The Bluest Changing-Look QSO SDSS J224113-012108

Changing-look active galactic nuclei (CLAGN) have been studied for more than four decades, since the first CLAGN NGC 7603 was reported in Tohline & Osterbrock (1976) with its broad Hβ becoming much weaker in one year. To date, there are about 40 CLAGN reported in the literature, according to the basic properties that spectral types of AGN are changing between type-1 AGN (apparent broad Balmer emission lines and/or Balmer decrements near to the theoretical values) and type-2 AGN (no apparent broad Balmer emission lines and/or Balmer decrements much different from the theoretical values). There are so far dozens of CLAGN reported in the literature.

Cohen et al. (1986) have reported the discovery of the CLAGN Mrk 1086, for which the type changed from type-1.9 to type-1 in 4 yr, and more recent results on the CLAGN Mrk 1018 can be found in McElroy et al. (2016). Storchi-Bergmann et al. (1993) have reported on the CLAGN NGC 1097 with a detected Seyfert 1 nucleus, though it had previously shown only LINER characteristics. Aretxaga et al. (1999) have reported on the CLAGN NGC 7582 with a transition toward a type-1 Seyfert experienced by a classical type-2 Seyfert nucleus. Eracleous & Halpern (2001) have reported on the CLAGN NGC 3065 with newly detected broad Balmer emission lines. Denney et al. (2014) have reported on the CLAGN Mrk 590 with its type changed from Seyfert 1 in the 1970s to Seyfert 1.9 in the 2010s. Shappee et al. (2014) have reported on the CLAGN NGC 2617, which was classified as a Seyfert 1.8 galaxy in 2003 and as a Seyfert 1 galaxy in 2013. LaMassa et al. (2015) have reported on the CLAGN SDSS J0159 classed as a type-1 AGN in 2000 and as a type-1.9 AGN in 2010. More recently, MacLeod et al. (2016) have reported ten CLAGN with variable and/or changing-look broad emission line features, and Gezari et al. (2017) have reported on the CLAGN SDSS J1554 for which type changed from type-2 to Type-1 in 12 yr, and Ross et al. (2018) have reported one new changing-look quasar SDSS J1100-0053 based on about 20 yr long spectroscopic variability in different wavelength bands. Stern et al. (2018) have reported a new changing-look quasar, WISE J1052+1519, found by identifying high mid-infrared variability, and Yang et al. (2018) have reported a sample of 21 CLAGN with the appearance or the disappearance of broad Balmer emission lines within a few years.

In order to explain the nature of CLAGN, different models have been proposed, such as the dynamical movement of dust clouds well discussed in Elitzur (2012), the common variations in accretion rates well discussed in Elitzur et al. (2014), the variations in accretion rates due to transient events as discussed in Eracleous et al. (1995) and Blanchard et al. (2017), and the magnetic torques in the freefall region inside the innermost stable circular orbit as discussed in Ross et al. (2018), Stern et al. (2018), etc. There is, so far, no clear conclusion on the physical mechanism of type transitions in CLAGN. Models of dust obscuration and variation of the central accretion process produce different effects on the spectral energy distributions (SEDs) of AGN and so SEDs could be examined to determine which model plays the key role in type transition. More interestingly, the detection of bluer changing-look AGN at the bright state could provide clear clues that would rule out dust obscuration, if intrinsic SEDs after correction for dust obscuration were unreasonably higher the observed SEDs at a bright state, such as is discussed in the results for CLAGN SDSS J2241 in this paper.

Among the reported CLAGN, central black hole (BH) masses have been estimated in SDSS J0159 and in SDSS J1554. The virial BH mass of SDSS J0159 has been estimated to be 1.6 × 10⁸ M_⊙ in LaMassa et al. (2015) and in Zhang et al. (2019) based on virialization assumptions for broad emission line regions (Peterson et al. 2004); however, Zhang et al. (2019) measured the stellar velocity dispersion to be about 80 km s⁻¹ in SDSS J0159, leading to a virial BH mass much larger than the value found using the M–sigma relation (Ferrarese & Merritt 2000; Gebhardt et al. 2000; Kormendy & Ho 2013; Savorgnan & Graham 2015). Meanwhile, as discussed for SDSS J1554 by Gezari et al. (2017), the virial BH mass found through properties of broad emission lines is about 2 × 10⁸ M_⊙, and the measured stellar velocity dispersion is about 176 km s⁻¹, which leads to the M–sigma relation expected BH mass of about 10⁸ M_⊙, well consistent with the virial BH mass. Therefore, as pointed out by Yang et al. (2018), CLAGN can provide perfect cases in which to study the connection between AGN and their host galaxies, using observations of properties in different states. The fact that CLAGN can have different BH masses from different methods could provide further interesting clues about the physical nature of type transition.

