Transient obscuration event captured in NGC 3227
II. Warm absorbers and obscuration events in archival XMM-Newton and NuSTAR observations

Active galactic nuclei (AGN) accrete matter onto a central supermassive black hole (SMBH) to produce intense broadband radiation, which can ionize and drive away the surrounding matter in forms of outflows. Many observational proofs have implied that outflows might play an important role in affecting the star formation and evolution of their host galaxies (see the review of King & Pounds, 2015). Ionized outflows can be detected via absorption features along the line of sight in the ultraviolet (UV) and X-rays, which usually have different types (Laha et al., 2021, and references therein) such as broad absorption lines (BALs; Weymann et al., 1981), warm absorbers (WAs; Halpern, 1984; Crenshaw et al., 2003), and ultrafast outflows (UFOs; Tombesi et al., 2010). UFOs might have an origin close to the central engine ($\sim 0.0003-0.03\ \rm{pc}$; Tombesi et al., 2012) with very high velocities ($\sim 0.03-0.3c$; Tombesi et al., 2010, 2012). BALs usually reside outside the broad line region (BLR) with high outflow velocities reaching $\sim 30000\ \rm{km}\ \rm{s}^{-1}$ (Trump et al., 2006; Gibson et al., 2009). Compared with UFOs and BALs, WAs have lower outflow velocities from about one hundred to a few thousands of ${\rm{km}}\ {\rm{s}}^{-1}$ (Kaastra et al., 2000; Ebrero et al., 2013), and they might originate in the accretion disk (e.g., Elvis, 2000; Krongold et al., 2007), BLR (Reynolds & Fabian, 1995), or dusty torus (e.g., Krolik & Kriss, 2001; Blustin et al., 2005). Although different types of outflows have overlaps in distance scales and outflow parameters, the direct connection between these outflows still remains unclear. In this work, we mainly focus on the properties of the WA outflows.

According to Tarter et al. (1969), the ionization parameter can be defined by

where $L_{\rm{ion}}$ is the ionizing luminosity over 1–1000 Ryd, $n_{\rm{H}}$ is the hydrogen number density of the absorbing gas, and $r$ is the radial distance of the absorbing gas to the central engine. WAs might be driven by radiation pressure (e.g., Proga & Kallman, 2004), magnetic forces (e.g., Blandford & Payne, 1982; Konigl & Kartje, 1994; Fukumura et al., 2010), or thermal pressure (e.g., Begelman et al., 1983; Krolik & Kriss, 1995; Mizumoto et al., 2019), and show a wide range of ionization parameter ($10^{-1}\leq\xi\leq 10^{3}\ {\rm{erg}}\ {\rm{cm}}\ {\rm{s}}^{-1}$) and hydrogen column density ($10^{20}\leq N_{\rm{H}}\leq 10^{23}\ {\rm{cm}}^{-2}$) (Laha et al., 2014). Investigating properties of WAs can help us to understand the formation of AGN outflows and their feedback efficiency to the host galaxy. These WAs have been found in about 50% of nearby AGN (e.g., Reynolds, 1997; Kaastra et al., 2000; Porquet et al., 2004; Tombesi et al., 2013; Laha et al., 2014), and the properties of WAs show differences among different AGN, such as the different ionization states, column densities, and outflow velocities (Tombesi et al., 2013; Laha et al., 2014).

Moreover, the X-ray spectra of some AGN present dramatic hardening accompanied by flux-drops on short timescales, which might be due to the X-ray transient obscuration events (Markowitz et al., 2014). Transient obscuration events can also cause absorption features in the soft X-ray and UV bands, which usually appear and disappear on shorter timescales compared with outflows. These obscuration events might be explained by discrete optically thick clouds or gas clumps crossing the line of sight, which are referred to as obscurers. These shielding gas clumps or obscurers might ensure that the radiatively driven disk winds in broad absorption line quasars are not over-ionized by UV/X-ray ionizing radiation and are accelerated further (Murray et al., 1995; Proga et al., 2000; Kaastra et al., 2014). The obscuration events may be triggered by the collapse of the BLR (Kriss et al., 2019a, b; Devereux, 2021). When the continuum radiation decreases, the BLR clouds will collapse towards the accretion disk; when the continuum brightens again, these collapsed clouds might be blown away as obscurers (Kriss et al., 2019b). X-ray obscuration events also have been found in many AGN, such as NGC 5548 (Kaastra et al., 2014), NGC 3783 (Mehdipour et al., 2017; Kaastra et al., 2018; De Marco et al., 2020), NGC 985 (Ebrero et al., 2016a), and NGC 1365 (Risaliti et al., 2007; Walton et al., 2014; Rivers et al., 2015). These obscurers may be located within the BLR (Lamer et al., 2003; Risaliti et al., 2007; Lohfink et al., 2012; Longinotti et al., 2013; De Marco et al., 2020; Kara et al., 2021), or close to the outer BLR (e.g., Kaastra et al., 2014; Beuchert et al., 2015; Mehdipour et al., 2017), or near the inner torus (e.g., Beuchert et al., 2017).

