Long-term multiwavelength monitoring and reverberation mapping of NGC 2617 during a changing-look event

NGC 2617 was first mentioned by Edouard Jean-Marie Stephan in 1885. It is a face-on Sc galaxy at a redshift of z = 0.0142 classified as Seyfert 1.8 (Sy 1.8) (Moran et al., 1996; Paturel et al., 2003) and located in Hydra. Shappee et al. (2013a, 2014) detected a dramatic change of the object’s spectral type that followed a significant brightening of the nucleus discovered by the All-Sky Automated Survey for SuperNovae (ASAS-SN) in April 2013. Prior to 2013, the object had been studied very rarely. Only two spectra were published: one using a slit spectrograph in 1994 (Moran et al., 1996) and the other with an objective prism in 2003 (6dFGS). These old spectra demonstrated the lack of the broad component in H$\beta$ and very shallow broad H$\alpha$ wings which resulted in a Sy 1.8 classification (Osterbrock, 1987). Comparison with more recent spectral observations allowed Shappee et al. (2014) to determine a change of spectral type (CL) from Sy 1.8 to Sy 1 between 2003 and 2013. The CL transition can be numerically indicated as -2, that is two classes from 1.8 through 1.5 to type 1, in the “magnitude” system introduced by Runco et al. (2016). The following IR, optical, UV photometry and X-ray observations revealed a strong outburst in all wavelengths. Specifically, the X-ray flux in 2013 was significantly brighter than measured by ROSAT (Boller et al., 1992). This triggered more intensive investigations of NGC 2617 from X-ray to radio wavelengths (Mathur et al., 2013; Shappee et al., 2013b; Tsygankov et al., 2013; Yang et al., 2013; Oknyansky et al., 2016a, b, c, 2017b, 2017a, 2018b; Giustini et al., 2017; Sheng et al., 2017; Fausnaugh et al., 2017, 2018; Sheng et al., 2017; Olson et al., 2020; Yang et al., 2021; Feng et al., 2021).

We began spectroscopic and photometric (IR $JHK$ and optical $BVRI$) monitoring of NGC 2617 in January 2016 to trace potential state changes (see Oknyansky et al. 2017a = Paper I, Oknyansky et al. 2017b, 2018b). Based on archival photometry (MASTER Global Robotic Net (Kornilov et al., 2012; Lipunov et al., 2010, 2019, 2022), MASTER hereafter), we proposed that the Seyfert type change had probably occurred between 2010 October and 2012 February. In 2016, while in the bright state, the object revealed a new emission component in the red wing of H$\beta$. Reverberation mapping analysis found that UV emission variations lag the X-ray by about 2-3 days as in the findings of Shappee et al. (2014). Uncorrelated X-ray/UV events were observed, which can be explained via X-ray absorption by Compton-thick clouds moving across the line of sight (see Paper I). This scenario is in agreement with an independent result for NGC 2617: the detection of a redshifted, high-velocity and highly ionised absorber with a column density $N_{H}\sim$ 10²³ cm^-2 (Giustini et al., 2017).

IR variations in the K band were found to be delayed with respect to the optical ones by $\sim$ 22 days, but variations in the J band were delayed by only $\sim$ 3 days. The time delay in the $K$ band is inconsistent with the value found by Shappee et al. (2014) (see Paper I). From May 2016 we also requested Swift observations of NGC 2617. In December 2017, the X-ray flux was detected to be the lowest since monitoring began in 1990-1991. The object brightened again in July 2018 (Oknyansky et al., 2018b) and demonstrated Sy1 type spectra at least until January 2019 (Senarath et al., 2021).

The first reverberation mapping for the optical emission lines was done for NGC 2617 by Fausnaugh et al. (2017) and an H$\beta$-to-continuum time delay of 6.4$\pm$ 0.5 days was found. The reverberation mapping for the continuum variations of the object at different wavelengths (optical, UV and X-ray) during 2014 was published by Fausnaugh et al. (2018). The lags were consistent with geometrically thin accretion-disc models, however, the observed lags were larger than predictions based on standard thin-disc theory by a factor of a few. The X-ray variability of NGC 2617 (among the number of selected Seyfert 1.8/1.9 objects) was investigated by Hernández-García et al. (2017) and its continuum was modelled by a combination of two power laws.