In this manuscript, based on multi-epoch Sloan Digital Sky Survey (SDSS) spectra, we report a new and, so far, the bluest CLQSO SDSS J224113-012108 (=SDSS J2241) at z = 0.059, of which both the host galaxy properties and the bright nucleus can be well determined during its type transitions, leading to virial BH masses that are quite different from the mass determined through the M–sigma relation (a detailed review can be found in Kormendy & Ho 2013), which will provide further clues about the physical mechanism of type transitions in CLAGN. The manuscript is organized as follows. In Section 2 and Section 3, the main results are shown on the spectroscopic properties and long-term photometric variability properties in SDSS J2241. In Section 4, the main discussions are given. Then, in Section 5, we give our main conclusions. In the manuscript, we have adopted the cosmological parameters of H₀ = 70 km · s⁻¹Mpc⁻¹, Ω_Λ = 0.7, and Ω_m = 0.3.

SDSS J2241, classified as a QSO in SDSS, has been observed in SDSS in 2011 September, 2016 November, and 2017 October. Figure 1 shows the SDSS spectra with plate-mjd-fiberid as 4381-55824-0952, 9218-57724-0423, and 9150-58043-0340. It is clear that there are apparent contributions of the host galaxy in the spectrum in 2011, and apparent blue AGN continuum emissions in the spectra in 2016 and 2017.

Zoom In Zoom Out Reset image size

The commonly accepted SSP (Simple Stellar Population) method has been well applied to determine the host galaxy contribution in the SDSS spectrum in 2011, similar to what we have done in Zhang (2014), Zhang & Feng (2016), and Zhang et al. (2019). More detailed descriptions about the SSP method can be found in Bruzual & Charlot (2003), Kauffmann et al. (2003), and Cid Fernandes et al. (2005). Here, the 39 simple stellar population templates from Bruzual & Charlot (2003) have been exploited, which can be used to well describe the characteristics of almost all of the SDSS galaxies, as detailed in discussions in Bruzual & Charlot (2003). In addition, a power-law component has been applied to describe the AGN continuum emissions, due to the apparent broad Hα line in the SDSS spectrum in 2011. After the emission lines have been masked out, the observed SDSS spectrum can be well described by the broadened SSPs (with the stellar velocity dispersion as the broadening velocity) plus a power-law component using the Levenberg–Marquardt least-squares minimization technique, leading to χ² = SSR/Dof ∼ 1.3 (where SSR and Dof are the summed squared residuals and degrees of freedom, respectively). Here, we have not only masked out all of the narrow emission lines with rest wavelengths between 3700 and 7000 Å with widths of about 450 km s⁻¹, mainly including [O ii] λ3727, narrow Balmer lines, [O iii] λ4364, the [O iii] doublet, the [O i] doublet, the [N ii] doublet and the [S ii] doublet, etc., but also masked out the optical Fe ii lines and the broad Hα and broad Hβ, which have been shown in the left panel of Figure 1. The best fit to the host galaxy contribution has been shown in the left panel of Figure 1. The determined stellar velocity dispersion is about σ_⋆ ∼ 87 ± 5 km s⁻¹, considering 60 km s⁻¹ to be the mean SDSS instrument broadening velocity. And the determined power-law AGN component has f_λ ∝ λ^0.06±0.12 in the SDSS spectrum in 2011.

Meanwhile, after subtraction of the host galaxy contribution from the SDSS spectra in 2016 and in 2017, the blue parts of the spectra at rest wavelengths less than 4000 Å can be well described by f_λ ∝ λ^{−3.83±0.04} and f_λ ∝ λ^{−5.21±0.02} in 2016 and in 2017, respectively. Among the reported CLAGN in the literature, SDSS J1100-0053, well discussed in Ross et al. (2018), is also a bright quasar at its bright state, with the blue part of the spectra at rest wavelengths less than 4000 Å described by f_λ ∝ λ^−2.45. Therefore, among the reported CLAGN, SDSS J2241 is the bluest changing-look QSO at its bright state.