Until now the relation between the WA outflows and the nuclear obscuration activity is not yet well understood. Whether the WAs and obscurers have the same origin or how shielding by the obscuration affects the WAs and their appearance is not well known. Studying WAs in targets that have transient obscuration enables us to probe these questions. Transient obscuration events have been studied simultaneously in UV and X-rays in only a few AGN that have WAs outflows, such as NGC 5548 (Kaastra et al., 2014), and NGC 3783 (Mehdipour et al., 2017), and also Mrk 335 (Longinotti et al., 2013; Parker et al., 2019). Studying NGC 3227 (a Seyfert 1.5 galaxy at the redshift of 0.003859¹¹1The redshift of NGC 3227 is obtained from the NASA/IPAC Extragalactic Database (NED). The NED is funded by the National Aeronautics and Space Administration and operated by the California Institute of Technology.) is a rare opportunity towards a more general characterization. NGC 3227 was one of eight suitable targets selected for the Neil Gehrels Swift Observatory monitoring/triggering programme (Mehdipour et al., 2017), which matches the following selection criteria: WAs in both UV and X-rays, suitably variable, bright in UV and X-rays, good archival spectra for comparing with the obscured spectra. Using our target of opportunity (ToO) monitoring programme of the Neil Gehrels Swift Observatory, we captured another X-ray obscuration event in NGC 3227 in 2019 (Mehdipour et al., 2021, hereafter Paper I), which was observed simultaneously with XMM-Newton, NuSTAR, and Hubble Space Telescope/Cosmic Origins Spectrograph (HST/COS) to get a deeper multi-wavelength understanding of the transient obscuration phenomenon in AGN. The studies of WA and the obscurer are interlinked, so without having a proper model for the WA, the new obscurer cannot be accurately studied. In this work (the second paper of our series), we aim to study a comprehensive model for the WA, and then use this WA model to investigate the obscuration events appearing in NGC 3227.

It should be noted that photoionization modelling strongly depends on the ionizing spectral-energy-distribution (SED). Therefore, to properly derive the ionization structure of the WA, having an accurate broadband SED model is important. A few papers have reported studies of the WAs in NGC 3227 (Komossa & Fink, 1997; Beuchert et al., 2015; Turner et al., 2018; Newman et al., 2021) and its nuclear obscurations activities (Lamer et al., 2003; Markowitz et al., 2014; Beuchert et al., 2015; Turner et al., 2018). However, the contribution of the SED components that dominate in the UV/optical band lacks consideration, which might affect the fitting results of the WAs and obscurers (see Paper I). The main effect of using different SEDs is that the derived ionization parameter $\xi$ would be different. The total hydrogen column density $N_{\rm{H}}$ of the WA (i.e. sum of the individual components) would be similar, but how $N_{\rm{H}}$ is distributed over different ionization components depends on the SED. We refer the readers to Mehdipour et al. (2016) where the effect of using different SEDs, and codes, is shown. With these considerations, we firstly built a broadband SED model from the near infrared (NIR) to hard X-rays for NGC 3227 in our Paper I. In this paper, we use this broadband SED model and a robust photoionization code (pion model; Mehdipour et al., 2016) in the SPEX package (Kaastra et al., 1996) v3.05.00 (Kaastra et al., 2020) to analyze the archival XMM-Newton and NuSTAR data taken in 2006 (Markowitz et al., 2009) and 2016 (Turner et al., 2018). SPEX is currently the only code that enables the SED and the ionization balance to be fitted simultaneously, while all other codes have to pre-calculate the ionization balance on a given SED. The pion model is a self-consistent model that can simultaneously calculate the thermal/ionization balance and the plasma spectrum in photoionization equilibrium. In this work, we focus on properties of the WAs and obscuration events of NGC 3227 with the archival 2006 and 2016 data. The detailed analysis of the 2019 obscuration events will be presented in Paper III by Mao et al. (in prep). The discussion about how the obscurer changes over the course of the 2019 observations with the XMM-Newton/EPIC-pn data will be presented in Paper IV by Grafton-Waters et al. (in prep).

