accreting near or around the Eddington limit. By applying a correction factor to its X-ray luminosity, we obtain Edd = 0.1 − 2.8 after taking into account the uncertainty of the BH mass measurements.
• At such a high accretion rate, the radiation pressure from the central region of the AGN may drive wind near the edge of the torus where column density is around 10 23 cm −2 . Future high-S/N, TES-based X-ray observations may reveal more spectral details of Mrk 1239, e.g. blueshifted absorption edge features.
• The X-ray data of Mrk 1239 are consistent with the optical polarisation model for polar-scattered Sy1s where our LOS intercepts with the upper edge of the torus. The LOS column density of Mrk 1239 is only a few per cent of the average column density of the torus in this source. The half-opening angle of the torus is estimated to be around 26 • , corresponding to a very high covering factor of 90%. Such a small opening angle makes Mrk 1239 near the lower end of the global torus opening angle distribution of AGNs.

ACKNOWLEDGEMENTS
This paper was written during the worldwide COVID-19 pandemic in 2020-2021. We acknowledge the hard work of all the health care workers around the world. We would not be able to finish this paper without their protection. J.J. acknowledges support from the Tsinghua Shui'Mu Scholar Program and the Tsinghua Astrophysics Outstanding Fellowship. M.B. acknowledges support from the YCAA Prize Postdoctoral Fellowship. This work made use of data from the NuSTAR mission, a project led by the California Institute of Technology, managed by the Jet Propulsion Laboratory, and funded by NASA, and data obtained from the Suzaku satellite, a collaborative mission between the space agencies of Japan (JAXA) and the USA (NASA). This research has made use of the NuSTAR Data Analysis Software (NuSTARDAS) jointly developed by the ASI Science Data Center and the California Institute of Technology.

DATA AVAILABILITY
All the data can be downloaded from the HEASARC website at https://heasarc.gsfc.nasa.gov.   Table A1. Best-fit model parameters obtained by fitting the reprocessed emission using mytorus. The half-opening angle of the torus is fixed at 60 • in this model. The high energy cutoff of power-law continuum emission is fixed at 400 keV. The scaling factor for the fluorescent line emission l is coupled with that for the scattered continuum S . S is the scattering fraction due to optically-thin matters outside the LOS.
fixed. The inclination angle of the torus is not constrained by the mytorus model. A 90% confidence upper limit of 70 • is obtained. In comparison, the upper limit of the inclination angle given by the borus02 model is 52 • when both and Cfact parameters are free to vary (see Section 3.4.1 for more information).

APPENDIX B: CONTRIBUTION OF DISC REFLECTION IN THE X-RAY BAND
Despite the obscured X-ray emission, the power-law component from the hot corona in the AGN of Mrk 1239 is like the one in a typical high-Edd NLS1 (Gallo 2018): a very soft power-law index of Γ = 2.6 − 3.0 is suggested by the data. Reflection off the innermost accretion disc region has been commonly seen in the Xray observations of other unobscured NLS1s. Prominent features of disc reflection spectra include broad Fe K emission and Compton Mrk 1239 Spectral Analysis 11 hump in the hard X-ray band. In particular, Seyfert 1 AGNs often show excess emission in the soft X-ray band, which could be part of disc reflection as well (e.g. Jiang et al. 2019). A strong supporting evidence of soft excess emission being reflection is the increasing number of discoveries of reverberation lags in the soft X-ray band (e.g. Kara et al. 2016). They are similar to the iron emission and Compton hump reverberation lags (e.g. Kara et al. 2015Kara et al. , 2016. The soft X-ray emission from the centre of the AGN in Mrk 1239 is unfortunately obscured by line-of-sight column density. The iron band of Mrk 1239 is dominated by fluorescence emissions of the torus. But some disc reflection may still exist (Buhariwalla et al. 2020) and make contribution in the hard X-ray band, the observed Compton hump in particular.
In this section, we estimate the upper limit of the contribution from a disc reflection component in the X-ray band. We add the relxill model (García et al. 2013;Dauser et al. 2016) to MO4. The relxill calculates reflection spectra of a relativistic thin disc. We use a power-law emissivity profile for the disc parametrised by an index . The inner disc inclination angle is linked to the inclination angle of the torus, although they might be different in reality. The Ecut and Γ parameters of relxill are linked to the corresponding parameters in borus02. Other free parameters include the normalisation parameter, the disc ionisation and the spin of the BH. The inner radius of the disc is assumed to be at the innermost stable circular orbit. By doing so, we only obtain an upper limit of the relxill component for both the Suzaku and the NuSTAR epochs. See Fig. B1 for the constraints of the normalisation parameters of relxill for two epochs.
The 3 upper limit of the normalisation parameters of relxill is approximately 10 −5 for both epochs. At the 3 upper limit, the relxill component produces an observed flux of approximately 3 × 10 −14 erg cm −2 s −1 in the 3-10 keV band for the Suzaku epoch and an observed flux of around 6.8 × 10 −13 erg cm −2 s −1 in the 3-78 keV band for the NuSTAR epoch. They respectively take up 4% and 8% of the total observed X-ray flux in corresponding total energy bands 3 . We conclude that the narrow Fe K emission and the Compton hump shown in the X-ray observations are likely to be dominated by reprocessed emission from the torus based on the obscured X-ray nature and observations in other wavelengths (Smith et al. 2004;Rodríguez-Ardila & Mazzalay 2006). The inner disc reflection makes little contribution to the data with a 3 upper limit of 8% during the NuSTAR epoch.
The lack of evidence for broad Fe K emission is interesting. Following are possible explanations: 1) the disc might be very ionised. As shown by Jiang et al. (2020), the discs in some very extreme, unobscured NLS1s with Γ > 2.5 show a high ionisation state of log( ) > 3. In comparison, typical Seyfert AGNs have a lower-ionisation disc of log( ) ≈ 1 − 2 (Walton et al. 2013). At such a high ionisation state, the Fe K emission is weak (e.g. García et al. 2013). 2) The high-Edd NLS1s in Jiang et al. (2020) and Mrk 1239 are all very soft X-ray emitters. The S/N of the current X-ray CCD data, e.g. from XMM-Newton or Suzaku, is not high enough to detect their broad Fe K emission. Future high-S/N observations, e.g. from Athena, might be able to do so according to the simulations in Jiang et al. (2020). 3) The innermost accretion region may not hold a thin disc geometry at a high Eddington ratio as in Mrk 1239. Due to the thickness of an optically thick disc, emissions from the innermost region might be obscured by the puffed-up disc (e.g. Ohsuga