Besides the SSP method, the apparent Ca ii triplet around 8500 Å, which is well detected in the SDSS spectra in 2016 and in 2017, can provide another method to estimate σ_⋆. Based on the 1273 template stellar spectra with a resolution higher than about 30 km s⁻¹ collected from the Indo-U.S. Coudé Feed Spectral Library (Valdes et al. 2004), the stellar velocity dispersion is remeasured based on the best fit to the Ca ii triplet using the Levenberg−Marquardt least-squares minimization technique, similar to what has been done in Rix & White (1992) and Greene & Ho (2006). Here, in addition to the broadened template stellar spectra, a three-order polynomial function has been applied to modify the continuum shape. The best matches have been shown Figure 2, with the calculated χ² values around 0.8 and 0.6 for the results of the spectra in 2016 and in 2017, respectively. After considering the spectral resolution of the template stellar spectra and the SDSS instrument broadening velocity of 50 km s⁻¹ around the Ca ii triplet, the remeasured stellar velocity dispersions are about 83 ± 10 km s⁻¹ and 86 ± 12 km s⁻¹ for the spectra in 2016 and in 2017. The remeasured σ_⋆ through the Ca ii triplet are quite consistent with the determined value found using the whole spectrum observed in 2011. Therefore, the mean value of σ_⋆ ∼ 86 ± 12 km s⁻¹ is accepted as the stellar velocity dispersion of SDSS J2241.

Zoom In Zoom Out Reset image size

Before proceeding further, as we know that both the ratio Q_h of the quasar luminosity to the host galaxy luminosity and the overall signal-to-noise ratio (SN) of the spectrum could probably have apparent effects on the measured stellar velocity dispersion of the host galaxy of a quasar, the following two methods are applied to well demonstrate that the measured stellar velocity dispersion about 86 km s⁻¹ in SDSS J2241 is reliable. First, properties of the Ca ii triplet are mainly considered as follows. Besides the best match to the Ca ii triplet shown in Figure 2, a new component described by the same stellar template with the same strengthening factor but with a broadened velocity of 100 km s⁻¹ is also shown in Figure 2. The large difference between the best descriptions and the new component with a broadened velocity of 100 km s⁻¹ strongly indicates that the measured stellar velocity dispersion is much smaller than 100 km s⁻¹ in SDSS J2241. Second, properties of the whole spectra of SDSS J2241 are mainly considered as follows. Based on the best matches of S_λ to the stellar emission shown in the left panel of Figure 1, a series of 100 synthetic spectra without the effects of noise can be well constructed by the S_lambda normalized by S₅₁₀₀ = 1 plus a power-law component ${Q}_{h}\times {\left(\lambda /5100\right)}^{\beta }$ (Q_h as a random value from 0.2 to 1; Q_h = 0.4 is the value for the results in 2011 shown in the left panel of Figure 1), which is applied to describe intrinsic AGN continuum emissions. Besides considering effects of Q_h , different values of SN from 20 to 50 ($\mathrm{SN}\sim 40$ for the spectrum observed in 2011 in SDSS J2241) are considered to add noise to the synthetic spectra. Then, the SSP procedure is applied to measure the stellar velocity dispersions of the 100 synthetic spectra with considerations of both Q_h and SN. The top left panel of Figure 3 shows one constructed synthetic spectrum (with Q_h = 0.6, β = −0.57, and SN = 49) and the determined best matches to the synthetic spectrum. The top right panel of Figure 3 shows the correlation between the SSP method determined slope of the continuum emissions β_SSP and the input value of β_input, with a mean value of β_SSP/β_input ∼ 1.03. The bottom panels of Figure 3 show the dependence of the measured stellar velocity dispersions around 86 km s⁻¹ on the Q_h and SN. It is clear that there are few effects of Q_h and SN on the measured stellar velocity dispersion. Therefore, the measured stellar velocity dispersion in SDSS J2241 is reliable enough, and the SSP method determined stellar emission can be subtracted in order to measure properties of the emission lines of SDSS J2241.