This paper is organized as follows. In Sect. 2, we present the archival data that are used in this work and the data reduction process. In Sect. 3, we introduce the spectral analysis based on the broadband SED model. In Sect. 4, we present and discuss the results about WAs and obscurer components. In Sect. 5, we give a summary of our conclusions. In this work, Cash statistic (Kaastra, 2017, hereafter C-stat) will be used to estimate the goodness of fit and statistical errors will be given at 1$\sigma$ (68%) confidence level. We adopt the following flat $\Lambda$CDM cosmological parameters: $\textit{H}_{\textrm{0}}$=70 km s${}^{\textrm{-1}}$ Mpc${}^{\textrm{-1}}$, $\textrm{$\Omega$}_{\textit{m}}$=0.30, and $\textrm{$\Omega$}_{\textrm{$\Lambda$}}$=0.70.

In Table 1, we list the archival data that are used in this work. These data include six XMM-Newton observations (Markowitz et al., 2009; Turner et al., 2018) and five NuSTAR observations (Turner et al., 2018). We do not use the archival XMM-Newton observations taken in 2000 and on 2016 November 29, because the spectrum of the 2000 observation has a lower signal to noise ratio (S/N) owing to its short exposure time, and the observation on 2016 November 29 shows a relatively unstable softness ratio curve (see Fig. 1 of Turner et al., 2018), which may bias the estimation of WAs parameters.

For the 2016 archival data, we will consider each set of XMM-Newton and NuSTAR observations taken on the same date as a single dataset (see Obs2 to Obs6 in Table 1), where XMM-Newton (OM UVW1 filter at $\sim$ 2910 $\AA$, RGS data in the 6–37 $\rm{\AA}$ wavelength range, and EPIC-pn data in the 2–10 keV energy band) and NuSTAR (combined FPMA and FPMB data in the 5–78 keV energy range) data are used simultaneously for the spectral analysis. For Obs1 that was taken on 2006 December 03, only XMM-Newton observation is available (see Table 1), and we used Obs5 to verify that the best-fit parameters of the WAs do not significantly change without NuSTAR observation. In Figs. 1 and 2, we show the XMM-Newton/EPIC-pn light curves, softness ratio curves (the ratio of count rates between 0.3–2 and 2–10 keV bands), and the correlation between the softness ratio and 0.3–10 keV count rate for Obs1 to Obs6. According to these results, we make a preliminary analysis for the state of each observation.

• •

  Obs1: This observation was considered to be in an unobscured state by Markowitz et al. (2009). However, compared with the average softness ratio of the six observations (see the dashed red line in the bottom panel of Fig. 1), a significant spectral hardening (softness ratio below the average value) occurred in the beginning of this observation (see the bottom panel of the first column in Fig. 1), which might indicate the existence of an obscuration event. After this spectral hardening period, the softness ratio of Obs1 returns to the average value, which might indicate that this obscuration event has disappeared. These phenomena mean that these two periods of Obs1 should be analyzed separately. According to the difference between the softness ratio of Obs1 and the average softness ratio of the six observations, Obs1 is subdivided into the following two slices: S1a and S1b (see the first column in Fig. 1). S1b is consistent with being in an unobscured state as it follows the correlation of Obs2 to Obs5 (see the right panel of Fig. 2). However, S1a has a harder spectrum that is similar to S6a and S6b (see the right panel of Fig. 2), which might be in an obscured state, but it was not reported in the previous works.

• •

  Obs2 to Obs5: According to Turner et al. (2018), Obs2 to Obs5 are in unobscured states, which show a linear correlation between the softness ratio (0.3–1$/$1–10 keV) and 0.3–10 keV count rate. We confirm this linear correlation for these observations (softness ratio is calculated between the 0.3–2 and 2–10 keV bands in this work) in the left panel of Fig. 2. It is worth noting that although Obs2 and Obs3 show a spectral hardening during some periods, these periods still follow the correlation of unobscured states (see the left panel of Fig. 2). Therefore, these periods might not be in obscured states.