The results of a high cadence velocity-resolved reverberation mapping campaign for the interval from 2019 October to 2020 May were recently published by Feng et al. (2021). The investigation found the time lags for H$\alpha$, H$\beta$, H$\gamma$ decreasing from the former lines to the latter. For H$\alpha$, and H$\beta$ the lag of the line cores were found to be longer than those of the relevant wings, and the peak of the velocity-resolved cross-correlations was slightly blueshifted.

The physical nature of the CL AGN phenomenon is still mysterious, particularly, the origin of the UV and optical variability, as well as its relation to the X-ray emission, which is not yet clearly established. Therefore variability investigations can be very helpful for understanding the central structures and physical processes in AGNs (see discussion and references in Runnoe et al., 2016; MacLeod et al., 2019; Oknyansky et al., 2017b, 2019b, 2021; Ruan et al., 2019; Wang & Bon, 2020). As soon as the first CL events in AGNs were discovered (Khachikian & Weedman, 1971; Pronik & Chuvaev, 1972; Lyutyj et al., 1984) two possible mechanisms were suspected to be responsible for the dramatic changes of the spectral type: strong change of the accretion rate and variable obscuration (see e.g. Malkan et al., 1998). It is possible that these two mechanisms are both involved. If the region of an AGN is obscured for the observer by dust clouds it will have close to a Sy2 optical spectrum. If at some moment a strong outburst happens, then the dust will be sublimated in part and so the dense clouds in the Broad Line Region (BLR) will become directly observable. If the accretion rate decreases after a while and the object stays in the low state for a few years then the dust can recover in some clouds along the line of sight and the object will not show broad emission lines due to low UV radiation and the obscuration by dust clouds in the line of sight. This possible situation was discussed in respect to NGC 2617 and can be considered a prediction (see details in Paper I and in Oknyansky (2022)).

Here we present our new investigations of the X-ray/UV/optical/IR continuum and optical emission-line variability of NGC 2617 during October 2016 – January 2022 using data obtained by the Swift, MASTER, SAI MSU (Russia), WIRO (USA) and Shandong (China) observatories. We also perform reverberation mapping (RM) of the data to investigate lags between the variable emission components at different wavelengths, including those associated with the H$\beta$ and H$\alpha$ broad emission lines.

In our previous study (Paper I), we reported on the results of spectroscopic and photometric monitoring of NGC 2617 during January – June 2016. In the present study we combine our old data and our new data obtained partly with the same telescopes, but with some differences and additions.

During 2017-2022 a large variation in the H$\beta$ intensity (see Fig. 2) was accompanied by a significant change in the line profile (see Fig. 8 and Fig. 9). The mean and root-mean-square (rms) spectra in H$\beta$ region (just WIRO data) for NGC 2617 are shown in Fig.10 (for details see Du et al. 2018). The narrow emission lines in the rms spectra are not seen, since there is no variability and that is an indication of correct calibration procedure. As seen from the rms spectra, the variability in the broad H$\beta$ line was significant and asymmetric. The examples of the H$\beta$ profiles for the maximum and minimum states are presented in Fig. 11 (WIRO) and Fig. 12 (CMO). We remove the narrow component of the H${\beta}$ and the [OIII]${\lambda}$4959 line which blend with the broad component by subtracting copies of the [OIII]${\lambda}$5007 line, scaled and shifted to the best fit. A similar procedure was performed for H$\alpha$ in the CMO data, where the [OI]${\lambda}$6300 line was used as a template. For the WIRO data, the H${\alpha}$ flux is simply corrected by a constant contribution from the narrow lines, which is estimated using overlapping CMO and WIRO data. The profiles of broad H$\beta$ in the high state are more asymmetric with the blue wing dominating. A double-peaked profile can be seen and the blue peak is closer to the narrow H$\beta$ component than the red one, and that is a reason why the red peak is seen more prominent in some spectra (see Fig. 8 and Fig. 9).

In Fig.1, we see a good correlation of fast variability in X-rays and $UVW1$, but in recent years (2017–2022) there has been a slow downward trend in UV that has been absent in X-rays. Due to this fact, there has to be some evolution of the UV/X-ray flux ratio. To reveal this evolution, we took into account the existence of time delay between X-ray and UV variations of about 2 days and selected only those X-ray and UVW1 data that satisfy this delay. In Fig. 14, the F($UVW1$) to F(X-ray) relations for two observational intervals (June 2016 – April 2017 and April 2017–January 2022) are significantly different, with a much lower UV/X(-ray) flux ratio for the second interval.