Zoom In Zoom Out Reset image size

After subtraction of the host galaxy contribution, emission lines, especially around the broad Hα (rest wavelength between 6200 and 7000 Å) and around the broad Hβ (rest wavelength between 4400 and 5600 Å), can be well measured by the following model functions. Around Hα, due to the double-peaked features of broad Hα and quite weak narrow lines of the [O i] and [S ii] doublets, three broad Gaussian functions are applied to describe the broad Hα, and only three narrow Gaussian components are applied to describe the narrow Hβ and [N ii] doublet, and a power-law component is applied to describe the continuum emission underneath the broad Hα. Around Hβ, three Gaussian components are applied to describe the double-peaked broad Hβ, one narrow Gaussian component is applied to describe the narrow Hβ, two Gaussian components are applied to describe the [O iii] doublet, one broad Gaussian component is applied to describe the broad He ii, the Fe ii template discussed in Kovacevic et al. (2010) is applied to describe the optical Fe ii lines, and a power-law component is applied to describe the continuum emission underneath the broad Hβ. Then, using the Levenberg−Marquardt least-squares minimization technique, the emission lines can be well described by the model functions, which have been shown in Figure 4. The determined parameters have been listed in Table 1 for the broad Gaussian components of the broad Balmer lines. Before proceeding further, there is one point we should note. There are determined narrow emission lines shown in Figure 4; however, the determined fluxes of the narrow emission lines are smaller than 5 times their corresponding flux uncertainties. Therefore, the line parameters of these narrow emission lines are not listed in Table 1.

Zoom In Zoom Out Reset image size

Based on the measured line parameters of the broad Balmer lines, the flux ratios of broad Hα to broad Hβ are about 6.98, 2.86, and 2.79 in 2011, 2016, and 2017, respectively. The apparent changes of the Balmer decrements in six years lead SDSS J2241 to be a new CLAGN. Meanwhile, based on the power-law component in the blue part of the spectrum shown in the bottom panel of Figure 1 with f_λ ∝ λ^{−5.21±0.02}, SDSS J2241 is the bluest CLQSO among the reported CLAGN.

Figure 5 shows the V-band photometric light curve of SDSS J2241 collected from the Catalina Sky Survey (CSS; Drake et al. 2009). The long-term variability over more than 8 years from 2005 July to 2013 October (MJD from 53,552 to 56,593) are very interesting. In the first 4.5 yr with MJD from 53,552 to 55,261, there is no apparent variability in SDSS J2241. However, since MJD = 55,261, the nucleus becomes brighter and brighter. Certainly, the long-term variability properties cannot be explained by the Damped Random Walk (DRW) process, which has become a preferred modeling process to describe AGN intrinsic variability (Kelly et al. 2009; Kozlowski et al. 2010; Andrae et al. 2013; Zu et al. 2013).

Zoom In Zoom Out Reset image size

4.1. Virial BH Mass and BH Mass through M–sigma Relation

Based on the measured line parameters, the second moments σ_B of broad Hα are about 2980 ± 270 km s⁻¹, 3480 ± 230 km s⁻¹, and 3520 ± 240 km s⁻¹ in 2011, 2016, and 2017, respectively. The FWHM of broad Hα are about 8494 ± 770 km s⁻¹, 6947 ± 470 km s⁻¹, and 6170 ± 420 km s⁻¹ in 2011, 2016, and 2017, respectively. The line luminosities of broad Hα (L_B) are about (2.12 ± 0.12) × 10⁴¹ erg s⁻¹, (6.42 ± 0.28) × 10⁴¹ erg s⁻¹, and (7.11 ± 0.25) × 10⁴¹ erg s⁻¹ in 2011, 2016, and 2017, respectively. It is clear that the decreasing FWHM with increasing line luminosity is consistent with results expected when the virialization assumption is applied to broad Balmer lines; however, the variation of σ_B is not. Therefore, instead of using σ_B, we prefer to use the FWHM to estimate the virial BH masses in SDSS J2241, through the following equation well discussed in Greene & Ho (2005) and after considering the more recent improved R−L relation in Bentz et al. (2013)

$$\begin{eqnarray}&&\displaystyle \frac{{M}_{\mathrm{BH}}}{{M}_{\odot }}=2.4\times {10}^{6}{\left(\displaystyle \frac{{L}_{{\rm{B}}}}{{10}^{42}\,\mathrm{erg}\,{{\rm{s}}}^{-1}}\right)}^{0.47}{\left(\displaystyle \frac{\mathrm{FWHM}}{1000\,\mathrm{km}\,{{\rm{s}}}^{-1}}\right)}^{2.06}.\end{eqnarray} \tag{ 1 }$$