• •

  Obs6: Turner et al. (2018) had observed a rapid obscuration event in Obs6. To restudy this obscuration event using a broadband SED model and the pion model in SPEX, we followed Turner et al. (2018) to subdivide Obs6 into three slices: S6a, S6b, and S6c (see Fig. 1). S6c shows an unobscured state as it follows the correlation of Obs2 to Obs5 (see the right panel of Fig. 2). However, S6a and S6b deviate from the correlation of Obs2 to Obs5 (see the right panel of Fig. 2), therefore, might be in obscured states.

Next we make the detailed spectral analysis for these observational data. We begin our spectral modelling by using a broadband SED model from the NIR to the hard X-ray bands for NGC 3227. We refer readers to our Paper I for full details about this SED model and we only give a brief introduction here. The main spectral components that are used in this work are shown below:

1. 1.

   The intrinsic broadband SED (see details in Paper I), which is composed of a disk blackbody component (dbb), a warm Comptonized disk component (comt) from the optical to the soft X-ray band, an X-ray power-law component (pow), and a neutral reflection component (refl) in the hard X-ray energy band. For pow of NGC 3227, we used 309 keV (Turner et al., 2018) as the high-energy exponential cut-off and used 13.6 eV as the low-energy exponential cut-off.

2. 2.

   The obscurer components, which heavily absorb the X-ray spectrum. We used the pion model in SPEX to fit their absorption features in the spectrum.

3. 3.

   The warm absorber (WA) components, which produce absorption features in soft X-rays. We also used the pion model to fit these absorption features.

4. 4.

   The Galactic X-ray absorption, which was taken into account by the hot model in SPEX with the hydrogen column density $N_{\rm{H}}=2.07\times 10^{20}\ {\rm{cm}}^{-2}$ (Murphy et al., 1996).

Obs2 to Obs5, S1b, and S6c are in unobscured states, so we will fit their spectra using the spectral components 1, 3, and 4. S1a, S6a, and S6b are in obscured states, so their spectra will be fitted with the spectral components 1–4. Next we will present the details of the spectral analysis for spectral components 1, 2, and 3.

The relation between WA outflows of AGN and nuclear obscuration activities is still unclear. NGC 3227 is a suitable target to study the properties of both WAs and obscurers, which might help us understand their correlation. To investigate the WA components of NGC 3227 in detail, we use a broadband SED model (Paper I) and the photoionization model in SPEX software to fit the unobscured spectra of the archival and previously published XMM-Newton and NuSTAR observations taken in 2006 and 2016. Based on the broadband SED and WAs parameters, we also study the X-ray obscuration events in the archival observations.

We detect four ionization phases for the WAs in NGC 3227 using the unobscured observations: $\log\ \xi\ ({\rm{erg}}\ {\rm{cm}}\ {\rm{s}}^{-1})\sim-1.0,\ 2.0,\ 2.5,\ 3.0$. The highest-ionization WA component has a much higher hydrogen column density ($N_{\rm{H}}\sim 10^{22}\ {\rm{cm}}^{-2}$) than the other three WA components ($N_{\rm{H}}\sim 10^{21}\ {\rm{cm}}^{-2}$). The outflow velocities of these WA components range from $\sim-100$ to $\sim-1300$ ${\rm{km}}\ {\rm{s}}^{-1}$, and show a positive correlation with the ionization parameter. Our estimates of the radial location of these WA components indicate that the WAs of NGC 3227 might reside over radii ranging from the BLR to the torus, even to the NLR.

We find an obscuration event in 2006, which was missed by previous studies. One obscurer component is required for this 2006 obscuration event. For the previously published obscuration event in 2016, we detect two obscurer components. A high-ionization obscurer component ($\log\ \xi=2.81_{-0.04}^{+0.15}$; covering factor $C_{\rm{f}}=0.29^{+0.04}_{-0.08}$) only appears in the 2016 observation, which has a column density around $10^{23}\ \rm{cm}^{-2}$, while both the 2006 and 2016 observations have a low-ionization obscurer component ($\log\ \xi\sim$ 1.0–1.9; $C_{\rm{f}}\sim 0.2-0.5$), which has a lower column density ($\sim 10^{22}\ \rm{cm}^{-2}$) than the high-ionization obscurer component. Assuming that the variations of flux is caused by the transverse motion of obscurers across the line of sight, we estimate the locations of obscurers to be within the BLR.

The obscurers of NGC 3227 are closer to the SMBH than the WAs and have a significantly larger hydrogen or electron number density than the WAs. In addition, the obscurers usually appear on a short timescale while the WAs can exist for a long time. These proofs indicate that the obscurers and WAs of NGC 3227 might have different origins.