To trace the evolution of the X-ray spectrum of NGC 2617 as a function of luminosity, we performed an analysis following the procedure described by Oknyansky et al. (2021). Namely, we selected two Swift data sets representing the states with minimal and maximal observed flux. For the low state, we took observations performed between December 4 and 17, 2017 (ObsID 00032812241–00032812246), whereas for the bright state we used the observation 00032812008 done on May 8, 2013. The fluxes in the 0.5–10 keV range were different in these states by a factor of 20. The following spectral analysis has been done using xspec package (Arnaud, 1996) and applying W-statistics (Wachter et al., 1979) after the energy spectra were rebinned to have at least one count per channel.

For the first step we fitted spectra in both states with the simplest absorbed power-law model (phabs$\times$po in xspec) with the absorption fixed at the Galactic value of 3.7$\times 10^{20}$ cm^-2 (HI4PI Collaboration et al., 2016). The unfolded spectra are shown in Fig. 15(a) with the best-fit parameters of the simple model presented in Tab. 5. As can be seen from panel (b) of the same figure, some evidence for a soft emission component can be noticed around 0.6–0.7 keV in the bright state. To improve the quality of fitting we introduced a mekal component with a temperature of $0.25\pm 0.02$ keV and normalization $5.2\pm 1.1$. This modification decreased the C-statistics value to 505.7 with 569 d.o.f. and made photon index slightly harder ($1.80\pm 0.4$). Unfortunately, the limited energy range and low count statistics do not allow us to make final conclusions about the composition of the intrinsic spectrum of the source even in the bright state. However, the hardening of the spectrum towards the low state is detected reliably. Similar behaviour was found for several CL AGNs and can be caused by different reasons (see, e.g., Ricci & Trakhtenbrot, 2022, for a review).

We can conclude from the available data that the soft thermal component in the bright state can not be explained by the contribution of the host galaxy. The significant difference detected in the photon index values (for high and low states) can be explained, e.g., by a substantial drop of the accretion rate in the low state and following change of the ionisation state in the material nearest to the SMBH (supermassive black hole), the Compton-thick clouds which partially obscure the X-ray source. In high states these Compton-thick clumps are highly ionised and in case of crossing the line-of-sight may produce temporarily colorless obscuring of the X-ray source (see, e.g. Paper I; Oknyansky et al., 2021; Turner et al., 2011). In low states these clumps are less ionised and may produce obscuration with change of the photon index.

NGC 2617 is one of the clearest cases where the appearances of both type 1 and type 2 AGNs have manifested themselves in the same object at different epochs. What must happen to make such a dramatic change possible? CL AGNs like NGC 2617 present a problem for the simplest unification model that considers the type 1 and type 2 AGNs to have the same nature and the difference in type to be due to the orientation to the observer. In this picture, we see a type 2 AGN if the BLR and AD are blocked from our view by obscuring dust surrounding the AGN perpendicular to the axis of symmetry (Keel, 1980). However, the orientation cannot change on the timescale of the observed type changes and hence some other explanation is needed. In a review by Ricci & Trakhtenbrot (2022), CL AGNs were divided in 2 classes: (1) changing-obscuration objects and (2) and changing-state AGNs. We have suggested (Oknyansky et al., 2015) that CL AGNs can have a combination of these two factors: i.e., a CL event can be due to a significant change in the accretion rate which leads to a variable absorption of the continuum and broad line emission by dust. We proposed a hollow bi-conical dust distribution model (Oknyansky et al., 2015, 2018a; Oknyansky & Gaskell, 2023)), where the sublimation or recovering of dust in some clouds along the line of sight in the hollow cone can be an additional factor to explain the CL phenomenon (see also investigations of pole dust obsucration in AGNs, Gaskell & Harrington, 2018; Buat et al., 2021). The model can explain not only the similar IR lags at different wavelengths (see results of numerical simulations performed by Oknyansky et al., 2015) but also the high correlation values for the IR and optical (UV) fluxes. Furthermore, the model can better (than in case of flat geometry) explain the observed luminosity in the IR due to a higher covering factor for UV radiation. The need for fast re-formation of the dust in order to explain the good correlation between UV and near-IR light curves was realized three decades ago (Barvainis, 1992) and later examples of the dust recovery were observed for some CL AGNs (Oknyanskij et al., 1999b, 2006, 2008; Oknyansky et al., 2014b; Kokubo & Minezaki, 2020). There are two main mechanisms that may be responsible for the recovery of dust after sublimation: (1) a condensation mechanism similar to the formation of dust in novae and supernovae; (2) delivery of dust from farther distances from the central source. Besides, it is possible that some amount of the dust survives in dense clouds during the outbursts being shielded from the UV radiation and then somehow moves from there to the outer parts of the cloud (Barvainis, 1992). In any case, this process takes some time. The estimates of the time required for the dust to recover were based on observations, and RM for different time intervals gave a value on the order of several years (Oknyanskij et al., 2006, 2008; Oknyansky et al., 2014b; Kokubo & Minezaki, 2020). Our results demonstrating the decrease of the near-IR time delay for NGC 2617, which started after a few years of being in the low state, are in agreement with the proposed recovery of the dust. In the conical outflow model, depending on the recent luminosity history, the opening angle of the hollow, dust-free conical region (as well as the same for the partially dust sublimated conical region) will be larger or smaller (see Fig. 1 in Oknyansky & Gaskell, 2023) due to dust recovery or sublimation processes. That can explain saturation in the relation between $F{{}_{\lambda}}(B)$ (which is proportional to $F_{UV}$) and $F{{}_{\lambda}}(K)$ at Fig. 6, which can be connected with variations of the covering factor. In Paper I, using the relation from Sitko et al. (1993) (see Eq. 1 and details there), we obtain a rough estimate for the dust sublimation radius (R_sub) at NGC2617 of about 32 light-days. Due to several not well known parameters, this estimate only provides a very preliminary value, but it is close to the time delay for $K$ band (this relation with the same parameters well fits the observaional data, Oknyanskij & Horne, 2001). In the conical dust outflow model, the sublimation radius R_sub also depends on the effective half-opening angle of the dust conical region due to axial symmetry of the UV radiation (see, e.g., Netzer, 2015) as well as due to absorption of the UV radiation by dust. The change in the mean UV luminosity (during 2016-2018 and 2019-2021) was very significant (factor of $\sim~4$) and that can explain variations of about a factor of two in the time delay, but we have to consider possibilities for fast dust recovery (see, e.g., Barvainis, 1992; Oknyansky & Gaskell, 2023). More detailed discussion of the conical dust outflow model will be done in future publications.