The virial BH masses are about (9.5 ± 2.5) × 10⁷ M_⊙, (10.5 ± 2.9) × 10⁷ M_⊙, and (8.7 ± 1.9) × 10⁷ M_⊙ in 2011, 2016, and 2017, respectively. Therefore, the mean value (9.6 ± 2.4) × 10⁷ M_⊙ is accepted as the virial BH mass of SDSS J2241.

Meanwhile, based on the measured stellar velocity dispersion σ_⋆ = 86 ± 12 km s⁻¹, the expected BH mass is about 2.8 × 10⁶ M_⊙ based on the M–sigma relation discussed in Kormendy & Ho (2013) and in Zhang et al. (2019). Therefore, the virial BH mass is about two magnitudes larger than the BH mass found using the M–sigma relation. The larger virial BH mass could provide further clues about special dynamical properties of broad line regions (BLRs). Furthermore, as shown in Figure 4, the line profiles of broad Hα and broad Hβ are very different: there are apparent double peaks in broad Hβ, but one peak plus one shoulder in broad Hα (especially in 2016 and 2017). We cannot find a natural explanation for the different line profiles, but they may indicate that there could be non-Keplerian components in broad Balmer emission regions. Further effort is necessary to check the variability properties of broad emission lines.

4.2. Moving Dust Clouds?

In the CLQSO SDSS J2241, if the moving dust clouds scenario was applied to explain the changes in the Balmer decrement from 6.98 in 2011 to 2.79 in 2017, E(B − V) = 0.8 could be estimated, assuming the theoretical Balmer decrement to be 2.80. Then, intrinsic reddening-corrected continuum emission could be checked in 2011, and they are shown as dashed dark green lines in Figure 1. It is clear that, when considering moving dust clouds as the cause of changes in the Balmer decrements, the expected intrinsic continuum emission would be unreasonably larger than the continuum emission in 2017 with no dust obscuration. Therefore, the moving dust clouds scenario cannot be applied in case of SDSS J2241. That said, we cannot totally rule out the moving dust clouds scenario to explain the CLQSO SDSS J2241, as the spectra redward of broad Hα have tiny variations in the spectral slope that contrast those expected from varying obscurations.

Actually, besides the established variations in the continuum emission of SDSS J2241, there are convincing methods that can be applied to rule out the moving dust clouds scenario in CLAGN, such as the timescale arguments that are discussed in LaMassa et al. (2015), Runnoe et al. (2016), and Ross et al. (2018) and the expected profiles of the broad emission lines discussed in Ruan et al. (2016). These two methods are applied in SDSS J2241 as follows.

First, we consider the timescale arguments as follows. As discussed in LaMassa et al. (2015), Runnoe et al. (2016), etc, it is necessary and interesting to check whether the timescale ${t}_{\mathrm{cross}}\sim 0.07{r}_{\mathrm{orb}}^{1.5}{M}_{8}^{-0.5}\arcsin ({r}_{\mathrm{src}}/{r}_{\mathrm{orb}})$ is short enough for foreground moving dust clouds in a bound orbit around the central BH in front of the continuum emission source and BLRs. Here, r_orb is the radius of the foreground moving dust clouds, M₈ is the BH mass in units of 10⁸ M_⊙, and r_src is the true size of the BLRs. Commonly, there are no accurate values for the parameters r_orb, M₈, and r_src. However, it is not difficult to estimate a minimum timescale t_cross when the estimated size of central BLRs is considered. Here, at the bright state of SDSS J2241 with a continuum luminosity at 5100 Å about 1.03 × 10⁴³ erg s⁻¹ in 2017, the size of the central BLRs is estimated to be R_BLRs ∼ 11 lt-day using the improved empirical R−L relation reported in Bentz et al. (2013). Similar to what has been discussed in LaMassa et al. (2015), Runnoe et al. (2016), etc., assuming r_orb ≥ k_o × R_BLRs and r_src ∼ k_s × R_BLRs, the crossing timescale can be estimated as

$$\begin{eqnarray}&&{t}_{\mathrm{cross}}\sim 2.6{M}_{8}^{-0.5}{k}_{o}^{1}.5\arcsin ({k}_{s}/{k}_{o})\mathrm{years}.\end{eqnarray} \tag{ 2 }$$

The estimated crossing timescale sensitively depends on both the central BH mass and the parameters of k_s and k_o . If we accept that k_o = k_s = 3, as has been set in LaMassa et al. (2015) and Runnoe et al. (2016), and that accepted BH mass is about 10⁸ M_⊙ (the virial BH mass), the minimum crossing timescale is about 21 yr. Moreover, if the central BH mass of 2.8 × 10⁶ M_⊙ found using the M–sigma relation is accepted, the minimum crossing timescale should be 125 yr. It is clear that the commonly estimated crossing timescale is quite longer than the observed transition time in SDSS J2241. Therefore, the timescale argument rules out the moving dust clouds scenario in SDSS J2241.

Second, we discuss the arguments based on broad line profiles as follows. As well discussed in Ruan et al. (2016), the moving dust clouds would have apparent effects on the profiles of broad emission lines, in that broad line emission from the outer, lower-velocity emission regions of the broad line region is more attenuated than the emission from the inner emission regions, leading to broad optical emission lines (such as the broad Balmer lines) that are broader in the dim state. However, in contrast to the findings for SDSS J2336 that are shown in Figure 4 in Ruan et al. (2016), the line profiles of broad Hα are more complicated in SDSS J2241. And, as the results indicate above, if the second moment of the line width is considered, the broad Hα become broader from 2011 to 2017; however, if the FWHM is considered as the line width, the broad Hα become narrower 2011 to 2017. Therefore, the properties of line profiles of broad emission lines cannot provide clear clues to rule out the scenario of moving dust clouds in SDSS J2241.

Based on this discussion, rather than moving dust clouds, accretion rate variations are preferred to explain the changing-look QSO SDSS J2241.

4.3. Accretion Rate Variations?

I give our main conclusions as follows. Based on the SDSS spectra observed in 2011, 2016, and 2017, the flux ratio of broad Hα to broad Hβ has changed from 7 in 2011 to 2.7 in 2017 in the QSO SDSS J2241, which identifies SDSS J2241 as a new CLQSO. Moreover, after considering the host galaxy contribution, the spectral index α_λ ∼ −5.21 ± 0.02 with λ < 4000 Å can be well determined in the 2017 spectrum, which makes SDSS J2241 the bluest CLQSO discovered so far. Based on the determined stellar velocity dispersion of about 86 km s⁻¹ and the properties of broad Hα, the virial BH mass in SDSS J2241 is about two magnitudes larger than the BH mass found using the M–sigma relation, which is not similar, as the clear case in the CLAGN SDSS J1554 which has virial BH mass is well consistent with the BH mass through the M-sigma relation. The long-term photometric variability curve shows interesting slowly and smoothly changing properties, which are not expected based on the DRW process that is commonly applied to normal AGN. Moreover, based on the properties of the continuum emission in 2017 assuming no dust obscuration, the consideration of only the moving dust clouds scenario cannot be explain the features of CLQSO SDSS J2241, because the expected intrinsic reddening-corrected continuum emission in 2011 were unreasonably higher than the unobscured continuum emission in 2017. Accretion rate variations are instead preferred, and the probable accretion rate variability could be probably due to a TDE in SDSS J2241.

I gratefully acknowledge the referee for reading the manuscript carefully and patiently, and for the constructive comments and suggestions that greatly improved the paper. I gratefully acknowledge the kind support of the Starting Research Fund of Nanjing Normal University and the financial support of NSFC-11973029. This manuscript has made use of the data from the SDSS projects. The SDSS-III website is http://www.sdss3.org/. SDSS-III is managed by the Astrophysical Research Consortium for the Participating Institutions of the SDSS-III Collaboration.