Changes in energy generation in CL AGNs are common, although it is not clear what drives them (see discussion and references on the topic in MacLeod et al., 2019; Oknyansky et al., 2017b, 2019b, 2021; Runnoe et al., 2016; Ruan et al., 2019; Hon et al., 2022). Mostly the discussed mechanisms responsible for such dramatic changes are various types of instabilities in the AD and tidal disruption events (TDEs). The TDE explanation has a problem of too low predicted cadence of such events that contradicts the fact that the observed CL phenomena are not too rare, as well as the recurrence of the events in individual objects (Oknyansky, 2022). A similar mechanism involving the tidal stripping of stars (Campana et al., 2015; Ivanov & Chernyakova, 2006) could lead to events more frequent (than TDEs) and recurrent events of significant brightening. The recurrence of CL events can find a natural explanation in models with AD instabilities (see, e.g., Sniegowska et al., 2020) and also by invoking the case of close binary black holes with low mass ratio Wang & Bon (2020). CL AGNs were proposed by Wang et al. (2022) to be some particular AGNs at a special evolutionary stage, such as a transition stage from different levels of fueling of the supermassive black hole. Dodd et al. (2021) investigated the properties of the host galaxies of CL AGNs and found suggested certain physical conditions leading to episodic accretion events responsible for triggering CL activity.

For NGC 2617, the duration of monitoring is too short to be able to fix the recurrence of CL events. It is unclear if the object had been in the low state characterized by spectral type Sy1.8 permanently or for a long time before this interval of high activity started, but we do see some evidence that the object is moving back to the low state. We have detected very deep minima in all wavelengths as well as in broad emission lines. It is useful to continue the monitoring which can help in selecting the best model to explain the CL phenomenon in this case.

We have found time delays between continuum variations at different wavelengths (X-ray, UV, optical, near-IR), as well as delays between continuum and broad emission lines, using our own statistical method MCCF. It is common to attribute the optical/UV emission to viscous heating of matter in an optically thick AD around a SMBH (Shakura & Sunyaev, 1973; Novikov & Thorne, 1973). The commonly invoked ‘lamppost’ model predicts an X-ray emitting ‘corona’ located on the rotation axis of the SMBH (see, e.g., Cackett et al., 2007) and wavelength-dependent time delays between the continuum at different wavelengths due to reprocessing of the extreme UV/X-ray photons into UV/optical radiation arising in the AD. A delay of $\sim$ 2 days found in this study for the lag of UV variations relative to the X-rays (which is in agreement with previously published results Shappee et al. (2014); Kammoun et al. (2021)), in combination with a very short delay between the UV and optical variations, is in contradiction with the lamppost-type X-ray reprocessing model. The same problem was noted before as a common property of CL AGNs (see, e.g., Cackett et al., 2007; Buisson et al., 2017; Edelson et al., 2017; McHardy et al., 2014; Shappee et al., 2014; Oknyansky et al., 2017b, 2020, 2021; Oknyansky, 2022; Lawther et al., 2023). The sizes of ADs measured from the microlensing data for some gravitationally lensed quasars are also significantly larger than those predicted by the reprocessing model (e.g., Morgan et al., 2010). A possible version of the lamppost-type scenario is to place the X-ray source at a large height (about 2 light days) from the SMBH, however this option was discussed by Noda et al. (2016) and rejected as highly unrealistic (see details therein). Two alternative explanations for the obtained results are: (1) a truncated AD in analogy with black-hole binaries (e.g, Done et al. (2007)) and (2) the model with a soft excess region at the inner edge of an AD which completely hides the hard X-ray corona (Gardner & Done, 2017; Edelson et al., 2017). Both possibilities were discussed by Noda et al. (2016) and they preferred the first explanation (1).

Meanwhile, our results for NGC 2617 (as well as for NGC 3516) involving soft variable black body components contradict the truncated AD model and are in agreement with the predictions of the second explanation (2) (see details in Oknyansky et al. 2021). Using our optical spectroscopy and multi-wavelength photometry (from near-IR to X-ray region), we have shown that NGC 2617 had been in the high state for several years (2013–2017) (see also Paper I), and from the beginning of April 2017 until the end of May 2018 the object had very low brightness and variability. In December 2017, the X-ray flux was the lowest since monitoring began. Despite the very low X-ray and UV flux observed in December 2017, the broad H${\beta}$ line was still prominent. As it was suspected in our publication (Paper I), the dust needs at least a few years of the object being in the low state to recover and to be able to obscure broad lines. Following this prediction the object was for a few years relatively low (2019–2021) before a new CL event was detected for the object (a change from Sy1 to Sy1.8 spectral type and this CL transition can be numerically indicated as +2, in the “magnitude” system introduced by Runco et al. (2016)). We have also found that the near-IR time delay became significantly lower than it was in 2016. Variable dust obscuration may explain the decrease of the UV/X-ray flux ratio in 2017–2022 found in our study. The tendency for a higher H${\alpha}$/H${\beta}$ ratio may also exist due to variable dust obscuration. All these facts support the assumption of dust recovering during the long low state of the object.

Using spectroscopy and multi-wavelength photometry, we have shown that NGC 2617 continued to be in the high state in recent years and to appear as a type 1 AGN, but more recently (2019–2021) the object was relatively low and the broad H$\beta$ line has become very weak. Perhaps as we proposed in Paper I, the object needs to be in the low state at least for a few years so that the dust can recover and obscure broad lines. The object has fulfilled the prediction and a new CL event was detected. We have also found that near-IR time delay became significantly lower than it was in 2016, which also supports the assumption of dust recovering. In the case of absorption, the UV/X-ray ratio also has to drop down. The duration of the high state and the continuing variability are not readily consistent with the type change being due to a tidal disruption event or a supernova. We propose that the original change of type was the result of increased luminosity causing the sublimation of dust in the inner part of a bi-conical dusty outflow. This leads to a much clearer visibility of the central regions. The recovery of dust occurring during the long low state is a reason for a new CL event.

We thank A. Cherepashchuk for supporting our research and observations, and the staffs of the observatories. We also express our thanks to the Swift administrators for approving the ToO observation requests and the Swift ToO team for promptly scheduling and executing our observations. We are grateful to M. Gaskell, P. Ivanov, H. Netzer and A. Laor for useful discussions. This work and MASTER equipment were supported in part by the M. V. Lomonosov Moscow State University Program of Development, by the National Research Foundation of South Africa. This research has been partly supported by Israeli Science Foundation grant no. 2398/19 and by the Center for Integration in Science of the Ministry of Aliyah and Integration. We thank WIRO engineers James Weger, Conrad Vogel, and Andrew Hudson for their indispensable and invaluable assistance. M.S.B. enjoyed support from the Chinese Academy of Sciences Presidents International Fellowship Initiative, grant No. 2018VMA0005. T.E. Zastrocky acknowledges support from NSF grant 1005444I.

The data underlying this study are available in the main body of the article and in online supplementary material.